Ublox-NEO-M8N GPS navigation signal amplify module for arduino Rasppery PI

[Johnny B Good]

It's generally assumed that OCXOs have a EFC input (unless otherwise specified) and they are usually referred to as just OCXOs and not VCOCXOs.

Also, retrace describes how an oscillator freqency changes after a disturbance like a power failure.  This change is usually not due to just temperature changes.

 That assumption (ASS-U-ME) on the part of the sellers might well explain my frustrated attempts at finding VCOCXOs on Ebay. 
I'll have another go, looking just for OCXO modules, and take a closer look at the results to see if any are actually VCOCXOs being misrepresented. Sadly, that assumption works both ways, i.e. the buyer assuming these OCXOs are electrically tunable OCXOs (VCOCXOs).

 Changes due to power interruptions are essentially thermal issues. Even brief interruptions can be sufficient, even though thermal inertia suggests otherwise, simply on account of the self heating effects of the oscillator's amplifier dissipating heat into the crystal as well as the milliwatts or micro watts of energy dissipated by the vibrational energy in the crystal itself. When we're dealing with stability requirements specified at less than 1ppb, it takes very little thermal disturbance of a hard won equilibrium to upset this level of stability, giving the retrace phenomena another chance to make itself known.

JBG

[texaspyro]

Even when/if the OCXO is back to the same temperature, the frequency will be different.  This is what is meant by retrace.  It is caused by lots of things (some not well understood).  Some factors are changes in the oscillator / crystal mecanical structure / stress changes,   micro-contamination settling on the crystal,  gremlins waking up and pooping on the gizmos,  etc.

Almost nobody uses the term VCOCXO... save yourself some grief and just look for OCXO.

[Johnny B Good]

Also cheap OCXOs are not known for having low phase noise, etc.   You get what you pay for... or not.  New, quiet, stable, high performance  OCXOs can easily set you back over $1000... some are over $30,000.  Price goes up exponentially with performance.,  a very few extra dB can easily double the price.  I'd say start with a cheap used one and then go from there.

You are not going to get quiet, low phase noise out of a cheap OCXO or wall wart.

 Those Connor-Winfield units looked to have decent phase noise figures. Do you happen to have a short list of the older VCOCXOs worth looking out for on Ebay with equal or better close in phase noise figures that you'd care to disclose?

 As for the cheap wallwarts, they can be a moveable feast, some terrible and others quite good for what they are. In any case, it's best to keep the smpsu out of the box and relegate it to a mains outlet and use the saved space to install a good quality screened LPF on the 5vdc input circuit to deal with that problem.

 A high quality low noise smpsu is essentially the noisy switching converter, screened off from the output filter section within the metal enclosure that prevents the rest of the appliance (GPSDO in this case) being polluted with unwanted EM noise. I'm just trying to split those two components so I can make use of a cheap switching psu separated from the expensive LPF within the box.

 I do appreciate your concern but this project is as much a way of gaining experience as actually producing a GPSDO to be proud of. It's my project so I'll be making all the pragmatic decisions in its development, including using cheap wallwarts without benefit of a quality LPF just to see for myself why this might not be such a good idea (i.e gain a much better understanding of the scale of the problem and the measures required to deal with it.) 

 As for a (relatively) cheap brand new VCOCXO not being a match to the thousand quid priced offerings or possibly even the old school high quality units that (apparently) can be had second hand for a song on Ebay (if you're lucky and know what to look out for), I'm not so sure the difference is as great as it once was these days, nor even if the difference is worth expending the extra effort and/or cash on.

 Obviously, I'm going to continue looking for second hand bargains on Ebay whilst I hold fire on those Connor-Winfield units until I've had some face time next month with the traders at that radioham rally I'm now committed to attending (the hotel room has now been booked). I might even buy a basic VCXO if the price is right and it can be shipped to me from a UK based warehouse within the week, just to have something to experiment with, possibly even test PLL/microcontroller/16 bit DAC configurations on.

 I suppose I should get to work making a short list of components to look out for at that rally. I've only got a month to get my shopping list sorted out. 

[Johnny B Good]

Even when/if the OCXO is back to the same temperature, the frequency will be different.  This is what is meant by retrace.  It is caused by lots of things (some not well understood).  Some factors are changes in the oscillator / crystal mecanical structure / stress changes,   micro-contamination settling on the crystal,  gremlins waking up and pooping on the gizmos,  etc.

Almost nobody uses the term VCOCXO... save yourself some grief and just look for OCXO.

 I see we're both on the same page regarding the retrace issue.    This is why Temperature Compensation is a poor second best to Oven Control in regard of frequency stability.

 As for your advice to use the search phrase "OCXO", I'm already on it.    I just haven't gotten round to an actual search just yet.

[texaspyro]

The Trimble 37625 has excellent phase noise but also some termerature sensitivity.  Other Trimble OCXOs are also usually rather good.

The HP-10811's are usually rather good.

The Morion MV-89A's can be rather good.  They do have a known issue where some of them have a poorly soldered output cap that can cause the output signal level to drop to unusable levels.

Some reading:
http://www.ke5fx.com/gpscomp.htm
http://www.ke5fx.com/tbolt.htm

[Johnny B Good]

The Trimble 37625 has excellent phase noise but also some termerature sensitivity.  Other Trimble OCXOs are also usually rather good.

The HP-10811's are usually rather good.

The Morion MV-89A's can be rather good.  They do have a known issue where some of them have a poorly soldered output cap that can cause the output signal level to drop to unusable levels.

