OCXO Design Questions

[Andrew_Debbie]

I am desiging a 10MHz and 1pps standard for my bench and other projects.  It will use a surplus OSC5AB02 OXCO.    I have a GPSDO, but want something I can use where GPS is not an option, like my lower ground floor (basement) work area.

I've got a couple of the cheap eBay modules, but we can do better, can't we?   I have attached a possible circuit for Vref, the adjust voltage.   On the two '5AB02s I checked.  Vref for 10.00000MHz is close to 2V.    A $6 ADR4540ARZ supplies a 4.096V ±4ppm/°C.  For an extra $2, the 'BRZ is ±2ppm but is that overkill?

4.096V goes into a voltage divider.  I'm using a $0.70 4 1K ±15ppm/°C resistor network  at the top and bottom of the voltage divider.   A 2K trimpot goes in the middle with the wiper arm providing Vref.

Reasonably priced  trimpots are 100ppm but cost less than £2.00.    A Vishay 1280G series foil trimmer is ±15ppm but is £24.   Is that overkill? 

Bourns recommend a minium  25µA wiper current and preferably 100µA to prevent oxide on the wiper arm.    That adds a ~68K resistor from Vref to ground.    Easy,  but it would add a small amount of thermal drift and noise.

Any suggestions?   Am I overthinking this? 

[guenthert]

I'm not familiar with the OSC5AB02 OXCO, but I recently spend (again) some quality time with the available documentation for the old HP 10544A and 10811A OXCO.  There electronic frequency control (EFC) allows for 1Hz (0.1ppm) adjustment using a control voltage of +/-5V (or so the marketing material states, the schematic suggest however 0...5V and that's what is used by HP's interface module).  There it is recommended that the EFC voltage has less of 100uV ripple in order to not compromise the performance of the OXCO.

You might want to find the specification for the OSC5AB02's EFC and the expected performance of the OXCO and then you'll be able to do the math.

[iMo]

The OCXOs have the "gain" in Hz/V, where Hz is the freq change per Volt at their EFC (for example 70Hz/Volt).
You may simply measure that around your 2V EFC "mid".
Then you may estimate the change caused by the TC of your divider.
For example, put the Vref and the divider into the LTspice, and do a MonteCarlo analysis with the resistor's and potentiometer TCs.. You will get an idea how it will perform.

[Andrew_Debbie]

I'm not familiar with the OSC5AB02 OXCO,

You might want to find the specification for the OSC5AB02's EFC and the expected performance of the OXCO and then you'll be able to do the math.

The 1 page datasheet is in Chinese.    I've got a rough translation here.    The datasheet claims the adjust pin is   2.0V,  +/-2.0V for -1.0ppm to +2.0ppm.     I have 5 of them, all pulls.    I haven't verified the range, but 10.00000000MHz is close to 2V.


There are several older threads about the OSC5A2B02 and the inexpensive modules on eBay and AE.      For example --> https://www.eevblog.com/forum/metrology/low-cost-lt1-ppb-10-mhz-oxcos-on-ebay.

I have two of the cheap modules.  One is awful.   The other isn't too bad, but I hope to improve on it.

[Andrew_Debbie]

For example, put the Vref and the divider into the LTspice, and do a MonteCarlo analysis with the resistor's and potentiometer TCs.. You will get an idea how it will perform.

Good suggestion.   

[dietert1]

You could measure frequency as a function of tuning voltage in the range from 0 to 4 V. Probably you need like five measurements (e.g. at 0 V, 1 V, 2 V, 3 V and 4 V) in order to see the nonlinear characteristic. Fitting a parabola (easy using a spreadsheet app) you find the slope in Hz/V at 2 V. From the datasheet specs mentioned above i would guess about 0.75 ppm/V = 7.5 Hz/V.
This can be used for error propagation estimates. With a 10 K ambient temperature change the 4 ppm/K reference will give you less than 80 uV error at 2 V. That will cause about 0.000 6 Hz frequency deviation. That is about 6E-11 of 10 MHz, well below the noise level of a crystal oscillator.

Regards, Dieter

[iMo]

..This can be used for error propagation estimates. With a 10 K ambient temperature change the 4 ppm/K reference will give you less than 80 uV error at 2 V..

.. and add the ADR4540ARZ epoxy drift/hyst. and TC of the divider.. But perhaps still ok for the purpose..

[trobbins]

You may want to browse the GPSDO thread by MIS42N (https://www.eevblog.com/forum/projects/budget-gpsdo-a-work-in-progress/), as it uses a OSC5A2B02 (I'm not sure what the difference is between that model and the indicated OSC5AB02).  The GPSDO thread indicates the voltage tuning used for the OCXO to stay within 1ppb.  You may find that OSC5AB02 drift and aging and ambient temp influence far outweigh the practical influence of some low tempco parts you are proposing.

[Dr. Frank]

I am designing a 10MHz and 1pps standard for my bench and other projects.  It will use a surplus OSC5AB02 OXCO.    I have a GPSDO, but want something I can use where GPS is not an option, like my lower ground floor (basement) work area.

