LeCroy PCIE6bus Decoder Software User manual

Category
Measuring, testing & control
Type
User manual
PCIE6bus D Software
Instruction Manual
PCIE6bus Decoder Software Instruction Manual
© 2023 Teledyne LeCroy, Inc. All rights reserved.
This document does not contain export controlled information.
Users are permitted to duplicate and distribute Teledyne LeCroy, Inc. documentation for internal educational purposes only.
Resale or unauthorized duplication of Teledyne LeCroy publications is strictly prohibited.
Teledyne LeCroy is a trademark of Teledyne LeCroy, Inc., Inc. Other product or brand names are trademarks or requested
trademarks of their respective holders. Information in this publication supersedes all earlier versions. Specifications are
subject to change without notice.
November, 2023
pcie6bus-d-im-eng_28nov23.pdf
Contents
Introducing the PCIE6bus D Option 1
PCIE6bus Acquisition for Decoding 2
Oscilloscope Requirements 2
Planning Inputs 2
Before Acquisition 3
PCIE6bus Input Set Up 3
PCIE6bus Equalizer Set Up 4
PCIE6bus De-embedding 8
Generating an Eye Diagram 11
Serial Decode 13
Decoding Workflow 14
Decoder Set Up 14
Correcting Poor Quality or Inverted Signals 16
Failure to Decode 17
Serial Decode Dialog 18
Reading Waveform Annotations 19
Serial Decode Result Table 20
Searching Decoded Waveforms 26
Improving Decoder Performance 27
Appendix A: Automating the Decoder 28
Configuring the Decoder 28
Accessing the Result Table 28
Reading the Structure of the Result Table 28
Modifying the Result Table 30
Technical Support 31
i
PCIE6bus Decoder Software Instruction Manual
About This Manual
This manual explains the basic procedures for using serial data decode software options for Teledyne LeCroy
oscilloscopes. It is assumed that you have a basic understanding of the serial data physical layer specifications, and
how to use the oscilloscope on which the option is installed. Only features specific to this product are explained in
this manual.
While some images may not exactly match what is on your oscilloscope display—or may show an example taken
from another standard—be assured that the functionality is identical. Product-specific exceptions will be noted in the
text.
Some capabilities described may only be available with the latest version of our MAUI®software. Updates are
available from the software download page at teledynelecroy.com under Oscilloscope Downloads > Firmware
Upgrades.
ii
Introducing the PCIE6bus D Option
Introducing the PCIE6bus D Option
PCI Express®(PCIe®) is a serial interface standard now pervasive in computing systems and embedded systems.
PCIe utilizes point-to-point connections and typically consists of multiple lanes of both transmit (Tx) and receive (Rx)
datastreams, with data being "striped" across multiple datastreams to achieve higher data transmission rate.
The PCIE6bus D option applies software algorithms to extract PCIe 6.0 serial data information from physical layer
waveforms measured on your oscilloscope. When displayed on oscilloscopes or in MAUI® Studio remote
oscilloscope software, the extracted information overlays the actual physical layer waveforms, color-coded to
provide fast, intuitive understanding of the relationship between messages and other time-synchronous events.
PCIE6bus D decodes 1x1, 1x2, and 4x1 packets.
Note: If you have installed other -DME or -TDME options, the dialogs for Measure/Graph and Eye Diagram
creation will appear when the decoder is open. They may or may not appear "grayed out." We do not support
operation of the functionality for this protocol.
1
PCIE6bus Decoder Software Instruction Manual
PCIE6bus Acquisition for Decoding
Because the use of SDA Expert (SDAX) equalizers is necessary to achieve sufficient signal quality for decoding at
PCIe 6.0 64 GT/s, it is worthwhile considering how you will acquire signals and input them to the SDAX software in
order to decode different lane configurations.
Oscilloscope Requirements
The oscilloscope used for PCIe 6.0 must support at least 30 GHz bandwidth per channel. It is not strictly necessary
to utilize high-speed DBI channels for acquisition, depending on the fundamental frequency of the signal and the
standard channel bandwidth; you may be able to use all four 33 GHz channels on WaveMaster 8000HD
oscilloscopes or 30/36 GHz channels on LabMaster 10 Zi-A oscilloscopes. In general, however, best signal quality
will be achieved by using DBI.
