EEVblog #647 - Agilent 53131A Frequency Counter Oven Upgrade

[EEVblog]

Dave upgrades his HP/Agilent 53131A frequency counter with an ovenised oscillator module from ebay.
http://www.ebay.com.au/itm/121330209846
http://www.thinksrs.com/downloads/PDFs/Catalog/PRS10c.pdf
http://www.2917.com/EBAY-images/NOX012166/STP2145A.pdf
http://cp.literature.agilent.com/litweb/pdf/5967-6039EN.pdf

[Vgkid]

I haven't watched this video yet. When watching the gravity vid I figured that you upgraded the timebase as yours were non ocxo oscillators.

[coppice]

Ovenized oscillators in frequency counters are a mixed blessing. They take so long to warm up and get the frequency right, and many of them don't even inform you when they are up to temperature. They are pretty much OK on a bench, where you leave them on all day. The are a pain when you carry them around.

The non-ovenised oscillator isn't the garbage you make it out to be. Many times the user doesn't care about the exact frequency, but only about jitter. A simple well designed crystal oscilator can have very low phase noise, and microphony can be pretty low with the right crystal. If the oscillator is pure you will get a stable measurement result to lots of digits, and that's all most measurements need - a stable result you can watch move around by small amounts as you perform experiments on your circuit. The need for an exact frequency measurement is actually quite a rare thing in the overall use of frequency counters. That's why most frequency counters sell with the simple crystal option.

Have you noticed how few labs these days even have a frequency counter? I go into labs with piles of expensive equipment of various types, but if I ask for a frequency counter they probably can't find one.

[dah5477]

Nice timing, Dave (no pun intended) - I actually just had some OCXO boards made for the HP 3325A/B function generators which I hope to start selling soon (still waiting for PCBs/parts to arrive).

That OCXO you have appears to be a Morion brand one (made in Russia!), specifically an MV89A or MV89B.  The "A" models differ in that they actually use a 5 MHz crystal with a frequency doubler to get them to 10 MHz (the attached image shows the 5 MHz harmonics within the crystal's output spectrum).  Both also use a double oven design.  Here's some more detailed info on the MV89: http://www.ke5fx.com/Morion_MV89A_IEEE_IFCS_2002.pdf

For the past couple days, I've been monitoring the sample MV89A I have along with a similar Trimble-branded OCXO, and also with an original OCXO board from the HP 3325B.  Both the Morion and Trimble seem to be a bit more stable in the short term against the 3325B (or at least have a lower tempco), however they may be drifting in the long-run faster - but that could be from me leaving the frequency adjust pin floating, as I only just found a trimpot to tune them a short while ago.  Unfortunately, I only have 0.001 Hz resolution, but for now it's good enough to measure drift.  If anyone is interested, I attached the data I have collected so far which are of the offsets from 10 MHz at 10 sec intervals for ~16 hrs (note: the Morion and Trimble OCXOs weren't tuned exactly at the time).

[EEVblog]

I haven't watched this video yet. When watching the gravity vid I figured that you upgraded the timebase as yours were non ocxo oscillators.

Yeah, the gravity thing was just a side comment I added to this video, but during editing I thought it was worthy of it's own video, so uploaded that first.

[bktemp]

It could be interesting to measure the tempco of the unit. I expect it beeing much worse than the oscillator specs itself because of the bad tempco of those cheap trimmers used to set the frequency. The dac option could be more stable.

[EEVblog]

It could be interesting to measure the tempco of the unit. I expect it beeing much worse than the oscillator specs itself because of the bad tempco of those cheap trimmers used to set the frequency. The dac option could be more stable.

The trimmer is just a supply divider, so tempco shouldn't come into it too much. You don't get one side of the trimmer magically higher temp than the other side, so the divider ratio stays pretty constant.

[JonnyBoats]

If anyone is interested in an alternate aftermarket source for a similar board see: http://gerrysweeney.com/update-diy-hpagilent-53131a-010-high-stability-timebase-option-pcbs-available/

Gerry Sweeney in the UK developed a project, and board, that can use various OCXOs and one can buy either a bare board or an assembled unit from him. Perhaps more importantly he blogged about the project, provided videos, and published full schematics and parts lists.

