Test Equipment Anonymous

This unit is about 40 years old and I've had it for 20 years. The frequency reference was getting tired and took hours to settle. Every time it was powered up, it eventually settled at a different frequency. It was within about 2ppm of where it should be. That's probably good enough for most things if a little irritating, and the main problem was the time it took to settle.

I decided to replace the OCXO reference with one based on an Aliexpress ex cell tower OCXO. The Racal reference is 5MHz and there's a frequency doubler arrangement to bring it to 10MHz. bd139 had a similar problem with a Racal counter and bypassed the doubler. I didn't want to mess with the main board of the counter and chose to divide the 10MHz down to 5MHz with a 7474. What happens with the 5MHz reference depends on the particular Racal counter; some double it to 10MHz, some divide by 5 to 1MHz. The output of the Racal references is a 1.5V peak to peak sine wave which goes through a single transistor amplifier to bring it to TTL levels. I couldn't see any problems with feeding a TTL level signal into it.

The PCB only carries the OCXO unit, the 7474 and a trimmer pot to adjust the frequency. It has holes drilled so it can be fixed in place of the Racal OCXO on standoffs and I had a bulkhead fitting 10K trim pot which could be fixed to the PCB and be adjustable through the hole at the back of the counter used to adjust the original reference.

With everything put back together the counter stabilised within about 10 minutes. It was fairly easy to adjust the frequency to agree with the GPSDO to within about 1 part in 10^7. Getting it spot on was like finding the sweet spot on a Cat's Whisker detector.

It settles down after no more than 10 minutes and always in the same place. Assuming the cheap GPSDO is perfect, or at least better than 1 ppb, the counter sits there for hours with the reading slowly varying within a range of 5ppb. It's a vast improvement.

Piccies:

The exquisitely crafted PCB.
The counter showing a 10MHz signal from the GPSDO using its maximum gate time.
Thoughts:

I'm very impressed with the very cheap, "previously enjoyed" OCXO.

I have two other Racal counters, a 9916 and a 9921. They settle to 1 part in 10^7 and don't wander much, but it takes hours for them to do it. I think I'll give them the same treatment. It needs a proper PCB with quality cermet multiturn trim pots, and maybe two to give coarse and fine adjustment, to avoid the safecracker fingers experience when adjusting it.

Maybe I need another GPSDO.

Nice job!

Thanks.

You made me aware of those £2.50 AliExpress ex cell tower OCXOs. These are CTI. I've been amazed at how good they are. The Racal 9906A with one fitted with a 7474 to halve the frequency to the required 5MHz, was set up to agree with a GPSDO, which has been left on for ages, and I assume is perfect for practical purposes. After about a week, it still agrees to within 5ppb. I took the adjustment voltage from the 5V supply line in the 9906A and didn't do anything to make it more stable or select a trimmer with a low tempco (it may have but it was out of something else).

What's more, it stabilises within about five minutes. The original Racal OCXOs in the other two counters are probably within the original spec, but they take hours to stablise and they appear to be nowhere near as good. They are going to get the CTI treatment.

Update.

After over a month, during which time the frequency counter has only been switched off overnight a couple of times, just to see how it reacted to being left to cool down and start from cold, it's still going strong. It warms up from cold to agreeing with the GPSDO in five minutes or so.

Mostly it shows the 10MHz signal from the GPSDO as 9,999,999.98 Hz. Occasionally the last digit drops to 6 or rises to 0. The fluctuations are very slow.

I'm impressed, because the adjustment voltage comes straight from the 5V supply of the instrument, with no extra filtering, and the preset pot was one I had in the collection and was taken out of something else, it wasn't specially chosen to have a low tempco.

I don't know whether that used OCXO was a particularly good one. I'm inclined to revisit this adding some refinements, and get some proper boards made up, with a view to replacing the OCXOs in my other Racal frequency counters.