Welcome to ham-interfaces.com
High‑Performance Racal Dana 19xx OCXO Upgrade
It’s 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 were the older 9913 and 9916 units, but my favourite was always the 1998 frequency counter, and I had long wanted one of my own.
A few years ago, I spotted a Racal Dana 1998 frequency counter for sale on eBay. It appeared to be in good cosmetic condition but was reported to suffer from the usual stuck‑button problem. Knowing that these buttons are notoriously difficult to source (although near‑equivalents do exist), I decided to take a different approach.
Around the same time, I came across someone selling parts from a 1998 counter, including the front‑panel PCB. Thinking I might be able to exchange working and non‑working buttons, I purchased the spare board.
When the counter arrived, it was covered in the usual asset and “out‑of‑date” calibration stickers and was a little dirty, but fundamentally sound. After a careful clean and some IPA to remove residual adhesive, what emerged was a counter in near‑mint cosmetic condition — a real find.
As described, however, almost every front‑panel button was jammed or non‑functional. This was no surprise.
The replacement front‑panel board, on the other hand, had perfectly working buttons throughout, all with a positive and consistent action. The only visible difference was that the failed buttons had black bodies, while those on the replacement board were white. Otherwise, they appeared identical. Had the supplier updated the design? Had I simply been lucky and acquired a newer production board? Time will tell.
In practice, it proved to be a simple case of swapping the boards over. My £25 spare from eBay worked first time, restoring the counter to full operation. The original board has been retained as a donor for future failures.
With the mechanical issues resolved, attention naturally turned to the counter’s 10 MHz timebase, which ultimately led to the development of the OCXO replacement module described below.
Over the past few years, the counter has been used regularly in my workshop and has provided sterling service. However, it quickly became apparent that the internal 10 MHz reference was the limiting factor. To make any reasonable measurements, the counter had to be externally locked to my Datum GPS‑disciplined 10 MHz reference.
A quick search revealed that Racal Dana offered a number of internal 10 MHz reference options for the 19xx series counters:
-
04A Ovened Oscillator – 3×10⁻⁹
-
04B High Stability Ovened Oscillator – 5×10⁻¹⁰
-
04C Un‑ovened Oscillator – 1×10⁻⁶
-
04E Ultra High Stability Ovened Oscillator – 5×10⁻¹⁰
-
04R Rubidium Oscillator – 5×10⁻¹¹
-
04STD Standard Crystal Timebase – 2×10⁻⁶
-
04T TCXO Timebase – 3×10⁻⁷
So which option was fitted to my counter? It turned out to be the 04C, a 1×10⁻⁶ un‑ovened crystal oscillator (pictured below). In practice, it was very disappointing. Adjustment using the trimmer capacitor was extremely fiddly, and even achieving the nominal 1×10⁻⁶ specification required a steady hand. Worse still, it would not hold adjustment for long.
Given that experience, one can only imagine how poor the cheaper 04STD option must have been.
This limitation ultimately led to the decision to design a replacement internal OCXO module for the Racal Dana 19xx series counters.
This got me thinking: could I design my own internal reference board that approached the performance of the Racal Dana Option 04A ovened oscillator, specified at approximately 3×10⁻⁹ stability?
Around that time, I was reminded of one of Tony Albus's YouTube videos in which he compared the performance of several surplus and ex‑equipment OCXOs. Tony’s work is well worth exploring; he consistently produces high‑quality content focused on frequency and time metrology. The particular video that influenced this work can be found here.
Based on Tony’s measurements and observations, I decided to investigate the CTi OSC5A2B02 OCXO. One of the reasons for choosing this specific variant was that the original Racal Dana reference module outputs a somewhat square‑wave‑like signal rather than a clean sine wave, and the CTi device produces a similar amplitude and waveform, making electrical substitution more straightforward.
The manufacturer’s specification for the CTi OSC5A2B02 OCXO module is shown below and is available separately as a datasheet.: 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 200mA. 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 (seen below) 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.
On paper, the CTi device appeared capable of meeting both the performance and compatibility requirements for a drop‑in Racal Dana replacement.
My next step was to investigate how far the CTi OSC5A2B02 OCXO could be electrically trimmed. The device includes a Vref control pin which, according to the manufacturer’s datasheet, can be adjusted between 0 V and 4 V DC to fine‑trim the 10 MHz output frequency.
Across a batch of five OCXOs, the required Vref voltage varied between 1.8 V and 2.2 V (depending on the individual module) in order to bring the output frequency into alignment with the 10 MHz reference from my Datum GPS‑disciplined oscillator.Selecting one OCXO from roughly the middle of this group, I connected it to my TinyPFA frequency and phase analyser to examine its trim‑voltage characteristics in more detail.
