Hi John, thanks for the comment. This was my very first UKposts video and I was rather doubtful whether I should make these videos and if anyone would watch and find it useful. It's comments like yours that make it worth it.

I put it up here: github.com/TheHWcave/GPS-FreqStd

I am afraid the CD4016 is not suitable for this as you have found out already. It is a switch for digital or analog signals. The in/out path through this chip doesn't improve the signal quality in any way (probably making it slightly worse) and the controls (to turn the output off or on) max out at 10 Mhz which isn't nearly fast enough. Try to get the 74AC14 for best results but even a plain 74LS14 will improve the output

I don’t think it feasible. Originally I was about to suggest what is effectively a non-retriggerable monoflop (see en.wikipedia.org/wiki/Monostable) with programmable delay. For example, 10 Mhz is made of a sequence of some pulses that are 63ns (8 Mhz) and some that are 40ns (12 Mhz) wide. You could feed the PPS signal into a monoflop with a pulse width of 50 ns. Regardless if it was triggered by, say the rising edge of a 40ns pulse or 63ns pulse, it will always produce a 50ns pulse on the output. The trouble is that while this now produces pulses that all have the same length, the gaps between them still vary, so its just another form of jitter. I am afraid, I think the best way is to use PLL, basically making your own oscillator (jitter free, and could even be sine instead of square wave) and use a PLL circuit to lock its frequency to the PPS signal. You might be able to do this all in software in a micro controller which would be an interesting but rather challenging bit of work! I never bothered with that. As I said the frequency (as in pulses per second) is always 100% correct, just the pulses are not always the same length. For testing the accuracy of counters or frequency readouts in multimeters, jitter doesn’t matter.

I agree that the module probably does more complex things internally at different clock speeds. I thought I made it clear in the video that the 24 MHz was just something I came up with to make determining useful frequency divisors easier. Apologies if this was misunderstood.

@@TheHWcave - zero apologies needed and nothing was unclear. I liked the video and the logic you used (pun intended). I just so happen to be reading the reference manual (for different reason) and went off to see if I could find more on this crystal frequency issue. Keep up the good work!

A very good question. I need to find a way to revive the UBLOX application on a suitable Windows machine to test all the variants of non-50% duty and locking or not locking it. I have not used the UBLOX software since making the video because I wrote my own little program for Linux that just sends the desired frequency to the chip. I'll have a look but it will be a some time before I can get to it.

I did a quick test, and changing the duty cycle did not help. In fact it can make things worse in that even jitter-free frequencies now start jittering

@@TheHWcave I looked how pll works. It assumes a symmetric distribution of the input signal around the lock, which is defined as 0. The input is low pass filtered to determine the vco correction. A duty cycle other than 50% would spoil this. So it would not work, unless some correction is applied

Thanks for the info. So unfortunately jitter is going to stay, but as I said, that doesn't stop the usefulness of the GPS module as reference standard (as long as you measure just frequency)

My thinking regarding using Skullcom's divider was obviously backward... We need a multiplier. Start with a safe frequency and increase it to the frequency that we want. I know that there are such devices, but nothing much of them. Looks like 2x and 3x factors are common?

To be honest, I'm not using this reference anymore, I "upgraded" to a GPSDO from Leo Bodnar, their mini model, which is pretty much what you are suggesting. It only has a TCXO but I keep it on permanently and so GPS keeps it nice and steady. This also contains a processor but just to control the PLL hardware and set dividers, not doing any maths. It provides a pretty good 3.3V signal selectable to (almost) anything from 400Hz to 810 MHz. I don't think a pure software solution is going to work with sufficient accuracy. Looking at the stuff inside the GPSDO, a proper solution is quite complicated in terms of special chips needed and registers to control. Using a couple of simple ripple dividers 7493 etc is a lot easier but then you only get frequencies that are divisible by a ^2. I decided on the GPSDO because I like that I can pretty much select any frequency, including much higher ones, however 99.9999% of the time, it is set to 10MHz and connected to my frequency counter's ext ref input. So, actually the UBlox could do the same for that. Actually, now that I mention it, that seems like a good idea for reusing that chip and maybe put it directly into the counter, a HP 5334B (it has just a simple 10MHz crystal oscillator as time source, the OCXO option is not fitted)

