Purchasing newer vs older OXCOs for GPSDO use, pros and cons?

[cdev]

If somebody is picking an OCXO for use with a GPSDO and both a newer and older unit have the frequency control method you want available, a newer oscillator, all other things being equal, is likely to have more pronounced aging and a steeper slope of drift at the beginning of its life, correct? (thats been my experience with other quartz crystal based devices)

They all must eventually fail and a very old or a broken OCXO (that still has output) may become less adjustable not respond as it should to changes in its control parameters?

With the ideal being linearity-

I could see reasons why older vs younger devices might get better - or worse, actually.

Also, the resistive element that is used to heat it, may I gather, some suddenly fail, like the filament in a light bulb? (although a resistor seems inherently much more likely to live a very long time than a filament in a bulb or vacuum tube)

So there new might be more life remaining in a new device, than an old one. (But that may also be an intentional wearing out built in by a manufacturer- the extreme example of something like that is the world wide light bulb cartel which existed pre WWII- which demanded of manufacturers that light bulbs never last "too long". Interesting story there.)

Did I miss anything or is any of the above plain wrong?

What are the pros and cons of getting 'new' or 'used' and new vs old ?

[jpb]

I think the heating is normally done with an active device - this is certainly true for the HP10811 and also for the rubidium sources I've taken apart.
Something like the HP10811 is repairable - I repaired the one I got and then broke it again - as it isn't hermetically sealed.
I've bought a lot of OCXOs off ebay, mostly going for old-new stock. I think the advantage is that they are well within tuning range and haven't been mistreated.

I have got a new one I bought for my GPSDO project (still in its box though).

I would say that carefully chosen ones off ebay will give you much greater bang for buck. Buying new and spending say £100 buys you a bottom of the range OCXO - it will be SC cut if you're lucky. On the other hand you can buy new/old stock OCXOs for around £10 especially at odd frequencies or perhaps more realistically at around £20 to £30. You can afford to get 4 for the same price and have 3 spare.
Some OCXOs on ebay are way over priced I think though. The price of Wenzel oscillators are very high as are the HP10811s now. But take something like this:
https://www.ebay.co.uk/itm/EDC-71842-VCO-OCXO-12V-F-OP-10-00000MHZ-TC-0-2PPM-FREQ-ADJ-1-0-TO-7-0V/113007584164?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2060353.m1438.l2649
no datasheet available but it is easy enough to gently experiment to find the connections.
or this one:
https://www.ebay.co.uk/itm/CORNING-MC834X4-010W-VCXO-16-384000-MHZ-NEW/182071424519?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2060353.m1438.l2649
(The 4 sold I think were bought by me - it is not advertised as an OCXO but I think it is one.
Or these which I think are versions of the Oscilloquartz 8663
https://www.ebay.co.uk/itm/UCT-OSA-MY-108663-01-10MHz-12V-Double-Oven-OCXO-Crystal-Oscillator/332933537147?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2060353.m1438.l2649
If you're building something for yourself you don't need guaranteed supply. If you're planning to commercially build GPSDOs you obviously have to go for new.

A disadvantage of old ones is possibly that they are generally much larger, have higher voltage and current requirements so if you're going for a low power GPSDO you may want to get a modern surface mount device say something like :
https://www.digikey.co.uk/product-detail/en/connor-winfield/DOC052V-010.0M/CW696-1-ND/2619589
but its specs aren't great (10^-9 short term stability is about 100 times worse than more expensive SC cut ones) or
https://www.digikey.co.uk/product-detail/en/abracon-llc/AOCJY-10.000MHZ-E/535-10094-ND/2089867
which is SC cut and better specs and in the UK is £105 including VAT.

BUT when buying OCXOs beware of the ones that are not VCXOs especially buying new, the cheaper ones often turn out to be the non-voltage controlled variant (they have a F in the code instead of V often).

What I'd like to get would be a BVA one. I did get a quote for one once out of curiosity - it was about £10,000!!!
http://rubiola.org/pdf-articles/conference/2007-ifcs-xtal-osa.pdf
http://www.sungwhatech.com/product/pdf/ocxo/8607.pdf

Rakon also do a nice one which probably never show up on ebay:
http://www.rakon.com/products/families/ocxo-ocso

[cdev]

Thank you!

Considering how incredibly beaten up some of the OCXOs I see on ebay look, new-old-stock certainly does sound preferable to 'been through hell'.

