Field notes from electronica 2026, Munich.
Disclosure: I work for UNI-T (UNI-TREND Technology EU GmbH). This demo ran on our own stand. The readings are what the instruments showed during the show, not datasheet claims.
The second station on the UNI-T stand was labelled "Ethernet Compliance: 100/1000BASE-T, Tx Test, Compliance". On the main screen was an eye that does not look like a textbook eye: three rails and two stacked openings. That shape tells you immediately which PHY is being tested, and it is a good starting point for explaining what an Ethernet transmitter test actually checks.
Three levels means 100BASE-TX
100BASE-TX sends 4B5B-coded data at 125 MBd using MLT-3 line coding: the output steps through +1, 0, −1, 0 and moves one step for every logical one. An eye diagram of MLT-3 therefore has three rails and two openings stacked vertically, exactly what was on screen. 1000BASE-T looks different again: PAM-5 on four pairs at once, with five levels. The fixture on the bench supports both PHYs, but the live test during my visit was the 100BASE-TX one.
The compliance software log made the sequence explicit:
- 15:47:21 "Compliance test started"
- 15:47:22 "UTP AOI template test started"
- 15:47:51 "Suspend test" (a demo pause)
The AOI template is the Active Output Interface eye mask from the 100BASE-TX twisted-pair PMD specification. Every transition must stay out of the forbidden regions of the mask. It is the single most visual pass/fail test in the suite.
What a 100BASE-TX transmitter test checks
The mask is only one item. The standard transmitter tests also check, among other things:
- Differential output amplitude, nominally 1 V peak (950–1050 mV), and the symmetry between positive and negative levels.
- Rise and fall times of 3 to 5 ns, plus the symmetry between rise and fall.
- Overshoot, limited to a few percent.
- Transmit jitter, about 1.4 ns peak-to-peak with a scrambled idle pattern.
These limits are generous by modern standards, which is why the oscilloscope's bandwidth is not really the constraint. On this screen, channel 1 used an active differential probe, and the channel badge read 1 GHz and 10×, so the probe sets the channel bandwidth. A 1 GHz system has a 10–90 % rise time of roughly 0.35 ns. Added in quadrature to a 3 ns signal edge, that changes the measured rise time by well under 1 %. The 12-bit vertical resolution matters more here than bandwidth, because amplitude symmetry and overshoot are percentage measurements on a 1 V signal.
The fixture is half of the test
The silkscreen on the fixture lists its sections: differential return loss, load and probe, common mode, a disturbing-signal section with DIST±, SCOPE± and DUT± connectors, and a balun. Each test needs a different electrical environment:
- Load and probe: the DUT drives a defined 100 Ω differential load, and the scope probes across it. Template, amplitude, rise time and overshoot are measured here.
- Return loss: measures how well the transmitter's output impedance matches 100 Ω across frequency. A poor match reflects energy back from the cable and degrades the link even when the eye looks clean.
- Common mode: checks the common-mode output, which matters for emissions.
- Disturbing signal: 1000BASE-T distortion tests inject a defined disturber, because in real operation all four pairs transmit and receive simultaneously.
The practical lesson is that "the eye passes" covers only one of several measurements. A transmitter can pass the template and still fail return loss.
Checklist for an Ethernet Tx compliance run
- Identify the PHY from the eye: three levels for 100BASE-TX, five levels for 1000BASE-T.
- Put the DUT into the right test mode or pattern (scrambled idle or a 1000BASE-T test mode) before measuring.
- Use the fixture section matching the test; do not measure the template on the return-loss path.
- Check that the probe, not the scope, sets the channel bandwidth, and that it is adequate (here more than enough).
- Keep the trace at full vertical scale: amplitude symmetry and overshoot are percentage tests.
- Run the full suite, not just the mask.
Takeaway
The three-rail eye on this station is MLT-3, so the live test was 100BASE-TX: the AOI template first, then amplitude, timing, overshoot and jitter. The oscilloscope's 8 GHz is not what makes this work. The probe sets a 1 GHz channel, which is ample for 3–5 ns edges. What matters is the right fixture section for each test, a DUT in the right test mode, and a full-scale 12-bit capture for the percentage measurements.
Notes: Observations are from the stand on 10 November 2026. Specification values are from the 100BASE-TX twisted-pair PMD requirements as commonly applied; check the current standard for your test plan. The rise-time estimate (0.35/BW) is my calculation. Pass/fail results were not visible in my photo.