If you have an old “Racal-Dana 199x” frequency counter or similar 10 MHz internally referenced gear with a poor tolerance “standard quartz crystal oscillator” or bit better “temperature compensated crystal oscillator” (TCXO) you could upgrade to a high stability timebase “oven controlled crystal oscillator” (OCXO) for under $25. [Gerry Sweeney] shares his design and fabrication instructions for a DIY OCXO circuit he made for his Racal-Dana frequency counter. We have seen [Gerry] perform a similar upgrade to his HP 53151A, however, this circuit is more generic and can be lashed up on a small section of solderable perf board.
Oven controlled oscillators keep the crystal at a stable temperature which in turn improves frequency stability. Depending on where you’re starting, adding an OCXO could improve your frequency tolerance by 1 to 3 orders of magnitude. Sure, this isn’t as good as a rubidium frequency standard build like we have seen in the past, but as [Gerry] states it is nice to have a transportable standalone frequency counter that doesn’t have to be plugged into his rubidium frequency standard.
[Gerry’s] instructions, schematics and datasheets can be used to upgrade any lab gear which depends on a simple 10 MHz reference (crystal or TXCO). He purchased the OCXO off eBay for about $20 — it might be very old, yet we are assured they get more stable with age. Many OCXO’s require 5 V, 12 V or 24 V so your gear needs to accommodate the correct voltage and current load. To calibrate the OCXO you need a temperature stable variable voltage reference that can be adjusted from 1 to 4 volts. The MAX6198A he had on hand fit the bill at 5 ppm/°C temperature coefficient. Also of importance was to keep the voltage reference and trim pot just above the oven for added temperature stability as well as removing any heat transfer through the mounting screw.
You can watch the video and get more details after the break.
[Dave Jones] gives an excellent overview of different crystal oscillators and their characteristics in EEVBlog episode about Rubidium Frequency Standards.
25 thoughts on “DIY High Stability Timebase Hack For ~$25. Why? Frequency Stability Matters!”
You might want to think twice for adding that 10 turns trimpot to the
oscillator. Temperature coefficient of a typical trimpot is +/- 100ppm
which is 20X of the voltage reference source.
What you should do is to use low tempco fixed resistors in series such
that the trimpot only looking at a small fraction of the range.
The trimmer here is used as a voltage divider. If the resistance of the pot is changing due to a temperature shift then both “legs” of the divider would shift equally percentage-wise. Thus keeping the voltage at the viper the same as before. Right?
This is of course only valid if the resistance of the adjustment input of the oxco is high enough to not influence the not perfectly stiff voltage.
But if the pot is heated unevenly all bets are off.
I would have agreed with what you said if the trimmer was composed of
only 2 fixed resistors on a small substrate and no mechanical moving
The divider ratio of a trimpot is dependent on the position of the
wiper. The lead screw that holds the wiper position, resistor substrate
and the molded case would have a different thermal expansion
coefficient. Wouldn’t that difference physically change the ratio of the
Since we are talking about ppm/C order of magnitude here, any of these
errors should be considered.
They don’t have the ratio tempco characterized in the datasheet, but
this support what I have said.
on page 184 (pdf page#):
>A slight variation in output voltage can be
caused by temperature-change effects on the ma-
terials used to fabricate the device. This may
result in minor mechanical movements. If the
wiper moves relative to the element, then Beta will
change. The same result occurs with a slight tem-
perature difference in two sections of the ele·
Although the resulting effects will usually be
negligible, they must be considered for critical
applications involving significant ambient tem-
Note: Beta = actual electrical travel
Yes, it seems like you’re right here. I didn’t consider mechanical movements due to temp changes. Interesting. Well, you learn something new everyday :-)
But considering that the trimmable range of most IsoTemp OXCOs are +/-1ppm when changing the trim voltage full range then the minuscule voltage change caused by mechanical expansion might be attenuated down to almost “nothingness” :-) I’ve looked for some appnotes related to this but I can’t really find anything relevant.
