MSO1000s vs MSO2000s

[FrankT]

Just a quick question about the Rigol MSO1000 series vs MSO2000 series.

The MSO1000 series has 4 analog channels, 16 digital channels, at 1GS/s, while
the MSO2000 series has 2 analog channels, 16 digitial channes, at 2GS/s.

But the MSO2000 is a lot more expensive MSO-1074Z $AU943, vs $1380, where I usually see 4 channel scopes costing more.

Is the 2GS/s over 1GS/s worth that much more $ or are there other significant differences I am missing?

[MarcelM]

Hello FrankT,

Some points to consider:

MSO2000A : 1 channel @ 2Gs/s
                   2 channels @ 1Gs/s

MSO1000   : 1 channel @ 1Gs/s
                   2 channels @ 500Ms/s
            3 or 4 channels @ 250Ms/s

So for me, the sample speed of the 2000 series is double that of te 1000 series, and once you start using those extra two channels on the 1000, the usable bandwidth drops significantly.

If you're like me, the 16 digital channels on the MSO severly limit the number of times I need more than 2 analog channels...

YMMV,

Just my €0,02

Cheers,
Marcel

[mjkuwp]

I have seen written elsewhere that the MSO2XXX series has a real-time clock and the other does not.  For me this would be a big difference because I would like to know the actual time at which I took data.  however, for others this may not matter at all.

I have not seen this explicitly laid out in Rigol literature so it is yet to be confirmed.

[FrankT]

If you're like me, the 16 digital channels on the MSO severly limit the number of times I need more than 2 analog channels...

Thanks, you did read my mind.

I want to upgrade my existing USB scope which has 2 analog channels and 16 digital, but a pitiful 64k of sample memory and 80MS/s.  It is the small memory that is giving me grief.

I need to try and remember whether I've been limited by analog channels, or sample rate.

[David Hess]

So for me, the sample speed of the 2000 series is double that of te 1000 series, and once you start using those extra two channels on the 1000, the usable bandwidth drops significantly.

Do these oscilloscopes support equivalent time sampling?  It is not always clear from their documentation.  I would hope the bandwidth is independent of sample rate.

[Mark_O]

Do these oscilloscopes support equivalent time sampling?  It is not always clear from their documentation. 

No, they do not.  None of the recent Rigols do.  For the most part, the current generation of econo-performance scopes have dropped this capability.

[Electro Fan]

Do these oscilloscopes support equivalent time sampling?  It is not always clear from their documentation. 

No, they do not.  None of the recent Rigols do.  For the most part, the current generation of econo-performance scopes have dropped this capability.

What is the definition of "equivalent time sampling"?

[Teneyes]

What is the definition of "equivalent time sampling"?

"Equivalent–time sampling works by constructing a picture of the input signal by accumulating the samples over many wave cycles. Because ETS samples the waveform over a number of cycles it can only be used to measure signals that are repetitive. ETS cannot be used for single–shot or non–repetitive signals."
See: http://www.tek.com/document/application-note/real-time-versus-equivalent-time-sampling

Rigol uses Real-Time sampling. (simple sampling)

[Slappy_g]

No, they do not.  None of the recent Rigols do.  For the most part, the current generation of econo-performance scopes have dropped this capability.

That said, if you look at the sampling rate on the MSO 2000A series scopes, their sampling rate does exceed the analog front-end bandwidth, so it appears there'd be no value in doing equivalent-time sampling on that series.  (Unless I am totally misunderstanding ETS, it aims to allow a slower sampling rate to get the most out of a high-bandwidth analog front-end.)

For reference, 2 GS/sec (on the MSO2000A series) = approximately 800MHz effective digital stage bandwidth, as per your linked article.

[David Hess]

No, they do not.  None of the recent Rigols do.  For the most part, the current generation of econo-performance scopes have dropped this capability.

That said, if you look at the sampling rate on the MSO 2000A series scopes, their sampling rate does exceed the analog front-end bandwidth, so it appears there'd be no value in doing equivalent-time sampling on that series.  (Unless I am totally misunderstanding ETS, it aims to allow a slower sampling rate to get the most out of a high-bandwidth analog front-end.)

Unfortunately there is a circumstance where this does matter.

If the input signal contains components which are close to but still below the Nyquist bandwidth, then non-linearity and aperture jitter from the digitizer will mix with them producing sidebands above the Nyquist bandwidth and aliasing will have occurred.  The result is commonly visible on fast edges or high frequency sine waves where each acquisition "wobbulates" in amplitude and time.  On a DPO, the signal looks smeared in amplitude and time which might be taken for jitter and noise which it is but it is jitter and noise from the digitizer and not the signal.  Sin(x)/x reconstruction makes it look even worse.

Equivalent time sampling ameliorates this problem by simply allowing much higher sample rates raising the Nyquist bandwidth.

Agilent has an application note discussing this among other things:

http://cp.literature.agilent.com/litweb/pdf/5989-5732EN.pdf

Since running across this issue, I have wanted to test an oscilloscope that uses digital triggering to see how it behaves under these circumstances but I have not had the chance.  I suspect that triggering on the signal after aliasing will make the problem even worse.

For reference, 2 GS/sec (on the MSO2000A series) = approximately 800MHz effective digital stage bandwidth, as per your linked article.

The problem occurs even with signals that only contain frequencies below the Nyquist bandwidth.

Next time I have the opportunity to evaluate a modern DSO, I will test it specifically for this which is easy enough with a fast transition time pulse generator.

[Slappy_g]

For reference, 2 GS/sec (on the MSO2000A series) = approximately 800MHz effective digital stage bandwidth, as per your linked article.

The problem occurs even with signals that only contain frequencies below the Nyquist bandwidth.

