Repair : TEMEX/SpectraTime LPFRS/LPRO 10MHz Rb generator [got GPSDO to cal]

[TheSteve]

Changed the caps on my first one. It proved much more difficult than anticipated because my desoldering tool was clogged up and I was in a hurry.

Put it back together with the heat sink back to front so I could get at the connectors and powered it up. The output was the correct level (0.5Vrms into 50 ohms) and the frequency started low and then went to something that seemed right. It measured a steady 10.00000050MHz on my counter - the error is in my counter I think but I'll need to get two going to compare and also compare to a GPS - perhaps at the weekend.

Altogether pleased so far.

Sounding pretty positive. If it isn't working the frequency error is generally enough to know its a problem. Have you connected a terminal/serial port to it yet, if not I recommend it so you can see the values.

[Orange]

Let it run for a few days, and check again. I had one that seemed OK during the first 12 hours, but then it lost lock, and stayed at 10.000130
The strange thing is that the lock indicator of this faulty one still said to be locked !!

It is the only one that is not OK, so 4 out of 5. Not bad.

In the mean time I have replaced the 85deg. caps for 105deg radial Panasonic FR caps (100uF/50Volts).

Changed the caps on my first one. It proved much more difficult than anticipated because my desoldering tool was clogged up and I was in a hurry.

Put it back together with the heat sink back to front so I could get at the connectors and powered it up. The output was the correct level (0.5Vrms into 50 ohms) and the frequency started low and then went to something that seemed right. It measured a steady 10.00000050MHz on my counter - the error is in my counter I think but I'll need to get two going to compare and also compare to a GPS - perhaps at the weekend.

Altogether pleased so far.

[jpb]

Sounding pretty positive. If it isn't working the frequency error is generally enough to know its a problem. Have you connected a terminal/serial port to it yet, if not I recommend it so you can see the values.

I plan to do so as soon as I get the time - it might not be until the weekend.

Let it run for a few days, and check again. I had one that seemed OK during the first 12 hours, but then it lost lock, and stayed at 10.000130
The strange thing is that the lock indicator of this faulty one still said to be locked !!

It is the only one that is not OK, so 4 out of 5. Not bad.

In the mean time I have replaced the 85deg. caps for 105deg radial Panasonic FR caps (100uF/50Volts).

That is a little worrying, I'll leave it running for longer.


I realise since posting that the frequency will be a little out as I've directly connected to the unit not via the little board which has a pot on it presumably adjusted to give an accurate 10MHz. My main concern is that it is steady. I'll worry about exact frequency when I get a chance to use one of my GPS units (I'm in rented accommodation whilst trying to move house so I don't have access to a lot of my stuff and can't really stick aerials up - though I can probably dangle one out the window.

[Orange]

Sounding pretty positive. If it isn't working the frequency error is generally enough to know its a problem. Have you connected a terminal/serial port to it yet, if not I recommend it so you can see the values.

I plan to do so as soon as I get the time - it might not be until the weekend.

Let it run for a few days, and check again. I had one that seemed OK during the first 12 hours, but then it lost lock, and stayed at 10.000130
The strange thing is that the lock indicator of this faulty one still said to be locked !!

It is the only one that is not OK, so 4 out of 5. Not bad.

In the mean time I have replaced the 85deg. caps for 105deg radial Panasonic FR caps (100uF/50Volts).

That is a little worrying, I'll leave it running for longer.


I realise since posting that the frequency will be a little out as I've directly connected to the unit not via the little board which has a pot on it presumably adjusted to give an accurate 10MHz. My main concern is that it is steady. I'll worry about exact frequency when I get a chance to use one of my GPS units (I'm in rented accommodation whilst trying to move house so I don't have access to a lot of my stuff and can't really stick aerials up - though I can probably dangle one out the window.

