Introduction

PCI Express® 6.0 features significant changes from PCIe® 5.0. In particular, PCIe 6.0 achieves its 64-GT/s data rate, double that of PCIe 5.0, by moving from non-return-to-zero (NRZ) signaling to four-level pulse-amplitude-modulation (PAM4) signaling. Consequently, PCIe 6.0 requires some new test methodologies and patterns, including a new PAM4 Compliance Pattern that finds use in multiple measurements.

The Compliance Pattern Signal

The new Compliance Pattern is used for calculating signal to noise and distortion ratio (SNDR), as well as ps21TX(the package insertion loss) and the transmitter ratio of level mismatch (RLM). In addition, it is used to measure transmitter equalization coefficients.

The new Compliance Pattern signal contains low-frequency patterns, such as strings of 64 0s, 1s, 2s and 3s, enabling a long run of each of the voltage levels found in PAM4. It also contains a big section of pseudorandom binary sequence (PRBS) data using PAM4. Shown in Figure 1:

  • Section 1 contains a long run of 2s and 1s.
  • Section 2 is the big section containing PRBS data.
  • Section 3 contains a long run of 3s and 0s followed by a clock pattern.
  • Section 4 contains another long run of 3s and 0s.

More PRBS data follows section 4, and the Compliance Pattern signal repeats itself. One or more of these sections in the Compliance Pattern signal can assist in the calculation of SNDR and other parameters.

Figure 1. New PCIe 6.0 Compliance Pattern divided into sections.

SNDR Measurement

SNDR refers to the Signal to Noise and Distortion Ratio. Figure 2 shows the breakdown of the SDNR equation in terms of the variables Pmax (the signal strength), sn(noise) and se(distortion).

Figure 2. SNDR equation variables.

The Pmax component of SNDR is calculated by using a pulse response, which is an extracted linear model of the transmitter signal. The pulse response is measured at the transmitter to describe the behavior of the transmitter as a filter if you apply the filter to a 1-UI pulse. Pmax is the maximum amplitude of the extracted pulse response (Figure 3), and it can be affected by various factors such as the channel bandwidth.

Figure 3. Pmax measurement.

The noise component of the signal, notated sn, is measured on the 61st UI of the 64-UI-long runs of 0s, 1s, 2s and 3s in the Compliance Pattern signal (Figure 4). The 61st UI is chosen for this measurement because, in theory, the signal has had plenty of time to settle down to its equilibrium state by the time the 61st UI arrives.

Figure 4. Noise variable snmeasurement.

The last component of the SNDR equation is se, which is the distortion variable. It is a measure of how much the average signal deviates from the extracted pulse-response shape and is measured on the PRBS section of the compliance-pattern signal. In Figure 5, you can see three traces in the top grid. The red trace is the ideal pattern; the blue trace, which is overlapping the red trace, represents the average of the captured pattern; and the green trace is the difference between the average and ideal patterns, which is equal to the distortion se.

Figure 5. Distortion variable semeasurement.

Combining all three components returns the SNDR measurement, but the basic SNDR equation of Figure 2 is missing an important variable—the oscilloscope noise (sscope). The oscilloscope noise impacts the SNDR of a real-life device, and it must be considered if you want to have an accurate SNDR measurement for your device. Figure 6 shows the equation for SNDR with the oscilloscope noise removed (SNDRNR).

Figure 6. SNDRNRequation with measurement results obtained using a Teledyne LeCroy oscilloscope.

Teledyne LeCroy's PCIe 6.0 Base Transmitter compliance test solution implements several noise-removal methodologies. The bottom of Figure 6 shows two measurements that were made on a Teledyne LeCroy LabMaster oscilloscope. The first is the SNDR without oscilloscope noise removed, and the second is the SNDRNRmeasurement with the oscilloscope noise removed.

RLM Measurement

Transmitter linearity is defined as a function of the mean signal levels V0, V1, V2and V3transmitted for PAM4 two-bit symbols, as shown in Figure 7.

Figure 7. V0, V1, V2and V3signal levels for calculating the transmitter RLM.

The transmitter RLM is defined by the following equations involving V0, V1, V2and V3:

Vmid= (V0+ V3) / 2

ES1= (V1– Vmid) / (V0– Vmid)

ES2= (V2– Vmid) / (V3– Vmid)

RLM = min((3 x ES1), (3 x ES2), (2 – (3 x ES1)), (2 – (3 x ES2)))

RLM is measured using the same section of the Compliance Pattern signal used for snfor the SNDR measurement, which is the 64-UI-long runs of 0s, 1s, 2s and 3s. The RLM measurement is straightforward. The goal is to have evenly spaced transition levels and to have RLM equal to 1.

ps21TXMeasurement

A key voltage measurement to make for PCIe 6.0 Base transmitter testing is the ps21TXmeasurement—the effective transmitter package loss. This measurement was also defined in the PCIe 5.0 Base specifications, so it's not a new measurement for PCIe 6.0. The package loss is measured by comparing the 64 0s and 64 1s voltage swing against a 1, 0, 1, 0 clock pattern. The ps21TXmeasurement is conducted with no transmitter equalization, and it is made by averaging over 500 repetitions of the Compliance Pattern signal to reduce noise. The measurement is illustrated in Figure 8, where V111is the amplitude of a low-frequency signal and V101is the amplitude of a high-frequency signal. The ps21TXmeasurementis then defined as follows:

Figure 8. Measurements for determining ps21TX.

