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

[PA0PBZ]

What are the date codes on these units that have failed caps?  I realize that these units run hot, but it still surprises me that these caps are failing and causing problems.  What brands are the failed caps?  Do the good units use the same brand?

the unit I replaced the caps on has serial # 7902, delivery date 24.11.03 according to the document.
The caps are black, 47uF, 35V 105C, not sure what brand they are, it has a rhombus shape with dots where the lines connect.

[shempe]

I've found this caps:
http://www.ebay.de/itm/231766325218?_trksid=p2060353.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT

Datasheet: http://www.mouser.com/ds/2/427/118aht-239891.pdf

Looks like a good, high quality replacement cap.

Nice find, thank you. I bought a few for my Units.
But beware, the wire´s are 0,8, better use some with 0,6 like the BX Series from nichicon in these units.
With a little bit creativity, a la "cut the old one out and and solder the new caps to the old wires" it works.   

Here another PDF about this Unit: https://www.denk-stein.com/pdf/fp/lpro-en.pdf

[TiN]

Bought simple GPS antenna.

UCCM-P >DIAG:LOG:READ:ALL?

Log status:  10 entries

L00:99/08/22.00:10:49 Log cleared
L01:99/08/22.00:12:05 Output mode: Active
L02:99/08/22.00:12:14 Output mode: Unlock
L03:99/08/22.00:35:38 Output mode: Active
L04:99/08/22.00:41:36 Output mode: Unlock
L05:99/08/22.07:39:13 Self survey started
L06:99/08/22.23:59:50 Jammer statistics: 10
L07:99/08/26.01:48:23 Antenna failure cleared
L08:16/01/08.16:58:58 Output mode: Active
L09:16/01/08.16:59:09 Output mode: Unlock

Code: [Select]

Command complete
UCCM-P >POSSTAT

---------------------------------------------
 01/08/2016  17:04:39
---------------------------------------------
 Position: LAT(N  25:57:29.658) LON(E 119:31:14.779) H(         +39.66 m MSL)
---------------------------------------------
 Geometry: PDOP(0.0) HDOP(0.0) VDOP(0.0)
 num of visible sats >  3
 num of sats tracked >  3
  -------- Receiver Channel State --------
 CH  0 > SateID(01) TrackMode(tracking ) SigValue( 0)
 CH  1 > SateID(03) TrackMode(tracking ) SigValue( 0)
 CH  2 > SateID(22) TrackMode(tracking ) SigValue(29)
 CH  3 > SateID(08) TrackMode(pos avail) SigValue(39)
 CH  4 > SateID(16) TrackMode(pos avail) SigValue(33)
 CH  5 > SateID(32) TrackMode(tracking ) SigValue(24)
 CH  6 > SateID(27) TrackMode(pos avail) SigValue(35)
 CH  7 > SateID(19) TrackMode(tracking ) SigValue( 0)
 CH  8 > SateID(09) TrackMode(tracking ) SigValue( 0)
 CH  9 > SateID(07) TrackMode(tracking ) SigValue(21)
 CH  10 > SateID(30) TrackMode(tracking ) SigValue( 0)
 CH  11 > SateID(11) TrackMode(tracking ) SigValue(20)
---------------------------------------------
 Rcvr Status(0):   
---------------------------------------------
 Antenna Voltage:  5000.00mV   , Antenna Current:    13.24mA
Location still incorrect, will see how it goes tomorrow morning..
Antenna just fixed 30 cm from window outside, don't have other way, it's apartment on 10th floor.

[TheSteve]

TiN,

Use the command - SYSTem:PRESet to get the GPS to perform a full survey of your location.

[TheSteve]

Here are some results of 8 hour time interval plots from my HP 53132A comparing my Nortel(Trimble) GPSDO to one of the rubidium units and comparing two rubidium units to each other. Each plot was run twice.

[edpalmer42]

It looks like something is degrading your results.  The 53132A has a resolution of 150 ps.  This should give you a one second result of 1.5e-10 but you're seeing more like 6e-10 and it's consistent through all the tests.  If you haven't done so, check the noise floor of your 53132A by connecting a signal to the Ch1 input through a T connector  and a cable to Ch2.  Set Ch1 to 1 Mohm and Ch2 to 50 ohm.  Measure the time interval from Ch1 to Ch2.  Make sure that the cable doesn't move or change temperature during the test.  If that test is good, now you're into the black arts of AlDev measurements.  Ground loops, power supply coupling, injection locking, external and internal interference, stray sunbeams or other temperature changes, etc.  After all, we're mucking about in the PPT area.  Be sure to think positive thoughts!  Try to stay at least 2 meters away from the equipment during the measurements.  No, I'm not kidding! (about the 2 meters - not sure about the positive thoughts, but it can't hurt  )

The Rb-GPSDO tests are about what's expected.  Looks like there was some warmup and stabilization between the two tests.  GPS performance can be broadly defined as a line that goes through 1e-10 @ 100 sec. and goes both up and down 1 decade for each time decade.  Notice that your results turn down and to the right to follow this line.  Different GPSDOs will take a slightly different path, but they all run parallel to the 'GPS line'.

