Yet another DIY GPSDO - yes, another one

[thinkfat]

The main reason to aim for long time constants is to make maximum use of the short term stability an OCXO offers against GPS. In other words, you want to disturb the OCXO only as required to meet your accuracy requirements (not stability!). Basically, you will want the OCXO to run freely without correction until it starts to drift away from the reference clock. A method to find this time constant is described in Lars documents about his GPSDO, but it similarly applies to all designs.

Herein lies a significant advantage of using a PLL over the counter method IMHO. A PLL let's you choose time constants as needed by the local oscillator only, while the counter demands time constants driven by the accuracy you want to achieve. Those two don't match most of the time.

[AndrewBCN]

Let me try to summary some GPSDO system design parameters we have discussed.

Counting the 10MHz pulses: 1 seconds gives 10e+7 count, 1000 seconds gives 10e+10 count. To reach 10e-10 accuracy you must at least count 1000 seconds to know if you are within the 10e-10 boundary. To correct to stay within 10e-10 you must count longer.

Measuring the phase of the PPS (once locked) from 2.5MHz can be done with at least 0.5ns accuracy, using the 10MHz for the phase may give 0.1ns accuracy. Without taking the fractional time of the GPS PPS into account, measuring one PPS with 0.1ns accuracy is sufficient to reach 10e-10

Including the 21ns fractional clock jitter of the PPS from the NEO-x modules gives 210 times less accurate time so you need to measure 210 times longer so to reach 10e-10 you need to measure the phase for 210 seconds. This 21ns clock jitter has exactly the same impact on counting the pulses but for 10e-10 accuracy using a 10MHz clock you already need to measure 1000 seconds so the 21ns PPS jitter has no impact when counting pulses.

This indicates that a FLL should count for at least 100 minutes to be able to adjust and remain within 10e-10 so the drift of the 10MHZ oscillator within 100 minutes should be well below 10e-10
Adding the phase measurement could reduce the measurement time needed to reach 10e-10 with a factor of 5. When using a GPS PPS without the 21ns PPS jitter this reduction could be much bigger.

The GPS position has an uncertainty in the order of 10ns leading to noise in the PPS of the same magnitude. Elimination of the position noise to reach 10e-10 requires at least 100 seconds measurement which is not relevant for the phase measurement method  (as its smaller then 210 seconds) and also irrelevant for the pulse counting method as the measurement time is already 1000 seconds to reach 10e-10

Is this a correct summary?

Erik, 
I would say that's a really excellent summary of the measurement accuracy/timing of PLL vs FLL GPSDOs, at least in theory and in ideal conditions. In theory, you are 100% correct. In practice, unfortunately, things are slightly more complicated.
1. As you have already noticed, all electronic circuits are sensitive to temperature variations. Thus, there are various parts of any GPSDO that are temperature sensitive: the power supply, the XO or TXCO or OCXO, the ADCs and DACs, the GPS receiver itself, the phase measurement circuit in the Lars design, etc.
2. Noise. Unfortunately noise is everywhere in electronic circuits, and when you are dealing with 1ppb or better measurements, it's impossible to ignore noise.
3. Random atmospheric propagation variations, and the fact that GPS satellites are always moving.
4. Aging effects of all the electronic components but most importantly, of the quartz itself.

A very interesting blog about building a version of Lars' GPSDO and the practical problems that can occur can be found here: http://www.paulvdiyblogs.net/2020/07/a-high-precision-10mhz-gps-disciplined.html

My approach to all these measurement uncertainty issues was to set for my GPSDO design an objective of 1ppb (10E-9) accuracy and stability while keeping the BOM and the components cost to a minimum, and being able to complete the project in less than four months. Thus my K.I.S.S. hardware and software approach. Also of course I wanted to "do it my way" instead of copying Lars' circuit, and most importantly, I believe any hobby project should be fun. And I admit I had great fun designing building and testing my STM32 GPSDO. 

You can always make a GPSDO more accurate, more reliable, and faster, but this always adds cost, complexity, and engineer.hours. And I guess at some point it's not fun anymore.

All that aside, I am very curious about your final GPSDO design which you have published on GitHub. I am going to get a good look at it in the coming weeks when I have more time. Good work! 

[AndrewBCN]

Another reason for averaging is that standard GPS modules receive one frequency band so don't fully compensate for ionospheric conditions. Advanced modules receive more than 1 band and are able to compensate by comparing them. Feed the NMEA data into a program like VisualGPS for a day and it shows deviations of several meters, each meter is about 3ns difference. Also for timing the best source is the satellite highest in the sky, but for position calculation it is better to have a scatter. So a standard GPS module is different from a timing module.

Manufacturers specify the stability of an OCXO after it has warmed up for a specified period, and that period is usually longer than 10 minutes. So expecting a GPSDO based on an OCXO to be OK after 10 minutes may be unrealistic. This might not be a problem with a TCXO because it doesn't have a warm up period. I have not used a TCXO.

Yes, I also think that TCXO's are too ambient temperature sensitive to be used in GPSDOs, at least when the target accuracy and stability are better than 10E-9. A used OCXO (< $10) provides the best "bang for the buck" for a DIY GPSDO, in my humble opinion.

[AndrewBCN]

The main reason to aim for long time constants is to make maximum use of the short term stability an OCXO offers against GPS. In other words, you want to disturb the OCXO only as required to meet your accuracy requirements (not stability!). Basically, you will want the OCXO to run freely without correction until it starts to drift away from the reference clock. A method to find this time constant is described in Lars documents about his GPSDO, but it similarly applies to all designs.

100% agree. That is why I wrote that for any control loop, the time constant can either be calculated by design, or found empirically. Lars' method to find a time constant is a good example of the empirical method.

Herein lies a significant advantage of using a PLL over the counter method IMHO. A PLL let's you choose time constants as needed by the local oscillator only, while the counter demands time constants driven by the accuracy you want to achieve. Those two don't match most of the time.

I disagree, because I wouldn't call that an advantage.   

The Lars' PLL control loop time constant must be adjusted individually for each build, and may even require periodic adjustments. It's an example of an empirical method. 

The FLL control loop I use measures over predetermined overlapping time intervals of 1s, 10s, 100s, 1,000s and 10,000s for predetermined measurement resolutions for a 10MHz oscillator of 10E-7 to 10E-11. This is an example of a deterministic method and it's much simpler to explain/understand for students and beginners, imho.

[erikka]

The Lars' PLL control loop time constant must be adjusted individually for each build, and may even require periodic adjustments. It's an example of an empirical method. 

It fairly simple to change this. My version already uses an adaptive strategy where the time between updates is automatically increase when proven stability over the current interval (and decreased when drift increases). There is currently an upper limit to the interval but but this is easily removed. After adding an absolute phase delta as minimum condition to act the oscillator will only be adjusted when needed.

[AndrewBCN]

The Lars' PLL control loop time constant must be adjusted individually for each build, and may even require periodic adjustments. It's an example of an empirical method. 

It fairly simple to change this. My version already uses an adaptive strategy where the time between updates is automatically increase when proven stability over the current interval (and decreased when drift increases). There is currently an upper limit to the interval but but this is easily removed. After adding an absolute phase delta as minimum condition to act the oscillator will only be adjusted when needed.

Erik,
That means your control loop algorithm uses a time variable (not a constant), but its empirical behavior doesn't differ in any essential way from Lars' control loop algorithm which has a fixed time constant.

For Lars' control loop algorithm: there is no update to the DAC value if the phase difference is below some arbitrary value. An update may or may not happen at the next time interval T-Lars.

For your control loop algorithm: there is no update to the DAC value if the phase difference is below some arbitrary value, and the update time interval T-Erik increases by delta. Again, an update may or may not happen at the next time interval (T-Erik + delta). At some point I assume the variable T-Erik stabilizes at some value which is just the maximum value for a given level of stability (and may not be the optimal value for your control loop, actually).

Lars' documentation provides an algorithm to find an empirical "optimal" value for the control loop time constant for T-Lars, your algorithm automates the task of monotonically increasing the initial time variable T-Erik up to some empirical stable time interval T-Erik(Max) where the measured phase difference is above some arbitrary value.

In both cases the "optimal" time interval is empirically determined, it changes from build to build, and it can (and probably will) change over time, as the quartz crystal and other components age.

Again, I fail to see any advantage here compared to the fixed overlapping time intervals for a given resolution of the FLL control loop in the STM32 GPSDO.

[FransW]

Trial & Error activities and using empirical methods are the signs of incomplete knowledge.
The true nature of the phenomenom is still hidden.

Whether the outcome is close to the true value or not remains uncertain.
We, all humans, have this deficiency.

Just a matter of evolutionary developments and they come slow.

Frans

[MIS42N]

Trial & Error activities and using empirical methods are the signs of incomplete knowledge.
The true nature of the phenomenom is still hidden.

Whether the outcome is close to the true value or not remains uncertain.
We, all humans, have this deficiency.

Just a matter of evolutionary developments and they come slow.

Frans

If there was a grain of useful information here, I missed it. The true nature of the phenomenon is well known and it has random elements. An adaptive response is appropriate to cater for this.

[bob91343]

My STM black pill module still hasn't arrived.  Worse, the antenna is delayed until September.  I fear this project will get buried and never be finished.

[FransW]

1. The Planck time is the length of time at which no smaller meaningful length can be validly measured due to the indeterminacy expressed in Werner Heisenberg's uncertainty principle. Theoretically, this is the shortest time measurement that is possible, roughly 10(−44) seconds, Wiki

2.  https://www.nist.gov/pml/time-and-frequency-division/time-services/utcnist-time-scale/performance-utcnist-and-utcnist
the NIST steering adjustments typically keep the root-mean-square deviation of UTC – UTC(NIST) to within < 3 ns, and the frequency instability of UTC(NIST) to below 2 ×10(−15)

As generally in physics, there are promising routes and there are waste of time routes.
The difference between 1. and 2. shows the lack of current knowledge and understanding.

Any time division improvement (a decade) takes about 10 years.
Trial & error or empirical efforts will not solve this. The contribution to a solution is still nill.
Individual satisfaction is the only gain. As for opinions.

Frans

[MIS42N]

1. The Planck time is the length of time at which no smaller meaningful length can be validly measured due to the indeterminacy expressed in Werner Heisenberg's uncertainty principle. Theoretically, this is the shortest time measurement that is possible, roughly 10(−44) seconds, Wiki

2.  https://www.nist.gov/pml/time-and-frequency-division/time-services/utcnist-time-scale/performance-utcnist-and-utcnist
the NIST steering adjustments typically keep the root-mean-square deviation of UTC – UTC(NIST) to within < 3 ns, and the frequency instability of UTC(NIST) to below 2 ×10(−15)

As generally in physics, there are promising routes and there are waste of time routes.
The difference between 1. and 2. shows the lack of current knowledge and understanding.

Any time division improvement (a decade) takes about 10 years.
Trial & error or empirical efforts will not solve this. The contribution to a solution is still nill.
Individual satisfaction is the only gain. As for opinions.

Frans

I believe you are indulging in a philosophical observation in a practical thread. We are interested in GPSDO construction and you contribute nothing toward it. Time is a human construct to make sense of events. Without events, time does not exist. Comparing events like the transitions between the spin states of the cesium nucleus in different gravitational fields leads to the conclusion time as we measure it is not absolute. It is just a convenient construct that allows humans to send payloads from Earth to Mars and have them arrive within meters of the target destination. Measuring "time" to a greater accuracy than is already done is not yet useful. GPSDOs are practical instruments that are used to calibrate other instruments or as a reference frequency for radio transmission and reception. It is irrelevant to us that we cannot do this to better than 10(−44) seconds. There are practical reasons for having frequencies that are accurate to 1 part in 10E-10 for low power transmission world wide, which in turn can be used for emergencies. And to calibrate them conventionally reference standards are 10 times more accurate. Individual satisfaction is a gain, but not the only gain.

[thinkfat]

Trial & Error activities and using empirical methods are the signs of incomplete knowledge.
The true nature of the phenomenom is still hidden.

Whether the outcome is close to the true value or not remains uncertain.
We, all humans, have this deficiency.

Just a matter of evolutionary developments and they come slow.

Frans

True, but that's what engineering is about: starting from a good guess rooted in experience and improving the solution empirically until it matches the requirements.

There is no "truth" in any science other than Mathematics. All we have is models that match what we can observe empirically and allow predictions. And lets not forget that Math is not realitiy, it describes reality, and it has its limits, too. Not everything can be "computed".

[FransW]

[/quote]
I believe you are indulging in a philosophical observation in a practical thread. We are interested in GPSDO construction and you contribute nothing toward it. Time is a human construct to make sense of events.
Individual satisfaction is a gain, but not the only gain.
[/quote]

I disagree strongly. There are plenty of subjects to spend positive time on.
Just follow the TEA thread.
All personal satisfaction and an enourmous sense of pleasure is derived from it.
The results is often working test equipment to the limits of the initial sales brochure.
Another thread to follow is the metrology section.
There you sometimes find ideas on "nutty" approaches: time-nuts, volt-nuts etc.
And the limits of relevant measuring equipment.

Frans

[AndrewBCN]

Hello,
I am still working on the STM32 GPSDO schematic in KiCAD, here is the latest version with the OLED display, the Bluetooth interface and the various sensors as optional modules. While these add modestly (< $10) to the cost and complexity of the circuit, I recommend adding them to the minimal version I published earlier.

Notes:
1. The final version will have (optional) output buffers for the 10MHz and 1PPS signals, and an (optional) onboard low-noise but still inexpensive linear 5V @ 1A power supply.
2. I am still debating whether or not to include an optional phase difference measurement circuit based on Erik's modified/simplified Lars' design, as well as a couple of op-amps for various optional features.
3. I got started on the PCB, using KiCad.
4. The final PCB design will include the footprints for all the optional modules.
5. All the KiCad files will be uploaded to the GitHub repository.

[thinkfat]

A few minor remarks based on a cursory look on your schematics:
* I see no pull-ups for the I2C1? Are they on the Black Pill module? If not, it's better to not rely on the weak pull-ups in the STM32, put some 4k7 resistors to VCC.
* MISO1 and MOSI1 could use some 22R resistors close to the driving pins, just take off the edge and prevent ringing. Only if not already on the Black Pill.
* Same for RX1, TX1, RX2, TX2.
* Add a bulk decoupling electrolytic capacitor 22µF close to the power plug, maybe add a series inductor of around 10µH before, for EMI purposes.

[AndrewBCN]

A few minor remarks based on a cursory look on your schematics:
* I see no pull-ups for the I2C1? Are they on the Black Pill module? If not, it's better to not rely on the weak pull-ups in the STM32, put some 4k7 resistors to VCC.
* MISO1 and MOSI1 could use some 22R resistors close to the driving pins, just take off the edge and prevent ringing. Only if not already on the Black Pill.
* Same for RX1, TX1, RX2, TX2.
* Add a bulk decoupling electrolytic capacitor 22µF close to the power plug, maybe add a series inductor of around 10µH before, for EMI purposes.

Hi thinkfat,
Many thanks for your remarks:

1. The pull-ups for the SDA/SCL are on the modules connected to the I2C bus. Ditto for the SPI bus (for the BMP280 module).
2. Indeed if the sensors/GPS/Bluetooth modules were mounted somewhere distant from the main circuit board, perhaps series resistors to avoid ringing would be recommended. In the case of mounting everything on a PCB, I don't think there is any significant ringing because the PCB traces are quite short (low inductance). I still have three prototypes assembled on breadboards, I am going to check the signals for ringing with my DSO.
3. There is an electrolytic decoupling capacitor close to the power plug on the 5V power supply, but I think your suggestion of a 10uH inductor to reduce EMI is a good idea, I am going to include the footprint for it on the final PCB design. Also note that the STM32 Black Pill has many decoupling capacitors, as do all the modules.

More generally, the fact that I am using prebuilt, breadboard-ready modules for the GPS, Bluetooth, MCU and sensor chips means that most/all of them already include LDO regulators and generous decoupling capacitors, as well as the pull-ups and series resistors where needed. In principle that helps coping with noise and signal transmission issues.

[bob91343]

Andrew, I still don't have the STM black pill.  I have sent a complaint to Aliexpress but so far no response.  If I can get a refund, the ordering process will start over.  Perhaps something will happen soon.

In case it does arrive, I am still at sea regarding my next step.  The antenna and GPS module and TCXO and so on are all sitting here in a little cardboard box.

I did run the rubidium standard to see how far my counter master oscillator had drifted, and it was amazingly close, so close in fact that I saw no need to adjust it.

On an unrelated (sort of) topic, I have been frustrated by the inaccuracy of the cheap digital watches.  I did find one that was almost perfect but it has disappeared.  I suspect one of my cats.  However, I bought a package of 20 32768 crystals to replace some of the poor timekeeper crystals.  On trying to measure the frequencies of these 20, I encountered a problem in that my source of signal for the test isn't stable enough to get a good reading.  I suppose my best move would be to make an oscillator and swap in the different crystals.  That would give me a stable reading which I could read out to 0.1 Hz resolution.  Right now, the best I can do is about 1 Hz.  Any ideas of a simple oscillator?  Maybe a dual AND gate or something similar.

[AndrewBCN]

Hi Bob,
I would suggest you order the WeAct STM32F411CEU6 Black Pill from their recommended seller on AliExpress:
https://www.aliexpress.com/store/910567080

Normal delivery in 10 days.

Testing 32.768 kHz watch crystals for accuracy would not be too difficult if you had a working STM32 GPSDO, since it works as a frequency/period meter with 0.001Hz/100ns precision. We can discuss that possibility once you have the GPSDO assembled and working.

The simplest 32.768 kHz XO oscillator possible requires a single unbuffered inverter gate from a 4069UB chip, an example here: https://reviseomatic.org/help/x-more/Crystal%20Oscillator.php

[bob91343]

The black pill I ordered was half the price posted by WeAct.  So I guess I will have to see if they refund or replace.  If refund, I'll reorder.

I will try to build an oscillator and find the closest crystal to nominal frequency.  Of course I don't know the oscillator parameters in the watches so each oscillator will operate on a slightly different frequency.  Trying it is the best way but I want to reduce the number of crystals to try.  Removing the current watch crystal and measuring it will get me a number that I can use to select the best crystal.

Most of the crystals are within 1 Hz of nominal, series resonance.

[AndrewBCN]

This is the second sheet of the GPSDO schematic, with an optional 5V @ 1A regulated low ripple, low noise power supply, and optional output buffers for the 10MHz and 1PPS signal. The footprints for these will be on the future PCB, but they can be left unpopulated if not needed, just like the sensors, OLED display, etc.

[bob91343]

Andrew, I eagerly await the chance to acquire a PCB to assemble this project, which seems to keep growing in cost and complexity with time.  All I want is 10 MHz, no other functions required.  Perhaps it will be easy for me to see how to get that.

[AndrewBCN]

Hi Bob,
Designing the PCB will take some time, but I am working on it, and of course I will post here when I am done. The STM32 GPSDO is very much a modular design, as I wrote before the guiding principle for this project is K.I.S.S. (keep it simple, stupid).
https://en.wikipedia.org/wiki/KISS_principle
Apart from the MCU board, the OCXO and the GPS module, everything else is pretty much optional, as clearly indicated in the schematic. So for a precise 10MHz +/- 1ppb frequency reference, basically you only need three small modules, an antenna and a few passive components for a total cost of less than $30. And even with all the optional modules the total cost should remain under $35.

And no, the project has not grown in complexity or cost, if anything it is simpler and cheaper than a couple of months ago, since now the MCP4725 12-bit DAC module can be replaced with a couple of resistors and capacitors to generate a 16-bit PWM voltage.

Please PM me when you finally receive the STM32F411CEU6 Black Pill and we can get started on detailed assembly instructions.

[bob91343]

Andrew, good news from you.  Yes I will send you a PM if and when the STM pill shows up.  I have opened a dispute with Aliexpress.  All the other stuff was ordered from them and arrived okay so it's the one part.  I don't necessarily blame the vendor; it has to pass through many hands.  In fact there is some evidence that it arrived in this country a few weeks ago so it's lost locally.  Nevertheless, I suspect Ali will refund my money, at which point I will reorder from a different vendor at double the price.  As they say, in for a dime, in for a dollar.  Or maybe the same vendor.

In the meantime I am trying to get a little radio working.  I designed it in about 1951 for my girl friend's high school science project and it worked (to the consternation of the teacher!).  I was about 19 at the time and essentially made her a kit.  It uses one tube, a 1N5GT and just a few other components.  I think it works but I have to hook up headphones to find out.  Somehow it got preserved for all these years and is mostly intact.

Two technological extremes.

[Johnny B Good]

Andrew, I eagerly await the chance to acquire a PCB to assemble this project, which seems to keep growing in cost and complexity with time.  All I want is 10 MHz, no other functions required.  Perhaps it will be easy for me to see how to get that.

 It seems to me that you're attempting to build your very first DIY GPSDO and trying to overcome your fear of starting a project with a seemingly high level of complexity that you're worried you might not be able to master without making a rather greater investment in both time and money than you can afford.

 In view of your stated requirement for a basic 10MHz reference, may I suggest that you start with the most basic of GPSDO projects as outlined by Gyro in this topic thread here:-

 https://www.eevblog.com/forum/projects/my-u-blox-lea-6t-based-gpsdo-(very-scruffy-initial-breadboard-stage)/msg929133/#msg929133

 As he mentions, it was inspired by James Miller's simple GPSDO (a small factoid I'd forgotten when I first described my own initial GPSDO projects as being inspired by this version, namely the circuit diagram he'd attached in the final thread posting).

 I'd substituted the 74HC4046 with a 74HC86 quad XOR IC on account I didn't have a 74HC4046 handy but I did have a few 7486s to hand to exactly emulate the PC1 portion of the HC4046 being utilised. Also I'd left out the dual RRO 5v cmos opamp and fed the output of the LPF direct to the the VFC pin on my 13MHz square wave output OCXO which I was feeding into a TTL based divide by 1.3 circuit to generate a 10MHz square wave locked to the OCXO.

 Whilst a 13MHz square wave OCXO can be used to make a 10MHz GPSDO reference source, it really has little to recommend it other than, in my case, it being a cheap way to get an OCXO that can directly drive logic gates without the need to include a voltage level shifter (as I discovered, much to my chagrin, when I first started using a 10MHz sinewave version of these AE CQE OCXOs -ex Symmetricon kit as I later found out).

 I was using a ublox M8N (my very first GPS module purchase and, rarity of rarities, a genuine one to boot!) in place of the LEA6T used by Gyro. I built the whole thing on a solderless breadboard, experimenting with various enhancements before transferring it onto vero stripboard and boxing it all up into a rather petite 110x110x50mm extruded aluminium enclosure which completed my MK I GPSDO which I powered from a cheap 12v plugtop psu (slightly modified to kill off the "Y" cap induced 90vac mains leakage 'touch voltage' that typically afflicts such smps based wallwarts).

 I'd made the choice of external psu power deliberately to facilitate rapid swap out of a failed psu to minimise the down time involved in disassembling it to troubleshoot an integrated mains psu, plus, this would also eliminate a significant source of internal heating.

 As a further measure to minimise internal waste heat, I'd used a low noise 7 to 24v input 5v output buck converter (designed to power the avionics in drones powered from 3 to 6s lipo battery packs) rather than the classic 7805 which would otherwise have required the additional complication of heatsinking to the aluminium case. This gave me another degree of freedom regarding my wallwart output voltage options since it can be powered from supply voltages ranging from 7 right up to a maximum of 24 volts and still only draw (OCXO at set temperature) 1.7 to 1.8 watts in the case of the MK I (the MK II only needs 1.3 to 1.4 watts over this range after the OCXO has come up to temperature).

 The main downside with using a navigation class GPS receiver module in such a basic James Miller based design is the 30 to 50ns phase wander due to ionospheric effects (space weather). Long term, with impractically long PLL time constants, it can match the much more expensive single and multiband timing modules. This where a cheap microcontroller and properly optimised firmware can win out but there are a whole bunch of temperature related issues that have to be dealt with which the basic James Miller design can neatly 'sweep under the carpet' as demonstrated in this interesting comparison here:-

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

 The significant improvements over the MK I GPSDO I'd made in the MK II were the use of a 10MHz (sine wave output) OCXO (same AE CQE brand) allowing elimination of all the noisy and power hungry TTL 'Magic' required to create a 10MHz clock from a 13MHz OCXO and a GPS Rx module upgrade to a ublox M8T which can be 'surveyed in' to put it into 'overdetermined mode' allowing it to retain a locked PPS output even when only a single valid SV signal is available (the navigation types such as the M8N require a minimum of four good SV signals to maintain a valid locked PPS output - not a problem when feeding it from a cheap active puck antenna mounted where it has a clear all round view of most of the horizon).

 With the MK II, I only see around a 6 to 7ns Pk-Pk phase wobble on a time scale of a minute or three which makes syntonising my LPRO 101 to the GPS atomic time reference much easier. I've connected a ten turn helipot to the C field input, padded out by a factor of ten with a pair of 22k resistors either side to provide a much finer adjustment of the Rubidium's frequency than is available from its internal ten turn trimpot.

 Even with such fine control, it only takes a degree or two of rotation to discern its effect after allowing the minimum of  two or three hours required to identify the drift against the GPSDO's background phase wobbles. The toughest part of this syntonising process is resisting the urge to fiddle with the helipot before enough time has passed by to properly identify how much, if any, drift has occurred. In this, patience is most definitely a virtue.

 Since encasing my fan cooled LPRO in a 2cm thick layer of polystyrene foam about three weeks ago to minimise changes to the internal thermal gradients due to variations in room temperature, I'm now routinely seeing some 30 to 60ns net daily phase shifts between it and the GPSDO as recorded by the infinite persistence displayed after a 24 hour run using an SDS1202X-E to compare the output waveforms.

 This shows that it's possible to syntonise a rubidium oscillator to better than 10E-10 with a very basic GPSDO. If you're not planning on doing anything more with your rubidium oscillator than simply mount a large passive heatsink to its baseplate to prevent it overheating as most video bloggers appear to have done and leave it at the mercy of room temperature variations, a simple GPSDO will more than suffice to syntonise it to around the 10E-9 mark. Even the more simplified version of Gyro's that I described should be good enough for the job.

 The component parts won't go to waste since the expensive bits will be needed anyway when you eventually do come round to adding a microcontroller into the mix and the experience gained will ultimately prove rather useful.

 I've attached four images which might prove inspirational (if inspiration was needed). 

[bob91343]

Johnny B Good I read your message and frankly it's mostly over my head.  All I want is 10 MHz but it appears that getting it isn't as simple as I expected.  While I have an engineering degree, it predates all this modern stuff by many years.  I have a reasonable grasp of the fundamentals and more, and have texts which I can peruse, but I am mostly lost in this morass of integrated circuits, digital technology, and software.  I don't hide behind my age, since I don't see that as a factor.  But my proclivities and talents lie elsewhere than what it seems to require to manage a project such as this.

So I was hoping Andrew would supply a circuit board and I could install the components and have a complete project.

If someone can give me a simpler route to my destination, I am ready to try it.