 I wonder, what might be wrong with them? They probably 'work' as far as outputting a signal - but... Maybe they are being removed from service because they are almost failing calibration or something like that.

[jpb]

Thank you!

Considering how incredibly beaten up some of the OCXOs I see on ebay look, new-old-stock certainly does sound preferable.

Also, I see a lot of TCXOs on ebay and wonder how they stack up as far as stability to the OCXOs.

Do you know of a good beginners introduction to the different (OCXO) control loop methods?

Do you mean PLL (phase locked loops) when you mean control loop methods? There is a lot of stuff on these around including some quite good app notes. There is quite a good section in the Art of Electronics. The only difference between an OCXO and an VCO built-in to a pll chip say is that the frequency/V gain of an OCXO is much smaller.

Here is Bill Riley's description of building a OCXO PLL unit:
http://www.stable32.com/A%2010%20MHz%20OCVCXO%20and%20PLL%20Module.pdf

If you're thinking of GPSDOs based on 1PPS then you really have to go digital and Lars stuff on this forum is very good
https://www.eevblog.com/forum/projects/lars-diy-gpsdo-with-arduino-and-1ns-resolution-tic/
as are a lot of other circuits published over the years such as the very early Brooks Shera, though the link to that seems to be broken - here is James Millers description of building a version of it:
http://www.jrmiller.demon.co.uk/projects/freqstd/frqstd.htm

I rather like this simple one by the famous Charles Wenzel :
http://www.techlib.com/electronics/GPSstandard.htm

James Miller's GPSDO is a very simple control circuit with just a XOR phase comparator which shows you don't need to do anything hugely complicated:
http://www.jrmiller.demon.co.uk/projects/ministd/frqstd0.htm

I wonder, what might be wrong with them? They probably 'work' as far as outputting a signal - but... Maybe they are being removed from service because they are almost failing calibration or something like that.

A lot of OCXOs on ebay do look very beaten up and I'd steer clear of them. They are often more expensive than much nicer ones of the same type that turn up occasionally. Generally new/old stock are safer as they have probably sat on a shelf for years and then been superseded. This is especially true of odd frequencies - I bought a stack of 13MHz ones which were going cheap, a fairly useless frequency but I couldn't resist a bargain! (I think it is a frequency used in GSM networks.)

[cdev]

You can read Brooks Sheras' old web pages by using the Wayback Machine at https://web.archive.org/web

Bookmark the wayback machine for future use, its invaluable.



***I hope Lars is okay***

If I build a GPSDO ideally I'd like to use an inexpensive ovened oscillator that also is not uncommon on the used market.

I think I am going to use Lars' design since it seems quite straightforward and I already have most of the parts, aside from the actual OCXO.

I'll likely use a Navspark Mini since I have one earmarked for it. Its very inexpensive and it does remarkably well as far as keeping a fix. I don't know how stable its PPS signal is but everything Ive read and my experience using it for NTP by itself, leads me to suspect its much better than average as non-timing GPSs go. 

It also has a 'stationary' kalman filter setting which is something that many other GPSs dont have although Ublox M8 series do have that too.

 

[jpb]

Thanks for the link to Brooks Shera - though I have a couple of paper copies and a pdf but it is handy to be able to get back to the original.

Kalman filters are normally used in hold over but I think I agree with Charles Wenzel in his design, for a hobbyist/home lab use I don't think hold-over is really needed or even that desirable.

[SoundTech-LG]

Or these which I think are versions of the Oscilloquartz 8663
https://www.ebay.co.uk/itm/UCT-OSA-MY-108663-01-10MHz-12V-Double-Oven-OCXO-Crystal-Oscillator/332933537147?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2060353.m1438.l2649

Testing one of these OSA 108663's currently. It very well could be the 8663-XS (have one to compare). They are brand new, (NOS) and are going to take some time to settle down. So far the results are mixed, but to be expected. Certainly the price is right. I'd say about a month more of on time, to get a real idea of it's performance.

[jpb]

Or these which I think are versions of the Oscilloquartz 8663
https://www.ebay.co.uk/itm/UCT-OSA-MY-108663-01-10MHz-12V-Double-Oven-OCXO-Crystal-Oscillator/332933537147?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2060353.m1438.l2649

Testing one of these OSA 108663's currently. It very well could be the 8663-XS (have one to compare). They are brand new, (NOS) and are going to take some time to settle down. So far the results are mixed, but to be expected. Certainly the price is right. I'd say about a month more of on time, to get a real idea of it's performance.