Using Your Oscilloscope to Make Compliance Pattern Measurements

With the installation of the SDAIII-PAMx option for SDAIII, the oscilloscope is equipped to calculate SNDR and RLM for custom PAM4 signals, and can plot the ideal Compliance Pattern (red Linear Fit), the average of the measured pattern (blue Pattern Avg), and the error difference between the average and ideal patterns (pink EFit), which is equal to the distortion se. However, these measurements do not include any oscilloscope noise removal methods, as is required for PCIe 6.0.

Figure 9. SDAIII Compliance Pattern measurements for custom PAM4 signals.

The further addition of the SDAIII-PCIE6 option for SDAIII calculates all the Compliance Pattern measurements at the limits defined by the PCIe 6.0 Base Specification. SNDR and its component measurements are calculated using the selected oscilloscope noise removal method (SNDRnr) or no noise removal. You may optionally apply a Bessel-Thomson Filter to the input signal or export the measurement data to a .csv file.

A full set of Base Specification Jitter Pattern, High Swing Pattern and Voltage measurements are also enabled by the SDAIII-PCIE6 option.

Figure 10. SDAIII Compliance Pattern measurements for PCIe 6.0 signals include noise removal options.

Three oscilloscope noise removal methods are available.

Manual Method

Manual uses the specified amount of oscilloscope noise for the variable in the SNDRnr formula. This method is useful if you have previously measured your oscilloscope intrinsic noise and know what value to enter.

Baseline Method

Baseline saves a reference measurement of the input terminated into 50 Ω, which is used to determine the oscilloscope's intrinsic noise floor. The calculated oscilloscope noise is then computed into the SNDRnr measurement results. To use this method, you must first:

1. Set the DUT to output the Compliance Pattern signal using preset Q0 with no emphasis.

2. Set the oscilloscope to acquire using:

  • DBI for 65 GHz bandwidth
  • 50% Edge trigger on C2B

3. Set the V/div and Offset appropriate for the signal. The signal should be reasonably centered vertically and occupying 6-7 vertical divisions of the grid.

4. Make a Single acquisition of the Q0 differential signal on C2B and C3B.

5. On the PCIe 6.0 Measure dialog, choose Noise Removal Method Baseline.

6. Without changing any settings, disconnect the input signal, terminate the oscilloscope inputs using 50 Ω terminators, and make a Single acquisition of the baseline noise signal.

7. Click Save Baseline Ref. The Noise Removal "LED" indicator should turn from red to green.

8.Without changing any settings, reconnect the DUT signal to the oscilloscope inputs and proceed to measure the SNDR. The “SNDRnr” parameter will display the SNDR with the oscilloscope’s noise compensated as per the baseline.

Note:The Baseline reference is only valid for the V/div and Offset settings at which it was acquired. If you change these settings, you will need to save a new Baseline reference. The Noise Removal LED will turn yellow to warn you if the reference is invalid due to age, changes to the setup, oscilloscope temperature, etc. In that case, repeat this procedure to save a new reference.

Attenuator Method

The Attenuator method is the most accurate noise compensation method, since a signal is being applied to the oscilloscope for both reference and measurement. It compares the SNR of a full-scale signal to the SNR of an attenuated signal (at the same full-scale input range) to calculate the oscilloscope's noise contribution to the SNDR as per the formula:

σFS2= σscope2+ σsignal2

σAtt2= σscope2+ (σsignal2/ K2)

σscope2= (K2σAtt2– σFS2/ K2– 1)

where K is the attenuation value.

This value is computed into the SNDRnr measurement results. To use the Attenuator method, you must first:

1. Set the DUT to output the Compliance Pattern signal using preset Q0 with no emphasis.

2. Set the oscilloscope to acquire using:

  • DBI for 65 GHz bandwidth
  • 50% Edge trigger on C2B

3. Set the V/div and Offset appropriate for the signal. The signal should be reasonably centered vertically and occupying 6-7 vertical divisions of the grid.

4. On the PCIe 6.0 Measure dialog, choose Noise Removal Method Attenuator.

5. Make a Single acquisition of the Q0 differential signal on C2B and C3B.

6. Disconnect the input cables from the oscilloscope.

7. Connect a pair or 6 dB or 10 dB attenuators to the oscilloscope inputs, then reconnect the input cables from the DUT to the attenuator, so that the attenuators sit between the oscilloscope and cables.

8.Without changing any settings, make a Single acquisition of the attenuated Q0 differential signal on C2B and C3B.

9. Click Save as Attenuator Reference. The Noise Removal "LED" indicator should turn yellow.

10. Check the box Apply Reference. If successful, the LED will turn green with the internal name of the reference beneath it.

11.Without changing any settings, disconnect the attenuators from the oscilloscope inputs, reconnect the cables directly to the oscilloscope and proceed to measure SNDR.

The "SNDRnr" parameter will display the SNDR with the oscilloscope's noise compensated as per the equation above.

Note:As with the Baseline reference, the Attenuator reference is only valid for the V/div and Offset settings at which it was acquired. The Noise Removal LED will turn yellow to warn you if the reference is invalid due to age, changes to the setup, oscilloscope temperature, etc. In that case, repeat this procedure to save a new reference.

Changing the Reference

Reference waveforms must be saved on the same day you make measurements, using the exact same equipment and settings. When you change any part of your setup, or start a new test session, choose Delete Reference and repeat the procedure to save a reference before proceeding with measurements.