Strangely Small Sampling rates with a 400 MHz Oscilloscope

[ThereIam]

I'm doing an experiment using a super fancy oscilloscope. I'm measuring a tiny output voltage that varies with time. I was instructed to make my sampling rate as small as possible (~500S/s).

Now the signal that I'm looking at is, in other applications when its being sent to a computer and not my scope, converted to a table of about 1 million values, generated about once a second. So I'm literally getting 500 of a million values, and I can't wrap my head around why.

Can you guys think of reasons why the sampling rate ought to be so low? (I do trust this guy, very knowledgeable, but he's on vacation). I get the impression that he did NOT think we were needlessly tossing away data, though having more data would obviously be better for what we're doing. It was a practical limitation of the way the scope was receiving the signal, or somethin' like that.

I'm an extreme novice when it comes to electronics.

Thanks.

 

[berkeman]

I'm doing an experiment using a super fancy oscilloscope. I'm measuring a tiny output voltage that varies with time. I was instructed to make my sampling rate as small as possible (~500S/s).

Now the signal that I'm looking at is, in other applications when its being sent to a computer and not my scope, converted to a table of about 1 million values, generated about once a second. So I'm literally getting 500 of a million values, and I can't wrap my head around why.

Can you guys think of reasons why the sampling rate ought to be so low? (I do trust this guy, very knowledgeable, but he's on vacation). I get the impression that he did NOT think we were needlessly tossing away data, though having more data would obviously be better for what we're doing. It was a practical limitation of the way the scope was receiving the signal, or somethin' like that.

I'm an extreme novice when it comes to electronics.

Thanks.

Maybe to limit the amount of data that has to be transferred and stored? What is the bandwidth of the signal you are monitoring?

 

[ThereIam]

Hi, berkeman, thanks for responding.

The bandwidth on the signal I'm monitoring is 625kHz.

 

[berkeman]

The bandwidth on the signal I'm monitoring is 625kHz.

And you are sampling it at 500Hz?

 

[ThereIam]

Yes, I am.

 

[berkeman]

Yes, I am.

Then the signal is undersampled, and you can get erroneous data because of aliasing and other problems:

http://en.wikipedia.org/wiki/Nyquist_frequency

You need to be sampling at least 2x the signal bandwidth, in order not to alias.

 

[ThereIam]

One thing that I neglected to mention which now makes much more sense to me is that the signal is passing through a preamp with a 3kHz lowpass filter on it... That ought to solve the aliasing problem, right?

 

[berkeman]

One thing that I neglected to mention which now makes much more sense to me is that the signal is passing through a preamp with a 3kHz lowpass filter on it... That ought to solve the aliasing problem, right?

No, 3kHz is too high to be of help here.

In sampled systems like A/D converters (which is what your oscilloscope has at its front end), you need to adjust your anti-alias filtering and sampling rate so that you do not alias your signal into frequencies that are not there.

You need to bandlimit the input waveform with a lowpass filter with a cutoff frequency B and a sharp skirt, and then sample the input waveform at a frequency at least 2B. If the skirt of the LPF is not very sharp, you will need to move the sampling frequency up above 2B to minimize the amplitude of aliased components.

So in your system, if your 3kHz LPF has sharp skirts, you should be able to sample around 6kHz without much aliasing. However, you have obviously lost much of the information in the original 625kHz signal by passing it through a 3kHz LPF, no?

http://en.wikipedia.org/wiki/Sampling_theorem

.