Some reading:
http://www.ke5fx.com/gpscomp.htm
http://www.ke5fx.com/tbolt.htm

 The NARSA radioham rally last Sunday was a bit of a disappointment but I did manage to snag a CQE 13MHz OCXO for just 4 quid, along with a (pitifully) few other odds and sods. After much searching I managed to track down a pin out diagram of sorts (see attached images) which allowed me to tentatively test it out on a 5 volt supply using croc clip leads with a solderless breadboard and its USB/9-12 volt wallwart adapter board powered by a cheap Poundland 1.2AH battery bank.

 Surprisingly, it actually fired up at this voltage and I could see the peak current draw of just 280mA courtesy of a cheap usb power monitor. The current eventually started to reduce after about 5 to 6 minutes from a cold power up. The transition to a steady 200mA was relatively quick (perhaps just 20 to 30 seconds) and I could tune 12Hz down and 5 or 6 Hz up from the 13MHz frequency I'd set my FY6600 to for comparison using a 50k pot across the 5v and ground to feed a tuning voltage to the Vfr control pin.

 The Vref pin only had about 3.3 volts on it which wasn't high enough to tune it to 13MHz. This fact, along with the relatively low ratio between cold and warmed up heater current levels, strongly hinted at the possibility that it may have been a 12 volt part that just happened to be able to function off a 5 volt rail but I knew that applying that much overvolting to a 5 volt part would almost certainly be fatal so I tread rather carefully in my experimental voltage probing attempts to determine whether or not it was 12v tolerant.

 I tried applying 6.8v from one half of my 13.8v 10A smpsu constructed from a pair of Advance 5v 10A smpsus, manufactured some time around the late 70s, stacked in series and maladjusted to 6.9v each via a half century old multimeter to monitor the current draw. The multimeter has a very sensitive 50K ohms per volt rating but the penalty for this is the need to drop a miilivolt per milliamp when on its highest DC current scale setting of 600mA FSD which left only some 6.5v during the warm up which, strangely, didn't go any higher than the 280mA it had drawn off the 4.9v I'd initially tested it with.

 I then tried a brief test using a NiMH cell held in series to bump the voltage up another 1.3v to see what sort of oscillator output voltage this would produce as well as to check on the oven heater current. The heater current still didn't go above the 280mA limit but the output voltage rose higher again than the initial 3 volts square wave I'd noted on the 4.9v supply and the circa 4.5 volts p-p on the 6.5v supply which was rather worrying.

 Even the 12v rated parts don't give more than 4v p-p on their oscillator signal output pin and this looked like it was trending towards providing a 10v p-p from a 12v rail suggesting the lack of a 5 volt LDO regulator chip implying the use of 74HCT parts powered directly from the 5 volt Vcc pin.

 I chickened out of applying any more voltage, what with discretion being the better part of valour and all that. However, as much as it seems to be a 5 volt OCXO (the rather low Vref voltage and the small margin between the run up from cold heater current (limited to 280mA) and the 'on temperature' current draw of 200mA suggest otherwise), I'd like a second opinion from the more experienced of the EEVBlogging community.

 Since a picture is worth a thousand words, I won't describe the OCXO in detail other than to mention that it takes a couple of minutes from cold to produce a 13MHz waveform on the 'scope. Prior to that, it looks like a much lower, unsynchronised signal, possibly the fundamental since it's not uncommon for third overtone SC cut crystals to be used in OCXOs.

 Anyway, now for the attached images I promised. The final image is one I downloaded from an Ebay listing for a 13MHz Vectron unit that looked like a relabelled version of mine. In this case, it's for a 12 volt OCXO, hence my request for thoughts on the question of voltage requirements since despite all my searches, I haven't been able to find a definitive answer. It has all the appearance of being unused NOS, possibly of some vintage when it was more common to power the oven heater from a 12 volt rail. For all I know, it may be of relatively recent manufacture when 5 volt only designs started to become more commonly available.

 I'd very much appreciate any thoughts and opinions you may have to offer on my chance purchase (I've already sorted out how to generate a 10MHz reference using this OCXO). 

Please note that the black fine tip marker writing on the underside identifying the pin functions, is all my own work, inspired by the information in the last image. When I bought it, it had no markings at all nor any signs that it had ever been marked up by a previous owner.

 Yet another thing to note is the 10 volts tuning range given in the final image. I discovered that the tuning range upper voltage limit appears to be clamped to the Vcc rail voltage when I connected a 9v PP3 to my 50k tuning potentiometer in order to see if I could extend the tuning range. Although I didn't check out the tuning range on the higher voltages, I did note a significant undershoot of the heater current down to a low of 50mA after dropping from the 280mA before recovering to 120mA resulting from the brief interruption of the supply voltage when inserting the 1.3v NiMH cell in series with the 6.9v supply. I noticed a similar undershoot effect when reconnecting back to the 6.9v supply (recovering to 150mA in this case) after removing the NiMH cell from the circuit. This effect was, afaicr, absent when powered from my 4.9v supply so this may be indicative of its 5 volt nature. Just a thought that, in this circumstance, I felt it was best to share with everyone.

[EDIT 20190504]

 Subsequent testing using these cheap power banks after making up a second NB3N502 clock multiplier SOT8 to DIP8 adapter module to drive the 74193  divide by 13 chip with an OCXO sourced 26MHz, also revealed the same heater current undershoot behaviour. It wasn't quite so pronounced but I guess reliance on a time shared digital display of volts and amps makes such behaviour a little easier to miss.

[Johnny B Good]

Further OCXO developments since my last post:-

 I've attached a couple of image files which may be of interest. The collection of 10MHz OCXOs shown in the first image was kicked off by the 13MHz example (also shown) that I'd snagged at that recent radioham rally for just 4 quid which I'd used to drive a 2 times clock multiplier to drive a divide by 13 (74193) which then drove a 5 times clock multiplier which gave me an ultra low jitter 10MHz square wave all from a 5 volt rail which gave me the tempting idea of installing it into my FY6600 to upgrade the previous TCXO upgrade on the generator's original and rather execrable 50MHz SMD XO chip.

