News for hackers. Stuff that matters.
Re:I stopped trusting USB-C cable labels and started testing... by ChrisMarshallNY on Friday August 07, 2026 @06:42AM (#49208397)
Time sampling isn’t enough.

You need to at least look at waveform integrity (which can be approximated, by doing things like looking for harmonics).

For example, you may send in a 40 GHz square wave, and get back a wave full of ringing, or a sine or sawtooth wave. It’s damn difficult to maintain clean square waves, at that frequency. The cable could also double as a microwave antenna.

Since it’s digital information, some waveform alterations are OK, but not too much. You need fairly clear state transitions. Ringing, or waveform distortion, can introduce extra (or fewer) transitions; thus, completely changing the data.

Re:I stopped trusting USB-C cable labels and started testing... by CamperBob2 on Friday August 07, 2026 @12:07PM (#49212589)
That's not really what I mean. You don't have to diagnose the signal, just the medium, which can tell you what will happen to the real signal. It requires a fast edge, but nothing much beyond that.

E.g., here's what a bad connector at the far end of an SFF-8654 cable looks like on a TDR: https://i.imgur.com/AiLnSuC.jpeg This type of plot can be interpreted as impedance versus distance along the line. It can be made with a small handheld instrument, but currently a human is needed to read the tea leaves. While I was making that measurement, I couldn't help but think that it would be nice if it didn't require either a 50 GHz VNA or equally-expensive training to read.

Re:I stopped trusting USB-C cable labels and started testing... by ChrisMarshallNY on Friday August 07, 2026 @01:44PM (#49213840)
Fair 'nuff.

As long as we don't get false positives. They are worse than false negatives, and is what happens, when you just read the markers from the chip. As we know well, lots of dodgy C.M.O.T. Dibbler-types will deliberately program the chips to emit bogus markers.