Yes its really helpful if you have an accurate time-base. I use a rubidium time-base to my counter (a HP5384A and a HP53132A), which is calibrated against a GPSDO (Z3805). So any discrepancy in the milli Hertz domain is notified. The LPFRS has +/- 1 miili Hertz accuracy (10 seconds)

[jpb]

After a bit of struggle, I've got my serial interface working - I had to reprogram my USB-RS232 module to invert the polarity and it still wouldn't work at all via PowerShell (which I use for my counter) so I had to resort to Putty.

Here is a selection of warmed-up output:

8D 50 BD B5 80 8C A4 DC
8D 50 BC B5 80 8D A4 DC
8D 51 BD B5 80 8D A4 DC
8D 50 BD B5 80 8D A4 DC
8D 4F BC B5 80 8D A4 DB
8D 51 BC B5 80 8D A4 DD
8D 4E BD B5 80 8C A4 DC
8D 4E BD B5 80 8C A4 DB
8D 4E BD B5 80 8C A4 DC
8D 4F BD B5 80 8C A4 DC
8D 50 BD B5 80 8E A4 DC
8D 38 BC B5 80 8D A4 DB
8D 4F BC B5 80 8D A4 DC
8D 4E BC B5 80 8D A4 DC
8D 51 BD B5 80 8D A4 DC
8D 4F BC B5 80 8D A4 DB
8D 38 BC B5 80 8C A4 DC
8D 51 BC B5 80 8D A4 DC
8D 50 BD B5 80 8D A4 DC
8D 50 BC B5 80 8D A4 DC
8D 4E BC B5 80 8D A4 DD
8D 50 BD B5 80 8D A4 DD
8D 38 BC B5 80 8C A4 DC

So the photo-cell DC voltage is 5V*(0xFF-0x8D)/256 = 2.23V which is in the range 2.0 to 3.5 but rather on the low side for my liking.

The Rb signal level is dropping to 0x38 which is 1.1V which is in the range 1 to 3.3V but again rather low. It does cycle though in the range 0x38 to 0x51 or (1.1V to 1.6V).

The VCXO voltage of 0xB5 is 3.53V so is just beyond the 3.5V upper limit.

The Frequency Adjust voltage is 0x80 because it is currently not connected and I reset the unit.

The Rb lamp heating limiting current  is 500mA*(0xFF-0x8C)/256 = 225mA and is well between 1A and E6.

The Rb cell heating limiting current is 500mA*(0xFF-0xA4)/256 = 178mA and is well between 1A and E6

The total current the unit is drawing at present is 438mA which seems consistent with 225+178+something for the rest of the electronics.

Finally the 90MHz power signal is fairly stable around 0xDC which is 4.3V which is just in the range 2 to 4.5V.

So my conclusion is the unit is functioning but probably the lamp is old and perhaps the crystal has shifted in frequency.

I have read about people rejuvinating lamps with a hot air gun - if I feel adventurous perhaps I'll try this at some point - meanwhile I've got another 4 units to sort out!

PS: my unit takes almost exactly 5 minutes to lock (I timed it at 4:55). How does this compare with your units everyone? It seems about what I'd expect.

The current draw on warm up is 1.3A which is rather high (the data sheet has 0.9A). 1.3A corresponds to 31W instead of the datasheet's 25W unless this is option code F or E but it doesn't say it is. The stable current is around 440mA which corresponds to 10.5W which is about right given that my house is relatively cold (around 18C).

I see from the failure document posted earlier that my unit:
7998 24.11.03 8592
is roughly 12 years old and was operated for less than a year (if it was on continuously).

The other units that I've yet to work on are:
7499 23.07.03 11568
7594 01.04.03 14280

and a couple not in the list:
662A
and
269C

[jpb]

I just opened up unit number 2 and was horrified to see the screws and even the sma connecter were corroded and rusted and the physics unit looked in a bad way. It looks as if water leaked into it.

So I decided to try it out before wasting any more time on it and amazingly the current draw etc seems normal and the after several minutes the lamp lit up! It doesn't lock but I've not yet changed the caps but it is looking much more hopeful than appearances would suggest.

Changed the caps and fired it up and it doesn't lock.

