Andrew Warren wrote:

>> To be fair, though, the 2230's digital mode rarely lies to me...
>> Its aliasing problems are nowhere near as bad as what I've seen in
>> my more-or-less limited experience with the Tek TDS220 and the HP
>> 54645D.

and I replied:

>> I'm not sure what you mean. The [TDS220] scopes have a sample rate
>> of 1 GS/s and an analog bandwidth of 100 MHz (60 MHz for the TDS
>> 210).
>>
>> Aliasing should not occur with frequencies less than 500 MHz.
>> Since the bandwith limitation would attenuate such a signal to
>> virtually zero, I cannot see how you would be able to experience
>> aliasing problems.

and then Andy wrote:

>    John Payson explained it well:  The problem has to do with the
>    fixed-number-of-samples-per-graticule-division sampling speed;
>    just because a scope CAN sample at 1 GS/sec doesn't mean that it
>    ALWAYS samples at that speed.

No, I'm well aware of this. The TDS 210/220 samples 2500 points
per "screen shot", which translates into 250 samples per horizontal
division. My original point was that by increasing the SEC/DIV setting
it should be possible to avoid the aliasing since the scope has
a maximum sampling speed of 1 GS/s and an analog BW of 60/100 MHz.

But that's not the point of this posting.
I found Andy's proposed tests quite interesting, so I sat down last night
with my TDS 210 and a PIC.

>    If a scope takes 100 samples per division, signal pulses shorter
>    than 1/100-division can occur BETWEEN samples, where they'll be
>    completely missed.

Agreed.

>    To test your scope for this behavior, program a PIC to generate
>    1-microsecond pulses spaced 5 milliseconds apart, then set your
>    scope to 20 milliseconds per division and see if it always shows
>    exactly four pulses in each division... Unless it has an
>    "envelope" mode which samples at a very high rate and displays
>    the minimum and maximum sampled values for each column of pixels
>    on the screen, it'll miss pulses.

I used an '84 running at 10 MHz, so I didn't get exactly these figures.
(More like 1.2 us pulses spaced 5 ms apart)

The scope does not have a 20 ms/div setting, so I used 25 ms/div instead.
The result was not very good at first. The scope's acquiring method was
"Sample" mode, and it showed almost no pulses at all.
Then I changed acquire mode to "Peak detect", and guess what?
No missing pulses whatsoever.

So this test is passed with flying colors...

>    If your scope passes the above test, you can move on to testing
>    for the noise-susceptibility problem to which John alluded...

>    Modify your PIC program to repeatedly generate a 2-millisecond
>    pulse followed by a string of 47 1-microsecond pulses spaced at
>    1-millisecond intervals... The idea is that you're simulating a
>    signal (a 2-millisecond pulse once every 50 milliseconds) in the
>    presence of noise (a 1-microsecond pulse every millisecond).

>    Set your scope to 200 milliseconds per division and see what it
>    displays.

I had more problems with this one...
Running the scope at 250 ms/div is useless. (Again, the scope doesn't have
a sec/div of 200 ms)

The screen just shows a horizontal black belt. (Correct amplitude, though)

I connected my analog scope (dual trace 20 MHz thing) and at 10 ms/div
I was able to see the signal and trigger correctly. (The slowest sec/div
of this scope is 50 ms)

Then I went back to the Tek and did some more experimenting.
At 10 ms/div I could see the signal, but it didn't trigger correctly.

Then I tried different trigger couplings, and when I used HF Reject
the scope was able to trigger correctly.

At slower sec/div settings (>25 ms if memory serves) the signal looked garbled.

>    It's a rare digital scope that passes both these tests...

The second test is partially passed. (In my opininon)

I've used quite a lot of different scopes, and I still think this little
Tek is one of the best when it comes to value for money.

>> BTW, I'm really pleased with the performance of these scopes. I
>> think they have an unbeatable performance/price factor.

>    I agree that they're great tools for the money... My only point
>    was that there's still a need, even when dealing with
>    relatively-slow signals, for an true analog scope.

I wonder, have any of you seen an analog scope (not analog storage)
that has a horizontal timebase of less than 50 ms/div?

And if so, do you find it useful at such a timebase?
Personally I think it's almost useless. You'll have to have a tube
with very high persistence, and I supect that might cause trouble
when running at faster timebases

So if the purpose of the second test Andy proposed was to compare
a plain analog scope with a digital counterpart, I think it should be
performed with another timebase.
At 10 ms/div I was able to see two of the "signal" pulses and 47 *2 of the noise
pulses. In many cases this would be sufficient to determine what causes the
problem. (Assuming the purpose was to solve a problem)

One more comment: If I had this much noise on a signal I would consider
a complete redesign of the circuit.....

And at last, a question for all you folks out there.
Why do they make scopes with for instance a maximum
sample rate of 100 MS/s sample rate and 400 MHz analog BW?

What's the use?

I'm aware that these scopes usually have a way of displaying
higher frequency signals by "skewing" the sample points, and thereby
building a picture of the signal, but this is in my opinion not very useful
unless you have a periodic signal. And this is certainly not always the case.


-Oyvind