Build a 10MHz OCXO for the shack for less than $40.00


Active member
Thanks very much for showing us that nice design Ray, it looks as though all the stray fields would really be nailed down. Looking forward to seeing the finished item one day. Good on you Josh.


New member
My design pictured above never got made as the previous version(same but without output filter) has been nice and stable enough. The two that I built up used the CTI-OSC5AB02 which were sub $10 on ebay for a while and were motivation for mine (and Rays) hp5385 OCXO upgrade.

I went a different route (medium cost) since there was only going to be two; I used a REF5040 reference, and matched thermal resistor network with solder jumpers for the coarse adjustment. Even though the coarse needs to be set with a soldering iron(JP1 and JP2) that's ok once you know how much you need to pull the ocxo it unlikely to drift outside the range of the fine adjust and the benefit of having a very stable divider is worth the hassle if it does. The fine adjust is 200 Ohm pot, which is one of the biggest sources for drift. I was going to spring for the nice pot, but since most of its error was being divided out by the RN i though I would try the medium TC ppm pot(i think it was 75ppm/C and a few $ ea).

The ref and the resistor network were pricey.
Network: MORNTA1001AT5‎ $5.50 in 1 off qty.
Reference: REF5040IDGK $10.89 in 1 off qty.

The outcome has been drift of 328mHz in first 5 months(compared to a rubidium std) and is expected due to the first 2000hr ageing of the ref. The last time I measured in Feb 2020 it was 63mHz low which is still too good for what I need once I gave it a little insulation.

If doing it again I would use a buffer on the VC input guard and also buffer the OCXO output, and add extra 50Ohm port to go to the back panel.
Last edited:


Active member
A very interesting problem Rob, on some boards we do see masses of ground plain through stitching , I suppose so as to produce isolated fields etc. I have often noticed tracks on boards have rows of grounding via's on either side of a track, I suppose that would control the field within the track itself . Shame you can't modify your beautiful boards. Good luck.
No big deal Bill, made a new board, along with another that has a buffer and four outputs


Just out of interest, here is an Old School crystal oven from Radio Australia. It is quite a bit bigger!!!


It is a bit cooked.


Active member
R TV & H produced an temperature regulated crystal oscillator circuit, I must look that up. I think it used a power transistor as the heating element strapped to the top of the crystal can . It might have used a thermistor or diode as the temp sensor/regulator. Could be interesting to reflect.


Active member
I suppose if designing a temp regulated oscillator it would be best to use two power supplies both with plenty of reserve and well regulated, both working as a low impedance source, one supply to drive the heater transistor and regulator, and the other to supply the oscillator itself. If using a style D crystal there is quite an amount of space between the can and the quartz plate so it could be assumed that it could take some time to reach a stable operating temperature and frequency. If the power transistor was mounted on one side of the crystal can and the thermistor on the other side we could envisage the can heating up reasonably evenly up to the setting of the feedback control. If the whole thing were mounted in a foam box it could be quite stable but obviously a bit power hungry. If I had more time it would be interesting to experiment with Josh, I would have to take it to you for testing.


Active member
Interesting comments you make Bill.
The construction of the OSC5A2B02 follows your ideas in part, the crystal along with the heating element (a transistor) and the temperature sensor are bonded on a common copper plate.

What surprised me is the lack of thermal mass, I think this is done to achieve a fast warm up time, with this the problem associated with temperature cycling that is observed as the heater turns on and off, with this device frequency deviation been a very small amount in the range of+/-0.002 Hz (Worst case)

The unit uses a common VCC rail, with a 16 pin IC (P180704), from what I can see on the board, it incorporates the oscillator and the temperature control, I could not locate any details on this IC. As long as the heater circuit that draws the most current is well decoupled from the supply oscillator rail there should be no issues to achieve a stable frequency from the oscillator as attested by the device specifications stating a drift of less than+/- 2 Parts per Billion with the supply variation of+/- 5%. This indicates excellent decoupling from a design point of view.

The two main drift factors from the module discounting external components caused drift is that of RF load that from 50 Ohms to 1 MOhms the frequency can shift by as much as 0.01Hz

Picture showing the bonding of the heater, sensor and crystal to the copper plate 1.5 mm thick (Thin plate at that )



Active member
Four port prototype assembled and tested, only one mistake on the PCB, a bad mistake and don't know how I missed it., I had the VDD and ground swapped over on the IC.

The tests results show :-
Load drift from 1 M to 50 Ohm Load = Zero frequency shift
Harmonics Less than 45 db at 1 MHz load, not measurable at 50 Ohms load
Frequency drift due to oven cycling +/- 0.003 Hz
Output into 50 Ohms 600 mV pp
Output into 1 M Ohm 1 V pp

Last edited:


Active member
This is the final evolution of this project.

Note the use of the high stability components in the Control voltage area, if using standard parts the stability is still excellent requiring a slightly longer warm up time.

High stability Parts (Low Temperature drift coefficient)
Ra, Rb SUSUMU SSM ARE350 100 Ohms
VR1, VR2 Bourns TT Electronics

Note that the OCXO are disposal recycled units, its unknown how long they have been in storage, to be sure that the units are as stable as possible a period of 7 days of continues power up should bring them back to the required stability, else you will be chasing the calibration as the unit drifts.

Explanation on conditioning Link

OCXO 4CH.jpg


Staff member
Nice work Rob, good job tackling the temperature stability issues. I look forward to the next in-person club meeting for a demo.


New member
I've read that 30 days is not an unsual amount of retrace time for a ocxo that has been off for 6 months.