Slightly OT: What good does oversampling provide in an oscilloscope?

[Joel Kolstad]

If I have a digital oscilloscope that has, say, a 1GHz bandwidth, presumably I
need to sample the input signal at something above 2Gsps if I want to capture
all the information present in the signal -- say 2.5Gsps given the anti-alias
filters I might be able to realistically build.

I see, though, that something like a Tektronix DPO4104 actually samples at
5Gsps. I understand how that can buy them another 6dB SNR (from noise
reduction), but other than this... does oversampling buy you anything on a
DSO?

Does anyone out there prefer a classic analog scope like a Tek 2465B over a
*modern* DSO (one that uses color or intensity variation to tell you something
about how the signal spends most of its time, like the brightness on an analog
scope does -- the old DSOs just plotted min & max voltages sampled at each
pixel...)?

---Joel

 

[Tim Wescott]

If I have a digital oscilloscope that has, say, a 1GHz bandwidth, presumably I
need to sample the input signal at something above 2Gsps if I want to capture
all the information present in the signal -- say 2.5Gsps given the anti-alias
filters I might be able to realistically build.

I see, though, that something like a Tektronix DPO4104 actually samples at
5Gsps. I understand how that can buy them another 6dB SNR (from noise
reduction), but other than this... does oversampling buy you anything on a
DSO?

Does anyone out there prefer a classic analog scope like a Tek 2465B over a
*modern* DSO (one that uses color or intensity variation to tell you something
about how the signal spends most of its time, like the brightness on an analog
scope does -- the old DSOs just plotted min & max voltages sampled at each
pixel...)?

---Joel

Joel, I'm not done with the article, so you'll have to wait for the long
answer.

The short answer is that Nyquist is a _theoretical_ limit, in practice
you usually need to go way above Nyquist or you need one big filter. If
you do use one big filter then you cut off way sharper than the
6dB/decade dominant pole that you see in most O-scopes.

I suspect the oversampling is there so they can have a mild slope, at
least at first, in their anti-aliasing filters.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

[[email protected]]

If I have a digital oscilloscope that has, say, a 1GHz bandwidth, presumably I
need to sample the input signal at something above 2Gsps if I want to capture
all the information present in the signal -- say 2.5Gsps given the anti-alias
filters I might be able to realistically build.

I see, though, that something like a Tektronix DPO4104 actually samples at
5Gsps. I understand how that can buy them another 6dB SNR (from noise
reduction), but other than this... does oversampling buy you anything on a
DSO?

Does anyone out there prefer a classic analog scope like a Tek 2465B over a
*modern* DSO

Always when I'm measuring noises. The Nyquist effect is still there
even you have a very good antialising filter.

greetings,
Vasile Surducan

 

[Phil Hobbs]

The short answer is that Nyquist is a _theoretical_ limit, in practice
you usually need to go way above Nyquist or you need one big filter. If
you do use one big filter then you cut off way sharper than the
6dB/decade dominant pole that you see in most O-scopes.

I suspect the oversampling is there so they can have a mild slope, at
least at first, in their anti-aliasing filters.

_What_ anti-aliasing filters?

Cheers,

Phil Hobbs

 

[Joel Kolstad]

Hi Phil,

_What_ anti-aliasing filters?

It had occurred to me that... hmm... building a DC-1GHz lowpass filter that's
reasonably flat is probably somewhat challenging... but you're saying many a
DSO doesn't have any AAF, eh? Interesting...