Agilent Transmitter MOI - InfiniBand® Trade Association

Agilent Transmitter MOI - InfiniBand® Trade Association
InfiniBand Trade Association
Revision 1.06
1/26/2014
IBTA MOI for FDR Devices
For Agilent 86100D DCA-X Oscilloscope Mainframe with 86108B Module
This material is provided for reference only. The InfiniBand Trade Association does not endorse the
vendor’s equipment outlined in this document.
This document is provided "AS IS" and without any warranty of any kind, including, without limitation,
any express or implied warranty of non-infringement, merchantability or fitness for a particular
purpose. In no event shall IBTA or any member of IBTA be liable for any direct, indirect, special,
exemplary, punitive, or consequential damages, including, without limitation, lost profits, even if advised
of the possibility of such damages
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 1 of 22
1/27/2014
Table of Contents
Introduction ..................................................................................................................................................... 4
Required Equipment ........................................................................................................................................ 4
Equipment Configuration ................................................................................................................................ 4
General Setup .................................................................................................................................................. 5
Test 1: Eye Mask test .............................................................................................................................................. 6
Test 2: Rise and fall time tests ................................................................................................................................ 8
Test 3: Vcm – AC common mode output voltage ................................................................................................ 10
Test 4: Vdiffpp – Differential p-p output voltage ................................................................................................. 12
Test 5: DDPWS..................................................................................................................................................... 14
Test 6: Normalized error “e”, “vf” and peak linear fit pulse response ................................................................. 15
Test 7: Calculating c-1, c0 and c+1 for other presets ........................................................................................... 17
Test 8: Jitter Measurements .................................................................................................................................. 18
Test 9: Far End Transmit Output Noise Measurements........................................................................................ 20
Test 10: Signal to Noise Ratio (Qsq) Measurements ............................................................................................. 22
Table of Figures
Figure 1 - Equipment configuration ........................................................................................................................ 4
Figure 2 - Eye Mask Test ........................................................................................................................................ 7
Figure 3 - Rise/Fall time test ................................................................................................................................... 9
Figure 4 - verify signal pattern.............................................................................................................................. 10
Figure 5 - ACVrms Measurement......................................................................................................................... 11
Figure 6 - Autoscale error message ...................................................................................................................... 12
Figure 7 - Vdiffpp measurement ........................................................................................................................... 13
Table of Tables
Table 1- Modification Record ................................................................................................................................. 3
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 2 of 22
1/27/2014
Table 1- Modification Record
Revision
1.01
Release Date
11/2/2012
Author
Rupert Dance
1.02
11/11/2012
Rupert Dance
1.03
9/16/2013
Alexander Rysin
1.04
10/7/2013
1.05
10/10/2013
1.06
1/26/14
Alexander Rysin &
Rupert Dance
Rob Sleigh & John
Dorighi
Rob Sleigh
Comments

Initial version of the document developed during Plugfest 22 in October 2012 with
the assistance of Daniel Fradkin and Vitaly Zavrajin from Mellanox.

Added more images, tables of content, figures and tables

Updated document format

Added details about Agilent Equipment

Added test scripts

Updated the DDPWS measurement method (Test 5), including new Matlab script

Updated Normalized error “e”, “vf” and peak linear fit pulse response (Test 6),
including new Matlab script.

Added transmit jitter measurements section (Test 7).

Added Far End Transmit Output Noise Measurements section (Test 8).

Added Signal to Noise Ratio (Qsq) Measurements section (Test 9).

Corrected inaccuracies in some of the test descriptions

Updated MOI to use FlexDCA user interface. Added a list of required equipment.


