Building one's own GPS-disciplined oscillator to generate 10 MHz cheaply?

[cdev]

I'd really like to have a precise source of 10 MHz precision goodness, but the substantial cost of commercial offerings makes it less attractive than a DIY project.

Recently there was a discussion about a PLL cleaning up device for cheap GPSs that output a relatively jitter prone square wave..  This may take a while but my usual way of getting to some capability is to ask a lot of questions and read and build up a picture of what I need to do bit by bit.

Recently there was a discussion of cheap GPSs and cleaning up their output to make a lab quality precision 10 MHz frequency standard..

I have those references. Basically I just want to ask if there is any standout way to go from a GPS 1PPS with the typical issues to a low jitter 10 MHz signal in people's opinion.   

The commercial offerings start at around $150.

I was thinking that my old Cobalt Qube 2 would make a good platform for a GPSDO because its a very solid mini server, way out of date, but built like a tank, (and also the cosmetic aspect, its pretty) it also (like RPI) has several GPIOs unlike most newer PC hardware. (It also seems that some recent ports to it arent getting the network performance they should, don't know why, it may be fixable)

The top is gently rounded, its not perfectly cubical..

A ceramic passive GPS antenna fits perfectly under the center of the plastic shell thats on its top - it has a solid metal internal case and a very pretty blue plastic outer shell, so the metal makes a good groundplane for my Taoglas passive GPS antenna and the plastic will prevent the frying of the little GPS's RF input from ESD which is a danger if you use a passive antenna, the antenna input needs to be protected from static.. (according to Skytraq implementation notes)

There is more than enough room inside the Cobalt for an oscillator and oven.

The GPS that I want to use is a Navspark Mini, which is a unique little GPS (around the size of a nickel or dime) that only costs $7. Its very sensitive and it uses a Leon 3 CPU. It has a bunch of GPIOs plus the standard UART connections plus it may be able to be configured to have an additional Tx data UART. Also the Leon 3 CPU I think has an analog input pin which is unsupported but there.

One of the GPIOs is 1PPS by default, but I think it may be possible to configure it to provide other frequencies.


All in all the Skytraq Navspark Mini and other Navspark hardware are clearly outliers in the GPS field in that they have a higher level of programmability.  What that programmability allows - I am still trying to get a feel for what may be possible. (If its not with this GPS I may switch to another one for this project, important to also state here that I'm not a programmer, but have managed several times to get tasks Ive needed to done, through reading a lot, then if I dont use it I forget it.


I think that so far the number of interesting things that people have done with them may not be so large because the field of cool new boards is very large and people only have so much time, but its possible that some neat things could come out of programmable GPS hardware especially if the data could be used to cotrol external factors to improve accuracy, and also to deliver different frequencies so that a frequency standard's cost could be lowered.

See:
https://www.febo.com/pipermail/time-nuts/2016-August/099953.html

This is a thread on the timing on the much much more expensive Venus (not Navspark) timing-module hardware..

https://www.febo.com/pipermail/time-nuts/2016-August/099713.html

This is interesting.. although sort of off topic its about means of doing division of squarewave signals from GPS hardware with cheap add ons.

https://www.febo.com/pipermail/time-nuts/2016-October/101087.html

These also look interesting:
http://www.gaisler.com/index.php/products/processors/leon3?task=view&id=13
http://www.skytraq.com.tw/A001.pdf
http://t.cn/8FbovJT  (GRIP/Leon 3)

http://t.cn/8Fboh6t

[edpalmer42]

if your GPS receiver can output a 10 KHz signal that's locked to GPS, you could use the James Miller circuit.

http://www.jrmiller.demon.co.uk/projects/ministd/frqstd0.htm

It's performance is comparable to commercial GPSDOs.

http://www.leapsecond.com/pages/gpsdo/

Ed

[NivagSwerdna]

Have you seen...  eBay auction: #252610556561 ?

If you want to DIY then go for it... but depending on how you value your time it might be $100 well spent.  These GPSDO modules are very discounted due to being surplus now... you get a lot of engineering per $.

[wkb]

This might be of interest:

http://qrp-labs.com/progrock.html

cheers, Wilko

[MosherIV]

Hi

I was thinking of doing something similar. You have not said what you are using for the base oscillator. Most people would use a OCXO off ebay, make sure it has voltage controlled adjustment.

