Transmit de-emphasis shapes a digital waveform before it travels through a lossy channel. In the simple NRZ case, the first bit after a transition has a larger magnitude than subsequent bits of the same value. Reducing the repeated-bit level makes transition content stronger relative to the slower-changing parts of the signal.
This relative shaping helps compensate a channel that attenuates high-frequency content. Read the waveform convention, measurement point and receiver settings together. More de-emphasis changes both the balance of frequencies and the available amplitude; the useful setting comes from evaluating the complete link.
Follow a short run of bits
Imagine the NRZ sequence 0, 1, 1, 1, 0, 0. Each symbol carries one of two logical values. The transition from 0 to 1 starts a run: the first 1 uses the transition level, and the following 1s use a reduced level. At the next transition, the first 0 uses the opposite-polarity transition level, followed by a reduced-magnitude 0.
TI’s SerDes Implementation Guide, section 9.3.6 describes repeated-bit amplitude relative to a bit preceded by a transition. This is a useful time-domain explanation: de-emphasis does not mean simply weakening the high-frequency content already attenuated by the channel.
On a narrow screen, scroll the table horizontally. Keyboard users can focus the table region and use the arrow keys.
| Bit position | Logical value | Differential level |
|---|---|---|
| 1, after a preceding 1 | 0 | −400 mV: transition |
| 2 | 1 | +400 mV: transition |
| 3 | 1 | +283 mV: repeated |
| 4 | 1 | +283 mV: repeated |
| 5 | 0 | −400 mV: transition |
| 6 | 0 | −283 mV: repeated |
These are fictional ideal levels referenced to zero differential voltage, with instantaneous edges omitted. Real transmitters use finite rise/fall times and may implement multiple filter taps, preshoot or protocol-defined presets. Those details need the actual device documentation.
Calculate the ratio with an explicit convention
Keep single-ended, differential, peak and peak-to-peak quantities consistent. In the example, +400 mV and −400 mV transition levels span 800 mV differential peak-to-peak. Mixing that span with a one-sided 283 mV magnitude gives a meaningless ratio.
Pre-emphasis often describes increasing selected transition content, while de-emphasis describes lowering repeated-bit content relative to a transition. The terms can overlap in transmitter documentation. TI’s TUSB501-Q1 data sheet, Figure 7 provides a device-specific waveform example; use its definitions only for that device and interface.
Keep the transmitter, channel and receiver together
Loss and dispersion spread the influence of one symbol into its neighbors, producing intersymbol interference. The result depends on the transmitted pattern as well as the board traces, connectors and cable. Transmit shaping anticipates part of that behavior; receiver equalization addresses the signal after it arrives.
Changing an output setting while leaving receiver equalization on automatic can change two parts of the experiment at once. Record whether equalization is fixed, trained or adaptive and whether a retimer or redriver sits in the path. Record firmware, lane, data rate and the full channel arrangement.
Teledyne LeCroy’s eye-diagram examples show improved receiver eye margins alongside reduced amplitude in specific test cases. Treat those as demonstrations of a tradeoff. They do not establish that the strongest available setting is best for every channel.
Compare settings with a controlled experiment
- Define the interface generation, data rate, lane and required test method. Use the documented reference planes and allowed transmitter settings.
- Save a baseline with the transmitter setting, receiver state, test pattern and fixture arrangement recorded.
- Change one permitted setting at a time. Allow any required link retraining or settling, then repeat the same acquisition.
- Evaluate the received signal and error behavior using the prescribed method. Include eye height/width, jitter or mask results only with their measurement definitions.
- Repeat promising settings across representative channel and operating conditions. Keep the observations that made you reject other settings.
A transmitter eye measured before the channel can look different from a receiver eye. Probes, fixtures and any mathematical de-embedding also affect the result. LeCroy’s channel-emulation note explains how fixture removal, channel models and receiver equalization represent different operations. Keep those models and settings with the saved waveform.
Interpret a clean run with its limits
An attractive eye diagram is one piece of evidence. Error testing needs the tested bit count, pattern, duration and operating conditions; zero observed errors in a short run leaves substantial uncertainty about rare failures. State the relevant standard’s acceptance criterion rather than inventing a universal test duration.
If every setting performs poorly, inspect continuity, impedance discontinuities, connector compatibility, reference clocks and channel loss before adding more correction. If only one lane fails, compare its physical path and settings with a working lane under the same conditions.
Use the worksheet to preserve a reproducible setting comparison. It includes the ratio calculation, reference plane and receiver state, which are the details most easily lost in a screenshot.
Keep a working record
Download the serial-link de-emphasis worksheet (editable text). Save a copy for each comparison or test. It includes the example assumptions, fields for source references and space for your results.