View AM7968AM7969_1357883.PDF datasheet online --- IC-ON-LINE

Datasheet File OCR Text:

taxichip tm integrated circuits transparent asynchronous transmitter/receiver interface am7968/am7969-125 am7968/am7969-175 data sheet and technical manual 1994
? 1994 advanced micro devices, inc. advanced micro devices reserves the right to make changes in its products without notice in order to improve design or performance characteristics. this publication neither states nor implies any warranty of any kind, including but not limited to implied warrants of merchantability or fitness for a particular application. amd a assumes no responsibility for the use of any circuitry other than the circuitry in an amd product. the information in this publication is believed to be accurate in all respects at the time of publication, but is subject to change without notice. amd assumes no responsibility for any errors or omissions, and disclaims responsibility for any consequences resulting from the use of the information included herein. additionally, amd assumes no responsibility for the functioning of undescribed features or parameters. trademarks amd and the amd logo are registered trademarks of advanced micro devices, inc. taxichip and taxi are trademarks of advanced micro devices, inc. product names used in this publication are for identification purposes only and may be trademarks of their respective companies.
table of contents iii table of contents am7968/am7969 taxichip integrated circuits am7968/am7969 data sheet 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . am7968/am7969 technical manual 50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . chapter 1 introduction 50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 the am7968 taxi tm transmitter 50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 the am7969 taxi receiver 52 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . chapter 2 using the taxichip set 52 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 data and command 52 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 operational modes: local, cascade and test 53 . . . . . . . . . . . . . . . . . . . . . . . . chapter 3 data encoding, violation and syncs 53 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 data encoding 53 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 violation logic 57 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 taxi pll characteristics 57 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . chapter 4 clock generation and distribution 59 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 taxi transmitter clock connections 59 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 local mode transmitters 60 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 taxi receiver clock connections 60 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 cascade mode receivers (am7969-125 only) 61 . . . . . . . . . . . . . . . . . . chapter 5 interfacing with the serial media 61 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 very short link, dc coupled 62 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 terminated, dc coupled 63 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 terminated, ac coupled 63 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 baseline wander and the ac coupling capacitor 64 . . . . . . . . . . . . . . . . . . . . . 5.5 interfacing to fiber optic transmitters/receivers 66 . . . . . . . . . . . . . . . . . . . . . 5.5.1 dc-coupled taxl-fiber optic transceiver interface 66 . . . . . . . . . . . . . 5.5.2 ac-coupled taxl-fiber optic transceiver interface 68 . . . . . . . . . . . . . 5.6 interfacing to coaxial cable 68 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7 interfacing to twisted-pair cable 70 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . chapter 6 board layout considerations 71 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 printed circuit board layout 71 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.1 rules for layout 71 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 layout using fiber optic data links 73 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
amd table of contents iv chapter 7 cascade mode operation 74 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 transmit cascaded data with a single taxi transmitter 76 . . . . . . . . . . . . . . . 7.2 receivers in cascade mode: connections (am7969-125 only) 79 . . . . . . . . . . . 7.3 auto-repeat configuration 81 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3.1 receiver connections in auto-repeat configuration 81 . . . . . . . . . . . . . 7.3.2 timing limitations of the auto-repeat configuration 84 . . . . . . . . . . . . . 7.4 unbalanced configuration (am7968/am7969-125 only) 85 . . . . . . . . . . . . . . . . chapter 8 test mode 86 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.1 transmitter connections 87 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 receiver connections 89 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3 timing relationships in test mode 89 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . appendix a optical components manufacturers 90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . appendix b error detection efficiency 91 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . appendix c taxi tips 94 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
publication# 07370 rev. f amendment /0 issue date: april 1994 advanced micro devices am7968/am7969 taxichip tm integrated circuits (transparent asynchronous xmitter-receiver interface) final distinctive characteristics n parallel ttl bus interface eight data and four command pins or nine data and three command pins or ten data and two command pins n transparent synchronous serial link +5 v ecl serial i/o ac or dc coupled nrzi 4b/5b, 5b/6b encoding/decoding n drive coaxial cable or twisted pair directly n easy interface with fiber optic data links n 32C140 mbps (4C17.5 mbyte/s) data throughput n asynchronous input using strb/ack n automatic mux/demux of data and command n complete on-chip pll, crystal oscillator n single +5 v supply operation n 28-pin plcc or dip or lcc general description the am7968 taxichip transmitter and am7969 taxichip receiver chipset is a general-purpose inter- face for very high-speed (4C17.5 mbyte/s, 40C175 mbaud serially) point-to-point communications over co- axial or fiber-optic media. the taxichip set emulates a pseudo-parallel register. they load data into one side and output it on the other, except in this case, the other side is separated by a long serial link. the speed of a taxichip system is adjustable over a range of frequencies, with parallel bus transfer rates of 4 mbyte/s at the low end, and up to 17.5 mbyte/s at the high end. the flexible bus interface scheme of the taxichip set accepts bytes that are either 8, 9, or 10 bits wide. byte transfers can be data or command signaling. block diagram am7968 note: n can be 8, 9, or 10 bits; total of n + m = 12. 07370f-1 strobe (strb) acknowledge (ack clock (clk) data mode select (dms) test serial in (tserin) test/local select (tls) strobe & acknowledge oscillator and clock gen. serial interface shifter data encoder encoder latch input latch media interface (serout+) serial out + (seroutC) serial out C data command nm x1 x2
amd 2 am7968/am7969 block diagram (continued) am7969 note: n can be 8, 9, or 10 bits total of n + m = 12 output latch data decoder decoder latch shifter (x1) (x2) oscillator and clock gen. media interface pll clock generator byte sync logic nm data command (vltn) violation (dms) data mode select (cnb) catch next byte (igm) i-got-mine (clk) clock (dstrb) data strobe (cstrb) command strobe serial in+ (serin+) serial inC (serinC) 07370f-2 connection diagrams top view am7968 1 2 3 4 28 27 26 25 5 24 23 22 21 20 19 18 17 16 15 6 7 8 9 10 11 12 13 14 serout- serout+ strb ack di5 di4 di3 ci0 ci1 di9/ci2 di8/ci3 di7 di6 clk v cc2 (ecl) v cc1 (ttl) v cc3 (ttl) reset dms tls tserin di2 di1 di0 gnd1 (ttl) x1 x2 gnd2 (cml) 16 15 28 27 26 25 24 23 22 21 20 19 18 17 13 14 1 2 3 4 5 6 7 8 9 10 11 12 dips lcc/plcc note: pin 1 is marked for orientation. 07370f-3 ci1 di9/ci2 ack strb serout+ seroutC v cc2 (ecl) v cc1 (ttl) v cc3 (cml) reset dms tls tserin ci0 di7 di8/ci3 di5 di4 di3 di2 di1 di0 gnd1 (ttl) gnd2 (cml) x1 x2 clk di6 07370f-4
amd 3 am7968/am7969 connection diagrams (continued) top view am7969 1 2 3 4 28 27 26 25 5 24 23 22 21 20 19 18 17 16 15 6 7 8 9 10 11 12 13 14 do0 do1 do2 do3 do4 do5 do6 dstrb cstrb vltn co0 co1 do9/co2 igm reset v cc1 (ttl) serin+ serin- dms do7 cnb x2 x1 clk gnd2 (cml) v cc2 (cml) do8/co3 gnd1 (ttl) lcc/plcc note: pin 1 is marked for orientation. 07370f-5 cstrb vltn do3 do2 do1 do0 igm reset v cc1 (ttl) v cc2 (cml) serin+ serinC dms dstrb co1 co0 do4 do5 do6 do7 cnb x2 x1 gnd2 (cml) gnd1 (ttl) clk do8/co3 do9/c02 16 15 28 27 26 25 24 23 22 21 20 19 18 17 13 14 1 2 3 4 5 6 7 8 9 10 11 12 dips 07370f-6 logic symbols am7969 do n /co m vltn dstrb cstrb igm clk cnb dms reset x1 x2 12 v cc = power supply (3) gnd = ground (2) 2 12 2 am7968 ack clk tls dms reset x1 x2 serin+ serout+ strb di n /ci m v cc = power supply (2) gnd = ground (2) 07370f-7 07370f-8 tserin
amd 4 am7968/am7969 ordering information standard products amd standard products are available in several packages and operating ranges. the ordering number (valid combination) is formed by a combination of: am7968 dc temperature range c = commerical (0 c to +70 c) package type d = 28-pin ceramic dip (cd 028) j = 28-pin plastic leaded chip carrier (pl 028) speed option -125 = max serial encoded transmission rate is 125 mhz -175 = max serial encoded transmission rate is 175 mhz device number/description am7968 taxichip transmitter am7969 taxichip receiver valid combinations valid combinations list configurations planned to be supported in volume for this device. consult the local amd sales office to confirm availability of specific valid combinations and to check on newly released combinations. valid combinations am7969 am7968-125 am7969-125 am7968-175 am7969-175 dc, jc C125
amd 5 am7968/am7969 military ordering information cpl products pkg temps (tc) v cc cpl part number smd part number apl part number lcc C30 c to 125 c 4.5 v to 5.5 v am7968-125/lkc lcc C55 c to 125 c 4.75 v to 5.5 v 5962-9052701m3a am7968-125v/b3a dip C30 c to 125 c 4.5 v to 5.5 v am7968-125/dkc dip C55 c to 125 c 4.75 v to 5.5 v 5962-9052701mxa am7968-125v/bxa lcc C30 c to 125 c 4.5 v to 5.5 v am7969-125/lkc lcc C55 c to 125 c 4.75 v to 5.5 v 5962-9052801m3a am7969-125v/b3a dip C30 c to 125 c 4.5 v to 5.5 v am7969-125/dkc dip C55 c to 125 c 4.75 v to 5.5 v 5962-9052801mxa am7969-125v/bxa amd products for aerospace and defense applications are available in several packages and operating ranges. cpl (controlled products list) products are compliant with mil-std-883c requirements with exceptions for v cc or operating temperature. the order number (valid combination) is formed by a combination of: temperature range k = C30 c to 125 c m = C55 c to 125 c package type d = 28-pin ceramic dip (cd 028) l = 28-pin ceramic leadless chip carrier (cl 028) device number/description am7968 C taxichip transmitter (local mode only) am7969 C taxichip receiver (local mode only) am7968 am7969 /l k valid combinations valid combinations list configurations planned to be supported in volume for this device. consult the local amd sales office to confirm availability of specific valid combinations and to check on newly released combinations. c = controlled product list c group a tests group a tests consist of subgroups 1, 2, 3, 7, 8, 9, 10, 11. -125 speed option -125 = max serial encoded transmission rate is 125 mhz valid combinations
amd 6 am7968/am7969 pin description am7968 taxichip transmitter ack input-strobe acknowledge (ttl output) ack high signifies that the am7968 is ready to accept new data and command. the timing of acks response to strb depends on the condition of the input latch (in given clk cycle). if the input latch is empty, data is immediately stored and ack closely follows strb. if the input latch con- tains previously stored data when strb is asserted, ack is delayed until the next falling edge of clk. note that for ack to rise strb must maintain high for both of the above conditions. ci0 C ci1 parallel command in (ttl inputs) these two inputs accept parallel command information from the host system. if one or more command bits are logic 1, the command bit pattern is latched, encoded, and transmitted in place of any pattern on the data inputs. clk clock (ttl i/o) clk is an i/o pin that supplies the byte-rate clock refer- ence to drive all internal logic. when tls is connected to ground (local mode), clk is enabled as a free-running (byte-rate) clock output which runs at the crystal oscil- lator frequency; this output can be used to drive the x1 input of taxichip receivers or other system logic. in test mode clk becomes an input . in test mode 1 clk is a byte rate input and in test mode 2 it is a bit rate input. di0 C di7 parallel data in (ttl inputs) these eight inputs accept parallel data from the host system, to be latched, encoded and transmitted. di8/ci3 parallel data (8) in or command (3) in (ttl input) di 8 /ci 3 input is either data or command, depending upon the state of dms . di9/ci2 parallel data (9) in or command (2) in (ttl input) di 9 /ci 2 input is either data or command, depending upon the state of dms . dms data mode select (input) data mode select input determines the data pattern width. when it is wired to gnd, the am7968 transmitter will assume data to be eight bits wide, with four bits of command. when it is wired to v cc , the am7968 transmitter will assume data to be nine bits wide, with three bits of command. if dms is left floating (or termi- nated to 1/2 v cc ), the am7968 will assume data to be ten bits wide, with two bits of command. gnd1, gnd2 ground pins gnd1 is a ttl i/o ground and gnd2 is an internal logic and analog ground. reset pll reset (input) this pin is normally left open, but can be momentarily grounded to force the internal pll to reactivate lock. this allows for correction in the unlikely occurrence of pll lockup on application of power. reset has an internal pull-up resistor which causes it to float high when left unconnected (50 k ohm nominal). if this board is driven by a board reset signal, an open drain (or open collector) style output should be used to insure the high level signal is at v cc . serout+, seroutC differential serial data out (differential open emit- ter ecl outputs) these differential ecl outputs generate data at ecl voltage levels referenced to +5.0 v. when connected to appropriated pull down resistors, they are capable of driving 50- w terminated lines, either directly or through isolating capacitors. strb input strobe signal (ttl input) a rising edge on the strb input causes the data ( di0 C di9) or the command (ci0 C ci3) inputs to be latched into the am7968 transmitter. the strb signal is nor- mally taken low after ack has risen. tls test/local select (input) tls input determines the mode of operation. when tls is wired to gnd, the am7968 transmitter assumes a local mode connection to the media. it will output nrzi encoded data, and will enable its clk output driver. the tls pin should always be grounded during normal operation. when tls is wired to v cc (test mode 1),the serial data is nrz, clk becomes an input, and ack timing is modi- fied. this mode is only used for automatic test equip- ment (ate) testing at full speed. when this input is left unconnected, it floats to an inter- mediate level which puts the am7968 transmitter into its test mode 2. in test mode 2, the internal clock
amd 7 am7968/am7969 multiplier is switched out, and the internal logic is clocked directly from the clk pin. test mode 2 is in- cluded to ease automatic test equipment (a.t.e.) test- ing by making the internal logic of the transmitter synchronous to the external clock instead of the internal pll. tserin test serial input (pseudo ecl input) this pin is left unconnected in local mode operation. tserin can be used to input serial data patterns into the shifter in test mode 1 operation. v cc1 , v cc2 , v cc3 power supply v cc1 , v cc2 , and v cc3 are +5.0 volt nominal power sup- ply pins. v cc1 powers ttl i/o, v cc2 powers ecl and v cc3 powers internal logic and analog circuitry. x1, x2 crystal oscillator inputs (inputs) the two crystal input pins connect to an internal parallel mode oscillator which operates at the fundamental fre- quency of the external crystal. the byte rate matches the crystal frequency. during normal operation, the byte rate is set by the crystal frequency. alternatively, x1 can be driven by an external ttl fre- quency source. in multiple taxi systems this external source could be another am7968s clk output.
amd 8 am7968/am7969 am7969 taxichip receiver clk clock (ttl output) this is a free-running clock output which runs at the byte rate, and is synchronous with the serial input. it falls at the time that the decoder latch is loaded from the shifter, and rises at mid-byte. the clk output of the re- ceiver is not suitable as a frequency source for another taxi transmitter or receiver. it is intended to be used by the host system as a clock synchronous with the re- ceived data. cnb catch next byte input (ttl input) if this input is connected to the clk output, the receiver will be in the local mode, and each received byte will be captured, decoded and latched to the outputs. if the cnb input is high, it allows the am7969 receiver to capture the first byte after a sync. the am7969 re- ceiver will wait for another sync before latching the data out, and capturing another. if cnb is toggled low, it will react as if it had decoded a sync byte. in cascade mode, cnb input is typically connected to an upstream am7969s igm output. the first am7969 receiver in line will have its cnb input connected to v cc . for am7969-175 applications, an inverter is required between clk and cnb for speeds above 140 mhz. see figure 3 and timing specifications t47a, t47b, t48, and t49. co0 C co1 parallel command out (ttl output) these two outputs reflect the most recent command data received by the am7969 receiver. cstrb command data strobe (ttl output) the rising edge of this output signals the presence of new command data on the co0 C co3 lines. command bits are valid just before the rising edge of cstrb . dms data mode select (input) dms selects the data pattern width. when it is wired to gnd, the am7969 receiver will assume data to be eight bits wide, with four bits of command. when it is wired to v cc the am7969 receiver will assume data to be nine bits wide, with three bits of command. if dms is left floating (or terminated to 1/2 v cc ), the am7969 re- ceiver will assume data to be ten bits wide, with two bits of command . do0 C do7 parallel data out (ttl outputs) these eight outputs reflect the most recent data re- ceived by the am7969 receiver . do8/co3 parallel data (8) out or command (3) out (ttl output) do8/co3 output will be either a data or command bit, depending upon the state of dms . do9/co2 parallel data (9) out or command (2) out (ttl output) do9/co2 output will be either a data or command bit, depending upon the state of dms . dstrb output data strobe (ttl output) the rising edge of this output signals the presence of new data on the do0 C do9 lines. data is valid just be- fore the rising edge of dstrb. gnd1, gnd2 ground gnd1 is a ttl i/o ground, gnd2 is an internal logic and analog ground. igm i-got-mine (ttl output) this pin signals cascaded am7969 receivers that their upstream neighbor has captured its assigned data byte. igm falls at the mid-byte when the first half of a sync byte is detected in the shifter. it rises at mid-byte when it detects a non-sync pattern. during local mode opera- tion the igm signal is undefined. reset pll reset (input) this pin is normally left open, but can be momentarily grounded to force the internal pll to reactivate lock. this allows for correction in the unlikely occurance of pll lockup on application of power. reset has an internal pull-up resistor (50 k nominal) which causes it to float high when left unconnected. if this board is driven by a board reset signal, an open drain (or open collector) style output should be used to insure the high level signal is at v cc . serin+, serinC differential serial data in (ecl inputs) data is shifted serially into the shifter. the serin+ and serinC differential ecl inputs accept ecl voltage
amd 9 am7968/am7969 swings, which are referenced to +5.0 v. when serinC is grounded, the am7969 is put into test mode; serin+ becomes a single-ended ecl input, the pll clock gen- erator is bypassed, and x1 determines the bit rate (rather than the byte rate). both pins have internal pull down resistors which cause unterminated inputs to stay low. v cc1 , v cc2 power supply v cc1 and v cc2 are +5.0 volt nominal power supply pins. v cc1 powers ttl i/o, and v cc2 powers internal logic and analog circuitry. vltn violation (ttl output) the rising edge of this output indicates that a transmis- sion error has been detected. it changes state at the same time doi or coi change and will be followed by either dstrb or cstrb . this pin goes low when the next valid byte is decoded. x1, x2 crystal oscillator inputs (inputs) these two crystal input pins connect to an internal paral- lel/mode oscillator which oscillates at the fundamental frequency external crystal. during normal operation, the byte rate is set by the crystal frequency. alternatively, x1 can be driven by an external frequency source. in multiple taxi systems, this external source could be a taxi transmitters clk output or an external ttl fre- quency source.
amd 10 am7968/am7969 functional description system configuration the taxichip system provides a means of connecting parallel data systems over a serial link (figure 2). in local mode (normal operation mode) each tx/rx pair is connected over a serial link which can be a fiber optic or copper media (figure 3). the am7968 transmitter accepts inputs from a sending host system using a simple strb/ack handshake. parallel bits are saved by the am7968s input latch on the rising edge of a strb input. the input latch can be updated on every clk cycle; if it still contains previously stored data when a second strb pulse arrives, data is stored in the input latch, and the second ack response is delayed until the next clk cycle. the inputs to an am7968 transmitter can be either data or command and may originate from two different parts of the host system. a byte cycle may contain data or command, but not both. data represents the normal data channel message traffic between host systems. commands can come from a communication control section of the host system. commands occur at a rela- tively infrequent rate but have priority over data. exam- ples include communication specific commands such as request-to-send or clear-to-send; or application specific commands such as message- address-follows, message-type-follows, initialize your system, error, retransmit, halt, etc. the am7968 transmitter switches between data and command by examining command input patterns. all 0s on command input pins cause information on the am7968s data input pins to be latched into the device on the rising edge of strb . all other command patterns cause a command symbol to be sent in response to an input strobe. the pattern on the data inputs is ignored when a command symbol is sent. in either case, if there is no strb before the next byte boundary, a sync sym- bol will be transmitted. the sync pattern maintains link synchronization and provides an adequate signal transi- tion density to keep the receiver phase-locked-loop (pll) circuits in lock. it was chosen for its unique pattern which never occurs in any data or command mes- sages. this feature allows sync to be used to establish byte boundaries. the sync pattern utilized by taxichip set keeps the automatic gain control (agc) fiber-optic transceiver cir- cuits in their normal range because the pattern has zero dc offset. the am7969 receiver detects the difference between data and command patterns and routes each to the proper output latch. when a new data pattern enters the output latch, dstrb is pulsed and command information remains unchanged. if a command pattern is sent to the output latch or if sync is received, cstrb is pulsed and data outputs remain in their previous state. reception of a sync pattern clears the command outputs to all 0s, since sync is a legal command. noise-induced bit errors can distort transmitted bit pat- terns. the am7969 receiver logic detects most noise- induced transmission errors. invalid bit patterns are recognized and indicated by the assertion of the viola- tion ( vltn ) output pin. this signal rises to a logic 1 state at the same time that data or command outputs change and remains high until a valid pattern is detected by the data decoder. the error detection method used in the receiver cannot identify bit errors which transform one valid command or data pat- tern to another. fault-sensitive systems should use ad- ditional error checking mechanisms to guarantee message integrity. am7968 transmitter the transmitter accepts messages from its parallel in- put pins (command or data). once latched into an am7968, a parallel message is encoded, serialized, and shifted out to the serial link. the idle time between trans- mitted bytes (evident by lack of strb) is filled with sync bytes. am7969 receiver receivers accept differential signals on the serin+/ serinC input pins. this information, previously encoded by an am7968 transmitter, is loaded into a decoder. when serial patterns are received, they are decoded and routed to the appropriate outputs. if the received message is a command, it is stored in the output latch, appears at the command output pins, and cstrb is pulsed; data output pins continue holding the last data byte and dstrb stays inactive. if a data message fol- lows the reception of a command, command output pins continue holding the previous command byte and cstrb stays inactive. the command outputs will retain their states until another command signal is received (sync is considered to be a valid command which, when decoded, sets command outputs to 0 and issues a re- sulting cstrb ). byte width the taxichip set has twelve parallel interface pins which are designated to carry either command or data bits. the data mode select ( dms ) pin on each chip can be set to select one of three modes of operation: eight data and four command bits, nine data and three com- mand, or ten data and two command. this allows the system designer to select the byte-width which best suits system needs.