Your Rb-Rb results are also about what's expected.  The manufacturer only specs out to 100 sec. as 2e-12 and you're in that ballpark from 1000 sec. on.

Did you do the Rb-Rb tests first?  I see that the first Rb-GPSDO test bottomed out at ~1e-12 while the second one kept on going down.  Something improved between the first and second test.  Did you make any changes in your setup, cables, power supply, anything?

The graph always shows the performance of the worst of the counter, the GPSDO, and the Rb.  So below about 20 sec., your counter is blind to the performance of either the GPSDO or the Rb.  All you're seeing is the counter's performance.  At about 20 sec. the GPSDO's performance becomes the dominant factor.  But the Rb is so good that you don't start to see it until you get to about 1000 sec. and even that may be misleading.  The red line that keeps going down at the right side of the graph is following the GPS performance.  The only way it could do this is if the Rb was better than the GPS.  Otherwise, the red line would turn and run horizontal like the others.

So, if you really want to find out the performance of your Rb units, more investigation is required.  You might discover that Rb 4 is somewhat better than the others.  Such things do happen.

Ed

[TheSteve]

Thank you for your feedback. There were no changes between any of the tests other then time of day and probably a 5 degree C temperature drift up and down throughout the day/night. Everything was powered on the entire time. However I was certainly closer then 2 meters and other gear was switched on and off, including signal gens. There are 8 sources of 10 MHz running right now - three GPSDO's and 5 Rb's all quite close. I also have no control over what my 20+ pound Bengal cat does while I'm at work. For all I know he could think the Rb's are paw warmers. The biggest factor is likely the 53132A itself. It is calibrated however the fine interval step was skipped as it needs a very specific HP setup - so that may be ruining the noise floor.

edit:

Looked into the fine time interval calibration further. It is a 4 step process with the HP 59992A time interval calibrator generating 4 signals. The first two are 10 Mhz square waves in phase and the second two are 10 MHz square waves out of phase. I am not sure what the specific differences are between the measurements which appear the same. The calibrator documents don't give any detail, it just says they are in phase signals. Anyway, this may be something I can generate at some point.

I also did more testing to try to get closer to the potential noise floor of the 53132A itself. Feeding in a test 10 MHz square wave cut the noise floor to 3e-10. If I feed in a test 1 MHz square wave it is 1.63e-10. I'll have to ponder how the input frequency relates to the noise floor. I think the biggest difference would be if all of my signals were square waves, right now all GPSDO's and Rb's are sine.

[TiN]

Replaced capacitors on two modules.
Rb cell value on one of them instantly jumped to 8A-8E value ranges.
Second one I trimmed to ~4x Rb cell value.

Adjusted both modules after lock to GPSDO, using scope (used both coarse and fine registers using UART) to get minimum phase drift.
But it's tricky. Also I'm not so sure GPSDO output is good, it sees only 4-5 sattelites. Perhaps I need better antenna?

Red - GPSDO, Blue and green - two RB's. 10 min, infinite persistance.



XY-mode:



RMS Jitter of RB units is ~1ns, while GPSDO output is ~2.2ns

[edpalmer42]

Looked into the fine time interval calibration further. It is a 4 step process with the HP 59992A time interval calibrator generating 4 signals. The first two are 10 Mhz square waves in phase and the second two are 10 MHz square waves out of phase. I am not sure what the specific differences are between the measurements which appear the same. The calibrator documents don't give any detail, it just says they are in phase signals. Anyway, this may be something I can generate at some point.

I also did more testing to try to get closer to the potential noise floor of the 53132A itself. Feeding in a test 10 MHz square wave cut the noise floor to 3e-10. If I feed in a test 1 MHz square wave it is 1.63e-10. I'll have to ponder how the input frequency relates to the noise floor. I think the biggest difference would be if all of my signals were square waves, right now all GPSDO's and Rb's are sine.

I'm not familiar with the details of the 53132A.  If the fine time interval calibration is intended to reduce the internal noise, then it's important in this situation.  If it ensures accuracy of the measurement then it probably won't have much of an effect.  Absolute values aren't important here.

Square waves definitely give better results when you're approaching (or exceeding!) the limits of your measuring equipment.  Trigger noise is something that you always have to keep in mind in these situations and square waves help minimize it.  You can also use the cable test that I described earlier to make sure that your counter doesn't have any white noise issues.  As the test runs, the graph keeps going lower and lower in a straight line.  If it gets down below levels you're concerned about then you're good.  If the counter has some white noise issues on the input, perhaps due to a wounded but not killed input stage, the graph will turn and become horizontal.