Added Equipment Configuration section with equipment setup graphic.
Added optional N1027A-PT2 to equipment list.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 3 of 22
1/27/2014
Introduction
This Method of Implementation (MOI) describes how to perform TX tests on an InfiniBand FDR device.
Required Equipment
Qty 1 Agilent 86100D DCA-X Wide-Bandwidth Oscilloscope with Options ETR/200/300/401
Qty 1 Agilent 86108B Precision Waveform Analyzer with Options LBW/216 or HBW/232
Qty 2 Agilent N1027A-PT2 (or Qty 1 86108B-PT2) 50 GHz Phase Shifters (optional)
Qty 2 Agilent N9398C/ N9398F DC Blocks (or equivalent)
Qty 2 Agilent 11901C 2.4mm (m) to 3.5mm (f) adapters (or equivalent) if using 86108B-HBW
Qty 1 Wilder QSFP+ Plug Test Adapter Model QSFP+ (or equivalent)
Qty 1 Device Under Test (DUT) – InfiniBand FDR
Equipment Configuration
Setup the test equipment as shown below in Figure 1 - Equipment configuration. The N1027A-PT2 50 GHz
phase shifters (optional) remove skew inherent in the system. The 11901C 2.4mm-3.5mm adapters interconnect
the Wilder cable fixture to the 86108B via a 50 GHz (2.4mm) DC Block. If a 26 GHz (3.5mm) DC Block is
used, swap the order of the DC block and 2.4mm-3.5mm adapters shown below.
Figure 1 - Equipment configuration
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 4 of 22
1/27/2014
General Setup
1. Select Run prior to making any test measurement.
2. Clock recovery setup procedure
a. Click on CDR (Clock recovery) button
b. Make sure Nominal Data Rate is set to the rate you want (for example 14.0625 Gb/s)
c. Clock Recovery Loop Bandwidth (LBW) Setup
i. Target LBW: Rate Dependent
ii. Rate Divider: Data Rate/1667 (8.4 MHz)
iii. Click on “Advanced” and configure the peaking setting:
1. Select Loop 1: 0dB (1st order PLL model)
Figure 2 – Clock Recovery Setup
d. Click on the Lock button – the indicator should turn GREEN; if the button is RED, then you are not
in synch.
e. Turn Precision Timebase (PTB) on.
3. Note: In Oscilloscope Mode, turn Pattern Lock ON to see the waveform, turn OFF (unlock) to see the eye
diagram.
4. Create a copy of the embedded spreadsheet for all ports you wish to test and select the appropriate tab for
each lane you wish to test.
a. There is a Tab labeled test Spec which provides details of the specification to be tested.
b. There is a Tab labeled example which provides some sample data.
c. There are Tabs for all 4 lanes of a 4X IB device
d. There is a Summary tab which summarizes the results for the core tests of all 4 lanes.
Test Results Spreadsheet
FDR Physical Layer
Measurements-v8-Template.xlsx
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 5 of 22
1/27/2014
Test 1: Eye Mask test
Purpose: Measure the eye and determine the Mask Margin for Hit Ratio of 5E-5.
Test Procedure
Note -To optimize measurement speed, it is best to acquire data first, and then turn on the mask test after the
waveforms have been acquired.
1) Vendor should set to PRBS31 and preset 0
2) Select Setup | Mode | Eye Mask Mode
3) Press the Channel button and select Diff (configures D1A difference operator)
4) Execute CDR (Clock recovery) procedure
5) Make sure that the Pattern Lock button is unlocked
6) Setup |Acquisition Setup…| Samples/Waveform  16200
7) Take at least 110 waveforms.
Select Measure | Limit Tests | Acquisition Limits | Check Enable Limits | Set Waveforms to 110.
8) Set mask test margins for desired hit ratio.
Select Measure | Eye/Mask | Mask Test Margins | Auto | Hit Ratio 5e-5 | Check Mask Test Margins On
9) Acquire the waveform data.
Click Clear | Auto Scale | Run
10)
11) Open the FDR mask.
Click the “Mask Test” tab. Select Load Mask File…. Navigate to the InfiniBand folder and open
InfiniBand_FDR.mskx
014.0625_Infiniband_FDR_Host_Tx.mskx
12)
13) The FlexDCA software will automatically grow the mask until it achieves a Hit Ratio of 5E-5. Record the
mask margin.
