What do you want from a GPSDO?

[texaspyro]

It all depends upon the GPSDO.  A lot of them have on-board switching supplies of who-knows-what quality.   The ones from reputable / knowledgeable makers usually use rather good ones.

[Electro Fan]

What would a typical spec be for ripple on the power supply for a reputable GPSDO?

On the order of 50 mV peak-to-peak, 15 mVrms, or something better?  What is the impact if the PS's ripple is higher?

[texaspyro]

What would a typical spec be for ripple on the power supply for a reputable GPSDO?

On the order of 50 mV peak-to-peak, 15 mVrms, or something better?  What is the impact if the PS's ripple is higher?

I dunno...  they are all different... check the GPSDO manual...  you do have proper manuals for all your undocumented GPSDOs?   

[jpb]

I was surprised by how much difference the type of supply I used had on ADEV measurements.

See this thread for a discussion on the subject:

https://www.eevblog.com/forum/metrology/rubidium-gpsdo-ocxo-use-of-switching-regulator/msg2291295/

[Electro Fan]

What would a typical spec be for ripple on the power supply for a reputable GPSDO?

On the order of 50 mV peak-to-peak, 15 mVrms, or something better?  What is the impact if the PS's ripple is higher?

I dunno...  they are all different... check the GPSDO manual...  you do have proper manuals for all your undocumented GPSDOs?   

Not sure what that was all about but fwiw I'm kind of new to GPSDO's.  I have a 2018-06-05 BG7TBL and the 1.1-D version of the manual, and I have a Star 4 that is on order that has not yet been delivered.  I've found some documentation on it from Oscilloquartz but I think that is at the oscillator and board level rather than the full GPSDO level.  Anyway, I appreciate all your good info and help on this forum.

[texaspyro]

Not sure what that was all about but fwiw I'm kind of new to GPSDO's.

Baically a jab at the lack of decent documentation on all but a very few GPSDOs...  most of the ones you find all are pulls from telecom equipment or rather amatureish devices.  The telecom stuff is usually pretty good, but were custom built for paying customers that want to keep their goodies and specs proprietary and secret.

[Johnny B Good]

 I've just posted a half marathon contribution to that other thread on the subject of smpsu issues, relating my experience with the FY6600 function generator by way of example (along with mention of my current distress at PLLing the 1KHz PPS of a cheap NEO-6M to a test setup using a 13MHz OCXO).

 The fact that the Oscilloquartz unit actually mentions ripple limits for its external smpsu requirement strongly suggests that they've taken account of the need for proper and effective supply rail filtering (and actually applied the appropriate measures). Other manufacturers that assume you use a perfect PSU, completely free of such ripple suggests they may have put no thought whatsoever into this aspect of their design (i.e. they've not tried hard enough to create a high quality product).

 When you think about this issue of PSU ripple, it's a sobering thought that this is something that's been on the minds of designers to varying degrees of concern for nigh on 80 years or more. It's not as though there's not been plenty of time for this issue to have been well and truly sorted out by now. Any shortfalls in this department are now largely down to pragmatic concerns over customer expectations and to what level of bean counteritus you allow your accountants to dictate to your design and production departments.

 Although manufacturing cost can't be allowed to run away unless you're making a very niche product, far too many manufacturers are taking this saving of pennies on product selling at hundreds and even thousands of pounds or dollars to a piss taking extreme which risks their credibility.

 Over the past 40 years or so, I've seen enough examples of rampant bean counteritus (and even one counter-example!) to convince me that most manufacturers of high tech electronic goods can be total and utter twats (I'm looking at you, Akai that once was in the business of high fidelity magnetic tape decks).

 If you can get hold of the service manual, complete with circuit diagrams, for your high tech electronic device (in this case, a GPSDO), you do at least have a chance of putting right the deficiencies of bean counteritus and sometimes even outright "skoolboy howlers" perpetrated in the design to turn your "Sow's Ear" into, if not a silk purse, then at at least a rayon purse facsimile of its intended function.

 Although more and more electronic products are hiding their inner workings from the gaze of the would be fettler by using proprietary designs within FPGA chips and the like, when it comes to sorting out PSU ripple issues, there are several approaches to solving such deficiencies of design including, if necessary, that of filtering out the ripple before it even reaches the device in question (either use of a better PSU or additional filtering between the existing PSU and the device's power rails). A quick 'n' dirty way to discover whether such an improvement to the PSU would be worth pursuing is simply to use a battery or batteries as a substitute for the original supply (removes the question of whether your hundreds of dollars worth of bench supply will introduce problems of its own or not).

 This all sounds like a lot of trouble but if you're motivated enough to be asking such a question, then it's not really that much trouble after all.

JBG

[Electro Fan]

A quick 'n' dirty way to discover whether such an improvement to the PSU would be worth pursuing is simply to use a battery or batteries as a substitute for the original supply (removes the question of whether your hundreds of dollars worth of bench supply will introduce problems of its own or not).

This all sounds like a lot of trouble but if you're motivated enough to be asking such a question, then it's not really that much trouble after all.

JBG

Ok, thanks for the good info and battery suggestion.  It does seem that an answer to determining how much impact external power supply ripple makes on the ADEV or other aspects of GPSDO operations is “try a battery.”  Of course, the results might be somewhat influenced by the internal/integrated power supply circuitry of the particular GPSDO design.  As previously noted, if for example, the internal board level design is spec’d at 150mV peak to peak ripple and noise, maybe a battery power supply won’t be able to show off it’s full impact.  But I think it’s worth a try.

Results of 1 baseline test with respect to some simple startup observations:  My BG7TBL seems to go from a cold start through a ~30 second period where it draws 1.2A and climbs to 1.3A; at about 30 seconds it starts drifting downward to less than .5A which it reaches in about 5 minutes and eventually (in about 60 minutes) it will settle at about 352mA.  I don’t know for sure which comes first, the Lock symbol on the display or 352mA but I think they occur about the same time.  At about 25 minutes the GPSDO reaches 360mA so the whole current descent process goes faster at the start and slower at the end.   By about 28 minutes the LCD shows 10,000,000.0000 for the first time, then the 10 MHz signal bounces up and down a bit and eventually settles out at or near 10,000,000.0000 (at least that’s what the display says – who knows?).

(I have developed a new theory that the GPSDO warm up cycle takes somewhat longer to reach 352mA and Lock when I’m measuring it with my handheld DMM than when I just observe it without the DMM measuring current.  Burden voltage?   I was pretty sure it only took about a half hour to reach Lock without the DMM inline.  With the DMM inline the last 8-10mA takes a relatively long time (~30 minutes?) before it reaches 352mA.  Update:  I now think the Lock occurs when the current settles to about 354mA, and then the GPSDO will bottom out at 352mA.) 

If you plug and replug the BG7TBL while it is warm it will come back to 352mA much faster (a few minutes?).  It consistently runs rock solid at 12.0V from startup to full warmup.  This is all using the seller supplied wall wart that is spec’d at 12V, 2.0A. 

