Re: [time-nuts] Z3805A cooling requirements?
Tom, Magnus, Ulrich:
Thanks for the comments and suggestions. They are appreciated and I now have an even better understanding of why ADEV measurements are not a good tool for characterizing the performance of oscillators that are subject to transient events or glitches.
Just to clarify a few points and ask a few questions:
My concern about not putting much emphasis on Adev data for Tau's of less than 80 seconds in the plots I’ve provided is driven by a belief that at shorter Tau's these ADEV plots are largely showing the noise of the counter (an HP5370B) vs the noise of the device being measured. Perhaps the 80 second cut off point is overly conservative but at some point I believe the counter noise will swamp the noise from the devices being measured.
My goal was not to try and use ADEV measurements to characterize the performance of the GPSDO in question while it was subject to fluctuations in air flow (or subject to other transient events..) I did include a frequency plot in my post that provides some insight as to what happened when air flow was added.
The goal was to see if operating the GPSDO in question with air flow changed the ADEV readings vs operating the GPSDO without air flow. I agree ADEV may not be the best tool for this but it is easy to collect and I have prior data to compare the results to. ADEV also seems to be a commonly used figure of merit for characterizing devices such as GPSDO’s. (I realize there are also other commonly used figures of merit.)
The lowest ADEV reading I have ever observed for the GPSDO in question came from analyzing a data set collected 45 thru 65 minutes after air flow was applied to that GPSDO in that particular circumstance. I found that result surprising although I agree the absolute difference in the ADEV figures is very small.
It's my understanding (based largely on comments I've read on this list over the years) that if you have roughly nx10 data points you can begin to draw inferences from ADEV plots for Taus <n. Is this a reasonable practice and or are there caveats one needs to be aware of ?
I agree that one test of this nature is in sufficient to draw any firm conclusions from and much more data is needed.
Regards
Mark Spencer
Message: 5
Date: Tue, 25 Dec 2012 03:12:51 +0100
From: Magnus Danielson magnus@rubidium.dyndns.org
To: time-nuts@febo.com
Subject: Re: [time-nuts] Z3805A cooling requirements?
Message-ID: 50D90BA3.5060205@rubidium.dyndns.org
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Hi,
On 12/24/2012 06:47 PM, Tom Van Baak wrote:
Hi Mark,
I wouldn't place much emphasis on the Adev data for
Tau's of less than 80 seconds.
Actually, just the opposite; the ADEV at short tau is
very close to correct.
I've collected some ADEV data as well but don't
entirely trust it yet.
Right, it's the ADEV for longer tau that is completely
misleading. Let me explain.
Realize that Allan deviation numbers are statistics;
essentially they predict
how constant the future frequency might be, based on a
sampling of measured
frequency in the past. For a statistic like this to be
relevant you want to
have at least 3, but more likely tens to hundreds of
past measurements in
order to have confidence in the prediction.
Here I want to point out what this is the statistics off,
and it is the
statistics noise, normalized to white frequency noise. It is
however not
statistics of systematic effects.
Plotting many Allan deviation statistics, each with a
different sampling
interval, on a log-log plot gives even more
information; the slope of the
line reveals noise types.
Which is the original intent of the ADEV/MDEV curves. MDEV
is being
preferred as it helps to distinguish two noise forms that
ADEV failed to
handle. For far-out noises, ADEV does just was well with
less processing
needed.
Now, there is no problem observing transient phenomenon
like temperature
changes (or phase jumps or frequency jumps or loose
cables or pets
jumping onto the bench). They show up dramatically in
phase or frequency
strip plots. You can see how quickly the effect occurs.
You can measure
the magnitude of the effect. You can measure how long
it takes to recover.
This is all useful: you get numbers like tempco or
thermal Q. But using
standard deviation or RMS or Allan deviation, or any
other statistic
on this data is not the right thing to do -- because
you have only a sample of one.
Consider if you wrap the time-sequence to re-occur at the
same period.
If this is the signal you have, then it is valid. If you
include more
"un-eventfull" time and wrap that, then this wrinkle has
less part of
the overall time, and thus is averaged out. Assume you keep
extending
with un-eventfull time to double each time you end up
averaging the
particular wrinkle out of the plot, but still only approach
an
approximation as a single wrinkle only occur once.
Those, the ADEV tool-set isn't going to give you very
meaningful
interpretation of that wrinkle.
On the other hand, if you encounter tens or hundreds of
these
transients over hours or days or months, then it is
perfectly
valid to use statistics like standard or Allan
deviation to
describe the probability of the transient occurring;
the
magnitude of the effect, etc. Now you have enough
events to
offer a future prediction based on many samples in the
past.
Here I don't agree. Re-occuring "wrinkles" is systematic
effects, and
the impact of systematic effects is different to those of
noise forms.
A sine modulation of frequency and the way we can estimate
it's impact
on future time is quite different from that of inherent
noise sources.
Also, it doesn't scale to the white frequency noise.
Similarly, other systematic effects should be separated out
of the data
before noise analysis.
Does this make sense? In your case "removing air flow"
is only one event.
Indeed. In my experience, forced air as such does not need
to be the
culprit, it just optimize the coupling between ambient
temperature
changes and the oscillator. Varying forced air rate also
counts as
inducing temperature gradients.