Planning Inputs
No special probes or fixtures are required to acquire PCIe 6.0 signals for decoding; it is user discretion how you will
input the signals.
However, if you wish to decode bidirectional (PCIE6-1x2 upstream+downstream) traffic on any one lane, each
stream will have to be acquired as a separate differential pair, which in turn form the differential inputs of one SDAX
data "Lane". In other words, one physical lane of bidirectional traffic will equate to two logical lanes in SDAX.
If your signal permits, you can acquire both streams simultaneously over four standard channels, using each pair of
channels as the inputs to SDAX Lane1 and Lane2.
If the fundamental frequency of the signal exceeds the standard bandwidth of each channel, you will have to acquire
each stream separately over two high-bandwidth DBI channels, saving at least one acquisition to internal memory so
that it can be displayed and processed simultaneously with the other stream by the SDAX software and the decoder.
Example: You acquire one upstream lane of data on DBI channels C2B and C3B. You save these two acquired
channels to internal memories M2 and M3. When setting up inputs in SDAX software, M2 and M3 are used as
the differential inputs for Lane1 (the upstream data). You then acquire the downstream data on C2B and C3B.
These channels are then used as the differential inputs for Lane2. Lane1 and Lane2 can both be enabled
simultaneously so they are viewed together in SDAX, and the equalized Lane1 and Lane2 waveforms output
by SDAX are then selected as the two Source inputs to the PCIE6-1x2 decoder, decoding upstream and
downstream data simultaneously.
Likewise, if you wish to decode four lanes of unidirectional traffic (PCIE6-4x1), acquire four differential pairs , saving
two or three pair to memory. Use the memories as well as the last pair of "live" channels as the inputs to SDAX
Lane1-Lane4. These four lanes of equalized data output by SDAX will make the four inputs to the PCIE6-4x1decoder.
Note: SDAX data lanes are labelled Lane1 to Lane4. There is no Lane0 in SDAX. You will have to keep track
of which of your PCIe lanes are displayed by which SDAX lanes.
2
PCIE6bus Acquisition for Decoding
Before Acquisition
Start by confirming that your DUT(s) are able to transmit by viewing the input channel traces, and adjust the
oscilloscope vertical scale accordingly. It is not necessary to display the input channel traces in SDAX, but any pre-
processing applied to them will be reflected in the test results.
Make all other usual acquisition settings on the oscilloscope, particularly the timebase and trigger necessary to
achieve your desired test record length.
Note: Acquisitions must be at least 10 µs/div long in order for the eye diagram to process.
PCIE6bus Input Set Up
The inputs used by the PCIE6bus decoders, including all signal conditioning such as equalization, are configured
within the SDA Expert software. See the SDA Expert Software Instruction Manual for full instructions on using all the
SDAX functionality included with your decoder. The most important features are briefly described here.
1. Choose Analysis > Serial Decode.
2. Select the decoder to configure (from Decode1 to Decode4) and Protocol to decode (e.g., PCIE6-1x1).
3. Select the Setup button then the Input & Equalizer Setup button to open the SDAX dialogs. The Serial
Standard, Nominal Rate and other selections (e.g., Clock PLL) should already be set for PCIe 64 GT/s.
4. Select and Enable the first logical Lane you will configure as a decoder input, continuing to configure each
lane needed as described below.
5. When you have finished setting up the inputs, set up equalizers. Equalization is necessary for a successful
decoding of PCIe 6.0 PAM4 waveforms.
Serial Data Inputs
With the Lane you are configuring selected, open the Signal dialog. If you have navigated from the decoder dialogs,
the Nominal Rate should already be locked to 32 Gbaud, 64 GT/s.
If you are using a differential probe, or a Diff Channel already calculated from two channels by a WaveMaster
8000HD, choose 1 Input (or Diff. Probe) and select the Input1 source. If using two, single-ended inputs, choose
Input1-Input2 and enter the two channels or memories that make up that logical Lane of data.
Note: There is no need to configure a math function to calculate the difference between the two inputs.
When in Input1-Input2 mode, or when using any math or memory traces, be sure that both traces have the
same record length and sample rate.
You can choose to Upsample the input signal(s) in order to provide a higher sample density for analysis.
