All the magic happens within this inner enclosure:
Sorry, I did not open the physics package, too afraid to break something. Anyway, the manual describes the operation priciple quite well, so I won't have to go into details here.
This thing is clearly made to last, and to run for prolonged time at elevated temperatures. If you inspect the photos carefully, you'll find IC's in ceramic or hermetic TO package, with extended temperature range (up to 125 degrees Celsius) specifications. This is quite necessary, since the innards of the physics package will be heated to temperatures in the 70 to 100 deg.C range.
There's no electrolytic capacitor that could dry over time, none of the modern solid tantalums that sometimes burst into flames. Looks like one wet slug tantalum is found on the power supply board, right below the heat sinked output transistor. Otherwise, lots of high reliability ceramic caps, and a rather large film capacitor. So the designers did a good job to avoid components that are known to fail at such operation conditions.
Discoloration of the foam attached to the crystal oven and on the top flex board proves this unit was long time operational at elevated temperatures.
Anyway, in my opinion, just the electronics alone is a piece of art.
This behaviour quickly changed to a permanent fail, the unit heats up, locks after 2 ... 3 minutes, wait maybe five more minutes and the lock is lost again.
As this unit was sold as tested and working, I've contacted the seller, told him about the faulty unit. Not a big deal, he said, you'll get a replacement unit, apparently this particular unit is a case of "shit happens".
This unit isn't the first thing I bought from him, he's always helpful and provides accurate information regarding the state of the equipment he sells. Since it's used equipment, one never can be sure there aren't any hidden issues left, but he's always helpful to resolve this. So I started to repair the unit instead of replacing it.
My first suspect was the rubidium lamp beeing worn out, since the lamp monitor voltage was rather low. See page B-3 in the manual, this states a range from 14 to 9 volts for a working unit, and says the units are expected to work with a lamp voltage as low as 6 volts. My particular unit was a the lower end, well below 9 volts, but still above 6 volts. So I wasn't assured if the lamp would be still OK or not.
To check if the lamp is still usable, one would have to disassemble the physics package and watch the light output for a purple haze according to this paper.
Anyway, checking the schematics revealed the lamp voltage output isn't derived from the lamp driver circuitry but rather from the photo detector on the other end of the physics package. So the fact that a voltage appears at all here implies the lamp must emit light at least. If that light contained the required spectral line by the vapourised rubidium - I can't determine this from the voltage here alone.
Otherwise, the unit does lock for a few minutes after power on, so the spectral line must have been there, and why should it disappear again? Operation of the physics package depends on the correct temperatures within, the paper mentioned above advices to check them. You'd have to disassemble the physics package to do so. Otherwise, the power consumption of the unit looks quite as expected, starting at 1.7 amps current draw cold and settling to 0.5 amps when fully heated up, so I suspected the heaters to be OK.
The paper has a checklist. So I did follow these instructions.
1. Check injected RF:
The scope picture shows the waveform after I'd adjusted the AGC voltage to 3.38V according to the checklist, using the right hand side (square shaped) pot. I don't remember what the waveform looked before, but the AGC set voltage (at the R25/R27 junction) was at 4.1 volts before I did the adjustment.
2. Check SRD (step recovery diode): OK (please don't expect to measure exact 0.745V as stated in the paper).
3. Check AGC and other voltages: Both mentioned (3.38V) voltages were at about 4.1V for my unit.
So I adjusted the set voltage to the mentioned 3.38 volts, using the square shaped pot (see left hand side picture for the location of the R25 / R27 junction). The other voltage, fed back from the SRD, followed that setting. It's neither written in the manual, nor in the paper, but the other (round shaped) pot places a resistive load onto the rectified voltage from the SRD, adjusting this pot has some influence on the voltage one can read at the R30 / R22 junction. So I adjusted this pot also a bit, trying to bring the voltage readings as close as possible to the proposed 3.38V.
While doing so, I noticed the unit locked again. Bingo!
