[time-nuts] Lady Heather's Tbolt oscillator auto-tune function
kb8tq at n1k.org
Sun Sep 11 09:12:31 EDT 2016
> On Sep 11, 2016, at 3:35 AM, Charles Steinmetz <csteinmetz at yandex.com> wrote:
> Mark wrote:
>> Here's a little info on Lady Heather's oscillator autotune function for the Thunderbolt GPSDO
> Thanks, Mark, that is very helpful.
> Accordingly, for people interested in best frequency accuracy and stability, I suggest (1) running autotune, then (2) manually setting damping to around 10, and (3) setting the holdover recovery parameters manually as I described in my last post (jam synch at 65-75nS and allow a large frequency error in recovery mode). The value autotune sets for loop time constant (500 seconds) is a good starting point for Tbolts with the Trimble p/n 37265 OCXO. A particular Tbolt may respond to further optimization by tweaking the TC to match the individual OCXO (by trial and error -- Plot ADEV, adjust TC, plot ADEV, adjust TC, etc.), but the improvement will most likely be subtle. You should also review the elevation mask settings and adjust if necessary.
>> Since the unit should be locked and stable, the current DAC setting is where the oscillator is at 10.0000000 MHz and will be set in EEPROM as the initial DAC setting. The tbolt uses this value to speed up locking the oscillator when powering up.
> Actually, for best results from a cold start, it is best to set the initial DAC voltage to whatever voltage produces 10.000000000 MHz *when the oven is cold*.
…… errr … that’s pretty far off :) DAC at 10 or 20 minutes is probably the target.
> Setting it to the voltage that produces 10.000000000 MHz when the oven is warm guarantees that it is set wrong when the oven is cold, so on a cold start the loop immediately goes into saturation to slew the DAC to the voltage that does produce 10.000000000 MHz. This slows down locking by quite a bit. Setting the initial DAC voltage so that the frequency is 10.000000000 MHz with a cold oven allows the loop to slowly adjust the DAC as the oven warms up, rather than racing off at full speed to meet the warming crystal. This speeds up locking from a cold start very substantially.
> It is not clear to me how one could automate this process -- I found the correct DAC settings by trial and error. Ideally, this would be determined every time the Tbolt does a cold start and the new value would take into account any crystal drift since it was last set. But LH has no way to tell that any given start is a cold start (AFAIK), and it would want to determine the correct cold-oven DAC voltage very quickly after power-up from a known cold start (say, within 10 seconds). It would be great if Tbolts had an oven on/off command, but to my knowledge they don't. LH could possibly give the user instructions ("OK, now power down the Tbolt for at least 30 minutes. When the Tbolt is fully cold, click the RESUME button below and then re-apply power to the Tbolt...”).
…. It’s a little worse than that :)
Temperature does indeed get into the act, what was the temperature when the DAC setting was recorded, what is it now, what is the frequency vs temperature on that assembly? One could do a temperature run (or series of runs) and save the data. There are OCXO’s done that way.
Exactly how long has the OCXO been off? There is a difference between 3 hours and 5 hours. Similarly there is a difference between one day and a week.
What temperature (and humidity and …) has the assembly been at while off? Temperature will impact the sealed OCXO. The parts on the board most certainly are sensitive to humidity.
Did something weird happen? Was the unit rotated? (2g tip does matter). Was it dropped? Was it shaken around? There are a number of weird things that *could* happen. Some OCXO’s get acceleration compensated to take care of this.
Is it the DAC at 5, 10, 15 or 60 minutes? The OCXO is moving pretty fast right after it turns on. For instance, the FE 405’s compensate for this in firmware.
That’s a short list, one could easily double it in length.
The *simple* answer is to use software. Rather than having the TBolt in one mode all the time, change it’s parameters when it first turns on. Give it a shorter TC and get it stabilized. Then feed it the “real” TC and let it go from there. I have absolutely no idea how a TBolt responds when you do this. I do have a lot of data on how other vendors parts respond when this takes place. In general it works pretty well.
> Best regards,
> 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.
More information about the time-nuts