High-speed signals oscilloscope

[kbecker]

Hi,
We are currently developing a camera system and want to set up our own lab. The camera data is pouring in with about 480 MBit. While we are not necessarily interested in finding that one odd data packet, we are very concerned about signal quality as the system those operates in are rather noisy, and there are quite a few connectors on a rather lengthly path.
What oscilloscope would you use for that job? (Not necessarily the brand, but rather spec wise.)
My current tendency would be to say that 500MHz/4GS would work (considering that eye-diagrams are quite helpful and the single data point is not that important), or would you say that the use-case justifies an even beefier 1Ghz BW one? From a financial point of few buying a >10k€ oscilloscope would be rather difficult, plus you need even fancier accessories. Any experience that you can share?

[tggzzz]

Hi,
We are currently developing a camera system and want to set up our own lab. The camera data is pouring in with about 480 MBit. While we are not necessarily interested in finding that one odd data packet, we are very concerned about signal quality as the system those operates in are rather noisy, and there are quite a few connectors on a rather lengthly path.
What oscilloscope would you use for that job? (Not necessarily the brand, but rather spec wise.)
My current tendency would be to say that 500MHz/4GS would work (considering that eye-diagrams are quite helpful and the single data point is not that important), or would you say that the use-case justifies an even beefier 1Ghz BW one? From a financial point of few buying a >10k€ oscilloscope would be rather difficult, plus you need even fancier accessories. Any experience that you can share?

You are probably more interested in the MHz spec than the GS/s spec.
500MHz is too low for a 480Mb/s signal; you should be looking at >1GHz if you want to see any detail in the edges.
Be aware of the difficulties of probing such signals. You won't be able to use bog-standard *10 high impedance probes; use active probes or resistive Z0 probes instead.

The reputable established scope manufacturers provide a very great deal of solid information about how to use (and misuse) their scopes and accessories for different purposes. USB is a standard, so somewhere or other they will state which of their products are suitable. If all else fails, ask one of their salesmen.

For pattern-sensitive "noise", you might like to consider using a TDR to look at the signal path. A TDR is irrelevant for impulsive EMI/EMC-type noise.

You can also consider renting test equipment. Last time I looked (many years ago) the rental was 10% of the list price per month.

[w2aew]

The general rule of thumb starting point is typically a minimum BW that is 2.5x the NRZ bit rate. This ensures that the 5th harmonic is no more than 3dB down. So, a 1GHz scope really isn't quite fast enough, and may mask/filter problems in the signal stream.

[kbecker]

This week at the embedded world I took a closer look at those oscilloscope, and we came to the conclusion that buying anything above 500MHz is just insanely expensive and mostly useless to us. (We still have access to an 6GHz 20GS Agilent at a nearby university if it becomes absolutely necessary).

But I found one thing that looks REALLY interesting. Aside from a small desktop oscilloscope for general purpose debugging, I found this:
https://www.picotech.com/oscilloscope/9200/picoscope-9200-sampling-oscilloscopes

It's a 12GHz BW "Oscilloscope" that is rather specific for eye-diagrams and it even contains TDR functionality.
Any thoughts on that one?
The price of 9k€ for the 9211A sounds definitely more reasonable than anything any of the major oscilloscope manufacturer would charge.

[T3sl4co1l]

You're probably going to spend >$2k on a 1GHz+ scope if you need to verify signal integrity of your PCBs, connectors, etc.

Or you can just kind of "suck it and see", which is what most high speed users do...

The reality is, most high speed circuitry has a nice wide noise band, so that signal bounce or rounding is often very tolerable.  Routing isn't very critical (if the impedance is within 30%, it's fine), as long as you don't go out of your way to trash the signals (say by running a diff pair over a switching supply node).

Tim

[rf-loop]

But I found one thing that looks REALLY interesting. Aside from a small desktop oscilloscope for general purpose debugging, I found this:
https://www.picotech.com/oscilloscope/9200/picoscope-9200-sampling-oscilloscopes

It is sampling oscilloscope!
General purpose scope usually need be real time scope.
9200:  Digitising rate    DC to 200 kHz maximum

[AndyC_772]

IMHO when it comes to measuring with a scope, there are two types of signal:

- common signals, which are slow enough to measure with an affordable scope, and
- really fast signals, for which the definition of "fast" changes with each incremental change in technology

If you're dealing with a signal that's 'fast', rent a scope that can measure it when you need to. Buying one for the purpose is expensive, and next year, when you're dealing with a signal that's twice as fast, your expensive scope will no longer be fast enough and you'll need an even more expensive one.

[kbecker]

It is sampling oscilloscope!
General purpose scope usually need be real time scope.
9200:  Digitising rate    DC to 200 kHz maximum

That is why I said aside from a general purpose oscilloscope. That picoScope would be used for eye-diagrams and TDR. As mentioned on the original post, we have a 480MBit signal that does show some signal quality issues. I am mostly curious on whether someone has experience with those picoScopes.

You're probably going to spend >$2k on a 1GHz+ scope if you need to verify signal integrity of your PCBs, connectors, etc.

If only if would be "just" $2k for 1GHz+. 2k is that is what the differential probe costs alone...

[nctnico]

For judging the quality of a signal path you can either use a TDR or look at return loss versus frequency (spectrum analyser + generator + directional coupler). An oscilloscope is not the tool for this job.

[kbecker]

For judging the quality of a signal path you can either use a TDR or look at return loss versus frequency (spectrum analyser + generator + directional coupler). An oscilloscope is not the tool for this job.

That is specifically why I found the PicoScope so interesting. It does have TDR built-in.

[Marco]

You can get an old Tektronix mainframe with TDR sampling plugins for a lot less if you don't mind taking some chances.