EEVblog Electronics Community Forum
Electronics => Metrology => Topic started by: JohnPi on September 22, 2017, 04:18:03 am
-
I have an external OCXO and wanted to use it to calibrate my 53132A. However, after I successfully calibrate, the 53132A then reads
10.000 000 002 4 or so. It is not that it is measuring the offset of the OCXO, but when calibrated (using the internal CAL: TIMEBAS capability) with a stable signal (visible dither is about +/- 3 counts in the last digit), but when measuring the exact same signal that it was just calibrated with, it reads high by a noticeable amount.
Has anyone else noticed this on their counter ?
-
If you split the 10MHz from the OCXO both to the Ext Ref at the back and into an input channel you should get an accurate number, with a spread corresponding to white phase noise that can be calculated from the counter specs.
For the 53230A the single-shot spec is 20ps for time-interval but in practice it does around 11-13ps, and that corresponds to an ADEV of around 2e-11 at 1s.
So with a 1s gate time, and the above setup with the same 10MHz to Ref and CH1, you'd expect values around 10MHz with a standard deviation of 2e-11 * 1e7 Hz = 200 uHz.
If you disconnect your OCXO from Ref and only connect it to CH1 it depends on the quality of the internal and external OCXO, you'd probably see a drift of maybe 1e-9 per day between the two OCXOs.
Anders
-
You need to repeat the calibration procedure a few times and with each time it gets closer to the nominal value and then it will stay stable.
I have noticed this requirement on my 53132A and 53131A counters.
-
The 53131/2 counters use a 12 bit DAC to set the OCXO EFC voltage. Depending upon your OCXO this gives a tuning accuracy of 1-2 mHz. Also, the tuning algorithm does basically a binary search of the EFC word to find the DAC value that puts the OCXO on freq. It starts with measuring the OCXO at min and max EFC words and refines the DAC value over a period of around 3 minutes. The initial min/max voltage swing can disturb the OCXO enough so that the OCXO later drifts further off frequency.
On the subject of HP-531xx counter calibration... the calibration of the counter time interval measurement can be either a crude "quick" mode or a sophisticated "fine" mode. The quick mode just does a measurement of channel A. The "fine" mode does four measurements between channel "A" and channel "B". The fine cal requires a high dollar pulse generator and a rather unobtainum HP-59992A J06 time interval calibrator.
Calibrating the channel gains requires a precise 5V source.
I have designed a board that replaces the pulse generator / HP-59992A. It also has a precise 5V reference for the channel gain cal. The time interval calibrator requires a high quality 10 MHz input signal (seems to work well with the 10 MHz output of the counter if it has the OCXO timebase).
The time interval calibrator works by squaring up the input 10 MHz signal, digitally multiplying it up to 20 MHZ, dividing it by two into complementary 10 MHz signals, buffering the outputs and level shifting them to 1V p-p, and routing various combinations of the true and complemented signals to the output BNC connectors. These correspond to the 1/2/3/4 buttons of the J06.
I plan to sell a kit (or possibly assembled units). I have parts and boards for 50 units on the way...
Hopefully the cost will be under $30.
-
I put one of Gerry Sweeny's OCXO boards in my 53132A. I used a UCT 8663 OCXO. I originally had the board set to the 0 .. 10V output range. The EFC sensitivity of the OCXO was around 0.75V / Hz and the best the timebase autocal could do was around 2 mHz. The OCXO was on freq with a 4.5V EFC voltage. I changed the OCXO board to 0 .. 5V and the autocal got the OCXO to around 1 mHz.
Over a month of continuous operation the OCXO drift rate settled down to less than 1 mHZ/day, but I was doing the autocal every day. After the initial month I did an autocal and did not change it for another month. The OCXO had drifted less than 1 mHz.
The reason I did the calibrator board was my 53132A came with early firmware that did not put commas between groups of three decimal points... very annoying to use. I upgraded the firmware to a much later version (ROM images are on the forum somewhere), but the firmware upgrade requires a full calibration. A member of the time-nuts list sent me a J06 for the cost of shipping from Australia, so I really didn't need my board... but I had already designed it and figured that other people would find it useful. HP has an ap-note that shows how you can use it to improve time interval accuracy for any counter.
-
I'm using Gerry Sweeney's OCXO also; I think it has a Trimble OCXO which is good.
A problem with the external OCXO I have is that it is unbuffered. the freq. shifts ~ 100 mHz when using a 50 ohm load vs not using one. Perhaps the 53132A is changing its internal 50 ohm on & off ? I plan on building a low noise buffer for my OCXO and will try with that.