amd 11 am7968/am7969 am7968 encoder/am7969 decoder to guarantee that the am7969s pll can stay locked onto an incoming bit stream, the data encoding scheme must provide an adequate number of transitions in each data pattern. this implies a limit on the maximum time allowed between transitions. the taxichip set encod- ing scheme is based on the ansi x3t9.5 (fddi) com- mittees 4-bit/5-bit (4b/5b) code. an ansi x3t9.5 system used an 8-bit parallel data pat- tern. this pattern is divided into two 4-bit nibbles which are each encoded into a 5-bit symbol. of the thirty-two patterns possible with these five bits, sixteen are chosen to represent the sixteen input data patterns. some of the others are used as command symbols. those re- maining represent invalid patterns that fail either the run-length test or dc balance tests. transmitters in 8-bit mode use two 4b/5b encoders to encode eight data bits into a 10-bit pattern. in 9-bit mode, transmitters use one 5b/6b encoder and one 4b/5b encoder to code nine data bits into an 11-bit pat- tern. in 10-bit mode, two 5b/6b encoders are used to change ten bits of data into a 12-bit pattern (see tables 1 and 2 for encoding patterns). the am7968 transmitter further encodes all symbols using nrzi (non return to zero, invert on ones). nrzi represents a 1 by a transition and a 0 by the lack of transition. in this system a 1 can be a high-to-low or low-to-high transition. this combination of 4b/5b and nrzi encoding ensures at least two transitions per symbol and permits a maximum of three consecutive non-transition bit times. the am7969 then uses the same method to decode incoming symbols so that the whole encoding/decoding process is transparent to the user. most serially transmitted data patterns with this code will have the same average amount of high and low times. this near dc balance minimizes pattern-sensi- tive decoding errors which are caused by jitter in ac- coupled systems. operational modes in normal operational mode, a single transmitter/ receiver pair is used to transfer 8, 9, or 10 bits of parallel data over a private serial link. (on the am7968, the tls pin is tied to ground and tserin is left unconnected). on the am7969, cnb must be connected to the clk output. the am7969 receiver continuously deserial- izes the incoming bit stream, decodes the resulting pat- terns, and saves parallel data at its output latches (see figure 3). local mode provides a fast and efficient parallel throughout because data can be transferred on every clock cycle. on the other hand, it is not necessary for the host to match the byte rate set by the transmitters crys- tal oscillator; the am7968 automatically sends a sync pattern during each clock cycle in which no new data or command messages are being transmitted. cascade mode (for C125 only) for very wide parallel buses, taxi receivers (commer- cial temperature parts only) can be cascaded. the am7969 receivers all have their serin+ and serinC pins connected to the media (or an optical data link). igm of each am7969 is connected to cnb of its down- stream neighbor or is left unconnected on the receiver farthest downstream. cnb of the first receiver is tied high, making this device the only receiver in the chain that can act on the first non-sync pattern in a message (see below). each taxichip receiver monitors the serial link and a special acknowledgment scheme is used to direct sym- bols into each of the am7969s. when a catch-next- byte ( cnb ) input is high, the receiver will capture the next non-sync symbol from the serial link. at this point, the device forces its i-got-mine ( igm ) pin high to tell the downstream receiver to capture the next symbol. the receiver then waits for the sync symbol or for its cnb to be set low before transferring the message to its output latch. igm is forced low whenever a sync byte is detected or when cnb goes low. this igm - cnb exchange continues down the chain until the last receiver captures its respective byte. the next byte to appear on the serial link will be a sync symbol which is detected by all of the cascaded am7969s. on the follow- ing clock cycle their messages are transferred to the output latch of each device and sent to the receiving host. igm pins on all receivers are also set low when the first half of the sync symbol is detected. asynchronous operation inputs to the am7968 transmitter input latch can be asynchronous to its internal clock. data strb will latch data into the am7968 transmitter and an internal clock will transfer the data to the encoder latch at the first byte boundary. data can be entered at any rate less than the maximum transfer rate without regard to actual byte boundaries. as data rates approach the taxi byte rate, care must be taken to insure that the 2 byte fifo inside taxi transmitter is not over filled. strb/ack handshake will assure that every byte is transferred correctly. at higher byte rates, where delays and setup/hold times make the strb/ack handshake impractical, strb should be synchronized with clk . synchronous operation the transmitter may be strobed synchronous by tying the strobe to the input clock. when doing this a provision should be make to inhibit the strobe periodically to en- sure proper byte alignment. in the absence of a strobe, syncs will be transmitted on the serial link which will al- low the receiver to re-align the byte boundaries. in addi- tion it is essential that the delay between the falling edge
amd 12 am7968/am7969 of the internal byte clock (clk) and the rising edge of strobe does not violate t bb specification shown in the switching characteristics section. the internal byte clock controls the flow of data from the input register through the shift register. the falling edge of the internal byte clock delineates the end of one byte from the start of the next. due to various tolerances in the pll, the period of the internal byte clock may vary slightly. this effect may cause a shift in the location of the byte boundary with respect to the falling edge of the clock. this variation may move the byte boundary and therefore creates a window during which the part should not be strobed. this window called the t 6 window, is shown in the figure below. if the part is strobed during the t 6 window data will not be lost however, a sync may be added and the transmitter latency will be increased by one byte time. clk C9/8(t1/n) + 9 ns 20 ns strobe stayout area (t 6 window) 07370f-9 nominal byte boundary sync acquisition in case of errors which cause am7969 receivers to lose byte/symbol sync, and on power-up, internal logic de- tects this loss-re-acquisition of sync and modifies the clk output. clk output is actually a buffered version of the signal which controls data transfers inside the am7969 receiver on byte boundaries. byte boundaries move when the am7969 receiver loses, and re- acquires sync. to protect slave systems (which may use this output as a clock synchronous with the incoming data) from having clocks which are too narrow, the out- put logic will stretch an output pulse when the pulse would have been less than a byte-time long. the data being processed just prior to this re-acquisition of sync will be lost. the sync symbol, and all subsequent data will be processed correctly. taxi user test modes tls input can be used to force the am7968 transmitter into either of the two test modes. if tls is open or termi- nated to approximately v cc /2 (test mode 2), the internal vco is switched out and everything is clocked directly from the clk input. the serial output data rate will be at the clk bit rate and not at 10x, 11x, or 12x, as is the case in normal operation. test mode 2 will allow testing of the logic in the latches, encoder, and shifter without having to first stabilize the pll clock multiplier. in test mode 1 ( tls wired to v cc ), the pll is enabled and the chip operates normally, except that the output is an nrz stream (clk is an input & ack function is slightly modi- fied). this will allow testing of all functions at full rate without needing to perform match loop tests to accom- modate the data inversion characteristics of nrzi. differential serin+/serinC inputs can be used to force the am7969 receiver into its test mode. this will allow testing of the logic in the latches, decoder, and shifter without having to first stabilize the the pll. if serinC is tied to ground, the internal v co is switched out and x 1 becomes the internal bit rate clock. the serial data rate will be at the clk bit rate, not at 10x, 11x, or 12x, as is the case in normal operation. in this mode, serin+ becomes a single-ended serial data input with nominal 100k ecl threshold voltages (referenced to +5 volts). these test mode switches make the parts determinate, synchronous systems, instead of statistical, asynchro- nous ones. an automatic test system will be able to clock each part through the functional test patterns at any rate or sequence that is convenient. after the logic has been verified, the part can be put back into the nor- mal mode, and the pll functions verified knowing that the rest of the chip is functional.
amd 13 am7968/am7969 oscillator the am7968 and am7969 contain an inverting amplifier intended to form the basis of a parallel mode oscillator. the design of this oscillator considered several factors related to its application. the first consideration is the desired frequency accu- racy. this may be subdivided into several areas. an os- cillator is considered stable if it is insensitive to variations in temperature and supply voltage, and if it is unaffected by individual component changes and aging. the design of the taxichip set is such that the degree to which these goals are met is determined primarily by the choice of external components. various types of crystal are available and the manufacturers literature should be consulted to determine the appropriate type. for good temperature stability, zero temperature coefficient capacitors should be used (type npo). the mechanism by which a crystal resonates is electro- mechanical. this resonance occurs at a fundamental frequency (1st harmonic) and at all odd harmonics of this frequency (even harmonic resonance is not me- chanically possible). unless otherwise constrained, crystal oscillators operate at their fundamental frequencies. a typical crystal specification for use in this circuit is: fundamental frequency 3.3 mhzC17.5 mhz 0.1% resonance: mode parallel load capacitor (correlation) 30 pf operating temperature range 0 c to 70 c temperature stability 100 ppm drive level (correlation) 2 mw effective series resistance 25 w (max) holder type low profile aging for 10 years 10 ppm it is good practice to ground the case of the crystal to eliminate stray pick-up and keep all connections as short as possible. reset am7968 or, am7969 x1 x2 c c power on reset (optional) c* = 220 pf for 4.0C12.5 mhz crystal, 150 pf for a 12.5C17.5 mhz crystal. *c determined by crystal specifications and trace capacities. values shown are typical. 07370f-10 figure 1. connections for 4.0 mhzC17.5 mhz
amd 14 am7968/am7969 table 1. taxichip encoder patterns 4-bit 5-bit 5-bit 6-bit hex binary encoded hex binary encoded data data symbol data data* symbol 0 0000 11110 00 00000 110110 1 0001 01001 01 00001 010001 2 0010 10100 02 00010 100100 3 0011 10101 03 00011 100101 4 0100 01010 04 00100 010010 5 0101 01011 05 00101 010011 6 0110 01110 06 00110 010110 7 0111 01111 07 00111 010111 8 1000 10010 08 01000 100010 9 1001 10011 09 01001 110001 a 1010 10110 0a 01010 110111 b 1011 10111 0b 01011 100111 c 1100 11010 0c 01100 110010 d 1101 11011 0d 01101 110011 e 1110 11100 0e 01110 110100 f 1111 11101 0f 01111 110101 10 10000 111110 11 10001 011001 12 10010 101001 13 10011 101101 14 10100 011010 15 10101 011011 16 10110 011110 17 10111 011111 18 11000 101010 19 11001 101011 1a 11010 101110 1b 11011 101111 1c 11100 111010 1d 11101 111011 1e 11110 111100 1f 11111 111101 * note: hex data is parallel input data which is represented by the 4- or 5-bit binary data listed in the column to the immediate right of hex data. binary bits are listed from left to right in the following order. 8-bit mode: d7, d6, d5, d4, (4-bit binary), and d3, d2, d1, d0, (4-bit binary) 9-bit mode: d8, d7, d6, d5, d4, (5-bit binary), and d3, d2, d1, d0, (4-bit binary) 10-bit mode: d8, d7, d6, d5, d4, (5-bit binary), and d9,d3, d2, d1, d0, (5-bit binary) serial bits are shifted out with the most significant bit of the most significant nibble coming out first. 4b/5b encoder scheme 5b/6b encoder scheme
amd 15 am7968/am7969 table 2. taxichip command symbols encoded hex binary symbol mnemonic hex binary 8-bit mode 0 0000 xxxxx xxxxx data no change no change (note 2) (note 2) no strb no strb 11000 10001 jk (8-bit sync) 0 0000 (note 1) (note 1) 1 0001 11111 11111 i i 1 0001 2 0010 01101 01101 tt 2 0010 3 0011 01101 11001 ts 3 0011 4 0100 11111 00100 i h 4 0100 5 0101 01101 00111 tr 5 0101 6 0110 11001 00111 sr 6 0110 7 0111 11001 11001 ss 7 0111 8 (note 3) 1000 00100 00100 hh 8 1000 9 1001 00100 11111 hi 9 1001 a (note 3) 1010 00100 00000 hq a 1010 b 1011 00111 00111 rr b 1011 c 1100 00111 11001 rs c 1100 d (note 3) 1101 00000 00100 qh d 1101 e (note 3) 1110 00000 11111 q i e 1110 f (note 3) 1111 00000 00000 qq f 1111 9-bit mode 0 000 xxxxxx xxxxx data no change no change (note 2) (note 2) no strb no strb 011000 10001 lk (9-bit sync) 0 000 (note 1) (note 1) 1 001 111111 11111 i i 1 001 2 010 011101 01101 t t 2 010 3 011 011101 11001 ts 3 011 4 100 111111 00100 i h 4 100 5 101 011101 00111 tr 5 101 6 110 111001 00111 sr 6 110 7 111 111001 11001 ss 7 111 10-bit mode 0 00 xxxxxx xxxxxx data no change no change (note 2) (note 2) no strb no strb 011000 100011 lm (10-bit sync) 0 00 (note 1) (note 1) 1 01 111111 111111 i i 1 01 2 10 011101 011101 t t 2 10 3 11 011101 111001 t s 3 11 notes: 1. command pattern sync cannot be explicitly sent by am7968 transmitter with any combination of inputs and strb, but is used to pad between user data. 2. a strobe with all os on the command input lines will cause data to be sent. see table 1. 3. while these commands are legal data and will not disrupt normal operation if used occasionally, they may cause data errors if grouped into recurrent fields. normal pll operation cannot be guaranteed if one or more of these commands is continuously repeated. command input command output am7969 receiver am7968 transmitter
amd 16 am7968/am7969 am7968 transmitter functional block description (refer to page 1) crystal oscillator/clock generator the serial link speed is derived from a master frequency source (byte rate). this source can either be the built-in crystal oscillator, or a clock signal applied through the x 1 pin. this signal is buffered and sent to the clk out- put when am7968 transmitter is in local mode. clk (input is multiplied by ten (8-bit mode), eleven (9-bit mode), or twelve (10-bit mode), using the internal pll to create the bit rate. the working frequency can be varied between 3.3 mhz and 17.5 mhz. the crystal frequency required to achieve the maximum 175 mbaud on the serial link, and the resultant usable data transfer rate will be: crystal am7968-125 input and am7969-125 internal mode frequency maximum parallel throughput divide ratio 8-bit 12.50 mhz 80 ns/pattern (100 mbit/sec) 125/10 9-bit 11.36 mhz 88 ns/pattern (102 mbit/sec) 125/11 10-bit 10.42 mhz 96 ns/pattern (104 mbit/sec) 125/12 crystal am7968-175 input and am7969-175 internal mode frequency maximum parallel throughput divide ratio 8-bit 17.50 mhz 57.1 ns/pattern (140 mbit/sec) 175/10 9-bit 15.90 mhz 62.8 ns/pattern (143 mbit/sec) 175/11 10-bit 14.58 mhz 68.5 ns/pattern (145 mbit/sec) 175/12 input latch the am7968s input latch accommodates asynchro- nous strobing of data and command by being divided into two stages. if strb is asserted when both stages are empty, data or command bits are transferred directly to the second stage of the input latch and ack rises shortly after strb . this pattern is now ready to move to the encoder latch at the next falling edge of clk . an input pattern is strobed into the first stage of the input latch only when the second stage is busy (contains previously stored data). the transmitter will be busy when strb is asserted a second time in a given clk cycle. contents of the first stage are not protected from subsequent strbs within the same clk cycle. at the falling edge of clk , previously stored data is transferred from the second stage to the encoder latch and the new data is clocked into the second stage of the input latch. if in local mode, ack will rise at this time. encoder latch input to the encoder latch is clocked by an internal sig- nal which is synchronous with the shifted byte being sent on the serial link. whenever a new input pattern is strobed into the input latch, the data is transferred to the encoder latch at the next opportunity. data encoder encodes twelve data inputs (8, 9, 10 data bits or 4, 3, 2 command inputs) into 10, 11, or 12 bits. the command data inputs control the transmitted symbol. if all com- mand inputs are low, the symbol for the data bits will be sent. if command inputs have any other pattern then the symbol representing that command will be transmitted. shifter the shifter is parallel-loaded from the encoder at the first available byte boundary, and then shifted until the next byte boundary. the shifter is being serially loaded at all times. as data is being shifted out of the transmit- ter, the shifter fills from the lsb. if parallel data is avail- able at the end of the byte, it is parallel-loaded into the shifter and begins shifting out during the next clock cy- cle. otherwise, the serially loaded data fills the next byte. the serial data which loads into the shifter is gen- erated by an internal state machine which generates a repeating sync pattern. media interface the media interface is differential ecl, referenced to +5 v. it is capable of driving lines terminated with 50 w to (v cc - 2.0) volts.
amd 17 am7968/am7969 am7969 receiver functional block description (refer to page 1) crystal oscillator/clock generator the data recovery pll in the am7969 must be supplied with a reference frequency at the expected byte rate of the data to be recovered. the source of this frequency can either be the built-in crystal oscillator, or an exter- nal clock signal applied through the x 1 pin. the refer- ence frequency source is then multiplied by ten (8-bit mode), eleven (9-bit mode) or twelve (10-bit mode) us- ing an internal pll. media interface serin+, serinC inputs are to be driven by differential ecl voltages, referenced to +5 v. serial data at these inputs will serve as the reference for pll tracking. pll clock generator a pll clock recovery loop follows the incoming data and allows the encoded clock and data stream to be de- coded into a separated clock and data pattern. it uses the crystal oscillator and clock generator to predict the expected frequency of data and will track jittered data with a characteristically small offset frequency. shifter the shifter is serially loaded from the media interface, using the bit clock generated by pll. byte sync logic the incoming data stream is a continuous stream of data bits, without any significant signal which denotes byte boundaries. this logic will continuously monitor the data stream, and upon discovering the reserved code used for am7969 receiver sync, will initialize a synchronous counter which counts bits, and indicates byte boundaries. the logic signal that times data transfers from the shif- ter to the decoder latch is buffered and sent to the clk output. clk output from the receiver is not suitable as a frequency source for another taxi transmitter or re- ceiver. it is intended to be used by the host system as a clock synchronous with the received data. this output is synchronous with the byte boundary and is synchronous with the receivers internal byte clock. byte sync logic is responsible for generating the inter- nal strobe signals for parallel output latches. it also generates the igm (i-got-mine) signal in test mode when the first byte after a sync symbol is transferred. parallel outputs are made on a byte boundary, after cnb falls, or when sync is detected. the i-got-mine ( igm ) signal will fall when the first half of a sync is detected in the shifter or when cnb goes low. it will remain low until the first half of a non-sync byte is detected in the shifter, whereupon it will rise (as- suming that the cnb input is high). a continuous stream of normal data or command bytes will cause igm to go high and remain high. a continuous stream of syncs will cause igm to stay low. igm will go high during the byte before data appears at the output. this feature could be used to generate an early warning of in- coming data. decoder latch data is loaded from the shifter to this latch at each symbol/byte boundary. it serves as the input to the data decoder. data decoder decodes ten, eleven, or twelve data inputs into twelve outputs. in 8-bit mode, data is decoded into either an 8-bit data pattern or a 4-bit command pattern. in 9-bit mode, data is decoded into either a 9-bit data pattern or a 3-bit command pattern. in 10-bit mode, data is de- coded into either a 10-bit data pattern or a 2-bit com- mand pattern. the decoder separates data symbols from command symbols, and causes the appropriate strobe output to be asserted. parallel output latch output latch will be clocked by the byte clock, and will reflect the most recent data on the link. any data pattern will be latched to the data outputs and will not affect the status of the command outputs. likewise, any com- mand pattern will be latched to the command outputs without affecting the state of the data outputs. any data transfer, either data or command will be syn- chronous with an appropriate output strobe. however, there will be cstrbs when there is no active data on the link, since sync is a valid command code. any pattern which does not decode to a valid command or data pattern is flagged as a violation. the output of the decoder during these violations is indeterminate and will result in either a cstrb or dstrb output when the indeterminate pattern is transferred to the output latch.
amd 18 am7968/am7969 note: n can be 8, 9, or 10 bits of parallel data; total of n + m = 12. 07370f-11 m n data signals data source message transfer control logic command source command signals command signals command destination data path control logic data destination n m transmission media ack strb cstrb vltn dstrb data signals am7968 am7969 figure 2. taxichip system block diagram
amd 19 am7968/am7969 notes: 1. dms = gnd = 8 bit mode tls = gnd = local mode pin 11 = dont connect = local mode 2. dms = v cc = 9 bit mode tls = gnd = local mode pin 11 = dont connect = local mode 3. two 8-bit local mode systems in parallel will result in an effective data rate of 200 mbps. 4. use inverter for operation above 140 mhz only. *alternatively, the x1 inputs may be driven by external ttl frequency sources. figure 3. taxichip system in local mode (note 1) serout+ seroutC strb ack clk taxi tx #1 tls dms x1 x2 message transfer control logic data source command source 8 3.3 mhz to 17.5 mhz 8 3.3 mhz to 17.5 mhz x1 x2 dms clock serin+ serinC cnb dstrb igm vltn cstrb taxi rx #1 4 do0C do7 co0 C co3 data destination command destination data path control logic message transfer control logic command source data source 9 3 serout+ seroutC strb ack clk tls dms x1 x2 taxi tx #2 to other stages serin+ serinC x1 x2 dms clock 3.3 mhz to 17.5 mhz igm vltn cnb dstrb do0 C do8 co0 C co2 cstrb 9 3 data destination command destination data path control logic taxi rx #2 4 * (note 4) (note 1) (note 2) (note 4) * * 07370f-12 di0 C di7 ci0 C ci3 di0 C di8 ci0 C ci2
amd 20 am7968/am7969 serinC serin+ serinC serin+ serinC serin+ crystal osc rx1 am7969 primary rx v cc cnb clk x2 x1 igm dms rx2 am7969 igm dms rx3 am7969 igm dms cnb cnb x1 x1 x2 x2 n/c from serial media 07370f-13 figure 4. cascaded receiver clock connections (commercial C125 only)
amd 21 am7968/am7969-125 am7968/am7969-125 absolute maximum ratings storagetemperature C65 c to +150 c . . . . . . . . . . . . ambient temperature under bias C55 c to +125 c . . . . . . . . . . . . . . . . . . . supply voltage to ground potential continuous C0.5 v to +7.0 v . . . . . . . . . . . . dc voltage applied to outputs C0.5 v to v cc max . . . . . . . . . . . . . . . . . . . . . dc input voltage C0.5 v to +5.5 v . . . . . . . . . . . . . . . dc output current 100 ma . . . . . . . . . . . . . . . . . . . dc input current C30 ma to +5.0 ma . . . . . . . . . . . . . stresses above those listed under absolute maximum rat- ings may cause permanent device failure. functionality at or above these limits is not implied. exposure to absolute maxi- mum ratings for extended periods may affect device reliability. operating ranges commercial (c) devices temperature (t a )0 c to +70 c . . . . . . . . . . . . . . . . . supply voltage (v cc ) +4.5 v to +5.5 v . . . . . . . . . . . . operating ranges define those limits between which the func- tionality of the device is guaranteed.
amd 22 am7968/am7969-125 dc characteristics over operating range unless otherwise specified am7968-125 taxichip transmitter parameter symbol parameter description test conditions (note 1) min max unit bus interface signals: di0Cdi7, di8/ci3, di9/ci2, ci0Cci1, strb, ack, clk v oh1 output high voltage v cc = min, i oh = C1 ma 2.4 v ack v in = 0 or 3 v v oh2 output high voltage v cc = min, i oh = C3 ma 2.4 v clk v in = 0 or 3 v v ol output low voltage v cc = min, i ol = 8 ma 0.45 v ack, clk v in = 0 or 3 v v ih input high voltage v cc = max (note 9) 2.0 v v il input low voltage v cc = max (note 9) 0.8 v v i input clamp voltage v cc = min i in = C18 ma C1.5 v i il input low current v cc = max, v in = 0.4 v C400 m a i ih input high current v cc = max, v in = 2.7 v 50 m a i i input leakage current v cc = max, all inputs 50 m a v in = 5.5 v except clk clk input 150 m a i sc output short circuit (note 4) C15 C85 ma current ack, clk serial interface signals: serout+, seroutC v oh output high voltage v cc = min ecl load v cc v cc v C1.025 C0.88 v ol output low voltage v cc = min ecl load v cc v cc v C1.81 C1.62 miscellaneous signals: x1, v cc1 , v cc2 , v cc3 v ihx input high voltage x1 2.0 v v ilx input low voltage x1 0.8 v i ilx input low current x1 v in = 0.45 v C900 m a i ihx input high current x1 v in = 2.4 v +600 m a i cc supply current serout = ecl load, dms = 0 v cc1 = v cc2 = v cc3 = max pin v cc1 (ttl) 20 ma pin v cc2 (ecl) 45 ma pin v cc3 (cml) 200 ma *see notes following end of switching characteristics tables.