Levels & Thresholds
Enter the percent amplitude or absolute voltage level that marks the crossing threshold for each level in the signal.
Once there is an acquisition in buffer, touch the Find Level button to allow the software to calculate the actual levels
from the input signal. The software finds the best fit estimates of the four levels and their mean thresholds.
3
PCIE6bus Decoder Software Instruction Manual
To prevent false determinations due to noise, etc., enter a Hysteresis value. The default hysteresis band is 500 mDiv
(half of a division), centered on the crossing threshold. Edges that fail to transit the entire hysteresis band are not
detected.
Test Pattern
You can allow the software to Auto Detect the signal pattern sent from the DUT, use one of the Built-In patterns or
upload a custom pattern From File.
In general, Auto Detect is sufficient for identifying the pattern from the DUT. However, you may find it helps with
pattern identification to choose from the many Built-in PCIe signal patterns available, matching the speed and lane
being decoded.
In cases where the signal was connected to the oscilloscope with inverted polarity, you can use the Invert Pattern
checkbox to correct it.
PCIE6bus Equalizer Set Up
Equalizers are used to improve signal integrity by conditioning the signal in ways that counteract Inter Symbol
Interference (ISI) and other types of jitter or noise caused by channel response.
A selection of Continuous Time Linear, Feed Forward and Decision Feedback equalizers can be applied to the inputs
prior to outputting the signals to decoders for analysis. All equalizers can be trained on the input signal following a
brief acquisition for efficient equalizer set up.
Note: Equalizing is essential to correctly decode PCIe 6.0 signals, and the EQ dialogs will always be active
whether you are decoding PCIe 6.0 or a lower speed grade of PCIe.
When you have finished equalizer set up for all sources, go on to set up any required emulation/de-embedding.
CTLE
Continuous Time Linear Equalization (CTLE) is a linear filter that attenuates low-frequency signal components,
amplifies components around the Nyquist frequency, and filters off higher frequencies. When this selection is
enabled, a first-order CTLE is implemented according to the PCIe standards.
To apply CTLE to the signal:
1. Select the Linear EQ block of the SDAX flow, then on the Linear EQ dialog select a Data Path to DFE of either
From CTLE or From FFE to include the CTLE among the processors applied to the data stream.
2. Select the CTLE block, then on the CTLE Setup dialog, check Enable and enter the desired Boost in decibels.
The default Auto selection automatically sets up the reference CTLE for the speed. Once there is an
acquisition in buffer, the software will calculate the optimal number of poles and zeros needed to yield that
DC Gain value at the Nominal Rate.
The CTLE tap values are displayed on the CTLE Details dialog, but they will not be editable with Auto selected.
Select the Edit/View CTLE Setup button to view them.
3. If you are not satisfied with the results, you can use the procedure below to determine the best DC Gain
setting, or switch to Custom and make your own CTLE configuration on the CTLE Details dialog.
4
PCIE6bus Acquisition for Decoding
Finding Optimal Boost
The default Boost is a good starting point. If you do not see good transitions at that setting, try increasing the value
incrementally. The following method can be used to determine the efficacy of different values.
1. Display the unequalized source signal along with the equalized signal by checking Show SDAX Input on the
SDAX dialog.
2. Zoom in on both signals until you can clearly see the upper and lower transitions. It is OK if this is only a very
small section of the original acquisition.
3. Move both zoom traces onto the same grid so that they overlap, then place vertical cursors on the zooms at
both the upper and lower transition levels.
4. Go to Math > Zoom Setup > MultiZoom and add both zooms to a MultiZoom group with Same zoom position.
5. Drag the zoom traces left or right, or use the Zoom Auto-Scroll controls, until you find a section of the
acquisition where the equalized trace fails to cross the thresholds. This would cause a decoder to fail.
6. Examine the zooms and increase the Boost dB value until the entire equalized trace clears the transition
thresholds, as shown below.
Manually Tuning the CTLE
To modify the CTLE tap values applied, touch Custom and Edit/View CTLE Details to open the CTLE Details dialog.
Enter the custom settings to be used.
Continue to reacquire and modify until there is no longer any drop out of the equalized trace. Leave the Custom
button selected to use your CTLE settings.
Tip: We recommend using Auto unless you have experience configuring CTLEs and know what results your
selections will yield. The software is designed to automate as many of these selections as possible
according to the bit rate of the signal.