So it would seem that with a 12V battery rated at 9Ah it might be possible to run a 12+ hour test.  Maybe not long enough to learn long term impacts but maybe enough to learn something about the impact of external power supply ripple (or the lack thereof in the case of a linear supply) on short term GPSDO performance.

As a side note, it’s funny how a mild time nut addiction can lead to purchasing a $200 GPSDO followed by a second GPSDO and maybe a better SMPS and maybe a 12V battery and then maybe lead to justifying why it would be useful to purchase a bench DMM capable of recording and displaying trends.  Stuff you previously didn’t know existed leads to the purchase of stuff you don’t really have a use for other than to observe and learn which leads to a need for more test equipment to measure stuff with more detail.  I guess it's a cycle.   

[Electro Fan]

While deciding on how to come up with a suitable battery I replaced the 12V, 2A wall wart with a linear PS set for 12V, 2A.  Below you can see the impact on the graph in LH.

Latitude is light blue
Longitude is orangish
Altitude is purple
DOP is maroon

[Electro Fan]

Other than less stair step effects on the lines in the LH graph I'm not sure what effects the linear supply has on GPSDO performance, but I can say that at a constant 12V the wall wart has less fan noise.

[jpb]

As a side note, it’s funny how a mild time nut addiction can lead to purchasing a $200 GPSDO followed by a second GPSDO and maybe a better SMPS and maybe a 12V battery and then maybe lead to justifying why it would be useful to purchase a bench DMM capable of recording and displaying trends.  Stuff you previously didn’t know existed leads to the purchase of stuff you don’t really have a use for other than to observe and learn which leads to a need for more test equipment to measure stuff with more detail.  I guess it's a cycle.

I justify similar expenses as being educational, though expensive for a hobby they are a lot less than say paying for a course at university. I also tell myself that I can sell off the kit when I'm finally cured.

[Johnny B Good]

A quick 'n' dirty way to discover whether such an improvement to the PSU would be worth pursuing is simply to use a battery or batteries as a substitute for the original supply (removes the question of whether your hundreds of dollars worth of bench supply will introduce problems of its own or not).

This all sounds like a lot of trouble but if you're motivated enough to be asking such a question, then it's not really that much trouble after all.

JBG

Ok, thanks for the good info and battery suggestion.  It does seem that an answer to determining how much impact external power supply ripple makes on the ADEV or other aspects of GPSDO operations is “try a battery.”  Of course, the results might be somewhat influenced by the internal/integrated power supply circuitry of the particular GPSDO design.  As previously noted, if for example, the internal board level design is spec’d at 150mV peak to peak ripple and noise, maybe a battery power supply won’t be able to show off it’s full impact.  But I think it’s worth a try.

Results of 1 baseline test with respect to some simple startup observations:  My BG7TBL seems to go from a cold start through a ~30 second period where it draws 1.2A and climbs to 1.3A; at about 30 seconds it starts drifting downward to less than .5A which it reaches in about 5 minutes and eventually (in about 60 minutes) it will settle at about 352mA.  I don’t know for sure which comes first, the Lock symbol on the display or 352mA but I think they occur about the same time.  At about 25 minutes the GPSDO reaches 360mA so the whole current descent process goes faster at the start and slower at the end.   By about 28 minutes the LCD shows 10,000,000.0000 for the first time, then the 10 MHz signal bounces up and down a bit and eventually settles out at or near 10,000,000.0000 (at least that’s what the display says – who knows?).

(I have developed a new theory that the GPSDO warm up cycle takes somewhat longer to reach 352mA and Lock when I’m measuring it with my handheld DMM than when I just observe it without the DMM measuring current.  Burden voltage?   I was pretty sure it only took about a half hour to reach Lock without the DMM inline.  With the DMM inline the last 8-10mA takes a relatively long time (~30 minutes?) before it reaches 352mA.  Update:  I now think the Lock occurs when the current settles to about 354mA, and then the GPSDO will bottom out at 352mA.) 

If you plug and replug the BG7TBL while it is warm it will come back to 352mA much faster (a few minutes?).  It consistently runs rock solid at 12.0V from startup to full warmup.  This is all using the seller supplied wall wart that is spec’d at 12V, 2.0A. 

So it would seem that with a 12V battery rated at 9Ah it might be possible to run a 12+ hour test.  Maybe not long enough to learn long term impacts but maybe enough to learn something about the impact of external power supply ripple (or the lack thereof in the case of a linear supply) on short term GPSDO performance.

As a side note, it’s funny how a mild time nut addiction can lead to purchasing a $200 GPSDO followed by a second GPSDO and maybe a better SMPS and maybe a 12V battery and then maybe lead to justifying why it would be useful to purchase a bench DMM capable of recording and displaying trends.  Stuff you previously didn’t know existed leads to the purchase of stuff you don’t really have a use for other than to observe and learn which leads to a need for more test equipment to measure stuff with more detail.  I guess it's a cycle.

 Wow! I can certainly relate to that final side note (as I'm sure most here would).    I believe it's something referred to as TEA addiction.

 For myself, it took the form of my purchasing a cheap Siglent DSO last November by way of returning from a 30 year sabbatical in my hobby interest in electronics enforced by marital and family responsibilities when the availability of modern test gear much cheaper and better specced than the Tektronix kit I couldn't afford second hand examples of three decades back when two or three hundred quid used to be 'real money', was pointed out in the SED news group just over six months ago, citing the FY6600 thread in this forum by way of example of just how much you could now get for your money if you set your sights below the more well known profit... manufacturers such as Tektronix and Keysight.

 Although this inspired me to buy a Siglent SDS1202X-E as a 'best bang for my buck' purchase (and that after doing without for thirty years, I really could now afford to treat myself to a piece of kit that could outdo the 2nd hand Tektronix offerings of three decades back for only slightly more than their then asking price), it took less than a week of ownership to realise that some sort of signal generator would now be a useful item to complement my 'scope purchase.

 Initially, I did briefly consider a cheap Siglent 2 channel 30MHz AWG until I started comparing it against the FY6600 at just a quarter of the price. Despite all the negatives against the FY6600 in an almost two years old thread, it still appealed to me, especially as most of the criticism was due to the readily fixable penny pinching induced shortcomings, most of which I've now fixed.

 One of the glaring deficiencies being the use of a "ten a penny" smd xo chip which would be of no detriment at audio frequencies but a severe annoyance for its lack of both accuracy and, even worse, stability for radio frequency work beyond the 100KHz mark, let alone, up in the Gods of the HF bands (3 to 30MHz).

 At the time when I read Arthur Dent's OCXO modification contribution, I thought he was going just a little over the top and so settled for just a 3 orders of magnitude improvement in the form of a neat little 50MHz 0.1ppm TCXO oscillator power board I'd initially planned on extracting the TCXO from to drop in place of the original and rather execrable smd XO chip until my IR thermometer revealed that it was running at 50 deg C due to being just 10mm away from the three LDO regulators which were showing a 70 deg C temperature on the main board. I landed up mounting the whole module above the cooling fan to act as a deflector plate to keep it within just two or three deg C above room ambient, connecting it to the vacated XO smd chip location via a short length of small bore co-ax.

