Racal Dana Counter/Timer OCXO Upgrade Project

Its strange how you get attached to bits of equipment you used when you were a young engineer. In the BBC Transmitter Department, we had various flavours of Racal Dana Frequency counters for our field visits. Some some were the older 9913 and 9916 units but my favourite was the 1998 Frequency counter and I always wanted one.

So a few years ago I spotted a Racal Dana 1998 Frequency Counter for sale on eBay. it looked in cosmetically good condition but was reported to have the usual stuck buttons problem. Now knowing that the buttons are notoriously difficult to get hold of (you can buy a sort-of equivalent) I decided to take a different route for repair. I also spotted someone selling parts from 1998 counter including the front panel PCB. Thinking that I may be able to exchange working and not-working buttons, I purchased the spare front panel as well.

When the counter arrived, it was covered in the usual asset and "out of date" CAL stickers and it was a bit dirty but it looked good. A good clean and some IPA to get rid of the bits of glue, what was revealed was a counter which was in near mint cosmetic condition. A real find.

But, as described, just about every front panel button was jammed or not working. No surprise there.

However, the replacement front panel board had perfectly working buttons. All of them in perfect condition with a nice positive action. The only difference is the failed ones had black bodies and the ones on the separate board had white bodies. Otherwise, they looked the same. Had the button supplier updated the design? Had I "lucked out" and simply purchased a newer board and this one will ultimately fail, who knows? Time will tell. However, it was a simple case of swapping the boards over and my £25 spare from eBay just worked first time. At least I have the old board (below) as a spare for other failures.

Over the past few years, the counter has been used in my workshop providing sterling service. But I quickly realised that the internal 10MHz reference was dreadful. So I needed to use the counter externally locked to my Datum GPS locked 10MHz source to make any reasonable measurements.

A quick google revealed there are a number of options for the internal 10MHz reference for the Racal Dana 19xx series counters.

04A Ovened Oscillator 3x10 E-9
04B High Stability Ovened Oscillator 5x10 E-10
04C Unovened Oscillator 1x10 E-6
04E Ultra High Stability Ovened Oscillator 5x10 E-10
04R Rubidium Oscillator 5x10 E-11
04STD Standard Crystal Timebase 2x10 E-6
04T TCXO Timebase 3x10 E-7

So what option was fitted to mine? Well it turns out to be the "04C", a 1E-6 unovened crystal oscillator (pictured below). And it was just awful. Really difficult to adjust with the trimmer capacitor and getting to the 1E6 specification took a steady hand and it would not stay there long. Heaven only knows what the cheaper "04STD" was like!!!

This got me thinking. Can I make my own board with MUCH better performance at least matching the performance of the Racal Dana option "04A" ovened oscillator with a stability of 3E-9?

I then remembered one of Tony Albus's YouTube videos where he was comparing some ex-equipment OCXOs. Tony's excellent video can be found here.

As a result of Tony's good work I decided to take a look at the CTi OSC5A2B02 OCXO. The reason I chose this version is that the original Racal Dana module outputs a "sort of" square wave and this model of the CTI OCXO also outputs a similar voltage square wave.

The Specification for the CTi OSC5A2B02 OCXO module (below) is here: CTi OSC5A2B02 Specification PDF

I went ahead purchased a batch of 5 CTi OSC5A2B02 OCXO modules and tested them in turn. In agreement with Tony's work, they seem to draw about 450mA whilst warming up and stabilise at around 240mA. As the counters PSU is already designed for ovened oscillators and rubidium plug-ins, there is no problem with the counter powering these little OCXOs.

Looking at the output of the original "04C" crystal oscillator option, the output was a highly distorted "square" wave of about 4V amplitude. The CTi OSC5A2B02 OCXO module output claims to be HCMOS and gave a much better square wave output of about 5v.. The yellow scope trace below is the OSC5A2B02 output and the purple trace is my 10MHz from my Datum time source.

Next test was to see how far you could adjust the CTi OSC5A2B02 OCXO. They have a Vref pin which according to the specification sheet can be adjusted between 0v and 4v DC which "trims" the output 10Mhz output frequency.

Vref voltage of 0v should pull the frequency -1 to -2 ppm
Vref voltage of 4v should pull the frequency +1 to +2 ppm
Vref of around 2v should set the frequency -0.2 to +0.2ppm

On the batch of 5 OCXOs that I had, the Vref needed to be set anywhere between 1.8V and 2.2V (module depending) to get the OCXO onto frequency compared to my good GPS disciplined source. This sets some good rules for any reference voltage circuit to set the tuning of modules.

But having the Vref stable is not the only consideration. How susceptible are they to supply voltage? The oscillator modules take the nominally 5V from the main PSU in the counter. But does supply tolerance make a difference?

