Crosstalk

Why use the Centellax PCB12500 for Crosstalk testing?

The Centellax PCB12500 Parallel Channel BERT has a number of unique features:

• Each PCB12500 can support up to 5 remote heads
• TG5P1A Pattern Generator heads have built-in de-emphasis
• TR2P1A Error Detector heads are highly sensitive
• All heads have fully independent AnyUI delay shift capability (-1000 to +1000UI in minimum 0.001UI steps)
• Heads can generate asynchronous data with the DelaySweep feature 

What's so special about DelaySweep?

The Centellax PCB12500 Parallel Channel BERT has a unique DelaySweep feature that can automatically sweep the clock delay of any or all channels at independent frequencies.  Sweeping is a triangular delay modulation with a range of 1UI, 2UI, or 4UI centered on transition edges.  Each channel sweeps at a unique rate, and all sweep rates are primes, so every combination of aggressor data edge is presented to the victim channel, ensuring that the 'worst case' scenario is tested.

Aggressor channel transitions induce a voltage waveform on the victim channel.	The induced voltage closes the victim channel eye, but not always at the central sampling point.
Multiple aggressors further impair the victim channel. Each aggressor impacts the victim channel differently.	Sweeping multiple aggressors guarantees the worst-case impairment of the victim channel because the effect is spread across the eye.

The benefit of DelaySweep is detailed further in Centellax AN24: Characterizing Crosstalk Effects in High-Density High-Speed Backplanes.

Why should I care about worst-case impairment of the victim channel?

Testing the worst-case phase relationship between aggressor channels and the victim channel ensures you’re measuring the crosstalk impairment that your DUT will be exposed to out in the wild. Measuring crosstalk with a fixed phase relationship between aggressor and victim doesn’t measure crosstalk!

Here’s an example of a bathtub curve measurement made on a qSFP-connectorized active optical cable with three near-end aggressors and four far-end aggressors. All patterns are PRBS31, all amplitudes are 1.2V-d. 30-second BER measurements give 1E-11 BER depth at 95% confidence interval.

With fixed phase relationship between aggressors and the victim channel, the eye looks wide open. No crosstalk impairment is detected. Everything seems fine.

This is the same measurement made with the Centellax PCB12500 DelaySweep feature turned on for all seven aggressor channels. DelaySweep automatically slews the phase alignment of the aggressor channels at a unique rate, approximating results that you would otherwise need eight fully-asynchronous clocks to measure.

\

This measurement shows that crosstalk impairment has created a BER floor at 1E-11 BER. This means a handful of errors in a 30-second measurement.