-
Repair : TEMEX/SpectraTime LPFRS/LPRO 10MHz Rb generator [got GPSDO to cal]
Posted by TiN on 18 Dec, 2015 17:30 -
As some may know already, due to f1lthy xmas time now I jumped into time-nuttery. What else it is, if one who don't even have frequency counter (any!) buys no less than five broken rubidium standards? Oh, well..
Units in topic are TEMEX / SpectraTime LPRO, which are basically LPFRS with heatsink and little coupling PCB. Quick googling revealed next useful links:
http://www.spectratime.com/products/isource/
http://www.egidy.de/temex_lpfrs_lpro/
https://www.eevblog.com/forum/projects/building-a-rubidium-freqency-standard-at-home/
http://time-nuts.febo.narkive.com/ssfEAnts/temex-lpfrs-01-analog-frequency-adjustement
Here's what I received today (lightning fast DHL shipment, yesterday I paid for them, today 3pm I got them delivered!).
Rather bulky and heavy things, 500gramms each. I still confused about purpose of those heatsink fins and frame, as whole thing is ovenized, so why to add extra mass?
All five are same units, with minor changes in labels, specced to power at +24VDC, have 10MHz output and RS232 TTL interface, as well as adjustment and lock pins.
Removing some screws allow to easily take mu-shield off:
Two rather dense populated PCBs interconnected by in-core flexible layer and mechanically locked together via Rb chamber outer case.
Chamber is heater from both sides with TO-220 FETs.
Here's look on PCB from outer sides:
And inner sides:
Motorola 68000 series MCU, DAC8800 and bunch of logic chips around.
Closeups..
There are two XO's glued to side of Rb chamber, 20.000MHz (bigger can) and 90MHz (smaller can in center). Some SMT trimpots and adjustable inductor(?) are nearby as well. Smaller part of chamber is lamp, bigger one is Rb cell with photodetector.
How do I know? Well, I turned this one in to see magic atomic glow...
In darkness:
It's not so bright, but definately visible in normal dim room lights.
I tried connecting serial console to it, set speed 1200 bps, 8N1 mode, but only garble came out:
Where should I start this repair? I tried 3 modules , two drift hundreds of hertz in just seconds, compared to my 33120A, and lock never come low, while one is locking (signal goes low) but still drifts.
Chassis getting about +50-53°C, and power current stabilize at about 0.515Amps with +20VDC (don't have powerful enough +24v source).
As usual, full worklog will be here. -
Cool that you also bought some
and did a teardown. So the possibility of successful repair increases massively for me, especially with your great PCB reversing skills.
Mine haven't arrived yet, but I have some nice tools to repair microwave electronics
They look quite nice and well laid out, from the inside. Maybe start with measuring ripple on tantalum/electrolytic caps? -
You reverse Rb side PCB, i reverse digital PCB, deal then!
I have nothing for microwave, just 4GHz scope and some 8GHz probes. I can try 16G scope in theory, to measure Rb cell output?
-
I'll try my best ,
I have HP 8566&8568 spectrum analyzer, EIP microwave counter, HP 8684 microwave cavity oscillator that covers the 6.835GHz rubidium transition frequency and some more lower RF frequency generators.
But not much digital gear, mostly stuck with my oldish analog Tek scopes -
Here's what happening in first tests:
-
I've never worked with that model of Rb standard, but all the other ones I've seen follow the same pattern so here goes.
The heat sink is necessary to keep the electronics cool for long reliability. There are typically two, maybe three ovens. The main two are for the Rb lamp (~140C) and the Rb cell (~85C). Those temperatures are higher than most other Rb standards, but they're listed in the manual. The light from the Rb lamp shines through the Rb cell to a photocell so they'll be at opposite ends of that brass cylinder. The point where you can see the light is the break between the two. If there's a third oven, it's a normal quartz OCXO. Most Rb standards don't bother with this. Check the temperatures of those ovens. If they're out of spec nothing will work right.
The second point is the Rb lamp. You've shown a picture that shows at least one of them has a working lamp so that's good. Whether it's quite right or not will have to wait until you get the serial port working.
