This is the result of the eye scan on differential clock:
I have to manually set the threshold to (100mV). The software always warns about invalid threshold setting outside what it thinks a valid range is when I start a capture.
The "EyeScan" tool has two parts:
The anchor point' (time = 0) of these measurements is dependent on the Clock having a fixed threshold.
If the Threshold on a signal is changed to a higher value, then the signal will be seen as crossing later on a rising edge, and earlier on a falling edge. The reverse is true if the threshold is changed to a lower value. At some point, the threshold could be set to a voltage that is outside the range of the signal, and there will be nothing for the EyeScan system to trigger on.
If the threshold is changed on the 'Clock', then that changes the 'anchor point' of all of the other measurements, and is impossible to know by how much it is changing.
For the most accurate measurement, we keep the threshold of the 'Clock' constant, while changing the thresholds of the other signals
If you want to get a full EyeScan on a signal that you are using as the 'Clock', you can double-probe that signal. The second instance of that signal will be treated the same way any other signal is treated, and you will be able to see the full EyeScan.
You had a comment, I have to manually set the threshold to (100mV). The software always warns about invalid threshold setting outside what it thinks a valid range is when I start a capture.
One last note... EyeScan is designed as a tool to help the user pick the best sample point (in time), and threshold for each signal, independently. It provides a qualitative view not a quantitative view. The signal passes through an RC isolation network (which acts as an attenuator and high pass filter), then up several feet of lossy cable (which acts as a low-pass filter) before it gets to the comparator. There can also be some channel-to-channel skew in the LA itself, and some inaccuracies in the delay lines that are used to adjust the relative timing of the signals. You should not be making any voltage measurements (Eye Height, for example), or timing measurements (Eye Width, Jitter, or Skew, for example) on the EyeScan data. It is only for improving the user's ability to make accurate State Mode Measurements. It does that very well, allowing us to make accurate State measurements on DDR Memory systems running at up to 4000 and beyond. In these systems, every signal could be at a slightly different voltage threshold, and the individual bits are at different delays from the clock. Also, the delays between Read and Write cycles will be very different (basically 90 Degree out-of-phase), and the threshold voltages could be different.
Thank you for the detailed response. You've provided a lot of additional insight. Regarding the EyeScan use case here: I wanted to leverage the view of the signal the Eye Scan provided to help me debug a potential issue between the interposer/DUT/and probe. I suspected I was only seeing one of the CLK differential pairs and thought this would come out if I had a better look at what the CLK signal looked like. The EyeScan qualitative view should be sufficient for that purpose.
I'm going to add a second instance of the CLK and run the EyeScan and see if I get the full view.
Thanks again for your detailed response!
Retrieving data ...