Ultra640 SCSI with Receiver Equalization, 25 meters into a Backplane with 6 loads. Russ Brown Quantum Corporation

Transcription

1 T1/-154r Ultra64 SCSI with Receiver Equalization, 25 meters into a Backplane with 6 loads Russ Brown Quantum Corporation SCSI Physical Working Group Meeting 7 March 2 Dallas, TX

2 U64 25 Meter Cable Test Overview 7 March 2 T1/-154r Russ Brown - slide 2! Quantum's goal for Ultra 32 SCSI is to have a solution that is so robust it could be extensible to Ultra 64.! In order to demonstrate that our Receiver Equalization scheme is extremely robust, we want to test it at conditions beyond the specified limits of Ultra 16.! The first of these was to test U32 using a 25 meter round cable into a fully loaded 6-slot backplane.! The second test was to use the same setup (25m round cable into a loaded 6-slot backplane) at U64 rates.! The signals were measured to find the eye opening with ISI and reflections.! The following describes the test and results.

3 7 March 2 T1/-154r U64 25m Cable Test Setup Russ Brown - slide 3! Margins were evaluated with the same techniques as used for our other Ultra32 data:! Transmitter driving voltage: +/- 4mV.! Because of the higher frequency the transmitted pattern was 1µs of a"111..." training pattern followed by 4µs random data.! The equalizer input signals were captured differentially with a Tektronix TDS694C oscilloscope by probing at the backplane.! The equalizer output signal is generated by Spectre, simulating linear models and using captured data as the input stimulant.! The boost used for the equalizer simulation was 3x! Crosstalk could not be measured for this test as our current pulse generator cannot generate a synchronized clock at 32MHz.

4 7 March 2 T1/-154r U64 25m Cable Test Schematic Russ Brown - slide 4

5 7 March 2 T1/-154r U64 25m Cable Test Configuration Russ Brown - slide 5! 25 meter Amphenol cable assembly using Madison 28AWG round shielded cable plus 6-slot backplane.! Waveforms 1Gs/s Balanced 1 Ω Termination 25m bp1 bp3 bp5 bd 25m 28AWG round shielded cable Backplane Single balanced transmitter, position zero bp2 bp4 bp6 supplied by Amphenol, bp1 bp6, receivers on backplane

6 7 March 2 T1/-154r Error Sources Russ Brown - slide 6! Error sources are used to define the range over which a receiver characteristic may typically vary from the ideal sample point, i.e., the actual sample point may lie anywhere within a box defined by 2 times -to-peak height and 2 times -to-peak width of the errors. 3 mv 1 ns* U64 Eye Mask *This has been scaled to be half the value as that used for U32 analysis! Amplitude error sources define height, and timing error sources define width, e.g., set-up time margin is measured as the distance from the eye diagram waveform to the box.

7 U64, Eq Input, 25m Cable, bp6 (raw) 7 March 2 T1/-154r Russ Brown - slide 7 Amplitude in olts U64 Unequalied Signal Amplitude in olts Time -6 Equalier Output U TIme -6

8 5.5 U64, Eq Input, 25m Cable, bp6 Unequalied Signal for U64 25m round cable + loaded 6-slot bp meters round cable to a full load 6 slots bac plane Output Amplitude (mv) Output amplitude KEY: pattern Random pattern - - U64 Eye Mask Time in ns Time (ns) - Conclusion: No Margin (Increasing amplitude would still fail) 7 March 2 T1/-154r Russ Brown - slide 8

9 U64, Rx Equalized, 25m Cable, bp Equalized signal for U64 25m round cable + loaded 6-slot bp meters round cable to a fully loaded 6 slots bac plane Output Amplitude (mv) Output amplitude U64 Eye Mask KEY: pattern Random pattern Time in ns x - Time (ns) Conclusion: Good Set-up Margin* (*Increased amplitude would improve margin) 7 March 2 T1/-154r Russ Brown - slide 9

10 Set-up and Hold vs Eye Opening, bp6 7 March 2 T1/-154r Russ Brown - slide 1 Eye Opening Amplitude (mv) Eye openning Amplitude Setup and Hold Time vs eye opening amplitude U64 d6 25m round cable + loaded 6-slot bp6 Transmit amplitude -4m Cloc req. 2MHz Rx Equalized Equalizer signal Output solid Solid line line setup = set-up time time dash Dashed line hold line time = hold time U64 Eye Mask Unequalized Unequalized Signal signal (no eye o eye opening) setup and hold time x - Set-up and Hold Time (ns)

11 7 March 2 T1/-154r U64 25m Cable Test Conclusions Russ Brown - slide 11! Though this test was a "rough cut", it demonstrates that a receiver equalization scheme is so robust, it can adapt a signal to having sufficient margin from a signal having no margin at the receiver input.! In addition, the data indicate that a receiver equalization scheme could be developed to operate at U64 transfer rates without changing the SPI specification for the maximum bus path length between terminators (25 meters point-topoint and 12 meters multidrop interconnect).