PCIe 6.0 Tx Equalization: Why Gen6 Uses Q0–Q10 Presets
If you have worked with PCIe Gen4 or Gen5 transmitter compliance testing, you are probably familiar with the preset list:
P0, P1, P2 … P10
These presets define standardized transmitter equalization settings. They give hardware vendors and test engineers a common way to configure and evaluate signal conditioning without working directly with arbitrary coefficient combinations.
PCIe 6.0 introduces a different naming convention:
Q0, Q1, Q2 … Q10
Changing the prefix from P to Q might look like a minor specification revision. It actually reflects a change in the transmitter equalization model.
PCIe Gen5 uses a three-tap feed-forward equalizer (FFE). PCIe 6.0, operating at 64 GT/s, introduces a four-tap arrangement with an additional second pre-cursor coefficient.
That extra coefficient gives the transmitter another degree of freedom when compensating for channel distortion. The change is particularly relevant because Gen6 replaces NRZ signaling with PAM4, which divides the available voltage range into smaller decision regions.
The result is a more demanding signal-integrity environment. The Q preset family identifies the new transmitter equalization model rather than simply renaming the existing Gen5 settings.
🔧 1. What Does a PCIe Transmitter Preset Actually Control? #
A high-speed electrical signal does not travel through a PCIe channel without distortion.
Between a processor and an NVMe SSD, for example, the signal may pass through:
- PCB transmission lines.
- Multiple PCB layers and vias.
- Connectors and cable assemblies.
- CPU and device package routing.
- PCIe risers or expansion cards.
- Retimers or other signal-conditioning components.
Each element introduces electrical effects that can degrade the transmitted waveform. Insertion loss attenuates signal energy, reflections arise from impedance discontinuities, and frequency-dependent channel behavior changes the shape of the received signal.
One particularly important problem is inter-symbol interference (ISI).
ISI occurs when energy from one transmitted symbol extends into neighboring symbol intervals. The receiver then sees a waveform influenced not only by the current symbol but also by previously transmitted symbols.
A simplified representation:
Transmitted: 1 0 1 1 0
Received: 1~~~ 0~~~ 1~~~ 1~~~ 0~~~
Residual energy from adjacent symbolsThe receiver must distinguish the intended symbol despite these overlapping contributions.
One way to reduce the problem is to modify the signal before it enters the channel. This technique is called transmitter equalization, commonly abbreviated as Tx EQ.
PCIe uses a feed-forward equalizer (FFE) at the transmitter. It adjusts the relative amplitudes of neighboring symbol contributions to compensate for the expected channel response.
Instead of sending an unmodified waveform and leaving all correction to the receiver, the transmitter deliberately shapes the signal so that it arrives in a more useful form.
Presets Are Coefficient Combinations, Not Power Levels #
PCIe defines standardized transmitter presets to make equalization configuration manageable across different implementations.
A preset represents a particular combination of FFE coefficients. Link-training procedures and compliance tests can refer to these predefined configurations instead of requiring every device to negotiate arbitrary values.
There is an important distinction here: a higher preset number does not necessarily mean stronger equalization or better signal quality.
For example, P10 is not ten times stronger than P0. Each preset represents a different operating point within the electrical limits established by the specification.
The best setting depends on the actual channel.
A preset that produces good results on one motherboard layout may perform poorly on another because the insertion loss, reflections, and frequency response are different.
⚙️ 2. PCIe Gen5 Uses a Three-Tap Transmitter FFE #
PCIe Gen3 through Gen5 use a three-tap transmitter equalization structure.
The three coefficients are:
- Pre-cursor (
C-1): Controls the contribution associated with the preceding unit interval. - Main cursor (
C0): Controls the primary contribution associated with the current symbol. - Post-cursor (
C+1): Controls the contribution associated with the following unit interval.
The arrangement looks like this:
Pre-cursor Main cursor Post-cursor
| | |
v v v
C-1 C0 C+1
Previous UI Current UI Following UIThese coefficients determine how the transmitter combines neighboring symbol contributions to produce its output waveform.
For PCIe Gen5 at 32 GT/s, this three-tap FFE forms the basis of the established P0 through P10 preset family.
It is tempting to think of the three coefficients as independent controls for sharpening different parts of a signal. That is a useful starting point, but the actual implementation has additional constraints.
The coefficients must satisfy the specification’s electrical requirements, including restrictions related to transmitter output swing and equalization behavior.
Consequently, the P presets are not arbitrary combinations of three numbers. They are standardized operating points designed to keep transmitter behavior within defined limits.
🚀 3. PCIe 6.0 Reaches 64 GT/s Through PAM4 #
PCIe Gen5 operates at 32 GT/s using NRZ signaling. PCIe 6.0 doubles the raw transfer rate to 64 GT/s per lane.
However, Gen6 does not simply double the symbol rate.
The key change is the transition from NRZ to PAM4 (Pulse Amplitude Modulation with four levels).