The HH value was way off but I reset to factory defaults and it is now more reasonable. I can see it sweeping up and down through 10MHz on a scope.

Some sample warmed up but not-locked log values are:
7B 00 B1 46 80 85 AF D9
7B 00 B0 3C 80 85 AE D8
7B 00 B0 32 80 84 AE D8
7C 00 B1 3D 80 85 AE D8
7C 00 B1 46 81 85 AF D9
7B 00 B0 4F 80 85 AE D8
7B 00 B0 59 80 85 AE D8
7B 00 B0 62 80 84 AE D9
7C 00 B1 6C 80 84 AE D8
7C 00 B1 78 80 84 AE D8

7C corresponds to 5V*(0xFF-0x7C)/256 = 2.56V which is better than my other unit and agrees with the fact that I physically saw the lamp light. The second 00 value corresponds to "warming up" but it has obviously reached temperature so it looks as if the Rb signal level is non-existant even though the dc level is ok??


No change after leaving on over-night, looks like the RF (modulated 90MHz?) is not there.

7C 00 B1 AD 80 86 AF D9
7D 00 B2 C3 80 85 AF D8
7D 00 B2 E2 81 85 B0 D9
7C 00 B1 CE 80 86 B0 D9

[Orange]

I just opened up unit number 2 and was horrified to see the screws and even the sma connecter were corroded and rusted and the physics unit looked in a bad way. It looks as if water leaked into it.

I also had one that had water damage, a lot of rusted stains all over the PCBs. Amazingly this one worked Ok after a CAPs change. I then cleaned it with alcohol.
The physics package still has some corrosion stains on it, which I cannot get rid of.
I think these things were stored in the rain somewhere.
BTW I saw in some of the Datum LPRO manuals, that if they are damp, sudden frequency jumps can occur. So a good bake-out is needed 
 

[Orange]

The current draw on warm up is 1.3A which is rather high (the data sheet has 0.9A). 1.3A corresponds to 31W instead of the datasheet's 25W unless this is option code F or E but it doesn't say it is. The stable current is around 440mA which corresponds to 10.5W which is about right given that my house is relatively cold (around 18C).

All my units start at 1.3 .. 1.36 A. They settle at 0.51 A. This when they are mounted on a bigger heat-sink in addition to the original one.
The data sheet is wrong, or these versions have an shorter lock time, which draws initially more power.

[uncle_bob]

All my units start at 1.3 .. 1.36 A. They settle at 0.51 A. This when they are mounted on a bigger heat-sink in addition to the original one.
The data sheet is wrong, or these versions have an shorter lock time, which draws initially more power.

Hi

Apparently these were made for "a major telcom customer" (according to the failure data pdf). It's not at all surprising that a tweak or two got made to the design to accommodate whatever the customer needed / wanted / thought up.

Bob

[jpb]

After changing caps on all five my current standing is:

Two good ones that lock at the right frequency (though one is getting rather old).

One that locks to a fixed frequency but is way out (by several kHz so perhaps the crystal is duff and it is never reaching the dip - it perhaps is locking on a flat part of the absorption curve well away from the resonant part).

Two ones that despite having good lamps don't lock at all and have zero readings for the rf lamp signal.

I'm working on a proper zero crossing detector, but in the mean time here is an Alan deviation plot for my two good ones. The two curves are the measurements (upper), the measurements over sqrt(2) assigning equal noise to each (the lower) and the three crosses are the values from the data sheet.

Given the crudity of the measurement I'm quite pleased with the agreement.

[TheSteve]

Have you tweaked any of the pots at all? One of mine didn't lock at all until I touched one of them. I guess I was lucky, all of mine work great and there was no water damage.

[jpb]

Have you tweaked any of the pots at all? One of mine didn't lock at all until I touched one of them. I guess I was lucky, all of mine work great and there was no water damage.

I messed around with the pots on one of the duff ones whilst looking at the output but it didn't improve it.
My next step will be to try changing the crystals on one or two.