I'd be interested in hearing how it works out. I have been looking at them for some time but I already have quite a few OCXOs and the postage to the UK together with 20% tax make them much more expensive for me than they are for someone buying in the States.

EDIT: as an update, I've now ordered myself 3 of them - a bit of late night extravagance on e-bay, I put in an offer and it was accepted. I tend to buy OCXOs in 3s so I can measure at least 2 against each other or do 3 cornered hat on 3 or have a spare as I don't have a Caesium reference or a Hydrogen maser one!

[SoundTech-LG]

I was becoming worried, as I saw quite a few spurs developing after about a week, or so...  these have settled down for the most part, and now, it appears it is close to, and heading to the nice values I have always seen on my used 8663XS. Better, and better as time goes on. This is my first new OCXO, I have always acquired used in the past. So, some getting used to.
If this continues, like you, I will offer for 3, or more. He's out til' the 26th.
Good luck with your tests.
Happy Harmonizing

[Leo Bodnar]

Does anybody have phase noise plots for these?

Ageing is a non-trivial, non-linear and, sometimes, discontinuous process.  I have overheard that whole batches and even part numbers were proactively removed from service due to accelerated ageing or frequency jumps.  I have a few used Wenzels that appear well looked after but refused to startup properly.  I have opened them up and after adjusting the feedback loop trim cap they started working again.

Leo

[cdev]

I read something similar.

Well, I suppose we should be happy that there still are lots of mysteries in the world that remain to be solved

[jpb]

I don't have any phase noise measurements on the samples I got - I don't yet have the kit to do so. All I did was check that they oscillated ok and that the frequency was adjustable roughly equal sides of 10MHz.

I think they are rebadged Oscilloquartz 8663s and the specs for those do have typical phase noise.

There are a lot of these around (two sellers on ebay) so I think they are good old new stock otherwise they must have abandoned large batches. Certainly just looking at the output on an oscilloscope and comparing the frequency to a GPSDO they seem very normal, nothing obviously wrong.

Leo, given that they are quite cheap at the moment perhaps you could get a few to play with and give us the results of measuring them with your phase noise kit?

The phase noise from the data sheet is :

Phase noise L (f) (BW = 1Hz) 
Frequency
10 MHz
Phase noise
1Hz         - 80 dBc
10 Hz    - 110 dBc
100 Hz  - 135 dBc
1k Hz   -  145 dBc
10k Hz  - 145 dBc
100k Hz - 145 dBc

[Leo Bodnar]

PN looks typical for an average OCXO, I will get a few to measure up.  It's difficult to get both good phase noise and long term stability since one demands high crystal drive and the other - low crystal drive.
Leo

Leo, given that they are quite cheap at the moment perhaps you could get a few to play with and give us the results of measuring them with your phase noise kit?

The phase noise from the data sheet is :
Frequency
10 MHz
Phase noise
1Hz         - 80 dBc
10 Hz    - 110 dBc
100 Hz  - 135 dBc
1k Hz   -  145 dBc
10k Hz  - 145 dBc
100k Hz - 145 dBc

[james_s]

The units on ebay look beat up because they are harvested off of ewaste that is shipped over to China in containers. I really do wish we had a better system of dealing with ewaste where stuff that is still usable or contains salvageable components is handled better and redistributed here in the US rather than just being dumped into bins but that's a whole different topic.

One aspect of crystal oscillators is aging, from my recollection the crystals get more stable with age and use, so in that sense an older unit may actually be better than a new one. At any rate I don't think I've ever had one fail, and if it did it's nit likely to kill anyone so I wouldn't worry about it.

[Leo Bodnar]

I bought a few of these UCTs and tested them about a week ago but wanted to make sure the results are correct before posting.
They take awhile to warm up - about 10 minutes during which they consume 0.6A at 12V, becoming 0.2A later.
They are, indeed, double oven jobbies.
Their phase noise levels are not the best around but not the worst either.
I have tested two identical units against each other, therefore applied -3dB correction. Spurs and 8Hz hump are probably deficiencies of my power delivery but these OCXO seem to be slightly more demanding to the purity of power supply rail then other OCXOs like Wenzels.
Leo

Leo, given that they are quite cheap at the moment perhaps you could get a few to play with and give us the results of measuring them with your phase noise kit?

The phase noise from the data sheet is :

Phase noise L (f) (BW = 1Hz) 
Frequency
10 MHz
Phase noise
1Hz         - 80 dBc
10 Hz    - 110 dBc
100 Hz  - 135 dBc
1k Hz   -  145 dBc
10k Hz  - 145 dBc
100k Hz - 145 dBc

[ArthurDent]

Other than servicing it to replace the battery pack a few times, I have a CFS-180A frequency standard made by the Manson Laboratory Division of Hallicrafters that has been running for over 35 years, which is pretty good reliability. The aging rate seems pretty consistent over the past few decades. I believe it uses a Bliley BG61AH-5, 5-MHz, 5th overtone, AT-cut resonator. The outer oven uses a mercury thermometer with a couple of electrodes so it acts as a switch and has been cycling 10 seconds on, 10 seconds off for all these years.

I don't think I own an oscillator less than ten years old.   

[texaspyro]

They are, indeed, double oven jobbies.
Their phase noise levels are not the best around but not the worst either.

That's pretty much what I found with them.   They are my favorite general purpose OCXO.  They are very temperature insensitive and once they age in (about a month) they tend to be wicked stable.  Here's my test of the one I put in an HP58132A.  This oscillator was a very old, ver used one... that had not been powered up for over 10 years.


UTC 8663 mounted on Sweeny DOCXO card in a HP 53132A counter.  The 8663 had
not been powered up for over 10 years.  Below is the frequency drift per day
as the 8663 aged in.  THe 8663 is spec'd at 1E-10 (1 mHz) per day aging. From
past experience, they are capable of much better than that (1E-11 to 1E-12 per
day).  They seem to be limited in this application by the EFC DAC / VREF
performance.

The 8663 was mounted on the Gerry Sweeny aftermarket OCXO board for the 531xx
counters.  For the first 30 days, the board was configured for the 0 .. 10V
EFC DAC range.  The Sweeny board powers the DAC with +/- 12V which is a bit
out of the spec'd minimum +/- 13.5V required for optimum performance on
the 0 .. 10V range.

The Sweeny OCXO board is based up on the HP design with a couple of nice mods.
First it has footprints for several common OCXO modules.  Second, it allows
the DAC range to be configured for 0 .. 5V,  0 .. 10V, and +/- 5V. 

The circuit feeds the OCXO output through a differential line driver.  The line
driver sends the OCXO to the HP counter mother board through a ribbon cable.
The motherboard makes the OCXO output available on a BNC on the counter
back panel.  Checking the ADEVs of the raw OCXO outout and the counter output
shows that the 53132A degrades the DOCXO output by around 1.5E-12 (raw DOCXO
value around 5.1E-12 at 200 seconds,  counter 10 MHz output BNC around
6.5E-12)

For the first 30 days the EFC DAC reading was recorded and the HP53132A was
then auto-cal'd and the new EFC DAC output was recorded.  (The DAC readings
for the first 10 days were lost when a momentary power failue occured and the
system was not shut down cleanly).

The initial 30 day run was done with the Sweeny board DAC configured
for the 0 .. 10V range.  Given the DOCXO EFC sensitivity (around 0.75 V / Hz)
and the 12-bit DAC, the minimum auto-cal resolution is 0.002 Hz.  Several
auto-cal cycles were performed each morning until one was within 0.001 Hz.

DAC readings were made by a Tektronix DMM914 4.5 digit DMM, not the best tool
for the task, but it was available. 


        drift (Hz/day)    DAC (after cal -> 24 hrs later)
day 1:    .298000 Hz       
day 2:    .057000 Hz       
day 3:    .033000 Hz       
day 4:    .022000 Hz       
day 5:    .017000 Hz       
day 6:    .011800 Hz       
day 7:    .009640 Hz      4.471V
day 8:    .008400 Hz      4.458V
day 9:    .006380 Hz      4.451V
day 10:   .003880 Hz      4.443V
day 11:   .002840 Hz      4.446V -> 4.438V  // power glitch caused reset
day 12:   .003470 Hz      4.437V -> 4.437V
day 13:   .002100 Hz      4.432V -> 4.433V
day 14:   .001590 Hz      4.430V -> 4.430V
day 15:   .001380 Hz      4.426V -> 4.428V
day 16:   .001660 Hz      4.424V -> 4.426V
day 17:   .000526 Hz      4.422V -> 4.423V
day 19:   .000284 Hz      4.421V -> 4.423V
day 20:   .000038 Hz      4.423V -> 4.424V
day 21:   .001019 Hz      4.420V -> 4.420V
day 22:   .001350 Hz      4.417V -> 4.418V
day 23:   .000402 Hz      4.417V -> 4.417V
day 24:   .000953 Hz      4.418V -> 4.419V
day 25:   .000304 Hz      4.417V -> 4.416V
day 26:  -.000441 Hz      4.417V -> 4.418V
day 27:   .000214 Hz      4.415V -> 4.415V
day 28:   .000517 Hz      4.415V -> 4.416V
day 29:  -.000998 Hz      4.415V -> 4.415V
day 30:   .001300 Hz      4.415V -> 4.414V
day 31:   .000423 Hz


