[Open-graphics] Re: Digital 'scope & spectrum analyzer

[James Richard Tyrer]

Dieter wrote:
>>>>> You need to be able to *capture* the data in real time, in 
>>>>> order to do a single sweep mode, for non-periodic signals. 
>>>>> The processing and display of that data don't have to be real
>>>>>  time.
>>>> Yes, that is true for some applications, but unless this is a 
>>>> real time spectrum analyzer ($$$$) -- which is useful for some
>>>>  applications such as looking for noise and intermittent 
>>>> distortion -- you are going to be working with a stable 
>>>> periodic wave form.
>>> That was in reply to:
>>> 
>>>>>> To be more specific, a digital oscilloscope displays a 
>>>>>> periodic signal on the screen and the refresh scans rather
>>>>>>  slowly from left to right.
>>> Single sweep mode is a very useful, often essential, feature for
>>>  an oscilloscope.
>> Yes, a one shot scope is a useful device.  With analog this was a 
>> scope with a storage CRT.  With digital this requires a very fast 
>> flash ADC (perhaps more than one).  I think that this is going to 
>> be rather expensive compared to a scanning type D-scope for stable
>>  AC signals.
> 
> So this "scanning" type would use a A/D with a fast sample time but a
>  long time between samples?

Actually not.  The time between samples would be just slightly more than
1/f -- (N + 1)/N.  At least at the maximum frequency.  I presume that
you could take multiple samples per cycle at lower frequencies but I am 
not familiar with that type of scope.

> It would collect data for a long time, and then calculate a time or 
> frequency domain graph?  So what sort of bad things happen when the 
> signal isn't periodic?

The same sort of things that happen with an analog scope.

> A time domain graph would be impossible, right?  Could you sample for
>  a long time and get a frequency domain graph that would be an 
> average?

Yes, there are commercial products that do this called digital signal
averagers.  They are basically a digital scope with a lot of memory.

>> The nice feature of one shot D-scopes is that they can record a lot
>>  more data than an analog one -- you can record a lot more data 
>> than will fit on the screen at once.
> 
> Very nice if the test is time consuming to set up, or worse, if it is
>  a destructive test.  And you don't need a storage CRT or camera. 
> Upload the data easily,  Remote access.  Could be used for automated
>  tests.

An inherent issue is that you can't take data faster than your ADC and 
faster ADCs aren't as accurate.  It depends on your application which is 
more important.  I presumed that you would be using a spectrum analyzer 
to accurately analyze the harmonic content of a constant and periodic 
signal.  An instrument good for that type of application isn't going to 
be as good for non-repetitive signals.

>> This has always been a serious issue with an analog spectrum 
>> analyzer. In theory, it should be a Gaussian distribution.  This is
>>  not realizable because it would have to extent to infinity.  But 
>> even taking a polynomial distribution, it is still impossible to 
>> exactly realize a band pass filter with that response function. 
>> And, the filter also needs to have linear phase response.  So, this
>>  -- the scan filter response shape -- is an important feature and 
>> something that is better in more expensive units.
>> 
>> I suspect that the band pass response shape in these devices is not
>>  anything near a Gaussian response or linear phase.
> 
> Does this band pass response shape issue go away with a digital SA? 
> Digital filters class was a long time ago, but IIRC you can get any 
> response you want since it is math.

There is no band pass response with a digital spectrum analyzer since it
assumes that the signal is periodic, with a known period, and computes 
the harmonics up to the Nyquist limit -- a filter isn't needed to do 
this.  An analog spectrum analyzer doesn't know the period so it can't 
work that way.

-- 
JRT