Another thing you should do in Timelab is check the phase and frequency graphs.  If there are anomalies or glitches there, your AlDev results are suspect.  Save your data and then try to edit the phase or frequency graph to remove the problems.  Once both graphs arre clean, your AlDev results should be valid.  It's also helpful to post the .TIM file so that others can see and play with your data.

Ed

[Orange]

I received a batch of five yesterday. First replaced all the caps, some of them where toast, mainly the black ones. The blue ones were not really bad, still high ESR of 0.8 ohms.
After some testing I have now 4 units working sort of. On one unit there is some LF phase noise...

I did some probing on the board, and I think I have found the Rb demodulated signal. At least this is what it looks to me. Referring to pot-meter designations.
I think the following functions are:
Pot A : Output signal level of the detector. I have set this to aprox 3..3.5 Vpp after tweaking all other pots
Pot B : I believe this the SRD diode bias, it behaves like this, and is fairly critical to set (CCW is a good starting point)
Pot C : Some sort of AGC for the RF level, fully clock wise is the safest I found
Pot D: No idea, perhaps an detector balance signal, but you can tweak it on the Rb signal.  It was set to mid position for largest Rb signal.

Update: One of the units did not get it to 10.000 MHz during the lock sweep. I changed the crystal for a new one (ordinary 20MHz type) and that made it work :-)
It was glued with Sikaflex (Swiss made polyurethane kit)....

Updated picture of TiN with potmeter locations

[jpb]

I've been following this thread with interest, wondering whether to join in or not (i.e. buy a set of 5).

The postage to the UK would be $80 so the cost for 5 would be $129 + $80 + 20% (Inport Duty or VAT) which is around $250. I guess replacing the caps/other bits would be around $10 per unit and not all of them would be repairable. So though the process of restoration looks fun and satisfying it is a largish investment and a bit of a gamble if they require more than just a change of caps to get them functioning properly.

Given that I have very little time at the moment and don't really have any space (I'm in between house moves living in small rented space with my equipment all in boxes) I suspect the sensible thing would not to acquire yet more stuff to carry around but you all seem to be having a lot of fun.

How are the restorations going and can anyone say what fraction of their lot of 5 are repairable, function ok?

[TheSteve]

All five of mine are working great. I've run them on and off to see how they lock from cold etc - always perform exactly as expected. If I log them with my 53132A they all look good. I'd think you have extremely good odds of easily getting 3-4 of the 5 working, if not all 5.

[jpb]

All five of mine are working great. I've run them on and off to see how they lock from cold etc - always perform exactly as expected. If I log them with my 53132A they all look good. I'd think you have extremely good odds of easily getting 3-4 of the 5 working, if not all 5.

Thanks for the feedback.

I feel very tempted - though if I get them they will probably need to sit in a box for a few months until I get my lab back.

[jpb]

What are the date codes on these units that have failed caps?  I realize that these units run hot, but it still surprises me that these caps are failing and causing problems.  What brands are the failed caps?  Do the good units use the same brand?

the unit I replaced the caps on has serial # 7902, delivery date 24.11.03 according to the document.
The caps are black, 47uF, 35V 105C, not sure what brand they are, it has a rhombus shape with dots where the lines connect.

Looking at TiN's photos the blue caps are 63V 85C / 40V 125C. So perhaps the later units had upgraded temperature ratings? Also It is hard to find Caps at that Voltage and temp that are 0.6mm lead diameter. Looking through Farnell there are some very expensive Vishay ones but they are only 0.6mm up to 35V and then are 0.8mm and they are very expensive (but 20,000 hours at 125C presumably mean they are pretty robust):

http://uk.farnell.com/vishay/mal212310479e3/cap-alu-elec-47uf-35v-axial/dp/2407388

The 0.8 mm 63V 125C ones are less than 1/4 the price and 8000 hours at 125C seems reasonable:

http://uk.farnell.com/vishay-bc-components/mal211838479e3/cap-alu-elec-47uf-63v-axial/dp/1695292

[jpb]

I received a batch of five yesterday. First replaced all the caps, some of them where toast, mainly the black ones. The blue ones were not really bad, still high ESR of 0.8 ohms.

0.8 ohms seems pretty low given that high temp (125C) Vishay caps list a ESR of over 2 ohms on the data sheet. So it looks as if the blue (40V 125C) caps were ok while the black (35V 105C) ones were not up to the job.