Note – refer to the FlexDCA Online HELP for more information on the Hit Ratio calculation.
(Total Margin Violations) / (Total Samples) ≤ Hit Ratio
where Total Samples is the # of samples in 1 UI; it is NOT the total samples acquired.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 6 of 22
1/27/2014
Figure 2 - Eye Mask Test
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 7 of 22
1/27/2014
Test 2: Rise and fall time tests
Purpose: Calculating the minimum rise and fall time for FDR Devices
Test Procedure
Note -To optimize measurement speed, it is best to acquire data first, and then turn on measurements after the
waveforms have been acquired.
1) Ask vendor to change to PRBS9 and Preset 0
2) Select Setup | Mode | Eye/Mask Mode
3) If the Diff (D1A) signal has not been configured…Press the Channel button and select Diff (configures D1A
difference operator)
4) Execute CDR (Clock recovery) procedure
5)
6) Setup # of samples/waveform.
Setup |Acquisition Setup…| Samples/Waveform  16200
7) Take at least 110 waveforms.
Select Measure | Limit Tests | Acquisition Limits | Check Enable Limits | Set Waveforms to 110.
8) Acquire the waveform data.
Click Clear | Auto Scale | Run
9) Configure 20%-80% rise/fall time
Click Measure | Measure | Configure Base Measurements. On the “Thresholds” tab, select “20%, 50%,
80%”.
10) Enable Measurements
On the “Eye Meas” tab, click:
a) Rise Time
b) Fall Time
c) Bit Rate (Optional). The displayed bit rate should be approximately the FDR bit rate.
11) Record the Values for both Tr and Tf
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 8 of 22
1/27/2014
Figure 3 - Rise/Fall time test
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 9 of 22
1/27/2014
Test 3: Vcm – AC common mode output voltage
Purpose: This test is used to measure the AC common mode output voltage
Test Procedure
Note: The test should be run with the device set to preset 0 and two other presets chosen at random
1) Ask the device vendor to change the device transmit PRBS9 and to the preset being tested
2) Select Setup | Mode | Oscilloscope Mode
3) Verify that you have good signal
a) If the Diff (D1A) signal has not been configured…Press the Channel button and select Diff (configures
D1A difference operator)
b) Execute CDR (Clock recovery) procedure
c) Enable Pattern Lock
d) Setup |Acquisition Setup… | Samples/Waveform  16200
e) Auto Scale
f) Display 511 bits onscreen.
Setup | Timebase Setup… |. Click “View Entire Pattern” button.
Figure 4 - verify signal pattern
4)
Setup Common measurement.
Select Measure | Waveform Signal Processing | Math Operators and define common mode operator as
shown below.
5)
a) Common (v+ plus v-)/2
b) Differential “D1” is v+ minus v6) Select the Amplitude Tab | Select RMS | Select entire display, AC and Function 2
7) Select Clear before recording the maximum value.
8) Record the value for ACVrms for Function F2.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 10 of 22
1/27/2014
Figure 5 - ACVrms Measurement
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 11 of 22
1/27/2014
Test 4: Vdiffpp – Differential p-p output voltage
Purpose: This test is used to measure the differential p-p output voltage with the driver disabled
Test Procedure
1) Ask the vendor to disable his device
2) Select Setup | Mode | Oscilloscope Mode
3) Configure Diff (D1A) signal (if not already selected)
Press the Channel button and select Diff (configures D1A difference operator)
4) Select the Amplitude Tab | Select RMS | select entire display AC and signal D1A
5) Select Autoscale
a) An error will occur because the signal is too small for the scope to synch on. Just click close.
Figure 6 - Autoscale error message
b) Record the value for ACVrms.
c) Record maximum value which we expect to be in micro volts but could be in milli volts. The
spreadsheet records mille volts so if the result is in micro volts, you need to convert. For example 428.46