Well the specification sheet suggests the output is good to ±2E-9 for a 5% supply variation. Presumably this is whilst holding the Vref pin very accurately. After I stabilised a module, I used the persistence mode on my scope with it locked to my good workshop 10MHz. I moved the supply voltage ±0.25V to see how stable the output was. The trace phase moved 60ns in 60 seconds. So that is 1ns per second or ±1E-9 and so twice as good as the specification.

Ok what about frequency stability? The specification suggests that the OSC5A2B02 short term stability is ±0.05ppb/s one hour after powering up. And from my testing, these modules do need a good hour to stabilise. So with a stable 5V supply and a very stable DC feed to the Vref pin to bring one onto 10MHz, what is the short term drift?

Short term wander or I suppose jitter was the next measurement. The specification is ±0.05ppb/s or 0.05ns/s. Could I capture this on the scope? Well I trimmed the OCXO in as accurately as I could to 10MHz and so it was only drifting across my scope trace very slowly with respect to my locking 10MHz feed. I put a 30 second persistence on the trace. So the short term wander is the width of the persistence trace divided by 30. The resultant width of the trace was 1ns over 30 seconds and so you could infer this was 0.033ns/s and so again meeting specification.

So ultimately, as an OCXO to be used in the counter how good is it?

Well this depends on how stable you can keep the temperature of the OCXO inside the counter and the ageing over time. At 25 Degrees Centigrade the OCXO claims to be stable to 10PPB or 1E-8. The ageing rate claims to be 0.5PPB per day or 5E-10 per day. Or at 10MHz, 5mHz per day.

To ensure you get close to these figures, any board design cannot rely on the 5V supply in the counter. Therefore the Vref frequency adjustment needs to be very stable and of course adjustable. So I decided to base my design around the Diodes Incorporated AS431 three terminal adjustable regulator choosing the better specified part with 0.5% voltage tolerance. But more importantly, this device is very temperature stable. And so in the sort of operating temperature range that will be found in the counter (say 20 Degrees C to 60 Degrees C, the 2.5V reference in the chip only moves 1mV or 0.04%.

Also, from my testing on the oscillator samples I had on hand, 1.5V to 2.5V of tuning voltage range gave you more than enough range to trim any oscillator over a good range.

The specification of the OSC5A2B02 stated that the Vref voltage cannot exceed 4V. But from my testing, you only needed between 1.5 and 2.5V to trim the oscillators. I settled with the circuit below. R8 (a 2K Bourn's 25 turn trimmer) gave plenty of adjustment and the ability to fine-tune the oscillator frequency.

Otherwise the design of the PCB was very straightforward. The original oscillator PCB used a 5 way 0.1" pitch board-to-board socket which was was a standard MultiComp Pro part from Farnell. The board just needed a little extra 5V supply decoupling on the PCB input. Testing of the OSC5A2B02 OCXO showed that the HCMOS TTL output was perfect for driving the clock conditioning input of the Racal counter.

The 3D render of the PCB (front and back) can be seen below. You will note that this is Rev 1.2. With the original development revision of the board I had used a user contributed footprint for the OSC5A2B02 OCXO and the pin spacing was wrong. Also the positioning of the adjustment trimmer just needed moving to centre it in the adjustment hole in the rear of the counter. Otherwise, the performance was perfect from Revision 1.1.

So what does the finished product look like? Below is the front and rear board views of the version I have had produced. it is a simple plug in replacement for the original simple crystal and TCXO options.

And a what the finished product looks like installed into my Racal Dana 1998 counter.

I suppose the final question is how good is the OCXO when stable and aligned in my counter?

I soak tested the board for 24 hours in my 1998 counter and then fed my known good 10MHz feed in from my Datum/Symmetricom ET6000 GPS Disciplined source. After 24 hours the OCXO was trimmed such that on the 20 second integration time, the frequency was 10.000000000 Hz. Observing over an 8 day period, the measured frequency had slowly moved down to 9.999999990 MHz or a drift of 10mHz in 8 days.

Comparing the reference output of the counter to the known 10MHz source on an oscilloscope, at the time of measurement, the OCXO drift was 30ns in 30 seconds or a 1ns/s drift. This equates to an accuracy of 1E-9. This was considerably better than our target specification.

According to the datasheet, the claimed ageing of the crystal is to be better than 0.5PPB per day. So on this 10MHz oscillator, the ageing should be better than 5mHz per day or 40mHz over 8 days. We have seen just 10mHz over this period and so I think this is just crystal ageing and it is over four times better than the claimed specification. Of course the results will more likely be a combination of all of the above. Ageing, temperature variations and the like. However, ultimately the performance is significantly better than expected.

If you would like to buy one of our Racal 19xx counter/timer OCXO boards, they are available from our Accessories Shop Page here