I recently purchased an inexpensive 10MHz OCXO reference clock module on eBay for less than $30.
Above is the seller’s description of the module.
The performance of this product is very much dependent on the OCXO component on the board.
Above is the OCXO, a product branded CTI. The component above sells new for about $13 inc shipping on eBay at the time of writing, so the specs are pretty grand for a very low cost component.
There are also plenty of these cheats… “Brand new” but at the same time “used and tested”.
Above is a screenshot from tinyPFA comparing the Plaza 10MHz OCXO with a Trimble Thunderbolt GPSDO which has been ‘locked’ for a very long time.
Frequency adjustment is by a 25 turn (3296) potentiomter. The oscillator frequency was adjusted after an hour warm up time, and it could not be adjusted lower than 1.5Hz high, so it is quite unusable as an accurate 10MHz clock source.
The specification of ±0.01Hz is equivalent to 1 part in 10^9 error, ambitious but not unrealisable in a good OCXO. But, this one has an error of 150 parts in 10^9 at the best adjustment.
If we assume that the module hardware properly operates the OXCO module and supplies correct ‘tuning’ voltage range, then this behavior speaks to the quality of the underlying CTI OCXO module. Nothing on the PCB can overcome a low quality OCXO module.
Even though the OXCO cannot be adjusted to 10MHz, a study of its stability / jitter is interesting.
Above is an Allan Deviation plot, again compared to the Trimble Thunderbolt GPSDO. As far as it goes, it is not too bad, if you had one of these modules and it was on frequency, then it is not too bad at all.
Defects with the whole PCB module:
- it cannot be adjusted to 10MHz;
- SMA connectors need resoldering; and
- the regulator gets quite hot running on 12V.
Though it is rated for 7-13V, it may withstand a little more, limited by 16V capacitors and regulator dissipation.
There are many similar modules available online (eg eBay and Aliexpress), and many if not most used a CTI OCXO.
The experience would bias me against purchase of any module with a CTI OCXO on board. If they let this out of the CTI factory, then any purchase is a lottery.
Needless to say I have opened a return case on eBay.
PS: it is also quite possible the PCB is populated with second grade parts, either used, or rejected during testing… but letting substandard parts out of the factory damns the maker. If they are used, were they pulled because they have aged badly? All of these hint quality problems.
Update 10/01/2023:
Investigation reveals that the adjustment range of the tune voltage on the OCXO component is quite small, and in this case not enough to get the thing on frequency. Since this is too high, connecting a suitable resistor from tune pin to ground reduces the tune voltage, in this case 68kΩ suited.
Above is a tinyPFA screen capture of the frequency error (magenta) which bounces within ±20mHz as the temperature control in the OCXO cycles.
It might just be the outcome of a bad module design that doesn’t accommodate the tolerance of the OCXO component, perhaps it needed coarse and fine adjustment pots.
Having calibrated it, more work is needed to characterise frequency stability with changing ambient temperature.
Update 11/01/2023:
I have run a data collection for 16h from before sunrise to after sunset.
Above, we can see the frequency variation, about 3e-9 or 30mHz peak to peak and ambient temperature varied from 18 to 26 to 23°.
This thing aged quite a bit for a few days, but it seems more stable now.
The chart above shows the startup time. The data capture is done with τ=20ms, so for frequency error greater than 25Hz there will be wrapping… and on power up it starts more than 100Hz off frequency. Within 240s it remains within 0.1Hz (1e-8).
Kudos to Erik, this could not have been done (by me) without tinyPFA.
Update 22/01/2023:
Frequency error from cold start was logged using the FA-3-6G counter to make record of the time for reasonable stabilisation from cold. The OCXO had been calibrated to within 100mHz some days before this test.
At powerup, the oscillator is 360Hz low (error -3.4e-5), at 3min error is 3.4e-8, at 5min, error is 1.2e-8, at 15min error is 5.1e-9.
…. a work in progress.