Taking Nyquist into account when using an oscilloscope

[DutchGert]

I feel realy n00b for asking this but recently we started developing products with high-speed serialisers and deserialisers in them (f.e: MAX9280 http://www.maximintegrated.com/datasheet/index.mvp/id/8067)

Since we want to do design verification on them we need a new oscilloscope to measure the serialised data which is transmitted at ~3,2GHz. Question is, can we use a 4GHz scope for the job (with the right probes, technique of course) or do you need to take into account the Nyquist frequency and do u need a scope of minimal 2 times the bandwith of the signal? In this case 6.5-8GHz?

In short, question is, can u measure a signal with upto the same bandwith as the specification of your oscilloscoop, has the manufacturer taking this into account?

Disclamer: until no we had a 1.5GHz scope with probes and only needed to do op to 1Gbit Ethernet verification which could be done perfectly so this I never had to consider this problem.

[jeremy]

Is the sample rate 4GHz? or is the 3db bandwidth 4GHz?

[lewis]

Nyquist is to to with the sample rate not the bandwidth. If a scope is specified at 4GHz, you will be able to see a 4GHz signal on the screen. But the sample rate has to be at least 8 gigasamples per second to see the waveform reliably (assuming real time sampling). In practice the sample rate is normally well above the Nyquist limit at 4 or 5 times the bandwidth.

The only thing is, with your application, you'll need to allow some bandwidth for harmonics and ringing on the serial line to get a reliable eye pattern. These nasties will be higher in frequency than the fundamental 3.2GHz. So a 4GHz scope might fool you into thinking your waveform is more sinusoidal than it really is. You'll probably need a spectrum analyser too

In short, no, you don't need to take Nyquist into account when buying a scope for a particular bandwidth.

[DutchGert]

Is the sample rate 4GHz? or is the 3db bandwidth 4GHz?

Ah good one, I think the specify the 3dB bandwith.
This is what we are considering: http://www.home.agilent.com/en/pd-1632466-pn-MSO9404A/oscilloscope-4-ghz-4-analog-plus-16-digital-channels?cc=NL&lc=dut
or : http://www.tek.com/datasheet/dpo7000/dpo7000-series-datasheet

[jeremy]

From the linked page: 20 GSa/s max. sample rate

So your theoretical frequency limit for the sampler is 10GHz. I would suggest that you try to get in contact with a salesperson to get a demo and then get some of your MAX9280 and run them in PRBS generating mode across various good and silly setups. That way there is no need to guess if it is enough!

[DutchGert]

Nyquist is to to with the sample rate not the bandwidth. If a scope is specified at 4GHz, you will be able to see a 4GHz signal on the screen. But the sample rate has to be at least 8 gigasamples per second to see the waveform reliably (assuming real time sampling). In practice the sample rate is normally well above the Nyquist limit at 4 or 5 times the bandwidth.

The only thing is, with your application, you'll need to allow some bandwidth for harmonics and ringing on the serial line to get a reliable eye pattern. These nasties will be higher in frequency than the fundamental 3.2GHz. So a 4GHz scope might fool you into thinking your waveform is more sinusoidal than it really is. You'll probably need a spectrum analyser too

In short, no, you don't need to take Nyquist into account when buying a scope for a particular bandwidth.

Now that you say it, it all falls into place, I knew this . Thanks

[T3sl4co1l]

If the bitrate is 4Gb/s then a sample rate of >= 4 GSa/s will just be able to resolve it, given ideal thresholds and all.  (No need to invoke brick wall filters or anything.)  The fundamental frequency of alternating bits at 4Gb/s is 2GHz, so Nyquist is already taken care of.  So to speak.

But talking about Nyquist in regards to a non-band-limited signal is rather perverse anyway.

Since we're talking information, the information-theoretic twist (nyTwist, you might say??!?...) is that, you can only observe *information* with a fundamental bandwidth <= half the sample rate.  Which is a baud rate (i.e., symbol rate -- only the bit rate if the symbols are binary) equal to the sample rate.

Now, if you want to do signal quality analysis instead, you need to observe harmonics, which means a rated bandwidth well into the 10s of GHz.

A bandwidth as low as 1.5GHz might be enough to observe an eye diagram within spec, but it wouldn't be representative of the actual situation (preventing any conclusions regarding BER and susceptibility).

Tim

[macboy]

I feel realy n00b for asking this but recently we started developing products with high-speed serialisers and deserialisers in them (f.e: MAX9280 http://www.maximintegrated.com/datasheet/index.mvp/id/8067)

Since we want to do design verification on them we need a new oscilloscope to measure the serialised data which is transmitted at ~3,2GHz. ...

Do you mean 3.2 Gb/s? This is not quite the same as 3.2 GHz.
And, what does one bit look like, signal wise? Is it a high for a 1 and a low for a zero? In that case, imagine the worst case of alternating 1's and 0's... you will have a 1.6 GHz 'square' wave, not a 3.2 GHz one. (The 'square' wave will be severely rounded off of course, no nice defined edges). Or is it a return-to-zero or something that requires even shorter pulses?

The more important thing I think is the risetime of the scope; that tells you how fast of a pulse it can display. Compare that to the risetime of the signals you are meaning to feed into it. If the risetime of your signal is less (longer) than the risetime of the scope, then the scope will/should be able to resolve the individual pulses. But if you want to see a truer representation of the shape of the signal, you will need to go much higher than that, so that you can see all the defects in the signal (ringing, overshoot, glitches, etc.) that would be effectively filtered out by a marginally adequate scope.

[ejeffrey]

The more important thing I think is the risetime of the scope; that tells you how fast of a pulse it can display. Compare that to the risetime of the signals you are meaning to feed into it.

++

You are interested in time domain behavior, look at the time domain spec.