Note: If you later change CTLE settings, repeat Find Rate and Nominal Levels and retrain the FFE and DFE.
FFE
Feed Forward Equalization (FFE) boosts the amplitudes of symbols surrounding transitions while keeping the
transmitted power constant. In principle, FFE should be able to invert ISI if the number of symbols modified—that is,
the number of “taps”—extends over the entire length of the pulse response.
To apply FFE to the signal:
1. Select the Linear EQ block of the SDAX flow, then on the Linear EQ dialog select a Data Path to DFE of From
FFE to include the FFE among the processors applied to the data stream.
2. Select the FFE block, then on the FFE Setup dialog check Enable.
3. Make an acquisition so that there is data in the buffer on which to Train FFE.
Training FFE
The Levenberg–Marquardt method is used to train on the optimal tap values to use in the FFEfilters. The variable
that is minimized is the range of the high and low voltages near the center of the eye. Minimizing this quantity
maximizes the eye opening. Complete the following settings to configure the training algorithm.
5
PCIE6bus Decoder Software Instruction Manual
In # Desired Taps, enter a value that spans the number of UIs the signal takes when settling to a final value after a
transition. The training algorithm will create a filter with the number of taps entered.
Specify how many of those taps should be precursor taps in # Precursor Taps.
Note: The number of precursor taps is often around half of the total number of taps.
When Auto find levels is selected, the training algorithm will find the optimal Levels of the equalized eye as well as
the determination Thresholds to use, and these fields are disabled, although they will show the levels calculated
from the input signal. The voltage levels of the minimized locations of high and low states will tend to match the
outmost voltage levels of the unequalized eye.
To specify these levels instead, deselect Auto find levels and for each logical level in the signal, enter the:
lThe expected voltage Level corresponding the signal state (first column)
lThe voltage Threshold marking the transition to the next state (second column)
Touch the Train FFE button to begin the FFE training process.
FFE Details
While the best way to use the FFE is to train it, you can also enter the exact tap weights (if known) on the FFE Details
dialog. Touch the Edit/View FFE Setup button to display it.
Specify the # Taps Used in the FFE filter and the # Precursor Taps. If you began your setup by training the FFE, these
controls now show the values set on the FFE Setup dialog. Changing values in either place updates them
everywhere. Enter the tap coefficients on each row.
The Clear Taps button sets all the taps to 0 except for the first tap, which is set to 1. This is the pass-through state
for the FFE.
DFE
Decision Feedback Equalization (DFE) helps to account for distortion in the current symbol that is caused by the
previous symbols. It does so by feeding a sum of logic or symbol decisions back to the software. Tap0 represents
the boost applied to the post-cursor following all decisions.
Note: DFE requires a clock, which is recovered from the data following the application of the CTLE and/or
FFE. If DFE is enabled, you may have to enable CTLE and/or FFE to ensure the software can accurately
recover a clock.
6
PCIE6bus Acquisition for Decoding
1. Begin by touching the DFE block of the SDAX flow, then Enableon the DFE Setup dialog.
2. Optionally, select the Show EQ Out checkbox to display the signal as it appears after equalization.
3. If there is not yet an acquisition in buffer, make an acquisition on which to Train DFE.
Training DFE
As with the FFE, the DFE can automatically train using the Levenberg–Marquardt algorithm. The variable that is
minimized is the range of the high and low voltages near the center of the eye. Minimizing this quantity maximizes
the eye opening.
Complete the following settings to configure the training algorithm.
In # Taps, enter a value to span the number of UIs the signal takes when settling to a final value after a transition.
The training algorithm will create a filter with the number of taps entered.
In Max UIs for train, enter the maximum number of unit intervals to use for training. We suggest leaving this value at
1000. Higher values make the training slower, and much lower values may reduce the training accuracy.
When Auto find levels is selected, the training algorithm will find the optimal Levels of the equalized eye as well as
the determination Thresholds to use, and these fields are disabled, although they will show the levels calculated
from the input signal. The voltage levels of the minimized locations of high and low states will tend to match the
outmost voltage levels of the unequalized eye.
To specify these levels instead, deselect Auto find levels and manually enter the Target Levels & Thresholds, same
as with the FFE.
Touch the Train DFE button to begin the DFE training process.
