| Electronics > Metrology |
| GNSS RB (chinese) |
| << < (11/11) |
| Johnny B Good:
I see there is still some interest in the use of rubidium frequency standards from the number of views all those images I'd attached to each post had gotten. >:D I have a few points to make, the first being that, in view of the exquisite sensitivity of the crystal oscillator to mechanical excitation demonstrated by all of these LPROs (and even the SA22.C according to another poster - I've yet to test this on my own unit), it's not a good idea to rigidly attach a cooling fan to a heatsink that's also firmly mounted onto a heatspreader to which the rubidium oscillator is in firm contact with. ::) The effect of random vibrations from the cooling fan may not be obvious in a scope trace (although likely to raise close in sideband noise that can be observed with a spectrum analyser) until you improve the cooling algorithm used to adjust the fan speed with so much precision and stability as to achieve base plate temperature regulation to within +/- 1mK of the set temperature. In this case, it then becomes entirely possible to phase modulate the crystal oscillator at precisely the wrong frequency for seconds long periods where it can then destabilise the corrective action by the physics package (143Hz adjustment rate AFAICR) leading to the crazy wobble of instability I'd observed. In hindsight, I'd been catching glimpses of this effect during the progression of my project all along. It was just that initially they had been so brief it had been a case of rubbing my eyes to check that it wasn't an optical illusion of some sort that I'd seen. Of course, over the now four year long project's development, this effect became ever more obvious to the point about a year back, I could no longer pass it off as an 'optical illusion'. In perfecting my temperature control algorithm to the point where it could maintain the temperature of the base plate to within less than a milli-Kelvin from the set point, I had become "Hoist by my own petard". :o All it took to point me to the fan as the culprit was the use of those replacement LPRO 101s I'd acquired. They'd shown no sign of this issue when parked on top of my rubidium frequency standard as a convenient temporary heatsink for their initial testing so it was a bit of a shock to see them behaving exactly the same as my original LPRO once mounted into my rubidium reference to allow me to check their lamp voltage. This left me with only one conclusion, namely that the fan cooling was the source of my mystery wobble in some way, leaving me to ponder just exactly what that mechanism might be. The most bleedin' obvious in hindsight cause being the last possibility I'd considered after disproving electrical noise imposed on the supply by the fan's commutation pulses by fitting a 16v 10mF capacitor across the fan connection to the driver board and magnetic interference from the unshielded 5.6mH filter choke or leakage from the fan's rotor by using a powerful HDD voice coil actuator magnet with a 2mm thick perspex shim to avoid the risk of saturating the mu-metal shield and magging it up, where only by careful placement, could I detect any effect on the ruby's output frequency (and that, despite the perspex shim where the magnet was still quite strongly attracted to the mu-metal cover. Only after discounting those two possibilities did I finally (like Sherlock Holmes) accept the one and only remaining possibility and removed the fan retaining screws for a final test did I finally solved the mystery. The aperture in the polystyrene foam plenum divider through which the fan poked, supported the fan in place to allow such a quick temporary test. The fan was a snug enough fit to hold it in place but I wrapped a couple of turns of electrical insulating tape around it afterwards to ensure a tighter friction fit to make it a more permanent fix. The second point being in regard of those A23A3 adapted LPRO-101 versions, namely which market segment they're targeting. It can't be the telecoms market since there's no means of disciplining it from a gpsdo via the external C field pin nor any means to monitor its health status via the lamp and crystal voltage monitoring pins, just an indication of atomic lock (BITE) which begs the question as to who would be satisfied with such a frequency reference oscillator? Oh hang about! :palm: I'd completely forgotten about the Audiophool market! Could that be the target market demographic do you think? Now that I've remembered the phrase, "An audiophool and his money are soon parted" the whole over-hyped description and premium pricing certainly fit the bill (although in this case, the pricing probably undercuts the more regular purveyors of such audio snake oil). Whilst writing this post, I noticed an upturn in the drift rate between the rubidium and the gpsdo so decided to wait for the full 24 hour run to complete before posting the final two screenshots. This seems to mimic previous behavior which I now rather think is due to an ageing rate of circa -2μHz per day calculated from the 10 year ageing data in the Datum spec sheet for the LPRO. I'd suspected as much with the MK II gpsdo results but that 33ns diurnal phase wobble had made it rather difficult to pick up on. The upgraded MK III with its ZED9-FT gps module (also a G3RUH type) has made this effect rather more obvious. I guess it won't be too long before I add the RTC module to the electronic baggage and rustle up another bunch of code to mitigate this ageing issue. Anyway, here's the last two screenshots, timed at 18:20 and 19:00 this evening. |
| Johnny B Good:
Here's another six screenshots, taken over the past four hours from 12:12 this afternoon. The first one shows the accumulated phase drift over the previous 24 hours. The second one was recorded within half a minute after clearing the persistence to start yet another 24 hour run. I'd adjusted the LPRO's frequency by +3μHz at the start of the previous run and by +2μHz at the start of today's run which was under heavy overcast conditions which had cleared up by the time of the fourth screenshot. It was the first time in over a week that I was able to see a clear blue sky and contrails of passing jet liners. Indeed, also my first view of a glorious red sunset in over a week! The screenshots seem to reflect that change in cloud cover. The extra propagation delay on the gps signals due to the earlier cloud cover had been reduced by some 2 or 3 ns, hence the apparent 2 ns 'backward' step on the rubidium trace. PS All images had been taken at hourly intervals (12 minutes past the hour) except for the fourth image which had been taken between those hourly images at 14:36 when I'd noticed the disappearance of the cloud cover. |
| Johnny B Good:
For anyone interested, I've just posted an update on my RFS project here: https://www.eevblog.com/forum/rf-microwave/gpsdo-phase-between-pps-and-clock/msg5768749/#msg5768749 Included in the attachments is a 12 hour movie condensed into a 2 minute timelapse showing ZED9FT and M8T based G3RUH type gpsdos along with the RFS and a spare LPRO mounted onto a passive heatsink with, naturally, the ZED9FT triggering the display as described. |
| bingo600:
Hi John Thanx for the "Tant pics" :-+ And nice trackdown of the "Fan vibrarion" issue :clap: |
| Johnny B Good:
--- Quote from: bingo600 on January 02, 2025, 09:02:47 pm ---Hi John Thanx for the "Tant pics" :-+ And nice trackdown of the "Fan vibrarion" issue :clap: --- End quote --- You're welcome. That repair turned into a bit of an epic adventure and in the end, resulted in no discernible improvement, just a bit of grief in almost totally destroying the replacement trimpot I'd fitted less than 12 months ago. The only interesting finding being that the original tantalum caps had remained in a remarkably good state of health despite the generally accepted 'wisdumb' that they'd be totally shagged after a decade's worth of service life. I've put that original unit to one side until I can remedy my previous (in hindsight) ill advised adjustments to the lamp exciter oscillator tuning to increase the lamp voltage from its original 4.9ish volts to a more reasonable 5.3ish volts. It does seem to be less stable than the replacements when it comes to percussive testing. Without the fan induced vibrations, I'd expect it to recover from a percussively induced loss of atomic lock without any need to power cycle it. I can't recall whether or not I'd tested this, so that remains something for me to explicitly test for when I do finally get around to it. BTW, what did you think of my timelapse movie 'footage'? |
| Navigation |
| Message Index |
| Previous page |