The quoted spec for 1pps jitter of the Skytraq Navspark Mini look impressive (10ns) but wait till you try it out.
I like the look of the Adafruit GPS (based on Global Top GPS module) or UBlox 6/7/8M
The performance of any of these is governed by the temperature behaviour of the crystal that drives the GPS, I looked at the UBlox devices, the T is the best but expensive and not available cheaply or easily.

There are a number of articales around now about doing digital PLL in software, that is basically what you will be tring to do phase lock 10MHz to 1PPS.

A simpler way round is to just average over a very lone time, look at this :
http://www.instructables.com/id/GPSDO-YT-Disciplined-Oscillator-10Mhz-Reference-Fr/

[FlyingHacker]

Pretty good video here:

[CJay]

Pretty good video here:

Not really, that GPS ha sa ton of jitter because of the way it generates frequencies.

The second user GPSDO

Have you seen...  eBay auction: #252610556561 ?

If you want to DIY then go for it... but depending on how you value your time it might be $100 well spent.  These GPSDO modules are very discounted due to being surplus now... you get a lot of engineering per $.

Those ex telecom GPSDO modules are a damn good buy. That type in particular appeared and started out selling at around $38 a few weeks ago.

There will be more, different models etc. so it's worth setting up a search that alerts you.

What they'll sell for is anyone's guess though.

[voltsandjolts]

What do you actually want to do with it?

For calibrating a frequency counter, this does the job just fine:

https://www.eevblog.com/forum/testgear/economical-option-for-precision-frequency-reference/msg754025/#msg754025
https://www.eevblog.com/forum/projects/ebay-u-blox-lea-6t-gps-module-teardown-and-initial-test/

[metrologist]

if your GPS receiver can output a 10 KHz signal that's locked to GPS, you could use the James Miller circuit.

http://www.jrmiller.demon.co.uk/projects/ministd/frqstd0.htm

It's performance is comparable to commercial GPSDOs.

http://www.leapsecond.com/pages/gpsdo/

Ed

Thanks for the links. I have one of the Jupiter modules and have been wanting to build this for years.

[mycroft]

Take a look at ttp://navsparkcommunity.ning.com/blog/more-ns-t-programmable-frequency-testing. It uses a PLL to clean the jitter but in doing so double the frequency, just add a divide by two flip-flop and you have a nice, clean and symmetric 10 MHz square wave.

[tautech]

Ha, I was looking into exactly this last night.
A couple of links found that might be handy for someone:
https://hackaday.io/project/11263-gps-locked-10mhz-lab-frequency-reference
http://www.hanssummers.com/gpsref.html

I was looking to cannibalize an old car GPS............
Following along with interest. 

[FlyingHacker]

Pretty good video here:

Not really, that GPS ha sa ton of jitter because of the way it generates frequencies.

I was under the impression that the modern versions of that unit were fine. Perhaps not the version he is using, though.

[cdev]

I would like to have a precise 10 MHz source for synchronizing test instruments and also for use as a means of tuning radios precisely. The higher you go in frequency the more off frequency you can potentially get unless you do something like that.

Most of my current devices are accurate to 1 ppm but that is not accurate enough, and my radios also drift with temperature and the crystals also age. I don't want to worry about crystals aging just because I'm leaving my receiver on much of the time.

[NivagSwerdna]

Those ex telecom GPSDO modules are a damn good buy. That type in particular appeared and started out selling at around $38 a few weeks ago.

$38.... I don't believe it.   These NEC ones have always been around $100 for the last year or so. There did seem to be a golden age in early 2015 when these things were super-cheap but they have climbed in price due presumeably to not turning up in mobile scrap anymore.

@cdev BTW.... avoid seller flyxy2015

...and read some of the other threads if you decide on a Symetricom or Trimble; I really like the look of the NEC... it is most modern GPSDO doing the rounds on ebay

FYI https://www.eevblog.com/forum/beginners/does-anyone-recognize-this-board-is-it-a-gpsdo/

[cdev]

That price sounds more like what Ive seen for the OCXO alone, not the GPSDO w/OCXO?

[CJay]

Those ex telecom GPSDO modules are a damn good buy. That type in particular appeared and started out selling at around $38 a few weeks ago.

$38.... I don't believe it.   These NEC ones have always been around $100 for the last year or so. There did seem to be a golden age in early 2015 when these things were super-cheap but they have climbed in price due presumeably to not turning up in mobile scrap anymore.

Mea culpa, not that specific one but there was a Furuno based board with OCXO for sale around May this year for $49.99 (not the $38 I remembered so, again, mea culpa)
http://www.ebay.com/itm/GPS-Receiver-board-from-TruePosition-GPS-locator-/152340314423?_trksid=p2047675.l2557&ssPageName=STRK%3AMEBIDX%3AIT&nma=true&si=059%252BT0IcF5f8%252BFx%252Fzruz8c%252BWN00%253D&orig_cvip=true&rt=nc

There are a few available from China still:

http://www.ebay.com/itm/TruePosition-10mhz-GPS-DISCiPLINED-CLOCK-GPSDO-AS-IS-untested-/302233354233?hash=item465e82fff9:g:0IAAAOSw4CFYrxG7

[NivagSwerdna]

...there was a Furuno based board with OCXO for sale around May this year for $49.99 (not the $38 I remembered so, again, mea culpa)
http://www.ebay.com/itm/GPS-Receiver-board-from-TruePosition-GPS-locator-/152340314423?_trksid=p2047675.l2557&ssPageName=STRK%3AMEBIDX%3AIT&nma=true&si=059%252BT0IcF5f8%252BFx%252Fzruz8c%252BWN00%253D&orig_cvip=true&rt=nc

That might work but is a bit crusty looking and is sold as not working.

The NEC comes with in-built LVDS goodness amongst other things.
(I have a Symetricom (which has Furuno GPS) but am still considering buying an NEC before they go un-obtainium).

[CJay]

Yeah, it's not the best example in the picture but for that money it's worth a punt if the Bliley OCXO is in working order.

The Furuno GPS has a decent reputation as a timing receiver so even just for parts it's not too bad a price if they can be persuaded to work.

I've broken with tradition (of being skint) and spent some money on one so I'm really hoping I'll be lucky and I'll report back when it arrives


*edit* Ah, my memory isn't that bad after all, these units were where the modules came from (ISTR they had bare boards too):

http://www.ebay.com/itm/TruePosition-LMU300-GPS-Navigation-model-101333-/162267042516?hash=item25c7de92d4:g:wEgAAOSwAvJW~UIi

Dirt cheap and there's information out there about them on a few websites, including a project to use one with an Arduino as controller:

http://www.packratvhf.com/GPS%20Project/A%20Packrat%20GPS%20Receiver%20Project-Web-upload-06-17-1.pdf

[cdev]

Hmm.. looks like Skytraq makes the s2525  "VCTCXO"  - this may be affordable and more modern (better satellite acquisition performance)

Had not seen this version of the S2525 before. I wonder how much this particular version costs. This could be a useful project for hobbyists if it could be used to make a stable, low jitter frequency standard in a small package.

http://www.skytraq.com.tw/datasheet/S2525DC8_v0.2.pdf

This is a TCXO and GPSDO with the associated jitter reduction circuitry, in module form format and likely also with low cost, not just another GPS.

So this might be comparable in cost to old telco equipment but higher quality output, and better satellite aquisition.

Thats the area where the skytraq GPSs stand out compared to my other GPSs. (Garmin and SIRFstar). They even work well with a passive antenna, indoors. (No LNA)  They will work with a (right handed) coiled piece of wire as the antenna.

Still dont own any ublox devices so I have not compared them to ublox.

[NivagSwerdna]

To throw a spanner in the works....

The GPSDOs need to be on continuously to meet their performance specifications, this can be a bit of a pain (especially since the OCXOs run hot and act a room heater!)... if you need a reliable standard but not continously then there are other options... e.g. Rb Standard   

[metrologist]

I bought one of those Rockwell Jupiter modules many years ago with the intent to build a GPSDO lab reference. The OCXO I got was spec'd to < 2E-10/day. I'd have to build the circuit and all, but now I see these completed modules seem so reasonably priced, I feel it would be a waste of time.

This module would be so much more versatile than what I had in mind:  http://www.sdr-kits.net/Webshop/products.php?106&cPath=5

"Short term signal quality is defined by internal TCXO clock source providing high-quality, low phase noise clock signal with sub-picosecond RMS jitter." 

Maybe I should sell of that Jupiter module and OCXO and put it towards an out of box solution?

