DSO bandwidth versus Sample Rate

[Pierre_Montreal]

Hi there,

I wanted to see if Dave has made a video explaining DSO bandwidth because I saw specs for the
Rigol DS2000 series here

http://int.rigol.com/download/Oversea/DS/Datasheet/DS2000_DataSheet_EN.pdf

and they all have the same sample rate for bandwidths of 70MHz, 100MHz, 200MHz. Why ? I tought it was related to the Nyquist sampling rate but there seems to be something else.

Any idea  or reference material ?

Thank you, bye.

[olsenn]

From my weak understanding of it, if the sampling rate is less than (or equal to) the Nyquist value (400MSa/s for 200MHz etc) then the waveform will be distorted and useless. However, oversampling, as far as I'm aware, is not meaningless. So the more the merrier.

The DS2000-Series has 2GSa/s sampling rate. That's 1GSa/s for each channel. Sampling is 5X max bandwidth (200MHz) >> 2X, so it should be good

[Pierre_Montreal]

Ok, but why do they have the same sampling frequency but different bandwidth ?

[Lightages]

They are all the same piece of hardware. When things are builtall the same way the easiest way to reduce the performance of the "lesser" models is to artificially reduce the bandwidth with input filtering. If they tried to make different sampling rates for each model it would be much more complicated due timing issues in hardware and software. The cheapest thing is a couple of caps and resistors thrown into the front end.

It is very common practice for many years now to design for the highest model and cripple the design for lower models.

[wraper]

Because that's how they make money by selling the same thing with different stickers by crippling the bandwidth (usually in software, hardware remains the same).

[AG6QR]

Ok, but why do they have the same sampling frequency but different bandwidth ?

They drive the A/D converters as fast as they can.  Their sample rates are 2GSa/s when in single channel mode, 1GSa/s in dual channel mode.

For a 200MHz signal, 2GSa/s is 10 samples per cycle at the rated analog bandwidth, which is enough to see what's going on reasonably well.  But 1GSa/s is only 5 samples per cycle, which is less than ideal.  The 200MHz version could probably benefit from a faster sampling rate in dual channel mode, if it were available.   For a 100MHz signal, that same 1GSa/s works out to 10 samples per cycle, so those scopes should do a good job of displaying 100MHz signals in dual channel mode.

[Pierre_Montreal]

The answer from Lightages is very interesting, I didn't know that I was expecting some advanced mathematical principle in sampling theory versus electronics components or else.

Thank you !




 

[larry42]

Because bandwidth is the analog bandwidth (at the "limit" the response is 0.707x down from the low frequency response). And the sampling rate is the sampling rate. The two are independent.

Early sampling scopes had analog bandwidths >> than the sampling rate and were therefore only useful for repetitive signals.



Sent from my HTC One S using Tapatalk 2

[larry42]

Ok, but why do they have the same sampling frequency but different bandwidth ?

They drive the A/D converters as fast as they can.  Their sample rates are 2GSa/s when in single channel mode, 1GSa/s in dual channel mode.

For a 200MHz signal, 2GSa/s is 10 samples per cycle at the rated analog bandwidth, which is enough to see what's going on reasonably well.  But 1GSa/s is only 5 samples per cycle, which is less than ideal.  The 200MHz version could probably benefit from a faster sampling rate in dual channel mode, if it were available.   For a 100MHz signal, that same 1GSa/s works out to 10 samples per cycle, so those scopes should do a good job of displaying 100MHz signals in dual channel mode.

The amount of oversampling required depends greatly on the frequency content of the signal.

With a sine wave input and a sinc filter then you only need about 2.5 times the analog bandwidth.

If there is no spectral energy above half the sampling frequency and you have a reconstruction filter, then I see no need for x5 or x10 oversampling.



Sent from my HTC One S using Tapatalk 2

[AG6QR]

Ok, but why do they have the same sampling frequency but different bandwidth ?

They drive the A/D converters as fast as they can.  Their sample rates are 2GSa/s when in single channel mode, 1GSa/s in dual channel mode.

For a 200MHz signal, 2GSa/s is 10 samples per cycle at the rated analog bandwidth, which is enough to see what's going on reasonably well.  But 1GSa/s is only 5 samples per cycle, which is less than ideal.  The 200MHz version could probably benefit from a faster sampling rate in dual channel mode, if it were available.   For a 100MHz signal, that same 1GSa/s works out to 10 samples per cycle, so those scopes should do a good job of displaying 100MHz signals in dual channel mode.

The amount of oversampling required depends greatly on the frequency content of the signal.

With a sine wave input and a sinc filter then you only need about 2.5 times the analog bandwidth.

If there is no spectral energy above half the sampling frequency and you have a reconstruction filter, then I see no need for x5 or x10 oversampling.

That's all true, but if you're probing an unknown signal, you can't always be sure in advance that those conditions will be met.  Hence the common rule of thumb that x10 oversampling can be useful.  It's a rough guideline of course, not an absolute.  There are situations where x10 is overkill, but that doesn't mean it's always wasted.

The analog input stage will have rolloff, but it will pass (with attenuation) frequencies significantly higher than the bandwidth if they're present in the input signal.  An oversampled signal that shows something at frequencies well above the rated bandwidth may not be a really accurate picture of the true waveform, but at least it can provide a clue that there's something going on at higher frequencies.

[nuhamind2]

Nyquist applies to bandlimited signal

[R_G_B_]

Dont just rely on the sin wave frequency responce of the scope. Take ino consideration how it displays a square wave 
slew rate Volts/ns, rise time etc etc. Sampling rate is meaningless on its own.

[olsenn]

Dont just rely on the sin wave frequency responce of the scope. Take ino consideration how it displays a square wave 
slew rate Volts/ns, rise time etc etc. Sampling rate is meaningless on its own.

I think the idea is that if you decompose the square wave into its frequency response (Fourier) you'll see that it is composed of harmonics of the fundamental frequency. If your DSO has an analog bandwidth of 100MHz, and the square wave you wish to view is of a 5MHz frequency, than your DSO will be able to show 20 harmonics (incl. fundamental).

(max) Rise time can be calculated based on bandwidth and vice versa.