uV is 0.42846 mV
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 12 of 22
1/27/2014
Figure 7 - Vdiffpp measurement
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 13 of 22
1/27/2014
Test 5: DDPWS
Purpose: This test is used to measure the Data Dependent Pulse Width Shrinkage
Test Procedure
1) Ask the device vendor to change the device transmit PRBS9 and to the preset 0
2) Enter Jitter Mode (86100D Option 200)
Setup | Mode | Jitter/Noise.
3) Configure Diff (D1A) signal (if not already selected)
Press the Channel button and select Diff (configures D1A difference operator)
4) Execute CDR (Clock recovery) procedure
5) Configure results in Unit Interval (UI).
Measure | Configure Jitter Mode Measurements…Select “Jitter Units” = Unit Interval
6) Select Meas Tab | DDPWS
7) Select Autoscale | Run
8) Record the DDPWS value into the spreadsheet.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 14 of 22
1/27/2014
Test 6: Normalized error “e”, “vf” and peak linear fit pulse response
Purpose: Test to calculate normalized error (linear fit) “e”, steady-state output voltage “vf” and peak linear fit
pulse response for preset 1
Test procedure:
1) Ask vendor to transmit PRBS9 using preset 1 and amplitude bit 1
2) Select Setup | Mode | Oscilloscope Mode
3) Configure Diff (D1A) signal (if not already selected)
Press the Channel button and select Diff (configures D1A difference operator)
4) Execute CDR (Clock recovery) procedure
5) Enable Precision Timebase
6) Setup Averaging
Setup | Acquisition Setup…. Select the “Smoothing” tab, select Average, # of Waveforms = 16.
Note - the spec requires you to average does not specify the exact number.
7) Pattern Lock the signal to view waveform
8) Configure Acquisition Setup
Select Setup | Acquisition setup…
a) Select Acquire Entire Pattern
b) Samples/Bit | Manual | 16 Total Record Length = 8176)
9)
10) Select File | Save | Waveform |
a) Under File Type, select: Pattern Waveform (*.csv)
b) Under Waveform Source, check Differential 1A
c) Save the file with the name waveform1.txt
d) Open file, ‘waveform1.txt’, in note pad and strip header information from waveform file so only the Y
values remain
e) Copy to a server and run the Matlab (Octave) script tx_equalizer_performance-v2.m with the
waveform1.txt and prbs9_sequence.txt in the same directory.
Matlab Script
PRBS9 Sequence
Sample Waveform data
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 15 of 22
1/27/2014
tx_equalizer_performance_v3.m
11) The interesting output from the script is “vf”, linear pulse fit peak, “e” and “w”
12) Note: vf is also known as DC Amplitude. Preset 1 is the basis for all other presets. Preset 1 is defined by its
amplitude (vf) and Linear Pulse Fit Peak - there is no pre-emphasis. Bigger the values indicate better signal
and less distortion. If you see a small value for Linear Pulse Fit Peak, it means the signal is distorted. You
don’t need to measure vf in other presets because spec does not demand it. This is because the value in the
other presets can be calculated from the values obtained in preset 1.
13) Also critical is the “w” vector which is used in other calculations.
a) Search for “w=” in the output and copy the numbers below into the script tx_equalizer_performancev3.m. Go to the line w_preset and horizontally put this into the script.
i) For example w_preset= [0.18258;3.42125;1.50535;0.67913;0.27759;0.16897;-0.01429];
b) Run script again and the c-1 and c+1 should be effectively zero and c0 should be effectively 1
vf = 0.83825
linear_fit_pulse_peak = 0.55555
rms_normalized_e = 2.9235e-004
c_minus1 = -2.3351e-006 – note this is effectively zero
c0 = 1.0000
c_plus1 = 2.0337e-007 - note this is effectively zero
c) This tells you everything is correct and w vector is correct.
14) Record the values for “vf”, linear_fit_pulse_peak and rms_normalized_e into the spreadsheet.