NRZ Versus PAM4 #
NRZ signaling uses two voltage states:
High = 1
Low = 0Each symbol represents one bit.
PAM4 uses four distinct voltage levels:
Level 3
Level 2
Level 1
Level 0Each symbol can represent one of four possible states, encoding two bits instead of one.
| Characteristic | NRZ | PAM4 |
|---|---|---|
| Voltage levels | 2 | 4 |
| Bits per symbol | 1 | 2 |
| Symbol rate for 64 GT/s | 64 GBd | 32 GBd |
| Signal encoding | Binary | Four-level |
This is how PCIe 6.0 achieves 64 GT/s while operating at 32 GBaud.
Because each PAM4 symbol carries two bits, the transfer rate doubles without requiring a corresponding increase in symbol rate.
There is another interesting consequence.
PCIe Gen5 at 32 GT/s NRZ has a nominal Nyquist frequency of approximately 16 GHz. PCIe Gen6 at 64 GT/s PAM4 also has a nominal Nyquist frequency of approximately 16 GHz because its symbol rate is 32 GBaud.
This helps explain why Gen6 does not simply require twice the channel bandwidth.
However, similar Nyquist frequencies do not make the two electrical interfaces equivalent. PAM4 creates a different signal-integrity problem because its voltage levels are much closer together.
📉 4. PAM4 Creates Three Eyes and Reduces Voltage Margin #
An NRZ signal has two voltage levels and one eye opening between them.
PAM4 has four voltage levels, producing three separate eye openings.
Conceptually:
PAM4 eye diagram
Level 3 ─────────────────────
Upper eye
Level 2 ─────────────────────
Middle eye
Level 1 ─────────────────────
Lower eye
Level 0 ─────────────────────The transmitter does not increase its total voltage swing fourfold to accommodate the additional levels.
Instead, the available voltage range is divided into smaller steps.
That leaves less voltage margin for each receiver decision.
Noise, crosstalk, jitter, insertion loss, and residual ISI can more easily push a received symbol across an incorrect decision threshold.
This is the central trade-off behind PAM4: it carries twice as many bits per symbol, but the receiver has a much smaller voltage margin when distinguishing between the possible states.
PCIe 6.0 Uses FEC and FLIT Mode to Manage Errors #
PCIe 6.0 addresses the higher raw error rate associated with PAM4 through several protocol-level mechanisms:
- FEC (Forward Error Correction): Corrects a defined number of transmission errors without requiring every corrupted transmission to trigger a retransmission.
- CRC (Cyclic Redundancy Check): Detects corrupted data.
- FLIT Mode: Organizes traffic into fixed-size flow-control units and provides the framing structure used by Gen6’s reliability mechanisms.
The raw bit error rate target for PCIe 6.0 PAM4 is commonly specified around \(10^{-6}\), compared with the \(10^{-12}\) target associated with earlier NRZ generations.
FEC and the associated reliability mechanisms help reduce the effective error rate seen by higher protocol layers.
However, error correction does not eliminate the need for good electrical signaling. A link that generates too many errors can exceed the correction capability of its FEC mechanism.
That makes transmitter waveform quality particularly important.
When the receiver has less voltage margin, the transmitter needs to control more of the distortion introduced by the channel.
🔬 5. PCIe 6.0 Adds a Fourth Tap: The Second Pre-Cursor #
This is where the Q presets become relevant.
PCIe Gen5 uses three transmitter coefficients:
C-1 C0 C+1
| | |
v v v
Pre-cursor Main cursor Post-cursorPCIe 6.0 expands the transmitter FFE to four taps at 64 GT/s:
C-2 C-1 C0 C+1
| | | |
v v v v
Second pre- First pre- Main cursor Post-cursor
cursor cursorThe additional coefficient is C-2, known as the second pre-cursor.
The four-tap arrangement gives the transmitter control over four consecutive symbol contributions.
| Coefficient | Function |
|---|---|
C-2 |
Second pre-cursor |
C-1 |
First pre-cursor |
C0 |
Main cursor |
C+1 |
Post-cursor |
The second pre-cursor provides another degree of freedom for compensating channel effects that extend beyond the immediately adjacent symbol interval.
This is more than an increase in output amplitude. The transmitter can now use an additional coefficient when shaping its output waveform.
The new structure is reflected in the PCIe Gen6 preset definitions.
Instead of the established P0 through P10 naming convention, Gen6 introduces 11 presets labeled:
Q0, Q1, Q2 … Q10
The different prefix helps distinguish the Gen6 equalization family from the earlier three-tap implementation.
For engineers working with actual waveforms, the distinction matters because the coefficient model used to reconstruct and evaluate the transmitter response has changed.
🧪 6. Why the Fourth Tap Matters on Real PCIe Channels #
A PCIe channel is not an ideal transmission line.
Consider a connection between a CPU and an enterprise NVMe SSD. The signal may travel through package routing, motherboard traces, multiple vias, connectors, and the SSD’s own PCB.
