Bench 10MHz Reference Project - Need OCXO Advice

[don.r]

I've decided to build a 10MHz bench reference and since Rubidium standards are now quite expensive (+US$130) and a little noisy, I decided to build one around a double oven OCXO.  Any recommendations for a reasonable (sub US$50) OCXO that has decent stability? I was looking at some Russian Morion units but have never used them. Are they any good?

[cdev]

You may be able to find a surplus "TruePosition" GPSDO that will give you a quite high quality 10 MHz reference for around that on ebay, a little less on taobao but you may end up with a unit that needs repairs. There are also symmetricom/trimble "UCCM" GPSDO boards for around $60 I think.

[don.r]

An interesting idea. I was thinking GPSDO's were out of my price range. How are the OCXO's they use on them for noise?

[cdev]

I have a Bliley on mine and I think its quite good. Mine has not been tested but that its good is what my understanding is.

There is a thread here about them.

[DaJMasta]

Usually not bad, the nice thing about the GPSDO over the normal ovenized oscillator is that you don't really have to deal with long term aging and drift, since it effectively works itself out.  Often a decent commercial surplus GPSDO will use a good OCXO, though not often a double oven in my experience, but even a moderate single oven will be good to around 10^-9 in short term stability, wheras any of them will be 10^-7 or so for a year of aging... so the GPS signal keeping it adjusted will keep your reference much more accurate compared to 'absolute' time as you use it - even the best ovenized oscillators don't match their normal short term spec with their aging characteristic, so the majority of the error is going to be that aging drift.

Especially with the easier to find GPSDOs, there is often good analysis from timenuts around the web with good measurement equipment, so take a look on ebay and see how the performance may be.  Usually these used telecom GPSDOs start around $50, and because they were originally designed for important timing applications, their oscillators tend to be fairly good.  The really top notch oscillators may not be that cheap, but a nice single oven will probably be in range.

[don.r]

Hmmm... this is morphing from a small project into a big one... not that that is necessarily a bad thing ... what is the power draw on the True Position?

[cdev]

right now my board is drawing around 250 ma @ 11 v. but at the beginning it draws at least 1.5 amps for around a second.

[don.r]

right now my board is drawing around 250 ma @ 11 v. but at the beginning it draws at least 1.5 amps for around a second.

Thanks. I gather that includes starting the heater for the OCXO. That is better than some double ovens by themselves!

[cdev]

It may vary quite a bit from unit to unit and also at different voltages. Thats just what its drawing for me right now.

(read off the analog meter on the front of my fairly new bench supply.. the 250-300 ma is after its been running for awhile. At the beginning it draws a hefty current, may be more than 2 amps but the analog meter just jumps up and then goes down again in around a second. That is most certainly the oven starting up.

I would put a fairly robust PS on it because I would expect that even with a voltage regulator that it likely needs a relatively stable voltage source to be stable. Its quite possible that some GPS conditions may require more current. Right now its in sort of a steady state - tracking several satellites. It may require more when it loses track of them or when its doing a survey, without doing that now I don't know. I would use a supply with a comfortable margin.

The TruePosition GPSDO definitely wants more than 5-6 volts.. The Packrats document claims it wants 13..

9 volts may be okay, right now I am feeding it 11.

[don.r]

I was going to use an Extron ADA 6 for the case since it would double as my distribution amplifier. Should be space in there for a bigger PSU along with the GPSDO and a small micro board.

[edavid]

To answer the original question, the time nuts used to like the Morion MV89, but they turned against it because it's a 5MHz oscillator with a doubler.  Now they like the Oscilloquartz 8663.

[4CX35000]

right now my board is drawing around 250 ma @ 11 v. but at the beginning it draws at least 1.5 amps for around a second.

Thanks. I gather that includes starting the heater for the OCXO. That is better than some double ovens by themselves!

Heater run up pulls around 1.5amp and the drops to around a steady 0.75amp on the GPSDO I have experimented with, the GPS takes a little longer to settle down once it has acquired the satellites. Some GPSDO require 5volt and others are 6volt supplies, so check with version you buy before buying a power supply.

Also remember there are these Chinese GPSDO with the uBlox NEO 6M GPS unit on eBay. See item number: 311943260991
Unfortunately I know very little about them and there performance, but the seller also does a 10 MHz distribution amplifier.

[technix]

I am wondering what prevented you from grabbing one of those second hand rubidium’s?

[don.r]

Two reasons. Cost (they are now well over $100 delivered) and, to some extent, noise. A good OCXO has a lower phase noise floor, particularly for lower frequencies. The FE-5680A appears to be particularly noisy, often 20-30db more than a quality OCXO.

[technix]

Two reasons. Cost (they are now well over $100 delivered) and, to some extent, noise. A good OCXO has a lower phase noise floor, particularly for lower frequencies. The FE-5680A appears to be particularly noisy, often 20-30db more than a quality OCXO.

I wonder if you can put the OCXO in some kind of slow-locking PLL's VCO loop. Before you decide to commit to a rubidium you can just use the VCO voltage input as a frequency calibration knob, breaking the PLL but not the oscillator. This way maybe you can use a FE-5680A to discipline the OCXO down the road, giving you the atomic class of long term stability. The noise from FE-5680A goes away in the PLL loop filter. And there are FE-5680 based modules that can be further disciplined by a GPS.

