Cheap 10MHZ reference

[BiOzZ]

I know rubidium standards are on ebay for ~$120 (nobid)

but i have had a concept running around my head for years for a high accuracy lab 10mhz reference and i want some input


Well first you would start with a 100MHZ oven compensated crystal with an external adjust pin (found one on ebay for $25 nobid)

you take that in an cheap and small FPGA that would count the frequency (i think thats the best way without having to deal with expensive ARMs or deal with clocks)

you have a fairly low cost GPS module like the highly configurable Venus638FLPx with a 1PPS output (that is not simulated in software i am told) that is fed into the FPGA

the FPGA counts how many pulses are outputted by the crystal between each high pulse of the GPS module
if the value is over 100million it would lower the frequency and viceversa via a 14 bit DAC (after heatup times and all of that ofcource)

than the output is divided by 10 and put out (converted to a sine wave if needed)

after calibration that can be done in software i think this would give a very cheap and powerful 10mhz reference

i know using GPS to reference oscillators is nothing new but i have no idea how those work ... maybe the same way?

but anywho what are your guys/gals thought on this? ... its so simple i think somthing is either wrong with it or it has been done before
... if you guys think the theory is sound i will start work on some prototypes!
thanks!

[ejeffrey]

This is quite common.  It is called a GPSDO: GPS disciplined oscillator.  Commercial ones aren't cheap of course, but building one as you describe shouldn't be terribly hard.  You don't even need an FPGA since the update rate is so slow.  You could do it all with discrete logic, or just a hardware counter and a MCU (which makes it easier to fiddle with feedback gain to get a good lock).

[BiOzZ]

This is quite common.  It is called a GPSDO: GPS disciplined oscillator.  Commercial ones aren't cheap of course, but building one as you describe shouldn't be terribly hard.  You don't even need an FPGA since the update rate is so slow.  You could do it all with discrete logic, or just a hardware counter and a MCU (which makes it easier to fiddle with feedback gain to get a good lock).

oh wow ... i might just build one up than under open hardware using an FPGA at first than seeing if i can lower cost with logic (for deving an FPGA would save me allot of heartake)

i dont have a rubidium standard to compare it to tho ... might buy one at some point tho

[ve7xen]

This naive approach will be okay, but you'll generally get quite a bit better performance with a reasonable resolution phase comparator instead, as I understand it. If you've already got an FPGA in the design this shouldn't be hard to realize, but the PLL will probably be a lot more difficult than if you just used a microcontroller.

There are several designs and articles for such a beast floating around, but no (good) open source software I'm aware of. Plenty to read about the topic on the time-nuts list though.

[pauln]

See: http://ve2zaz.net/GPS_Std/GPS_Std.htm
Almost exactly what you describe.  Except using 10mhz ocxo.  I found a cheap Datum 10mhz ocxo on eBay - but 24V.

[BiOzZ]

Thanks for all your helps ... i have a GPS module, 10mhz OCO and a CPLD on order ... GPS module for the Venus638FLPx (chosen for a very short 1PPM trace) cost $35, 10MHZ OCO cost $22 new from china with great reviews, ATF750C CPLD for $8 and an LTC1655 16bit DAC  for $8 .... also a ANT-555 magnetic mount active GPS antenna with a nice long cord so i can shove it out a window and still have room to work with it .. tho i dont THINK you need allot of locks to keep a good 1PPS ... im not an RF guy ...

Notes: I chose a 10mhz instead of 100mhz because 1hz resolution is good enough for me right now at 10mhz
I may need to get a higher end CPLD or one of those fancy all in one AT94S05ALs

i will include a small micro ... maybe a small AVR ... to drive the CPLD and DAC and a few indicators with a few knobs for cal or even go as far as PC software cal ... one step at a time

[JJalling]

Hi, great project.
I'm sorry if I'm stating the obvious, but you want to look for at timing GPS receiver and not a navigation one.

BR Jonas

[calmtron]

I'm in the process of building a GPSDO 10 MHz reference myself at the moment, using a Morion MV89 double-oven ultra precision crystal oscillator (cool name, eh?) and a Trimble Resolution T timing GPS receiver.
Both are available on eBay pretty cheap, watch out for the datecode on the MV89s though, they have a very tight VCO pulling range (+-2.5 Hz) which means they might have drifted out of the 10 MHz adjust window after 10 years or so.
It is an interesting project, I've never tangled with this high-precision stuff before.

As JJalling stated, you'll really want a timing GPS receiver for projects like this. They will have a more accurate 1pps output (or maybe a 10kHz one oslt), and a bunch of features that might not be available on on a standard (position) GPS (sawtooth correction, stationary location mode, single satellite mode, ...).

Regarding precision, you are not limited to +-1Hz just because you use a 10 MHz reference, use some digital filtering and you should be able to pull of mHz resolution (milli, not mega).