I don't mind one duff one as I want to experiment with taking the bits out separately and perhaps using a separate ocxo (with some sort of interface circuitry) and it is easier to do that with one that is broken anyway. But 3 duff ones is a little high.

I have reasonable hopes of getting the one with the off frequency working and perhaps the last two will end up just providing spares.

[uncle_bob]

Hi

A quick poke with a scope will tell you if the crystals are still running in the oscillators. The same sort of check will tell you if it's on frequency or not (maybe not quite a poke, but the same idea).

Best approach is to take a working one and the dead one. Poke around and see what's not the same between them.

Bob

[Orange]

Buying these things is for sure a gamble. I have 4 working consistently after a lot of 'burn-in' hours. You are just unlucky with your batch.
One of mine  has a problem with the heating of the lamp and Rb cell. It jumps up and down with the temperature of both the lamp housing and also the Rb cell housing, and at some point completely runs away....
Symptoms of this are an oscillating current draw from 0.6 to 0.23

Since there are no schematics, it becomes almost impossible to do some fault finding....

[jpb]

Hi

A quick poke with a scope will tell you if the crystals are still running in the oscillators. The same sort of check will tell you if it's on frequency or not (maybe not quite a poke, but the same idea).

Best approach is to take a working one and the dead one. Poke around and see what's not the same between them.

Bob

I did have a look with the scope and the crystals themselves seemed ok but were a little low in frequency (especially the 90MHz) but then the scope probes are providing capacitive loading so it is hard to know. I didn't do a direct comparison with a working unit because I didn't really want to take it to bits again in the time I had available.

Looking at another thread on this forum I see that others have suffered from duff Step-Recovery-Diodes, if it is this then there probably is little hope in fixing it.

But it is early days, I don't have time to do anything during the week and very little time at weekends so progress is rather slow - still it is educational!

[jpb]

Buying these things is for sure a gamble. I have 4 working consistently after a lot of 'burn-in' hours. You are just unlucky with your batch.
One of mine  has a problem with the heating of the lamp and Rb cell. It jumps up and down with the temperature of both the lamp housing and also the Rb cell housing, and at some point completely runs away....
Symptoms of this are an oscillating current draw from 0.6 to 0.23

Since there are no schematics, it becomes almost impossible to do some fault finding....

You are right that it is a gamble but not a bad one. Even in my case I hope to have 3 good units at a cost of (excluding my time of course) of under £200 including the excessive postage and VAT whilst most units on ebay (which may not be that good anyway) seem to go for more than £100 each now-a-days so I have not really lost a lot and it is a very educational experience.

Those with 4 or 5 good units have done very well.

[jpb]

Further Alan Deviation measurements. Leaving the devices warmed up for a day and measuring over 48hours at 1Hz offset seems to give better results (I've yet to build my zero crossing detector but results aren't bad without it).

I've just started a 48 hours measurement of the ref vs itself to get the noise floor.

[uncle_bob]

Hi

Not quite sure if this is a dupe or not (posted before it seems to have vanished ... if it is ...sorry !!).

My load of Rb's got here today via DHL. They made it fine with no "lost in a snow drift" incidents. They were well packed and nothing seems to have been damaged in transit.

One of the ones I got took a major hit to the connectors at some point in it's life. The DB-9 is pretty well caved into the front of the unit. I think I'll save that one to work on last. One other one has a couple of dings and may be rusty. I'll know better when I get it open. The serial numbers are all over the place. They seem to have been pulled at random from the same pile as everybody else's parts.

A package from Mouser also got here today. I bought a batch of EGXL350ETD221MJ16S electrolytic caps. They are 35V 220uf 125C parts with a  1050 ma ripple current rating at 100 KHz. At 30 cents each, they were the best thing I could find. Based on a quick look, I think they will fit ok on the boards. They are radial rather than axial. I had zero luck finding anything with a rational ripple rating in axial (at least with my price limits). They have 0.6mm diameter leads so they should drop in fairly easily.

======

So after a bit of fun and games:

Seven of the units came open. The one with the crushed front was one of those.