Every transition contributes to the overall channel response.
The resulting distortion is not necessarily limited to the immediately preceding or following symbol. A transmitted pulse can leave residual energy extending across several unit intervals.
This behavior is described by the channel’s impulse response.
A useful analogy is dropping a stone into water. The initial impact creates ripples, but the resulting wave pattern continues beyond the point where the stone entered.
A high-speed electrical channel behaves differently in physical terms, but the general idea is similar: the response to one transmitted event can persist after the event itself has ended.
More Taps Provide More Control Over the Waveform #
A three-tap FFE can compensate for the most immediate neighboring contributions.
Under PAM4 signaling, however, residual distortion can consume a larger proportion of the available eye margin.
The fourth coefficient, C-2, gives the transmitter another way to shape the waveform before it enters the channel.
This additional control can help reduce the impact of channel memory extending across multiple symbol intervals.
It does not mean that four taps automatically eliminate ISI. The result still depends on several factors:
- Channel insertion loss and frequency response.
- Impedance discontinuities and reflections.
- Transmitter output characteristics.
- Receiver equalization capability.
- PCB routing and connector quality.
- The actual coefficient values selected during link configuration.
The additional tap also does not replace receiver equalization or proper channel design.
It expands the transmitter’s ability to compensate for distortion. It does not turn a poor electrical channel into a good one.
For platform engineers, this distinction matters. Gen6 compliance still requires careful channel budgeting, PCB simulation, signal-integrity analysis, and validation across the supported operating conditions.
📊 7. What Changes in a PCIe 6.0 Compliance Test? #
For engineers working with oscilloscopes, BERTs, and PCIe compliance equipment, the transition from P presets to Q presets affects more than the labels displayed in a test interface.
A Gen5 test environment works with:
- The
P0–P10preset family. - The three-tap transmitter equalization model.
- NRZ waveform analysis.
- Gen5-specific electrical limits and compliance patterns.
A Gen6 test environment must account for:
- The
Q0–Q10preset family. - The four-tap transmitter equalization model.
- The additional second pre-cursor coefficient.
- PAM4 waveform analysis.
- Gen6-specific electrical limits and compliance patterns.
The measurement process also changes because PAM4 has four voltage levels instead of two.
An oscilloscope must capture and analyze the relevant PAM4 waveform characteristics. The transmitter’s equalization response must then be evaluated against the applicable Gen6 requirements.
Measuring the Four-Tap Response #
One approach described in Teledyne LeCroy’s measurement documentation uses both equalized and non-equalized compliance patterns.
The general workflow is:
- Configure the device under test to transmit the required compliance patterns.
- Capture the transmitted waveforms using an appropriate oscilloscope.
- Process the captured data to reconstruct the equalized response.
- Apply candidate equalization coefficients to determine the best-fit response.
- Compare the measured and reconstructed responses against the required limits.
The objective is to determine whether the transmitter’s actual behavior matches the expected equalization characteristics.
This is why changing a test script from P to Q is not sufficient.
The measurement algorithms, waveform interpretation, compliance patterns, and coefficient limits must all match the signaling generation being tested.
The broader hardware transition is already appearing in next-generation server and storage platforms. For example, KAD’s coverage of PCIe 6.0 and CXL 3.1 controller development examines how the higher interface bandwidth is being incorporated into enterprise storage designs.
As PCIe 6.0 controllers move toward wider deployment, validating the electrical interface becomes just as important as meeting the advertised bandwidth target.
P Presets Versus Q Presets #
The distinction can be summarized as follows:
| Characteristic | PCIe Gen5 | PCIe 6.0 |
|---|---|---|
| Signaling rate | 32 GT/s | 64 GT/s |
| Modulation | NRZ | PAM4 |
| Symbol rate | 32 GBd | 32 GBd |
| Tx EQ structure | Three-tap FFE | Four-tap FFE |
| Pre-cursor coefficients | C-1 |
C-2, C-1 |
| Main cursor | C0 |
C0 |
| Post-cursor | C+1 |
C+1 |
| Preset family | P0–P10 |
Q0–Q10 |
The letter change is a useful indicator when reading compliance reports, configuring test equipment, or reviewing transmitter equalization behavior.
P0 through P10 identify the established Gen5 preset family and its three-tap transmitter equalization model.
Q0 through Q10 identify the Gen6 preset family, incorporating the additional second pre-cursor coefficient in the four-tap FFE structure.
PCIe 6.0 did not just increase the transfer rate. Its move to PAM4 changed the electrical constraints under which the interface operates.
Gen5’s three-tap FFE was designed around NRZ signaling at 32 GT/s. Gen6’s four-tap arrangement provides another coefficient for controlling inter-symbol interference while operating at 64 GT/s.
That is the real reason behind the transition from P presets to Q presets: a different transmitter equalization model, built to address the tighter signal-integrity margins of PCIe 6.0.