The Clear Taps and Skew button resets all the taps to 0. This is the pass-through state for the DFE.
DFE Details
While the best way to use the DFE is to train it, you can also manually enter the exact tap weights (if known) on the
DFE Details dialog. Touch the Edit/View DFE Setup button to display it.
You will see the # Taps Used by the DFE filter. Enter the coefficients for each tap. Use Clear Taps to remove the
previous data.
DFEs are prone to burst errors (meaning, once errors are created, they run for a long time). The DFE's propensity to
create burst errors is based on the effectiveness of the DFE to aid in the accurate decoding of bits, which depends
on its ability to accurately decode bits - a case of circular logic. Depending on the strength of DFE applied, a single bit
error may lead to a long run of errors. Erasure DFE is used for this problem and improves the situation. Erasure DFE
effectively sets a band around the threshold value. When the signal falls inside the band at the time of slicing, it
indicates an uncertainty in the bit decoding. The receiver, although obliged to decode the bit, can then decide not to
apply this bit to the decision feedback, since the decoding was not certain enough. In operation, if erasure DFE is
utilized and the signal is within the voltage delta about the threshold specified, a voltage value that is the average of
the ideal +1 and -1 values is applied to the DFE delay taps, causing the bit to have no effect on decision feedback.
The Erasure DFE feature excludes certain bits within the specified Erasure Delta from the decision feedback. Use the
Erasure Delta control to set the indecision band around the DFEthreshold.
Note: As with the FFE, it is advisable to allow the software to train rather than try to manually configure the
DFE, unless you are experienced at doing so.
Clear and re-train whenever the input levels change.
7
PCIE6bus Decoder Software Instruction Manual
PCIE6bus De-embedding
Fixture De-embedding allows you to move the measurement reference to a point before the fixture, such as directly
to the transmitter output, by removing the effects of the fixture.
Channel Emulation / De-embedding allows you to to move the measurement reference to a different point in the
system than the probing point—either to the end of a channel by emulating the channel effects or to the start of a
channel by de-embedding the channel (similar to fixture de-embedding).
To perform either operation, touch the Emulate / De-embed block of the SDAX dialog flow to access the Emulate/De-
embed dialog, then check Enable to turn on the functionality.
When you have finished configuring any required emulation/de-embedding, go on to generate an eye diagram to
confirm the settings will yield a good eye, which usually indicates that the signal will decode properly.
Configuration
The Configuration control on the Emulate /Deembed dialog offers the following three configurations.
Emulate Channel /De-embed Fixture
This configuration de-embeds a fixture and emulates a serial data channel at the same time. It is best used when
you are using a fixture to measure at the transmitter side of the serial data channel. It first moves the measurement
reference to the output of the transmitter, and then moves the reference to the far side of a serial data channel.This
allows for emulating the effects of the serial data channel on your signal with the fixture effects removed.
De-embed Channel / De-embed Fixture
This configuration de-embeds both a fixture and a serial data channel at the same time. This is best when using a
fixture to measure at the receiver side of the serial data channel. It moves the measurement reference back to the
output of the transmitter. You then see what the serial data signal looked like at the transmitters output with the
effects of the channel and fixture removed.
De-embed Fixture Only
This configuration removes the effects of fixtures from you measurements. You can then see what the serial data
signal looks like at the transmitters output.
8
PCIE6bus Acquisition for Decoding
Bandwidth Limit
When de-embedding a fixture or channel, a bandwidth limit is required and must be set.
Note: When emulating a passive fixture or channel, the S-Parameter system does not have any boost, and
the bandwidth used is the highest bandwidth possible.
The Bandwidth Limit control imposes a reasonably sharp low pass filter in addition to the S-Parameter system
response. This is useful when de-embedding a lossy channel, to limit the amount of boost applied. When a channel
is de-embedded, high frequency response must be boosted to compensate for the high frequency attenuation in the
channel. However, if the signal has been attenuated into the noise floor, boosting the signal on the oscilloscope
makes it impossible to distinguish between the signal and the noise. The system boosts the noise along with the
signal. The BandwidthLimit setting can limit the overall response to the lower frequencies where signal components
are detectable above the noise. If this value is set to zero (the default) then no bandwidth limit is applied.