The third main point is that you should see the output sweeping up and down in frequency over a range of a few hundred Hz or so. It might take 30 - 60 seconds for a sweep. If the quartz crystal has aged enough that this sweep range doesn't include 10 MHz, the unit won't lock. Look for a trimmer to adjust the frequency. It looks like this model uses the 20 MHz crystal and divides it by two. They control the crystal temperature by gluing (soldering?) it to the brass cylinder. Actually, the 85C temperature of the Rb cell is in the temperature range for an OCXO, so maybe that part of the brass cyclinder is the Rb cell. Drift on this oscillator is the most common failure I've seen in the units I've played with. I just looked at your video and it looks like the sweep is happening.
The serial port may have been customized to a different speed. Use a scope to check the length of a bit to see if it's actually 1200 baud.
Ed
-
Checked serial, it is 1200 Hz indeed, but inverted (idle bus stays low). I'll add inverter and try tomorrow. 3am here already...
-
According to this document, your LPFRS SN #7694. #7731 and #7894 must be in perfect working order
http://www.spectratime.com/documents/lpfrs_failure_rate.pdf
SN first op h
7695 27.06.03 12192
7731 20.06.03 12360
7894 24.11.03 8592
But that was 10 years ago
Also interesting:
http://www.spectratime.com/documents/life_mtbf1.pdf
-
... with +20VDC (don't have powerful enough +24v source).
I think you should correct that before anything else. -
Quote
I tried connecting serial console to it, set speed 1200 bps, 8N1 mode, but only garble came out:
Try looking at it with a scope - if there is any weird processor communication timing issue (or a different baud rate), the baud rate might be well of. The fact that it sent out SOMETHING means that something is working there. Also, if the unit wants 24V I suggest you give it to it - the heater may not be powerful enough to work of such a derated voltage... or something else might disagree.
Offtopic: Seeing your previous (and amazingly beautiful) precision voltage sources, I'm guessing that your next purchases will include a dewar, 70+GHz frequency source referenced to 10MHz and a bunch of Josephson junctions?
-
Serial comms fixed, with simple 74HCT14 between Rb and UART-USB.Code: [Select]
TNT/LPFRS01 Rb Version 2.05 1/2/02 Checksum: 1D1A/E981
00
F8
?
00
F8
80 84 62 38 1E 93 A5 E4
7E 7F 62 39 1F 8C A0 D4
7D 79 62 39 1E 93 9A D3
7C 7B 61 39 1E 93 9A D3
80 81 62 39 1E 97 A0 E4
80 6A 62 39 1F 99 A2 E4
7D 78 62 39 1E 9A 9B DB
7E 6A 62 39 1E 98 99 DE
80 7A 62 39 1E 9A A3 E6
7E 7D 62 39 1E 8A A3 DA
I also took apart all modules. Nothing visibly blown inside. They are made in 2002-2004 per datecodes, and 4 are older revision, while one is newer (bit changed in layout, photos above are for newer one).
Lamps on 4 modules are OK (at least I can see them glowing). (did not test 5th yet).
Crystals 20MHz and 90MHz are working..
Heading now to local market for 24V power supply.
Temperature on lamp side (smaller section of physics chamber) is about +95C, photocell side about +75C (open case, no cover).
-
QuoteCode: [Select]
80 84 62 38 1E 93 A5 E4
7E 7F 62 39 1F 8C A0 D4
7D 79 62 39 1E 93 9A D3
7C 7B 61 39 1E 93 9A D3
80 81 62 39 1E 97 A0 E4
80 6A 62 39 1F 99 A2 E4
7D 78 62 39 1E 9A 9B DB
7E 6A 62 39 1E 98 99 DE
80 7A 62 39 1E 9A A3 E6
7E 7D 62 39 1E 8A A3 DA
VCXO Control (4th param, 38-39) seems stable, so I wonder if the output was also stable at that point.
-
Bought Meanwell 24V/5V 4.6A dual PSU, now output seem to be 10.000000MHz (at least very close matching that value on HP 33120A generator now).
Capturing serial logs for modules (status command every second), and will hook four Rb's together on scope to see if I can sync them...
Maybe they are not so broken after all?
-
Maybe they are not so broken after all?
That would be great news, TiN!
Thanks for your efforts! I'm waiting for my standards.. -
Hooked Rb's to SMPS DC/DC (Meanwell RD-125B, 24V 4.6A + 5V 4.6A).
