[time-nuts] Chinese Scopes

[Jim Lux]

On 4/17/12 6:56 AM, shalimr9 at gmail.com wrote:
> I agree that memory depth is an under appreciated parameter, but even 2,500 points like what's available on the cheap Tek scopes is quite useful.
>
> On the other hand, I had a few LeCroy with 50k deep memories and there are cases where that is very useful too. I can't imagine real life use cases when I would need multiple MB. It would be nice to have but seldom used.
>

oddly enough, I had a case where very deep memory was useful last fall. 
  It was an issue with logic that was switching from one clock source to 
another where the clocks were orders of magnitude different frequency 
(10Hz and 300kHz or something like that), and it was the relative timing 
of the edges that was important, so you needed a bunch of cycles of the 
low frequency clock (i.e. record length of half a second or so), but 
enough samples to see the timing of the 300kHz at the same time.

Another deep memory use was when I used a fast 20GHz sample rate Tek 
scope a few years back (2007) debugging a radar target simulator (for 
the landing radar that's going to be used to land on Mars in August) and 
deep buffers were nice there, because we essentially needed to capture 
multiple pulses that were 4 ns to several microseconds long.   The 
requirement was that the delta phase (and time) of successive pulses be 
within a certain value (the radar used what's known as "two pulse 
doppler") following a pre-programmed simulated descent profile.  We also 
wanted the pulse timing after the trigger to be accurate to, as I 
recall, 0.5 or 1 ns.

The PRF is pretty high, so you don't have time to unload the memory in 
between pulses.

So we did something like 500 pulses, captured 16,384 samples at a time 
at 20GS/sec to make a dataset of 16 million samples.

You learn a lot about what's hidden in the specs on inexpensive signal 
generators like the Agilent E4421B when you start comparing phase for 
1600 pulses 1 microsecond apart.