All,

The following comment appeared on this list recently and it scared 

me a little:

I assume this instability is due to the instability of the internal
frequency standard.  There are two options for the SR620: the standard
oscillator and an ovenized oscillator.  In fact, in our measurements, we
plan to use a Cesium frequency standard as the timebase to our SR620.
Does this anecdotal warning apply generally to the instrument or mainly to
the use of the internal standard oscillator?

We are using the SR620 to measure the interval between 1PPS signals from
two clocks.  One is the Septentrio PolaRx4 GPS receiver and the other is a
Rubidium clock.

Many Thanks,
Paul

--
Paul DeStefano

Hi

Some of the problem comes from the OCXO. Some of it also comes from the stability of the internal circuitry. They do some interesting things with delay lines and the like in the SR620. Best bet is to let it warm up for a while before you need it to perform at it's top level of performance.

Bob

On Dec 2, 2012, at 2:45 PM, Paul DeStefano paul.destefano@willamettealumni.com wrote:

Paul wrote:

I concur with the comment above that the thermal design of the 620
could have been better -- the sensing thermistor is in an "exhaust
stack" between the fan (which is blowing out) and the rear enclosure
wall.  This means that, instead of trying to maintain the internal
instrument temerature at a constant level, it tries to maintain the
exhaust stack temperature constant with a viciously fast response
time that leads to instability at startup.  I have more than once
considered moving the thermistor to a location near the TCXO, but
since the fans always run up to full speed rather quickly at room
temperature anyway, I have never bothered to try to improve the fan circuit.

Additionally. the TCXO remains powered during standby, but not
exactly on frequency because the DAC that adjusts it during operation
is not powered.  So, there is some settling from that adding to the
temperature drift.  Note also that the DAC steps are not very fine,
so you cannot expect to get the internal oscillator trimmed to better
than e-9 or so.  SR apparently thought that most users would connect
620s to external standards, so there was no reason to make them pay
for a high-precision internal standard they would not use.

IME -- operating with an external reference that is better than the
specified accuracy of the 620 -- they meet SR's specifications within
a few minutes at most after switching on from room temperature
storage.  (The trigger circuitry may drift a bit as it warms up, so
you may want to check the trigger drift if your application involves
slowish sine waves.  I have not investigated this.)  Ideally, you
would let the instrument warm up for at least an hour and then
perform an internal calibration before starting your measurements.

All that said, the only way you will know for sure how your
particular instrument and standard will perform is to characterize
them before you start your mobile measurements.  In doing so, you
should observe a protocol that resembles the actual travel between
measurements, at least with respect to time and temperature.  I
strongly urge you to do this so you can have confidence in your measurements.

Best regards,

Charles

On 12/02/2012 10:25 PM, Charles P. Steinmetz wrote:

Never thought about that part, other than the fact that the fan is
annoying like hell.

Which is why a high stability reference is an option, like most.

In contrast it is interesting to note that the HP5370 had a lower
stability oscillator as option, so removing the 10811 was thus using a
negative option.

Another source of temperature dependence is the analogue interpolator.

This is good advice. Consider what you need to do. You might want to
consider having a rubidium doing your hold-over. for instance. A PRS-10
would be an interesting option for instance.

Cheers,
Magnus

Hi,

I was the one who wrote those nasty things about a fantastic instrument.

I bought such a counter a few weeks ago. When I first peeked into it
(because the fan was running at top speed after a few minutes) I
noticed, that the arrangement of the components was anything but optimal.

Examples: The osc is only one or two inches away from the linear voltage
regulators. These are "cooled" by a 1mm iron sheet. The iron sheet has
some cooling slots where the air can flow through, but these slots are
miles away from the regulators. The sheet gets so hot that you won't
touch it voluntarily. If at least the air could flow directly along the
regulators - but they sit in a corner and the air flows diagonally
through the case, avoiding contact with the regulators.

Since I use an external osc, that doesn't bother me too bad.

However, while trying to adjust the trigger circuits, I noticed, that
they are very sensitive to temperature variation. Unfortunately, the
warmed air (which is coming from the cooling slots) flows over these
parts... see picture.

The thermistor that controls the fan is located amidst the air flow -
thus resulting in a very nervous control loop. The fan is continously
speeding up and down. It's not to hard to imagine, that this has effect
on all parts where the air flows (-> Trigger...).

If I had to construct the thermal design for a 4000 Dollar counter I
would do totally different.

That's why I feel a little displeased.

If you own a counter, that you can't modify (due to calibration seals)
heat it up and do not switch it off. Let the cooling air flow by giving
the counter enough space around it. Stabilize room temperature and don't
place your counter near a door or a window or a place where air
temperature is often varying.