 That gave me ample stability and accuracy to let me observe in sufficient detail what was happening with the NEO M8N module I'd bought as a starter item for my GPSDO project - I've since upgraded from the TCXO to a 10MHz OCXO with, as per Arthur Dent's OCXO upgrade, a 3N502 low jitter clock multiplier to generate the required 50MHz clock drive. That's now resulted in a free running OCXO calibrated within 300ppt with a drift rate over the past week or so of less than 200ppt (although I haven't been able to compare against a GPS reference for almost a week now). Suffice to say, it's been accurate and stable enough for me for use as a secondary reference in developing my DIY GPSDO project.

 I mention this on account this new found interest in test and measurement kit has set me off on a seemingly never ending quest for better and more accurate gear, aka a GPSDO I can rely on and be proud of. Lately, in earlier postings here, I've been bemoaning my problems with getting a 5 volt 13MHz OCXO to phase lock against my cheap NEO-6M's 1kHz PPS (it's a make do for now substitute for the original M8N module I damaged with an injection of 12 volts into its PPS line a couple of weeks back).

 All this time I'd been blaming first the XOR gate phase detector circuit and then the 4046 phase detectors for not doing 'their job' and, only now between this and the previous post, discovered a rather strange fact about the PPS signal from a NEO-6M (or at least this NEO-6M) when it's programmed to output a 1ms second pulse with a 50% duty cycle (1KHz sq wave).

 Up until just prior to this post, I'd only ever quickly checked out the PPS and divider chain  1KHz output waveforms, preferring to monitor the 13MHz OCXO and FY6600 outputs when testing my crude PLL setup. It's only since I decided to take a much closer look in detail at the 1KHz signals going into the 4046's phase detectors, whilst monitoring the output voltage, that I've finally discovered the root cause of my crudely lashed up on prototyping bread board GPSDO test setup's failure to lock onto the GPS reference. It turns out to have been all to do with the NEO-6M's strange habit of slowly  wittling away at the positive going pulse, shrinking its width until it disappears up its own backside to instantly reincarnate itself as a full fat sq wave, remaining so for the following five or so minutes before starting another cycle of shrinkage over the next 5 or so minutes.

 If it hadn't been for the fact that I'd upgraded the FY6600 to an OCXO clock, I doubt I'd have been able to hand adjust the tuning pot on the OCXO to keep it close enough to the leading edge of the 1KHz PPS pulse to observe such weirdness (I'm not sure the original 0.1ppm 50MHz TCXO upgrade would have been enough of an improvement to have pulled off such a stunt).

 It's no wonder the XOR gate failed to produce any meaningful failure behaviour let alone actually function as intended other than for relatively brief moments when it seemed to be locking in an offset fashion. The type 2 phase detector in the 4046 should have gotten round the non-square waveform issue but I'm guessing it was triggering on the trailing falling edge which would have been slowly happening at earlier and earlier times in the cycle for half of this 10 or so minute cycle before jumping back to a sq waveform for another 5 or so minute round of fun and games. No wonder it was slowly driving me round the bend these past few evenings (including midnight to 5 or 6 in the am if truth be known  ).

 Anyway, that's just a taster of the sort of process that sucks you ever deeper into the mad mad world of DIY test gear and electronic projects in general. BTW, I think, now I've got the use of the type 2 phase detector in that 4046 which doesn't rely on the waves being square, I'll have a go at reprogramming the NEO-6M to generate the narrowest pulse possible (10 microsecond, afaicr) and see how it behaves at that setting.

 Obviously, I could try an inverting gate on the PPS line but I don't have any room on my prototyping breadboard for another TTL chip. Paraphrasing the famous line out of the movie, "Jaws", I think I'm going to need a bigger board. I think I might have one in the attic somewhere so I might have a look for it later on.

 I've just taken a look and found it straight away! It's actually a "Microprocessor Trainer model ET-3400" made by Heathkit and it's complete with original manual. Taking another look at it now, I see the only item of value to me is the older style breadboard block that's simply stuck down onto the main circuit board to allow additional analogue circuits to be assembled (the 50/60Hz 120/240v transformer supply, according to the circuit diagram has two centre tapped secondary windings, one biphase rectified to produce circa 8vdc for the 5v regulator IC and the other with bridge rectifier to, presumably, provide a +/- 18 to 20v dc for the 12v + and - regulators which, according to the manual, are only for use with the prototyping breadboard that had been stuck, centre stage on the mainboard/panel.

 They'd designed this 'Trainer' around a Motorola MC6800 cpu so it's no wonder I never even bothered to fit a plug onto its captive mains lead. If they'd more ambitiously based it around a Zilog Z80 or even its kakamaimee cousin, the Intel 8080, I might have deigned to fit the mains plug and have a play with it. With the association of the words "trainer" and "potty" in my mind and the fact that I was probably already writing Z80 assembler code to run on my S100 bus desktop machine at the time, it's no wonder it spent 30 years or so gathering dust in the attic.

 I think it was a novelty item even back when I picked it up for cheap (or possibly even as a freeby) at some hamfest in the early to mid 80s. I was probably attracted to the prototyping breadboard (and possibly the plugged in TTL ICs) more than anything else. I suppose the case might be repurposed to stick a bunch of prototyping breadboards down on a replacement panel to the original mainboard and the transformer supplies replaced with more modern smpsu modules extracted out of redundant wallwarts and so on but that seems it might consume too much of my time right now and it may well have some value as a "Collectable" for all I know (it seems to be pretty well complete, only lacking its "Original Packaging"(tm)  ). I might prise the breadboard off for my immediate needs since it's an easy enough 'restoration job' should it prove to have any value as a 'collectable'.

 As it happened, I did use the right tool for the job, a flat bladed screwdriver, just not the right way since it proved to be screwed down to the panel rather than, as I'd assumed, stuck down with double sided adhesive tape. Still, it did save me having to remove a whole bunch of screws just  to separate the top and bottom halves of the case to gain the required access to use the screwdriver as Ghod had intended (there was enough flex in the ventslots to grab hold of the loose screws with long nose pliers and with careful jiggling to extract them for reuse to attach the breadboard onto a plastic baseplate to prevent the contact strips being pushed out the underside.

 I'm pretty certain it's a twin to another prototyping breadboard I already had but which I had managed to misplace at some time during the past 35 years or so. Making up a baseplate for my liberated breadboard will be worth the modicum of time and effort required since I can use it as an annexe to the existing board to support the luxuries of extra logic gates, hex Schmitt trigger inverters, CMOS RRO opamps and such like. 

 I've just taken the time to time the pulse width cycling on the PPS line with a stopwatch and it cycled in 11 minutes and 12 seconds during which time it would remain at a 50% duty cycle for a few seconds longer than 5 minutes, just as I suspected was the case. I think the addition of a hex inverter gate in the PPS line should fix the phase locking issue that's been torturing me these past three days or so.

