[time-nuts] Fluke/Pendulum PM6680/CNT-80 measurement instability/outliers

Bill Byrom time at radio.sent.com
Thu Dec 14 23:08:01 EST 2017


I have a few questions about your setup:
(1) What is the waveform shape you are measuring? Are you measuring a
square wave (or some other duty cycle waveform with fast rise and
falling edges)? Or is it a sinewave with low harmonic content?
(2) How is the counter voltage threshold set with respect to the peak
excursions? Is it set to the 50% point or near the bottom or top of the
waveform?
(3) How are you connecting the source to the counter input? Is it a 50
ohm RG-58 cable with BNC connectors? Are there any TEE connectors or
feed-through terminations? What is the cable impedance? You can easily
find 75 ohm cables with 75 ohm BNC connectors which will mate (with a
bit of difficulty in some cases) with 50 ohm female connectors. If you
have a 50 ohm counter impedance, a 75 ohm cable can cause difficulty.
(4) What is the output impedance of the source you are measuring?
(5) What is the input impedance of the counter (as it is set up for your
test)?

Here are a few observations about your results and my guess of the root
cause:
(A) As presented, the outliers are all positive offsets at multiples of
1 count (1, 2, 3, 4, 5, and 6 extra counts).
(B)  So the reason for the multiples of 1 Hz offset is obvious. With a 1
second gate interval, occasionally the counter measures 1 to 6 extra
counts.
(C) My belief is that you have some "noise" source or waveform
distortion which is generating the extra counts. It's a marginal case
where the interference is only occasionally counted. For example, the
noise or distortion might be passing through the counter input voltage
threshold without enough overdrive to reliably generate spurious counts. 
(D) If the source waveform is a square wave (or any pulse type waveform
with fast rising or falling edges) and there are any reflections due to
poor load match (the impedance of the counter not matching the coax
cable impedance), you will get a reflection. If the source is also
poorly matched (such as a low impedance drive from a digital logic
buffer), you will get a second reflection. This second reflection will
arrive at the counter input delayed from the original edge by twice the
propagation delay in the cable.
(E) If you use a 50 ohm RG-58 cable with a solid dielectric, the
velocity factor will be around 66%. So the propagation velocity is about
2 x 10^8 m/sec. This results in a round trip (2 m distance) delay of 10
ns.
(F) So if there is a mismatch at each end of the cable you will see a
second edge arriving at the counter input with a delay of 10 ns per
meter of cable length.  If it's a 100 MHz bandwidth counter, it will
count these extra transitions which pass through the input voltage
threshold.

Some potential causes for this condition:
(i) Poor source and load impedance match. If the counter only has a 1 M
input resistance, use an inline 50 ohm broadband terminator. I work for
Tektronix as an Application Engineer, and for over 50 years we have sold
inline 50 ohm terminators, such as this one:
https://www.bmisurplus.com/products/54615-tektronix-011-0049-01-termination-50-ohm-2-watt
You can also find many other sources for inline 50 ohm terminators. But
be careful -- if the counter already has a 50 ohm input impedance
(internal termination), use of an external 50 ohm terminator will result
in a 25 ohm combined impedance and a reflection. The same issue is found
at the source, but if you can get the load properly matched you don't
need to worry much about the source.
(ii) If the source is a pure sinewave (with low harmonic distortion) you
won't see this issue with matching, since there are no fast edges.
(iii) If the counter voltage threshold is set to a value near the upper
or lower waveform levels, it's easier to get spurious counts. Noise or
spurious ingress (broken or poor quality cable shield) can cause
problems. 
--
Bill Byrom N5BB

On Thu, Dec 14, 2017, at 10:58 AM, Attila Kinali wrote:
> Hi,
> 
> Just for fun, I've set up a PM6680 to measure a 10MHz OCXO (Trimble
> 34310).
> First I used the 10MHz output of an Trimble UCCM as reference (green
> trace
> in the attached file, drift removed, offset to match second trace),
> measuring
> with a gate time of 1s, one measurment every approximately 1.1s.
> As you can see, there are a couple of outliers, evenly spaced in 1Hz
> steps.
> To make sure, these weren't comming from the GPSDO (it's more noisy than
> I thought it would be), I switched to the PM6680's internal reference
> (violet trace) and got basically the same outliers.
> 
> If these outliers would have been only at 1Hz, I would have said they
> are metastability issues in the coarse counter (which runs at 10MHz).
> But getting metastability to last more 100ns is exteremly unlikely,
> yet I get outliers that are 6Hz(=600ns) out. The probability for
> metastability
> lasting that long is lower than being struck by a meteorite....much
> lower!
> 
> So, the question is, what am I looking at? What causes those outliers?
> Does someone have a good idea?
> 
> 				Attila Kinali
> -- 
> It is upon moral qualities that a society is ultimately founded. All 
> the prosperity and technological sophistication in the world is of no 
> use without that foundation.
>                  -- Miss Matheson, The Diamond Age, Neil Stephenson
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
> Email had 1 attachment:
> + pm6680-instability.pdf
>   938k (application/pdf)


More information about the time-nuts mailing list