[BiOzZ]

I'm in the process of building a GPSDO 10 MHz reference myself at the moment, using a Morion MV89 double-oven ultra precision crystal oscillator (cool name, eh?) and a Trimble Resolution T timing GPS receiver.
Both are available on eBay pretty cheap, watch out for the datecode on the MV89s though, they have a very tight VCO pulling range (+-2.5 Hz) which means they might have drifted out of the 10 MHz adjust window after 10 years or so.
It is an interesting project, I've never tangled with this high-precision stuff before.

As JJalling stated, you'll really want a timing GPS receiver for projects like this. They will have a more accurate 1pps output (or maybe a 10kHz one oslt), and a bunch of features that might not be available on on a standard (position) GPS (sawtooth correction, stationary location mode, single satellite mode, ...).

Regarding precision, you are not limited to +-1Hz just because you use a 10 MHz reference, use some digital filtering and you should be able to pull of mHz resolution (milli, not mega).

Thats exactly the one i got for this project!

but yes after research i will be buying a Motorola ONCORE M12+T timing gps to use for this ... under $40 its not a bad choice for its age ... i hope to also try and reverse engineer it to see if i can make one for cheep if i can use an FPGA

[JJalling]

but yes after research i will be buying a Motorola ONCORE M12+T timing gps to use for this ... under $40 its not a bad choice for its age ... i hope to also try and reverse engineer it to see if i can make one for cheep if i can use an FPGA

I've made a small circuit with a uC and a programmable delay line. It uses the saw tooth correction to lower the jitter on the PPS.
I can find the schematic if you are interested.

BR Jonas

[bingo600]

I've made a small circuit with a uC and a programmable delay line. It uses the saw tooth correction to lower the jitter on the PPS.
I can find the schematic if you are interested.

BR Jonas

Please upload the schematic Jonas , i'd like to have a look.
I might even get a M12T+ , if the comps are obtainable for a decent price.

Might be a fun backup to my thunderbolts

/Bingo

[JJalling]

Please upload the schematic Jonas , i'd like to have a look.
I might even get a M12T+ , if the comps are obtainable for a decent price.

Might be a fun backup to my thunderbolts

/Bingo

Here you go - hope you find it interesting. I couldn't find the firmware in the first go, but I will look again if people want it.
The GPS has a parameter, containing the negative sawtooth time error of the next PPS pulse. The uC reads the parameter, and uses the delayline to correct the error.
The uC can't communicate with the GPS, only read its output, so the Time RAIM status message has to be enabled manually (@@Hn).

BR Jonas

[bingo600]

Please upload the schematic Jonas , i'd like to have a look.
I might even get a M12T+ , if the comps are obtainable for a decent price.

Might be a fun backup to my thunderbolts

/Bingo

Here you go - hope you find it interesting. I couldn't find the firmware in the first go, but I will look again if people want it.
The GPS has a parameter, containing the negative sawtooth time error of the next PPS pulse. The uC reads the parameter, and uses the delayline to correct the error.
The uC can't communicate with the GPS, only read its output, so the Time RAIM status message has to be enabled manually (@@Hn).

BR Jonas

Thnx Jonas

And as i understand , the uC is just listening "in" on the M12T - TX pin , and captures the sawtooth correction messages. Then programs the delayline accordingly ... Correct ?


/Bingo

[JJalling]

Thnx Jonas

And as i understand , the uC is just listening "in" on the M12T - TX pin , and captures the sawtooth correction messages. Then programs the delayline accordingly ... Correct ?

/Bingo

Np!

And yes - that was what I was trying to say

BR Jonas

[mark03]

Here is a nice discussion of some of the issues encountered when locking a xx MHz oscillator using a GPS 1-PPS signal:

http://gpsdo.i2phd.com/

This seems to be why most amateurs have used old Motorola or Trimble timing GPSes with higher-frequency reference outputs.  The thing is, the only ones you can get at a reasonable price are truly ancient.  I mean no exaggeration...  like 1980s or 90s!!  They fall short of modern GPS modules in power consumption and voltage requirements (5V, really?? where am I going to get that?) .

I would love to get my hands on some of the contemporary u-blox timing GPS modules, e.g. NEO-6T or LEA-6T, but the niche market keeps prices super high ($199 in single quantities ).  AFAIK the timing modules cost twice as much only because of the magic firmware, and the price the market will bear.

[JJalling]

Here is a nice discussion of some of the issues encountered when locking a xx MHz oscillator using a GPS 1-PPS signal:

http://gpsdo.i2phd.com/

This seems to be why most amateurs have used old Motorola or Trimble timing GPSes with higher-frequency reference outputs.  The thing is, the only ones you can get at a reasonable price are truly ancient.  I mean no exaggeration...  like 1980s or 90s!!  They fall short of modern GPS modules in power consumption and voltage requirements (5V, really?? where am I going to get that?) .

As I read the discussion from your link, the problem is the long "settle time", because of the slow 1Hz pulse compared to 10MHz.
Could a solution be to start out with the 100Hz output of the M12+T and then switch over the the 1Hz output when the two were synced?
Of course the pll circuit needs to support both frequencies.