 I monitored the output voltage on the output of the LPF connected to the PD2 output line and it has been cycling from zero to 2.37 volts, shadowing the phase difference between the trailing edges of the OCXO divider and the PPS outputs. It would seem I have finally solved the mystery of this phase locking failure. As to what the best cure is, I've yet to investigate my options. An inverter should get me at least 99% of the way to a fix (possibly even 99.9% or better). I shall see.

JBG

[Johnny B Good]

A quick 'n' dirty way to discover whether such an improvement to the PSU would be worth pursuing is simply to use a battery or batteries as a substitute for the original supply (removes the question of whether your hundreds of dollars worth of bench supply will introduce problems of its own or not).

This all sounds like a lot of trouble but if you're motivated enough to be asking such a question, then it's not really that much trouble after all.

JBG

Ok, thanks for the good info and battery suggestion.  It does seem that an answer to determining how much impact external power supply ripple makes on the ADEV or other aspects of GPSDO operations is “try a battery.”  Of course, the results might be somewhat influenced by the internal/integrated power supply circuitry of the particular GPSDO design.  As previously noted, if for example, the internal board level design is spec’d at 150mV peak to peak ripple and noise, maybe a battery power supply won’t be able to show off it’s full impact.  But I think it’s worth a try.

Results of 1 baseline test with respect to some simple startup observations:  My BG7TBL seems to go from a cold start through a ~30 second period where it draws 1.2A and climbs to 1.3A; at about 30 seconds it starts drifting downward to less than .5A which it reaches in about 5 minutes and eventually (in about 60 minutes) it will settle at about 352mA.  I don’t know for sure which comes first, the Lock symbol on the display or 352mA but I think they occur about the same time.  At about 25 minutes the GPSDO reaches 360mA so the whole current descent process goes faster at the start and slower at the end.   By about 28 minutes the LCD shows 10,000,000.0000 for the first time, then the 10 MHz signal bounces up and down a bit and eventually settles out at or near 10,000,000.0000 (at least that’s what the display says – who knows?).

(I have developed a new theory that the GPSDO warm up cycle takes somewhat longer to reach 352mA and Lock when I’m measuring it with my handheld DMM than when I just observe it without the DMM measuring current.  Burden voltage?   I was pretty sure it only took about a half hour to reach Lock without the DMM inline.  With the DMM inline the last 8-10mA takes a relatively long time (~30 minutes?) before it reaches 352mA.  Update:  I now think the Lock occurs when the current settles to about 354mA, and then the GPSDO will bottom out at 352mA.)

 The question of "Burden voltage" might have some credence (ITYMTS, burden load from the DMM's 10 or 11 Mohm voltage sensing circuit impedance) since I noticed a definite detuning effect on the PLL when disconnecting the meter from the output of the LPF feeding the OCXO's tuning pin after it had finally established a rather fragile lock. Obviously, a better measuring point would have been on the output of a buffer amp between the LPF and the tuning control pin but, since I hadn't been able to add such a luxury item to my breadboard lashup for the want of space, that simply hadn't been an option. 

If you plug and replug the BG7TBL while it is warm it will come back to 352mA much faster (a few minutes?).  It consistently runs rock solid at 12.0V from startup to full warmup.  This is all using the seller supplied wall wart that is spec’d at 12V, 2.0A. 

So it would seem that with a 12V battery rated at 9Ah it might be possible to run a 12+ hour test.  Maybe not long enough to learn long term impacts but maybe enough to learn something about the impact of external power supply ripple (or the lack thereof in the case of a linear supply) on short term GPSDO performance.

 An alternative to that scheme, if you feel the need for protracted days long runs, would be to use a pair of 7AH alarm batteries (the cheapest SLA battery option if you can have a friendly chat with your local alarm shop's proprietor  ) and connect them via Schotky barrier rectifier diodes so you can attach a freshly charged alternate when the first starts to show signs of exhaustion, allowing a breakless changeover to put the changed out battery on charge. Thinking about it, you might want to swap the batteries between load and charger on a shorter cycle to avoid going beyond a 50% DoD and the risk of premature wear out.

 The "Try using batteries" suggestion works fine when, for example, testing to see whether weird noise pollution on a signal generator's output is a power supply artefact or not when such testing can be over and done with in less than half an hour but the cost of sufficient battery capacity to power a GPSDO for several day's worth of testing makes the building of an over specified high grade LPF to put between the cheap wallwart and your precious GPSDO rather more appealing.

 The big appeal of using a battery is the complete elimination of switching noise which trumps the obvious downsides of slowly drooping voltage during discharge (which usually isn't an issue in most cases) and the inconvenience of having to swap out or recharge exhausted battery packs.

As a side note, it’s funny how a mild time nut addiction can lead to purchasing a $200 GPSDO followed by a second GPSDO and maybe a better SMPS and maybe a 12V battery and then maybe lead to justifying why it would be useful to purchase a bench DMM capable of recording and displaying trends.  Stuff you previously didn’t know existed leads to the purchase of stuff you don’t really have a use for other than to observe and learn which leads to a need for more test equipment to measure stuff with more detail.  I guess it's a cycle.

 I've already said more than enough on that subject.

 The latest report on my own travails with trying to get a 13MHz OCXO to lock to the "One KiloHertz PPS" signal from a cheap (and, as it turned out, NASTY) NEO-6M is that I did manage to get it to lock for a good ten to fifteen minutes yesterday before it suddenly threw a wobbly (the 1 microsecond wide pulse disappearing up its own backside to reincarnate itself as another cycle of 1 microsecond wide pulses at a 1KHz repetition rate. It seems that even setting a 1 microsecond pulse width doesn't stop it being wittled right down to nothing over the final half second or so of the 10 to 15 minute cycle (it varies - presumably a side effect of the sawtooth corrections to keep the original one pulse per second's leading edge to within half a cycle of the free running on board 48MHz XO synchronised to GPS time).

 I tried various concoctions of inverting buffer configurations on the PPS and divider chain output lines but since the process of getting the 1KHz pulse edges lined up takes ten thousand times longer than it would at 10MHz, it's not so much a case of waiting for paint to dry so much as waiting for it to peel.  In the end, I just left it running to see what tricks it was going to get up to whilst I dealt with other stuff.

 By then, last night, I was rather resigned to waiting for my cheap NEO M8N to eventually arrive any time from the middle of next week onwards so I was rather pleasantly surprised when it, along with an order for ten 74HC390s, turned up in this morning's post.

 I set the new M8N up with a USB to serial adapter and hooked it up to the winXP VM to check it out with u-blox's u-centre application. I tried programming the TimePulse but that doesn't apply with the M8N as I had a deep suspicion would be the case.

 There's a shed load of programming options which seems to be a superset of all the options that have ever applied to the various GPS receivers the u-centre app can control. Unfortunately, this means you can never be certain which options work and which ones don't because the particular receiver you're using simply doesn't happen to support that function.

 This wouldn't be so bad if the configuration options not supported by your specific u-blox module were greyed out or you were offered some other means to indicate the futility of programming an unsupported feature. Anyway, with my suspicions confirmed over the slapdash programming effort that went into the u-centre UI, I turned to the TP5 item which I knew from experience was the correct menu option. Although I can program any frequency from 1 to 24,000,000Hz in 1 Hz steps, I decided to program a 1KHz 50% duty cycle pulse so I could both carry on testing my existing PLL setup and also look out for any of the symptoms of a shrinking pulse that had plagued the NEO-6M.