There isn't a 24 crystal per say, all I found is that mathematically, it behaves as if the safe frequencies are full multiples of 24 MHz or 12 or 6 or 48, your choice. What really happens is that the chip allows you to insert additional pulses within the frame of the GPS 1 PPS - pulse-per-second. The PPS is normally a highly accurate 1s signal synced and controlled with the GPS time. Using the software, you can ask the chip to put out n additional pulses within the 1 second frame. As I found, the chip uses pulses of the same width if it happens to be a safe frequency but mixing longer and shorter pulses if they are not. So what is guaranteed is that if you asked for say 12345 Hz, you will get 12345 pulses per second (within a 1s PPS frame) precisely. However, these pulses will not all be 1/12345 = 81.004... us long. Some will be shorter, some will be longer, but there will be exactly 12345 of them and they repeat exactly every 1s with GPS precision. Using this as a frequency standard works only if you count pulses per second. If you use shorter gate times, you should use only safe frequencies. As I said, the chip tries to do a good job spreading the desired number of pulses evenly across the fixed boundaries of the 1s PPS but to illustrate the principle, it might just as well bang out a short train of 12345 pulses of 1us length (0.5us high, 0.5us low), followed by 987.655 ms of total silence. Same result, "12345 Hz" .This is very precise but very different from a normal crystal oscillator. In my experience it works brilliantly for testing and adjusting frequency counters for all frequencies up to 10 Mhz and a bit beyond, but for signal purposes, I would generally stick with safe frequencies only.

@@TheHWcave Thank you so much for your reply. It has been incredibly useful. So concluded, the chip can only guarantee an accurate 1Hz output. Anything above that is digitally generated in the chip and may be full of jitters and uncertainty. It is useful and good for testing and calibrating equipment with a >2sec gate counter, but not suitable as a clock source or a crystal oscillator alternative.

Hi Dragan, you are right, using the GPS PPS to drive a clock gen. chip is a much better solution. In fact I got one already, using a MAX M8 to drive a SI5328. This came from Leo Bodnar (mini GPS ref. clock) and can produce 400Hz to 810 MHz.

Hi David, finding a crystal freq. of 26 MHz confusing. I would have expected 24 as it seems all frequencies that divide 24 with no remainder are jitter free, like 12, 8, 6, 4 ... but of course not 10. I am very curious about what happens if you try injecting an external frequency. Please update the comment or email me with the result.

@@TheHWcave hi, i desoldered the on board 26Mhz tcxo and connected si5351 output#3 directly in its place. i started with 26Mhz. the neo-7 works as usual. then i changed the frq to 21.66666667Mhz and set the TP5 register for outputting 12Mhz and i got jitter free 10Mhz from the neo-7. however serial communication with the neo and i believe that the chipset (other then the pll divider) isnt working. it needs the 26Mhz as a clock. i will put back the tcxo and connect the neo7 as external clock to the si5351. that way i will have a steady gps clock for the si5351 and could get 10 mhz with no jitter and any other frq on any of the 3 outputs.

Very interesting. Which variant of the si5351 are you using? The C version CLKIN could also be driven by 12 MHz which is jitter free from the neo7. I have not looked at the programming of the si5351 and that would make a big difference in the choice of output frequencies.

@@TheHWcave i have the "A" version. the "C" version is a beast :) you can get almost any frq you want jitter free with the SI5351. it also has a clock builder application for desktop to help you find the correct dividers to get jitter free frq.

Those Silicon Labs chips are great, no doubt but don't get me started on those desktop apps that work only with Windows and then sometimes only when connected to the proper development board.... and no documentation in data sheet or application notes on the algorithm for those without the board and running Linux. With lots of effort I managed to work out my own algorithm needed to set the register combinations for the Si5328 to do (almost) any frequency jitter-free and GPS locked from a few kHz to 800 MHz. The hardware uses a board from Leo Bodnar based on a u-blox M8 GPS module driving a Si5328. So, conceptually very similar to what your are doing. I am using it with my own software, One of these days I plan to make a video about it but too busy with my day job at the moment.

The micro USB provides power and is used to set frequency. As an additional benefit, you also get GPS signal quality (number of satellites.., position (lat/long) + height, + UTC time and date.

Sehr interessant. Ich sehe dass du dasselbe Jitter Problem mit dem 10 MHz GPS Signal hattest. Gute Idee einen speziellen 10 MHz Ausgang mit eigenem Filter einzusetzen. Für alle anderen Frequenzen hast du dann aber immer noch dasselbe Problem wie ich, und nur die Frequenzen die ohne „Nach-Komma Rest“ aus 24 MHz geteilt werden können sind wirklich sauber. Mein GPS Video was mein aller erster UKposts Versuch und ist entsprechend… Werde mir deine anderen Videos im Lauf der nächsten Zeit ansehen. (PS stelle fest: habe komplett verlernt Deutsch zu schreiben)