 However, in my search for a data sheet for this 5 volt part, I chanced upon a uk based ebay trader offering those 10MHz 12 volt OCXOs for just a penny shy of a fiver each. I bought three just over a week ago then another four a few days later, hence the collection. At just a fiver each, one can never have too many 10MHz OCXOs. 

 One is now fitted to the AWG (along with an additional 12v 300mA smpsu board recovered from a wallwart) . Whilst the existing 5v rail in the FY6600 would probably have handled the 280mA warm up load of the 13MHz unit ok (dropping to 170mA or so once up to temperature), I felt the additional 70 to 90mA from the required 74193 and a pair of NB3N502 clock multiplier chips would prove to be a back breaking straw for my FY6600 camel, hence the use of one of those 10MHz OCXOs with its own dedicated 12v smpsu board.

 The additional loading once the OCXO is up to temperature is just a mere 1.3W which isn't a thermal issue for the generator since I'd long ago added the small cooling fan it had been so desperately in need of to begin with. Assuming I'd have been able to get away with using the existing 5 volt rail to power my 13MHz OCXO sourced 10MHz (substituting the first NB3N502 with a less precious dual input XOR gate and RC delay doubler), I'd most likely have seen a similar increase in power consumption in a swings and roundabouts choice.

 The use of one of those 10MHz OCXOs along with the complication of an additional 12v smpsu did allow me to keep the OCXO at temperature whilst still plugged into the mains supply when the generator is switched off. The alternative would have required it to be put into standby to achieve the same effect at the cost of an additional 5 watts or so of "standby" consumption.

 Now that I have a "Lifetime's Supply" of 10MHz OCXOs, I can disclose my source for the benefit of anyone here, looking to DIY a GPSDO on the cheap.  He's still got "more than ten available"    I must point out that the description is less than honest in that they're actually ex-equipment rather than NOS (I could see the odd through plated hole remnants attached to a pin or two) but, at that price, I could hardly complain since they're otherwise in very good condition with plenty of tuning control range left in them.

<https://www.ebay.co.uk/itm/CQE-CRYSTAL-OSCILLATOR-10MHz-REDUCED-TO-CLEAR/253081992039?hash=item3aecdcbb67:g:x1gAAOSw~qNZh2rl>

<https://tinyurl.com/y6pl2xvk>

 The second image is my attempt at showing the effect of using a couple of 10MHz Xtals to filter the rather jittery and saw tooth corrected 10MHz square wave output on the PPS line of that u-blox M8N based arduino.Rpi GPS module. For a still image, it's a pretty fair representation but what you can't see is the rather toned down sawtooth correction jitter (4 per second to one every few seconds - even tens of seconds).

 Although not perfect, it's a surprisingly huge improvement over the raw PPS signal I'd previously been using as a frequency calibration source. My "filter" simply comprises of a pair of 10MHz crystals in series with a 4K7R forming the vertical of the Tee, about as minimalistic as you can get but very effective just the same. 

PS Forgot to mention: These 10MHz OCXOs seem a close match to the details in the data sheet I found for that 13MHz Vectron unit, except for the fact that the output is an ac coupled 4Vpp sine wave rather than a dc coupled square wave. The control voltage range is also 0 to 10 volts (will go a little higher in fact) and is of positive polarity.

[Johnny B Good]

Hello All,

 This morning, I'm feeling rather sorry for myself (I know, self pity is not the best of emotions to admit to but I'm basically, at heart, an honest person).      Some ten hours earlier, I'd done a rather stupid thing and blew out the PPS line with a 12 volt jolt and my "navigation signal amplify module" (the u-blox M8N custom chip to be precise) is now useless as a discipliner of XOs. 

 Curiously, the u-centre app remains blithely unaware of this fact (and the rather high temperature (67 deg C according to my IR thermometer) that it is running due to the additional 330mA it now draws from the USB port) and it still shows good satellite reception and position data. It's not completely broken but the only use it can now serve apart from being a donor for parts, is as a GPS antenna testing device. Either way, it's headed for the "Possibly Useful Spares" (PUS?) box where it can also provide a memento of my fallibility.

 Basically, I'd decided to try a quick experiment to test out some options for "Injection Locking" one of the seven 10MHz twelve volt CQE OCXOs I'd fortuitously acquired dead cheap last week. I'd worked out that I had three possible ways to inject the 10MHz PPS signal into the OCXO module, one of them being to apply a sample of the 10MHz PPS to the frequency control pin (the other two being to couple the injection signal onto the 12v supply pin or the Vref pin).

 Initially, I just took the 10MHz sine wave on the output of my crude xtal filter which I knew would provide the required DC isolation for these experiments. Unfortunately, in my haste to repeat my experiments (which had failed to offer the slightest hint of any response) with an even stronger 10MHz signal by bypassing the filter, I'd overlooked the fact that I'd also be losing the filter's DC blocking effect (silly me!  ) and forgot to interpose a coupling capacitor for the following tests when I tried feeding the PPS output directly to the frequency control pin.

 This wasn't harmful to begin with since the voltage, fed from a 50K pot, was only around 3 1/2 volts to start with. Unfortunately, I forgot that this 12Kish impedance 3 volt supply could (and did) land up as a low impedance 12 volt source in the event that I were to turn the pot to the hot end of its range. Of course (and with "The Lord Murphy" looking over my shoulder) this is exactly what I did! 

 Thus do the "Mighty" (of arrogantly self perceived intellect in my case) fall.    This is the point in my "Tale of Woe" where you, my readership, can take pause to deal with your emotional responses to my misfortune. Some of you no doubt, will be shedding a tear or two, either out of empathy for my plight or else, simply for the sheer joy of schadenfreude. Whatever the case, I'm in no position to judge since I've enjoyed all of these emotions myself in the past so, "knock yourself out" as those "Goddam Yanks" are wont to say. 