By way of disclosure, I never met him but did buy a bare board from him. I have not yet assembled it.

[max_torque]

With the caveat that i am not a proper time junkie, i'd suggest that you can't actually determine much about this new OCXO from the minimum data taken in the video, other than it IS possible to trim it to a "one off" high accuracy representation of the 10MHz signal from your Rb clock?

I.e. all the figures on the data sheet for short/long term and Temp stability are just that, figures on a datasheet, that may not apply to the actual device you have fitted to your Freq counter?

Surely,  before you can award the new OCXO a full "bobby dazzler" award, you need to do some statistical measurement and some long term repeatability measurement??

[Bengt]

Just a quick note, the STP2145A is not made by Morion, but Rakon/C-MAC. You can get a spec sheet here: http://www.febo.com/pipermail/time-nuts/2011-March/055144.html

Oh, and the LED goes on when the temperature regulator first hits its setpoint, it takes a few minutes from the time the LED goes on until the power consumption has stabilized.

/Bengt

[EEVblog]

With the caveat that i am not a proper time junkie, i'd suggest that you can't actually determine much about this new OCXO from the minimum data taken in the video, other than it IS possible to trim it to a "one off" high accuracy representation of the 10MHz signal from your Rb clock?
I.e. all the figures on the data sheet for short/long term and Temp stability are just that, figures on a datasheet, that may not apply to the actual device you have fitted to your Freq counter?
Surely,  before you can award the new OCXO a full "bobby dazzler" award, you need to do some statistical measurement and some long term repeatability measurement??

Of course, but unless it' faulty it's going to be streets ahead of the internal oscillator which was the entire point. I have little doubt it will meet it's spec.
Yes, further testing required.

[gerrysweeney]

All,

I noticed that on the board Dave used the calibration was done with a trimpot, what what I did not see on that board was a temperature stable voltage reference which is really needed IMHO, on the many OCXO's that I tested some had an alleged stable reference voltage output but I found the output impedance of this to be insufficient to drive the Fcontrol pin so would need buffering.  I noticed that the trimmer on the board Dave used seems very touchy, you can achieve much better precision than that with the right circuitry, my meter will consistently calibrate to less than ten one thousandths of one cycle at 10Mhz using the front panel calibration capability. 

In the end I used the original circuitry for calibration so you can cal off the front panel.  All of the information I found about the various OCXO's, tests I done and so on are all documented in the following articles along with some video content too. I also built a one-off OCXO upgrade for a Racal Dana 1999 meter where I used a voltage reference and a trimpot, again with lots of testing and stability validation, I have included the relevant links to the articles/videos below for anyone who is interested. I also had some PCB's made which are available if anyone wants them, I have sold quite a few but have quite a few left. I also have some ready made and tested boards if anyone wants one, I have sold more than 60 of these too so they ended up being more interesting to people than I expected. 

Original Article
http://gerrysweeney.com/diy-hpagilent-53131a-010-high-stability-timebase-option/

Follow-Up Article (get bare boards and fully assembled ones)
http://gerrysweeney.com/update-diy-hpagilent-53131a-010-high-stability-timebase-option-pcbs-available/

Racal Dana 1999 Upgrade
http://gerrysweeney.com/racal-dana-199x-diy-high-stability-diy-timebase-hack-for-under-25/

Thanks @JohnnyBoats for posting a link back to my project.

Gerry

[David Hess]

I noticed that on the board Dave used the calibration was done with a trimpot, what what I did not see on that board was a temperature stable voltage reference which is really needed IMHO, on the many OCXO's that I tested some had an alleged stable reference voltage output but I found the output impedance of this to be insufficient to drive the Fcontrol pin so would need buffering.  I noticed that the trimmer on the board Dave used seems very touchy, you can achieve much better precision than that with the right circuitry, my meter will consistently calibrate to less than ten one thousandths of one cycle at 10Mhz using the front panel calibration capability. 