Powered down, replaced 53132A fan, changed DAC range to 0 .. 5V, and
the 53132A was put back into case (hence no more DAC readings).  The 0 .. 5V
DAC range reduced the auto-cal minimum step size to 0.001 Hz.   Hopefully
it should also improve the DAC stability.

Warmup frequency error after power up:
   1 Hz     - 3m     
   0.1 Hz   - 3m 25s
   0.01 Hz  - 5m 45s
   0.001 Hz - 6m 15s

After 24 hours the unit was auto-cal'd once and left alone.  The frequency
error for the next week:

day 32:   .000152 Hz  (0.00009 Hz/day for last 18 hours)
day 33:  -.000825 Hz
day 34:  -.000730 Hz
day 35:  -.000287 Hz
day 36:  -.000144 Hz
day 37:  -.000086 Hz
day 38:  -.000083 Hz

[jpb]

Thank you Leo and Texaspyro for the measurements.

The phase measurements seem a lot better than the slightly conservative spec sheets.

They seem to be very good value at least from the supplier with the reasonable postage costs that I linked above. Wenzel OCXOs might be better but they are stupid prices on ebay and with no guarantee that they are still OK.
A second hand Wenzel is about ten to twenty times the price of one of these new/old stock ones.

If it wasn't for the fact that I probably don't need more than the 3 I've already got, I'd be tempted to stock up on them while they are still around.

[texaspyro]

If it wasn't for the fact that I probably don't need more than the 3 I've already got, I'd be tempted to stock up on them while they are still around.

Once upon a time a seller was offering used ones at 10 pcs for $100.   I stocked up and have used most (17?) of them.  Also bought three of the shiny new ones.  If you don't need world class phase noise they are hard to beat.

[tkamiya]

I recently bought 13 rubidium modules from eBay.  As to what might be wrong with them, they were actually remarkably good.  I had a few with unstable lock until heating up, and one that wouldn't lock at all.  Problem showed up only after a week or so of power cycling and running for few days simulating actual usage.

The one that didn't lock at all after a week had a sign of pretty strong impact. 

I also had many cases of identical unit configured differently by jumper placement.  That caused issues on testing but performance wasn't affected. 

Another unexpected difficulty was ability to obtain connectors.  In many cases, I can still buy it from manufacturers but they are incredibly expensive.  Luckily, I found someone who had a lot, found a right combination on Digikey, and a willing Chinese vendor.  If the unit does not come with a connector, I would research that first. 

As to pros and cons, PRS-10 currently sells for $1500.  Add few options and it will quickly reach $2000.  I spent just a little more than that, but my purchases include 3 units that are cased and portable.  Why buy so many?  I have no idea....  they just....  you know....  showed up on my doorstep....

[kj7e]

Has anyone experimented with an external layer of thermal insulation around the 8663 type ovens?  I would presume, the more stable the thermal environment around the metal can the better.  Provided there is still some thermal loss in order for proper internal regulation.

Anyway, I made a small closed cell foam box to fit round my UCT 108663's.  Monitoring the GPSDO EFC voltage, it stable to + or - 500uV over the last week.

[Leo Bodnar]

Here is quick phase noise plot for ENE3311E.
Not great, not terrible.
There is no Vref output on this OCXO so make your own voltage reference.
Leo

[GigaJoe]

i use heavy insulation on Chinese crystals, it significantly decrease power usage from 140ma to 40-30ma. (+5V )    an insulation 2in Styrofoam box. all  V-ref and adjustment voltage circuit also inside the box.   haven't seen any problems with that ... probably it increase stability; but hard to judge.