For this particular device, achieving close to zero frequency offset relative to the GPSDO required a Vref setting of 1.916 V.I then applied step changes of ±10 mV and ±100 mV to the trim voltage and measured the resulting frequency deviation.
The results of these measurements are shown below.This confirmed that the OCXO offered sufficient trim range and resolution to be used as a practical replacement for the original Racal Dana timebase.
Ignoring the absolute values — which will naturally vary from module to module — a useful rule of thumb emerges: the output frequency shifts by approximately 10 mHz per millivolt of Vref adjustment. This is helpful guidance for anyone considering these OCXOs for their own projects.
This immediately sets some practical constraints on the reference‑voltage circuitry used to tune the modules.
However, Vref stability is not the only consideration. How susceptible is the OCXO to variations in its supply voltage?
The oscillator modules are powered from the counter’s 5 V rail, derived from the main power supply. The datasheet suggests that the output frequency remains within ±2×10⁻⁹ for a ±5 % supply variation, assuming the Vref pin is held constant. To verify this behaviour, I stabilised an OCXO and then used TimeLab together with my phase and frequency analyser to observe frequency changes as the supply voltage was varied.
Please excuse the frequency‑difference chart below — it is somewhat busy due to the annotations.
The test was carried out over 1000 seconds. Starting at 5.0 V, the supply was stepped down to 4.75 V, returned to 5.0 V, raised to 5.25 V, and finally brought back to 5.0 V. Throughout the test, the Vref pin was held constant at 1.916 V.
Following each supply‑voltage change, the OCXO exhibited a settling period of approximately 90 seconds. Over the entire test, the observed peak‑to‑peak frequency variation was only 22 mHz, indicating that supply‑voltage sensitivity is very small.
From these measurements, a +0.25 V change in supply voltage resulted in a frequency shift of approximately +4×10⁻¹⁰ after settling, while a –0.25 V change produced a shift of around –7×10⁻¹⁰. Both figures are comfortably within the device’s specified tolerance of ±2×10⁻⁹.
These results suggest that careful design of the Vref circuitry is more critical than aggressive regulation of the OCXO supply rail.
To assess short‑term frequency stability, I allowed the OCXO module to run for several hours and then trimmed the reference voltage so that it was aligned with my GPS‑disciplined oscillator (GPSDO). Using TimeLab, I then monitored the frequency and phase drift over a period of one hour.
Over the duration of the test, the OCXO exhibited some short‑term wander in the frequency‑difference measurement. However, this was very small, with a peak‑to‑peak variation of just 5.15 mHz. Applying a trend fit to the data showed that the average drift over the full hour was only 0.5 mHz.This level of stability — yes, 0.5 millihertz — is remarkably good for such an inexpensive OCXO module.
For context, this places the short‑term stability firmly in the same class as Racal’s higher‑grade internal options.
Another way to view the same behaviour is to measure the phase difference between the OCXO and the GPS‑disciplined reference over time. In the measurement shown below, the result is zero‑based, meaning that at T = 0 the two signals are treated as being in phase. The phase analyser then measures the accumulated phase change over the one‑hour test period.
In this case, the total phase change over the hour was 727 ns. Expressed as a slope, this corresponds to a fractional time drift of approximately –2×10⁻¹⁰ s/s.
This aligns well with the frequency‑domain measurements and confirms the excellent short‑term stability of the OCXO once trimmed and settled.
Expressed another way, a phase drift of just 727 ns over an hour corresponds to a remarkably small –2×10⁻¹⁰ s/s, which is exceptionally good performance for a low‑cost OCXO module.
The next task was to install the OCXO into the Racal Dana 19xx‑series counters using the same physical form factor as the original crystal reference modules.
To achieve performance close to that measured above, the board design cannot rely directly on the counter’s 5 V supply rail to generate the frequency‑trim reference voltage. While the supply itself is reasonably stable, it is not sufficiently well‑controlled for fine frequency trimming at the millihertz level.
As a result, the Vref frequency‑adjustment voltage needs to be both very stable and easily adjustable. I therefore based the design around the Diodes Incorporated AS431 three‑terminal adjustable regulator, selecting the higher‑grade variant with a 0.5 % voltage tolerance. More importantly, this device exhibits excellent temperature stability. Over the expected internal operating range of the counter — approximately 20 °C to 60 °C — the AS431’s 2.5 V reference shifts by only around 1 mV, or 0.04 %.
