On Sat, Jul 15, 2023 at 05:47 PM, Paul Amaranth wrote:

I think part of the problem is the start and stop signals overlap each
other.

Does the manual specify those timing details? (start/stop timing, max speeds, etc.) I would think that it does...(?). Is it working "as designed"?

As for the LFSR polynomial, I remember that the HP5005 manual shows it. I don't know about Tek.
If the polynomials are the same, then it's just a matter of how the states are clocked in with HP vs Tek.

Pete

Interesting stuff

So the signature gets unstable at higher frequencies. It seems to be
missing a stop signal and the gate window is twice what it should be.

I stole the design from the Tek 308, so I hooked it up and, what do
you know, the 308 gets an unstable signature at higher frequencies with
the same unstable sigs. (The 308 also does not work when the stop and
start signals are from the same signal line).

The HP 5004a does not do that but the gate control design is completely
different.

I think part of the problem is the start and stop signals overlap each
other. If they were single discreet pulses separated in time this
situation wouldn't happen (which is probably how you would set up an
instrument). So in practice, this may not matter much, but ...

I'm going to have to think about how to revamp the gate control. The
goal is to be at least as good as a 5004. Merely being as good as
a Tek 308 may not be good enough :-)

Paul

I wouldn't argue for SA on new builds, I think it was a production tool
whose time has passed.
But if you like it and are writing software for a new device it's probably
possible to calculate the results of specific test patterns, or simulate
them.
I do think it's simpler to work with a scope or logic analyser and debug
intelligently, but still, it wouldn't be impossible.

On Tue, Jul 11, 2023 at 12:46 PM, John Griessen wrote:

So, even with your early experience using SA in a test bed, you
think the speeds are a killer to this approach?
What if Harvey puts it in CPLDs, would that go fast enough?

To me, having SA available on a current design with chip radios
could be good. You could trigger a startup of some of the code,
then do SA on bitstreams going to the radio and maybe even coming
from it -- if you had the right trigger on the start of an
external radio signal being received...

Using modern parts, you can certainly make SA work at high speeds. There's no inherent speed limitation.
But I think the main point is that SA is only useful in certain very specific cases that are (overall) very rare.
- For existing equipment, you must have known-correct documented signatures (or capture them yourself from a known-good unit). Since setting up an appropriate loop can be non-trivial, by the time you do it, you could have (perhaps) diagnosed the fault with traditional methods.
- For something you build yourself, you have to make it work *before* you can capture known-good signatures (which is then mostly pointless).
- For gear that includes digital and analog circuitry (like a DVM or radio for example), SA won't help with the analog parts.
- If it's modern gear with a microcontroller, in nearly all cases you get much better results by investing in diagnostic firmware rather than SA. I could be wrong, but I think that NOBODY invests in SA for modern gear any more.

Pete