The unit with the rust signs is indeed rusty. The screws are rusted in place. That's going to take some work to get apart.

Two other units have screws that just don't want to come out. Not sure if it is rust or if somebody decided to play with thread locker.

Of the seven open units, about half have the black caps and half have the blue ones. A bit less than half have the gold plated boards. The rest have tin plated boards.

That's all for now..

Bob

[jpb]

The unit with the rust signs is indeed rusty. The screws are rusted in place. That's going to take some work to get apart.

Two other units have screws that just don't want to come out. Not sure if it is rust or if somebody decided to play with thread locker.

Bob

One of my units I couldn't shift the screws and ended up stripping the head so I had to drill them out.

[uncle_bob]

The unit with the rust signs is indeed rusty. The screws are rusted in place. That's going to take some work to get apart.

Two other units have screws that just don't want to come out. Not sure if it is rust or if somebody decided to play with thread locker.

Bob

One of my units I couldn't shift the screws and ended up stripping the head so I had to drill them out.

Hi

I got the rusty unit open. It's not all that bad inside. I lost the "outside" board in the process. No big loss there. I still have two other units that I may need to drill. The most interesting thing I noticed was that the Philips head screws at the "bad" locations were head stripped when I got the units. I suspect that somebody at the factory got a little heavy handed.

Bob

[jpb]

I've now finished my 48hour measurement of 7998 against itself to get a noise floor and, considering the crudity of the setup, it is quite good and well below my other measurements which is ok.

So here is my final (for the time being) Alan Deviation plots. They are a bit flat, start off lower than expected but then don't drop as much as some other measurements (see for example:
http://www.ke5fx.com/rb.htm
)
Of course, assuming the standards are equal the individual results should be shifted down by LOG10(sqrt(2)) = 0.15 but this is not a big change.

What I find interesting is how having a stable reference (7998) gives a much better noise floor than using the internal oscillator in my Agilent 33522A. The mean measured frequency came to exactly 1Hz which was satisfying as well - though perhaps not surprising as I reconstruct the 20nsec clock counts from the measurements on my counter so that they are integer values.

Tomorrow I plan to have another go at fixing my broken ones.

[jpb]

It is not quite the same model but looks close enough to be useful (I don't know if it has already been posted) :

http://www.radcomms.net/EFRATOM%20LPRO%20101%20Repair%20Guide.pdf

[uncle_bob]

It is not quite the same model but looks close enough to be useful (I don't know if it has already been posted) :

http://www.radcomms.net/EFRATOM%20LPRO%20101%20Repair%20Guide.pdf

Hi

I have a pretty big pile of those as well. It's a *very* different design from Efratom rather than Temex. The common element is that the cell base station guys (Lucent etc) standardized on the Efratom LPRO early on for their internal production of timing modules. When Temex wanted to break in, apparently the "big guys" said - make it the same as an Efratom !!! The result is the gizmos we are all happily hacking away at. It's a Temex LPFRS on a heat sink that mounts on an LPRO footprint. It has a board that converts the LPFRS pinout over to a "close enough" LPRO pinout.

Yes that all does sound crazy. Consider that starting in the early 90's the cell guys made up maybe 90% of the Rubidium market. There ar a relatively small number of players in that market. Getting "on board" with one or more of them would have been a really big deal.

Just for reference, my LPRO pile came out of Lucent / AT&T base station salvage. I bought them as full (but badly damaged) modules. They were built like a tank so damage to the outside was not the same as damage to the guts.

Lots of fun.

If you get an LPRO .. . wait for it .... check the caps on the power supply.

Bob

[jpb]

I don't seem to be having much luck (or rather I don't have much skill/experience) with my lot.

I changed a couple of crystals which made no difference. Adjusted pots a bit - again to no effect.

I decided to swap the working lamp from the water damaged one into the much newer one whose lamp didn't fire (the lamp voltage being down at 0.7V on the serial port output).