Auto BW allows the software to determine the limit, using a specified Max Boost and filter Order. Max Boost has the
same effect as Bandwidth Limit, but instead of setting the frequency, you set the maximum boost allowed. The
software looks at the S-Parameter responses and sets up the low-pass filter at the frequency where one of the
outputs has more boost than the specified Max Boost. The resulting bandwidth is reported to the user.
Bandwidth Limit is only required if the configuration actually includes de-embedding. If you choose the Emulate
Channel / De-embed Fixture configuration, but leave the De-embed Fixture block set to "Ideal" (pass through), the
bandwidth limit has no effect on the signal.
Upload and Map S-parameters
To emulate/de-embed a fixture or channel with loss, after choosing the configuration and setting bandwidth limits,
touch the De-embed Fixture or Emulate Channel blocks that appear in the flow. On the respective Fixture and
Channel dialogs, check Use Ideal Fixture or Use Ideal Channel. This eliminates the block from the process (it has no
effect on the signal), even if it appears in the configuration flow. No further configuration is necessary.
To emulate/de-embed a fixture or channel with loss, map the ports in the S-parameter files that define the loss
characteristics for each block where they are used.
Note: S-Parameter files should cover a frequency range up to (at least) 1/2 the oscilloscope's sample rate.
For example, if the S-Parameter file covers up to 10 GHz, the oscilloscope sample rate should be 20 GS/s.
When a fixture or channel is being emulated, there is little effect. However, if it is being de-embedded, then
the S-Parameter matrix must be inverted. S-parameters that do not have data up to half the oscilloscope's
sample rate can cause problems, particularly when de-embedding.
1. Touch Browse and navigate to the S-Parameter file representing the fixture or channel.
2. Choose the S-Parameter Format in use, either Single-Ended or Mixed Mode.
3. Assign to each Port a data column from the selected S-Parameter file. (This is useful for re-mapping from the
same S-Parameter file.)
4. Optionally, use View Response to plot the S-parameters.
9
PCIE6bus Decoder Software Instruction Manual
Apply Model
When you have completed the setup for each block in the configuration, touch Apply.Apply builds the system
description for the circuit and compiles it. The color indicator next to the Apply button shows the compilation status:
lGreen if everything is properly setup.
lYellow if the settings have changed and recompiling is needed.
lRed if something is wrong with either the S-Parameter file or port assignment. Choose View Response to plot
S-parameter Magnitude or Phase to help debug your issue.
Optionally, select Show Chann(el) Out to display the input signal as it appears following the emulation/de-embedding
processes, but before any other signal conditioning.
10
PCIE6bus Acquisition for Decoding
Generating an Eye Diagram
Use the PAM4 Eye dialog to generate eye diagrams of the equalized signals, confirming good signal quality for
decoding. To open it, touch the PAM4 Eye Meas. block of the SDAX dialog flow.
Modify your equalizer settings until you have achieved a satisfactory eye diagram, then choose the Decoder button
from the far right of the SDAX flow to return to the decoder dialogs and finish configuring the decoder.
Eye Diagram Set Up
1. On the PAM4 Eye dialog, select Enable Eye Meas.(ure) to turn on eye diagram functionality. Deselecting it will
turn off all eye diagrm views and parameters.
2. Select Show Eye to create the eye diagram for all displayed lanes.
3. Choose an Eye Scale setting that maximizes use of the vertical grid.
lNormal uses the vertical scale (V/div) of the data signal input to the Eye diagram algorithm. This
typically will be the V/div of the source when using 1 input, or the sum when using two inputs.
lAuto Fit All Lanes adjusts the vertical scales to a value and offset that places the top level at +2.5 div,
and the bottom level at -2.5 div.
lLock All Lanes sets the vertical scale of all eye diagrams to the values entered in the adjacent VerOffset
and Ver Scale entry fields.
Note: All eye diagrams generated for all lanes will share the same eye scale and style.
4. Optionally, enter a Delta X value in thousandths of a UI (mUI) to shift the horizontal position of the eye
diagram on the grid.
5. Choose an Eye Style:
l
With color-graded persistence , pixels are given a color based on the pixel's relative population and
the selected Eye Saturation. The color palette ranges from violet (lowest) to red (highest).
l
With analog persistence , the color used mimics the relative intensity that would be seen on an
analog oscilloscope.
