Eye-diagramm oddness

[TheUnnamedNewbie]

Hi all. I didn't quite know if I should put this here or in the test-equipment section.

I was playing around with connecting my gear to my pc for control. I thought a good first experiment would be to generate a PRBS with matlab, send it to my arb, and then display the eye-diagram on my scope.

When I did, all was well untill I went to a bit higher in frequency (For my gear). When I go past a clock of 10MHz, I start to notice the eye-diagram splits up into two eye diagrams with a slight offset (see picture attached). I am puzzled what would cause this. I would expect that the clock jitter of both instruments becomes more visible as we go up in frequency. However, when this is the issue I would just expect the eye to close. What I have here is almost like 2 clean eyes overlapping - like there is a binaty jitter, or some clock-jumping going on. What is up with this? Is there any obvious reason for this to happen that I am not understanding (like the way the clock or trigger works?)

This is the first time I'm measuring an eye myself, so I could be making some stupid mistakes.

[toli]

looks like its about 4ns time shift. I'd guess you have an arb running at 250MSPS, and this is the inherent jitter you have there. There are ways to improve this (like they do in most arb's nowadays), or by selecting the appropriate frequency to run the signal.
try reading a bit more about it, its been talked over the forum more than once.

[TheUnnamedNewbie]

Interesting, I wasn't aware of this. I think the issue lies in the control. Am I correct in assuming that were I to tell the arb to use sinx/x interpolation it would fix the issue?

[toli]

this is something very basic. the device can only operate in 4ns increments (250MSPS). if your signal doesn't split into 4ns increments, it'll be be modified either too early or too late. as this accumulates the signal will jump between being high for N cycles and being high for N+1 cycles (and similarly for the low value). that's exactly what you are seeing.

you can read a lot about it here on the forum. also, each of the manufacturers has its own brochure for the newer line of arbs that try to deal with this problem.
just look on google for the sdg2082x (or similar model) from siglent, or the rigol z series (they call it sifi), or keysights truearb, etc.

[tggzzz]

"Playing around with gear" to convince yourself that you understand what's going on is always useful, and a characteristic of a good engineer.

The ARB 4ns point made by other posters is obviously valid, but I still don't understand your display.

If you are triggering from the input (i.e. not a separate clock), I would like to see all input traces coincide at the triggering point and triggering voltage, and diverge away from that point.

When looking at points (pun intended) like this, I prefer to turn off any display interpolation in the scope, and just show the sampled points. That lets you see what the scope is seeing, and lets you use your mark 1 eyeball do the interpolation. That can reveal some traps for the unwary!

[AndyC_772]

The trigger point is off the screen to the left, so the point at which the traces exactly overlay is out of view.

As expected, we're seeing the fact that each edge is nominally (4ns * N), or (4ns * (N+1)) away from the trigger point, for some integer value of N.

[tggzzz]

The trigger point is off the screen to the left, so the point at which the traces exactly overlay is out of view.

As expected, we're seeing the fact that each edge is nominally (4ns * N), or (4ns * (N+1)) away from the trigger point, for some integer value of N.

OK.

Personally I always like to see the trigger point, since that is a critical part of the displayed waveforms. Obviously that doesn't apply if the points of interest are far from the trigger point, but I don't believe that's the case here.

In this case, seeing the trigger point at the LHS would probably have indicated that there were two different periods to the next zero crossing, which is always A Big Clue. Doubly so if there if the waveform doesn't diverge further afterwards.

[Berni]

Yep that is DDS signal generator jitter.

Just stick to multiples of your signal generators sample rate. So for 250MS/s go for things like 10MHz 12.5MHz 25MHz 50MHz etc.

The better signal generators have a fractional PLL creating the DAC clock. That way the signal generator can run the DAC at a sample rate that is exactly a multiple of the frequency you are asking for so that the data points in your arb waveform precisely line up with the sample points going out.

[coppice]

Yep that is DDS signal generator jitter.

A DDS doesn't jitter like this, but this is not a signal being generated by a DDS. Its a simplistic repetitive replay of a stored pattern, which will jitter like this.

[cdev]

I am just curious whether the large chunk of hardware panel real estate at the bottom of the picture that they have dedicated to a marketing slogan comes standard or is an extra?

My old Tek 2211 doesn't have anything like that.

Hi all. I didn't quite know if I should put this here or in the test-equipment section.

I was playing around with connecting my gear to my pc for control. I thought a good first experiment would be to generate a PRBS with matlab, send it to my arb, and then display the eye-diagram on my scope.