a) Note: the script reports the values as Volts whereas the spreadsheet assumes mille volts. So for example
0.55555 should be entered as 555.55
15) Note: there is no reason to record the “c” values for preset 1 because they should always be 0, 1, 0 (this is
the definition of preset 1.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 16 of 22
1/27/2014
Test 7: Calculating c-1, c0 and c+1 for other presets
Purpose: Calculating the tap weights c-1, c0 and c+1
Test Procedure
1) Ask vendor to transmit PRBS9 using another preset and amplitude bit 1
2) Select Setup | Mode | Oscilloscope Mode
3) Configure Diff (D1A) signal (if not already selected)
Press the Channel button and select Diff (configures D1A difference operator)
4) Execute CDR (Clock recovery) procedure
a) Possible problem: if the signal does not look correct, select pattern lock (to unlock) and then lock it
again.
5) Enable Precision Timebase
6) Pattern Lock the signal to view waveform
7) Configure the waveform Acquisition Setup
Select Setup | Acquisition setup…
a) Select Acquire Entire Pattern
b) Samples/Bit | Manual | 16 (Total Record Length = 8176)
8) Select File | Save | Waveform | a) Under File Type, select: Pattern Waveform (*.csv)
b) Under Waveform Source, check Differential 1A
c) Save the file with the name waveform1.txt
d) Open file, ‘waveform1.txt’, in note pad and strip header information from waveform file so only the Y
values remain
e)
9) Copy to the server and run the Matlab (Octave) script with the changed w vector.
10) Record the values from the output for rms_normalized_e, c-1, c0 and c+1 into the spreadsheet.
11) Record the values for Vcm for the other preset using the test procedure described in Vcm test section.
12) Important: Go back to step one of Calculating c-1, c0 and c+1 and ask the vendor to change to another
preset and execute all these steps again.
Important: Go back to Step one of Calculating normalized error and ask the vendor to change the amplitude
bit to zero (0) and execute all these steps again.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 17 of 22
1/27/2014
Test 8: Jitter Measurements
Purpose: This test is used to measure the device transmitter jitter
Test Procedure
Note: The test should be run with the device set to preset 0 and two other presets chosen at random.
1) Ask the vendor to configure the device to transmit a PRBS9 pattern with the preset being tested.
2) Enter Jitter Mode (86100D Option 200)
Setup | Mode | Jitter/Noise.
3) Configure Diff (D1A) signal (if not already selected)
Press the Channel button and select Diff (configures D1A difference operator)
4) Execute CDR (Clock recovery) procedure and check to make sure the precision timebase is active
5) Configure results in Unit Interval (UI).
Measure | Configure Jitter Mode Measurements…Select “Jitter Units” = Unit Interval
6) Press Auto Scale
7) Lock in the RJ value
Measure | Configure Jitter Mode Measurements | Advanced | Stabilization | RJ Stabilization
a) Select ‘Get Measured RJ’
b) Select ‘Fix RJ’ checkbox
8) Select Setup | Mode | Eye/Mask Mode
9) Turn Off (i.e. Unlock) Pattern Lock
10) Change Device Pattern to PRBS31
11) Select Autoscale | Run
12) Select Advanced Eye Tab
13) Select Advanced Eye Setup….
a) Change Jitter Units to ‘Unit Interval’
b) Activate Jitter Measurements for TJ(1e-12), DJ(dd), RJ , J2 and J9
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 18 of 22
1/27/2014
c) Accumulate data and wait until the jitter result stabilizes (typically ~ 30 to 60 seconds)
Note: The RJ value measured in Jitter Mode for the PRBS9 pattern is fixed and used as part of the DJ and TJ
measurement with Advanced Eye Analysis /401 for the PRBS31 pattern. The Advanced Eye Analysis requires
N1010A revision greater than or equal to A.02.50.
14) For preset 0 only, go to the “Configure the TJ Probability” and set the Total Jitter P1 to “J2” and the Total
Jitter P2 to “J9”. For any other preset set the Total Jitter P1 to “TJ”