M12+t needs 3.0V, but yes. I would like to get a NEO-6T as well.

BR Jonas

[mark03]

As I read the discussion from your link, the problem is the long "settle time", because of the slow 1Hz pulse compared to 10MHz.
Could a solution be to start out with the 100Hz output of the M12+T and then switch over the the 1Hz output when the two were synced?

I'm not familiar enough with oscillator-locking and PLLs to know if this is a good idea or not.  Maybe try asking on the time-nuts mailing list?  I think a proper analysis treats the problem as a feedback control system, looking at the loop bandwidth etc. but I've never done that before.

[Sigmoid]

Welll... I was thinking about how I'm gonna calibrate my new analog scope, and that's when I found THESE HERE:
http://www.mouser.com/Passive-Components/Frequency-Control-Timing-Devices/Oscillators/Standard-Clock-Oscillators/_/N-7jdva?P=1z0ybztZ1z0yyf9

I really like the form factor! Nice assembly, you just need to tack an SMA-to-BNC adapter to the business end, and an SMA shrouded power cable to the other. I don't know if the 15PPM 25PPM max. error is particularly good or bad though.

[ve7xen]

I really like the form factor! Nice assembly, you just need to tack an SMA-to-BNC adapter to the business end, and an SMA shrouded power cable to the other. I don't know if the 15PPM 25PPM max. error is particularly good or bad though.

25ppm is pretty typical for a bog-standard quartz oscillator, like the DIP can type. For a few bucks you can get an order of magnitude better 2.5ppm TCXO (in SMD package only these days). For less than $20 you can get a <1ppm one. So it's not all that impressive, but it's an interesting way to package one up in a handy way. For a DIY-er you'd get a better spec with the TCXO in a Pomona 2391 BNC case for about the same total cost though, I think.

Such an FLL as discussed here should get into the single-digit ppb fairly easily.

[Sigmoid]

BTW, did anyone try the $80 Chinese Rubidium standards?

[ve7xen]

Gerry Sweeny has been working on building a standard out of one. I think Dave did a video on one a while back too. Their performance is pretty much the best you'll get as a standalone device (other than a caesium or H-maser), but long-term not as good as GPS - it's hard to compete with a constantly corrected constellation of caesium and rubidium standards.

The risk is that these are working pulls, and rubidium cells have a finite life. They've been taken out of service probably because of a replacement interval on them as a 'disposable' part. Should still have years left in them, but it's worth taking note of. Also apparently you can refresh the cells by revapourizing the rubidium or something.

I think this is the first of Gerry's videos:

[Robaroni]

I know this is an old thread but I've decided to revamp my GPSDO so I thought I'd get feedback and give some impressions with OCXO's that I've worked with.
A couple of years ago I bought two or three Milliren 260 OCXO's. These go for $1k US a pop so getting them on eBay for ~30 bucks seemed like a good deal. Milliren advertises these as up to Rubidium standards. The ones I got from China (fully tested....Yeah right!) were hit and miss. One couldn't be voltage adjusted and another one had a voltage reference drift, then after a day on the bench the heating element quit for awhile and  then came back on and off and on.... This makes sense, at the price they go for nobody wants to change them out more than they have to. The seller sent me a replacement for the first one and after testing for several days I now know I have two good ones. These are 5Mhz, by the way and I double the frequency with a Cypress chip.
The work very well, when they work, but look at the dates if you buy them. I noticed they're mostly past their MBTF dates (about 17 years if I remember Milliren telling me). The date is the bottom line, something like this

"990718-2"

99 is the year, 07 is the month and 18 is the day. The "-2" is the number in the line up that day out of the total number made.

Even the best ones still had some ref drift but not enough to make any real freq. difference.

The  Morion 89 uses more power, I think it came in around ~380ma at 12V as opposed to 250ma for the 260. I got hotter too (around 125F) but the voltage ref seemed to be more stable, after a couple of days the histogram showed a few uV's drif, much better than the 260. Both had superb frequency stability. I have a Rubidium standard that's calibrated but it's handy to see drift. I turn it on for a day, adjust the OCXO and then shut if off to save it. The uC 12 bit DAC controlled circuit I built (manual adjust using the 260's ref ) was within a few parts in 10 gigs over the few months since I checked. I didn't test the Morion but I would expect, after watching it for a few days on the bench, that it would be pretty close to the 260.

I  tested a few more OCXO's, a CMAC for one, but it was no where as good as the two above.

I've decided to build a uC controlled 1 sec GPS GPSDO. Seeing the long term stability of these OCXO's I figure I can go quite a while before freq,. corrections. I know people have done it but I use Atmel mostly with an external 12 bit DAC. I like this arrangement so I'd like to stick with it as much as possible, maybe tweaking a design. I do have a one second timing GPS, I had a Jupiter 10Khz but it died but with these OCXO's I think I don't need the, now scarce, 10Khz outputs.

Thanks for your help,
Rob