...

Any suggestions?   Am I overthinking this?

I've got a basement lab (for temperature stable precision DCV measurements) in our 3-story house, as well.

I mounted the GPS antenna on the roof top, and ran its cable on the outside of our house down into the basement, drilling an appropriate hole into the window frame.

So, I have the Thunderbolt running downstairs, or an EFRATOM FRS Rubidium clock, or the diverse OCXOs (hp 10811A) in my plenty counters.
If required, I adjust latter clocks to the GPSDO. The FRS features an outside 20 turn trim pot for 2x10^-12 adjustment (resolution), which trimming voltage is derived from its internal reference supply, in a ratio configuration, so that the drift of the components does not influence the stability that much .

The 10811As can be fine tuned by an external voltages, as well, but I always use the internal cap trim only.
The stability is sufficient about several parts in 10^-9 per month, or so.
The influence from the external components can be calculated from its trimming sensitivity parameter, I guess.

Frank

[GigaJoe]

tuned such ocxo to -12 , it dropping to around -10 in 6 month, so approx -9 per year
but this is AC cut

Vref to OCXO:
i shunt 68uf polymer then few ceramic - minimize jitter
stability should be at least 0.05mV per year ,  so construct divider ,  as coarse ,and fine adjustments , fine need outside and minimal impact per temp change.
RC filter for Vref chip advised
would be advisable to keep Vref chip close or direct contact with OCXO,  as a plastic SOIC , it will minimize temperature and humidity effects.

OCXO stable  more less after around 2 Month,  Vref chip .... can be 6 months,  physically finalize a place where it should be , then do adjustment ;  movement can slightly affect output, or significantly by gravitational waves. depend on axis positions. 

my personal construction was  - whole board in 2 inch Styrofoam box  - minimize outer temperature affect to Vref chip, volt dividers, and OCXO itself.
TinyPFA are very useful tool for stats , adjustments.

SC cut are bit better

[MIS42N]

I have built multiple GPSDOs around the OSC5A2B02 OCXO. Their tuning are all around 10Hz/V @ 10MHz. My GPSDO design measures this automatically when commissioning a new OCXO and uses the value in the control loop (saves any need for manual tweaking). They can take quite a while to settle down and not jitter or drift. My advice is power the OCXO continuously and don't move or vibrate it. Some can take a week before they come good (in my mind, good is consistently less than 0.01Hz deviation from 10MHz when disciplined), and it may be weeks before the drift slows to the long term value. Some exhibit retrace problems, if turned off then on again they settle to a slightly different frequency (so need a slight change to the control voltage). The only one I've got running long term has settled around 1.6V for 10MHz. Most are closer to 2V.

[jonpaul]

In basement lab: Leo Bodan GPDSO>>>long SMA cable>>


At top floor/roof;

SMA cable>>GPS antenna.

Ask leo the way to extend the active antenna

j

[Solder_Junkie]

Like the reply from MIS42N, I have high stability OCXOs in test equipment that is only used occasionally. If I check the ovens in that equipment it will be surprisingly “unstable” unless it has been left powered for several hours. For calibration it should be powered for at least 24 hours. The ovens are fitted to Racal equipment and are option 04B.

The specification is quite detailed:

The ageing rate is 5 parts in 10^10 averaged over 10 days after 3 months continuous operation.
Temperature stability of +/- 6 parts in 10^10 per degree centigrade averaged over a range of 0 to +45C
Warm up to +/- 1 part in 10^7 within 20 minutes.

I would look to use a GPSDO rather than a free running OCXO. If you can install an outdoor antenna and a low loss coax feed to the basement, you won’t then have to bother about accuracy. Check coax loss at 1.5 GHz, it is quite significant but the outdoor “mushroom” GPS antennas tend to have a high enough output to cope with several metres of coax. Some roof mounted antennas seem to be over priced, the Leo Bodnar one with 30m of coax costs 3 times the cost for the same antenna with 10m of coax (they only use common/cheap RG58 coax).

SJ

[rhb]

The noise level of your PSU is critical.  The Vctl sensitivity and noise spec for those modules matches an LM399, 24 uVrms.   I have two Chinese new PCB units that are 0 1 ppb after about 15 minutes of on time.

I have 20 bare modules to test and select from. Notion is to measure the small residual drift's, model them and apply corrections to the tuning voltages using a set of 3-4 OXCOs>

These are better than the OXCO in my HP 5386A OXCO.  As received, after adjustment and enclosures.

Setting one to match a GPSDO is not easy.  And small air currents in the case can affect the frequency.  You use a scope to match the frequency.

Have Fun!
Reg

[Leo Bodnar]

Like the reply from MIS42N, I have high stability OCXOs in test equipment that is only used occasionally. If I check the ovens in that equipment it will be surprisingly “unstable” unless it has been left powered for several hours. For calibration it should be powered for at least 24 hours. The ovens are fitted to Racal equipment and are option 04B.