 I'm not seeking any sympathy here since I can only blame myself for blowing away some 21 quid's worth of my hard earned in a moment of utter carelessness. The only consolation is that my 21 quid investment hadn't been a complete waste since I did manage to get some use out of it and learn a thing or three about the issues involved in establishing accurate and stable frequency references based on readily accessible GPS timing data (given a fully functioning GPS module) during the short 3 1/2 months it had been in my care.

 I'm aware of the well known saying, "You learn from your mistakes". In my case, what I've realised is that, at less than 22 pence a day, I can afford to maintain this rate of GPS module attrition for the rest of my life even if I were to perversely ignore the obvious lesson.

 As I mentioned, I'm not seeking any sympathy (just some shopping advice now that I need another GPS module). However, since one of the many lessons learned is that I could have chosen a GPS module better suited to the task of  disciplining an OCXO, I am now seeking advice on  candidate replacement GPS timing modules. I promise to take better care of my next one... honest!

 Please don't let my rank carelessness discourage any of you who may otherwise be so inclined from offering advice or making recommendations. Although I've come to realise that a good OCXO is the primary consideration in building a DIY GPSDO with the GPS module relegated second in importance (virtually any GPS module with an accessible PPS line will suffice), I'd still prefer to choose one that more readily lends itself to this task if possible. Obviously, even a cheap 'n' cheerful module will allow me to experiment further in my new found interest in GPSDOs. 

 In the meantime, I have attached a photograph of the injured module for "Education and Training Purposes"  There's no obvious sign of damage to the u-blox custom chip which directly drives the PPS line (sadly, no intermediate discrete transistor or commodity driver chip to be swapped out or bypassed in an attempt to effect an economic repair. ).

[Johnny B Good]

 
Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.

 Ignoring the above notice (It's my thread after all ), this is intended as a final follow up to provide closure for the sixty or so intererested enough to view the attached image of my delidded M8N module since that last posting way back in May.

 First off, I have to say that that "Ublox-NEO-M8N GPS navigation signal amplify module for arduino Rasppery PI" I had been seeking opinions on from anyone who may have bought one, turns out to have been that rarest of beasts, an actual genuine U-blox module, unlike the three other U-blox based navigation modules I've subsequently purchased off Ebay since then. In hindsight, it had proved to be the best of the lot (although the final unit I bought earlier this month just might have the edge in terms of marginally improved position fixes as plotted by U-centre's deviation map).

 The first replacement module was a NEO-6M at just 4 quid delivered. Surprisingly, the flash ram worked (no need to add a CR2032 for long term (circa 18 months) backup of settings) and it came supplied with the tiniest of active ceramic patch antennas I've ever seen (pictures attached). It had only an IPEX connector with not even the solder lands for an SMA socket and only a serial interface which at least, did include a PPS pin.

 Once I'd gotten it interfaced to my winXP VM via an FTDI usb2serial adapter, I discovered, via U-blox's rather klunky u-centre application program, that the PPS option was limited to a minimum time period of just one millisecond (effectively 1KHz maximum pulse frequency - Oh, JOY!!!  ). Undismayed, I lashed up a basic PLL with a 10MHz OCXO on a solderless breadboard using two 74HC390 ICs to divide the OCXO frequency down by 10000. This was then followed by a week or so of frustration as I tried to get this lashup to phase lock for more than ten minutes at a time. With a maximum tuning offset of just 10Hz or so and a 10000 frequency ratio, the process of waiting for lock was rather glacial to say the least.

 Eventually, when it finally occurred to me to monitor the PPS signal itself, I discovered a very peculiar "quirk". Effectively, with a 50/50 mark space ratio, the 'mark' half of the PPS pulse would start slowly shrinking in width about 7 to 10 minutes from startup (the repetition rate stayed at 1KHz throughout) and then disappear briefly up its own fundament another 7 to 10 minutes after that before reincarnating itself as a 'Full Fat' 50% ratio square wave to start the next cycle of 'weirdness'.

 No wonder I'd been slowly going round the twist trying XOR and 4046 PLL techniques to get the OCXO to lock to this bastard GPS module! I don't know whether it was just some weird peculiarity of this particular (or batch of) module(s), possibly only occurring when the default PPS rate was altered from its normal one or ten millisecond wide pulse at 1PPS but since I was never going to entertain the even more glacial prospect of phase locking to a 1Hz pulse, I never bothered investigating this weirdness any further. I did start a thread topic on this but no one ever came back admitting to being caught out by this peculiarity.

 Obviously, that was a case of "Scratch one cheap bastard M6N module and be a little less cheap about my next U-blox module purchase.", hence my spending a little over 15 quid on an even smaller M8N based module. Despite its small size, it actually possessed both an IPEX and an SMA antenna socket. Unfortunately, it was an out and out fake (no flash, nor it seemed, any SAW filter). However, the lack of flash was just a matter of piggybacking a CR2032 cell holder onto the module and accepting its weird response to signals from the external active antenna and an indoor active unit I'd picked up at the Blackpool amateur radio rally (Hamfest) which showed very similar signal strengths when I knew for a fact that the external antenna was at least 10dB up on the indoor antenna.

 Despite this signal strength anomaly, it did seem to perform quite well, and being an M8N unit, I could at least program the PPS to output 100KHz by which to phase lock the OCXO using just a single 74HC390 to get the required divide by 100 which seemed the optimum middle ground between using a 1KHz pulse (I was curious enough to try this with a module that wasn't going to sabotage the proceedings by having the PPS disappear up its own backside every ten to twenty minutes) and a 10MHz pulse feeding my 7486 XOR gate based phase detector now that I was free of the need to use a crappy 4046 PLL detector as I'd been forced to resort to with that bastard M6N module.

 I actually got as far as laying out a vero-board for mounting into a neat aluminium case and soldering a support spacer to the serial port pins of the module to mount it on end right next to the FTDI adapter, fitting the aforementioned CR2032 battery holder and removing the SMA connector (I was going to use an SMA socket with IPEX ended flylead on the back panel for this build). However, before I got around to this stage of the build, I spotted another neat looking M8N navigation module for just over £7.50 delivered (there was postage on this one) which looked worth taking a punt on.

 This unit was also a fake but only in so far as it too lacked working flash. The signal strength reporting behaviour was a lot closer to that of the original M8N on both antennas but something like about 6 dB down rather than the 16dB down (on the external antenna) of the previous fake. This came with the bonus addition of a rather neat 1 inch square active patch antenna with a one inch length of IPEX connector terminated co-ax.

============================================================================================
[EDIT 20191129]

 I decided to connect this first M8N module to the PC (it rather conveniently includes its own built in micro USB port) to refresh my memory of witnessing 55 to 56 dB C/N signal strength reports that I would occasionally observe. This time, however, the signal strength reports seemed little to no better than my latest module's reports. Also, I discovered it had lost its ability to save configuration settings to the FLASH (or the I2C-EEPROM or the SPI-FLASH). I'm absolutely certain it was able to save its settings without selecting the bbRAM option otherwise I'd remember lashing up a CR2032 to it which I'm certain I hadn't felt any need to do.

 This came as a bit of a shock to me since it brings my powers of recall into question. Quite possibly the loss of this feature may be down to collateral damage that went unnoticed at the time when I blew the PPS output with a 12v jolt, possibly a latent fault which only now has finally revealed itself. Anyway, even when I bypassed the 5 metre antenna extension cable to maximise the signal strength reports, I never saw it go above 53dB. By comparison, when I did this test with the current module, I only saw brief reports of 52dB peak strengths so not as big a difference as I thought.

 Possibly, in the intervening 6 months, the mag mount patch antenna and its 5 metre co-ax may have deteriorated due to damp ingress in spite of such antennas normally being weather-sealed for use as a car roof antenna. However, the impossibly ultra low loss behaviour of my 5 metre RG174 extension cable (just 2dB when the loss per metre figures for this cable indicate an expected 6dB +/-1dB of insertion loss) still seems to apply with both the oldest and the newest examples of M8N based navigation modules.

 I've been running u-centre with this first module connected on the end of the antenna extension cable for the past 12 hours or so and my impression now is that its signal strength reporting is a pretty close match to this last fake module which I'll be using in my GPSDO project rather than being some 6dB down as I originally stated.
===========================================================================================

 Although it lacked an actual SMA socket, it did at least have the solder lands to take one. Since this was a little bulkier than the other module, I decided to rearrange its mounting in the case and solder a right angled SMA-F connector to it to mount it onto the back panel. I'd acquired a few different types of SMA sockets just after I'd purchased the genuine M8N module and this seemed a good way to make some use of this stock of PCB mount SMA sockets.

 I've been bogged down with this DIY GPSDO project over the matter of how best to implement the necessary supply of 5 volt DC required. I've been testing a couple of dc-dc converter modules bought from Banggood a month or so back with this project in mind since I want to avoid the need to bolt any heatsink tabs on analogue regulators to the case and the effect of varying levels of heat when powering it from 9 to 15 volt DC wallwarts.

 That and the fact that I don't want to have to deal with more than an absolute minimum of fixing screws whenever the need to repair/modify this homebrewed GPSDO arises. Indeed, being able to slide the whole board out of its case and be able to power it up with less fear of overheating a voltage regulator during testing is the prime motivation to finding an effective voltage stabilisation solution.

 Initially, I was just going to use a well filtered dc-dc converter module to directly provide the required 5 volts but I'm now thinking of using low noise LDO regulators from a dc-dc converter providing somewhere between 6 to 7 volt output. However, that plan is now in hiatus since I blew the small, low ripple output converter in a meter probing accident and the other, larger unti was generating almost 1 volt pk-pk ripple on its output (until I solder blobbed an open circuit jumper bridge to connect the Shottky diode to the output inductor, reducing the level to a mere 400mV).

 Hopefully, this was just a rogue bad part (I've ordered a couple more similar units to experiment with along with another four examples of the smaller module I burnt out with careless probing). In the meantime, I've decided to connect just the FTDI usb2 serial converter and the GPS modules together which I can power over the usb link, allowing me to run some tests whilst I wait on those dc-dc converters, an "LCR-T4 12864 LCD Graphical Transistor Tester Resistance Capacitance ESR SCR Meter" bought simply because it was so cheap at just £4.73 delivered, an "Original Hiland DIY M12864 Graphics Version Transistor Tester Kit LCR ESR PWM" (a high end version at £14.28 with discount),  an "Original Hiland DIY M8 LC Tester Digital Inductance Capacitance Meter Kit" (12 quid with discount - it doesn't test ESR but it covers an extended range down to 1pF and 0.1μH values which the ESR meters don't touch) and a pack of three terminal PCB mount screw terminal blocks that are not supplied in that LC tester kit.

 All the above items were ordered just a week ago from Banggood. I placed individual orders for each item. You don't get any extra discount by integrating them all into a single delivery so it's best to keep them separate to avoid needless processing and shipping delays due to any items not being in stock. With a bit of luck, I might start seeing a trickle of small parcels by the end of this week.

============================================================================================
[EDIT 20191129]

 For those of you wondering why I even mention my BG orders of kits and components, the reason is simple - I realised, whilst trawling through my 'salvage' extracted from smpsus and motherboards looking for capacitors and inductors to make up filters to reduce the ripple noise of these dc-dc converters (and that originating in the 9 to 15 volt wallwarts I plan on using), that a convenient way to measure inductance and capacitance (and also ESR) would facilitate my picking out the required parts from my 'salvage' which undoubtedly will include pretty well all the required items, saving me having to specify then order brand new parts at both monetary cost and, perhaps more distressingly, waiting time for the damn things to be delivered.

 Also, of course, I can use this kit to measure the inductance of ferrite beads and home made inductors as well as test suspect components, including diodes and transistors. For DIY electronics such as this GPSDO project of mine, they're a surprisingly desirable investment which will likely recoup its capital cost in rather short order.
============================================================================================

 Harking back to a comment I made near the start of this missive, Since I've been running tests with the latest fake M8N module over the past two days, I've noticed much less wander in the position deviation map display than any of the other modules, including the original genuine module since I reduced the elevation filter setting from the previous "optimum" value of 10 deg (default having been 5 deg) down to just 1 deg. Surprisingly, this seems to reduce the range of wander to just around the 2 to 3 metre (radius) mark versus the more typical 5 to 7 metre wander.

 My externally mounted active mag mount patch antenna has a pretty good all round view of the horizon. It is still slightly obstructed to the East (uphill slope with ground clutter from neighbouring properties and some localised obstruction from the gable ended attic bedroom, the window of which gives me access to the roof of the bay window below upon which a ballasted metal draw with eight foot of antenna pole clamped onto it to clear as much of this obstruction as possible is sat).
 I'd like to raise it another 18 inches but until I can reroute the 5 metre cable to allow the SMA connection to a 5 metre extension cable to be kept in the dry, this desire has to remain on hold for the time being.

 I've come to the conclusion that when the GPS antenna has a good all round view of the horizon, it's better to reduce the elevation angle filter value to a minimum in spite of the perceived wisdumb that such low elevation signals are so degraded by ground reflections as to render them unreliable. I've actually found the reverse to be the case. It's not quite so obvious with the 5 degree default or even with the "Best Compromise 10 degree" setting but once you go for "Better Kwalitee" satellite signals using a 30 or 40 degree elevation filter, the deviations get worse not better.

 A 30 to 50 degree setting might help when the antenna is in a bit of a canyon where unwanted reflections can create chaos with the satellite signals (Timing is everything in these GNS systems) but this can only gain an improvement in an otherwise intolerable situation. If you've been able to site your GPS antenna to gain a good sky view, don't waste the extra satellite signals by needlessly filtering out those close to the horizon.

 Also, it's worth taking note of the fact that atmospheric bending (mostly an effect of the ionosphere I understand) means the actual signal paths will arrive a degree or two higher than the satellite's reported position data would lead the navigation engine in your GPS module to report (I've noted elevation angles of -1 and -2 degrees being reported just prior to the SVs disappearing off the plot).

 If you observe the way satellites hove into view as they come up over the horizon with signal strengths well above the minimum required to be decoded whilst still being reported as at zero degrees of elevation, the conclusion is that such signals are way better than the perceived "Wisdumb" in this matter implies. Just consider that at the moment the Sun's disk just contacts a true horizon at sunset, it has actually just completely dipped below the horizon as observed in the absence of any atmosphere. Anyway, that's my tuppence worth on the subject; make of it what you will.

 I've taken several photographs this afternoon and evening which may be of some interest, showing some U-centre screens as well a few of the modules mentioned above which I will attach to this post. From past experience with this blog's web forum posting UI, they may not appear in the submitted order so if the following descriptions look confused, check the filenames.

 The first, 9930, is worth a close look at R22's stats. The next three (possibly the first three from past experience) numbered 9942, 9943 and 9944 are just showing the state of play after about a six hour run. The fifth (or 4th), 9949, shows the last GPS module mounted on stand-offs over the OCXO module on the half built GPSDO PCB with the supplied active patch antenna parked on top of the OCXO simply to include it in the shot.

 The module is being powered from my desktop PC via the FTDI usb2ser adapter just above the GPS module with the backup battery (CR2032 cell in its holder) just to its right. The OCXO and the other ICs aren't yet connected to this temporary arrangement. The yellow wire coming from the CR2032 Cell holder is threaded through the bottom right TPH with a short length of silicone rubber insulation (also yellow) for protection and is soldered onto an SMD diode which is soldered to the 1K resistor hanging off a solder joint onto the end of the 1K smd resistor that connects to the Vbk pin (pin 22) of the M8N module itself.

 The original 80 milli Farad supercap represents a vampire loading of 1.2μA on my backup battery but since the draw by the module is some 14.7μA anyway, it's no big deal so I've resisted the urge to remove or disconnect the supercap - I could end up doing more harm than good.

 The diode leakage when powered up seems to be around 0.1μA in this circuit which eventually raised the coin cell voltage from this afternoon's start point of 3.128v up to 3.152v over an 18 hour period. However, it had reached this voltage several hours ago so appears to have reached its limit. ISTR reading that such cells in backup use could tolerate charging currents of half a μA indefinitely but after spending the past hour or more on a fruitless search of CR2032 datasheets, the best I could track down was this little nugget from a Maxell datasheet claiming the limit on such inadvertent charging current being 10mA ! 

 That's all a far cry from my recollection from the days when manufacturers used to offer more comprehensive data than the typical sales blurb sheets you see today! According to Maxell, my 1K resistor is a 5 times larger resistance value than needed. That was based on using a 5v supply - in this case it's a 3.3v supply so it might actually be a hundred times the resistance requirement based on the voltage difference calculation they used. It seems I can now take note of Dr Strangelove's philosophy and stop worrying about exploding CR2032 cells.

 The final image shows the three previous GPS modules I'd used. Displayed in the order left to right, is the original M8N module as per the subject line of this topic, then the next one is that bastard NEO-6M that was not fit for purpose, finishing with the very petite fake as a nine bob note M8N module, all prepped up to be stood on end into the main PCB, replete with a CR2032 coin cell holder and cell stuck to the top of the U-blox module on its backside with double sided tape and wired to the 80mF supercap via a 1K resistor and smd Shottky diode with the SMA socket removed from the PCB.

 On the LHS of the picture, the passive 1 inch square ceramic patch antenna is laid out alongside of its donor module whilst at the top of the picture, the smallest of cutdown ceramic patch antennas is shown next to its partner in crime. The most expensive of the M8N fakes, BTW, wasn't supplied with any such accessory.

 All in all, my quest to build a cheap DIY GPSDO has worked out a little more expensive than anticipated. Still, it's been quite a learning experience at relatively little cost despite all of that and the lessons still aren't over. I've yet to complete my little "Basic GPSDO" project and I've no doubt I'll still be learning more once I have it up and running.

JBG

[Johnny B Good]

 Hello All,

 After over a year since my last update of keeping this page tab open with no further activity other than minor corrections for spellings and typos, I've decided this topic has finally run its course. This posting will be my last by way of "Closure" for the benefit of anyone who chances upon what is now ancient history and is even just a little curious about the final outcome.

 I've moved on since showing my part built GPSDO displayed in the penultimate image. That MK I unit did finally get built a few months after but is now just a collection of recovered parts having re-spun that design into a MK II replacing the original 13MHz "Five Volt" CQE OCXO (which finally did prove to be a 12v part very forgiving of being severely undervolted) with a 10MHz CQE OCXO powered off the buck converter's 5.17v rail via a 5 to 12 volt boost converter which retains the MK I's ability to be powered from a 5W rated wallwart with voltage outputs in the extreme range of 6.8 to 24 volt DC.

 The direct use of a 10MHz sine output OCXO (buffered with a 74HC14 driving a 5 element LPF to recreate a +14dBm sine wave output) eliminates the 400mW's worth of energy consumed by a selected 'Old Skool TTL' 74193, a couple of NB3N502 clock multipliers and an unnecessary 7427 triple 3 input nor gate I thought might help by evening up the duty cycle of the 2MHz feeding the final NB3N502 from an 8:5 ratio to a 7:6 one (I don't think it really mattered to the 3N502 though).

 The result of shedding these four ICs from the MK I design was a warmed up energy consumption reduction from 1.8W down to a mere 1.4W in spite of the addition of a boost converter to supply the same amount of OCXO heater power at 12v rather than the 5.17v I'd used for the undervolted (I know now!) 13MHz OCXO.

 Not only did shedding these four ICs reduce energy consumption, it also eliminated a storm of noise on the Vcc rail which made it very difficult to detect the buck converter's 1.2MHz contribution (and to think, I'd been concerned about the noise and ripple (non) issue of directly using the output of a buck converter without the "benefit" of an LDO to "filter out" switching noise  ).

 Mind you, it does help considerably to choose a small (7805 sized) lightweight buck converter designed to power the control electronics of drones and other RC kit off a 3s to 5s LiPo pack where all the buck converter chip's manufacturer circuit board design rules have been followed to the letter (and beyond) to keep such switching noise to its absolute minimum. I used the 1.3A (max) rated Mini 360 synchronous rectified output 1.2MHz switched buck converters sold by Banggood (and no doubt elsewhere) for the MK I and then its (just ever so slightly) bigger 3A max rated 500KHz switched brother in the MK II.

 The other major change between the MKs I and II was the switch from veroboard (stripboard) construction to squashed or deadbug on plain copper clad board since my experience with the high levels of TTL noise currents on the Vcc rail had amply demonstrated the major weakness inherent to stripboard wiring.

 This also improved the effectiveness of the low pass filtering I used on the buck converter's output, rendering the use of an LDO pointless - above 100KHz they start looking like an HPF, making the problem worse, especially at switching speeds of 500KHz and above, along with the switching transients which extend hundreds of MHz beyond.

 With the IC count now reduced to just four, I elected to mount them conventionally by drilling chamfered holes to mount them onto the copper clad board and run point to point wiring on the underside (it looks a lot tidier and I'd much prefer to still be able to read the mounted IC part numbers). Of course, in the case of a larger IC count, I'd resort to squashed or dead bug style simply to avoid having to drill more than fifty holes in a copper clad board.

 The final icing on the cake being an upgrade of GPS receiver module from the M8N types I'd been using to an M8T. Whilst this doesn't eliminate the major component of the GPS system's deficiencies, mainly the ever varying propagation delay imposed by the ionosphere, a timing module, once "Surveyed in" (or set to timing mode using manually entered known to be accurate co-ordinates) will provide a more stable timing pulse output, reducing the +/- 20 or more nanosecond excursions, seen even with a well located antenna, down to +/-5ns or less.

 Although the MK I design had been inspired by Gyro's shown here:-

https://www.eevblog.com/forum/projects/my-u-blox-lea-6t-based-gpsdo-(very-scruffy-initial-breadboard-stage)/msg1493431/#msg1493431

 the result of this upgrade had morphed the design to one that now looks rather like a modern day re-spin of the James Miller GPSDO that had been reviewed here:- http://www.leapsecond.com/pages/gpsdo/

 The notable difference between the above unit and mine being the substitution of the Jupiter T with a modern day u-blox NEO-M8T, allowing the use of a 100kPPS and elimination of a 74HC390 dual divide by 10 ripple counter IC. The 74HC4046 I've used can be replaced with a 74HC86 quad XOR gate since, after some experimentation, the PC1 (XOR type) had proved to be my best choice (I'd only used a 74HC4046 in the first place to widen my choice of PLL PC options).

 If anyone's interested, I posted images of the hand drawn circuit sketches here:-

https://www.eevblog.com/forum/projects/ocxo-frequency-dividers-and-pll-multipliers/msg3146758/#msg3146758

 The MK I is shown first, followed by the MK II. I've not shown the buck converters and the boost converter used to supply the 10MHz OCXO in the MK II nor the details of the external voltmeter DC offset voltage chain that allows me to observe changes in the EFC in 100uV increments with a cheap 9999 counts DMM in the MK II and the TL431 based 3V offset  circuit used in the MK I for the same purpose.

==========================================================================================
[EDIT 2020-09-15] I almost forgot to include the FTDI232 serial to usb converter module in that list of parts not shown in those circuit diagram sketches.

 I was only reminded of this omission having finally tracked the culprit for an antenna configuration corruption issue that had been plaguing my MK II for the past several weeks, down to a problematic FTDI232 used to convert the 3.3v serial interface on the GPS module to a USB connection standard.

 This had led me on a merry dance which had finally culminated in my opening the MK II up just last night to temporarily hook up the old FTDI232 I'd recovered from the MK I to the M8T module in order to subject it to extensive power cycling stress tests using the worst possible scenario of not even disconnecting u-centre from the serial port connection before unplugging the usb cable (the FTDI232 is only powered via its connection to the PC USB port).

 Only then did I repeat this testing regime with another of my brand new FTDI232 modules (the recovered one's main chip was an anonymous unlabelled IC, hence my repeating the tests with the later modules which were all sporting FTDI labelled ICs). Only once this new module had also proved itself free of this corruption issue did I dare to taunt Messrs Murphy and Sod by removing the suspect module and fit the new tested module in its place and box it all up (but not until I'd repeated a few more test cycles before taking the final boxing up step).

 After some more power cycling tests of the fully assembled MK II, I decided to power it down overnight to reconfirm the presence of flash memory in my cheap but genuine NEO-M8T modules by virtue of it hopefully powering up the next morning with all user settings intact which, I'm pleased to say, it did and with flying colours - it locked onto the GPS satellites a mere 20 seconds from power up as opposed to the more usual 36 seconds typically seen with u-blox modules after such protracted powered down states. To say I was feeling a little pleased with myself would be a major understatement.

===========================================================================================

 You may note that (based on Gyro's methodology for calculating the PLL's LPF time constant) that the MK I used a 500s TC which I used initially in the MK II before going to a 2000s TC and then a 5000s TC before dialling it back to around the 1100s value I'm currently running it at.

 Unlike the failure to achieve a stable lock with a 100s TC in the MK I until I dialled that back to circa 38 seconds and then, with the aid of a biasing resistor, up to 100 and finally 500 seconds, I saw no such instabilities with any of the TCs used in the MK II, just a whopping 3 1/2 hour run up time with the 5000s TC which my add on accelerator circuit shortened to a mere 25 to 30 minutes (now a remarkably swift 10 to 15 minutes with the 1100s TC to get within a few mHz with another 20 minutes to get within a few tens of uHz.

 If you read the whole of Gyro's topic thread, you'll see mention of non-integer clock divide and sawtooth correction issues, all of which in the case of a simple hardware PLL based design mean diddly squat - they're a 'Red Herring' in this type of design since the necessary LPF  will average both these effects down below the noise floor, leaving all the other (and plentiful) GPS system deficiencies such as variable propagation delay due to the ionosphere and, about an order of magnitude less, due to the troposphere (precipitable water vapour content) plus orbital and clock drift errors and so on to "Spoil your day".

 Non-integer clock jitter and sawtooth corrections only become issues when you try to micro-manage them with, well appropriately enough, a micro-controller running DSP filtering algorithms with Kalman filtering thrown into the mix for good measure to put yourself into a "Catch 22" situation regarding reliance on the disciplined oscillator following a predictable trajectory by which to refine the Kalman filtering for periods extending to 24 hours and beyond to get close to the 10E-14 limit of accuracy offered by the atomic clocks used to drive the GPS master clock. Even the best quality double ovened OCXOs can't be guaranteed to always age gracefully.

 However, provided your DSP/Kalman filtering algorithms cover all possible contingencies, including OCXOs suddenly deciding to go off piste, then your micro-controller based design has a fighting chance of not ruining your day. If you want to try rolling your own microcontroller algorithms, a good way to avoid grief is start with something easy to discipline like a rubidium frequency standard (RFS). Once you've stabilised the base plate temperature and compensated for barometric pressure changes, you're essentially left with just a well behaved ageing effect to contend with.

 I think that covers everything I learnt from my experiences with GPS receiver modules kicked off by my purchase of the one mentioned in this topic thread subject line along with the art of using them to discipline an OCXO into submission as a reasonably accurate frequency standard. That being the case, as promised (barring any last minute replies), these will be my final words (offered to all and sundry as 'Food for Thought') in this topic thread.

JBG

P.S  As a final parting 'gift', I've attached a couple of pictures of the recently finished MK II, the front of which is identical to that of the MK I . The upper red LED is a simple power indicator. The green LED below shows GPS lock when solidly lit (100kPPS rate) otherwise it blinks at 4Hz 50% duty during acquisition to confirm that the user settings are still intact or, hopefully never, remain unlit until acquisition of GPS lock when it blinks at 1PPS (the default settings).

 In such a basic design devoid of anything more complex than an XOR phase detector PLL with a 1100s TC LPF for control of the OCXO, intended to run 24/7 week in week out for years at a time, there's no need of anything else to clutter the front panel with - all the 'exciting stuff' is on the rear panel.