I'm shure, you will run a great instrument.

If I own a totally atypical counter and Time Nuts have different
experience please stop me from spreading rubbish.

Volker

Am 02.12.2012 21:12, schrieb Bob Camp:

Volker,

On 12/02/2012 11:36 PM, Volker Esper wrote:

I am also not very impressed with the location of the linear regulators.

This forms a nice interesting relationship between line voltage and room
temperature. As line voltage shifts, the rectified voltage varies, and
then the voltage that the linear regulator needs to "burn off" varies,
and thus their heat contribution varies with the line voltage. Now, the
ambient temperature then controls how easy we get rid of this excess
heat. The way the airflow goes, this cooling makes sure it couples very
well with the electronics, so fan speed amongst other things depend on
ambient temperature... and line voltage. All this assist to give us an
"interesting" temperature dependence.

In contrast, I was impressed by how the HP10811A was located in a
turbolence free corner of the HP5070A/B assembly. While itself also not
particularly well designed in heat context, at least the heat of the
linear regulators was supposed to mostly affect an external
self-convection stream. Forces convection (fan) to stabilize the
temperature of the heat sink would be a good thing there.

Cheers,
Magnus

Hi Paul,

1. If you are making frequency measurements, the warm-up of the internal oscillator is the major factor limiting accuracy. This doesn't mean you have wait 12 hours. For example, if all you need is 6 digits of precision, a one-minute warm-up may be sufficient. For each further digit of precision you wait longer. You can probably get 9 digits with 1/2 hour of warm-up. It depends on the oscillator. Plotting digits of precision as a function of warm-up time would make a very educational graph you could tape to the top of your SR620.

2. If you are making time interval measurements and using an external standard, the warm-up time will also affect the accuracy of your TI measurements, but to a far lesser degree. Here are informal results for TI (time interval) mode after a 5 minute power-down (see also attached plots):

• if you need 1 ns accuracy, you can use the SR620 immediately after power-up
• if you need 100 ps accuracy, wait 2+ minutes
• if you need 10 ps accuracy, wait 15+ minutes
• if you need 1 ps accuracy, you need a seriously stable lab environment or a different counter.

Given that you plan to use the SR620 with high-end GPS gear I would suggest you try this quick experiment for yourself to see what your SR620 does, with your inputs, in your environment. Your numbers will come out different than mine; but the methodology is the same. Your procedures can then be based on measurement and confidence instead of guesswork and folklore.

Note also if your data collection is automated, there's really no reason to wait after power-up at all. Just collect data as soon as you can and skip the predetermined number of samples. I can send you the SR620 GPIB scripts I used for my test. This way you have a record of the warm-up settling time itself, which further gives you confidence in the data that follows.

/tvb

Tom,

I agree. Since Paul want's to use an SR620 I presume he needs precision.
Otherwise almost any TIC with a fairly stable osc would do, for example
one with a battery backup. So I further presume that he needs nearly the
full accuracy / stability. But that's just speculation, surely Paul can
answere this question?

Volker

Am 04.12.2012 14:11, schrieb Tom Van Baak:

I would be interested in those scripts as well. as soon as the replacement
TCXO for my sr620 arrives I can put it back together and this would be fun
to try.

-eric

On Tue, Dec 4, 2012 at 5:11 AM, Tom Van Baak tvb@leapsecond.com wrote:

--
--Eric

Eric Garner

On Tuesday, 4 December 2012, Tom Van Baak wrote:

Tom & Co.,

Thank you!  These plots are excellent and will be very helpful. 

You are quite right; we should do the test ourselves.  We will definitely
do that.  Obviously, there is not need to worry, as we can characterize
the instrument behavior ourselves, which is probably necessary anyway if
we're going to publish these measurements with error values.

Many Thanks,
Paul

--
Paul DeStefano

Volker,

That's a great question and I'm afraid I don't have a good answer 

for you.  If pressed, I would estimate less than 100ps.
The error of this measurement contributes to the error in our
final measurement which has many components.  I haven't worked out an
error budget for each contributor.  Our goal is a final error of less than
10ns.  But a few contributors are already expected to be around 1ns, so
this error really needs to be less than 1ns, preferably less than 100ps.

Cheers,
Paul

On Tuesday, 4 December 2012, Volker Esper wrote:

--
Paul DeStefano

I'm curious, which way you went and which accuracy you achieved... :-)
Can you tell us?
Volker

Am 06.12.2012 19:10, schrieb Paul DeStefano: