I've noted the interest in those photos and screenshots I'd attached to my previous post, so to satisfy everyone's curiosity, here's some more of the same (at the end of this post).
My GPSDRO project is still a work in progress but I
have made some (progress, that is!). I've modified that frankensketch into more a work of my own. I've added ambient temperature cooling compensation to reduce overshoot (and the consequential undershoot), modified the temperature control algorithm so it now detects when the temperature starts responding to the fan's cooling effect and shuts it off without having to wait for the temperature to drop below the set threshold as it did in the original sketch which mitigates the undershoot effect considerably.
The temperature now varies over a much smaller range, circa 0.15 *C typically, versus the 0.35 *C of the original frankensketch, still a little dependent on the ambient temperature (ambient compensation isn't perfectly tuned right now) and it still spends a little more time below than above the set point. There's little point in fine tuning everything anyway at this stage until it's finally boxed up in a properly ducted and insulated enclosure.
I replaced that bi-colour oled with a single colour version (I'd purchased two white oleds from banggood). Unfortunately, I blew the rice grain sized 3.3v LDO on the first one I'd tried due to ASS-U-ME effect (I'd made the mistake of assuming it was a direct drop in replacement without verifying that the Vcc and GND connection placements matched
).
Luckily, said LDO had 'thrown itself upon this particular hand grenade' that I'd thrown at it and I was able to (snugly) fit a slightly larger one (extracted from yet another piece of my collection of salvaged parts) in its place. This now displays both ambient and baseplate temperatures as well as absolute barometric pressure.
Despite the improved temperature control, I was still seeing an increasing need to calibrate its frequency with the heli-pot. The problem proving to be down to the internal pot having been 'maladjusted' close to its (normally unused) upper frequency end of its range to allow me to shift the external tuning voltage from circa 2.5v down to circa 250mV to attenuate the temperature drift of the AMS1117-5.0 used in the YRobot breadboard psu module by a factor of 10 to 1. It's a neat trick, but only if the internal pot doesn't get upset by this adjustment to 'parts exotic' of its resistance track. I'd had to 'exercise it' around this new setting to polish that bit of the track clean before I could set the external tuning voltage to 270mV (middle of the new zero to 515mV range that I'd padded the heli-pot out to) and re-syntonise the LPRO.
The high drift rate problem now appears to been eliminated but it's early days. When you're chasing this level of stability, it needs more than just 24 hours for everything to settle down after even a brief power interruption. I'd been powering the 12v CPU cooling fan from my bench supply to test it at 18.8v to extend the lower usable PWM control range downwards from 22% on 12v to just 14% at a 17v test minimum to verify it would still start reliably using a one diode volt drop less than the 19v powering the LPRO. The psu count is now back down to just two (19v laptop charger and a 12v wallwart powering the breadboard in the absence of a usb connection to the PC).
Ultimately, the whole GPSDRO will be powered from a couple of redundant 19 volt laptop charging bricks powering a 7808 powering a 7805, both thermally bonded to the baseplate for temperature stability to minimise the effect of the 7805's voltage tempco on the Vcc rail.
Having an extra DC input socket for a second 19v PSU will come in handy for testing in low ambient temperature conditions (basement or outside on a dry winter's day) since I can plug the redundant laptop charger into a very long mains extension lead plugged into a downstairs socket, allowing me to physically move it from my 1st floor hobby lab without interrupting power.
However, from previous experience, a simple swap over from a lab mains socket to a very long mains extension is unlikely to disturb its hard won stability if done in just a few seconds long time frame. The LPRO can recover lock after a few seconds of power loss within just a few seconds of restoration of power. My GPSDO (a modern day version of the G3RUH design), otoh, can take minutes to stabilise even after the briefest of interruptions. Notably though, once it has stabilised, the phase relationship it had beforehand with the LPRO is always resumed, seemingly exactly from where it had left off.
Considering that the PPS has been programmed to output 100KPPS to phase lock at 100th the OCXO's frequency, My instinctive expectation had been to see a random phase displacement after such a disruptive event. I mean, any one out of the individual 100KPPS is as good as any other for the PLL to phase lock at any one of the 100 cycles of the 10MHz it takes to make a single full cycle at 100KHz. Perhaps I'm simply failing to see the wood for the trees to figure out this conundrum.
Looking at that sequence of screenshots over a 10 hour period, it seems to be in need of more settling time and some more tweaking of the heli-pot to syntonise it to the GPSDO.
PS I've just realised that the time labels on the screenshots could be mistaken for run time. They're just the time of day they were grabbed via the web control panel I've only just reacquainted myself with - no dynamically updated screen image feature, just a Refresh Screen button - probably why I left this interface option to lie unused these past couple of years.
That first screengrab was made about half an hour after I'd lined up the traces for a casual test run (honestly? I'd simply forgot to take a screengrab
)