 After waiting for 15 minutes to confirm that, even after disturbing the divider chain to induce a better alignment of the pulses, the PLL was behaving in a rational manner (tuning voltage glacially dropping from 4.6v as the divider chain pulse edge approached the balance point, around a phase lag of 125 deg, I left it for an hour or two to its own devices whilst me and the missus nipped out for a bit of shopping, returning to find it had finally settled at said balance point with a tuning voltage of 3.22v verifying it had finally phase locked the OCXO.

 It seems stable in that any disturbance, which detunes the OCXO, once removed, causes the PLL to glacially restore the balance rather than, in the case of the NEO-6M, send it tumbling out of control (LPF voltage nose diving or climbing to the Vcc limit). Basically, I've just proved that all my efforts at getting an OCXO of any sort to phase lock to that NEO-6M's PPS had been doomed from the outset - who knew? Seriously, did anyone know?

 Anyway, "Honour being satisfied" as it were, I plan on programming the PPS to output a 100KHz 50% duty cycle pulse by which to phase lock one of the 10MHz OCXOs (although I may still ultimately decide to utilise the 13MHz one with its convenience of a 5 volt power requirement).

 Several points of interest I noticed when testing the replacement M8N module were that the SV signals were down by around 18dB or so on the original M8N module but, by eliminating the 5m antenna extension cable, I saw the expected 6 to 7dB boost, unlike the inexplicably low 1 to 2 dB loss it had displayed with the original module. I'd surmised at the time, that the incredibly low loss from 5 metres of RG174 at 1.6GHz was more to do with a fortuitous impedance mismatch at the RF input of the original whilst now, I'm possibly just witnessing the effect of a proper impedance match (or possibly a Lo-Z mismatch because a 270 ohm antenna feed resistor had been substituted in error by a 27 ohm resistor [Edit:- Belay that! a more likely possibility would be a shorted out inductor] - I'll take a closer look later, leave no stone unturned and all that - you never know what you might discover if you don't take a very close look).

 The other difference being that despite configuring to receive the SBAS geo-synchronous SVs over 6 hours ago, I have yet to see any signs of one. As for the Galileo constellation, I'd already given them up as a lost cause since, apart from a fleeting appearance during the first month since I acquired an external GPS antenna after buying the first module, there's been no further signs, presumably due to a protocol change in the Galileo system requiring a firmware update to get the M8N to recognise them again.

 The damaged M8N was still showing the SBAS SVs whenever I programmed the module to display their presence until quite recently so that's another test to run. It's possible I might have a 'fake' M8N but whatever its shortcomings, it still appears to be just as useful as the original module for making up a GPSDO (it's kept the OCXO locked to the 1KHz square wave output from its PPS line for the best part of the past 6 hours.  ).

 Now that I have the option once again, I'm going to reconfigure the PLL for 100KHz operation and shift my life into the 'fast lane' (FVSVO "fast"). I'll report any further progress once I've made some.

JBG

[Electro Fan]

"An alternative to that scheme, if you feel the need for protracted days long runs, would be to use a pair of 7AH alarm batteries (the cheapest SLA battery option if you can have a friendly chat with your local alarm shop's proprietor  ) and connect them via Schotky barrier rectifier diodes so you can attach a freshly charged alternate when the first starts to show signs of exhaustion, allowing a breakless changeover to put the changed out battery on charge. Thinking about it, you might want to swap the batteries between load and charger on a shorter cycle to avoid going beyond a 50% DoD and the risk of premature wear out."

That's a pretty neat idea.

[Johnny B Good]

"An alternative to that scheme, if you feel the need for protracted days long runs, would be to use a pair of 7AH alarm batteries (the cheapest SLA battery option if you can have a friendly chat with your local alarm shop's proprietor  ) and connect them via Schotky barrier rectifier diodes so you can attach a freshly charged alternate when the first starts to show signs of exhaustion, allowing a breakless changeover to put the changed out battery on charge. Thinking about it, you might want to swap the batteries between load and charger on a shorter cycle to avoid going beyond a 50% DoD and the risk of premature wear out."

That's a pretty neat idea.

 Thank you, it's not particularly original but at least it gives you a means of checking whether the switching ripple noise from a cheap wallwart is going to compromise the GPSDO's performance.

 The 7AH rating is based on the 20 hour discharge at 350mA which, in my case with those CQE OCXOs settling down to around the 80 to 90mA mark and my simple GPSDO (basically Gyro's design here: https://www.eevblog.com/forum/projects/my-u-blox-lea-6t-based-gpsdo-(very-scruffy-initial-breadboard-stage)/msg1493431/#msg1493431) taking maybe only another 100mA all in (fake NEO M8N - no flash[1]  and an FTD232 usb adapter, just one '390 and a '86 for the PD, not forgetting the obligatory 74HC14), a pair of those batteries would only reach a DoD of 34% on a 12 hour cycle turnaround so in my case, it would be entirely do-able.

 If your GPSDO draws rather more than the 200mA estimate I made for mine you might need to swap the batteries around on an eight hour cycle if you don't want to wear them out prematurely. With a nod to Peukert's law, a 400mA draw will get you to a 50% DoD in just about 8 hours. Any higher than that, to misquote "Jaws", and you're gonna need a bigger battery.

 Mind you, if you can prove the point in a 16 hour session, you can get away with a shorter turn around cycle if you have a suitable constant voltage charger to keep pace with the higher discharge and recharge cycles involved. Alternatively, if you can borrow the use of a car starting battery, you could save on that investment in a couple of cheap alarm batteries. After all, you're looking at the most pragmatic way to take the questionable quality of a 'cheap wallwart' out of the equation.

 Getting back to the latest development with my own little GPSDO project, after swapping from the 5v 13MHz 3Vpp sq wave out OCXO to one of the 12v 4Vpp sine wave output ones by using my liberated prototyping breadboard as a 12v annexe to mount it alongside of the original rat's (Vipers?) nest wiring lashup last night, I decided to live dangerously and rebuild it all onto the one breadboard.

 Before I could rebuild a now much simplified configuration (two less ICs to find room for), I had to figure out a safe way to provided a +12v feed to the OCXO. I soldered a short 2 wire female header plugged flylead (the sort used to connect a PC's front panel soft on/off switch to the MoBo header pins) onto the switched 6.5 to 12 volt feed from the wallwart socket of the YwRobot MB102 breadboard power supply using hot melt glue to fix it in place and provide strain relief. The ends of the jumper wires fit quite nicely into these female header pin connectors making it a neat way to safely provide a 9 or 12 volt feed directly from whatever wallwart I choose between the 9 or 12 volt ones I've kept aside for powering my breadboard concoctions.