amd 23 am7968/am7969-125 am7969-125 taxichip receiver parameter symbol parameter description test conditions (note 1) min max unit bus interface signals: do0Cdo7, do8/co3, do9/co2, co0Cco1, dstrb, cstrb, igm, clk, cnb, vltn v oh output high voltage v cc = min, i oh = C1 ma 2.4 v v in = 0 or 3 v v ol output low voltage v cc = min, i ol = 8 ma 0.45 v v in = 0 or 3 v v ih input high voltage v cc = max (note 9) 2.0 v v il input low voltage v cc = max (note 9) 0.8 v v i input clamp voltage v cc = min, i in = C18 ma C1.5 v i il input low current v cc = max, v in = 0.4 v C400 m a i ih input high current v cc = max, v in = 2.7 v 50 m a i i input leakage current v cc = max, v in = 5.5 v 50 m a i sc output short circuit C15 C85 ma current (note 4) serial interface signals: serin+, serinC v ihs input high voltage (notes 9, 21) v cc v cc v serin+ C1.165 C0.88 v ils input low voltage (notes 9, 21) v cc v cc v serin+ C1.81 C1.475 v tht test mode threshold v cc = max 0.25 v serinC v dif differential input voltage 0.3 1.1 v v icm input common mode (note 6) 3.05 v cc v voltage C0.55 i il input low current v cc = max, v in = v cc C1.81 v 0.5 m a i ih input high current v cc = max, 220 m a v in = v cc C0.88 v miscellaneous signals: x1, v cc1 , v cc2 v ihx input high threshold x1 2.0 v v ilx input low threshold x1 0.8 v i ilx input low current x1 v in = 0.45 v C900 m a i ihx input high current x1 v in = 2.4 v +600 m a i cc supply current v cc1 = v cc2 = max pin v cc1 (ttl) 50 ma dms = 0 v pin v cc2 (cml) 300 ma
amd 24 am7968/am7969-125 switching characteristics (note 20) am7968-125 taxichip transmitter (notes 10, 13, 22) parameter no. symbol parameter description test conditions min max units bus interface signals: di0Cdi7, di8/ci3, di9/ci2, ci0Cci1, strb, ack, clk 1t p clk period 8n 25n ns 2t pw clk pulse width high 30 ns 3t pw clk pulse width low 30 ns 4t pw strb pulse width high (note 7) 15 ns 5t pw strb pulse width low 15 ns 6t bb internal byte boundary to clk 20 ns (note 11) 9t s dataCstrb setup time 5 ns 10 t h dataCstrb hold time 15 ns 11 t h ack - to strb hold (note 8) ttl output load 0 ns 12 t h ack to strb - hold ttl output load 0 ns 13 t pd strb - to ack - (note 18) ttl output load 40 ns 14 t pd strb to ack ttl output load 23 ns 15 t pd clk to ack - (note 18) ttl output load ns serial interface signals: serout+, seroutC (note 2) 22 t sk = serout skew ecl output load C200 +200 ps 23 t r = serout output rise time ecl output load .45 2 ns 24 t f = serout output fall time ecl output load .45 2 ns 26 t pw = serout pulse width low ecl output load ns 27 t pw = serout pulse width high ecl output load ns miscellaneous signals: x1 (note 15) 29 t pw x1 pulse width high (note 12) ttl output load on clk 35 ns 30 t pw x1 pulse width low (note 12) ttl output load on clk 35 ns 32 t pd x1 - to clk - ttl load 32 ns 33 t pd x1 to clk ttl load 32 ns 3t 1 n + 33 t 1 n C 5% t 1 n + 5% t 1 n + 5% t 1 n C 5% C9t 1 8n +9
amd 25 am7968/am7969-125 am7969-125 taxichip receiver (notes 13, 14, 22) parameter no. symbol parameter description test conditions min max unit bus interface signals: do0Cdo7,do8/co3,do9/co2,co0Cco1,dstrb,cstrb, igm,clk,cnb,vltn 35 t p clk period (note 24) 8n 25n ns 36 t pd data valid to strb - delay ttl output load ns 37 t pd clk to strb - ttl output load ns 38 t pd clk - to strb ttl output load ns 38a t pd strb - to clk - (note 23) ttl output load ns 39 t pd clk to data valid delay ttl output load ns 40 t pw strb pulse width high ttl output load ns 41 t pw clk pulse width high ttl output load ns 42 t pw clk pulse width low ttl output load ns 43 t pd serin to clk delay ttl output load ns 44 t pd clk - to igm ttl output load ns 45 t pd clk - to igm - ttl output load ns 46 t pd cnb to igm ttl output load 20 ns 47 t s cnb - to clk - setup time ns (note 5) 47a t s cnb to clk - setup time ns (note 19) 48 t h cnb to clk - hold ns 49 t pw cnb pulse width low ns serial interface signals: serin+, serinC 57 t j = serin peak to peak input jitter 5 ns tolerance (note 16) miscellaneous signals: x1 (note 15) 60 t pw x1 pulse width high 35 ns 61 t pw x1 pulse width low 35 ns 2t 35 n - 2t 35 n 2t 35 n +15 t 35 n C7 5t 35 2n 5t 35 n C15 5t 35 n C15 t 35 2n +17 2t 35 n +26 2t 35 n +7 2t 35 n +10 C32 t 35 n C31 - t 35 n + - 23 5t 35 n 2t 35 n +5 2t 35 n 3t 35 n C14
amd 26 am7968/am7969-175 (page intentionally left blank)
amd 27 am7968/am7969-175 am7968/am7969-175 absolute maximum ratings storagetemperature C65 c to + 50 c . . . . . . . . . . . . ambient temperature under bias C55 c to +125 c . . . . . . . . . . . . . . . . . . . supply voltage to ground potential continuous C0.5 v to +7.0 v . . . . . . . . . . . . dc voltage applied to outputs C0.5 v to v cc max . . . . . . . . . . . . . . . . . . . . . dc input voltage C0.5 v to +5.5 v . . . . . . . . . . . . . . . dc output current +100 ma . . . . . . . . . . . . . . . . . . . dc input current C30 ma to +5.0 ma . . . . . . . . . . . . . stresses above those listed under absolute maximum rat- ings may cause permanent device failure. functionality at or above these limits is not implied. exposure to absolute maxi- mum ratings for extended periods may affect device reliability. operating ranges commercial (c) devices temperature (t c )0 c to +70 c . . . . . . . . . . . . . . supply voltage (v cc ) +4.5 v to +5.5 v . . . . . . . . . operating ranges define those limits between which the func- tionality of the device is guaranteed.
amd 28 am7968/am7969-175 dc characteristics over operating range unless otherwise specified am7968-175 taxichip transmitter parameter symbol parameter description test conditions (note 1) min max unit bus interface signals: di0Cdi7, di8/ci3, di9/ci2, ci0Cci1, strb, ack, clk v oh1 output high voltage v cc = min, i oh = C1 ma 2.4 v ack v in = 0 or 3 v v oh2 output high voltage v cc = min, i oh = C3 ma 2.4 v clk v in = 0 or 3 v v ol output low voltage v cc = min, i ol = 8 ma 0.45 v ack, clk v in = 0 or 3 v v ih input high voltage v cc = max (note 9) 2.0 v v il input low voltage v cc = max (note 9) 0.8 v v i input clamp voltage v cc = min i in = C18 ma C1.5 v i il input low current v cc = max, v in = 0.4 v C400 m a i ih input high current v cc = max, v in = 2.7 v 50 m a i i input leakage current v cc = max, all inputs 50 m a v in = 5.5 v except clk clk input 150 m a i sc output short circuit (note 4) C15 C85 ma current ack, clk serial interface signals: serout+, seroutC v oh output high voltage v cc = min ecl load v cc v cc v C1.025 C0.88 v ol output low voltage v cc = min ecl load v cc v cc v C1.81 C1.62 miscellaneous signals: x1, v cc1 , v cc2 , v cc3 v ihx input high voltage x1 2.0 v v ilx input low voltage x1 0.8 v i ilx input low current x1 v in = 0.45 v C900 m a i ihx input high current x1 v in = 2.4 v +600 m a i cc supply current serout = ecl load, dms = 0 v cc1 = v cc2 = v cc3 = max pin v cc1 (ttl) 20 ma pin v cc2 (ecl) 45 ma pin v cc3 (cml) 200 ma *see notes following end of switching characteristics tables.
amd 29 am7968/am7969-175 am7969-175 taxichip receiver parameter symbol parameter description test conditions (note 1) min max unit bus interface signals: do0Cdo7, do8/co3, do9/co2, co0Cco1, dstrb, cstrb, igm, clk, cnb, vltn v oh output high voltage v cc = min, i oh = C1 ma 2.4 v v in = 0 or 3 v v ol output low voltage v cc = min, i ol = 8 ma 0.45 v v in = 0 or 3 v v ih input high voltage v cc = max (note 9) 2.0 v v il input low voltage v cc = max (note 9) 0.8 v v i input clamp voltage v cc = min, i in = C18 ma C1.5 v i il input low current v cc = max, v in = 0.4 v C400 m a i ih input high current v cc = max, v in = 2.7 v 50 m a i i input leakage current v cc = max, v in = 5.5 v 50 m a i sc output short circuit C15 C85 ma current (note 4) serial interface signals: serin+, serinC v ihs input high voltage (notes 9, 21) v cc v cc v serin+ C1.165 C0.88 v ils input low voltage (notes 9, 21) v cc v cc v serin+ C1.81 C1.475 v tht test mode threshold v cc = max 0.25 v serinC v dif differential input voltage 0.3 1.1 v v icm input common mode (note 6) 3.05 v cc v voltage C0.55 i il input low current v cc = max, v in = v cc C1.81 v 0.5 m a i ih input high current v cc = max, 220 m a v in = v cc C0.88 v miscellaneous signals: x1, v cc1 , v cc2 v ihx input high threshold x1 2.0 v v ilx input low threshold x1 0.8 v i ilx input low current x1 v in = 0.45 v C900 m a i ihx input high current x1 v in = 2.4 v +600 m a i cc supply current v cc1 = v cc2 = max pin v cc1 (ttl) 50 ma dms = 0 v pin v cc2 (cml) 300 ma
amd 30 am7968/am7969-175 switching characteristics (note 20) am7968-175 taxichip transmitter (notes 10, 13, 22) parameter no. symbol parameter description test conditions min max units bus interface signals: di0Cdi7, di8/ci3, di9/ci2, ci0Cci1, strb, ack, clk 1t p clk period 5.7 n 8 n ns 2t pw clk pulse width high 20 ns 3t pw clk pulse width low 20 ns 4t pw strb pulse width high (note 7) 15 ns 5t pw strb pulse width low 15 ns 6t bb internal byte boundary to clk 20 ns (note 11) 9t s dataCstrb setup time 5 ns 10 t h dataCstrb hold time 15 ns 11 t h ack - to strb hold (note 8) ttl output load 0 ns 12 t h ack to strb - hold ttl output load 0 ns 13 t pd strb - to ack - (note 18) ttl output load 40 ns 14 t pd strb to ack ttl output load 23 ns 15 t pd clk to ack - (note 18) ttl output load ns serial interface signals: serout+, seroutC (note 2) 22 t sk = serout skew ecl output load C200 +200 ps 23 t r = serout output rise time ecl output load .45 2 ns 24 t f = serout output fall time ecl output load .45 2 ns 26 t pw = serout pulse width low ecl output load ns 27 t pw = serout pulse width high ecl output load ns miscellaneous signals: x1 (note 15) 29 t pw x1 pulse width high (note 12) ttl output load on clk 24 ns 30 t pw x1 pulse width low (note 12) ttl output load on clk 24 ns 32 t pd x1 - to clk - ttl load 32 ns 33 t pd x1 to clk ttl load 32 ns 3t 1 n + 33 t 1 n C 5% t 1 n + 5% t 1 n + 5% t 1 n C 5% C9t 1 8n +9
amd 31 am7968/am7969-175 am7969-175 taxichip receiver (notes 13, 14, 22) t 35 n t 35 n parameter no. symbol parameter description test conditions min max unit bus interface signals: do0Cdo7, do8/co3, do9/co2, co0Cco1, dstrb, cstrb, igm, clk, cnb, vltn 35 t p clk period (note 24) 5.7 n 8 n ns 36 t pd data valid to strb - delay ttl output load ns 37 t pd clk to strb - ttl output load ns 38 t pd clk - to strb ttl output load ns 38a t pd strb - to clk - (note 23) ttl output load ns 39 t pd clk to data valid delay ttl output load ns 40 t pw strb pulse width high ttl output load ns 41 t pw clk pulse width high ttl output load ns 42 t pw clk pulse width low ttl output load ns 43 t pd serin to clk delay ttl output load ns 47a t s cnb to clk - setup time ns (note 19) 47b t s cnb - to clk setup time 29 ns 48 t h cnb to clk - hold ns 49 t pw cnb pulse width low ns serial interface signals: serin+, serinC 57 t j = serin peak to peak input jitter 2 ns tolerance (note 16) miscellaneous signals: x1 (note 15) 60 t pw x1 pulse width high 21 ns 61 t pw x1 pulse width low 21 ns 2t 35 n 2t 35 n +15 t 35 n C5 5t 35 2n 5t 35 n C7 5t 35 n C4 t 35 2n +17 2t 35 n +26 + - 23 5t 35 n 2t 35 n 3t 35 n C10 C31 - 2t 35 n C3 C2
amd 32 am7968/am7969-125 militar y (page intentionally left blank)
amd 33 am7968/am7969-125 militar y am7968/am7969-125 military absolute maximum ratings storagetemperature C65 c to +150 c . . . . . . . . . . . . ambient temperature under bias C55 c to +125 c . . . . . . . . . . . . . . . . . . . supply voltage to ground potential continuous C0.5 v to +7.0 v . . . . . . . . . . . . dc voltage applied to outputs C0.5 v to v cc max . . . . . . . . . . . . . . . . . . . . . dc input voltage C0.5 v to +5.5 v . . . . . . . . . . . . . . . dc output current 100 ma . . . . . . . . . . . . . . . . . . . dc input current C30 ma to +5.0 ma . . . . . . . . . . . . . stresses above those listed under absolute maximum rat- ings may cause permanent device failure. functionality at or above these limits is not implied. exposure to absolute maxi- mum ratings for extended periods may affect device reliability. operating ranges military (smd) devices 5962-9052701m3a 5962-9052701mxa 5962-9052801m3a 5962-9052801mxa temperature (t c ) C55 c to +125 c . . . . . . . . . . . . supply voltage (v cc ) +4.75 v to +5.5 v . . . . . . . . . . . military (cpl) devices am7968-125/lkc am7968-125/dkc am7969-125/lkc am7969-125/dkc temperature (t c ) C30 c to +125 c . . . . . . . . . . . . . supply voltage (v cc ) +4.5 v to +5.5 v . . . . . . . . . . . . operating ranges define those limits between which the func- tionality of the device is guaranteed.
amd 34 am7968/am7969-125 militar y dc characteristics over operating range unless otherwise specified (for cpl prod- ucts group a, subgroups 1, 2, 3 are tested unless otherwise noted) am7968-125 military taxichip transmitter parameter symbol parameter description test conditions (note 1) min max unit bus interface signals: di0Cdi7, di8/ci3, di9/ci2, ci0Cci1, strb, ack, clk v oh1 output high voltage v cc = min, i oh = C1 ma 2.4 v ack v in = 0 or 3 v v oh2 output high voltage v cc = min, i oh = C1 ma 2.4 v clk v in = 0 or 3 v v ol output low voltage v cc = min, i ol = 8 ma 0.45 v ack, clk v in = 0 or 3 v v ih input high voltage v cc = max (note 9) t c = C30 to +125 c 2.0 v t c = C55 to +125 c 2.1 v v il input low voltage v cc = max (note 9) 0.8 v v i input clamp voltage v cc = min i in = C18 ma C1.5 v i il input low current v cc = max, v in = 0.4 v C400 m a i ih input high current v cc = max, v in = 2.7 v 50 m a i i input leakage current v cc = max, all inputs 50 m a v in = 5.5 v except clk clk input 150 m a i sc output short circuit (note 4) C15 C85 ma current ack, clk serial interface signals: serout+, seroutC v oh output high voltage v cc = min ecl load v cc v cc v C1.165 C0.88 v ol output low voltage v cc = min ecl load v cc v cc v C1.81 C1.62 miscellaneous signals: x1, v cc1 , v cc2 , v cc3 v ihx input high voltage x1 v cc = max (note 9) t c = C30 to +125 c 2.0 v t c = C55 to +125 c 2.1 v v ilx input low voltage x1 0.8 v i ilx input low current x1 v in = 0.45 v C900 m a i ihx input high current x1 v in = 2.4 v +600 m a i cc supply current serout = ecl load, dms = 0 v cc1 = v cc2 = v cc3 = max pin v cc1 (ttl) 30 ma pin v cc2 (ecl) 45 ma pin v cc3 (cml) 215 ma *see notes following end of switching characteristics tables.
amd 35 am7968/am7969-125 militar y am7969-125 military taxichip receiver parameter symbol parameter description test conditions (note 1) min max unit bus interface signals: do0Cdo7, do8/co3, do9/co2, co0Cco1, dstrb, cstrb, igm, clk, cnb, vltn v oh output high voltage v cc = min, i oh = C1 ma 2.4 v v in = 0 or 3 v v ol output low voltage v cc = min, i ol = 8 ma 0.45 v v in = 0 or 3 v v ih input high voltage v cc = max (note 9) 2.0 v v il input low voltage v cc = max (note 9) 0.8 v v i input clamp voltage v cc = min, i in = C18 ma C1.5 v i il input low current v cc = max, v in = 0.4 v C400 m a i ih input high current v cc = max, v in = 2.7 v 50 m a i i input leakage current v cc = max, v in = 5.5 v 50 m a i sc output short circuit C15 C85 ma current (note 4) serial interface signals: serin+, serinC v ihs input high voltage (notes 9, 21) v cc v cc v serin+ C1.165 C0.88 v ils input low voltage (notes 9, 21) v cc v cc v serin+ C1.81 C1.475 v tht test mode threshold v cc = max 0.25 v serinC v dif differential input voltage 0.3 1.1 v v icm input common mode (note 6) 3.05 v cc v voltage C0.55 i il input low current v cc = max, v in = v cc C1.81 v 0.5 m a i ih input high current v cc = max, 220 m a v in = v cc C0.88 v miscellaneous signals: x1, v cc1 , v cc2 v ihx input high threshold x1 2.0 v v ilx input low threshold x1 0.8 v i ilx input low current x1 v in = 0.45 v C900 m a i ihx input high current x1 v in = 2.4 v +600 m a i cc supply current v cc1 = v cc2 = max pin v cc1 (ttl) 55 ma dms = 0 v pin v cc2 (cml) 335 ma
amd 36 am7968/am7969-125 militar y switching characteristics over operating range unless otherwise specified (note 20) (for cpl products group a, subgroups 9, 10, and 11 are tested unless other- wise noted) am7968-125 military taxichip transmitter (notes 10, 13, 22) parameter no. symbol parameter description test conditions min max units bus interface signals: di0Cdi7, di8/ci3, di9/ci2, ci0Cci1, strb, ack, clk 1t p clk period 8 n 25 n ns 2t pw clk pulse width high 25 ns 3t pw clk pulse width low 25 ns 4t pw strb pulse width high (note 7) 20 ns 5t pw strb pulse width low 20 ns 6t bb internal byte boundary to clk 25 ns (note 11) 9t s dataCstrb setup time 10 ns 10 t h dataCstrb hold time 15 ns 11 t h ack - to strb hold (note 8) ttl output load 0 ns 12 t h ack to strb - hold ttl output load 0 ns 13 t pd strb - to ack - (note 18) ttl output load 45 ns 14 t pd strb to ack ttl output load 25 ns 15 t pd clk to ack - (note 18) ttl output load ns miscellaneous signals: x1 (note 15) 29 t pw x1 pulse width high (note 12) ttl output load on clk 35 ns 30 t pw x1 pulse width low (note 12) ttl output load on clk 35 ns 32 t pd x1 - to clk - ttl load 32 ns 33 t pd x1 to clk ttl load 32 ns 3t 1 n + 43 C9t 1 8n +3
amd 37 am7968/am7969-125 militar y am7969-125 military taxichip receiver (notes 13, 14, 22) t 35 n parameter no. symbol parameter description test conditions min max unit bus interface signals: do0Cdo7, do8/co3, do9/co2, co0Cco1, dstrb, cstrb, igm, clk, cnb, vltn 35 t p clk period (note 24) 8 n 25 n ns 36 t pd data valid to strb - delay ttl output load ns 37 t pd clk to strb - ttl output load ns 38 t pd clk - to strb ttl output load ns 38a t pd strb - to clk - (note 23) ttl output load ns 39 t pd clk to data valid delay ttl output load ns 40 t pw strb pulse width high ttl output load ns 41 t pw clk pulse width high ttl output load ns 42 t pw clk pulse width low ttl output load ns 43 t pd serin to clk delay ttl output load ns serial interface signals: serin+, serinC 57 t j = serin peak to peak input jitter 5 ns tolerance (note 16) miscellaneous signals: x1 (note 15) 60 t pw x1 pulse width high 35 ns 61 t pw x1 pulse width low 35 ns 2t 35 n 2t 35 n +15 t 35 n C7 5t 35 2n 5t 35 n C15 5t 35 2n C15 + - 23 5t 35 n 3t 35 n C14 t 35 2n +17 2t 35 n +26 note: clk (pin 19) must be connected to cnb (pin 24).
amd 38 am7968/am7969 1 2 v cc notes:* 1. for conditions shown as min or max use the appropriate value specified under operating range. 2. the clock fall to serial output delay is typically 3 bit times. 4. not more than one output should be shorted at a time. duration of the short circuit test should not exceed one second. 5. if the cnb - to clk - setup time is violated, igm will stay low. 6. voltage applied to either serin pins must not be above v cc nor below +2.5 v to assure proper operation. 7. t 4 guarantees that data is latched. ack (t 11 ) timing may not be valid. 8. if t 11 is not met, ack response and timing are not guaranteed, but data will still be latched on strb - (see t 4 ). 9. measured with device in test mode while monitoring output logic states. 10. for the taxi transmitter, n" is determined by the following table: 11. t 6 (internal byte boundary to clk ) is created by the variation of internal strb propagation delays relative to internal byte boundaries over temperatures and v cc . the internal byte boundary determines the byte in which data will come out (serout ). if strb occurs before the byte boundary, then the data will be sent out two bytes later. if strb occurs after the byte boundary, then the output data will be delayed by one additional byte. 12. x1 pulse width is measured at a point where clk output equals t 2 or t 3 . 13. for the taxi transmitter, `data' is either di0 - di7, di8/ci3, di9/ci2, ci0 - ci1. for the taxi receiver, `strb' is either cstrb or dstrb and `data' is either do0 - do7, do8/co3, do9/co2, co0 - co1. 14. for the taxi receiver, `n' is determined by the state of the dms and serin- inputs. when serin- is held below v tht max or left open, n=1. when serin- is held above 0.25 v and when: 1 2 gnd open or open open open n = 1; 8 bit test mode 2 n = 10; 8 bit local/test mode 1 n = 1; 9 bit test mode 2 n = 11; 9 bit local/test mode 1 n = 1; 10 bit test mode 2 n = 12; 10 bit local/test mode 1 dms tls n" gnd open or > 2.5 v n = 1; 8 bit test mode n = 10; 8 bit local mode n = 1; 9 bit test mode n = 11; 9 bit local mode n = 1; 10 bit test mode n = 12; 10 bit local mode dms serin- n" > 2.5 v > 2.5 v < v thtmax or open < v thtmax or open < v thtmax or open gnd/v cc gnd/v cc gnd/v cc v cc v cc v cc
amd 39 am7968/am7969 15. jitter on x1 input must be less than 0.2 ns to ensure that automatic test equipment can properly measure device switching characteristics. the x1 input frequency will determine the byte rate reference for the receiver byte clock. 16. this specification is the sum of data dependent jitter, duty cycle distortion, and random jitter. 18. ack delay is determined by t 13 when the input latch is empty or by t 15 when the latch is full (busy mode). also note that ack will not rise if strb does not remain high until ack rises. 19. if t 47a (cnb ? to clk setup) is violated, then output data will occur one byte time later. 20. all timing references are made with respect to +1.5 v for ttlClevel signals or to the 50% point between v oh and v ol for ecl signals. ecl input rise and fall times must be 2 ns 0.2 ns between 20% and 80% points. ttl input rise and fall times must be 2 ns between 1 v and 2 v. 21. device thresholds on the serin ( +/C ) pin(s) are verified during production test by ensuring that the input threshold is less than v ihs (min) and greater than v ils (max). the figure below shows the acceptable range (shaded area) for the transition voltage. 22. switching characteristics are tested during 8-bit local mode operation. 23. the limit for this parameter cannot be derived from t 37 and t 42 . 24. this specification does not apply during reacquisition when clk stretch can occur. = this parameter is guaranteed but is not included in production tests. * notes listed correspond to the respective references made in the dc characteristics and the switching characteristics tables. v cc v cc = 0.88 v v cc = 1.165 v input threshold transition voltage v cc = 1.475 v v cc = 1.81 v
amd 40 am7968/am7969 switching test circuits notes: 1. r1 = 500 w for the i ol = 8 ma 2. all diodes in916 or in3064, or equivalent 3. c l = 30 pf includes scope probe, wiring and stray capacitances without device in test fixture. 4. amd uses constant current (a.t.e.) load configurations and forcing functions. this figure is for reference only. notes: 1. c l < 3 pf includes scope probe, wiring and stray capacitances without device in test fixture. 2. amd uses automatic test equipment load configurations and forcing functions. this figure is for reference only. 2.4k 30 pf v out v cc r1 07370f-14 v out c l v cc C 2 v 50 w 07370f-15 ttl output load ecl output load
amd 41 am7968/am7969 switching test waveforms 3.0 v 2.0 v 1.5 v 1.0 v 2 0.2 ns 2 0.2 ns 07370f-16 2 0.2 ns v cc C 0.9 v 80% 2 0.2 ns 50% 20% v cc C 1.7 v 07370f-17 ecl input waveform ttl input waveform 0 v key to switching waveforms must be steady may change from h to l may change from l to h does not apply dont care any change permitted will be steady will be changing from h to l will be changing from l to h changing state unknown center line is high impedence off state waveform inputs outputs ks000010
amd 42 am7968/am7969 switching waveforms 30 29 1 3 2 9 10 4 12 5 6 13 11 x1 clk data in strb ack serout+ serout- data or command 23 24 27 22 26 note 2 23 24 27 26 33 32 14 15 07370f-18 am7968 taxichip transmitter ac note: 2. the clock fall to serial output delay is typically 3 bit times.
amd 43 am7968/am7969 switching waveforms 35 61 60 57 43 111101010 010 010010101100010001 41 39 37 36 38 48 49 x1 serin+ clk data out strb cnb dstrb or cstrb do or co data=02 data=34 data=sync (jk) 1 1110 1 0 1 0 0 10 01 data=02 data=xx data=02 data=34 command=0 data=02 47a n m 40 42 47b 38a igm 07370f-19 am7969 taxichip receiver ac note 1 note 2 note 1 note 3 44 46 45 notes: 1. igm rises because cnb = 1 and serin = first half of non-sync byte. 2. igm falls because cnb falls. 3. igm falls because serin = first half of sync byte. this diagram illustrates how timing relationships are measured. functional operation is clarified on following pages.
amd 44 am7968/am7969 switching waveforms 1 23456 int clk* clk output data 1 data 2 data 3 data 4 data/command input strb input ack out (note 1) data 1 data 2 data 3 data 4 input latch* data 1 data 2 data 3 data 4 sync sync data 1 data 2 sync data 3 encoder latch* shifter* nrz data* serout serial output data sync sync data 1 data 2 sync data 3 sync sync data 1 data 2 sync data 3 1 000 1 10001 1 000 1 10001 1 000 1 100 0 1 1 000 1 1 00 01 strb to serout timing (8-bit local mode) *internal signals 07370f-20 taxichip transmitter note: 1. the input latch is busy when the second strb comes in; the internal strb-ack is delayed until the next clk window. refer to figure 3.
amd 45 am7968/am7969 switching waveforms 1 2345 6 sync data n data 1 cmd 1 data 3 data 4 data n sync data 1 cmd 1 data 3 data 4 data n-1 data n sync data 1 cmd 1 data 3 data n-2 data n-1 data n no change data 1 no change no change no change no change cmd 0 no change cmd 1 internal clock* serin serial data nrz data* clk out = cnb decoder latch* data out dstrb out command out cstrb out 1000 1 1 00 01 1000 1 10001 taxichip receiver 07370f-21 taxichip receiver timing (8-bit local mode)
amd 46 am7968/am7969 switching waveforms (note 4) (note 3) clk out = cnb igm decoder latch* command out cstrb out data out dstrb out internal clock* serin serial data nrz data* 01234567890123450123456789012345678901234567890123456789 data n sync data n sync data 1 data 2 data 3 data 1 data 2 data 3 (notes 1 & 2) *** *** 1000 1 10001 1000 1 10001 (note 5) no change sync no change no change no change data n (lost) sync data 1 data 2 data 3 data n-1 data n no change data 1 data 2 (note 7) (note 6) taxichip receiver timing (8-bit mode/local) showing external effect of sync error 07370f-22 notes: *internal signals 1. sync detected in shifter, but not synchronized with internal state machine. 2. state machine re-cycled to new sync position. 3. clock output delayed to new position. 4. the low time or high time gets stretched depending on what state of the internal machine is reset. 5. igm rises at the 6.5th state of the state machine. 6. strobe falls at the rising edge of the clock out. 7. strobe may be shifted one bit time if the state machine is reset at state 1.
amd 47 am7968/am7969 switching waveforms 1 2345 6 sync data n data 1 data 2 no change no change no change no change command 0 no change no change no change no change data n-1 no change no change internal clock* serin serial data clk out command out dstrb out data out 1000 1 100 01110 00100 01 sync sync 1000 1 1 00 01 data n data 1 data 2 sync sync 1000 1 100 01110 00100 01 1 00 0 1 1 00 01 nrz data cnb taxi #1 = 1 igm taxi #1 = cnb taxi #2 cstrb out no change no change no change no change no change no change no change no change no change data n no change no change command out dstrb out data out cstrb out taxichip receiver 07370f-23 taxichip receiver timing (8-bit cascade mode) *internal signals taxi #1 taxi #2
amd 48 am7968/am7969 physical dimensions* cd 028 28-pin ceramic dip (measured in inches) .050 .065 .005 min .100 bsc .160 .220 .015 .022 .015 .060 1.435 1.490 .565 .605 06837d bz13 cd 028 1/8/91 c dc .125 .160 .008 .012 .590 .615 0? 15 .700 max end view side view top view 1 .098 max .150 min pl 028 28-pin plastic leaded chip carrier (measured in inches) .050 ref .042 .048 .450 .456 .485 .495 .042 .056 .165 .180 .090 .120 .009 .015 .300 ref .013 .021 .390 .430 .020 min .025 .045 06751f bv 8 pl 028 12/31/91 c dc r .026 .032 .450 .456 .485 .495 top view side view *for reference only. all dimensions measured in inches. bsc is an ansi standard for basic space centering.
amd 49 am7968/am7969 physical dimensions clt028 28-pin ceramic leadless chip carrier (measured in inches) .015 min .006 .022 .050 bsc .150 bsc .300 bsc .022 .028 .150 bsc top view index corner .020 x 45 ref. (optional) .040 x 45 ref. (3x) (optional) .430 max .442 .458 .430 max .442 .458 bottom view .054 .065 .064 .075 plane 2 plane 1 07703d cs47 clt 028 04/28/94 ae side view .045 .055 .300 bsc
50 publication# 12330 rev. e amendment /0 issue date: april 1994 taxichip tm integrated circuits technical manual 1.0 introduction modern electronic systems move data from point-to-point across physical layer bounda- ries using either serial or parallel data links. parallel data links provide fast data transfers and are compatible with most computer architectures. however, conventional parallel data links are burdened with cost/performance issues such as costly multi-con- ductor cables, crosstalk, rfi, bit-to-bit skew and other concerns associated with multiple wire interfaces. serial data links, although simpler and less costly, have not provided sufficient bandwidth to compete with the high data transfer rates of parallel links. recent technological advances have altered the cost performance trade-off between serial and parallel data transfer techniques. a new chip set from advanced micro devices offers a high performance integrated alternative to traditional serial/parallel data transfer techniques. the taxlchip set (transparent asynchronous xmitter-receiver interface) provides the means to establish a transparent high speed serial link between two high performance parallel buses. the taxlchip set consists of a transmitter, which takes parallel data and transmits it serially at up to 175 mhz, and a receiver, which converts the serial data stream back to parallel form. taxlchips provide a simple parallel interface through a high speed serial link, while maintaining the data bandwidth required by the system. 1.1 the am7968 taxi transmitter the taxi tm transmitter consists of an input latch, an encoder, a parallel to serial shift register, a multiplying phase locked loop (pll), and some control logic (figure 1-1). data are input to the latch, encoded, and shifted out at the serial data rate. the encod- ing used is the efficient 4b/5b scheme specified for the ansi x3t9.5 fiber distributed data interface (fddi specification). this encoding divides an 8-bit byte into two, 4-bit nibbles. each nibble is encoded into a 5-bit symbol. the 10-bit encoded byte is format- ted into an nrzi data stream for output to the media. this 4b/5b encoding is 80% efficient, using a 125 mbaud transmission rate to send 100 mbps of data. the am7968 transmitter has differential pseudo-ecl (referenced to +5 v) outputs which can drive 50 w lines. this capability makes it easy to directly interface with shielded twisted pair or coaxial cables. the pseudo-ecl outputs are also compatible with the ecl interface of optical compo- nents used to drive fiber optic cable. in addition to providing high bandwidth and low attenuation, fiber optic data transmission also offers noise immunity, eliminates rfi and provides data security. declining optical components costs are bringing the advantages of fiber optic data transmission to an ever wider range of applications, from process control to avionics. the taxlchip set is the ideal complement for fiber optic interfaces.
amd 51 taxichi p inte g rated circuits technical manual figure 1-1 am7968 taxi transmitter block diagram note: n can be 8, 9, or 10 bits. total of n + m = 12. 12330e-1 strobe (strb) acknowledge (ack clock (clk) data mode select (dms) test serial in (tserin) test/local select (tls) strobe & acknowledge oscillator and clock gen. serial interface shifter data encoder encoder latch input latch media interface (serout+) serial out + (seroutC) serial out C data command nm x1 x2 figure 1-2 am7969 taxi receiver block diagram note: n can be 8, 9, or 10 bits. total of n + m = 12. output latch data decoder decoder latch shifter (x1) (x2) oscillator and clock gen. media interface pll clock generator byte sync logic nm data command (vltn) violation (serin+) serial in + (serinC) serial in C (dms) data mode select (cnb) catch next byte (igm) i-got-mine (clk) clock (dstrb) data strobe (cstrb) command strobe 12330e-2
amd 52 taxichi p inte g rated circuits technical manual 1.2 the am7969 taxi receiver the taxi receiver (figure 1-2) accepts the encoded data stream into a serial-to-paral- lel converter, decodes and outputs the received data with an accompanying strobe. an on-chip data tracking pll performs the necessary clock recovery from the input serial data stream. 2.0 using the taxichip set the current taxlchip set has a maximum effective data throughput of 140 mbps, over ten times faster than the data rate of conventional rs-422 drivers and receivers. the taxlchip set has a frequency range of 40 mhz to 175 mhz corresponding to a parallel data transfer rate of 4 to 17.5 mbyte/sec. data rates of less than 4 mbyte/sec are accommodated by the automatic insertion of sync symbols in the absence of new data or commands. the taxi transmitter accepts parallel input data with a simple strobe/acknowledge handshake, while the receiver asserts an output strobe when data is available at its parallel outputs. the high speed serial-to-parallel conversions and data encoding/de- coding are transparent to the user, who sees only an effective parallel transfer rate of up to 17.5 mbyte/sec (see figure 2-1). appendix c, taxi tip # 89-07 addresses the use of synchronous and asynchronous strobes. it is important to note here that the user is not forced to supply data at the maximum byte rate equivalent of the serial data rate. data or commands are sent down the serial link only when the user strobes the transmitter. for those byte clock cycles when the transmitter is not strobed, it automatically sends a special sync symbol down the link. the sync symbol is a unique bit pattern which cannot be confused with any other valid pattern. for this reason, sync is used to establish byte framing at the receiver. see appendix c, taxi tip # 89-03 proper use of taxi sync. it also keeps the link active when no other symbols are being sent, maintaining receiver pll lock. sync will not over-write data already present in the receivers output data latch. the serial rate is, therefore, truly transparent to the user; at input data rates less than the equivalent serial bit rate, the taxi transmitter will fill the gaps with syncs, which do not disturb receiver output data. 2.1 data and command the am7968 taxi transmitter and the am7969 taxi receiver interface directly to an 8, 9, or 10 bit data bus. each taxlchip has 12 parallel interface lines which are designated as either command or data bits. command bits implement user defined system supervisory functions, such as initialize your system, re-try, halt, or error which cannot be embedded in the ordinary data path.
amd 53 taxichi p inte g rated circuits technical manual figure 2-1 basic taxichip operation 6. the data comes out the rx and data strobe is raised am7968 transmitter am7969 receiver 1. parallel data is entered by the user 2. the data is strobed in by the user 3. the data is encoded and then converted into a serial stream 4. the serial data is sent out 5. the re- ceiver takes the serial data and converts it back to parallel data and then decodes it data out data strobe data in strobe acknowledge 12330e-3 three different widths are possible: 8 data and 4 command bits, 9 data and 3 com- mand bits, and 10 data and 2 command bits. this choice of data and control bus widths allows flexibility to meet different system bus width requirements, while providing the capability of merging control and data into a common data stream. 2.2 operational modes: local, cascade and test a taxichip set point-to-point link can be operated in one of three modes: local, cascade, or test. local mode consists of a single transmitter communicating with a single receiver over the serial medium. cascade mode for am7968/7969-125 consists of a single transmitter driving two or more daisy chained (cascaded) receivers over a single serial medium. cascade operation for am7968/am7969-175 consists of a single transmitter driving a single receiver as shown in appendix c, taxi tips #13 and #14. cascade mode permits direct interface with 16-bit, 32-bit and wider busses. the link may be operated in any of the above modes using the taxls internal pll for bit rate generation and tracking, or the link may be run in test mode with external frequency multiplying and data tracking plls. 3.0 data encoding, violation and syncs 3.1 data encoding any form of serial data transmission requires some form of encoding before the data are output to the transmission medium. encoding is the process of converting a set of m data bits to a set of n code bits. the purpose of the encoding operation is to include clock information in the data stream. without this timing information, the receiver would not be able to distinguish adjacent bits of the same value. for example, if we transmit a thousand ones followed by a zero, the receiver might detect only 999 ones, or perhaps 1001 ones, followed by one or two zeros. an accurate clock is needed to tell the receiver when to sample the incoming bit stream to determine if the bit is a one or a zero. since the transmitter and receiver have only one data path between them, the clock (timing) information must be included in the serial data stream.
amd 54 taxichi p inte g rated circuits technical manual the taxichip set uses 4b/5b or 5b/6b coding, so that m is either 4 or 5, and n is either 5 or 6. in 8-bit mode, each 4-bit nibble is presented to one of two 4b/5b encoders to produce 10 code bits, 5 from each encoder. in 9-bit mode, the more significant 4b/5b encoder is replaced with a 5b/6b encoder to yield a total of 11 code bits. in 10-bit mode, both encoders are replaced with 5b/6b encoders, yielding a total of 12 code bits. the taxichip set can encode two types of data: either 8, 9, or 10-bit data, or com- mands. commands are special symbols which are typically used as control functions at the receiving end of the link. commands may be four, three, or two bits wide, corre- sponding to a data width of eight, nine, or ten bits respectively. the presence of any non-zero bits on the command inputs when strb is asserted will cause a command symbol to be sent, regardless of the state of the data lines. the command bits are encoded into 10,11, or 12 bit groupings which are special cases of the 4b/5b or 5b/6b code not used for data. in the absence of data or commands, a unique symbol (sync) is automatically gener- ated to maintain link synchronization. if the user has not supplied a strb during a byte, a sync symbol is sent. nrzi stands for non-return to zero, invert on one. logic ones are indicated by a transition, while logic zeros produce no transition. further encoding the 4b/5b encoded data in this way ensures that the receiver pll will get a transition at least every three clock times (the maximum number of zeros in the 4b/5b code). since a pll can make a phase comparison and initiate a correction only at a transition, maximizing the number of transitions helps to keep the loop solidly in lock.
amd 55 taxichi p inte g rated circuits technical manual table 3-1 taxlchip encoder patterns 4-bit 5-bit 5-bit 6-bit hex binary encoded hex binary encoded data data symbol data data* symbol 0 0000 11110 00 00000 110110 1 0001 01001 01 00001 010001 2 0010 10100 02 00010 100100 3 0011 10101 03 00011 100101 4 0100 01010 04 00100 010010 5 0101 01011 05 00101 010011 6 0110 01110 06 00110 010110 7 0111 01111 07 00111 010111 8 1000 10010 08 01000 100010 9 1001 10011 09 01001 110001 a 1010 10110 0a 01010 110111 b 1011 10111 0b 01011 100111 c 1100 11010 0c 01100 110010 d 1101 11011 0d 01101 110011 e 1110 11100 0e 01110 110100 f 1111 11101 0f 01111 110101 10 10000 111110 11 10001 011001 12 10010 101001 13 10011 101101 14 10100 011010 15 10101 011011 16 10110 011110 17 10111 011111 18 11000 101010 19 11001 101011 1a 11010 101110 1b 11011 101111 1c 11100 111010 1d 11101 111011 1e 11110 111100 1f 11111 111101 * notes: hex data is parallel input data which is represented by the 4- or 5-bit binary data listed in the column to the immediate right of hex data. binary bits are listed from left to right in the following order. 8-bit mode: d7, d6, d5, d4, (4-bit binary), and d3, d2, d1, d0, (4-bit binary) 9-bit mode: d8, d7, d6, d5, d4, (5-bit binary), and d3, d2, d1, d0, (4-bit binary) 10-bit mode: d8, d7, d6, d5, d4, (5-bit binary), and d9,d3, d2, d1, d0, (5-bit binary) serial bits are shifted out with the most significant bit of the most significant nibble coming out first. 4b/5b encoder scheme 5b/6b encoder scheme table 3-2 taxichip command symbols
amd 56 taxichi p inte g rated circuits technical manual encoded hex binary symbol mnemonic hex binary 0 0000 xxxxx xxxxx data no change no change (note 2) (note 2) no strb no strb 11000 10001 jk (8-bit sync) 0 0000 (note 1) (note 1) 1 0001 11111 11111 i i 1 0001 2 0010 01101 01101 tt 2 0010 3 0011 01101 11001 ts 3 0011 4 0100 11111 00100 i h 4 0100 5 0101 01101 00111 tr 5 0101 6 0110 11001 00111 sr 6 0110 7 0111 11001 11001 ss 7 0111 8 (note 3) 1000 00100 00100 hh 8 1000 9 1001 00100 11111 hi 9 1001 a (note 3) 1010 00100 00000 hq a 1010 b 1011 00111 00111 rr b 1011 c 1100 00111 11001 rs c 1100 d (note 3) 1101 00000 00100 qh d 1101 e (note 3) 1110 00000 11111 q i e 1110 f (note 3) 1111 00000 00000 qq f 1111 9-bit mode 0 000 xxxxxx xxxxx data no change no change (note 2) (note 2) no strb no strb 011000 10001 lk (9-bit sync) 0 000 (note 1) (note 1) 1 001 111111 11111 ii 1 001 2 010 011101 01101 tt 2 010 3 011 011101 11001 ts 3 011 4 100 111111 00100 ih 4 100 5 101 011101 00111 tr 5 101 6 110 111001 00111 sr 6 110 7 111 111001 11001 ss 7 111 10-bit mode 0 00 xxxxxx xxxxxx data no change no change (note 2) (note 2) no strb no strb 011000 100011 lm (10-bit sync) 0 00 (note 1) (note 1) 1 01 111111 111111 ii 1 01 2 10 011101 011101 tt 2 10 3 11 011101 111001 ts 3 11 notes: 1. command pattern sync cannot be explicitly sent by am7968 transmitter with any combination of inputs and strb, but is used to pad between user data. 2. a strobe with all os on the command input lines will cause data to be sent. see table 3-1. 3. while these commands are legal data and will not disrupt normal operation if used occasionally, they may cause data errors if grouped into recurrent fields. normal pll operation cannot be guaranteed if one or more of these commands is continuously repeated. command input command output am7969 receiver am7968 transmitter 8-bit mode
amd 57 taxichi p inte g rated circuits technical manual 3.2 violation logic the taxi receiver logic has been designed to detect the most common types of transmission errors. it detects these errors by completely decoding the incoming data patterns, and recognizes the following types of violations: 1. illegal, reserved or unused data patterns (pure violations). 2. unused command combinations. 3. command in upper half and data in lower half pattern. 4. data in upper half and command in lower half pattern. type 1 and 2 violations are decoded and interpreted as either data or command outputs with the appropriate strb output. type 3 & 4 violations are decoded as command outputs with a cstrobe output (even though one half would have been transformed data/command or command/data by an error), since there is no information available to the taxi receiver to indicate where the error lies. the user needs to be aware of this possible transposition (possible with all four types of viola- tion), since the system must account for it. violation will always be the flag for these detectable errors. this method of detection is not 100% effective. as appendix b shows, it will detect approximately 50% of the possible double bit errors in data. double bit errors in command will be detected 99.8% of the time or more, depending upon the pattern width. appendix b contains a more detailed treatment of the efficiency of the violation logic for the various data bit modes. the method of detecting violations, is effective enough to be used to give an early warning of transmission problems before the hosts error detection system would detect the errors. it should not be used alone in fault sensitive systems, since it misses a significant number of transmission errors which cause one valid data pattern to alias to another valid data pattern. 3.3 taxi pll characteristics the phase locked loop in the taxi receiver is used to recover the data encoded in the serial bit stream sent by the taxi transmitter. in order to ensure accurate data recovery, the receiver pll must lock on to the underlying code rate of the transmitter, and must track minor changes in frequency and phase while rejecting noise superim- posed on the bit stream. this noise includes both amplitude and phase/frequency disturbances. amplitude variations are dealt with in the receivers input amplifier (serin+/-), and are not passed through to the pll, except for phase effects. phase/frequency noise, or jitter, can come from many sources, and can have many different characteristics. jitter can be introduced by the transmitter, the receiver, the media interface or by the media itself. examples of media induced jitter include reflec- tions and edge perturbations caused by improper line termination, pulse width spreading due to frequency dependent cable attenuation, and pulse dispersion caused by fiber optic cable effects. examples of media interface jitter include low light effects in optical receivers and pulse width distortion caused by baseline shift (changing dc offset) in ac coupled amplifiers. the taxi pll has been optimized to allow correct data recovery in the presence of the largest jitter possible. to this end, the pll parameters, most notably loop bandwidth, have been chosen to enhance the jitter tolerance of the taxi receiver.
amd 58 taxichi p inte g rated circuits technical manual this optimization is at the expense of lock-up time. in taxi systems, lock-up time is relatively unimportant, since the system must achieve lock only during system power-up. if the pll achieves proper lock within a few tens, or even hundreds of microseconds, its startup will be similar to the start-up characteristics of the system power supply. the actual time to lock begins during power-up, when both transmitter and receiver are marginally powered and the entire link is marginally functional. transient effects other than pll characteristics, which typically occur during power-up, can either lengthen or shorten the apparent lock time. these effects are a function of actual implementation and are not discussed here. the discussion which follows assumes that both transmitter and receiver are fully powered, and that the link is fully operational. the only effects included are pll transient effects. if there is no data on the link (if the transmitter is off, or if there is a quiet line) the data recovery pll will drift to its natural oscillation frequency. this frequency is determined by component values and tolerances inside the am7969 receive pll, and will vary slightly from both the receiver reference frequency (at x1 of the receiver) and the transmitter data frequency (x1 of the transmitter). when data appears on the line, the receive pll must achieve phase lock from its resting frequency. the structure of the pll used in the taxichip set ensures that this resting frequency will be no more than a few percent (typically less than 3%) from the reference frequency applied at x1. this is in addition to the specified transmitter/re- ceiver frequency mismatch allowed by the crystal tolerance specification of +0.1%. figure 3-1 calculated receiver lock-up time 80 70 60 50 40 30 20 10 0 12330e-4 lock-up time ( m s) percent offset frequency at 125 mhz 0.1 1 2 hq jk ii neglecting frequency variations in the transmitter and jitter in the data stream, the time to lock is related to the pll loop bandwidth and damping factor, and to the transition density. the loop parameters are set by the internal component values and tolerance of the taxichip set. a plot of calculated lock-up time vs transmitter to receiver frequency offset and transition density is given in the figure 3-1. note that low transition density causes longer lock times. in fact, at very low transition densities (1 transition per 10 bit times of the hq symbol), and large offset frequencies, the pll may not be able to acquire lock at all, even though the lock equation used to produce the graph seems to indicate a solution. as the limits are approached, lock time may grow to several times the value predicted by the lock equation.
amd 59 taxichi p inte g rated circuits technical manual 4.0 clock generation and distribution the serial baud rate for the am7968 transmitter is derived from a byte rate frequency source. the taxi receiver must run at the same frequency as the taxi transmitter. the relationship between serial baud rate and data byte rate depends on the width of the transmitted data. for 8-bit data, the byte rate is multiplied by 10 to obtain the serial clock rate. since the maximum operating frequency is 125 mhz and the minimum frequency is 40 mhz, the byte rate frequency range for 8-bit data is between 4.0 and 12.5 mhz. the multipliers for 9 and 10 bit data widths are 11 and 12 respectively. the following table summarizes the byte rate frequency ranges for each data width selected. am7968/am7969-125 data width pll multiplier byte rate 8 10 4.00 C 12.50 mhz 9 11 3.64 C 11.36 mhz 10 2 3.33 C 10.42 mhz am7968/am7969-175 data width pll multiplier byte rate 8 10 12.5 C17.5 mhz 9 11 11.37 C 15.90 mhz 10 12 10.42 C 14.58 mhz the source of byte rate frequency can be either from the built-in crystal oscillator or from a ttl clock signal. the maximum allowable mismatch between transmitter and receiver frequency sources is 0.1%. this tolerance is derived from the pll architec- ture in the taxi receiver, and from considerations of crystal accuracy. more information on crystal specifications and available distributors can be found in appendix c, taxi tip #89-05, taxichip set crystal specification . when there is no incoming data, the receiver pll has no serial data stream to track. this situation can arise if the transmitter has not been powered up, or if the transmis- sion medium is disconnected. in this case the vco will drift to a frequency determined by internal component tolerances. when data appears at the receiver serial input, the loop must acquire lock from this resting frequency. the worst case frequency offset and the capture range of the pll are designed to allow frequency mis-matching between transmitter and receiver of 0.1%, since this accuracy is achievable with inexpensive available crystals. 4.1 taxi transmitter clock connections the byte rate frequency source drives a multiplying pll to create an internal bit rate clock which is used for timing all internal logic. the x1 and x2 pins are used to input the byte rate frequency source to the transmitter. their exact usage will vary, depending on type of frequency source (crystal or external ttl) and mode of taxi transmitter operation (local or test).
amd 60 taxichi p inte g rated circuits technical manual 4.1.1 local mode transmitters in local mode, x1 and x2 are the crystal oscillator inputs. the external component connections are shown in figure 4-1. zero temperature coefficient capacitors (type npo) should be used for good temperature stability. typical crystal specification fundamental frequency 4.0 mhz C17.5 mhz +0.1% resonant mode parallel load capacitor (correlation) 30 pf operating temperature range 0 c to 70 c temperature stability 1.00 ppm drive level (correlation) 2 mw effective series resistance 25 w (max) holder type low profile aging for 10 years 10 ppm figure 4-1 taxlchip crystal connection reset am7968 or, am7969 x1 x2 c c c = 150 pf for a 12.5 C 17.5 mhz crystal, 220 pf for a 4 mhzC12.5 mhz crystal 12330e-5 the transmitter may also be run in local mode by applying a ttl frequency source to x1 and grounding x2. the ttl source may be either from a crystal oscillator module, or from a neighboring taxi transmitter clk output. in local mode, clk is the buffered output of the internal crystal oscillator. connecting the clk output of a taxi receiver directly to the x1 input of a taxi transmitter is not recommended, because the transmitters clock stability and jitter requirements are not satisfied by the receiver clk output. 4.2 taxi receiver clock connections the considerations and connections for the taxi receiver are similar to those for the taxi transmitter. the receiver x1 and x2 inputs connect to an on-chip oscillator, whose frequency is determined by a parallel resonant crystal, or is driven by an external ttl frequency source. the oscillator provides the reference, which sets the expected center frequency for the data synchronizing pll. the synchronizing pll tracks the incoming data and generates a bit clock from the serial data stream. all of the internal taxi receiver logic, including the logic that generates the clk output, runs on this bit rate clock. this recovered clock is as stable as possible in both frequency and phase, as it tracks the incoming data stream. in addition to the bit synchronization accomplished by the pll, the logic will maintain byte synchronization (framing) with the incoming data
amd 61 taxichi p inte g rated circuits technical manual using the sync symbol to define byte boundaries. if the byte boundaries must be re-aligned (on power-up or re-acquisition of signal), the logic will ensure that the clk is stretched (never shortened) upon re-sync to the new byte alignment. due to this behavior, the clk output from the receiver is not suitable as a direct frequency reference for another taxi transmitter or receiver. clk is intended to be used by the host system as a clock synchronous with the received data. figure 4-2 cascaded receiver clock connections serinC serin+ serinC serin+ serinC serin+ 12.5 mhz crystal osc rx1 am7969 (primary receiver) v cc cnb clk x2 x1 igm dms rx2 am7969 igm dms rx3 am7969 igm dms cnb cnb x1 x1 x2 x2 n/c from serial media 12330e-6 4.2.1 cascade mode receivers (am7969-125 only) when using an on-board ttl clock source, receivers which are in cascade mode should have their x1 pin tied to the crystal oscillator and their x2 pin grounded. figure 4-2 shows a typical cascaded receiver clock connection. the frequency source for the local mode receiver should be either a crystal oscillator (as shown) or another external ttl source. it should not be the clk output of another receiver. as discussed above, the clk output from the receiver is not suitable as a frequency source for other taxi receivers. 5.0 interfacing with the serial media the am7968/am7969 taxlchip set is capable of providing a high speed point-to-point serial link over fiber-optic, coaxial, or twisted pair media. the choice of the appropriate medium depends primarily on line length and data rate. this chapter discusses the issues involved in media choice and the requirements for driving different types of media. any taxichip set to media interface design must first take into account the electrical properties of the taxi transmitter and taxi receiver. the transmitter serial output drivers are open emitter, emitter followers which generate pseudo-ecl (pecl) levels when terminated by pull-down resistors to a voltage more negative than v ol . pecl is ecl referenced to the +5 v supply, so that v oh = (5C0.8) and v ol = (5C1.8) volts. a safe termination voltage which guarantees meeting v ol is 3 v or less. the receiver input is a long-tailed pair which will switch on 50 mv differential input voltage, with a large
amd 62 taxichi p inte g rated circuits technical manual common mode range. the average dc value of the input signal is therefore relatively unimportant. there are three broad classes of taxl-to-media interface: 1. very short (<3 2 link length), usually dc coupled. 2. terminated, dc coupled. 3. terminated, ac coupled. the short link is typical of a taxichip set to optical components connection. the terminated cases are used for driving cables, also optical or other components with incompatible power supply and/or logic level requirements may sometimes need circuits and layout that exceed 3 2 . 5.1 very short link, dc coupled for dc coupled inter-connections in which the distance between the serial pins and the next device is less than 3 2 , transmission line terminations are not necessary. all that is required is an appropriate pecl pull-down resistor, re. elimination of reflections is not required for these short line lengths because the round-trip propagation is significantly less than the 2 ns taxichip set rise and fall time. the effect of media mismatch in this case is distortion and slowing of the transition due to the addition of the reflection to the still changing edge. figure 5-1a standard load circuit v oh = 4.1 v v cc C2 v = 3 v i oh 50 w 12330e-7 figure 5-1b pull-down with i oh matched to standard load i oh re v oh = 4.1 v c 12330e-8 the lower limit for re is that value which produces the maximum value of l oh . in a standard pecl load circuit (figure 5-1a) l oh max is given by: (v oh C (v cc C2))/50 = (4.1C3)/50 = 22 ma if we return re to ground instead of 3 v (figure 5-1b), the minimum value of re becomes 4.1 v/22 ma, or 186 w . reflections due to mismatch can be minimized by locating the pull-down resistor at the end of the line, rather than the source. a mismatched line termination will give a reflection coefficient less than one while leaving the end of the line open will give a reflection coefficient of one (maximum reflection). note that the supply voltage and logic
amd 63 taxichi p inte g rated circuits technical manual level of the optical components must match those of the taxichip set in order for the dc connection to work. if the supply voltage or the logic levels are incompatible, an ac connection must be used. 5.2 terminated, dc coupled the parallel termination shown in figure 5-2 may be used for a dc connection of a taxi transmitter to a taxi receiver. the parallel termination provides both the line termina- tion (r1llr2 = z0 = line characteristic impedance) and a pull-down voltage. the thevenin equivalent of this termination is z0 pulled down to v cc C2 v, assuring both matched termination and adequate v ol . using v cc and a voltage divider provides pull-down voltage without the need for a separate power supply. figure 5-2 parallel termination r2 12330e-9 v cc r1 r2 (r1 + r2) v cc = v cc C 2 v r 1 | / r 2 = z 0 thevenin equivalent r2 v cc r1 serin taxi rx serout taxi tx pseudo-ecl driver copper media z0 5.3 terminated, ac coupled ac coupling is the connection of choice for many taxichip set applications. the typical arrangement for an ac coupled link is shown in figure 5-3. re is returned to ground to provide the pecl (pseudo-ecl) pull-down for the driver. the capacitor c blocks the dc voltage, and r1 and r2 terminate the transmission line and provide a dc bias level for the receiver. since only ac variations are passed through the coupling capacitor, the bias level at the termination should be set to the midpoint of the signal swing expected by the receiver input stage. note that this bias level is not the same as that which is recommended for the dc coupled case. the minimum value of re was previously established as 186 w , to avoid exceeding i oh max. the maximum value of re must be small enough to supply the transmission line with enough current to avoid cutting off the output driver. when switching from a high to the low state, the transmission line may cause the emitter voltage of the driver to fall more slowly than the base voltage, causing the output transistor to turn off. when the output transistor turns off, its output impedance becomes very high, causing the falling edge rate to be controlled by the external load (re and the transmission line). this variation in edge rate cannot be tolerated until the falling edge crosses the
amd 64 taxichi p inte g rated circuits technical manual threshold level of the receivers differential amplifier. once the receiver recognizes the state change, variations in the falling edge are not significant. to avoid edge rate variations due to driver turn-off, we must equate the voltage to which the driver is taken at turn-off with a point in the logic swing which will guarantee that the receiver changes state. since pecl logic swings are 800 mv, we may safely choose a 500 mv change at the driver (100 mv past the midpoint) as a guaranteed state change at the receiver. if the driver turns off instantly, we require the voltage divider formed by re and z0 to produce a 500 mv change from v oh . we can write: v oh re/(re + z0) = v oh C 0.5 4.1 x re/(re + z0) = 3.6 re = 7.33 z0 as a general rule, we may then say that: 186 < re < 7.33 z0 figure 5-3 pull-down and termination for ac coupled link 12330e-10 r2 v cc r1 serin taxi rx serout taxi tx pseudo-ecl driver copper media re c figure 5-4 serial link with output driver model 12330e-11 r2 v cc r1 serin taxi rx swi taxi tx pseudo-ecl driver (model) copper media re c v oh r0 b 5.4 baseline wander and the ac coupling capacitor the 4b/5b and 5b/6b data encoding schemes which are used by the taxichip set are run-length limited to a maximum of 3 consecutive low states (non-transitions in nrzi). this type of encoding ensures that on average there will be less than 10% variation in the dc component of the encoded data. when the encoded data is passed through an ac coupled link, the high-pass filtering of the ac coupling will introduce jitter because of the fluctuating threshold caused by the variation in dc component. this undesired side-effect of ac coupling is often described
amd 65 taxichi p inte g rated circuits technical manual as baseline wander effect and is illustrated in figure 5-5. in figure 5-5a, the average dc fluctuates between 40% and 60% of the maximum level (+10% of midpoint). after the signal is capacitively coupled (figure 5-5b), the average dc component is lost due to high-pass filtering, causing an undesired shift in the signal levels. this shift in the signal levels, coupled with non-zero rise and fall times of the serial stream cause pulse width distortion and thus apparent jitter and possible increased error rates. this dc shifting effect can be minimized if the values of the ac coupling components are chosen appropriately. the dc level of the data will fluctuate at a data-dependent frequency, f b , called the baseline wander frequency. the 3 db corner frequency of the ac coupling, f 3db =1/(2 p rc), should be chosen below the minimum baseline wander frequency of the data. this allows most dc variations to pass through the ac coupling high-pass filtering, minimizing the dc shift in the signal. to minimize f 3db we must maximize r and c. the resistance r is generally determined either by the termination required by the transmission line or by biasing requirements on both sides of the link. hence, only the coupling capacitor c can be maximized to keep f 3db as low as possible. the largest value capacitor that can be used is limited by the fact that it must be an rf capacitor. rf capacitors are generally of the ceramic type (npo and x7r dielectrics) and are limited to a maximum value of approximately 1.0 m f. 0.1 m f capacitors have proven to be sufficient in laboratory tests of taxichip set systems. for a 0.1 m f capacitor, we must verify that the capacitive reactance at the lowest fundamental frequency possible is less than 1 w . the lowest fundamental frequency possible is the frequency that results when the taxichip set is running at its lowest baud rate (40 mbaud) and the command or data pattern with the least number of transitions is being sent. this pattern turns out to be the hq command (fddi terminol- ogy) which has only 1 transition per command, or 1 transition per 10 bits when the command is encoded. if a continuous stream of hq commands are sent at 40 mbaud, the resultant fundamental frequency of the signal is 2 mhz. at 2 mhz, the capacitive reactance of a 0.1 m f capacitor is calculated as follows: xc = = 2 p fc 2 p (2*10 6 ) (0.1*10 -6 ) 11 0.8 w = hence, in the worst case a 0.1 m f capacitor will give a reactance of less than 1 w , as desired. in summary, the largest value rf capacitor available should be used to optimize the performance of the taxlchip link. figure 5-5 baseline wander 12330e-12 average dc level varies with data pattern a) data before ac coupling varying dc is filtered out causing an undesired dc shift in the data b) data after ac coupling
amd 66 taxichi p inte g rated circuits technical manual 5.5 interfacing to fiber optic transmitters/receivers the taxlchip set can be used in conjunction with optical components and optical fiber to form a simple fiber optic communication link. optical transmission has many advan- tages over conventional electrical transmission. these include: immunity to emi/rfi, low attenuation, electrical isolation, data security, and wide bandwidth. because of these features, use of optical fiber as the serial media will result in optimum performance of the taxichip set link. depending on the type of fiber and the optical components used, taxi links using optical fiber can cover distances of up to several kilometers. figure 5-6 shows a block diagram of a complete taxi fiber optic link. the optical components transmitters and receivers can be obtained from one of the sources listed in appendix a of this manual. the interface between the taxlchips and the optical components will be the subject of this section. figure 5-6 taxl-based fiber optic link optical detector: pin or avalanche photodiode 12330e-13 optical source: led or laser diode source driver electronics 8, 9, 10 4, 3, 2 fiber optic transmitter am7968 transmitter fiber optic cable receiver electronics fiber optic receiver 8, 9, 10 4, 3, 2 am7969 receiver taxi /optical interface data command optical connectors transceiver data command 5.5.1 dc-coupled taxl-fiber optic transceiver interface when passing data between the taxichip set and an optical module, care must be taken to assure that the logic levels of the taxichip set and the optical components are matched. if the supply voltages of the optical components do not match those of the taxichip set, then the logic levels will probably differ and the interface will require ac coupling to isolate these different levels. however, if the power supply requirements match those of the taxichip set (i.e. v cc = +5 v, v ee = gnd) and if the two components are connected to the same power and ground planes, a dc coupled interconnection may be sufficient.
amd 67 taxichip integrated circuits technical manual for dc-coupled interconnections in which the distance between the taxichip set and the optical module is less than 3 2 , transmission line terminations are not necessary. all that is required is the appropriate ecl pull-down as shown in figure 5-7 (1) . on these short line lengths, elimination of line reflections is not critical. however, without any increase in complexity or power consumption, line reflections can be reduced simply by locating the pull-down resistor, r e , at the end of the line instead of at the beginning. this reduces the reflection coefficient at the end of the line, and therefore, reducing the magnitude of the reflections. figure 5-7 dc-coupled taxl-fiber optic interface (unterminated) 12330e-14 taxi tx + C odl tx + C re re l < 3 2 , z 0 +5 v 186 re 7.2 (z 0 ) taxi rx + C odl rx + C l < 3 2 , z 0 +5 v re re optical fiber note: if the dc-coupled interconnection is longer than 3 2 , transmission line terminations are necessary. for this case, the suggested configuration is shown in figure 5-8. note that the line termination network also provides the desired pull-down to v cc C 2 v, sufficiently below the output low level of v cc C 1.8 v figure 5-8 dc-coupled taxl-fiber optic interface 12330e-15 taxi tx + C odl tx + C r2 r2 l>3 2 , z0 +5 v taxi rx + C odl rx + C +5 v r2 r2 optical fiber r1 r1 r1 r1 r1r2 r1+ r2 = z0 5r2 r1+ r2 = 3 v note: if the optical and taxi power and ground planes are decoupled as shown in chapter 6, ac coupling is always recommended to allow for variations in power and ground plane voltages. ac coupling is discussed in section 5.5.2. l>3 2 , z0 (1) adequate bypass capacitors have been omitted from this and the following figures to simplify the drawings.
amd 68 taxichi p inte g rated circuits technical manual 5.5.2 ac-coupled taxl-fiber optic transceiver interface some applications will require the taxichip set to optical transceiver interconnection to be ac-coupled. ac coupling should be used in the following situations: a) when the taxichip set and optical components are driven by a common power supply but the supply pins are decoupled using the scheme recommended in chapter 6, and b) when the taxichip set and optical components operate on different power supplies. ac coupling via capacitors along with the necessary design equations is shown in figure 5-9. in this configuration, re is the ecl output pull-down resistor, c provides the ac coupling, the connection is made with a transmission line (coax, twisted pair, microstrip) of length l and characteristic impedance, z0, and r1 and r2 provide a matched line termination and voltage bias to the midpoint of the optical components logic swing (v bb ). the configuration shown in figure 5-9 is recommended for any line length, l, which separates the taxichip set and the optical module. although the matched line termina- tion is not necessary for l<3 2 , the v bb bias voltage is always needed for ac-coupled links. therefore, even for line lengths where matched line terminations are not neces- sary (less than 3 2 ), the resistors r1 and r2 can be chosen to give a matched load without any added complexity. figure 5-9 ac-coupled taxl-fiber optical interface 12330e-16 taxi tx + C odl tx + C r2 r2 l, z0 +5 v r1 r1 c c v ee v bb re re v cc taxi rx + C odl rx + C re re l, z0 +5 v c c v ee r4 r4 v cc r3 r3 r1r2 r1+ r2 = z0 5r4 r3+ r4 = 3.7 v r3r4 r3+ r4 = = midpoint of pseudo- ecl signal swing v bb = midpoint of odl signal swing = v ih + v il 2 = v ee + (v cc C v ee ) r2 r1 + r2 186 < re < 7.2 (z0) c = largest rf capacitor available optical fiber 5.6 interfacing to coaxial cable in many applications, system cost can be reduced by using coaxial cable as the serial media. unlike optical fiber, which requires optical components between the fiber and the taxichip set, coaxial cable can be connected directly to the taxi serout pins, giving lower system costs.
amd 69 taxichi p inte g rated circuits technical manual because of the resultant lower system costs, coaxial cable is the recommended serial medium for short-to-moderate length links. at longer lengths, the advantages of fiber optic transmission (low attenuation, immunity to emi and ground loops, etc.) make it the media of choice. the maximum length possible for a coaxial cable taxi link depends on the type of coaxial cable used and the data rate. higher data rates will tend to limit link lengths because attenuation and pulse dispersion on coaxial cable increases with frequency. many different types of coaxial cables are available. some have far less attenuation than others however, with low loss generally comes increased size and rigidity. rg-58 is a commonly used, readily available type of coaxial cable. in lab tests using this type of cable, it was found that the taxlchip link could operate with a byte error rate of better than 10 C10 with a confidence limit of 95%, at byte rates of up to 12.5 mhz, at distances of up to 200 feet. the confidence limit accounts for the statistical nature in which errors occur in a digital system and it implies that we can be 95% sure that, under the given circumstances, the byte error rate will be 10 C10 or better. note that a byte error could have been due to a single bit error or more hence, the bit error rate may not be equal to the byte error rate divided by ten. using the taxlchip set in conjunction with coaxial cable as the serial media is quite simple. appropriate line terminations are required and ac coupling is strongly recom- mended to eliminate ground loops. the recommended configuration, including the necessary design equations, is shown in figure 5-10. each of the components that make up the interface serve the same purpose as in the ac-coupled taxl-fiber optic interface shown in figure 5-9. note that two coaxial cables comprise the link, one for each of the differential pseudo ecl signals. these two lines should be calibrated for a propagation delay difference of less than 0.2 ns. figure 5-10 coaxial cable interface 12330e-17 r1 taxi tx re l, z0 c c c re +5 v taxi rx r2 r1 + C + C +5 v r2 r1r2 r1+ r2 = z0 5r2 r1+ r2 = midpoint of pseudo-ecl signal swing = 3.7 186 re 7.2 (z0) c = largest rf capacitor available
amd 70 taxichi p inte g rated circuits technical manual sample values using rg-58a/u, 50 w coaxial cable, a successful taxi link was established using the following component values: r1 = 68 w r2 = 200 w re = 300 w c = 0.1 m f 5.7 interfacing to twisted-pair cable another low cost alternative twisted pair cable. twisted pair cable is generally more lossy than coaxial cable making it suitable only for short distances. to reduce the possibility of noise being induced along the line, the shielded twisted-pair cable is recommended. using the taxlchip set with shielded-twisted-pair as the serial medium is very similar to using it with coaxial cable. the recommended configuration is shown in figure 5-11, where each of the components that make up the interface serve the same purpose as in the ac-coupled taxl-fiber optic interface shown in figure 5-9. note that with shielded-twisted-pair, only one cable is required to form the link. the twisted-pair conductors carry the differential pseudo-ecl signals and the shield is grounded at the receiver. figure 5-11 shielded-twisted pair cable interface 12330e-18 r1 re l, z0 c c c re +5 v r2 r1 + C + C +5 v r2 r1r2 r1+ r2 = z0/2 5r2 r1+ r2 = 3.7 v = midpoint of pseudo ecl signal swing 186 re 7.2 (z0) c = largest rf capacitor available taxi tx taxi rx
amd 71 taxichi p inte g rated circuits technical manual sample values using ibm type 1 stp 150 w shielded twisted pair cable, a successful taxi link was established using the following component values: r1 = 101 w r2 = 291 w r3 = 300 w c = 0.1 m f 6.0 board layout considerations while the taxichip devices are digital in application, they are essentially analog parts, containing high frequency analog phase locked loops. reliable operation in a high frequency analog and digital environment requires that some simple board layout rules be followed. for example, most taxi applications which are laid out on a wire wrap board will not work reliably. because they have at most one power and ground plane, most wirewrap cards have insufficient separation between small signal current and digital switching current. digital switching noise can couple into the analog pll, causing phase errors and loss of synchronization. the preferred realization of a taxi application is on a printed circuit board, where the user can control the layout of power and ground planes. 6.1 printed circuit board layout 6.1.1 rules for layout the following rules should be followed to ensure minimal noise coupling: 1. use a pc board with separate gnd and v cc planes. 2. use two capacitors which differ by at least a factor of ten in value to decouple the devices. the reactance of large capacitors has a significant inductive component at high frequencies. because of this inductive component, a single large capacitor is not very effective against high frequency noise. two capacitors, one typically of 1 m f and one of 0.1 m f are more efficient at decoupling than a single large capacitor of 1.1 m f. the recommended layout is as shown in figure 6-1.
amd 72 taxichi p inte g rated circuits technical manual figure 6-1 transmitter and receiver decoupling layouts v cc1 (ttl) v cc2 (ecl) gnd1 v cc3 (cml) gnd2 (cml) am7968 ab c2 c1 c3 6 5 22 c1 c3 7 21 c1 c3 leads must be very short (less than 1/4 2 ) c1 = 0.1 m f (ceramic) c2 = 1 m f tantalum c3 = 0.01 m f (ceramic) v cc plane gnd1 v cc2 (cml) gnd2 am7969 d c2 7 20 8 21 c1 c3 transmitter receiver c3 to further decouple the taxichip set, it is highly recommended that ferrite beads be inserted at locations a, b and d. 12330e-19 c1 v cc plane v cc1 (ttl) figure 6-3 jogs and glitches in the clock line normal overshoot < 0.5 v clk or x1 jog or glitch normal undershoot < 0.5 v 12330e-20 3. keep all bypass capacitors as close to the power pins of the device as possible. lead lengths should be minimized. 4. use high quality rf grade capacitors such as type cog or x7r. use of z5u capaci- tors is not recommended. 5. ensure that the power supply does not have more that 100 mv of peak-to-peak noise at any of the taxi vcc pins. make this check while the taxls are sending random data. 6. while clk can drive four x1 inputs or several ttl loads, the highest performance can be achieved by reducing the load on the clk pin. care should be taken to en- sure that no jogs or glitches occur in the clk signal as shown in figure 6-3. if pre- sent, these glitches will be passed onto the pll and cause an occasional error. serial lines 7. run serial outputs parallel to each other, or one on top of the other at all times and route them away from the transmitter. do the same for serial inputs on the receiver. running these serial traces adjacently will minimize noise caused by these extremely fast signals on other traces. use of strip lines for serial signals is recommended.