Next time I have the opportunity to evaluate a modern DSO, I will test it specifically for this which is easy enough with a fast transition time pulse generator.

And clearly, my somewhat stale EE knowledge is eclipsed by someone who knows more about this topic than me... 

[David Hess]

For reference, 2 GS/sec (on the MSO2000A series) = approximately 800MHz effective digital stage bandwidth, as per your linked article.

The problem occurs even with signals that only contain frequencies below the Nyquist bandwidth.

Next time I have the opportunity to evaluate a modern DSO, I will test it specifically for this which is easy enough with a fast transition time pulse generator.

And clearly, my somewhat stale EE knowledge is eclipsed by someone who knows more about this topic than me... 

I would not consider it a well known issue and in most applications it is not a problem.  It is just another way DSOs can be misleading.  I noticed it 20+ years ago and guessed the cause but never assigned much importance to it because equivalent time sampling was available to avoid or at least detect it.  With newer DSOs however if equivalent time sampling is not available, this issue may have increased importance.  As the Agilent application note mentions, some DSOs have much better digitizers than others and since it is easy to test for, I would do that when evaluating a DSO now that I am aware of it.

I noticed something similar a couple years ago when I evaluated a Tektronix MSO4000 series oscilloscope but I suspect that was Gibbs phenomenon caused by the DSP bandwidth limiting:

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

I assume the two are related but it is not entirely clear to me how.

[Teneyes]

I noticed something similar a couple years ago when I evaluated a Tektronix MSO4000 series oscilloscope but I suspect that was Gibbs phenomenon caused by the DSP bandwidth limiting:
http://en.wikipedia.org/wiki/Gibbs_phenomenon

Interesting,
Here is the display of  Discontinuous data (manually set) on the DS2000
Is this the Gibbs phenomenon  in the  Fourier sum for Sin(x)/x interpolation of a step?

[David Hess]

I noticed something similar a couple years ago when I evaluated a Tektronix MSO4000 series oscilloscope but I suspect that was Gibbs phenomenon caused by the DSP bandwidth limiting:
http://en.wikipedia.org/wiki/Gibbs_phenomenon

Interesting,
Here is the display of  Discontinuous data (manually set) on the DS2000
Is this the Gibbs phenomenon  in the  Fourier sum for Sin(x)/x interpolation of a step?

I think it could be either as I have seen both and they look pretty much identical which is why I suspect a direct link between the two.  If it is caused by aliasing before sin(x)/x interpolation, then I would expect the displayed waveform to change shape and amplitude with different asynchronous single shot acquisitions.  If it is the Gibbs phenomenon, then I would expect it to be static.

Since the Rigol and other recent cheap DSOs lack equivalent time sampling and apparently implement bandwidth limiting for market segmentation in firmware, it would not surprise me if they visibly suffered from both problems simultaneously.

It is worth noting that an completely analog(!) oscilloscope can produce similar results with a fast edge if the phase response of the vertical signal chain is not linear which usually crops up as a problem with the phase and amplitude compensation for the vertical delay line.

[Teneyes]

To be clear. I patched the waveform data to be an exact step , with 140 data points at 500ps/sample.
And reloaded into the DSO.  There is no Analog circuits involved with that display

[David Hess]

To be clear. I patched the waveform data to be an exact step , with 140 data points at 500ps/sample.
And reloaded into the DSO.  There is no Analog circuits involved with that display

That makes it the Gibbs phenomenon then since no sampling occurred unless the internal representation is resampled for some reason.  I think finite DSP precision could cause this as well.

Is the DSO applying the DSP bandwidth filter they use for market segmentation to patched waveforms?  That could make some sense although I am not sure I would want to justify that to a customer.

Update: The sin(x)/x reconstruction filter will produce exactly that result and it is the Gibbs phenomenon.  The solution is to increase the sample rate or bandwidth limit the signal.

[Teneyes]

Is the DSO applying the DSP bandwidth filter they use for market segmentation to patched waveforms?  That could make some sense although I am not sure I would want to justify that to a customer.

In the DS2000 the front-end fIlter is software programmable BW amplifier (LMH6518 )

I will redo the test, at different BW options installed (70,200,300) to see if that will affect the Interpolation program, I think NOT.

As a point of interest, here are 2 displays where I adjusted the data to be less discontinuous to see how smooth I could get the displayed trace and still be a fast step. I think it's Gibbs
1st the Sample Dots ,then the Interpolated 'Vector' display

[David Hess]

Is the DSO applying the DSP bandwidth filter they use for market segmentation to patched waveforms?  That could make some sense although I am not sure I would want to justify that to a customer.

In the DS2000 the front-end fIlter is software programmable BW amplifier (LMH6518 )

I will redo the test, at different BW options installed (70,200,300) to see if that will affect the Interpolation program, I think NOT.

I wonder how they get 70 MHz when the LMH6518 supports 20 MHz and 100 MHz but not between them.  Maybe a custom version?

As a point of interest, here are 2 displays where I adjusted the data to be less discontinuous to see how smooth I could get the displayed trace and still be a fast step. I think it's Gibbs
1st the Sample Dots ,then the Interpolated 'Vector' display

I do not think there is any mystery here.  The sin(x)/x reconstruction filter results in overshoot if the edge has signal components above the Nyquist bandwidth at the sample rate used.

[Teneyes]

I wonder how they get 70 MHz when the LMH6518 supports 20 MHz and 100 MHz but not between them.  Maybe a custom version?

That is correct, Some owners have tested the 100Mhz option and found it to be the same as 70Mhz 

[Mark_O]

I wonder how they get 70 MHz when the LMH6518 supports 20 MHz and 100 MHz but not between them. 

Advertising.