It was a rather delicate operation - the wires are very brittle and after everything the working lamp no longer worked in the new unit so the lamp wasn't the problem and I damaged the lamp I removed (wire broke off and impossible to solder to what is left).

So I now have two units that are probably dead beyond my skills to repair.

This leaves me with two working units (one getting towards end of life from the voltage readings over the serial interface) and a mysterious third unit which thinks it is locked but the frequency is far too low. I changed the crystal and this made it worse. I don't know if the lamp fires because I've not run it without the case.

The unit with the lamp that doesn't fire does seem to come up to temperature and it produces a good signal on one lamp coil at 135MHz and a much smaller signal on the other end of the coil (?) at 135MHz but only a few tens of mV. Someone else reported higher frequencies for the lamp coil but it may be that my scope probe capacitance is loading things.

[uncle_bob]

Hi

Ok, here's what you are looking at:

At one end you have a neon bulb that's filled with Rb instead of Ne. Rather than fire it up with high voltage they fire it up with RF. (A neon bulb will fire on RF as well). If the RF end of things does not have enough power (or the bulb is shot) you will get no glow. The RF is running a resonant circuit to increase the field intensity. If that is out of whack, no glow.

Once you have a glow, you have a couple of atomic lines being excited. That light goes through another cell that takes out most of what you don't want (nothing is ever perfect). With nothing else in the cell, it would pass through to the photo detector and give you a nice DC voltage on the detector.

Toss in another cell (or monkey with the gasses) and you have a bit of magic. Hit the combo with microwaves at just the right frequency and the light to the photo cell goes way down. You have a notch filter. It's  a bit nutty because the notch is RF in and light out. The why has to do with quantum mechanics ... we'll skip that for now. The next trick is how to lock to a notch.

The normal way to lock to the notch is to modulate the microwave signal and look at how the recovered light responds. Tone in and only tone out = tuned right. Tone in and tone + second harmonic out = tuned wrong. With a bit of work you can create a PLL that will keep things centered up.

So what can go wrong:

1) No RF into the bulb or frequency way off from correct (= no field).
2) One or more bulbs broken
3) Photo detector broken or covered in slime
4) SRD (or other multiplier) not putting out microwaves
5) Cavity not tuned to SRD output (again no field, just at a very different frequency)
6) A whole raft of problems in the synthesizer or MCU.
7) A copule of other things relating to the quantum mechanics stuff we skipped (C field levels etc)

To make it even more exciting:

Back in the "good old days" the cells had to be pretty much right on or the (narrow) synthesizers of the day would not lock to the resonance. Along came DSP and DDS and a bunch of FPGA's. Now you do a cheap synthesizer with a much wider range. A greater range of cells can be used (like over 100's of KHz). The magic numbers get stored in an eeprom or flash somewhere and the unit chugs happily on it's way. Does everybody do a Rb that way? Who knows, there certainly are some people who do that.

How all this relates to which test points in the LPFRS, no idea beyond what has been posted here. This is not anything I designed. Many of the fine details of how this or that is done vary between manufacturers. The way Temex does it may not be how FE does it or how Efratom once did it or how Symmetricom does it today.

edit ... ====================

Got a chance to play with 8 of the 10 Rb's today (still need to drill out the other two).

5 of the eight lock up properly and appear to be ok.
1 of the eight locks up, but it is about 180 Hz low
1 of the eight sweeps past 10 MHz without ever locking
1 of them starts low and never sweeps as high as 10 MHz, it also has odd voltages.

None of this is with any cap changes. No pot tweaks. I simply opened them all up, looked around in them, put a DVM on the 9.2 V supply and let them warm up. I watched them on a 53181. I did not bother to hook up a serial port to them.

So here's a question:

There are two versions of the CPU board. That has been mentioned before. There is also a version of the RF board that has a daughter board on it. The daughter board has the crystals attached to it. Anybody else seen one? With the daughter board it seems to warm up more like an AT cut based oscillator. The normal parts are quite strange on warmup (very little shift).

All for now, more fun in the morning.

Bob