On Mon, 31 Dec 2012 17:04:16 -0500, Chuck Harris <cfharris@...>
wrote:
However, any competently designed DSO won't allow signals
faster than the Nyquist limit into the sampling stages...
They typically have a brickwall lowpass filter set at some
point significantly below the Nyquist limit to prevent such
aliasing.
Could you point out any examples of DSOs that do this?
I'm not sure that I can point to any that don't.
The Tektronix MSO5000 I evaluated last year certainly did not unless
you count the DSP bandwidth filtering which had to be specifically
enabled and was *not* useful as an antialiasing filter. It also had
the unfortunate side effect of screwing up the pulse response.
While it took a little more work because of its deep acquisition
memories, I had no trouble at all creating full scale aliasing of a 4
MHz reference signal with that 10 GS/sec oscilloscope. I never
considered that a flaw since it had all of the right acquisition modes
available to prevent what I did but antialias filtering was not among
them.
Unfortunately I have not gotten a chance yet to evaluate any modern
Agilent DSOs. The recent Rigol I tested lacked peak detection.
All of the modern ones I am familiar with rely on high real time
sample rates and post processing to prevent aliasing if the record
length is insufficient. Designing a clean analog filter that tracks
the stored sample rate seems like an exercise in futility that would
just result in poor impulse response and even worse, a varying impulse
response at different decimated sample rates.
The DSO's I am aware of sample at their maximum rate all the time,
and have a simplish analog roofing filter to keep the aliasing to
a minimum. They simulate the lower sample rates by digital post
processing.
Decimation in one form or another is usually a better choice than
variable sample rates because it is difficult to keep the later from
altering the characteristics of a digitizer. Even the old enough to
vote Tektronix 2230 and 2232 which I like to use as examples have
constant maximum rate ADC conversion clocks and variable decimation
rates.
All of my old DSOs (they are all old enough to drink) will operate in
equivalent time sampling mode and support peak detection. The lack of
the later is the major reason I never picked up the discussed 7D20.
There are many ways to get a waveform using sampling. All of those
that sample waveforms that are higher bandwidth than the sampling rate
are storing only small parts of many, many, repetitions of the signal
under test. In the case of the 7D20, and the 7854, you may be looking
at snippets of hundreds of repetitions of that signal, just to get a look
at a single copy. In the days of old, these were called sampling
oscilloscopes.
My old Tektronix catalogs always refer to them as digitizers or
digital storage oscilloscopes. The term sampling was always
associated with instruments that had actual sampling front ends.
The confusion of DSO (digital storage oscilloscope), DSO (digital
sampling oscilloscope), and sampling analog to digital converters is
unfortunate.
The 70MHz bandwidth spec isn't specmanship. It tells you that
the amplifiers will pass a 70MHz signal with 3dB attenuation.
This spec gives you some assurance that the analog stages won't
be distorting the signals being sampled significantly.
The 7D20 is a little weird but in equivalent time sampling mode, it
can indeed acquire a 70 MHz or faster signal without aliasing using
its 40 MS/sec digitizer. Its maximum equivalent time sample rate,
limited by its record length, is 2 GS/sec.
Yes, and it requires about 100 triggers of the waveform under test
to assemble that scan... worthless for most purposes.
The only way I use my 7D20 is for single shot waveforms. If I had
access to hundreds of repetitions of the waveform under test, I would
just use an analog scope.
I never picked up a 7D20 because it lacks peak detection but the slow
waveform regeneration rate of my 2230 has only rarely been a problem.
David Hess