The paper describes this as an indicator for good Rb light, but it is also an indicator for the proper adjustment of the injected RF amplitude. This signal's amplitude varies with the setting of the above mentioned pots within the AGC loop. Observing the waveform, I found out the SRD's operating point apparently drifted away from its inital setting. Maybe R24's or R30's wiper suffered from time and elevated temperature operation, leading to a drift in the SRD bias current, in turn leading to not enough output signal from the demodulator to achieve a stable lock. Maybe some other component drifted away, can't tell that. For now, I've made it working again by repeatedly adjusting R24 and R30 to find the operation point of the SRD. Maybe one wants to replace these pots, maybe these are cured now.
I didn't find any detailed instructions of how to adjust these pots, so I did some experimenting. The sine wave amplitude at servo board's TP1 can be adjusted well above 3.5Vpp. From the manual's illustration I took the goal of 2Vpp amplitude around 5V. The DC offset is rather constant and not influenced by this adjustment, so it's a don't care as long as it is about 5 volts.
First, one has to get the unit to lock, this adjustment doesn't work as long as the unit isn't locked. As a starting poing, set R24 to a rough 3.38 volts reading for the set point (left picture above). Watch for the unit to lock, the demodulator signal should appear. If it doesn't, try adjusting R30 for more bias current and wait for the unit to lock again. Now adjust R30 for a rather low amplitude (about 1Vpp), and find a maximum amplitude point by adjusting R24. This should end up near the mentioned 3.38V. Now adjust R30 again for a 2Vpp demodulated signal amplitude, and you're done. Watch the lock state of the unit while adjusting, since the observed signal disappears in the unlocked state.
Note, this procedure is experimental, the manufacturers procedure might be different, there's a little hint in the manual,
having an other starting point.
This is the Efratom FRS sitting on the bench while beeing monitored:
So at first I suspected the pots had gone drifty again. Did a re-adjust and replaced them by fixed resistors (circled green).
Worked for a short amount of time, then the same symptoms appeared again. While tinkering and testing, I've noticed one of them was out of adjustment range. This could be fixed by changing the resistor that was intended for this purpose. Apparently there're quite a few components that are intended to be selected while test.
Anyway, this didn't fix the main issue (loosing lock and drifting). Some more testing showed the sync loop loosing demodulator signal amplitude when the units reach their final thermal equilibrium. By lucky chance, I've noticed the amplitude changing while probing the 60MHz output (marked ASP1) with a 10:1 scope probe. Interestingly there's a space for a capacitor C30 "SIT" (select in test) unpopulated. Adding a capacitor (one unit took 10pF, the other one 18pF) significantly increased the demodulator signal amplitude.
To find this issue, I had to assemble the units (except the outer cover) and let them warm up while monitoring the 60MHz output at "ASP1", and the 125Hz servo loop signal as described above. While the unit was heating up and reached its final thermal equilibrium, the 125Hz signal was fading away until lost lock while the 60MHz waveform changed its shape significantly.
So I soldered these capacitors (circled red in the above schematic) into place, and the units are locking and outputting a stable frequency again. 125Hz signal stayed stable and at a healthy amplitude. I haven't noticed any more issues during some frequency drift measurements, so I'm assured they're finally fixed.
It's an interesting failure mode as apparently something in the guts of the physics package changes it's impedance (capacitance) while the unit is heating up. There's not a lot of components involved here, basically a feedthrough capacitor, a resistor and the SRD (step recovery diode) that produces the harmonics to energize the microwave cavity resonator. The SRD needs a certain amount of drive level to operate correctly. The drive level is provided by the output network of Q5 that most probably gets de-tuned by the change of capacitance of (presumably) the feedthrough capacitor. I can't figure any other cause for this de-tuning. The added capacitancy at C30 fixes this. Don't know if this is a permanent fix, as I didn't dig into the physics package to check the mentioned components. One might assume there's some ageing happing to the feedthrough capacitor that hasn't ended. Time will tell.