The specification is quite detailed:

The ageing rate is 5 parts in 10^10 averaged over 10 days after 3 months continuous operation.
Temperature stability of +/- 6 parts in 10^10 per degree centigrade averaged over a range of 0 to +45C
Warm up to +/- 1 part in 10^7 within 20 minutes.

I would look to use a GPSDO rather than a free running OCXO. If you can install an outdoor antenna and a low loss coax feed to the basement, you won’t then have to bother about accuracy. Check coax loss at 1.5 GHz, it is quite significant but the outdoor “mushroom” GPS antennas tend to have a high enough output to cope with several metres of coax. Some roof mounted antennas seem to be over priced, the Leo Bodnar one with 30m of coax costs 3 times the cost for the same antenna with 10m of coax (they only use common/cheap RG58 coax).

SJ

This is what our supplier charges us. I am happy to find a good source of cheaper antennas if you know one.

I have found that you can extend the GPS antenna cable with pretty much anything. Modern GPS chipsets have very low intrinsic noise so most of the noise is external and attenuating the signal with the cable does not reduce C/No as dramatically as if with high frontend noise receiver. Even 20m RG-58 extension will work fine if you start with decently placed antenna.

All I can say is give it a try. We take back anything you are not happy to keep.

Leo

[Andrew_Debbie]

An antenna on the roof here is not an option. I don't have permission.   I might be able to get enough signal if I stick it out the window.   I'll have to see if can gen enough satelites.

Otherwise, I'll have to carry everything home, adjust the OCXO and then carry it back on battery power.   I can't have an antenna on the roof at home either, but I can stick one out an open skylight for as long as the weather and my wife permit.   Besides Our house is 800 sq. ft.   No room for a bench there, at least not if I want to stay married.


I have a Racal Dana 1998 with Option 04B.  Over the summer, I checked it against my LBE-1420.  After a few hours it settled to reading 40-60mHZ high. I was able to let it run for a few days before it started to rain. I did not touch the adj. screws.    Just yesterday the 1998 started to randomly switch to the ext. std. even with nothing or a 50ohm terminator on the input.  Repair is on my todo list.

I will probalby use an oscilliscope to fine tune the OCXO. 

[Andrew_Debbie]

stability should be at least 0.05mV per year ,  so construct divider ,  as coarse ,and fine adjustments , fine need outside and minimal impact per temp change.
RC filter for Vref chip advised

Maybe use 5 or 10ppm fixed resistors for coarse? I've got an old Dana bench DMM that did that. Bit messy.     If I use a reasonably priced 100ppm trimpot, it doens't really matter. 

would be advisable to keep Vref chip close or direct contact with OCXO,  as a plastic SOIC , it will minimize temperature and humidity effects.

I plan to slot the board around the OCXO and encase just the OCXO in a hand made foam "box".   I can change that so that the VRef parts are inside the thermal island.

[MIS42N]

An antenna on the roof here is not an option. I don't have permission.   I might be able to get enough signal if I stick it out the window.   I'll have to see if can gen enough satelites.

Here's the current GPSDO antenna - the standard el cheapo puck with a 3m coax. It is on a window ledge, window closed, ground floor of a 2 storey house. The building next door is just over 1m away. The GPS receiver is a NEO-7N (also cheap).

This gets enough signal for the GPSDO to lock within 1 part in 10^-10 (i.e. 1mHz). Which is as good as any standalone setup costing 10 or more times as much. I have intention to try a mushroom with a 10m cable (already bought) but it's been sitting for a few months because the current system works.

So worth a try, then think of something more elaborate next if it doesn't work out. You can find good signal spots using most mobile phones or tablets and an app. I use one called GPS Test on a Samsung tablet.

[Andrew_Debbie]

For example, put the Vref and the divider into the LTspice, and do a MonteCarlo analysis with the resistor's and potentiometer TCs.. You will get an idea how it will perform.

R1,R2,R5,R6 are spec'd to be within ±15ppm of each other.    The trim pots are specified for the entire resistance.   For the trimmer I tried 15, 50 and 100ppm for each 'half'  which is likely worse than the actual device at very low current.

With a 4K temp range and the trim pot exactly centered I get  2.048 ± 240µV  with the 100ppm part.   50ppm gets ±189µV.     The 15ppm trimer is ~ ±100µV.

At 10Hz per volt, that is ±2.4mHz with the 100ppm trimpot over a 4 degree temperature range.  Unless I made an error, I don't need an expensive trimmer. 

[Andrew_Debbie]

an EFRATOM FRS Rubidium clock,

I worked part-time for Efratom California somewhere around 1980 while I was a student.   I set up an automated ageing rack for the crystal ovens in FRKs.  It used a Comador Pet, a Racal Dana RF switch and a Racal Data frequency counter.    Before that, someone used to manualy check each partially built FRK  every day and write down a number on a pad.

  I also did random rework and whatever else the downstairs technicians asked of me .

That was ages ago and I don't remember much else.