amd 73 taxichi p inte g rated circuits technical manual 8. when terminating serial lines to or from the taxls ensure that the vcc rail or ground tap is not at a noisy location. resistors can couple noise from a power supply rail into the serial lines. vcc to ground decoupling adjacent to the resistors is recommended when using pullup/pulldown terminating resistor setups as shown in figure 6-4. when using only a pulldown, do not use decoupling as this could add more vcc noise into the serial signals. figure 6-4 decoupling terminations 12330e-21 r1 taxi tx re re +5 v taxi rx r2 r1 + C + C +5 v r2 no decoupling capacitor required decoupling capacitor gnd 6.2 layout using fiber optic data links because of their small signals levels, fiber optic data links require some care in layout. fiber optic data link receivers consist of a photo sensitive diode and an amplifier. the photo-diode converts light pulses into currents of around a few hundred nano-amps. this signal current is then amplified and translated into an ecl signal. taxi receivers and most digital chips switch hundreds of milliamps. if switching noise from the digital section of the board gets coupled into the optical data link, the signal from the light pulse data can be corrupted. to prevent the coupling of the optical data link output with other digital signals, the user must ensure that small signal and digital switching currents do not flow in the same path. this is done by separating both the optical v cc plane and the optical ground plane from the v cc and ground planes used by other digital circuitry. see figure 6-5.
amd 74 taxichi p inte g rated circuits technical manual figure 6-5 fiber optic data link decoupling note: this connection includes a ferrite bead in the v cc circuit of the fiber optic components. fodl gnd plane taxi v cc plane ferrite bead (part# 2743002111 by fair-rite) 0.1 m f 4.7 m f 4.7 m f fodl v cc plane as appropriate logic v cc plane power supply ground 12330e-22 logic gnd plane as appropriate as appropriate taxi gnd plane .1 m f 4.7 m f 0.1 m f taxi gnd plane 7.0 cascade mode operation the taxichip set can be cascaded to send multiple byte words over a single serial channel. cascade operation is the loading of n bytes of data into a taxi transmitter, and the serial transmission of that data through the transmitter to the media. data and command bytes are multiplexed into the local mode transmitter. detailed block diagram and explanation of the data multiplexing method are described in section 7.1. also see appendix c, taxi tip #14. for am7969-125 taxi receivers, the connection is parallel for data and daisy-chained for control. that is, the serin pins of all taxi receivers are connected together and to the media. the daisy chain of igm to cnb control signals determines which receiver latches which incoming byte. the receiver whose cnb input is connected to +5 v is the primary receiver, supplying the initial igm, as well as a reference frequency for the x1 inputs of the down stream taxi receivers. cascade mode does not increase data rate or throughput. the maximum data rate is 100 mbits per second, cascaded (2) or not. the advantage of cascading lies in the fact that the width of the data word is transparently maintained. if the application requires the transfer of 32 bits of data, cascade mode allows taxi transmitters to latch all the data, and send it over a single serial channel, and receive the data in proper order in four taxi receivers at the other end of the link. another advantage of cascade mode is that if a byte gets corrupted, the system can be reset by just sending a sync, ensuring that the first taxi receiver gets the first byte, the second receiver gets the second byte, and so on. performing this reset operation with latches would require additional logic to decode a sync command which in turn would reset all the latches. for am7969-175 taxi receivers, see appendix c taxi tip #13 for single receiver cascade operation. (1) actually, in non auto-repeat cascade mode, the throughput is less than 100 mbits/s due to the need to send syncs.
amd 75 taxichip integrated circuits technical manual figure 7-1 cascaded taxi system 8 4 12330e-1 clock dms x1 x2 serin+ serinC cnb dstrb do0Cdo7 co0Cco3 cstrb taxi rx #1 clock dms x1 x2 serin+ serinC dstrb do0Cdo7 co0Cco3 cstrb cnb igm vltn igm vltn taxi rx #2 taxi tx #1 ack data path control logic data path control logic mixed data sources data destination command destination data destination command destination 84 from an external ttl frequency source 12.5 mhz to next stage (note 2) 12.5 mhz v cc (note 1) taxi rx #1 * 84 x1 x2 dms tls serout+ seroutC di0Cdi7 ci0Cci7 strb ack clk
amd 76 taxichi p inte g rated circuits technical manual 7.1 transmit cascaded data with a single taxi transmitter for systems that require data transfer wider than a single byte, a single taxi transmit- ter can be used to cascade the multiple bytes. this operation allows the data to be multiplexed onto a single serial link, and then automatically demultiplexed and restored to the original word width. the taxi receiver performs this demux operation automati- cally when connected in the cascade configuration illustrated in the taxlchip data sheet and section 7.2. the circuit shown in figure 7-2 illustrates the basic technique that may be used to control multiplexing of word-wide data into a single taxi transmitter. this circuit assumes that the data to be transmitted is stored in appropriate registers that are all loaded simultaneously. while many systems will already include these storage elements, in the diagram these registers are shown as 74als374 octal d flipflops. they could be any register with the appropriate number of bits for the data, and a three-state controllable output. the registers are connected in a tristate mux configuration wherein each output can be selected individually. to clarify the illustration of the technique, the command lines are not used, and have been tied low. in systems that send commands as part of the data stream, these lines would be buffered in the same way as the data, except that the unused bits (or bytes) need to be held low when data is to be sent. the controller for the automatic multiplexer consists of a shift register that can be loaded with a 0 that shifts through and selects each data register in sequence, and strobes the taxi transmitter. in the attached figure, this shift register is a 74ls174 , but any collection of flip-flops would serve as well. the shifter is loaded with a 0 when strobe, the signal that loads data into the registers, is a 1 . the nand gate (u1) at the input of the first flip-flop assures that only a single 0 is possible while the registers are being selected. strb for the transmitter is derived from the clk output of transmitter, and is gated by the same signals that select the data. it is important that no glitches appear on the taxi strb input, since that will cause false data to be sent, and will disrupt the information transfer. to assure that any race -caused glitches appearing at the output of the four input nand gate (u2) are suppressed, the counter must be clocked on the falling edge of the clk. this assures that, during the time the outputs are changing, the low on the clk input of the two input nand gate (u3) will suppress anything happening on the other input. when clk rises, it will be the only signal active, and there should be no false strobes. this configuration also assures the longest possible setup time for the output of the data registers, since the strb happens immediately before the outputs change, and a full byte time before they change again. the other gates (u4, u5, u6) are only buffer and inverters used to assure proper signal sense, and fanout. they may not be needed in all systems. only four stages of shift register are required to select the four data registers, and the fifth stage shown in the figure is used to provide the sync character required for some cascade systems. the output of the fifth stage (ack1 stands for one sync) is used to ack systems that require a sync between data words. the output of the fourth stage (ack0 stands for no sync) can be used for ack in systems that expect to send contiguous data, and no syncs between words (auto-repeat cascade). either of these outputs can be connected back to the data strb input if the system is to run automati- cally, as in data sampling systems.
amd 77 taxichi p inte g rated circuits technical manual figure 7-2 cascade with one taxi transmitter dq ck dff dq ck dff dq ck dff dq ck dff dq ck dff dq ck buffers dq ck dq ck dq ck oe- oe- oe- oe- buffers buffers buffers 74als374 74als374 74als374 74als374 8 8 8 8 byte1 byte2 byte3 byte4 8 4 2 clk clk strb serout ack data command loaden u3 u2 8 8 8 8 32 u5 u4 u6 clk2 load4- load3- load2- load1- 74ls174 u7 u1 strbin data strobe ack0 ack1 clk1 serout clk buffer; may not be required for low fanout system four-byte cascade mode logic taxi 12330e-24 7.2 receivers in cascade mode: connections (am7969-125 only) unlike transmitters, all cascaded receivers are directly connected to the media, via the two serial input data lines. all receivers see the same serial data at the same time. the primary receiver always receives the first byte of serial data after a sync. the signals used by the upstream receiver to tell the downstream receiver that it has captured a byte are igm (i got mine) and cnb (catch next byte). after receiving its byte, the upstream receiver raises its igm signal, telling the next receiver in line that it is to catch the next byte on the serial line. in this way each succeeding receiver down the line catches each succeeding byte. the second receiver waits for the primary receiver to capture data before capturing its data (the second byte). similarly, if there were a third receiver it would wait until the second receiver had captured the second byte before capturing the third byte. the connections of the cascaded (downstream) taxi receivers are as follows (see figure 7-3): the cnb input of the cascaded receiver is tied to the igm of its upstream neighbor. the cnb input of the first upstream or primary receiver is tied high. the igm output of the last downstream receiver is left unconnected normally. (this pin is used differently in auto-repeat configuration, discussed in section 7.3).
amd 78 taxichi p inte g rated circuits technical manual x1 is connected to a common crystal oscillator or a ttl clock source. it is not recommended that x1 be connected to another receivers clk output. x2 is grounded. the dms pins of all taxls must be tied in the same state as the dms pins on the transmitters. the cstrb/dstrb pins on each of the receivers are all active simultaneously. a timing description for cstrb and dstrb is included in appendix c, taxi tip #89-10 taxi receiver cstrb and dstrb pulse width . the vltn pin has timing that is identical to the timing of the data out and the com- mand out lines. its connections are specific to each users applications. if cnb is high, the receiver will catch the next valid byte of data and hold it. it will not attempt to catch any more data until it sees a sync command from the transmitter or until its cnb goes low and then high again. figure 7-3 receivers in cascade mode cstrb cmd data dstrb vltn cnb dms clk x2 x1 igm v cc serin+ serinC serin+ serinC cstrb cmd data dstrb vltn cnb dms x2 x1 igm cstrb cmd data dstrb vltn cnb dms x2 x1 igm n/c serinC serin+ rx1 rx2 rx3 am7969 am7969 am7969 primary rx 12.5 mhz crystal osc d23-d16 d15-d8 d7-d0 *transmission line terminations not shown. 12330e-25 the following section describes the functionality of individual pins: the dstrb pin any one of the dstrbs may be used as the users dstrb to his system. when an entire word has been received (signified internally by a sync from the transmitter) the data in the receivers are latched out to the output ports and all the dstrbs are raised (simultaneously) one cycle later. likewise, if commands are sent as part of the cascade word, the cstrb/dstrb connections must be made appropriately. timing description for receivers in cascade mode is included in figure 7-4.
amd 79 taxichi p inte g rated circuits technical manual figure 7-4 receiver timing8-bit cascade mode internal clock* serin serial data nrz data* clk out igm taxi #1 = cnb taxi #2 data out command out dstrb out command out data out data n data n sync data 1 data 2 sync sync sync 3 2 no change command 0 11000100011100010001 no change no change no change no change cnb taxi #1 = 1 no change no change no change no change data nC1 no change no change no change no change data n no change 1 456 1100010001 data 1 data 2 11000100011100010001 cstrb out no change no change no change no change no change cstrb out dstrb out * internal signals 1100010001 sync sync no change no change taxi #2 taxi #1 12330e-26 if cnb is high, the receiver will catch the next valid byte of data and hold it. it will not attempt to catch any more data until it sees a sync command from the transmitter or until its cnb goes low and then high again. if cnb is held low, the receiver will not attempt to capture any data. when the primary receiver rx1 catches a valid data byte it will raise its igm (i got mine) so the next receiver rx2 can catch the next byte and so on down the line. after all the receivers in the system have received their bytes a sync must be sent or the next byte of data will be lost (3) . referring to figure 7-5 for a system of cascaded receivers. (3) in the auto-repeat configuration, a sync is not required.
amd 80 taxichi p inte g rated circuits technical manual figure 7-5 cnb and igm propagating down cascaded receivers time sync data 1 data 2 data 3 sync serial data cnb1 = v cc igm1 cnb2 igm2 cnb3 igm3 = n/c t 46 a b c d note: half of the byte is sufficient for the receiver to decide whether the byte is a sync or data. 12330e-27 when cnb on the first receiver is raised (figure 7-3 it is tied to vcc), its igm does not follow until the first half of a non-sync byte is detected in its serin. note that if a sync is detected, igm does not go high, since it is a sync that makes igm fall. the igm on rx1 rises when it sees a non-sync byte, then since it is tied to the cnb of rx2. rx2 will now be ready to accept the next byte of data. rx2 will now wait for the next non-sync byte to come down the serin lines. during this time all the other downstream receivers will ignore the data on the serin lines because their cnbs are still low. in the same way the upstream (primary) receiver will ignore the serin lines because it has already caught one byte and thus it will continue to ignore the data until it sees another sync. the igm on rx2 rises when it sees the second non-sync byte. in this fashion, each receiver will sequentially get ready to receive data as the cnbs propagate down the igms. when the first receiver sees a sync, it will lower its igm which is connected to rx2s cnb which will lower its igm and rx3s cnb and so on. in this way the low igm will also propagate down all the cascaded downstream receivers. cnb falling to igm falling is t 46 ns. in normal cascade mode, the cnb on rxl is tied high and thus, a sync has to be sent after all the receivers are full to ensure that rx1 is reset to accept the next byte of data.
amd 81 taxichi p inte g rated circuits technical manual the data out lines when a receiver sees the sync symbol it sends the data byte it just received from its input latch to its decoder latch, and then the receiver lowers its igm. one more clock cycle is required for the data to go to the output latch. at this point dstrb is raised. in this way all data bytes are output simultaneously from all receivers, two clock cycles after the first sync (or two clock cycles after a low cnb). the dstrbs of all the receivers rise simultaneously as well. the vltn (violation) pin in cascade mode the vltn pin acts exactly like a data out line. the timings are exactly the same. violations do not change the output of the igm pin. i.e., a receiver that gets a vltn will still raise its igm signal as if it received a valid data byte. 7.3 auto-repeat configuration 7.3.1 receiver connections in auto-repeat configuration in auto-repeat configuration the igm of the last receiver on the line is inverted and tied to the cnb of the primary receiver. this connection eliminates the need to send a sync between each data word. in a 3-receiver cascade system, igm3 is inverted and tied to cnb1. when the igm of the last receiver goes high, cnb1 goes low. cnb1 going low ripples through the chain pulling each igm low (t46 ns) until finally the last igm goes low again, pulling cnb1 high resetting rx1 to receive new data. in figures 7-6 and 7-7, as each receiver decodes its data byte, it raises its igm and thus the next receivers cnb. figure 7-6 taxi receivercascaded in auto-repeat serin+ serinC cnb igm dstrb cstrb cnb igm dstrb cstrb cnb igm dstrb cstrb cnb igm dstrb cstrb 12330e-28
amd 82 taxichi p inte g rated circuits technical manual figure 7-7 receiver timing in auto-repeat configuration sync data 1 data 2 data 3 serial data cnb1 igm1 cnb2 igm2 cnb3 igm3 = cnb1 note: only when a receiver has a cnb = 1, can it accept new data. it then raises its igm when it sees a non-sync byte. it wont accept another data byte until its cnb has gone low and high again. 12330e-29 when igm1 goes high, cnb2 goes high. this allows rx2 to decode the next byte and raise its igm. igm2 is connected to cnb3 and rx3 is now allowed to decode the next byte and raise its igm. in figure 7-8 since igm3 = cnb1 , cnb1 goes low. when cnb1 goes low, rx1 is reset and it pulls its igm low (t 46 ns). since igm1 is connected to cnb2, rx2 is reset and pulls its igm low t 46 ns later. cnb3 = igm2 goes low, which causes igm3 to follow it low t 46 ns later. igm3 going low makes cnb1 go high again and rx1 is now set to receive the next byte of data on the serin. see figure 7-9. thus, the cycle starts over again.
amd 83 taxichi p inte g rated circuits technical manual figure 7-8 receiver timing in auto-repeat configuration sync data 1 data 2 data 3 serial data cnb1 igm1 cnb2 igm2 cnb3 igm3 = cnb1 t 46 note: when igm3 goes high cnb1 goes low. thus, igm1 = cnb2 goes low t 46 ns later, and igm t 46 ns after that. this will ripple down to igm3. 12330e-30
amd 84 taxichi p inte g rated circuits technical manual figure 7-9 receiver timing in auto-repeat configuration sync data 1 data 2 data 3 serial data cnb1 igm1 cnb2 igm2 cnb3 igm3 = cnb1 data 4 data 5 data 6 note: igm3 = cnb1 so rx1 is now ready to receive new data. the cycle can now be repeated. 12330e-31 7.3.2 timing limitations of the auto-repeat configuration note, however, that the t 46 delay adds up as it ripples through the daisy chain. if the total delay from the first to the last receiver in the cascade is greater than 1 byte time, parallel data will output 1 byte time later on some receivers than on others. the following example is for t 46 = 20 ns and a 12.5 mhz byte rate, the time between the start of one byte to the start of the next is 80 nanoseconds. when igm on the last receiver goes high forcing the cnb1 on the first one to go low, it will take 20 x r ns (where r is the number of receivers in cascade) before the last igm goes low again, (allowing cnb on the first receiver to go high). in order for the first receiver to capture the next byte its cnb cannot remain low for more than x ns (where x must be less than 1 byte period). x = (20 x r1) + (inverter delay) + (cnb to clk set-up) (r1 is the number of receivers that can be connected in cascade in this format) the cnb to clk set-up time is specified as t 47 = [(byte time/n) C32 ns] in 8 bit mode at 12.5 mbyte/s, cnb to clock setup = - [(80/10) C32] = 24 ns figure 7-10 demonstrates an alternative scheme which will allow a virtually unlimited number of receivers to be cascaded. the fan-out of the inverter dictates the number of and gates that can be driven. multiple inverters can be connected to the last igm output if needed. using this scheme guarantees that all of the receivers in cascade will
amd 85 taxichi p inte g rated circuits technical manual output data at the same time this also guarantees that the cnb on the first receiver goes active (high) within 2 gate delays + 20 ns after it goes low. this leaves enough time for the first receiver to capture the (r+1)th byte of data. figure 7-10 taxi receivercascaded in auto-repeat configuration. configuration 2 12330e-32 cstrb dstrb cnb igm cstrb dstrb cnb igm cstrb dstrb cnb igm cstrb dstrb cnb igm serin + serin - figure 7-11 taxi receivercascaded in auto-repeat configuration. configuration 3 serin+ serinC cnb igm dstrb cstrb cnb igm dstrb cstrb cnb igm dstrb cstrb cnb igm dstrb cstrb 12330e-33 in practice, all the and gates are not required. using the above equation for x we can calculate a value of r1 for which x is less than 1 byte period at the appropriate fre- quency of operation. then if the number of receivers to be cascaded is greater than r1, an and gate is needed for every (r1+1)th receiver in cascade. the other receivers can be directly connected as shown in figure 7-11. syncs in auto-repeat configuration and recovering from errors a sync in auto-repeat configuration acts much like a sync in normal cascade mode. it resets all the receivers and their igms so the upstream (primary) receiver receives the next non sync byte of data. this remains as the method of recovering from byte framing errors. 7.4 unbalanced configuration (am7968/am7969-125 only) in reality there is no difference in connection between balanced and unbalanced configurations. the name only indicates that the number of transmit bytes and the number of receive bytes are unequal. the taxi receivers do not care how many data bytes the transmitter is sending to them. one data byte can be transmitted to several receivers. the only limitation here is
amd 86 taxichi p inte g rated circuits technical manual the drive capability of the transmitter and the termination circuit for multidrop transmis- sion lines. similarly, several transmit bytes can be multiplexed to one receiver. there are no drive considerations in this case. figure 7-12 shows an example of an unbalanced mode of operation in which one transmitter is connected to three receivers. figure 7-12 unbalanced configuration example: one transmitter to three receivers serin cnb x1 igm 8, 9, 10 am7969 v cc serin cnb x1 8, 9, 10 am7969 serin cnb x1 igm 8, 9, 10 am7969 + C serout 8, 9, 10 x1 x2 am7968 ttl data in clock osc 12330e-34 ttl data out note that in the unbalanced configuration, attention has to be given to where a sync will be needed. either the auto-repeat receiver configuration should be used or a sync must be provided every (r + 1) bytes, where r is the number of receivers cascaded together. more information on proper use and requirement of sync, refer to appendix c, taxi tip #8903. 8.0 test mode the phase locked loops (plls) in the taxlchips are designed to run within a fre- quency range that has been set for maximum efficiency and accuracy. the lower limit of this frequency range is 40 mhz. in test mode, the plls of the transmitter and the receiver are disconnected and the internal clock is applied from an external source. this allows the taxls to function at a much slower speed. this mode was designed to simplify the testing of taxls in an automatic testing production environment. a by-product of test mode is that it allows the user to run the taxls in systems that are slower than 4 mhz (the minimum byte rate). in this mode there is no minimum frequency. a system that needs to transfer data at low byte data rates can normally be imple- mented without modifying the standard setup, and test mode need not be used. when there is no data to be sent, the taxlchips will keep the line active by sending syncs.
amd 87 taxichi p inte g rated circuits technical manual the serial link will operate in the 40 to 175 mhz range as determined by the byte rate clock, but the byte data rate will be determined by how often the user strobes the taxi transmitter. the transmission speed is transparent to the user. some applications may have a serial link bandwidth limitation. typically, this means that the media connecting the transmitter to the receiver can only handle serial data rates that are lower than 40 mhz. the user can run the taxlchips in test mode in order to overcome the 40 mhz lower frequency limitation. for convenience in the following discussions, encoded data width n has been set to 10, corresponding to an 8-bit input byte, (i.e. dms = low). since the multiplying pll is turned off in test mode, an external clock source must be supplied to the taxls. in normal (non-test) mode, the transmitter pll multiplies the byte clock by 10. the new 10x clock is called the bit clock or bitclk , and is used to transmit the serial data. the receiver pll generates the same type of bitclk to decode the incoming data and to track and follow any fluctuations in the transmission frequency of the incoming data. in test mode the transmitter pll is disconnected and the internal clock multiplier is switched out. the internal logic is now clocked directly by the signal applied to the clk pin. the input to the clk pin now becomes the bitclk and must be supplied by the user. on the receiver side, the internal data tracking pll is disconnected in test mode. an external clock recovery circuit must be used to allow the receiver to track the incoming serial data stream. this recovered bitclk is supplied to x1. either a digital pll or an analog pll (for faster rates) can be used for clock recovery as shown in figure 8-2. the transmitter and receiver test mode connections and functionality are given in the following section. 8.1 transmitter connections refer to figure 8-1. the tls pin is left floating. this is the pin that puts the transmitter in test mode. the reset pin is left floating. reset pin function is described in appendix c, taxi tip #89-02. the x2 pin is grounded. serin is left floating (d/c = do not connect). the dms pin is set in the appropriate state for 8-, 9- or 10-bit mode as desired by the user. the clk is now an input for bitclk (the bit rate clock). this means that if the serial transmission rate is to be 1.5 kbits/s, clk must be 1.5 khz. the ack pin is raised only when a sync byte is detected in the transmitters shifter latch (note that if strb is lowered before ack is seen, ack will be suppressed. see the strb/ack description in section 7.1). the x1 input is the reset pin for the internal state machines and can be left unconnected in operational systems. for testing purposes, the following steps are to be taken upon power up or initialization. 1. x1 should be kept high and the transmitter bitclked about 15 times 2. x1 should be lowered and the transmitter bitclked about 200 times.