Even multi-turn trimmers have poor stability compared to fixed resistive dividers.  A great improvement in stability may be had by restricting the trim range.  Appendix C on page 29 of Linear Technology application note 42 has a good discussion about trimming techniques:

http://www.linear.com/docs/4133

Is there any documentation about how the Morion MV89 OCXO Uin and Uref pins are suppose to be used?  With a reference output I would think that the control voltage is ratiometric.

[bktemp]

The trimmer is just a supply divider, so tempco shouldn't come into it too much. You don't get one side of the trimmer magically higher temp than the other side, so the divider ratio stays pretty constant.

I thought the same before I did some quick&dirty measurements using a 24bit adc: I adjusted a trimmer to about half the adc reference voltage and started the measurement. After a few days I had a curve that showed exactly the trend of  the room temperature. I could not belive the bad tempco and reversed the trimmer pins. Now the curve went into the opposite direction. The estimated tempco was about 50ppm/°C. I think it has a mechanical cause.
In addition, the lead screw types are quite sensitive to flexing the pcb, the worm gear type used for the frequency adjustment is much more immune, because of its smaller footprint.
I could not find any data on the adjustment range of the frequency, but let's assume +/-10ppm.  This gives a total tempco of 1E-9. This is one order of magnitude worse than the oscillator itself.

[max_torque]

With the caveat that i am not a proper time junkie, i'd suggest that you can't actually determine much about this new OCXO from the minimum data taken in the video, other than it IS possible to trim it to a "one off" high accuracy representation of the 10MHz signal from your Rb clock?
I.e. all the figures on the data sheet for short/long term and Temp stability are just that, figures on a datasheet, that may not apply to the actual device you have fitted to your Freq counter?
Surely,  before you can award the new OCXO a full "bobby dazzler" award, you need to do some statistical measurement and some long term repeatability measurement??

Of course, but unless it' faulty it's going to be streets ahead of the internal oscillator which was the entire point. I have little doubt it will meet it's spec.
Yes, further testing required.

The one thing that puts me off "upgrading" my Freq Counter to an OCXO is the necessary "wait" after power on.  I can think of loads of times i just want a quick single measurement, and unfortunately i haven't got the shelf/bench space to leave the thing out and switched on the whole time.
 Could you measure the "accuracy" vs time after switch on?  Even if you just write a few numbers on a piece of paper say every 30secs?  Be interesting to see how long you need to wait to get a better result that the build in oscillator??

[edpalmer42]

The one thing that puts me off "upgrading" my Freq Counter to an OCXO is the necessary "wait" after power on.  I can think of loads of times i just want a quick single measurement, and unfortunately i haven't got the shelf/bench space to leave the thing out and switched on the whole time.
 Could you measure the "accuracy" vs time after switch on?  Even if you just write a few numbers on a piece of paper say every 30secs?  Be interesting to see how long you need to wait to get a better result that the build in oscillator??

I have a Racal-Dana 1992 counter.  It's not as good as the Agilent 53131 or 53132 counters, but can measure 10 MHz to .001 Hz with a 10 second gate time.  I wasn't happy with the performance of the internal OCXO.  Even though the oscillator was running continuously, the readings drifted for about 30-60 minutes after I turned the counter on.  So I set up a stand-alone Rubidium as a house frequency standard and used it as the external reference.  I can now turn on the counter and immediately make measurements that are stable to .001 Hz.  No warmup is required.  Dave's double oven oscillator would probably react better, but for me, the external reference is the only thing I ever use.  I'm actually considering downgrading the internal OCXO to a TCXO to reduce the power drain.  I can't remove it completely because the counter requires an internal clock to enable it to recognize the presence of the external clock and switch over to it.

Ed

[David Hess]

I have the 1992 with the internal TCXO and never gave the warmup time a thought because it occurs fast enough that I have never been able to easily observe it so I will have to reconsider updating it with an OCXO.  My HP5315A is similarly fast with its TCXO.  My older Tektronix counters which have unexceptional OCXOs do have 15 or 30 minutes of warm up time but they are pretty low performance.

[craigh]

Thanks for another interesting video.  Say, what do you need that kind of accuracy for?