Three modules lock in and stay very close together, according to scope. Triggered on CH3.
Two others are way off.
Sorry for lab mess, I was tinkering with all this whole night... Going bed now (7:47am).
~0.5Amps taken from PSU after warmup. Rb cell temperature +85-88°C (no case).
Here are serial logs:
Module S/N 340A
Module S/N 7011
Module S/N 7011 adjustment
Module S/N 7731
I got some video footage on this too, will cut it and upload later. -
Did you leave the working ones running overnight with the scope set to infinite persistence?
-
No, i will put them into cases first.
-
3458A is upside down in hope to reverse the drift?
-
Did you leave the working ones running overnight with the scope set to infinite persistence?
Nice if you have such a great scope
Does anyone know a program which can do webcam snapshots periodically (WinXP, or better Linux)? My idea is to mount a webcam in front of my old Tek7904A with 2x 7A26 in chopped mode and then take snapshots periodically. Then use imagemagick to logic OR all images => analog digital persistence scope
-
Does anyone know a program which can do webcam snapshots periodically (WinXP, or better Linux)? My idea is to mount a webcam in front of my old Tek7904A with 2x 7A26 in chopped mode and then take snapshots periodically. Then use imagemagick to logic OR all images => analog digital persistence scope
Raspberry PI + camera and Tektronix C-5C
Set up a cron job to take a snapshot as often as needed.
http://www.netinstructions.com/automating-picture-capture-using-webcams-on-linuxubuntu/
-
Raspberry PI + camera and Tektronix C-5C
Set up a cron job to take a snapshot as often as needed.
http://www.netinstructions.com/automating-picture-capture-using-webcams-on-linuxubuntu/
I have no operational RPi and noTek C-5C, but I have Thinkpad X61s, some black cardboard, some old craptastic Logitek webcam and my awesome shellscript skript kiddie skills
I found fswebcam to work quite nicely, so I used it for the shellscript sample below: (you have to install some libs, if you install the latest 2014 version of fswebcam from source on old linux like my Ubuntu 10.04)
So, here the shellscript sample, it even worksCode: [Select]initialtimestamp=$(date +%s)
piccount=1 #image counter, set 1 initially, dont set zero, or it will crash!
picdelay_sec=10 #delay between images, fswebcam needs some time for each pic, say 10-15s
a=1 #need this shit apparently so that nested loops work :)
b=2 #otherwise first "done" breaks down everything :(
while [ $a -le $b ]
do
timenow=$(date --rfc-3339=seconds)
timestamp=$(date +%s)
customtime=$(date +%Y%j%H%M%S)
next_pic_time=$(( initialtimestamp + piccount *picdelay_sec ))
(( piccount++ ))
echo "time $timenow"
echo "timestamp $timestamp"
echo "custom timestamp $customtime"
echo "picdelay $picdelay_sec"
echo "initial timestamp $initialtimestamp"
echo "next pic time $next_pic_time"
while [ $timestamp -le $next_pic_time ]
do
timestamp=$(date +%s)
done
echo "Cheeze :) $timestamp"
echo "next pic time $next_pic_time"
fswebcam -r 640x480 --jpeg 80 test$timestamp.jpg
done
It takes unix timestamp, compares that to some timestamp in the future, waits until that time is reached and then takes a picture with fswebcam. Loops restarts after that. With 40kB per image (640x480), it can take lots of scope images until HDD is full.
Depending on how fast your computer is, you can set pic delay as low as 2 or 3 seconds, my old thinkpad needs some more time.
-
Hi Guys!
My rubidiums arrived today (Germany) -that was really fast.
I only had time for a fast check. 2 Rubidiums lock in after a few Minutes.
The other three ones not. No time for more testing at the moment
-
Cool. Are they same as mine?
I see at least someone have freq. counter.
Seems I got hooked, BG7TBL GPSDO board on it's way... -
Cool. Are they same as mine?
I see at least someone have freq. counter.
Seems I got hooked, BG7TBL GPSDO board on it's way...
Which one did you get? The original one isn't a perfect 10 MHz, it is closer to 9.999 999 999 8 Hz which is very easy to see compared to a Trimble GPSDO etc. -
Pic on eBay shows one with Trimble labelled OXCO. Will see what one exactly will arrive.