[DaJMasta]

It's my understanding that that's how good rubidium based sources generally work.  The rubidium is for long term stability, the OCXO is for low phase noise and short term stability, the GPS is for preventing aging (until the point where you can't adjust it out).

It's more expensive than the OP's requirements, but if you're looking at rubidium modules, look at the Stanford Research PRS10.  It's a nice rubidium source, but it has that PLL described above already built in, and it accepts incoming PPS disciplining signals - so it makes it very easy to drop into a design.  Not only that, but the equipment used in the loop is pretty good - 1ns timestamping for the PPS input, I think a 12 bit dac for adjusting the rubidium source and a 22 bit dac adjusting the OCXO.  Rather than some of the other modules which are just a rubidium reference, you only need a GPS input to make a complete GPSDO out of the PRS10.  Not sure if there are other similar modules available, but I was surprised to learn it was so integrated.

[don.r]

Yeah, even the FE-5680A's are too rich for this project but who knows where this will lead?  I guess this sort of thing is like a time-nut black hole, once you are in its tough not to get in deeper. I'm still trying to climb out of my volt-nut black hole. 

[don.r]

In case anyone is looking at this thread, I have managed to get a TruePosition GPSDO for not much more money than an OCXO (it comes with an OCXO as well). I am now deciding whether to put the GPSDO into the Extron case or have a separate case for the GPSDO. The Extron will need a new PSU anyhow as the original one isn't working and wouldn't provide enough current for the GPSDO anyhow.

[texaspyro]

The Truposition GPSDO's were pulled from cell site location measuring units like the LMU300.   These units fed them with 12V.   I know 22V will kill them...

The PRS-10 is a very nice rubidium, but the internal time interval counter circuit has a basic resolution of 400 ns (runs of a 2.5 MHz oscillator).  They get 1 ns res using an analog interpolator.  That interpolator is subject to things like temperature changes and component drift.

Attached is a plot of the TT time tag of it measuring the 1PPS output of an HP-5071A cesium beam oscillator.  Ignore the large scale variation as it is due to the PRS-10 discipline routine still locking up to the 1PPS input.  Look at the small wobbles that correlate with temperature swings due to air conditioning.

[texaspyro]

My favorite OCXO is the UCT-8663.   I bought a bunch of them for dirt cheap (10/$100) a few years ago.

They run off if 12V, have an EFC input, and a stable reference voltage output.   After a couple of months of continuous operation after being removed from a board by blow torch wielding neanderthals in China, tossed in a bucket, shipped to the US, and stored for 10 years mine have a drift in the low E-12 parts per day   I put one into my HP-53132A counter using a Gerry Sweeny OCXO board.  It drifts around 1 mHz/month (3E-12 parts/day)... better than some cheap rubidiums.

I've put them in a lot of Tek DC5010 counters.  I glued a quality 25 turn 20K trimmer to the OCXO with thermal epoxy.  The trimmer is fed with the ref voltage output with the wiper to the EFC input.  Those counters were still within 10 mHz after a couple of years.

I'm not sure how the Oscilloquartz 8663's compare.  I've been told that they are the same device.  Like all things precision oscillatey YMMV and performance depends upon your luck of the draw.

[don.r]

The Truposition GPSDO's were pulled from cell site location measuring units like the LMU300.   These units fed them with 12V.   I know 22V will kill them...

Yeah, no kidding. The shielding can on the Furuno receiver is totally covered in surface rust. Works a treat though. I feed it 9.5V and it gives me a healthy 10MHz signal but at 9.4V the signal is malformed. I suspect the floor voltage is down to the OCXO and each one may be different. The rubidiums are going for crazy money now, especially since the tubes can have a limited lifespan if left on 24/7 and are more power hungry than the GPSDO.

[medical-nerd]

Hiya

A rubidium at $130 is relatively cheap. A good double-oven HP10811 can cost about that now.
I paid about that for a good FRC-S 10Mhz rubidium about 12 years ago when the sub $100 ones dried out.
I only use it occasionally so the bulb is still good.
A good price for the Z3801a 10 years ago was $250 before they hit silly prices - $400 to $500
So when a new telecoms surplus GPDSO becomes available, grab a couple before the supply dries out.
I run a Z3801a continuously supplying my test equipment.

Cheers

[CJay]

The Truposition GPSDO's were pulled from cell site location measuring units like the LMU300.   These units fed them with 12V.   I know 22V will kill them...

Yeah, no kidding. The shielding can on the Furuno receiver is totally covered in surface rust. Works a treat though. I feed it 9.5V and it gives me a healthy 10MHz signal but at 9.4V the signal is malformed. I suspect the floor voltage is down to the OCXO and each one may be different. The rubidiums are going for crazy money now, especially since the tubes can have a limited lifespan if left on 24/7 and are more power hungry than the GPSDO.

I'm really curious why you're using only 9.5V if 0.1V difference makes the output waveform distorted, doesn't make sense to me not give it some headroom to work with?

[blackdog]

Hi don.r

Yust for your information, take a look at these 2 topics on this Dutch forum.

https://www.circuitsonline.net/forum/view/message/1730147

https://www.circuitsonline.net/forum/view/126898


Google translate is your friend 


Kind regarts,
Blackdog