From earlier testing of the OCXO samples, it was clear that a Vref range of 1.5 V to 2.5 V provided more than enough adjustment margin to trim any of the oscillators.
Although the OSC5A2B02 datasheet specifies that the Vref input should not exceed 4 V, practical testing showed that values well below this limit were sufficient. Based on this, I settled on the circuit shown below. R8, a 2 kΩ Bourns 25‑turn trimmer, provides ample adjustment range together with the resolution needed for precise frequency alignment.
Taking into account both the temperature‑drift characteristics of the regulator and the limited tuning range used, the expected variation on the OCXO’s Vref pin between 20 °C and 60 °C is approximately 0.4 mV. Using the earlier rule of thumb, this corresponds to a frequency variation of about 4 mHz, or 4×10⁻¹⁰, due solely to internal temperature changes within the counter.
From this point on, the PCB design itself was fairly straightforward. The original Racal oscillator module uses a 5‑way, 0.1″ pitch board‑to‑board connector, which is a standard MultiComp Pro part available from Farnell. The replacement board therefore only required minor additions, chiefly some extra 5 V supply decoupling at the PCB input.
Testing of the OSC5A2B02 OCXO confirmed that its HCMOS/TTL‑level output was entirely suitable for driving the clock‑conditioning input of the Racal counter without any additional buffering.
A 3D render of the PCB (front and back) is shown below. You will notice that this board is marked Revision 1.2. During the original development revision, a user‑contributed footprint was used for the OSC5A2B02 OCXO, and the pin spacing proved to be incorrect. In addition, the position of the frequency‑adjustment trimmer needed to be shifted slightly to better align with the access hole in the rear panel of the counter.
Aside from these small mechanical corrections, the electrical performance of Revision 1.1 was already meeting expectations.
So what does the finished product look like?
Shown below are the front and rear views of the production version of the board. It is a simple plug‑in replacement for the original crystal oscillator and TCXO reference modules fitted to the Racal Dana 19xx series counters.
And finally, the finished OCXO replacement module installed in my Racal Dana 1998 counter.
The final question, then, is how good the OCXO is when fully settled and aligned in the counter?
The board was soak‑tested for 24 hours installed in my Racal Dana 1998 counter, after which it was aligned against my known‑good 10 MHz reference derived from a Datum / Symmetricom ET6000 GPS‑disciplined oscillator. After the soak period, the OCXO was trimmed such that, with a 20‑second integration time, the measured frequency was 10.000 000 000 MHz.
Observed over an eight‑day period, the measured frequency slowly drifted down to 9.999 999 990 MHz, corresponding to a total change of 10 mHz over that time.
Comparing the counter’s reference output directly to the 10 MHz source on an oscilloscope showed a phase drift of approximately 30 ns over 30 seconds, equivalent to 1 ns/s. This corresponds to a fractional frequency error of around 1×10⁻⁹, which is considerably better than the original target specification.
According to the manufacturer’s datasheet, the specified crystal ageing is better than 0.5 ppb per day. For a 10 MHz oscillator, this equates to 5 mHz per day, or 40 mHz over eight days.
The observed drift of 10 mHz over the same period is therefore more than four times better than the stated ageing specification.
In reality, the measured behaviour is almost certainly a combination of crystal ageing, small temperature variations, and residual environmental effects. Nevertheless, the overall performance comfortably exceeds expectations.
Availability
If you would like to purchase one of our Racal Dana 19xx counter/timer OCXO replacement boards, they are available from the Accessories Shop section of this site.
Fitting Instructions
Fitting is straightforward and typically takes around 10 minutes. All that is required is a small Posi‑Drive / Phillips screwdriver. Refer to the photos above to see the correct installation position inside the counter.
-
Unplug the counter from the mains supply.
-
Remove the rear bezel (two screws).
Note the orientation of the bezel, as it will only refit correctly one way. -
Slide the outer case off the counter.
-
The original reference PCB is mounted using a single fixing screw and plugs into a 5‑pin header on the main PCB.
Remove the screw and carefully withdraw the original reference board, taking care not to lose the screw and washer, as these will be reused. -
Install the new OCXO reference PCB as shown in the photos above and secure it using the original screw.
Ensure that the rear‑mounted adjustment trimmer is facing towards the rear panel. Take care to align all pins correctly between the main board and the socket on the OCXO PCB before fully seating the module. -
Reassemble the case and refit the rear bezel.
Each OCXO PCB is soak‑tested for 24 hours and aligned against our GNSS‑disciplined 10 MHz reference before shipping.