 Whilst I'd not had any stability issues with the 13MHz OCXO, I couldn't say the same for the 12v 10MHz units. I found myself having to hold onto then slowly release the wire taking the oscillator output to the 74HC14 I was using as a buffer between the 74HCT390 dual divide by ten IC to get a rather fragile lock. In the end, it occurred to me that maybe I just needed to bias the HC14's input to the mid point since, unlike the 13MHz unit which outputs a DC coupled 3Vpp square wave, the 10MHz outputs a capacitively coupled 4Vpp sine wave.

 Once I'd managed to retrieve a couple of 2K2 resistors from my 'salvage' collection and fitted them, all was sweetness and light once more (although it did help improve phase locking performance a tad when I remembered to replace the earthing link wire to reconnect the ground side of the LPF  ).

 Anyway, the new setup with the 10MHz OCXOs now demonstrates the same remarkably stable behaviour that, considering the rat's nest breadboard wiring involved, I'd seen with the 13MHz unit. It even works just fine when using the 9v smpsu wallwart instead of the 12v one (it just takes a little longer to warm up, maybe five minutes versus three on the 12v supply). The cheap 'n' cheerful wallwarts don't seem to be cause for concern right now as far as I can tell. That, of course, may well all change once I've had time to do more thorough testing on the finalised boxed up version. 

 TBH, apart from the effect on the warm up time, I don't think the 12 volt supply to these OCXOs is particularly critical. They all include a well filtered low noise LDO regulator to power the really critical oscillator and oven controller circuit  which regulates the heating current drawn from the 12v rail in the power transistor that acts as the heating element. Its collector voltage doesn't matter so much as long as there's sufficient voltage to allow the maximum current limited flow to generate enough heat.

 I've had them get up to temperature when I was testing the first batch of three off a 5.15v supply. It took a good eight minutes or so and the output was a little on the weak side although at the correct frequency. Since all the indications were that I wasn't supplying enough voltage (that, and the fact that I now had another two spares), I went straight to a 9 volt supply (a five volt part would have been damaged almost straight away at another two volts beyond the 7 volt absolute maximum limit typically quoted for five volt parts) and then to the full twelve volts once my test unit had managed to survive this potential voltage abuse for an hour or so.

 I was treading very carefully since the vendor had absolutely no data on these OCXOs and my first 13MHz example had proved to be a 5 volt part once I'd realised that my 5 volt supply was being dragged down to less than 4.75v due to the warm up demand, just below the +/-5% tolerance limit I'd have expect to have been applied in this case. As soon as I fixed the low 5 volts issue, it would then spring immediately into life unlike before when it had taken several minutes before even any sign of activity became apparent on the sub-standard supply voltage which had eventually reached the dizzy heights of 4.82 volts.

 Whilst I'd expect the supply voltage to be of super-critical importance with a 5v OCXO, there's a lot more leaway in this regard for the 12v units as my current lashup is so clearly demonstrating with an 8.98v supply as I type this little missive.  You could probably get away with using a low noise 10v LDO to clean up the ripple from a cheap 12v wallwart and save the expense of an ultra low ripple 12v supply which savings can be used to stock a couple of 'cheap spares' against the risk of your one and only suitable PSU expiring unexpectedly.

 I was observing the effects of positional deviations on the phase between the signal generator's "10MHz" (I have to tune down by some 7 or 8mHz to get a 'stationary' display of the two waveforms) which can turn any calibration attempts into a game of "Chase Wil o' the Whisp", if you overlook this phenomena. This is what must drive the need for extremely long time constants and the desire to get hold of Timing GPS Modules. TBH, since I was expecting to see this effect, I wasn't at all surprised and it rather confirms that my basic GPSDO test rig is functioning as intended, considering the relatively short time constant being used in the PD's output filter.

 I'm actually using 22 and 220μF aluminium electrolytics rather than the  33 and 330μF tantalum caps shown in Gyro's circuit. I'm thinking of substituting these with a 47 and a 470μF cap by way of an experiment. TBH, I'm surprised at how well this filter is working considering the issue of leakage in normal electrolytics. I'd normally only consider the use of non-electrolytic types in filters but needs must and electrolytics were the only type in these values I had to hand.

 Unlike Gyro's circuit, I've missed out the cmos rro opamp buffer and taken the filter's output directly to the OCXO's control pin. I might strap one of the ten 5v cmos rro opamps, that had finally arrived nearly three weeks ago, onto a SOT adapter board so I can test out a higher impedance version of the LPF (1MR with a 2.2μF and a 22μF with 100KR series resistor combination). TBH, I'm surprised Gyro didn't take advantage of the buffer amp himself to reduce the caps down to a more manageable size. Since he posted that circuit diagram over a year ago with no further follow ups in that thread, it's a little late in the day to query him on that. I might get an answer to that when I come to add the buffer amp and rescale the filter values.

 Anyway, that's the latest situation regarding what I want from a GPSDO which is simply a basic functioning frequency reference good enough to feed the external reference inputs of test gear and HF transceivers and such like (I'm not planning on working with GHz rated kit - at least not just yet). It makes a pleasant change for once, to be be able at long last, report some real progress. 

[Notes]

[1] What this means is that I'll have to add a CR2032 coin cell holder to get round the limitations of the tiny 80 milli-Farad super-cap that provides less than an hour's worth of data backup (about an hour when only used to maintain the RTC which is the primary purpose of the BBRam). This should give me at least 250 days of data retention (220mAH and a 30μA drain current).

 Considering the fault with the NEO-6M's shrinking PPS, that did at least have the option to save the config data into either the SPI-Flash or the I2C-EEPROM as well as the Flash and the BBRam. I found that if I selected all four options, the settings would remain intact even overnight. I'm guessing it was either the SPI-Flash or the I2C-EEPROM which was saving the data but ICBA trying to figure which it was. I simply selected all four options whenever I wanted tp make sure the settings didn't evaporate overnight. Unfortunately, the same trick doesn't work with this fake M8N module, hence the additional complication of wiring up a lithium coin cell holder to complete the project. 

JBG

[jimmc]

An alternative way of long term saving and setting the configuration is to program a small MCU to set-up the module via the serial port.
This is only needed at start-up and the MCU is then free to be used as a divider (if clocked from the OCXO).
 
I did this with a NEO-7M and an ATtiny85 to give a 10kHz comparison frequency, the ATtiny85 also turned on the NMEA output only when an external USB-RS232 converter was powered.
This was done to reduce the possibility of ripple on the power rail at the1Hz rate of the data bursts when monitoring not required. (Simple lock LED used in normal operation.)

As an aside a second ATtiny85 (again clocked from the OCXO @10MHz) was programmed to give an output frequency, switch selectable,  from 5MHz, 2.5MHz, 1MHz down to 1Hz (with 1:1 M/S ratio).

Jim

[Johnny B Good]

An alternative way of long term saving and setting the configuration is to program a small MCU to set-up the module via the serial port.
This is only needed at start-up and the MCU is then free to be used as a divider (if clocked from the OCXO).
 
I did this with a NEO-7M and an ATtiny85 to give a 10kHz comparison frequency, the ATtiny85 also turned on the NMEA output only when an external USB-RS232 converter was powered.
This was done to reduce the possibility of ripple on the power rail at the1Hz rate of the data bursts when monitoring not required. (Simple lock LED used in normal operation.)

As an aside a second ATtiny85 (again clocked from the OCXO @10MHz) was programmed to give an output frequency, switch selectable,  from 5MHz, 2.5MHz, 1MHz down to 1Hz (with 1:1 M/S ratio).

Jim

 Hi Jim,

 Thanks for the suggestion. I did consider making use of an Arduino nano I'd bought at the beginning of February just after I'd bought the first (genuine but now broken) NEO M8N module I'd purchased right at the end of January, but adding a CR2032 seemed the more pragmatic approach in view of my lack of Arduino programming expertise. I'll be incorporating the FD232 USB to serial adapter that's currently on the prototyping board lashup to let me monitor and configure the fake M8N module so even if I allow the CR2032 to go flat after some 250 to 300 days worth of not being powered up, it'll only be a minor inconvenience to replace the coin cell and reconfigure it.

 I've not even bothered unsealing the anti-static bag the Arduino nano had arrived in although I do intend to try my hand at programming it in the not too distant future. The last time I did anything like this was 35 years ago writing the control routines in Z80 assembler to adapt a Philips solenoid controlled data cassette drive as a floppy disk substitute (332KB formatted storage each side of a C60 with a worst case end to end search time of 14 seconds and 19200 bps read/write speed).

 This had allowed me to load the 8K TCL BASIC, which had been originally supplied on a set of four 2KB eproms which I'd relocated into ram space, in just 5 seconds flat - faster than that crappy waferdrive system of Clive's at a time when 3.5 inch SD floppy disk drives had just dropped in price to a mere hundred quid a pop (and S100bus FDC cards also happened to be around the 100 quid plus VAT mark).

 I'd had plenty of financial incentive at the time to put these surplus decks into use as a poor man's alternative to floppy disks - all the convenience of a floppy disk drive system but at a slower pace. Of course, this all changed over the next five years as the IBM PC created a mass market to drive competition which drove prices right down to mere 'chump change'.

 I've still got that old S100 bus kit but the original main board with its dozens of socketed chips appears to have succumbed to the damp basement conditions it had been kept in for some 30 years or so and I'm in the middle of swapping to a spare (but unmodified) board which had been stored in a much warmer and drier part of the basement.

 However, I've been in the middle of this swap out job for over a year now since I couldn't maintain my enthusiasm to bring it back to life. Still, at least it's now parked in a more benign environment for the time being so I may yet finish the repair project if the urge to get it going again ever returns.

 Anyhow, my GPSDO setup seems to be working just fine on the breadboard setup. TBH, it's a wonder it worked at all when I realised I'd actually forgotten not just one earthing strap wire in the LPF circuit but two! Even more remarkable was that the 74HC14 was able to perform at all after I'd misplaced the ground wire onto the adjacent output pin when I'd fitted a 100nF disk ceramic cap over the IC to improve the supply decoupling the day before when I'd been trying to get to the bottom of the weird instability on the oscillator drive signal which had proved to be the lack of dc biassing resistors on the input of the gate I'd selected to buffer the OCXO's ac coupled output.

 I wouldn't have spotted this Obe Won Kenobe error if I hadn't been looking to use one of the spare inverters to buffer the oscillator feed to the FY6600 counter's input/trigger sockets just to take the opportunity to take pictures of it showing an exact reading of 10,000,000.00Hz on its display at the 100 seconds gate time setting (maximum digits).

 The frequency counter function built into the FY6600 is not something I've ever bothered with before other than for a few brief tests to verify that it actually worked. It seemed a good opportunity, now that the internal reference is comfortably within 1ppb of exactitude, to test just how it would display a frequency more accurate than its resolution by at least one more order of magnitude.

 Funnily enough, it would display the correct frequency about half the time and under-read it by 0.01Hz the other half whilst the 1 and 10 second gate times resulted in either an exact reading about two thirds of the time or else show a "1" in the final digit position for the other third.

 Considering that I've seen much higher specced counters displaying even larger discrepancies on the least significant digit(s) when hooked up to a 10MHz GPSDO reference in various youtube vids, this doesn't seem so bad, especially considering the price difference involved.

 Still, I had rather hoped to be bored by an unremitting display of all zereos after the Leading "1". I guess it's all down to how well the gating circuitry has been implemented. Incidentally, I saw the same results using either the front panel input socket (AC coupled) or the rear panel 'trigger' input socket (DC coupled).

 I'd like to recalibrate the signal generator's OCXO reference to better than the +600ppt it  currenly seems to be set at but the apparent short lived frequency drifts that arise out of the phase shifts on the GPS's PPS line, due to the wandering in the calculated position data as displayed in the deviation map, lends an element of confusion to the calibration process. I'll resist the urge to fiddle right now until I have a better setup in regard of a longer time constant filter to minimise this effect (and possibly a relocation of the GPS antenna to get a less obstructed view of the whole sky - it's currently blocked on a large part of the eastern sky).

 The fact I can now confirm my suspicions over the effect of a wandering fix as per the deviation map display is a sign that I've made some useful progress in my basic DIY GPSDO project. Now all I have to do is rebuild what I've got onto a PCB and fit it into a case. I haven't entirely nailed down the details (I'll add one of those 5v cmos rro opamps so I can make a longer time constant filter with higher resistor values rather than use larger value caps) but I think it's more or less there as a cheap DIY GPSDO. I've got another half dozen of those CQE OCXOs from which to develop and build more ambitious GPSDO projects. 

JBG

[Johnny B Good]

 I've managed to sort the breadboard lashup out with a 5v cmos RRO opamp to buffer the Phase detector's LPF output so I could experiment with using a 1MR in place of the 100KR series element. This worked but it didn't really solve the issue with the plot error induced phase shifts (as correlated with the deviation map plot).

 Since all it seemed to do was extend the lockup settling time by an extra 5 minutes or so, I decided I'd be better off reverting to the original filter value and look to improving the external patch antenna location to reduce the excursions in the deviation map plot (at least there's less lag between what's shown in the deviation plot and the resulting phase shifts in the 'scope trace of my signal generator and the GPSDO reference).

 BTW, when I added the buffer amp, I was expecting to eliminate the phase shift effect of connecting and disconnecting the 10MR loading from the DMM I'm using to monitor the EFC line but, strangely, the loading effect hadn't gone away. After wiring up the second amp in the dual RRO opamp to provide a buffered DMM feed from the output of the opamp driving the EFC, largely to avoid leaving the second amp with floating terminals and the possibility of random oscillations, I was still experiencing the same effect.

 Eventually, I realised it wasn't the 10MR loading from the DMM but just the fact of adding extra wired connections to my lashed together GPSDO that was the culprit. In the end I realised that  I only needed to put my hand near the OCXO's output link wire to induce this phase shifting effect. It seems my concoction is suffering from 'Theremin Syndrome'(tm)  It's not a major issue with something lashed up on a prototyping breadboard but it had me scratching my head for a while after I thought I'd sorted the initial and expected effect of loading the unbuffered EFC with the DMM's 10MR voltmeter impedance by adding the opamp.