15) For preset 0 – Record the J2, J9 and RJ values into the spreadsheet. Repeat the test with the “Total Jitter P1”
set to “TJ”, subtract the DDJ value from the TJ value, and record the result to the spreadsheet.
16) For presets other then 0 - Record the RJ value into the spreadsheet. Subtract the DDJ value from the TJ
value, and record the result to the spreadsheet.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 19 of 22
1/27/2014
Test 9: Far End Transmit Output Noise Measurements
Purpose: This test is used to measure the far end transmit output noise
Test Procedure
Note: The test should be run with the device set to preset 1 with amplitude bit set to 1. The test should be
performed for two cables – a low loss cable, having an insertion loss of 6 dB ±1.5 dB at 5.15625 GHz, and a
high loss cable, having an insertion loss of 15.7 dB ±1.5 dB at 5.15625 GHz. Prior to the measurements, the far
end insertion loss of the cables should be measured using the following procedure:
a. Measure the near end and far end crosstalk noise of the selected cables, using a vector network
analyzer (VNA). The frequency range of the measurements should be 10 MHz – 15 GHz.
b. For each lane, create a CSV worksheet arranged in the following manner:
 The first column will contain the frequency points of the VNA measurements in GHz.
 Columns B-E will contain the measured near end crosstalk values, in dB.
 Columns F-H will contain the measured far-end crosstalk values, in dB.
c. Save the attached Matlab (Octave) script icn.m in the directory of the created CSV worksheets.
icn.m
d. Open the saved Matlab (Octave) script icn.m. Adjust the following parameters:
 The filename of the CSV worksheet.
 “Ant” value should be set to the output amplitude of the tested device’s transmitters in mV.
 “Tnt” value should be set to the rise time of the tested device’s transmitters in ns.
 “Tft” value should be set to the rise time of the signal source for the aggressor lines
(crosstalk generator) in ns.
e. Run the Matlab (Octave) script. Record the calibrated far end crosstalk (denoted as sigma_fx).
1) Ask the vendor to configure the tested transmitter to transmit a low frequency square wave pattern
(0xFF00) with the other transmitters of the same port set to transmit a PRBS31 pattern.
2) Select Setup | Mode | Jitter Mode
3) If the Diff (D1A) signal has not been configured…Press the Channel button and select Diff (configures
D1A difference operator)
4) Execute CDR (Clock recovery) procedure
5) Control | Autoscale
6) Turn on Amplitude Analysis in Jitter Mode
a. Select ‘Amplitude Tab’
b. Press ‘Amplitude Meas ON’
7) Record the RN(rms) Ones and RN(rms) Zeros Value. This is the standard deviation of the noise on the
one and zero level with deterministic amplitude interference contributions removed.
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 20 of 22
1/27/2014
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 21 of 22
1/27/2014
Test 10: Signal to Noise Ratio (Qsq) Measurements
Purpose: This test is used to measure the transmit signal to noise ratio
Test Procedure
Note: The test should be run with the device set to preset 0.
1) Ask the vendor to configure the tested transmitter to transmit a low frequency square wave pattern
(0xFF00).
2) Select Setup | Mode | Jitter Mode
3) If the Diff (D1A) signal has not been configured…Press the Channel button and select Diff (configures
D1A difference operator)
4) Execute CDR (Clock recovery) procedure
5) Control | Autoscale
6) Turn on Amplitude Analysis in Jitter Mode
a. Select ‘Amplitude Tab’
b. Press ‘Amplitude Meas ON’
7) Record the Q Value in the Amplitude Analysis Window. Q is the ratio of the signal level to random
noise on the signal. Q = (Vone – Vzero) / (RNone + RNzero)
IBTA-FDR-Device-Tx-MOI-Agilent-v1.06.docx Page 22 of 22
1/27/2014
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising