Budget GPSDO - complete

[trobbins]

MIS42N, I just received the pcb in the post - much appreciated.  A few more parts to come in and I will be able to start the build and testing 

My incentive was initially to just go for an OCXO, but then reading through the GPSDO options on this forum provided a relatively simple option to go one step beyond in certainty/calibration.  For me this is related to a vintage Systron Donner counter with 0.1/1Hz display resolution, and a venerable HP3325A with control capability to 0.001Hz - both of which have just basic crystal oscillator internal references.  So having a 10MHz GPSDO reference that can directly interface with the HP provides the ability to experience more 'zeroes' than I rightly have a need for (the >7 zeroes club), and for which I would never experience during the calibration of any other type of bench top equipment I'm ever likely to have.

Cheers, Tim

[MIS42N]

but filtering mostly depend on load of filter , would it be make sense to add opamp , between the filter and the control pin ?

Thought of that. That is why there are only 1kΩ resistors between the 10µF capacitors and there are 3 capacitors (most designs I've seen have 2). The source is only 12kΩ (10+1+1), the OSC5A2B02 input impedance is quoted as >100kΩ. An op amp there has to be a good one. Easier to design it out.

[GigaJoe]

curious, why opamp should be good one ?
it basic follower , and approx +2V out , simple OP07 , will do it,  ( i'm wonder , adding opamp together with ocxo  in foam isolated box , eliminated temp and humidity on opamp )
but if  100K then more troubles i guess ...

[MIS42N]

Is it time to rename this project Budget GPSDO - working?

The circuit was run for a day in what I consider a worst case scenario. The GPS receiver was a NEO-6 with an active antenna. The active antenna was placed on a window ledge (inside the window) of a downstairs side window in a two storey residence. The house next door is less than a meter away so the sky view is a narrow strip of sky between the two buildings. The power supply was a 5V charger supplied with a Samsung tablet.

VisualGPS was used to see what the GPS receiver was doing. It took quite a while to get a fix. After getting a fix, the number of usable signals was small, four or five satellited being common with one or two having a reading over 30 (signal to noise? - it's the figure in the NMEA data). In contrast, the antenna on the roof had 8 or 9 satellites over 30, some over 40.

In the 24 hour test period the circuit was able to maintain the desired 1ppb accuracy (10MHz+-0.01Hz) most of the time. There was one wild excursion for about 4 minutes of about 0.2Hz.  Apart from that, there were several minor deviations of less than 0.02Hz.

This demonstrates that the unit is still usable in less than ideal circumstances. If a user was using the unit for calibrating another instrument, they would watch the LED and wait until it indicated the unit was within specification.

[conrad550]

I have a question. Why not just reprogram the Neo-6 with uBlox u-center software to put out a 10MHZ signal?

[thinkfat]

Because it is heavily jittering

[AndrewBCN]

I have a question. Why not just reprogram the Neo-6 with uBlox u-center software to put out a 10MHZ signal?

The whole purpose of a GPS disciplined oscillator (aka GPSDO) is to combine the short term stability of an OCXO with the long term accuracy of the GPS signal.

In the short term, for example over a 1 second interval, most GPS receivers exhibit jitter of around 100ns in their 1PPS signal, and the very same jitter is there when you reprogram them to "put out a 10MHz signal". 100ns is 1 period of a 10MHz signal meaning the Neo-6 10MHz signal would have a precision of +/- 1Hz, whereas the OCXO 10MHz can be adjusted to a precision of +/- 0.01Hz or even 0.001Hz. That's 100 or even 1000 times better.

[iMo]

With Neo 6/7/8 modules the output 1PPS has jitter at 10MHz caused by the internal frequency division mechanism (in order to get 10MHz output you have to divide by 4.8 and 4.8 is not an integer).

The 1PPS output is "clean" when the output frequency is set to:

freq_out = 48MHz/N, where N is integer, N=4, 5, 6, 7, 8..., 98.000.000

Thus 12MHz, 9.6MHz,  8MHz, 6.8571MHz, 6MHz etc. output frequencies are ok.

This issue has nothing to do with GPS.

[MIS42N]

I have a question. Why not just reprogram the Neo-6 with uBlox u-center software to put out a 10MHZ signal?

Because that facility isn't available for Neo 6.

Even if it was available, imo AndrewBCN and thinkfat are all correct. The 10MHz output of a GPS module has limitations. The purpose of a GPSDO is to remove the limitation.

Each project has some starting point. The starting point of this project was to discipline an OCXO with as little extra circuitry as possible. It morphed to being at lowest cost, because less parts usually means less cost. So I am choosing the cheapest readily available components and try to get the best bang for buck. So I try working with the NEO 6.

It seems successful. I created a few kits to populate the PCB (as shown in Documentation02.pdf). I can mail that in Australia for $20AU including the postage. Already dispatched one.

[conrad550]

Thanks for the info. I don't need a 10MHZ ref. Any frequency will work for my calibration needs. Do the clone
GT-U7 and clone NEO 7/8 units have this capability?

[MIS42N]

Thanks for the info. I don't need a 10MHZ ref. Any frequency will work for my calibration needs. Do the clone
GT-U7 and clone NEO 7/8 units have this capability?

I don't know. I know it is talked about in a few other topics on this site, but it wasn't relevant to what I'm doing so I didn't bookmark them. IIRC some clone modules work, most don't retain changes, some don't work at all. You'll have to do your homework.

[MIS42N]

A new challenge. The software is basically finished, I have flashed a few units with a PICkit3 and no problems. However it seems PICkit3 is not that common and people have trouble with that final step of getting the software into the PIC. At first I thought, well that's OK I'll just mail out a loaded PIC, no problem. But what if the software changes?

Looking around, a few quite reputable companies (Bosch, Cisco, etc.) used Tera Term and Xmodem to update their software. Could this work for the PIC? Technically it seems so but I couldn't find an assembler implementation so I am writing one. So far, so good - the Xmodem logic has been implemented - the Tera Term Xmodem sends the PIC a file (which in the final version will be the hex file output by MPLAB X) and the PIC receives it. Tera Term thinks it has completed and so does the PIC. Now all that has to be done is decode the hex file (it is well documented), flash the program. Should keep me going for a week or two.

I sent a unit to Silicon Chip magazine to see what they make of it. Also getting some good feedback on a few PCB improvements. The current version is V0.0, maybe there will be a V0.1 if there's enough interest.

[de_light]

Quick update. I have received one of the trial kits from @MIS4N which came very quickly and well-packaged. After some struggles borrowing a Pickit programmer, I used an Arduino nano to flash the PIC which worked fine. I have installed it in a plastic Jiffy box available locally, but one of the suggestions I made was whether the board layout could be optimised a little and whether the board could fit in some readily available housing (Australia Jaycar jiffy boxes or an extruded aluminum housing or something).

Total current draw is about 450mA when the OCXO is heating and this drops to about 250mA after a few minutes. I created a serial port outlet with a JST connector and this works well with PuTTY. The LED works well and clearly displays the operation. MIS4N has written a very comprehensive manual that is highly useful. I have an active antenna that has arrived today so hopefully can properly fire the device up.

[de_light]

I'm pleased to say that this device works exactly as intended. I had a few hiccups along the way, mainly with supply of recycled NEO-6M devices from Aliexpress (poor soldering/shorted chips, shorted voltage regulators...you name it).

Apart from that, the device works very well. It is sitting, stable, on my desktop. It does lose calibration a few times a day (been running 48 hours) but quickly regains it with the inbuilt algorithm. It also seems sensitive to the activation of other test devices (esp clock-driven) on the bench, so could do with some sort of shielding.

Overall: excellent project. The fact that this is working for the price is mind-boggling. I feel this thread needs some traction.

[HB9EVI]

Thanks for the info. I don't need a 10MHZ ref. Any frequency will work for my calibration needs. Do the clone
GT-U7 and clone NEO 7/8 units have this capability?

neither the clones nor the NEO7 or NEO8 give a jitter-free signal @ 10MHz; and like the others said - a GPSDO is like the 'best of both worlds' - short term stability via OCXO and long term stability by disciplining it to GPS.

the signals from those GPS modules may be accurate in frequency, but also jitter-free frequencies like 12MHz or 2MHz are not clean enough to be used as a reference clock for other devices, while most OCXO have a rather high spectral purity. you just have to watch out to not 'mess up' things when disciplining a OCXO

[cdev]

One can buy a six pack of Navspark Mini, which is a very sensitive Skytraq dual GNSS GPS/Beidou board that has some unique connectivity options and is exceptionally sensitive, more sensitive than Ublox. It runs on  a Leon 32 bit microprocessor some of which is made available to the user via an API. Anyway, when bought in a six pack the cost per GPS is around $6. They are very small, around the size of a dime. (The US's smallest coin.) They are provided in an eight pin DIP board. The current draw is around 35 ma or so (please double check) Its very modest. I have several of these and I use them for NTP timekeeping on my home network. They make excellent cheap timekeeping GPSs in "stationary" mode set via the Skytraq GPS Viewer program. They seem able to maintain time well even in the basement of my house with a amplified or passive antenna. Generally they do much better than he Ublox performance with weak signals. They are not a timing GPS and Sytraq takes care to make it clear that they are not a timing GPS. 

I have two decent GPSDOS and would like to be able to compare them with their 1PPS output. Maybe a PIC would be the best way to do that. They all seem to be well synchronized on my oscilloscope. Unfortunately apart from a PIC eight digit counter I dont own a better frequency counter yet.   I can likely using PICs leverage my existing equipment to make an accurate time and frequency measuring device given as I have a known decent 10 MHz frequency standard GPSDO, a Samsung- also a  TruePosition (which I think is very likely to be as good or perhaps even better than the Samsung. But I dont know. . )

I don't know what the additional older version Beidou sats that the Mini receives add to the mix. Maybe they add some improved accuracy in the vertical plane (altitude) (which is not inaccurate for me, the altitudes they return seems more accurate than the altitudes sometimes returned by other GPSs for me. I have never seen more than one Beidou sat in the sky here in the US using my Mini. But the path as plotted in the GPS Viewer program seems very stable, with very few if any cycle slips once the Sktraq Mini has fully settled down. (which improves over a few hours time, maybe even 24 hours.. for maximum accuracy you should always turn a GPS on as long as you can before you plan to use one. Most importantly, given even  marginal antenna signal they seem to maintain a fix for months without any interruption or significant glitches in the 1PPS output. Also all Skytraq GPSs Ive used Venus 5, 6 and 8  )  are fairly configurable via GPS Viewer, on the Windows platform.. The Mini is Skytraq 6 family GPS it seems.. Which seems ideal for use with a GPSDO. Another person, maybe @jbeale or Dan Drown has written a fair bit on his blog about their PPS characteristics, and I think maybe has some additional code available on his github. A Gogle + github search on the Navspark Mini should find all of it.

Anyway, for $6 they are hard perhaps impossible to match elsewhere. The skytraq is quite a bit more sensitive and its location output seems smoother and more consistently steady than any other low to mid price GPS Ive used. When used with NTP it seems as if I am doing about as accurate as NTP can measure. Very low jitter as reported by ntpq.

[bingo600]

One can buy a six pack of Navspark Mini, which is a very sensitive Skytraq dual GNSS GPS/Beidou board that has some unique connectivity options and is exceptionally sensitive, more sensitive than Ublox.

URL please

/Bingo

[cdev]

http://navspark.mybigcommerce.com

You may have to dig a bit to find it. I suspect that whatever frequencies can be generated using the Arduino like (Navspark is a family of Skytraq Venus GPSs that add additional capabilities that involve the 32 bit Leon CPU are not jitter-free signals. Others here know more about them than I, so I defer to them as to what they can do.  .

Some capabilities of the Leon Sparc-like CPU allow use of it for functionality, for example, MCU-like features,  logging, geofencing, (doing math on the positioning results), RTKlib level accuracy down to a few mm (with other more costly GPS engines)  I sincerely doubt produced time and frequency relevant signals would be free of the problems described above (shown by Ublox) without the use of an external quartz oscillator.  They might be able to reduce the additional parts required to build a GPSDO in some creativge way, butI think the chance one could eliminate that need is likely zero.

They have discussed this! (something other companies would never do, IMHO)

 To state the obvious, They dont want to kill sales of their timing GPS, which is a good one, and a good value for someone looking for  a timing GPS to turn into a GPSDO.   Like most timing GPSs its a bit more expensive, and produces the correction signals needed for building a top notch timing solution. That's quite reasonable.

They are a maker-friendly Taiwanese company that deserves our community's enthusiastic support. 

[bingo600]

I can get Ublox LEA-5T (Timing) for approx 8..9$ on ebay
Older lady but still Timing.

Will have to look at this one

Thanx
Bingo

[cdev]

Where? I'd like to get one if its that cheap.

Although my Samsung GPSDO contains an LEA6T internally.

The fake M8N that I have although its a dual GNSS with both GPS and Glonass is really a fairly crappy GPS.  The RTKKit but older Skytraq GPS I have (S1315F / Venus 8 series based) is superior to the Ublox in positional accuracy by a large margin. Even though it is only a single system (GPS) unit. I dont think its just different Kalman filter settings like some have implied.

Its probably the lack of a TCXO on the (demonstrably counterfeit) Ublox.

This is what I got for trying to save money.

[cdev]

Basically the fake Ublox is useless for what I wanted it for. (RTKlib) I bet timing too. Maybe I will use it for a bedside clock.

I think a few microseconds fast or slow there will make little difference.

[cdev]

The cheaper units typically have just a single "time pulse" output which by default is set for (either) 2 Hz 50% duty cycle (bllinking light) OR 1 PPS. It could also be set to produce a fairly wide range of frequencies at other duty cycles (such as 50%) but its a dirty signal.