-
Theoretically you can get better performance with OCXO boards that use a pot to set the frequency than the DAC. But that means you need a VERY stable (time and temperature wise) EFC reference voltage and pot (plus probably some range limiting resistors). Probably not an easy thing to achieve in the real world. If I did that, I would mount the pot on the OCXO board so the adjustment screw could be accessed through a hole drilled in the case to that you could adjust it without opening the case and messing up the thermal environment.
I've put a lot of 8663's into counters like the Tek DC5010. The 8663 has a nice reference voltage output. I used some very stable pots that I glued to the OCXO case with thermal epoxy. That worked rather well.
-
I plan to sell a kit (or possibly assembled units). I have parts and boards for 50 units on the way...
Hopefully the cost will be under $30.
Please add me to the mailing list for when you go live with it - I'll happily buy a kit for this (in fact I might get two).
TonyG
-
I'll take one too.
-
I plan to sell a kit (or possibly assembled units).
Are you willing to publish the circuit diagram?
-
It looks like the pinouts of the rotary switches that I received do not match the sample... bastards! Looks like I may need to re-spin the PCB.
-
I am interested in one.
-
I am also interested in one of your kits
-
A quick update on the time interval calibrator board... the Chinese board house has finally shipped the boards DHL express (I think they mis-placed the original order). Should be here in a few days, US customs is apparently giving everything from China a triple, un-lubed body cavity search looking for drugs. But, I'm 99% sure I need to re-spin the board to match the pinouts of the rotary switches that I received vs the original samples I got in... grrrr...
I found a local assembly house and a couple of people willing to do assembly. It will probably raise the cost $5-$10 over a kit, but it does allow me to do some testing on them. Setup costs for small quantity SMD assembly are a bitch.
-
Interested, add me to the list, please.
-
An update on my HP-531xx calibrator board:
The boards finally arrived! The board house I use in China normally take 7 to 8 days to get boards into my hands. This time it took 3 weeks... grrrr.
Anyway, as expected, I need to re-do the boards to match the rotary switches that I got in. I built one up and bodge wired the switch in... SUCCESS! I ran the calibration routines on my 53132A and all steps passed the first time.
-
Need one, please add :clap: :-+
-
Hi all, (newbie here)
Add me to the list of those interested in a time interval fine calibration module.
A question on that module, I don't see any ECL in the overlay, only HC, or is this only from the library used to create the pcb but you're looking at using fast 3.3V logic equivalents such as the AUC family or AVC or ALVC? I ask as I would have thought the 8130A and 59992 suggested by the manual would have edges much faster than HC logic? Would that have any effect on the fine cal?
While I'm here I'm wondering if anyone can help with an issue I have with my 53132A.......
I was (am) looking at a homebrew 12.4GHz frequency extension and I tried to change a few eeprom values that have been posted in a yahoo group. In replacing the eeprom with a socket I managed to corrupt (erase?) the eeprom and lost all data. I was able to recal, except for the time interval fine, and when I put in the modified values to supposedly enable the 12.4GHz module divide ratio I still get a divide by 128 no matter what value is in the 3 eeprom locations. After I recal'd those 3 locations had FF in them.
So I'm thinking one of three things, 1) when I erased the eeprom I lost some critical info, or 2) not only does the eeprom need changing but the software also needs to change to enable the 12.4GHz module or 3) there's more to it than that and the fpga also needs to change, but that seems overkill to change the divide ratio from by 128 (for the 3ghz extender) to divide by 512 (for the 12.4ghz extender). I tried all four possible code0, code1 combinations and got two which gave divide by 128 and two which gave no channel 3.
Would anyone have an eeprom dump that they could share so I can compare what I have now to what a factory eeprom should look like? My unit is a no option unit from the factory.
Has anyone had success in incorporating a non factory 12.4Ghz extender that uses a divide by 512 ratio into a no option counter?
Thanks
Greg
-
Add me to the list, one board please.
-
A question on that module, I don't see any ECL in the overlay, only HC, or is this only from the library used to create the pcb but you're looking at using fast 3.3V logic equivalents such as the AUC family or AVC or ALVC? I ask as I would have thought the 8130A and 59992 suggested by the manual would have edges much faster than HC logic? Would that have any effect on the fine cal?
The output buffer is a 74AC244 and produces signals with a rise time of less than 1 ns (which gets degraded to around 2ns through the switch and cables). The actual rise time and duty cycle is not all that critical since the calibration firmware uses various combinations of the P1/P2, P2/P1, N1/N2, N2/N1 measurements to null out all the various offsets and differences.
The cal routine measure the signals at the zero crossing so the rise time is not all that critical as long as the signals are stable over the measurement interval.