amd 88 taxichi p inte g rated circuits technical manual this serves to flush all extraneous data from the buffers and reset all internal state machines. once this is completed the transmitter may be strobed. x1 should be left in the low state upon completion of the initialization. the strb input must now be strobed only once every n = 10 bitclk pulses or more. this will allow time for an 8 bit wide byte to be encoded to 10 bits and shifted out one bit every clock pulse. the parallel data input pins are provided with new data every 10 bitclk pulses. setup and hold times remain the same as in non-test mode with respect to strb. (in the non-test modes, the clock rate is the byte rate and a new data word and a strobe is provided every clock pulse. in test mode, the clock rate is the bit rate so the new data word and strobe are provided every n clock pulses). in test mode the receiver expects only single ended data. thus only one of the serout lines from the transmitter is used. however, both lines must have pulldown resistors to electrically balance the outputs. figure 8-1 transmitter test mode connections ack strobe data in 8, 9, 10 command in 4, 3, 2 serout+ seroutC x1 x2 reset clk cls dms 300 w n/c = test mode bit rate clock generator divide by n or byte rate clock am7968 n/c tls 300 w media interface can be set for 8, 9, or 10-bit mode to receiver 12330e-35
amd 89 taxichi p inte g rated circuits technical manual 8.2 receiver connections refer to figure 8-2. grounding serinC puts the receiver in test mode. serin+ is a single ended 100k ecl nrz input. the x1 pin now becomes the bit rate clock input (bitclk), just like the clk pin on the transmitter. the clk pin remains a byte rate clk out. 8.3 timing relationships in test mode the timing parameters in test mode are similar to the parameters in standard mode. propagation delay values remain the same, however bit time relationships are now calculated with respect to the new bit times. for example, using a bitclk = 1.0 khz, which is a 1 ms period, the byte time t 35 = 10 bits x 1 ms = 10 ms. in the same way t 37 , which is the clk falling to strb rising delay is now [2 (t 35 /n) + 15 ns] = 2.015 ms. note that setup and hold times for serin to x1 are not specified and must be determined for each application. figure 8-2 receiver test mode connections dstrb data out 8, 9, 10 command out 4, 3, 2 vltn serin+ x2 reset clk dms clock recovery circuit digital or analog pll am7969 normal function cnb dms can be set for 8, 9, or 10-bit mode normal function x1 serinC igm cmd strobe data strobe single ended input from transmitter byte rate clk out cstrb 12330e-36
90 taxichi p inte g rated circuit technical manual appendix a fiber optic data link manufacturer list presented below is a partial listing of fiber optic data link suppliers that manufacture or market optical components in a data rate range compatible with the taxichip set. several of these components have been demonstrated in a bench level evaluation in conjunction with the taxlchip set. amp 1 (800) 552-6752 (416) 475-6222 (canada) at&t microelectronics (800) 372-2447 bt&d technologies delaware corporate center 11 suite 200 2 righter parkway wilmington, de 19803 (800) 545-4306 hewlett packard customer information center (800) 752-0900 sumitomo electric 777 old saw mill river rd. suite 230 tarrytown, ny 10591-6725
91 taxichi p inte g rated circuit technical manual appendix b error detection efficiency when a received data pattern does not represent a valid coding symbol, the taxi receiver asserts the vltn pin to indicate that the current data contains an error. the receiver cannot detect the occurrence of a bit error that transforms one valid symbol into another valid but incorrect symbol. this means that the transition error can change a valid data symbol into a different valid data symbol, or in certain cases a valid command symbol and not be flagged by the violation pin. a single noise event on the serial link can cause at a minimum a double bit error. single bit errors are assumed to be impossible (or at least rare) because nrzi encoding would require that the voltage level on the link be inverted after the event. there is no known error mechanism external to the taxichip set which could cause this condition. having confirmed that all errors are at least 2 bits wide, let us examine the location at which these errors can exist. consider the 4b/5b encoded data pattern for the taxichip set in the 8-bit mode. the output corresponds to two five bit nibbles for each eight bit data byte. shown below are four nibbles, or two bytes of encoded data output, with six possible locations for double bit errors within nibble 1 of byte 2. figure b-1 b 9 b 8 b 7 b 6 b 5 b 4 b 3 b 2 b 1 b 0 b 9 b 8 b 7 b 6 b 5 b 4 b 3 b 2 b 1 b 0 byte 2 byte 1 nibble 2 nibble 1 nibble 2 nibble 1 msb lsb msb lsb msb lsb msb lsb a b c d e f notes: error location a corresponds to a double bit error occurring in the least significant bit of nibble 2 and the most significant bit of nibble 1. error locations b, c, d and e occur within the nibble between adjacent bits, and, error location f occurs between the lsb of nibble 1 (byte 2) and the msb of nibble 2 (byte 1). 12330e-37
amd 92 taxichi p inte g rated circuit technical manual for example, consider transmitting hex b [1011], encoded as 10111. error e occurs changes bits b0 & b1, resulting in encoded pattern 10100, which is hex 2 [0010] 2 bits changed, and the run length the error = 4 bits, 1 0 1 1 0 0 1 0 becomes 2 bits changed, and the run length the error = 4 bits, a double bit error can change valid data into a violation, a valid command byte, a 1-bit, 2-bit, 3-bit,or 4-bit data error. a summary of the occurrence of these errors for the six error locations for 4b/5b encoding is summarized below in table b1. table b-1 error type abcdef v=5 v=5 v=5 v=3 v=3 v=1 c=5 c=3 c=5 c=3 c=5 c=1 1b=4 1b=0 1b=2 1b=4 1b=0 1b=14 2b=2 2b=8 2b=4 2b=6 2b=6 2b=0 3b=0 3b=0 3b=0 3b=0 3b=0 3b=0 4b=0 4b=0 4b=0 4b=0 4b=2 4b=0 table b-2 similar reasoning for the 5b/6b encoding scheme results in seven possible error locations, and the summary of the occurrence of these errors is listed below: abcdefg v=13 v=16 v=12 v=6 v=9 v=10 v=5 c=3 c=2 c=2 c=4 c=3 c=4 c=3 1b=10 1b=0 1b=0 1b=0 1b=8 1b=2 1b=22 2b=6 2b=14 2b=12 2b=14 2b=12 2b=12 2b=2 3b=0 3b=0 3b=4 3b=6 3b=0 3b=0 3b=0 4b=0 4b=0 4b=2 4b=2 4b=0 4b=4 4b=0 utilizing this information one can determine the efficiency of the violation logic in the taxi receiver. figure b2 summarizes the violation effectiveness, as well as depicting the number of bits in error in the undetected corrupted data. this information can be extremely useful in determining what, if any, additional error detection schemes should be implemented. figure b3 graphically represents the run length of the corrupted data for the undetected errors. as shown in this figure, there are a small percentage of unlimited run length errors. this is due to the few data patterns, which, when corrupted will cause a false sync pattern to be generated. this pattern will cause a running error which will continue until the next valid sync realigns the byte edge to its proper position. while these false syncs occur very rarely, these are the most dangerous errors in a taxi system, this very well may dictate the maximum user packet size.
amd 93 taxichi p inte g rated circuit technical manual figure b-2 60 50 40 30 20 10 0 percent of error events 8 bit 9 bit 10 bit violation 1-bit error 2-bit error 3-bit error 4-bit error 12330e-38 figure b-3 50 40 30 20 10 0 percent of undetected error events 1 bit 2 bit 3 bit 4 bit 5 bit 12330e-39 6 bit 0.104 0.027% 0.57 unlimited run length of error in corrupted data 8 bit 9 bit 10 bit
94 taxichip integrated circuits technical manual appendix c taxi technical information publications the taxi applications team has documented questions and answers that are general purpose in nature and applicable to a wide range of applications. this documentation has taken the form of taxi technical information publications (tlps), and have been incorporated in this revision of the technical manual. the contents of this appendix are as follows: taxi tip # 1: subject: receiver response to loss of input signal taxi tip # 2: subject: taxlchip reset pin function taxi tip # 3: subject: proper use for taxi sync taxi tip # 4: subject: taxi pll lock-up during power-on! taxi tip # 5: subject: taxichip set crystal specification taxi tip # 6: subject: taxi for fddi applications taxi tip # 7: subject: synchronous vs. asynchronous strobe taxi tip # 8: subject: taxi receiver lock time taxi tip # 9: subject: taxi bridge: bidirectional taxi communication taxi tip # 10: subject: taxi receiver cstrb and dstrb pulse width taxi tip # 11: subject: using receiver clk output to run a taxlchip transmitter taxi tip # 12: subject: taxlchip pins internal circuit taxi tip # 13 subject: demuxing a taxichip receiver to output multi-byte words taxi tip # 14 subject: 32-bit multiplexed cascade with the taxichip transmitter taxi tip # 15 subject: general device information for 125/175 mhz taxichips taxi e.b. nov 89: subject: taxlchip error rate example
95 taxichip integrated circuits technical manual taxi tips taxi t echnical i nformation p ublication #89-01 subject: receiver response to loss of input signal question: it is desired that the taxi receiver outputs be predictable and stable during conditions when the taxi transmitter may cease transmitting (power-off) or is disconnected. how can a system designer predict the taxi receiver outputs or use the taxi receiver in a system where the taxi receiver must appear inactive under these conditions? this applies to both coaxial and fiber-optic systems. answer: the key to this problem is interpreting the loss of incoming tx data as a quiet-line- state and either flagging the system accordingly or gating the taxi receiver outputs with an inactive flag. in a coaxial coupled system the loss of incoming signal drive will cause the taxi receiver inputs to rest at the input termination bias voltages. the differential serial inputs will normally be at equal potential. to properly interpret a loss of signal as a constant quiet state the input termination bias voltages must be altered slightly from each other. to allow reliable interpretation of the offset as a constant logic state, the offset voltage should typically be set to about 50 mv. the taxi receiver will interpret the quiet-line-state differently depending upon the operation mode (8-bit, 9-bit, or 10-bit). in 8-bit mode the receiver will generate continu- ous cstrbs with the command outputs at f hex, or all high. in the 9-bit and 10-bit modes there is no defined interpretation of an incoming quiet data stream. this will cause the taxi receiver to generate continuous cstrbs and the violation output will be continuously high. a one-shot may be used to determine violation=high duration and then generate an inactive flag, or the system may interpret the violation output directly as an inactive or invalid condition flag and halt data-dependent system opera- tions during any byte with violation=high. in a fiber-optic coupled system two methods may be used, depending on the optical receiver construction. if the optical receiver has a carrier-detect output this signal may be used to flag an inactive state. if there is no carrier-detect , one may be gener- ated using an ecl one shot arrangement which will detect loss of edges after a predetermined period of say, for example, 80 bit times. the actual time may vary depending on the desired response to loss of data. in either condition, if the receiver recovered byte clock (clk) is used as a system clock, then the system must be able to tolerate a shift in the clk frequency of typically about +/-3% to allow for rx-pll offset and drift during this period. (if the optical receiver begins oscillation when the fiber is dark, the taxi receiver pll may attempt to track the oscillation resulting in an indeterminate rx recovered clock frequency. optical receiver dark response thus becomes a possible system concern.)
amd 96 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-02 subject: taxlchip reset pin function question: how long must the reset pin be held low in order to insure that the taxichip has reset? answer: the reset pin is level sensitive and after a low input level is asserted it instantane- ously forces the phase lock loop (pll) to its lowest possible frequency (approximately 5 to 10 mhz). a 1 ms low pulse should allow sufficient time for the pll to reach a stable state. preliminary tests conducted in the lab reported that for the full taxi frequency and temperature range, the time required to recover from a reset was less than 100 m s. resetting is intended to allow graceful recovery from the rare occurrence of a pll lock-up due to noise bursts on the serial data lines, as may occur when light is removed from certain optical links. in a fiber-optic coupled system, loss of optical signal may cause the optical receiver to oscillate, causing the taxi receiver to track the oscillation to an indeterminate frequency. care must be taken to avoid the oscillation, or a reset can be used to recover from it. after reset, the pll begins tracking incoming data, and the byte boundary remains undefined until the transmitted data includes a sync (jk). the sync is a unique bit pattern which forces the taxi receiver to align itself to the correct byte boundary. in a coaxial system when a loss of incoming signal drive occurs, there will be no data for the taxi receiver to track. this quiet state will be interpreted as a continuous data pattern. the receiver decodes this quiet-line-state differently depending upon the operational mode selected: 8-bit, 9-bit, or 10-bit. in 8-bit mode the taxi receiver will generate continuous cstrbs with command outputs all high ( f hex ). in 9-bit or 10-bit modes there is no defined interpretation of an incoming quiet data stream thus generat- ing continuous cstrbs and forcing the violation output (vlin) to be continuously high. further information on the effects of incoming signal drive loss is available in taxi tip #89-01, receiver response to loss of input signal.
amd 97 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-03 subject: proper use for taxi sync question: what is the proper use for sync? how often is a sync needed? answer: when a transmitter has no data to send, it sends sync. this symbol allows the receiver pll to maintain phase and frequency lock with the transmitter, exactly as would a normal data stream. it has the additional special property of being a unique bit pattern that can be used to discover the byte boundaries in an otherwise continuous bit stream. the receiver pll takes some time to achieve phase and frequency lock (as described in section 3.3). after lock is achieved the taxi receiver must align the incoming data to the proper byte boundaries. the receiver logic compares the incoming bit stream (bit by bit and without regard for byte boundaries) with the pattern for sync, and when it is found, forces an internal bit counter to 0 . the internal counter then continues to count bits and run the byte rate logic without further reference until another sync is found. from this description of the sync function, it is obvious that only one sync symbol is ever really required to define the byte boundary for the taxi receiver if the internal counter continues to count correctly. it is unlikely that the internal logic function will make a mistake, and therefore the counter will continue to count off the proper number of bits per byte forever. however, there is some chance that noise can corrupt data into a pattern that looks exactly like the sync symbol. (the chance is about 0.13% of all possible error types.) when this happens, the byte boundary is forced to an incorrect position, and all data following is decoded incorrectly. taxi receiver violation detection logic may or may not flag the errors, but the receiver cannot distinguish properly framed data from incorrectly framed data. the only thing that can correct this running error is another sync. the minimum number of syncs required in a user data stream is dictated by the system sensitivity to running errors, and the systems built in error detection mechanism. however, it is a good practice to send a sync every 1000 bytes. systems that send packetized data, should allow a sync between each packet. this will assure that if an error occurs, it will be terminated at the end of the packet and will not corrupt the succeeding packets. systems that send data at a rate slower than the taxi byte rate will have sync automatically inserted as pad characters, so the user may not need to specifically insert them. systems that send byte or short phrase data (commands or control words, for example) might send a sync before each byte or phrase to assure that the message is not missed because of an earlier framing error. from this discussion it should be clear that there is no right number of syncs to send with taxi data. the correct number is dependent on the type of data the user is sending, and the system sensitivity to running errors.
amd 98 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-04 subject: taxi pll lock-up during power-on! question: is there a recommended power-on sequence for the taxichips to prevent pll lock-up? answer: early versions of the -70 taxichips did have some sensitivities associated with hot- plug-in lock-up, fast v cc rise time, and/or power supply sequencing during power-on causing occasional pll lock-up. these potential problems were addressed in the -125 taxichips with an improved circuit design. the -125 taxichips do not exhibit any known power-on problems, but a pll reset function is available on a package pin and may be used to restart the pll if problems occur. use of the external pll reset pin should not be necessary with the -125 taxichips, but there may be specific situations where its use may be helpful with the taxi receiver. there are possible situations where large amounts of noise may occur on the receiver serial data inputs. under these conditions the rx-pll may attempt to track the noise to an indeterminate frequency. this deviation in frequency may adversely affect data recovery when a good data stream reappears. system requirements may also place a restriction on the allowed clock frequency deviation. use of the pll reset function on the receiver may assist in the containment of some of these effects. all v cc s may be powered-on simultaneously with a common supply. it is only recom- mended that the taxi gnd pins be connected to a common ground and the v cc pins be separately filtered and decoupled to that ground. variations in power-on times due to separate v cc filtering and decoupling are not a problem. use of a common v cc supply is sufficient as long as noise filtering of the supply at the taxi v cc pins is adequate, therefore use of separate power supplies for the v cc s is usually not necessary. refer to section 6.1, for more information concerning power supply layout and decoupling. further information is also available in taxi tip #89-02, taxi receiver reset pin function.
amd 99 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-05 subject: taxichip set crystal specification question: what are the design considerations for crystals used with taxichip set? answer: the taxichips parallel mode oscillator uses a 4.0 mhz C 17.5 mhz crystal with a frequency tolerance of 0.1%. any crystal will oscillate in either series or parallel mode depending upon the type of oscillator used. by specifying parallel mode and the load, the manufacturer will calibrate the crystal in parallel mode at the desired frequency. this could be important in applications where frequency tolerance is critical due to the fact that the resonant frequency in parallel mode is typically 0.02% above the series mode frequency. as discussed in section 3.3, the time required to synchronize the data recovery circuit in the receiver is proportional to the delta between the transmitter frequency and the receiver pll resting frequency. the receiver resting frequency is typically less than 1% away from its own crystal frequency, due to mismatch between internal circuits. the crystal specification of 0.1% is meant to be small enough to cause a negligible effect on lock time while not increasing the crystal cost (0.025% crystals are quite common). sometimes tighter frequency and tolerance specifications may be necessary to meet the users system requirements (i.e. fifo depth, etc.). more information on crystal specifications for the taxichip set may be found in section 4.1.1. a partial list of vendors follows: typical 12.5 mhz distributor part number phone number i.e.a. hc18u 12.500 (408) 435-1000 calmax corp. um-12.5 (714) 957-1299 monitor products mm 49xle12a-12.5 (619) 433-4510 anderson electronics 011-668-03260 (814) 695-4428 target electronics cc025a-12.5 (408) 733-0384
amd 100 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-06 subject: taxl for fddi applications? question: can the taxichip set be used for fddi physical layer applications? answer: the taxichip set is code compatible with the fddi physical layer but there are restric- tions in the design which would cause difficulty in using the taxichip set for the physical layer of an fddi node. the taxichip set by itself cannot be used to build a fully compliant fddi node, although it provides several of the functions required. the taxi transmitter is compatible with fddi at the physical layer electrical interface and can send all codes specified by fddi. an exception to the encoding is that quiet- line-state (qls) is defined as fiber-dark for fddi, requiring a static serout=low, and the transmitter defines the equivalent of qls, as command f , as no-transitions , with no control of the static logical state. the taxi receiver is also compatible with fddi at the physical layer electrical interface and can recognize the codes specified by fddi, with restrictions. the restrictions concern master-line-state (mls), halt-line-state (hls), and the carrier detect function. mls and hls are terms describing a data stream composed of a consecutive string of hq and hh symbols respectively, representing a line-state condition. the receiver will decode these symbols, but it does not count them to signal line-states as required by fddi. mls and hls are relatively long run-length signals with 10 and 5 bit-times between transitions respectively, as compared to a maximum limit of 3 bit-times for data. the receiver pll was designed for wide operating frequency range, with tradeoffs in the ability and time required to capture long run-length data sequences. the fddi specifica- tion allows 100 m s for the receiver to lock upon and detect mls following a long period of qls. a typical taxi receiver will meet these criteria but the production parts are neither tested nor guaranteed for this condition. there are no problems associated with tracking the mls signal once the pll has acquired lock. hq and hh, within the taxi receiver, require proper byte framing for detection. mls and hls as specified by fddi are not framed, therefore the transition may be located at any of the ten bit locations. the result, as decoded within the taxi receiver, will be as follows: mls: 00100 00000 = hq cmd-a 10% probability 00000 00100 = qh cmd-d 10% probability all other = violation 80% probability hls: 00100 00100 = hh cmd-8 20% probability all other = violation 80% probability the fddi line state definition does not preclude the insertion of an occasional sync into the mls or hls data stream for proper framing, solving the recognition problem. if full fddi compliance is required, mls and hls must be detected external to the taxichip set. the carrier detect function, as specified by fddi, requires the flagging of a qls to the mac layer as long as the fiber is dark . the taxi serin inputs must be static for this condition to be met by the taxi receiver. this problem must be addressed directly by the optical receiver or gating of its outputs. functions of the fddi mac layer interface are not directly addressed in the taxi designs.
amd 101 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-07 subject: synchronous vs. asynchronous strobe question: when should synchronous vs. asynchronous strobing be employed? answer: inputs to the taxi transmitter can be strobed asynchronously, but with some limita- tions. in local mode, the strobe edge can rise at any time, without regard for place- ment within any particular byte. the data associated with the strobe (strb) will be latched into the transmitter, and will be transmitted at the earliest opportunity. if some system limitation insures that a second strb cannot ever be within the same byte time (80.0 ns at 125 mhz), then there is no need to observe the ack output of the taxi transmitter. if no such guarantee exists, then ack must be used to insure that no more than two strbs lie within a single byte. again there is no restriction on strobe placement within a byte, since ack will always insure that the capacity of the taxi input buffer is not exceeded. the taxi input buffering can handle two strbs within a single byte, but this additional buffer must be flushed by the absence of a strb in a given byte before another two-strobe byte is encountered. in an early revision of the 70 mhz transmitter, there was a particular placement of strobe with respect to the falling edge of clk, that could cause loss of data. this lead to several revisions of technical literature mentioning the prohibited zone , and uncer- tainty zone . this restriction in strobe placement has been removed in the 125 mhz version of the transmitter. t6 in the data sheet, now refers only to the exact placement of the internal byte boundary. knowledge of this time is only important for synchronous systems to predict in which byte the data will appear. the only restriction on local mode asynchronous strb input would be in systems which require contiguous data output from the taxi receiver. it is possible that when strobing asynchronously, there will be an occasional byte with no strobe, and another nearby byte with two strobes. the transmitter has sufficient buffering to handle this condition, but will pad the missing byte with a sync character. for systems that have this restriction, strb should be made synchronous with the clk of transmitter.
amd 102 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-08 subject: taxi receiver lock time question: in a fully operational system in which both the transmitter and receiver are powered on, how long will it take for the receiver to lock to new data after a quiet line? answer: when data transmission stops and the link becomes quiet, the taxi receiver pll will drift to its natural resting frequency which by design, is less than 3% away from the reference frequency applied at the x1 pin. when data appears on the line, the receiver pll will achieve phase lock in a time which is proportional to the incoming data edge density and pll loop bandwidth. because this lock time is dependent on the data being transmitted, the time it takes for the receiver to lock will depend on the specific system application. in section 3.3, three types of data are represented and their calculated lock times are shown. by dividing the lock time for a specific data pattern by the x1 clock period, the number of bytes to lock the pll can be calculated. because time to lock is dependent on many variables, it is represented as a typical time. if time to lock is critical to the specific application, we suggest you allow at least the times shown. although there is no guaranteed specification for time to lock, a test is run (using a jk pattern) as part of amd outgoing tests to ensure that all devices can achieve lock within a reasonable time. the test is performed by sending jks for 640 m s, and then without interruption, a full rate functional test is run. for the test to pass, the pll must lock to the jk pattern and then track the incoming data perfectly.
amd 103 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-10 subject: taxi receiver cstrb and dstrb pulse width question: what is the maximum cstrb and dstrb pulse width? answer: the internal logic of the taxi receiver determines the pulse width of cstrb and dstrb based on the timing of an internal clock (bit clock). under normal conditions, the pulse width will be 4-bit times wide in the 8-bit mode, and 5-bit times wide in the 9- and 10-bit modes. an exception to this typical width is upon re-sync which can cause the pulse to be expanded by up to 5 bit times as the byte boundaries are re-aligned to the incoming data stream. the number of bit times used to represent data differs based on the operational mode; in 8-bit mode, data is encoded into 10 bits, in 9-bit mode 11-bits, and in 10-bit mode 2 bits. for example, a receiver operating with a 12.5 mhz crystal and utilizing 8-bit mode will have a clock period of 80 ns (1/12.5 mhz = 80 ns). internally the receiver divides this period by 10, forming the internal bit boundaries used to represent the encoded data. this example yields a 8 ns (80 ns/10 = 8 ns) bit period, which translates to a internal clock rate of 125 mhz (1/8 ns = 125 mhz). figure 11. shows a timing diagram of a taxi receiver internal clock and its relationship to clk, data, and strobe outputs. the receiver utilizes this divided clock to define its internal logic states. the cstrb and dstrb signals are generated by using these logic states and have a fixed relationship to the incoming encoded data. the figure shows that from the beginning of the byte (state 0), the cstrb or dstrb delay is two internal clock periods before going high, and the signal remains high for four internal clock periods then returns to a low logic level. actual pulse width will vary from this ideal width due to signal rise and fall delay, propagation delay and effects of loads external to the receiver. the data sheet parameters reflect these delays and normal manufacturing guard bands.
amd 104 taxichip integrated circuits technical manual figure 11 (8-bit mode example) taxi receiver internal clock distribution 90123456789012 internal bit clock internal clock (byte rate) external clock (clk) internal cstrb & dstrb external cstrb & dstrb internal data external data 12330e-40
amd 105 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-11 subject: using receiver clk output to run a taxi transmitter question: is it possible to use the receiver clk output to drive the x1 input of a taxi transmitter? answer: to assure accurate transmission of data, the transmitter must have a stable, jitter free, byte rate reference to its multiplying pll. this is typically derived from a crystal and can be connected to any crystal controlled and noise free ttl source. the receiver synchronizes its internal clock with the incoming signal and recovers data and clock for use by the receiving host system. in the process of sending high speed data over typical serial links, the data may be affected by noise from various sources. the pll in the receiver removes this noise and delivers a synchronized clock to the receiver logic and to the host system. however, some of the noise may feed through the pll and appear on the clk output. the clk output can jitter as much as 2 ns when recovering data from a noisy link. this will not typically affect normal logic functions, and can be ignored. if the receiver must realign its byte boundary, it will stretch clk to the new alignment position and thus protect the host logic from shortened clk cycles. these noise and phase jumps make the receiver clk output undesirable for use as a transmitter frequency source. for systems that must use synchronized clocks (for example to avoid fifo re-timing logic) it is possible to filter the receiver clk output and make an adequate reference for the transmitter. there are two basic approaches to provide this filter. the first is to use a crystal filter (figure 12). when placed between the receiver clk and transmitter xl, the crystal filter can be effective in attenuating system jitter to levels nearly comparable to crystal controlled reference clock levels. by the nature of a crystal filter, as the frequency of the crystal used in the filter and the data rate frequency vary, the phase of the output varies. this will make the filter seem to have a variable delay (+ or -) which must be accommo- dated by the users logic. the second method is to use a pll tracking filter (figure 13). the jitter attenuation through the pll is less than that through the crystal filter because the pll has a bandwidth several orders of magnitude larger. the pll provides a solution whose merits lie between the simplicity of the crystal filter and the need for tight crystal tracking and matching. the pll filter is relatively straight forward. attention to proper grounding and board layout should be followed. the pll filter is more tolerant of component and environmental variations than the crystal filter.
amd 106 taxichip integrated circuits technical manual figure 12 crystal filter lt1016 am7968 transmitter serout + serout - x1 x2 clk am7969 receiver serin + serin - x1 x2 clk 300 300 3 2 1 7 +5 v 4 56 300 am7968 transmitter serout + serout - x1 x2 clk am7969 receiver serin + serin - x1 x2 clk clk gnd 1 m f crystal filter note : all crystals used are of same type. 4 pf + - 300 12330e-41 figure 13 pll filter lmc660c +5 v gnd 1 m f notes: 1. filter components were chosen for the following loop parameters: noise bandwidth = 10 khz damping factor = 0.5 natural frequency = 3.18 khz 2. refer to motorola mc4024, mc4044 and national lmc660c data sheets for specifications. + - v r v cc u1 d1 mc4044 gnd vcx2 out 2 mc4024 am7968 taxi transmitter serout + serout C x1 x2 clk am7969 taxi receiver serin + serin C x1 x2 clk gnd v cc gnd v+ v- 15k 15k 15k 15k 4.7k 4.7k 50 cx2 vcm2 gnd 1 m f 50 gnd gnd 50 pf 12330e-42
amd 107 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-12 subject: taxlchip pins internal circuit question: what do the taxichip i/o circuits look like? answer: there are five different input circuits and two different output circuits in the taxichip set. each i/o circuit has electro static discharge (esd) protection circuit attached to it. the five input types are: ttl input, ecl input, three state input (tsi), high threshold input (hti), and oscillator (osc). the two output types are: ttl output and ecl output. each i/o circuit and the esd circuit are shown on the following page. the following table lists all transmitter and receiver pins and their i/o type. pin pin name i/o type pin pin name i/o type 1 ack ttl out 1 do3 ttl out 2 strb ttl in 2 do2 ttl out 3 serout+ ecl out 3 do1 ttl out 4 seroutC ecl out 4 do0 ttl out 5v cc2 5 v 5 igm ttl out 6v cc1 5 v 6 reset hti 7v cc3 5 v 7 v cc1 5 v 8 reset hti 8 v cc2 5 v 9 dms tsi 9 serin+ ecl in 10 cls tsi 10 serinC ecl in 11 serin ecl in 11 dms tsi 12 ci0 ttl in 12 dstrb ttl out 13 ci1 ttl in 13 cstrb ttl out 14 di9/ci2 ttl in 14 vltn ttl out 15 di8/ci3 ttl in 15 co0 ttl out 16 di7 ttl in 16 co1 ttl out 17 di6 ttl in 17 do9/co2 ttl out 18 clk ttl out 18 do8/co3 ttl out 19 x2 osc 19 clk ttl out 20 x1 osc 20 gnd1 gnd 21 gnd2 gnd 21 gnd2 gnd 22 gnd1 gnd 22 x1 osc 23 di0 ttl in 23 x2 osc 24 di1 ttl in 24 cnb ttl in 25 di2 ttl in 25 do7 ttl out 26 di3 ttl in 26 do6 ttl out 27 di4 ttl in 27 do5 ttl out 28 di5 ttl in 28 do4 ttl out transmitter receiver
amd 108 taxichip integrated circuits technical manual gnd esd v cc 22k 11k gnd ttl in esd ecl in gnd 300 300 50k cc v cc ecl in esd ttl in esd dms/cls in v cc 17k 9k gnd three-state in 12330e-43 12330e-44 12330e-45 12330e-46
amd 109 taxichip integrated circuits technical manual v cc esd 3k 3k 48k reset ref. gnd esd esd 5.3k 125 125 125 gnd x1 x2 gnd high threshold in oscillator v cc 50 esd gnd ttl out ttl out esd ecl out v cc 300 300 gnd ecl out esd ecl out 12330e-47 12330e-48 12330e-49 12330e-50
amd 110 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-13 subject: demuxing a taxichip receiver output to multi-byte words question: how can a single taxi receiver be used to receive multi-byte words? answer: introduction for systems that require data reception wider than a single byte, a single taxi receiver can be used to cascade the multiple bytes. this operation allows the data to be demultiplexed from a single serial link and used by an external system. in the following example, data is captured sequentially and output in the form of four 8-bit words. commands, which can also be transmitted are not used in this example in order to clarify the basics of the technique. some simple modifications to include commands will be presented at the end of this technical note. the circuit shown in figure 14 illustrates the logic configuration that has been built and tested in the laboratory using nominal commercial parts. the circuit handles blocks of data typically ranging from four bytes to 64k bytes. the taxi receiver converts the serial information that is received by the serin+/C inputs to the data information that is output by the d0Cd7 data pins, c0Cc3 command pins, and the dstrb, cstrb, and vltn pins. this data information output by the single receiver is used by the controller to capture the incoming data and output it four bytes at a time. functional description controller circuit: the controller consists of a shift register constructed of four d flip-flops and a 3-input nor gate. the shifter is loaded with a 1 that progresses through the flip-flops sequen- tially clocking the first column of four registers which capture the incoming data. when the 1 is shifted through the fourth flip-flop, it raises the pco signal for the clkout d flip-flop. on the following rising edge of the /clk signal the bytes of cascaded data are simultaneously clocked out through the second column of four registers that buffer the cascaded data to the outside system. controller clock: the clock for the controller circuit is generated by or-ing dstrb and cstrb. this ensures that the dstrb signal is captured for output to the external system. these signals also prepare the way for a simple upgrade to allow the use of commands (explained later). sync commands: when not receiving blocks of data, sync commands (bytes) are received which keeps the taxi receiver locked onto the correct byte rate and byte boundaries. this ensures proper capture of the data at the beginning of the next block. in addition, before a block of data is to be sent, a sync command must be received to reset the counter to the proper byte alignment and initialize the system. the sync commands are sent by default in the system because they are automatically inserted whenever a byte time passes without a strb (no data to send) pulse at the transmitter. it is important to note that the receiver generates a cstrb and outputs zeros on the command lines when a sync command is received.
amd 111 taxichip integrated circuits technical manual 12330e-51 figure 14 logic diagram of cascaded data with one taxi receiver dstrb vltn d0 - - - - - - d7 oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp dstrb vltn d0 - - - - - - d7 oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp dstrb vltn d0 - - - - - - d7 oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp dstrb vltn d0 - - - - - - d7 oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp oe d0 d1 d3 d2 d4 d5 d6 d7 d8 d9 gnd vcc y0 y1 y3 y2 y4 y5 y6 y7 y8 y9 cp co0 co1 co2 co3 dstrb cstrb igm clm vltm 008 001 002 003 004 005 006 007 vcc1 vcc2 x1 x2 vcc nc clr gnd osc serin+ serin- cnb dms /reset dq clk qb clr prb dq clk qb clr prb dq clk qb clr prb dq qb clr prb dq clk qb clr prb dq clk qb clr prb taxi rx am7969 chmndo buffers vcc vcc taxi_clk /clk clk_cntr clr_cntr sync clk1 clk2 clk3 clk4 controller block pcd data capture registers data output registers byte 8 byte 1 byte 2 byte 3 clkout 4-byte demux cascade receiver 12330d-61
amd 112 taxichip integrated circuits technical manual the circuitry that handles the sync commands or sync bytes generates several signals. the cmnd0, clr_cntr, sync and pco are the signals that are generated by sync command logic. the clr_cntr signal is generated from the cmnd0 and the cstrb signal which signify a sync command has been received. clr_cntr clears the controller and then is latched by the rising edge of the receiver clk to form the sync signal. the sync signal then generates an active pco signal. the clkout is then driven high on the following rising edge of clk if clk4 has not already driven the clkout signal high. the sync command only clocks out the data when it is received before the fourth byte of data has been received. in all other cases, the data is clocked out by the logic involved with the fourth state of the controller. the sync commands that follow this sync command hold the clkout signal high to effectively hold the control- ler circuitry in a constant state of reset with no change to the output data. buffering: the buffering of signals should also be considered for this design. in this example, the data outputs from the taxi receiver drive the first column of four low power registers. this design does not exceed the driving capacity of the receiver, but if different parts are used, load calculations should be redone. this system should work with any standard logic, although logic families should not be mixed unless timing considerations have been made. this particular example uses low power schottky devices with relatively fast low power output registers. timing considerations some critical timing considerations must be met to ensure the proper operation of this design. in order to capture the dstrb signal, the timing of dstrb going active and the rising edge of the clkx signals from the controller must agree with the setup and hold times of the first column of registers. to ensure capture of sync commands, the clr_cntr signal becoming active and the rising edge of the receiver clk must agree with the setup and hold times of the sync flip flop. to prevent glitches on the clkx signals and the potential capture of incorrect data, the timing between clk_cntr rising and clr_cntr becoming active must be considered, clr_cntr needs to become active at a time before clk_cntr can effect the output of clkx. the timing diagram is shown in figure 16. figure 16 shows the timing of the system where one sync command is received between data blocks being received. the premature sync command is not shown, but can be derived by following the given timing diagram and known responses of the logic given in figure 15. upgrade notes command line handling: to add the capability to receive commands in this design, only a few additions are necessary. since this design uses 8-bit data mode, 4-bit commands can be used. it will be necessary to add command storage registers four bits wide as well as command output registers four bits wide to output these command lines correctly. the clkx signals as well as the clkout signals for the existing registers need to be connected to these new registers. the clkx signals may need to be buffered to meet fanout limita- tions of the controller circuitry. control signals: the signals that need to be output by the new features do not add to the logic. the circuitry to capture the cstrb signal is already designed into the system. the dstrb signal can be used as a cstrb indicator, active low, as well as a dstrb indicator,
amd 113 taxichip integrated circuits technical manual active high, without any additional logic. the vltn signal is used for both command and data violations. buffering of the dstrb and vltn signals may be necessary as illustrated in figure 15 to meet the drive requirements of the first column of registers. data/command output note: in this system, when a nonsync command byte is received the data line values corresponding to that byte will change to the values last output from the taxi receiver data lines. conversely, when a data byte is received the command line values for that byte will change to the values last output from the taxi receiver command lines. this is a characteristic of the example given and depends on how the command and data information is latched. altering the number of output bytes: the above described design is an example of a four byte cascaded serial data receiving system. the same design techniques can be used to expand or reduce the number of bytes output at a time by the system. the considerations that should be taken into account for an altered system deal with board space and part cost in accordance with the requirements of the situation at hand. for board space note see the pal usage section below. pal usage: in this system, the use of a pal could greatly reduce the amount of board space used. the pal could incorporate all of the flip-flops and buffering logic as well as the first column of registers that capture the system information. the reason a pal was not used in the system described above was to help ensure the understanding of the design concept. for this application it is recommended that the timing considerations men- tioned before should be investigated to ensure proper operation of the system.
amd 114 taxichip integrated circuits technical manual figure 15 timing diagram of 4-demux cascaded receiver taxi clk /clk dstrb cstrb rcvr data clkcntr clk1 clk2 clk3 clk4 clkcntr pco sync clkout data out byte 4 byte 1 byte 2 byte 3 byte 4 byte 1 byte 2 byte 3 41234 sync 123 valid data (for all four bytes) 12330e-52
amd 115 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-14 subject: 32-bit multiplexed cascade with the taxichip transmitter question: how can a single taxichip transmitter be used to send n-byte data words? answer: i. introduction many systems have data/cmd paths wider than the twelve lines available per taxi transmitter. amd taxi applications has designed a circuit which economically multi- plexes data words longer than eight bits using one taxi transmitter. the following discussion is specifically for an am7968 transmitter with a 32-bit data word, but is also applicable to systems with shorter or longer data words (with or without commands). ii. advantages there are several advantages to using the multiplexed data scheme utilizing one taxi transmitter as opposed to a system using several transmitters: 1) to implement the mux circuit for 32 bits requires one am7968 taxi transmitter and three relatively small integrated circuits. a 32-bit wide data path without multiplexing requires four am7968 taxi transmitters. 2) four am7968 taxi transmitters require more board real estate than three ssi parts and one taxi transmitter. 3) four am7968 taxi transmitters will dissipate about 3.5 w, while one am7968 and three ssi chips dissipate only 1.25 w. the power saving is even more dramatic if op- tical data links are being used. a design using four taxi transmitters and four am79h1000t optical data links would dissipate over 5 w of power. the same 32-bit wide system using the multiplexing circuit would dissipate only 2.6 w! iii. implementation implementation of the 32-bit multiplexed transmitter circuit is straightforward. (see figure 11). in addition to the am7968 taxi transmitter, the following parts are required: (1) 74ls00 (1) 74ls20 (1) 74ls174 a group of buffers with tri-state outputs (four am29c821s in this example), would likely be required in any type of point to point communication application and might already be available in the host system. additionally, a number of termination resistors are required. the number and values are dependent upon the type of coupling and the media used. iv. operation referring to figure 15, the data to be transmitted is assumed to be simultaneously loaded into the buffers when a strobe pulse is input to the system. the controller for the mux is the 74ls174, which is wired as a shift register. as a 0 (which occurs on strobe) is shifted through the register, each buffer is enabled in turn. the nand gate (u1) at the input of d1, ensures that only a single 0 is possible while the registers are being selected. the taxi clk signal, which is used to clock the 74ls174, is inverted to provide set-up time to ensure that no false strobes reach the taxi transmitter. the other four-input nand gate (u2) enables the two-input nand gate (u3), so that the transmitter will be strobed while there is data available in the buffers.
amd 116 taxichip integrated circuits technical manual jumpers are provided on the outputs of q4 and q5 to be inverted (u7) and fed back to the first nand gate (u1). if q4 is shorted back to the strobe input, the system will run in auto-repeat ack 0 mode. this means that there will be a strobe on every clock cycle. in this mode a sync will never be sent. if the output of q5 is shorted back to the strobe input, the system will run in auto-repeat ack 1 mode. this means that a sync will be automatically inserted in between each group of four data bytes. on the receiver end of the link, the option is left up to the designer to either use four am7969 taxi receivers or to demultiplex the data and use only one receiver. figure 16 has been included to give a detailed schematic of the circuit with an am7969 taxi receiver on-board to complete the data path. figures 17, 18, and 19 show typical outputs for auto-run ack0, auto-run ack1 and normal run modes. v. conclusion to increase the length of a data word beyond eight bits, multiplex the data into an am7968 taxi transmitter. this method uses less power, less board space, and lowers the parts cost of the system.
amd 117 taxichip integrated circuits technical manual 12330e-53 figure 16 32-bit multiplexed transmitter circuit 21 20 19 18 17 16 15 strb d0 d1 d2 d3 d4 d5 d6 d7 d8 d9 d10 d11 d12 d13 d14 d15 d16 d17 d18 d19 d20 d21 d22 d23 d24 d25 d26 d27 d28 d29 d30 d31 gnd am29c821 gnd am29c821 gnd am29c821 gnd am29c821 1a 1b gnd 1y 1d 1c vcc 2d 2y 2a 2b 2c 1 2 4 5 6 7 14 13 10 9 8 12 gnd d1 d2 d3 d4 d5 d6 clk clr 3 5 13 11 14 9 1 4 2 12 10 15 5 d1 d2 d3 d4 d5 d6 28 27 26 25 24 23 22 d15 d14 d13 d12 d11 d10 gnd1 gnd2 x1 x2 d18/c13 d17 clk d18 ack strb serout+ serout- vcc2 vcc1 vcc3 reset/ dms cls d19/c12 c11 serin c10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1a 1b gnd 2b 2a 2y vcc 4b 3y 3a 3b 4a 1y 4y 1 2 3 4 5 6 7 14 13 12 11 10 9 8 v c c c l k gnd ub dsc g n d 2 gnd 3 pwr s2 138 pwr gnd 4 s2 138 serin- serin+ gnd u2 u10 am7966 74ls0o 74ls2o u9 u1 74ls174 ack0 ack1 u11 jmp2 u12 jmp3 4 4 4 4 4 4 4 4 1 3 5 6 7 8 9 10 11 12 24 23 22 21 20 19 18 17 16 15 14 13 oc/vcc d0 d1 d2 d3 d4 d5 d6 d7 d8 d9 gnd y0 y1 y2 y3 y4 y5 y6 y7 y8 y9 cp u6 2 4 1 3 5 6 7 8 9 10 11 12 24 23 22 21 20 19 18 17 16 15 14 13 oc/vcc d0 d1 d2 d3 d4 d5 d6 d7 d8 d9 gnd y0 y1 y2 y3 y4 y5 y6 y7 y8 y9 cp u5 2 4 1 3 5 6 7 8 9 10 11 12 24 23 22 21 20 19 18 17 16 15 14 13 oc/vcc d0 d1 d2 d3 d4 d5 d6 d7 d8 d9 gnd y0 y1 y2 y3 y4 y5 y6 y7 y8 y9 cp u4 2 4 1 3 5 6 7 8 9 10 11 12 24 23 22 21 20 19 18 17 16 15 14 13 oc/vcc d0 d1 d2 d3 d4 d5 d6 d7 d8 d9 gnd y0 y1 y2 y3 y4 y5 y6 y7 y8 y9 cp u3 2 4 to coax or fiber media 12330d-63
amd 118 taxichip integrated circuits technical manual figure 17 autorun ack 0 (no sync between data bytes) 12330e-54 d 7 d 6 d 5 d 4 d 3 d 2 d 1 d 0 strb txclk dstrb cstrb figure 18 autorun ack 1 (one sync between every four data bytes) txclk 12330e-55 d 7 d 6 d 5 d 4 d 3 d 2 d 1 d 0 strb dstrb cstrb
amd 119 taxichip integrated circuits technical manual figure 19 normal run mode (transmission of syncs depends on host) 12330e-56 d 7 d 6 d 5 d 4 d 3 d 2 d 1 d 0 strb txclk dstrb cstrb
amd 120 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-15 subject: general device information for 125/175 mhz taxichips this t.i.p. provides general information about the design and manufacturing of the 125 mhz and 175 mhz taxichips. the information is separated into three categories: design, wafer fab, and assembly/packaging. design: transmitter (tx) receiver (rx) product p/n: am7968-125 am7969-125 am7968-175 am7969-175 die number: 4768 4769 chip dimensions: 170 x 167 mils 2 196 x 187 mils 2 # npn transistors: 3386 4384 # pnp transistors: 14 24 # resistors: 2504 3556 # diodes: 87 76 equiv. gate count: 595 720 i/o schematics: see taxi tip #89-12 supply currents: typical values (ma), v cc = 5.5 v process=nominal, cd 028 package, temp. forced with moving air flow (approx. theta jma = 20 c/w) C55 c0 c25 c70 c 125 c transmitter: 230 215 203 187 178 receiver: 272 250 237 212 182 wafer fab: location: fab 2a, san antonio, tx (formerly fab 11) process id: bipolar imoxCs2: 402lC1156 metal one: tiw (barrier metal): 1800 a nom. thickness alcu: 1.0% cu, 8000 a nom. thickness pitch = 4 m metal two: alcu: 1.0% cu, 15500 a nom. thickness pitch = 8 m passivation: silox/nitride dual layer. 7500 a nominal thickness nitride: 6800 a nominal thickness
amd 121 taxichip integrated circuits technical manual assembly/packaging: cerdip (cd 028) lcc (clt028) plcc (pl 028) assembly location: manila bangkok ld. frame material: cd: alloy 42 copper bond wire: 1.25 mil 1.25 mil au al/si (1% si) bonding method: ultrasonic ball bonding die attach: ag glass ag filled epoxy molding compound: n/a sumitomo 6300h lead finish cd 028 comm.: tin plate cd 028 mil.: solder dip clt028 mil.: solder dip pl 028: solder plate thermal impedance: cd 028 1 clt028 2 pl 028 2 q ja tx : 41 c/w n/a 53 c/w rx: 43 c/w n/a 52 c/w q jc tx, rx: 4 c/w 10 c/w 12 c/w 1 socketed 2 surface mounted
amd 122 taxichip integrated circuits technical manual taxi t echnical i nformation p ublication #89-nov '89 subject: taxichip error rate example introduction a method was devised to establish a baseline taxichip set error rate. a series of tests were conducted at a transmission rate of 125 mhz at room temperature, and various power supply voltages. the data collected will be used to determine fiber optic and wire interconnect ber (bit error rate) tests to be completed at a later date. method the test method used the taxi transmitters and receivers to transfer data continu- ously for at least one thousand hours per pair with different v cc conditions. to imple- ment this test, five taxi k2 boards were used. each board includes a taxi transmitter with a rom data source, and a receiver with a rom data checker to test data integrity on every byte. they were set up according to the diagrams in figure 20, and intercon- nected with ac coupled short coax lines. in setup 1, a single power supply with 5 v v cc was attached to taxi k2 board #1. board #1 ran independently with the serout+/C connected to its serin+/C with 50 w coaxial cables. setup 2 had oscillating voltages (4 v to 6 v) connected to the v cc of taxi k2 boards #2 and #3. the serout+/C of board #2 were connected to the serin+/C of board #3, and the serout+/C of board #3 connected to the serin+/C of board #2, with 50 w coaxial cables, forming two test setups with continually varying power supply voltages. setup 3 also uses two power supplies with one set at 4 v and the other at 6 v. these two power supplies were connected to taxi k2 boards #4 and #5. the serout+/- and serin+/- were connected in the same configuration as boards #2 and #3 in setup 2 with 50 w coaxial cables. the power supply voltages used (4 and 6 v) are outside the data sheet specification for the taxichip set. this test was intended to stress the parts and to simulate extreme temperature and operating conditions. these five boards were checked regularly, and the error counts were recorded. to verify that these boards were still running correctly, they were made to fail intentionally and then reset. result the tests were completed after each board ran more than 1,000 hours. the table below summarizes the results. board# hours errors notes 1 1,606 0 2 1,623 8* errors occurred between 438C558 hours 3 1,082 0 4 1,607 2* errors occurred between 438C558 hours 5 1,082 0 . * one error can cause multiple error counts. these were assumed to be one error event.
amd 123 taxichip integrated circuits technical manual board #4 failed on another occasion after the errors indicated above, but this failure was due to a power supply failure. failure time was subtracted from the total run time, and errors were not indicated in the total. also, boards #2 and #4 ran over 1,000 hours each without any error after the only noted error occurrence. conclusion the fives sets of taxi transmitters and receivers have run a sum total of 7,000 hours (3.15 x 10 14 bytes) with two error events. figure 20 taxi k2 board transmit/receive section taxi tx rom taxi tx rom and comparator ab 12330e-57 v cc v cc note: taxi k2 board includes both taxi tx and taxi rx . each half may be used independently or with other boards with matching rom data patterns. for the test described above, fddi ddj rom patterns were used. taxi tx taxi rx setup 1 5 5 setup 2 4C6 variable 4C6 variable 4C6 variable 4C6 variable setup 3 4 6 64 v cc conditions

▲Up To Search▲

Price & Availability of AM7968AM7969

	To Download AM7968AM7969 Datasheet File
If you can't view the Datasheet, Please click here to try to view without PDF Reader .