Back in another era, I used to re-channel and tune SSB tube (valve) radios used on ships at sea.  I had to wait for the oven in the radio to cool down before I could open it and change the crystal (one crystal per channel).  It was way to hot to handle while at operating temperature.  Then I had to wait again for it to come back up to operating temperature before I could start working on the radio.

I used an older HP counter (don't recall the model) to check and adjust the radio's crystal frequency (a trimmer cap for each channel).  That old counter sat on the bench always plugged in.  If I remember correctly, it took over an hour to stabilize if it had been shut off for a while.

[German_EE]

I use something similar to Dave's module as my workshop frequency standard, however:

1) The module is fed with a power source that is first regulated down to 18V and then down to 12V using low noise regulators.

2) Although ovened the entire unit is held in an expanded polystyrene 'brick' which is about 10cm x 10cm x 20cm. This protects from abrupt temperature changes and offers some protection from vibration.

3) The output is opto-coupled so that changes in load do not affect the frequency. Output amplifiers are on a separate supply.

4) After playing around with the 10-turn potentiometer I discovered that the best setting was about 6,5K on one leg and 3,5K on the other. I then replaced the potentiometer with a 6K ohm and 3K ohm resistor with a 1K potentiometer in the middle and adjustment is now much easier.

The oscillator is run 24/7 and once a month I calibrate it to DCF77 which is only 30 Km away from me, now that the unit has settled down adjustments are tiny and appear to be in one direction.

[bingo600]

I have a Racal 1991 that takes ext. 10Mhz ,and even the  Real ocxo in the draver.
I never bothered to install it , as i run all my instruments of a 10Mhz trimble tbolt.

I know Dave has a PRS-10 , and wonder why he's bothering.

My PM-6680 and 5370B that has quite Nice ocxo's in them are also run off the tbolt.

So i just ask why .. Spend the extra heating ??

/Bingo

[papo]

Hello

Oh, and the LED goes on when the temperature regulator first hits its setpoint, it takes a few minutes from the time the LED goes on until the power consumption has stabilized.

Is that the 'oven alarm' pin in the datasheet you linked? I tried to figure it out from the video but I don't get a good look at the board... I'm asking because I purchased two Trimble OCXOs some time ago for a similar purpose. Getting that Lock-LED would be nice, but I don't think this is available on the parts I have. I have only 5 pins anyway, and I figure the pins must be ground, then obviously the output, input, a reference that is used to derive a calibration voltage which is then fed into the fifth pin.

Also I'm not sure about the voltage range I am supposed to feed into the calibration pin (obviously between ground and the reference output, but in what range?) and I thought since you managed to not only recognize the OCXO Dave is using, but also provide a link to the datasheet, chances are that you know something about these parts and might be able to help, so it was worth a shot. The sticker says: Trimble, *0346-6945-T*, *34310-T*. fc = 10M000000.

Regards and thank you in any case (and Dave, great video as always!)

Matt

[EEVblog]

I know Dave has a PRS-10 , and wonder why he's bothering.

I thought I answered that in the video?
For standalone use when it's not convenient to use with the external rubidium standard.

[jahonen]

Apropos, here are the warmup transients of my FE-5680A rubidium reference and the OCXO which came with it.

Regards,
Janne

[gerrysweeney]

The one thing that puts me off "upgrading" my Freq Counter to an OCXO is the necessary "wait" after power on.  I can think of loads of times i just want a quick single measurement, and unfortunately i haven't got the shelf/bench space to leave the thing out and switched on the whole time.
 Could you measure the "accuracy" vs time after switch on?  Even if you just write a few numbers on a piece of paper say every 30secs?  Be interesting to see how long you need to wait to get a better result that the build in oscillator??

What I have found using the Trimble OCXO's is that you need to wait only 10 minutes to get 99.9% of the way there, its not that bad at all really. I would honestly say its worth it, without a doubt its worth it, all of my counters now have OCXO's in them, once you know the extreme difference between an ovenized OCXO and a non-ovenized one you would not go back I am sure

Gerry

[SeanB]

If you just want a quick measure you probably are only interested in a ballpark accuracy in any case, so the start up drift would not be a worry in any case. If you want the precision you will need to wait in any case for both the counter and the DUT to stabilise in any case.