[DaJMasta]

To throw a spanner in the works....

The GPSDOs need to be on continuously to meet their performance specifications, this can be a bit of a pain (especially since the OCXOs run hot and act a room heater!)... if you need a reliable standard but not continously then there are other options... e.g. Rb Standard   

Eh, a decent GPSDO with an OCXO will run maybe 10-15W tops (and less after it's warmed up), while a Rubidium standard doesn't have to be always-on to be super stable, it does if it's going to be GPS disciplined (if absolute frequency is important, rather than just high stability), and it still will take half an hour or more to warm up.  Plus, the rubidium will be consuming 10-20W after being warmed up and sometimes 50W+ while heating.  Sure, it's still a savings, but it's not going to be an instant-on reference, and you can get to a pretty stable point with a GPSDO in just a couple hours, even if it's highest performance is usually after a week or so.

I agree that the DIY route, while interesting, is not all that cheap compared to what's available.  It's also really nice to have stats specified when you're using it as a reference for your equipment.  While the basic PLL isn't so complicated, there's a huge variation in the performance of the loop given component choice and additional smoothing algorithms.  GPS receivers can vary greatly in their ability to generate a clean, stable reference frequency, and if you're not using a slow-reacting highly averaged smoothing algorithm for your PLL, those small changes in its stability will throw off the stability of your OCXO.  To get a really stable, disciplined output, you need to be averaging that output pulse for a long time to get rid of the jitter from the GPS frequency.  Even the very best GPS timing units will have 15ns RMS or so in their PPS output timing when fully locked and stabilized, so if you only use the output of the module to discipline your loop without any additional smoothing, the variance will likely be visible using a good OCXO.


My advice would be either get a surplus GPSDO module from a telecom or similar application or just pick up a GPS receiver that can make a 10MHz reference.  A lot of them with the higher frequency output options actually have a TCXO that's being disciplined by the GPS PPS signal, so while it's not quite as stable as a really nice OCXO, it's a good source in its own right in a very small and power efficient package.

[NivagSwerdna]

Eh, a decent GPSDO with an OCXO will run maybe 10-15W tops (and less after it's warmed up), while a Rubidium standard doesn't have to be always-on to be super stable, it does if it's going to be GPS disciplined (if absolute frequency is important, rather than just high stability), and it still will take half an hour or more to warm up.  Plus, the rubidium will be consuming 10-20W after being warmed up and sometimes 50W+ while heating.  Sure, it's still a savings, but it's not going to be an instant-on reference, and you can get to a pretty stable point with a GPSDO in just a couple hours, even if it's highest performance is usually after a week or so.

You are, of course, correct.  It really depends on what you are trying to achieve.  For my purposes as a hobby electronics person I don't tend to run a frequency standard on a regular basis, it is more a couple of times a year calibration exercise (or just a timenut indulgence).  The advantage of something that converges to a properly defined standard (e.g. the atomic clocks in the GPS constellation) is hard to argue against if you have the time and the electricity bill.   

[David Hess]

More complex designs rely on jitter correction from the GPS so if we ignore those, then we have typically 10s of nanoseconds of jitter in the pulse from the GPS because the clock generating the output pulse is asynchronous to GPS time in addition to the timing error from the GPS solution.  That leads to the simplest GPSDO solution being to divide down the output from the disciplined oscillator so that the maximum jitter from the GPS pulse does not allow cycle slipping when a phase locked loop is used to compare them.  At that point, performance depends on the time constant of the phase locked loop filter and even an analog filter can achieve 10s to 100s of seconds.

The Jupiter-T solution relies on a 1kHz reference and the James Miller circuit relies on a 10kHz reference from the GPS but there is no reason this cannot be done with the more common 1PPS reference.  Note that many or even most GPSes which provide a high frequency output are just phase locking it to the 1PPS solution so the higher reference frequency provides no advantage.

[tggzzz]

Still dont own any ublox devices so I have not compared them to ublox.

Even the cheap (£10) ublox neo6m modules can have their "1pps" output changed to a maximum of 500kHz. You do need a reasonable view of the sky with the supplied 4cm*4cm active antennas.

They have a 48MHz clock, so a 10MHz output would fall across clock boundaries (would need a neo6t).

A ublox app note implies 8kHz is a good frequency because it is an integer fraction of 48MHz, but doesn't say why 10kHz isn't just as good.