I think I’ll have to run some tests myself. Take a 10 turn trimmer connecting it to a x1000 gain opamp with some heavy lowpass filtering and measure the voltage with my 6.5 digit Agilent and see what happens when if I heat the tpot to 100C.
You are right about the insensivity of the trim voltage to frequency. OP
is essentially throwing away a 5mV/C tempco reference for an application
that don’t need it.
There are other ways of providing a trim with a tempco that is
parameterized on a datasheet e.g. a high precision DAC with dipswitch/uP
if you are fancy. (Digital pot have awerful tempco.)
I am interest with the results of your thermal test.
Yes I agree, it would have been better to use a couple of low tempco resistors in series and limit the range of influence to lower the effects of temperature effect. However, in my case, the problem was I did not have any pots of the right size, the Vfc pin is about 10KR so its source needs to be lower than that. As @Mats says the temp drift on the pot is just about even (save for the mechanical drift), the voltage (in my case) is about 2.4v when calibrated so that puts the pot quit near the middle of its range. I accept there is drift due to temperature but, the influence the Vfc pin has is +/-0.9ppm on a 0-4v voltage control range so drift is attenuated considerably – not that I have done the math to be fair…. In practice, I have powered the meter up three times from cold, and all three times after about 30 minutes of warm-up I was back to 10,000,000.00 on the display so I think its probably good enough for general day to day use, I would not use it for calibrating something else though – I would simply plug in the Rfs source :)
Thanks for making the point though, its a very good one…
Why do you force the temperature of the crystal? Why do you not just measure the temperature of the crystal and compensate for deviation since you have to measure temperature anyway to keep it constant so heating is adding another source of error? This would be lower power for things like clocks.
because wihout heating first you cant cool it down later
The requirement* is to keep the oscillator crystal at a stable temperature; it’s not strictly necessary to heat the crystal, you could cool it, or keep it at (a fixed)room temperature, but a heater is much simpler to implement than a refrigerator(even peltier) or heater/refrigerator combination. By keeping the temperature elevated, you can just reduce power to the oven when the crystal gets warmer than your calibration point.
*aside from ageing effects and annealing, charge carrier drift, etc.
They try to heat the crystal at a temperature that the crystal is least sensitive to small changes in temperature.
Not much point going for a surplus OCXO, you can get a 1ppm TCXO in smd for about $10. Have a look at Element 14 etc…
The Isotemp OCXO is rated in terms of ppb not ppm.
After the response Gerry got last time, I’m surprised he sent something into hack-a-day again. Good on him.
Thats @punkerdood, think skin I am afraid. It always seems to be that those that get on a high horse and complain about someone else’s hack are often not themselves content creators, so someone has to create the content in order to give others a purpose, even if that is to simply complain :) Thanks for your support
As GotNoTime says, OCXO’s are in a different class altogether. I am pretty (but not 100%) sure that what I took out of the counter is a TCXO
I want to see a DIY full frame time base corrector for NTSC analog video. Why? Because I have a ton of old tapes to digitize and the prices for even real old TBCs that need repair are insane.
low cost time base correctors are available for less then $300 usd
like the AV Toolbox AVT-8710 (can be found for $250)
the current pro level equipment starts in the $700 range
but the av toolbox unit will do everything you need
DTCXO’s provide basically the same reliability as OXCO’s for MUCH smaller cost and power requirement.
Pretty good explanation from Dave.
Ah, didn’t see the link tucked in there. oh well.
You really should link to the EEV Blog site and not a youtube link though!
Jeeze, after the respone Gerry got here last time, I’d think he say FU to hack-a-day. Good on him for not.
@Bobsyouuncle No, still happy to do projects and publish them even if that upsets some peoples sensibilities about theoretical perfection. thick skin and still willing to share with people that have an interest :) Thanks for your support though.
A good addition to this would be adding a phase change material, like a sealed capsule of wax who’s melting point is at the desired stable temperature. Think of it as a hot ice cube. It keeps you from overshooting the temperature or from the outside conditions changing the temp.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)