[Orange]

The design got upgraded during their life. The old units have black caps of really poor quality, they are also 85 deg types with 0.6mm leads. The PCB is also designed for this. You can fit 0.8mm caps in it, but need to suck the solder very good out of the hole.
The latter production ones have bigger holes, adapted for the 0.8mm leads.
I used 100uF/63 volts, 85 deg. types from Philips, which I had at hand. the ESR of those are 0.22 Ohms. I don't mid to replace the caps in a few years from now if they should go bad.

Old units can be identified by the stickers on it with a Tekelek Aitronic logo.   

Mind you that the design requires low ESR types, since some of these caps are used in a switched mode set-up with a 125KHz freq. (See manual, general description)
I'm not sure if the Farnell types can be placed in this design looking at their ESR (2.xx Ohms). If all of this becomes an issue you can also put in radial caps from Nichicon or Panasonic. The voltage on one of the caps is 24 Volts, other run at 16 volts... So 35Volts for the caps are a minimum I would say.


 

[TiN]

I used radial 105C caps instead, there is still plenty space inside Rb to fit standing radial caps.
Will post more details after get back to it, bit busy with voltnut projects for now.

[jpb]

I see the same seller also sells a lot of 5 without the heat sinks for the same price. These have the same label on the front as the 2011 datasheet - I don't know if this means they are any newer?

I now can't decide which of the two sets to order.

http://www.ebay.com/itm/Lot-of-5-LPFRS-01-10M-Rubidium-Frequency-Standard/252232382829?_trksid=p2047675.c100005.m1851

http://www.ebay.co.uk/itm/272086453625?_trksid=p2060353.m1438.l2649

Everyone on here has ordered the ones with heatsinks - was this just a random choice?

[edpalmer42]

Same price for 5 units.  Same stupid shipping cost even though the weight will be substantially less.  But now you have to find your own heatsink.  Doesn't seem like a good deal to me.

Ed

[Orange]

The ones without heatsink are the older ones, since they have the Tekelek Airtronic logo on them.

At one point Tekelek Airtronic sold the TEMEX subsidary (TEMEX Neuchatel Time TNT)
http://www.spectratime.com/product_downloads/4g_sro_pr.pdf

TNT became TEMEX Time

TEMEX Time was again renamed to Spectratime....
http://www.orolia.com/pdf/TemexTime%20_changes_name_toSpectraTime.pdf

[TheSteve]

When we bought them originally the non heatsink version wasn't even listed. Being the heatsink version is not only newer but also that a heatsink is required I'd still go for the heatsink version.
And at this point I think its fair to say most can be easily fixed.

[jpb]

Thanks everyone for the info.

So the heatsunk ones are the better buy.

[jpb]

I have just ordered and paid for a set. By the time postage and tax/import duty is added the price doubles but it should still be a bargain as long as at least a couple of them can be got working.

[jpb]

Mine arrived today, but sadly it may be a week before I can get to play with them properly.

Meanwhile I've been setting up an Alan Deviation kit in terms of my old TTi TF930 counter (my new Tek one is hundreds of miles away at my mother-in-law's house), my Agilent 33522A, a mini circuits mixer and low pass filter and some home-brew software. I'm quite pleased with the noise floor (using the "reference" as input) see attached. Of course things might be very different when I try doing it with two real devices. (And I don't yet know if I can get them working.)

The plan is to use one as a reference for the 33522A and the counter and the other one fed into the mixer and the output of the mixer going via the filter to the counter the difference frequency of 10Hz seems to work best. (The plot was this setup except using the 33522A's internal reference for it and the counter and feeding this via a splitter to the mixer as well.)

Edit: I add a photo of the setup.

Edit : EdPalmer42 has kindly pointed out that my numbers are probably unbelievably good, and I agree with him that they need investigation. My 1 second value of 1E-13 in particular. Please note though that I have plotted LOG(tau) so that 1 on the x axis is 10 seconds not 1 second.

I'll remove the plot until I can come back with more realistic results or at least have thoroughly checked my code and workings.

Edit again - Correction! (Thanks Ed). I've rechecked my code which I originally wrote for frequency measurements but then put an option for period measurements. In one place I failed to do this correctly so I ended up having a factor of 100 error (roughly dividing by 10 instead of multiplying by 10). The graph is now rather more realistic - it is the same shape but shifted up by 2 (on the Log10 scale). I attach it.
The ADEV noise floor is 10E-11 at 1 sec. My counter is quantised at 20 nsec (50MHz) but I gain a factor of 1E6 from the mixer but then lose some on the whole process. The 10Hz signal period is 100 mHz which is measured to around 10E-5 (orders of microseconds) so with the superhet factor this comes to around 1E-11 which all makes rather better sense.

Hopefully when I get my good counter back (30psecs with back to back measurements of each period) I should be able to do a bit better.

[jpb]

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.