6. Use the Eye Saturation slider to adjust the color grading or intensity. Slide to the left to reduce the threshold
required to reach saturation.
7. The Upsample factor increases the number of sample points used to compose the eye diagram. Optionally,
increase the value from 1 to a higher number (e.g., 5) to fill in gaps.
Note: Gaps can occur when the bitrate is extremely close to a submultiple of the sampling rate, such
that the sampling of the waveform does not move throughout the entire unit interval. Gaps can also
occur when using a record length that does not sample a sufficiently large number of unit intervals.
Using record lengths of ≥ 1 million points is recommended in order to acquire tens of thousands, if
not hundreds of thousands, of unit intervals.
11
PCIE6bus Decoder Software Instruction Manual
Eye Views
Show Eye displays the eye diagram of the input signal.
Show Levels visually annotates the measured voltage values of the 0, 1, 2 and 3 levels on the eye diagram.
Show IsoBER superimposes an extrapolated eye contour at a single BER onto each of the three eye openings. The
measurement locations of the eye heights and widths are annotated on these contours. Enter the BER value for
which you would like to show contours in the IsoBER plot.
Eye Descriptor Box
The Eye diagram descriptor box shows the V/div, Time/div, and total number of UIs represented
by the eye. If vertical cursors are placed on the eye, cursor readout values appear in subsequent
lines.
12
Serial Decode
Serial Decode
The methods described here at a high level are used by all Teledyne LeCroy serial decoders, differing only slightly for
signals with an embedded clock and separate clock and data signals.
Bit-level Decoding
The first software algorithm examines the embedded clock based on a default or user- specified vertical threshold
level. Once the clock signal is extracted, the algorithm examines the traffic to determine whether a data bit is high or
low. The default High and Low levels are automatically determined from a measurement of the amplitude of the
signals acquired by the oscilloscope. Alternatively, they can be manually set by the user. The algorithm intelligently
applies a hysteresis to the rising and falling edge of the serial data signal to minimize the chance of perturbations or
ringing on the edge affecting the data bit decoding.
Note: Although the decoding algorithm is based on a clock extraction software algorithm using a vertical
level, the results returned are the same as those from a traditional protocol analyzer using sampling point-
based decode.
Logical Decoding
After determining individual data bit values, another algorithm performs a decoding of the serial data message after
separation of the underlying data bits into logical groups specific to the protocol (Header/ID, Address Labels, Data
Length Codes, Data, CRC, Parity Bits, Start Bits, Stop Bits, Delimiters, Idle Segments, etc.).
Message Decoding
Finally, another algorithm applies a color overlay with annotations to the decoded waveform to mark the transitions
in the signal. Decoded message data is displayed in tabular form below the grid. Various compaction schemes are
utilized to show the data for the duration of the acquisition, from as little as one serial data message acquisition to
many thousands. In the case of long acquisitions, only the most important information is highlighted, whereas with
the shortest acquisition, all information is displayed with additional highlighting of the complete message frame.
User Interaction
Your interaction with the software in many ways mirrors the order of the algorithms. You will:
lAssign a protocol/encoding scheme and data sources to one of the four decoder panels on the Serial Data and
Decode Setup dialogs. Each decoder can utilize different protocols or data sources, or have other variations,
giving you maximum flexibility to compare different signals or view the same signal from multiple perspectives.
lComplete the remaining subdialogs required by the protocol/encoding scheme. Once there is an acquisition in
buffer, you will see a result table and an annotation overlay on the waveform trace showing the decoded data.
lWork with the annotated waveform, result table and other functionality to analyze the decoding.
13
PCIE6bus Decoder Software Instruction Manual
Decoding Workflow
We recommend the following workflow for effective decoding:
1. Set up the decoder using the lowest level decoding mode available, but do not yet enable it.
2. Acquire at least one complete transmission reasonably well centered on screen in both directions, with
generous idle segments on both sides.
Note: See Failure to Decode for more information about the required acquisition settings.
3. Stop acquisition, then enable the decoder. It will operate on the acquisition in buffer.
4. Use the various decoder tools to verify that transitions are being correctly decoded. Tune the decoder
settings as needed to produce a satisfactory decoding.
