Transmitter and Dispersion Eye Closure for PAM4 — the single number that says how good an optical transmitter is.
What TDECQ Measures
TDECQ asks a deliberately practical question: how much noise would a defined reference receiver tolerate with a perfect transmitter, and how much does it tolerate with this one? The ratio of those two noise levels, in decibels, is the TDECQ. A perfect transmitter would score 0 dB; a real one scores the penalty it imposes.
It is therefore a margin measurement rather than a shape measurement. Nothing about the eye is compared against a drawn mask. The test simply establishes how much of the link's noise budget this transmitter has already spent before the signal has entered the fibre.
Why It Replaced the Old Tests
On NRZ a transmitter was judged by an eye mask and an extinction ratio minimum, and those two numbers between them described the signal well enough. PAM4 broke that. Three stacked eyes are not equally affected by the same impairment, the four levels are not evenly spaced because modulators are not linear, and every real receiver runs an equaliser that partly undoes the damage.
TDECQ folds all of it — level nonlinearity, jitter, noise, rise and fall time, and how well the signal responds to equalisation — into one figure, and it does so against the same 4.8e-4 symbol error target that KP4 FEC is dimensioned for. That is why the number is directly meaningful rather than merely indicative.
What a High TDECQ Actually Does to You
TDECQ is a penalty, so it comes straight out of the link budget. A module at 3.3 dB against a 3.4 dB limit passes, but it has handed almost the whole allowance to the transmitter, leaving nothing for a connector that is slightly dirty, a patch panel more than the design assumed, or a receiver at the far end sitting at the low end of its own tolerance.
In the field that shows up as a link that works on the bench and produces a rising pre-FEC BER once it is installed in a real channel, or one that is fine in winter and marginal when the row runs warm. Two modules with identical reach ratings and identical launch power can behave completely differently for this reason alone.
What to Look For in a Test Report
Ask for the measured TDECQ per lane, not a statement that the part is compliant. Compliance means ‘below the limit’; the number tells you by how much, and on a parallel module the worst lane is the one that will fail first. A report showing 1.5 to 2.5 dB across all lanes is a comfortable part; one showing 3.2 to 3.4 dB is a part with no margin left to give.
A supplier who cannot produce per-lane TDECQ for a 400G or 800G module is not testing it properly, whoever's name is on the label. This is one of the concrete technical questions in the OEM versus third-party comparison and in our supplier checklist. Every module in our transceiver range is characterised before shipment.
TDECQ reference numbers
| Full name | Transmitter and Dispersion Eye Closure for PAM4 |
|---|---|
| Units | dB — lower is better, 0 dB would be ideal |
| Reference equaliser | 5-tap feed-forward; P802.3dj drafts propose a longer FFE plus a 1-tap DFE |
| Target symbol error ratio during the test | 4.8e-4 |
| 400GBASE-DR4 and FR4 limit | 3.4 dB |
| 800GBASE-DR8 limit | 3.4 dB |
| LR-class single-mode limit | 3.9 dB |
| PAM4 multimode SR limit | 4.5 dB |
| What it replaced | eye-mask plus extinction-ratio pass/fail from the NRZ era |
Related Terms
PAM4, the extinction ratio it replaced, and the standard lookup for the limit that applies to a given PMD.