 Anyway, that was last night's good deeds (of work which didn't go unpunished) which included my having another go at calibrating the OCXO upgrade to the FY6600, seemingly with a lot more success than the last time I'd tried this with the rather jittery 10MHz output from the previous M8N module.

 However, it's no longer an issue of non-integer division of a 48MHz TCXO down to a 10MHz reference, nor is it the damned "Sawtooth" corrections to hold the PPS to within 20.8333ns of the nearest 48MHz clock edge. When it comes to averaging over periods of thousands of seconds (and rising), such minor deficiencies eventually dilute themselves into utter insignificance. What I find most troublesome now, are the nanoseconds (5 to 30 of them at a time) of phase shift due to position plotting errors as evidenced in U-blox's U-centre application's deviation map plots.

 When you're aiming for better than 30ppt accuracy, such variations swamp the hoped for very slow drift between the 'scope traces, making it difficult in the extreme to achieve the desired precision that is now a real possibility given a stable enough reference.

 Since I knew that my external active mag mount patch antenna location (on a ballasted biscuit tin on the flat roof of my office bay window with obstructing brickwork to the East) wasn't helping this situation, I decided to use an old steel drawer, ballasted with four used car batteries, to support the antenna on a 'jamjar' lid screwed into the wooden plug into the top of an 8 foot aluminium pole clamped to one corner of said ballasted drawer which now sits on the bay window roof in place of the biscuit tin arrangement.

 As per usual with these DIY projects, more time is spent in preparation which, in my case, is largely spent trying to track down the various items I knew should still be lying around somewhere in the basement than on the actual construction. I managed to sort out the major issues last night, leaving the fiddly details to be dealt with this afternoon. Suffice to say, after some effort hoiking the bits (including those four car batteries!) to the second floor bedroom above my office from where I could access the bay roof via the window, I finally managed to elevate the antenna into a much less obstructed view of the open sky.

 Mind you, the ridge of the roof is still a good foot higher but an extra foot on the length of the pole would have meant having the join between the end of the 5 metre antenna cable and the 5 metre extension outside which I wanted to avoid. In any case, the combined length of the antenna cabling is barely sufficient as it is and not very much of the Eastern sky is being blocked compared to how it had been previously. Indeed, I was seeing some 19 to 20 SVs (after adding the GLONAS constellation to my previously GPS exclusive setup) "In the Green" (right now at 22:55 BST, it's down to 15 'greens'). The sky plot, even after just six hours or so of data, is looking far better balanced all round with the obstructing effect only showing for SV elevations in the lower 10 deg section of the SE quadrant pair.

 I'm still seeing deviations of position but they seem to be spread over a smaller diameter, typically around 4 metres versus the previous 6 to 8 metre diameter 'dance floor'. I can still see variations in phase that correlate to the deviation map display which swamp the extremely slow drift due to the discrepancy in frequency between the free running OCXO in the signal generator and the long term precision of a GPS reference.

 It's difficult to be precise on account of these phase shifts but since it's taken some 5 hours for the signal generator's 10MHz to drift out of phase by a whole cycle, I reckon, more by pure dumb luck than anything else, that it's now only adrift by a mere 5.5ppt. I suspect it's passing through the 0ppt error point at possibly some 5 to 15 ppt of drift per day. I'll have a better idea when I check it tomorrow.

 This issue with phase shifts due to plotting errors with navigation only GPS units such as the M8N can be minimised by using timing GPS modules. However, I don't know whether the expense of such timing modules could be justified in a DIY GPSDO project. My question in this case being, "Just how much of a reduction in such phase errors can I expect to see with a timing GPS module?" If it's as much as a tenfold reduction, then it can allow me to continue relying solely upon a simple PLL arrangement without resorting to the extra complication of microprocessor control. Obviously, some means of tracking the EFC voltage to record the current trend and maintain it in the event of loss of lock would be desirable, implying the use of a micro-controller in any case.

 However, for a cheap DIY GPSDO, the Luddites amongst us can always record this voltage on a daily then weekly basis from which to manually trim the OCXO during such loss of lock events. Of course, this would only be putting off the inevitable upgrade to a micro-controller. In the meantime, I'll put up with these variations of phase until I get to grips with utilising that Arduino nano that's still sealed within its protective anti-static packaging.

 As for my signal generator's OCXO, it appears to have halted and reversed its frequency drift. However, I suspect it might be as much to do with room temperature changes as an ageing effect. When you're down to parts per trillion accuracy levels, ambient temperature changes as small as just one or two degrees, despite the moderately insulated ovenised XO, will start to show up. The phase variations from a basic GPSDO just make such changes look all the more dramatic, hence my new found annoyance. TEA anyone? 

JBG

[Johnny B Good]

 A quick follow up to my previous post:

 A check late this morning, after timing three cycles showed the signal generator had drifted from its previous 5.5ppt error figure to an error of 155ppt. Some six hours later and it would appear to have dropped to 110ppt. It does look as though ambient temperature is influencing the drift rate as I suspected despite it only being a matter of a couple of degrees (Celsius, FTAOD in the minds of the Leftpondian community).

 The skyplot does indeed show a marked improvement after yesterday's antenna relocation exercise. No doubt, an extra 18 inches of height would have provided some additional improvement but not a lot more in view of the upward sloping terrain to the east, along with the ground clutter effect of the dwellings located thereon.

 I've increased the default 5 degree elevation mask setting to 10 degrees which seems to have helped reduce the size of the 'dance floor' as it appears in the deviation map plots. This has reduced the number of usable SVs from a typical count of twenty down to eighteen (GPS and GLONAS constellations). We've had the odd heavy shower with variations between sunny blue skies and heavy overcast which does appear to have an effect on SV reception (possibly as much as a 5 to 6dB variation) which no doubt makes some small contribution to the "navigational fix errors" which impose a phase shifting effect upon the GPSDO output.

 With regard to GPS patch antennas, I am a little bit puzzled over an oddball unit I picked up at the Blackpool radioham rally some seven weeks ago. It's slightly larger than the mag mount gps 'puck' antenna I'm currently using (and isn't mag mount) and has two (not one!) co-ax cables exiting from a corner located strain relief grommet. One is about two metres in length, sporting the usual male SMA plug whilst the second 15cm one sports an in-line SMA socket. I'm guessing this second one is for GSM use rather than to allow another GPS antenna to be daisy-chained on but search as I might in the 'images' category, I've not seen any other examples that look remotely like it.

 Perhaps someone here might recognise it from the attached images and verify what the short co-ax flylead connection's intended function is. For added value, I've also included the ridiculously tiny active patch antenna that came with the NEO-6M module, and the salvaged passive patch antenna I'd rescued from the original NEO-M8N module. All three antennas are fully working (for ever increasing values of "working" going from left to right).

 Incidentally, the groundplane of that rescued patch antenna appears to have been 'tinned' with that low melting point desoldering alloy since a momentary contact with the dried out solder blob on my soldering iron bit resulted in an instant accidental tinning event showing itself as that marking on the opposite side of the centre pin to where I'd soldered the braid of the co-ax flylead.

JBG