5. Once you are correctly decoding in one mode, continue making small acquisitions of five to eight
transmissions and run the decoder in higher level modes.
6. Finally, run the decoder on acquisitions of the desired length.
When you are satisfied the decoder is working properly, you can disable/enable the decoder as desired without
having to repeat this tuning process, provided the basic signal characteristics do not change.
Decoder Set Up
Use Decode Setup and its subdialogs to configure decoders.
1. Choose Analysis > Serial Decode from the oscilloscope menu bar.
2. On the Serial Decode dialog, enable the decoder by checking On next to the decoder number. This may be
done any time, although we recommend having an acquisition in buffer before enabling the decoder.
3. Click the Setup button at the end of the row to open the Decode Setup dialog.
4. Select the Protocol to be decoded and the inputs (sources). This selection drives the other fields that appear.
5. Define the level of decoding on the subdialogs to the right of the Decode Setup dialog.
Note: The PCIE6bus decoders are integrated with SDA Expert to provide equalizer and eye diagram
capabilities. Following decoder selection, you set up the actual inputs within SDAX. Decoding is performed
on the equalized output of SDAX.
14
Serial Decode
Inputs and Equalization
Select Input & Equalizer Setup to open SDA Expert and be sure to:
1. Choose the correct Lane number to configure and Serial Standard to decode.
2. Define inputs.
3. Make equalization and emulation/de-embedding settings.
4. Generate an eye diagram to confirm the signal quality is sufficient for decoding.
5. Use the Decoder button at the far right of the SDAX framework dialog to return to the decoder dialogs.
6. On Decode Setup, for each SDAX Lane that is to be used as a decoder source, select the correct SDAEqDig
equalized waveform.
Note: When you open the Source Selector popup, first choose Category SDA and Subcategory
corresponding to the SDAX Lane (L1-L4), then choose the SDAEqDig Source waveform. Note that
you will need to configure a different SDAX Lane for each PCIe 6.0 data lane or bidirectional stream
to be decoded, and for each Lane you will need to input a different SDAEqDig waveform.
PCIe Decoder Subdialog
The subdialog should already show the locked 64 Gbit/s signal Bit Rate.
If the signal utilizes precoding, check Precoded.
15
PCIE6bus Decoder Software Instruction Manual
Correcting Poor Quality or Inverted Signals
It is important to provide a high-quality signal when using a Serial Decode package; this is true for PCI Express
decoders as well as all other types. If bits cannot be interpreted correctly, the decoding will be bad. Follow these
guidelines to ensure good signal quality:
lDo not capture in the middle of the bus (at the PCIE connector) as reflections may seriously degrade the signal.
lEqualize at higher speeds. Signals traversing a significant length of printed circuit board material above 5 GT/s
may warrant some equalization, depending on the quality of the probe and where it was connected. Above
8 GT/s, equalization is most definitely required.
lUse correct polarity signal. An oscilloscope cannot automatically determine if the signal is inverted and
compensate like real receivers do. Therefore, the decoder tries to parse the signal supplied – it has no choice
because the acquisition may begin long after link initialization. Therefore it is important to give the decoder a
signal with correct polarity. If the decode table shows many UNRECOGNIZED and single bytes of IDL (basically
junk), then the signal probably needs to be inverted. Use the Invert checkbox on the input channel dialog.
16
  • Page 1 1
  • Page 2 2
  • Page 3 3
  • Page 4 4
  • Page 5 5
  • Page 6 6
  • Page 7 7
  • Page 8 8
  • Page 9 9
  • Page 10 10
  • Page 11 11
  • Page 12 12
  • Page 13 13
  • Page 14 14
  • Page 15 15
  • Page 16 16
  • Page 17 17
  • Page 18 18
  • Page 19 19
  • Page 20 20
  • Page 21 21
  • Page 22 22
  • Page 23 23
  • Page 24 24
  • Page 25 25
  • Page 26 26
  • Page 27 27
  • Page 28 28
  • Page 29 29
  • Page 30 30
  • Page 31 31
  • Page 32 32
  • Page 33 33
  • Page 34 34
  • Page 35 35
  • Page 36 36

LeCroy PCIE6bus Decoder Software User manual

Category
Measuring, testing & control
Type
User manual

Ask a question and I''ll find the answer in the document

Finding information in a document is now easier with AI