This HP product note describes how it is done. Note that the server seems to be slow and flakey lately:
https://rubidium.dyndns.org/~magnus/instruments/hp/J06-59992A/Better_than_100_ps_Accuracy_in_HP_5370B_Time_Interval_Measurements_Through_Bias_Error_Reduction.pdf
This is the manual for the J06-59992A:
https://rubidium.dyndns.org/~magnus/instruments/hp/J06-59992A/J06-59992A.pdf
And the patent:
https://rubidium.dyndns.org/~magnus/instruments/hp/J06-59992A/US4627268.pdf
-
An update on the 531xx calibrator board...
I have sent the boards for the next version out to the fab house. Should be here in a couple of weeks (although the last ones took over three weeks to arrive, they normally take 7-8 days). Solder paste stencil was shipped today.
The new version fixes the rotary switch footprint issue. Also, the boards are now 1mm thick with a power connection flood fill on the top. This allows traces with a 50 ohm impedance to be used. I also tweaked some component values.
I have all the parts in and have lined up a person to do the assembly... he needs to build some first to determine the price.
Attached is a photo of the current board (with switch bodge wires). Ignore the 20 MHz DIP oscillator on the board. That is a population option that allows for stand-alone operation without an external 10 MHz signal. Those DIP oscillators are not the most stable things. The counter does pass calibration with it, though.
The board runs on 6V - 12V. I'm including a 9V battery snap/barrel connector. The channel A/B BNC connectors are spaced to match the counter input connectors so that you can use male-male adapters instead of BNC cables. If your counter has the OCXO option, using the counter ref output as the calibrator input seems to well. The ref output does have a slight duty cycle bias, but the calibration routine nulls that out. If your counter uses the ultra-super-crappy-utterly-useless standard oscillator, don't bother calibrating the thing.
-
I would also like a board, if available!
-
The channel A/B BNC connectors are spaced to match the counter input connectors so that you can use male-male adapters instead of BNC cables.
What is the correct way to perform a CAL:TI FINE?
1. Use short connections for A and B. This "calibrates' the instrument itself.
2. Use the actual cables to be used for a measurement and calibrate for each test setup. This will also calibrate the cables compensating for different lengths etc.
-
I would also like a board, if available!
-
The channel A/B BNC connectors are spaced to match the counter input connectors so that you can use male-male adapters instead of BNC cables.
What is the correct way to perform a CAL:TI FINE?
1. Use short connections for A and B. This "calibrates' the instrument itself.
2. Use the actual cables to be used for a measurement and calibrate for each test setup. This will also calibrate the cables compensating for different lengths etc.
For "1." the length of the cables should not matter. The cable length gets calibrated out. I do see a little degradation in rise time on a scope due to cable capacitance, but it does not seem to matter.
I suspect "2." could give more accurate measurements, but may not be worthwhile unless you are trying to do super critical measurements. The time interval calibration is more oriented towards removing instrument biases. If you are measuring the time interval between channel A and channel B, the interval at the end of the cables should be the same regardless of their lengths. The cable length would affect when the signals arrive at the counter, but not appreciably affect the time difference between them. The 53132A has a 150 ps resolution and there could be some cable effects that affect the time interval at that level.
-
I guess "2" is only worthwhile if A is generated say 10 meters away and B is local and you only have one long coax.
-
The board house has said the new boards have shipped DHL express.
-
Bare boards have arrived... they look good.
Now to build one up and see if the new layout and 1mm board thickness / 50 ohm signal traces make any measurable difference.
-
Now to build one up and see if the new layout and 1mm board thickness / 50 ohm signal traces make any measurable difference.
Oh, yeah... it made a difference. A horrible difference. Massive ringing on the outputs. Massive coupling of crapitude onto the precision 5.000V gain cal signal and elsewhere. Damn... looks like yet another spin of the board... back to the original layout with only the switch connections fixed. I'm also changing the 50 ohm input terminator to a 1 watt one. We'll see how it turns out in a couple of weeks...
-
I have send the next rev of the calibrator off to the fab. I replaced the output driver chip with a 74ACT245. This allowed for a MUCH cleaner layout with shorter traces and better power distribution. Also, all the SMT passives are oriented in the same direction which will make the assembler happier.
-
If there is a spare kit, please count me in.
Thanks, Rob.
-
If there is a spare kit, please count me in.
I've got boards and parts to build 50 units...
-
Outstanding. Please let me know how much plus shipping to 98056 and I'll shoot that over to your PayPal account.
Looking forward to it.
TonyG
-
Outstanding. Please let me know how much plus shipping to 98056 and I'll shoot that over to your PayPal account.