The cheap M8N which I have doesn't even break out the 1 PPS properly. It sends a signal to an LED which (from memory) may be 2 Hz or 1 pps. People who want to use it for other stuff have to solder a wire to that pin. Also as I remember the DC supplied to the GPS has to be fairly clean for the GPS to work well. Glitchy DC results in terrible GPS problems.

Thanks for the info. I don't need a 10MHZ ref. Any frequency will work for my calibration needs. Do the clone
GT-U7 and clone NEO 7/8 units have this capability?

I don't know. I know it is talked about in a few other topics on this site, but it wasn't relevant to what I'm doing so I didn't bookmark them. IIRC some clone modules work, most don't retain changes, some don't work at all. You'll have to do your homework.

[HB9EVI]

The fake M8N that I have although its a dual GNSS with both GPS and Glonass is really a fairly crappy GPS. 

cannot confirm that from my experience. I built 2 GPSDOs, one with an NEO-7M and a more recent one with a NEO-M8N. they both work alike, with my measurement possibilities I cannot see a drift between the two. I constantly receive both GPS and GLONASS satellites with the M8N, together always between 12 and 16.

and very likely both are just clones and not genuine Ublox

although I'm not using 1PPS but 1MHz to give on my PLLs as reference, while I divide the OCXOs by 10

[cdev]

my m8n receives a lot of sats too, but the output plot jumps around a lot with lots of cycle slips. It seems to me. I'll have to record the traces and compare it to the aforementioned S1315F or even the mini.

Maybe I'm wrong, this is just a seat of the pants guess.   Something seems broken on the Ublox.

It seems that Ublox doesn't ever completely stop jumping around even when the antenna is in the clear and is totally stationary.  WTF?

I blamed this on the lck of TCXO - but maybe its something else.


With the other GPSs I can lay out say a one meter square in my back yard and then move the GPS around the square and - it looks like a square. The tracks they generate can show you what side of a street you are on, or small turns, it records them. Same with trails and hiking. I record most of my GPS traces while hiking. Still seeking what is the best antenna for hiking.

Whan the GPS is totally still the display becomes fairly stationary, almost.

Yes, they work. Thank you.

[MIS42N]

No updates for a while because the design works. The current test rig has been running for over a month, the OCXO is not ageing much. The program update via Tera Term Xmodem is working in test. The output is usually within 1 part in 10E-10 of 10MHz as far as I can tell. I can't absolutely prove this because I haven't found a local standard better than this for comparison.

I live in Newcastle, Australia - anyone reading this willing to help with calibration?

Based on feedback from several people, there is room for improvement The following are under investigation:

1) Design for an enclosure - I haven't figured what is a standard - UB2, UB3 etc sold by several suppliers but these are plastic. Are these a standard? If the user wants metal because they are putting the device in an electrically noisy situation are there UBx enclosures in metal? More research needed. Any advice?

2) Sensitivity to external influences.
- I notice when the TTL to USB converter is plugged into the PC it causes a glitch - probably an earth loop. I've done some experiments with 4N25 opto couplers, and that fixes the problem for less than $2. The plan is to incorporate them onto the PCB and have the USB TTL converter in the case with a micro USB connector exposed to connect by cable to PC USB.
- power supply variations. I've tried various sources of 5V with "interesting" results. The new plan is to have a somewhat complex system using a buck converter to drop an external source (nominal 12V) to around 7V, then use separate 5V LDRs to feed the OCXO and the rest of the circuitry. Also a boost converter to kick the 7V up to (nominally) 15V to power a precision voltage supply from which the control voltage is derived. The 15V will also be used for 3). Awaiting deliveries of buck/boost from O/S at the moment.

3) Use a line driver/receiver between the GPS module and the GPSDO, connected via Ethernet cable. I have been using this combination to get PPS and NMEA data from a GPS receiver near the roof (antenna on the roof) for over 6 months. It works flawlessly. Heaps of satellites received, HDOP below 1 over 99% of the time. The GPS and line driver to be supplied by a buck converter from the nominal 15V (see 2.). The rig at the moment uses a 5V regulator but this requires the full current between GPS receiver and GPSDO. This results in an offset voltage between driver and receiver due to voltage drop in the earth return. This has not proven to be a problem up to 20 meters, but not all cables are made equal. A buck converter runs cooler and reduces the offset voltage problem.
A variant on this solution is to use one of the GPS modules with patch antenna inside a weatherproof enclosure, eliminating the need for an active antenna. Under construction at the moment.
Provision to be made for those people who are able to mount the receiver inside the GPSDO. It is easy to leave out the line receiver and put the GPS module in its place.