[time-nuts] More Noise Floor Nonsense
Bob kb8tq
kb8tq at n1k.org
Wed Feb 10 17:18:55 UTC 2021
Hi
Back a few years ago, Keysight did one of their PR videos:
https://www.youtube.com/watch?v=4DfLkxVe7Lk <https://www.youtube.com/watch?v=4DfLkxVe7Lk>
It shows a fairly simple setup to see what your two channel
counter is doing. A lot of counters have a built in standard
deviation function so it applies to more than just the counter
he’s showing. People have been doing this sort of test for many
decades.
The one thing he’s not to clear on is what sort of signal levels
are being used. “Square wave” is about as specific as he gets.
I decided to try the same test with a 100 MHz sine wave at
2.5 dbm. Sine wave is nice since it has a well defined slope
at the zero crossing ( which does matter in this test). It also
happened to be an available output on the FS-740.
If you repeat the YouTube test with this signal, you get the same
13 ps one sigma and 9 ps resolution on a 53230A. That suggests
to me that this is a “fast enough” signal to be useful. Same test
gave the same results on a couple of 53230’s.
So … what happens if we try this on a few other counters?
A 53132A comes in right around 140 ps one sigma to give it
a 100 ps resolution. That’s nice compared to a 150 ps spec.
The “surprise” candidate is the 53220A. One would *think* that
with a 100 ps resolution spec, it would be 5X worse than the
53230A. Not so much. It comes in at 14 ps one sigma and 10 ps
resolution. ( Usual disclaimers apply …. used counter of
unknown (China / eBay) origin …. one sample tested…. YMMV ….).
Does this test tell you *everything* about your counter? No
of course it doesn’t. There are a lot of bits and pieces in there,
this is only looking at some of them. Does it tell you what you
will see on a 10 MHz sine wave at 10 dbm? Nope, that’s a
slower edge. The one sigma is roughly 24 ps on a 53230A.
What it *does* provide is a quick check of how your counter
is doing. Most of what gets killed in a counter is in the input
channels (as noted in a previous post). This gives you a fast
check on how they are doing. It also checks a bit more than
that ….
To be really useful, you need to run the test with *your* source
on *your* counter(s). (It does not have to be 100 MHz and 2.5
dbm) Note the number(s) you get and file them away. Come back
a year or three later scratching your head about some readings
….. you have something to check against. Checking against spec
… not so much ….
The key thing in this test as shown in the video *is* the fast edge.
For a lot of years many of us have simply been using a 10 MHz
test signal. There’s nothing wrong with doing that. One might
*guess* that’s what the designers of the 53230 did. Turns out
that a faster edge is a better edge, even at RF frequencies.
Who knew …. ( other than anybody who watches YouTube …. :) )
Could you extend this test by playing with various levels to come
up with a good idea of what to expect at this or that level
and this or that frequency? Sure you could ….
Might a faster signal produce better results? I have a nasty
suspicion that the video was done with a signal that was
fast enough to show the best numbers possible on the ‘230.
That’s just a guess …..
Does the 53220 “surprise” also carry over to the 53210?
There’s no way to know. This test simply does not work
on a single channel counter. The 53180 is in the same
“twilight zone”.
How does a recently aligned and fully functional 5370
compare on a “fast edge” test. We have lots of data with
10 MHz…. Same question on the SR-620 and CNT-90.
Grubby details on the setup:
Both inputs AC coupled / 5V range
Auto Level turned off on both channels
Channel A set to 1 M ohm
Channel B set to 50 ohms
Both trigger levels set to zero volts
Measure TI from A to B
My coax was about 2.5 ns long.
External reference applied to all counters
External ref not coherent with the test signal.
It took you longer to read all that nonsense than it will to
run the test :)
Fun !!!
Bob
More information about the Time-nuts_lists.febo.com
mailing list