The new boards should be here in a couple of weeks... hopefully.
I'll know more about the price when I get a new quote from a different assembly house. The one that built my Symmerticom X72 breakout boards screwed me royally (quoted soldering the through hole parts and then refused to do them... I spend all weekend soldering a BUNCH of connectors and de-fluxing the boards).
I'm still shooting for something around $30 assembled.
-
New PC boards have been shipped. Should be in early next week. All the new parts and a stainless SMT stencil have arrived.
-
New PC boards have been shipped. Should be in early next week. All the new parts and a stainless SMT stencil have arrived.
New layout boards arrived today. I built one up and it seems to be working well. I'm still trying to get a non-stupid quote from some assembly houses.
I thought I had a problem because it was failing to calibrate the counter. I then checked my older rev boards and those were failing too. Turned out to be a coax with a broken wire... now I remember why I colored that BNC with a red marker. |O Enough signal was getting through the gap to trigger the input stage and the outputs looked stable on a scope but were miserable when evaluated by an ADEV analysis.
-
I have two more 53132A for repair and I am looking forward to your finished product.
-
I did a test of the newer calibrator board signal stability. I fed the input with the 10 MHz sine wave from an HP-5071A cesium beam oscillator and fed the output to a TAPR TICC. I used Lady Heather to calculate the xDEVs of the output signals. At 10,000 seconds the ADEV value was in the 6E-15 range... not too shabby!
The outputs do have ringing on them, but the counter's calibration routine uses the zero crossing point which is quite stable.
The HP docs say for the gain calibration to input a 5.000V +/- 1mV signal. The LM4040A voltage reference chips that I am using seem to be within +/- 2 mV. Should not be an issue since the A/D and D/A chips in the counter are not that accurate. You can always use your own voltage source for the gain cal.
-
Just wanted to check in and see how it is going?
TonyG
-
Just wanted to check in and see how it is going?
I have the latest PCBs in from China. I built one up and it seems to working well. I sent it to TVB for his very expert evaluation. He is quite busy this month. I am currently looking for somebody to assemble them for a reasonable price.
-
Just picked up a 53132A with the OEM installed options 10 and 30. Interested in both doing a firmware update and a full cal, so put me on the list for one of the cal boards please.
-
I'm still waiting for TVB to do some higher end testing of the completed board. He has been tied up on a long-ish project for the last several weeks. That is apparently completed and hopefully he will be able to get to the board soon.
-
Put me on the list as well. No rush ;)
-
I just came across this while searching for a J06-59992A. Is your board available at this time? I'm transferring some 12GHz boards from a 53132A and am having to do a complete recalibration of the unit.
-
I just came across this while searching for a J06-59992A. Is your board available at this time? I'm transferring some 12GHz boards from a 53132A and am having to do a complete recalibration of the unit.
Just reposting my interest as well - Those J06-59992A seem to be as rare as hen's teeth.
DF20Bug - I hear you're shutting down the Texas facility and retiring (was talking to a friend in Denmark that mentioned it) - Enjoy the retirement it was great doing business with you previously.
TonyG
-
I have the bare PCBs and parts but have been waiting for TVB to do some testing. Also, I've been tied up (and out of town) on the Neverending Project From Hell for most of this summer.
I hope to these built by the end of August.
-
I have the bare PCBs and parts but have been waiting for TVB to do some testing. Also, I've been tied up (and out of town) on the Neverending Project From Hell for most of this summer.
I hope to these built by the end of August.
Hope this thread will be still read ???
@texaspyro
If you have finished the boards I would take two of them...
Regards Chris
-
I have the bare PCBs and parts but have been waiting for TVB to do some testing. Also, I've been tied up (and out of town) on the Neverending Project From Hell for most of this summer.
I hope to these built by the end of August.
Hate to resurrect this thread but where you able to make any progress on these calibrator boards?
While I certainly appreciate the time you've put into the development, can't it (or the obsolete J06-59992A) be replaced with a modern waveform generator as long as you have a stable 10MHz & 5V source?
-
Hi,
I am very intersted in a populated board.
My age, hands and eyes are not what they used to be.
Please add my name to the pist.
Thanks, Frans
-
I am also interested in a populated board.
Add me to the list, one board please, if available.
Thank you very much
-
Same here as well. Thank you.
BTW, can you simulate the 59992a output with dual output signal generator?
-
I used a dual output gen to do that stage of the the cal, but can't verify if it was better or not. It did pass fine though.
-
Good to know, Steve. Thanks.
What settings did you use for modes 1 through 4?
-
Hi mate, just wondering if the calibration kit is still available?
Thanks.