Hi

There really is no practical way to get an accurate temperature sensor inside the oven / on the crystal, without upsetting the operation of the device.

You really do need the “magic” setting process that loads and saves a number into the PIC.

Bob

Well, tweaking all the PIID parameters in turn feels like an NP-complete problem so I am certainly not going to try that.

I hope, and that is mostly founded on wishful thinking I might add, is that you can change the temperature setpoint without changing all the other parameters. Rationale being that the control loop parameters for the best part are determined by the physics of the oven. Which are, I would assume, for reaons of manufacturability "the same" for all ovens manufactured. The variability appears, to me at least, primarily in tbe crystal. Resulting in a unique temperature setting for each individual crystal.

Well... that is my interpretation of the design papers, I could very well be wrong of course.

Rick, if you could be so kind, can you please shed your light on this subject?

cheers,
Wilko

Well, the three thermistors are an accurate way to measure the temperature, otherwise the PIID ovencontroller would be a non-starter.

A good correlation between whatever the thermistors measure and the temperature on the crystal plane, I imagine that can be dealt with during the oven design process.
So, measured thermistor voltage is X, that equates to Y degrees centigrade.

Wilko

Hi

The suggestion was to use a thermometer to measure the temperature of the oven. Adding a thermometer and it’s probe to the setup was what was being talked about.

Bob

Hi

One very common way to “calibrate” an SC based design is to switch the oscillator from the normal C mode over to the B mode. This gives you a frequency that rapidly changes with crystal temperature. The advantage is that you are looking directly at the properties of the crystal blank. There is no guessing about offsets or gradients.

If you are using precision thermistors, a “calibration” run would be something done in the design phase. Once the offsets between the crystal and thermistors are known, there is no need to do it again. In a typical circuit, you need to make parts changes to get it on the B mode. You also need to calibrate the crystal (over in “crystal land”) to know what the frequency to temperature data looks like. All of this makes it a bit difficult to use in the field.

Bob

On Thursday, June 26, 2025 3:47:48 PM Central Daylight Time Bob Camp via time-
nuts wrote:

Do you really need to measure the temperature independently?  I don't fully
understand the control scheme and the differences between C mode and B mode,
but if you get the oven to the right temperature, will that not set the
oscillator output back to within spec?  I recall the original description was
the measurement was with EFC connected to 0V, so I guess you would need to
know what the nominal value of the EFC should be for on-target frequency.
Am I thinking too simplistically?  Is there a range of oven temperatures where
the output is close to nominal but the temperature coefficient is not optimal?

--
Chris Caudle

Hi

If you want to map “some reading” on the thermistor to 103.0 C on the crystal blank, you
need a way to know the crystal blank temperature. You do a sweep run of some sort. You
log the thermistor bridge readings vs frequency. You then convert this to temperature based
on the data on that specific crystal.

The B mode is “linear” over temperature. It has a nice steep slope of frequency vs temperature.
This makes it ideal for checking oven temperature. Since you need a calibrated crystal and
circuit changes, there are practical limits using it.

One would assume this was done “way back when” on the HP design. The data probably got
saved somewhere in the PIC code. It may well have been updated when thermistor vendors
changed.

If you want to use the labeled temperature on the crystal to set the oven, you need that
mapping. Why do it this way? It could save you a lot of “fun” in production.

Long long ago folks would do a “turn hunt” on the crystal. They then would ship the part.
Eventually they found that you actually needed to do that turn hunt at multiple temperatures
due to a range of issues. Every OCXO design is different and exactly how this or that one is
processed will be a “that depends” sort of thing.

Bob

hello Chris,

You are correct, the EFC is shorted right now. So, if there is a nominal EFC voltage !=0 V  required to obtain 10 MHz this explains why my oscillator is off-frequency.

In case somebody on the list can measure the EFC voltage with the oven being GPS disciplined we can get this clarified Inthink.

I unfortunately do not have the GPSD board(s), I just have the O of the whole GPSDO system. In case someone has a GPSD board available for the good cause I would be interested, for obvious reasons.

cheers,
Wilko

Hi,

"Fun in production", yes, exactly.

In the mists of time there was likely a manufacturing process/procedure that characterised the xtals as part of the oven assembly production. That resulted in the "x degree C" sticker on the oven.
The main PCB got married to an oven, and the PIC on that main PCB was duly programmed with the temperature indicated on the oven label. Doing it like this once could produce oven assy and main PCB assy in fully independent  production lines.

In a repair process a similar thing might have happened: new oven assy mounted, main PCB setting updated based on label of new oven assy. Etc.

All speculation, but this is roughly what I have seen in production. i.e. when I was in hardware design (when dinosaurs roamed the planet, and all that...)

Cheers,
Wilko

NB: I updated the web page with my latest progress.

-----Original Message-----
From: Bob Camp via time-nuts time-nuts@lists.febo.com
Sent: 1 July, 2025 15:31
To: chris@chriscaudle.org; Discussion of precise time and frequency measurement time-nuts@lists.febo.com
Cc: Bob Camp kb8tq@n1k.org
Subject: [time-nuts] Re: repair of an HP E1938A OCXO

Hi

If you want to map “some reading” on the thermistor to 103.0 C on the crystal blank, you need a way to know the crystal blank temperature. You do a sweep run of some sort. You log the thermistor bridge readings vs frequency. You then convert this to temperature based on the data on that specific crystal.

The B mode is “linear” over temperature. It has a nice steep slope of frequency vs temperature.
This makes it ideal for checking oven temperature. Since you need a calibrated crystal and circuit changes, there are practical limits using it.

One would assume this was done “way back when” on the HP design. The data probably got saved somewhere in the PIC code. It may well have been updated when thermistor vendors changed.

If you want to use the labeled temperature on the crystal to set the oven, you need that mapping. Why do it this way? It could save you a lot of “fun” in production.

Long long ago folks would do a “turn hunt” on the crystal. They then would ship the part.
Eventually they found that you actually needed to do that turn hunt at multiple temperatures due to a range of issues. Every OCXO design is different and exactly how this or that one is processed will be a “that depends” sort of thing.

Bob

time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe send an email to time-nuts-leave@lists.febo.com