I know it's easy to make ADEV/MDEV plots using Plotter
or Timelab
but that doesn't mean it's appropriate in every case.
When your
data has an accidental data glitch or an intentional
transient,
it's best not to use statistics to describe that one
event.
In fact, looking at SP 1065 for instance, cleaning your data
of such
events is assumed normal procedure.
Cheers,
Magnus
time-nuts mailing list
time-nuts@febo.com
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
End of time-nuts Digest, Vol 101, Issue 169
Mark,
On 12/26/2012 06:24 PM, Mark Spencer wrote:
Tom, Magnus, Ulrich:
Thanks for the comments and suggestions. They are appreciated and I
now have an even better understanding of why ADEV measurements are
not a good tool for characterizing the performance of oscillators that
are subject to transient events or glitches.
Good. You do get a gold star for your ADEV over time analysis, also
known as dynamic ADEV. It helps to see where in time a certain ADEV
wrinkle occurred so the time-plot makes sense. Trouble is, you already
have to have a clue to get to that point.
Just to clarify a few points and ask a few questions:
My concern about not putting much emphasis on Adev data for Tau's of
less than 80 seconds in the plots I’ve provided is driven by a
belief that at shorter Tau's these ADEV plots are largely showing the
noise of the counter (an HP5370B) vs the noise of the device being
measured. Perhaps the 80 second cut off point is overly conservative
but at some point I believe the counter noise will swamp the noise
from the devices being measured.
I agree with you, but rather than not showing it, show it and point out
that this is counter-noise. Then that little slope remaining has a
natural explanation and you also get a good line to follow down and
understand where the DUT noise takes over.
It would be cool if we could artificially "remove" that limitation and
see the added noise only.
AVAR_lim(tau) = (ADEV_lim(tau0)/tau)^2
ADEV_corr(tau) = sqrt(AVAR(tau)-AVAR_lim(tau))
It should be fairly simple to fit ADEV_limit to the curve, and it will
represent the white phase noise limit. Seeing that number should also
help to see where trigger noise etc. could be improved.
In a similar sense could other noise-forms be removed, so that you would
have a residual ADEV plot. This is after all what ADEV was developed
for, to establish the levels of noises and have them in separated form.
My goal was not to try and use ADEV measurements to characterize
the performance of the GPSDO in question while it was subject to
fluctuations in air flow (or subject to other transient events..)
I did include a frequency plot in my post that provides some
insight as to what happened when air flow was added.
The goal was to see if operating the GPSDO in question with air
flow changed the ADEV readings vs operating the GPSDO without air
flow. I agree ADEV may not be the best tool for this but it is easy
to collect and I have prior data to compare the results to. ADEV
also seems to be a commonly used figure of merit for characterizing
devices such as GPSDO’s.(I realize there are also other commonly
used figures of merit.)
It all comes down to how "quiet" your ambient air is to your GPSDO/OCXO.
Forced air-flow improves the thermal connectivity between the ambient
air and the GPSDO/OCXO. For many professional buildings and computer
halls, the AC/heating system is not as quiet as you would like. I've
killed several good measurement runs of free-running oscillators just by
walking up to the lab-bench and with it a wall of colder air sweeps over
it...
That's why I try to measure things in a cardboard box just to get
somewhat less airflows on the oscillators, and it works very well.
Forced air as such poses some issues, but ambient air is in my
experience the real killer.
The lowest ADEV reading I have ever observed for the GPSDO in
question came from analyzing a data set collected 45 thru 65 minutes
after air flow was applied to that GPSDO in that particular
circumstance. I found that result surprising although I agree the
absolute difference in the ADEV figures is very small.
Which I could very well believe.
It's my understanding (based largely on comments I've read on this
list over the years) that if you have roughly nx10 data points you
can begin to draw inferences from ADEV plots for Taus<n. Is this
a reasonable practice and or are there caveats one needs to be
aware of ?
Having spent many times watching the data coming into timelab, seeing
the high end flap like a whip until it settles down, I'd say that x10 is
still very unstable, but by all means look at it. The reason you want to
see real confidence intervals on your measure is to know where about the
real value could be compared to the value you currently see.
How tight you want your confidence interval to be depends on what form
of conclusion you want to take. I'd say that even more conservative
values like 100 time samples could be viewed as incorrect for some
applications. This is where you need to decide what you need. Sit down
and see the curve vary for a tau until it settles, that way you learn
where your confidence in values lie.
I agree that one test of this nature is in sufficient to draw any
firm conclusions from and much more data is needed.
It's more about building experience of what matters.
Temperature changes rather than temperature as such affects you, as long
as the oven is operating in linear state.
For one oven I once saw an interesting case, and I realized that the
oven took a "nap" to cool down and then started heating up again. In
effect, during the nap, the crystal was cooling down in an unregulated
environment, and then it was being heated up by a jolt of energy.
Another oven had a self-oscillation in the oven controller, which was
visible from power-on. It also had my current digits flopping around and
current measurement gave the controller away finally. That design was
built on a ceramic brick rather than FR4 board, so it lacked the thermal
mass to remain stable. When the vendor understood the issue, they kept
that design running arguing that the other customers didn't complain. Ah
well.
Cheers,
Magnus