Antenna for GPS frequency standard?

[Solder_Junkie]

From a time point of view, indoor vs outdoor GPS antennas are quoted as 500 uS vs 20 uS accuracy. I am trying to improve the short term stability of a GPS disciplined oscillator and am currently using the supplied magnetic mount antenna on a Window sill.

Am I likely to see an improvement in short term stability if I fit an outdoor “mushroom” type of antenna? I am seeing at least 8 satellites and am just using the GPS ones. Most antennas do not cover Galileo satellite frequencies, is it worth using them compared to GPS?

I am obtaining +/- 2 parts in 10^10 over a 5 minute period, maybe that’s as good as a low cost GPSDO can manage. I am not bothered about longer term stability.

SJ

[Gyro]

I've found that the performance of the standard magnetic mount antennas improves significantly (not surprisingly, given the intended mounting position) if you put them on a piece of steel sheet / groundplane. No need to go overboard on the size, mine sits in the middle of a nice decorative black painted sheet with rubber feet, approx 100mm x 120mm (what I had to hand).

[mendip_discovery]

I know there is issues if you are trying to get accurate time as cable length and how much cable is inside and outside etc.

I would say external should be better but no specific sceintific reasoning for it. My thoughts are even if you can see 8 satellites it doesnt mean you will get a good constant pulse as the the signal might not be strong so it swaps between all 8 and they differ in distance so causing differences in the travel time of the pulse. I may be completely wrong.

[Solder_Junkie]

My thoughts are even if you can see 8 satellites it doesnt mean you will get a good constant pulse as the the signal might not be strong so it swaps between all 8 and they differ in distance so causing differences in the travel time of the pulse.

That sounds to make sense and is probably the reason one receiver spec shows a huge difference in timing “jitter” between an outdoor antenna with a good sky view, vs an indoor one. Whether that equates to an improvement in the 10 MHz stability, I’m not sure, as the GPS controls an ovened crystal oscillator over a long time constant.

SJ

[mendip_discovery]

All gps provides to the oscillator is a 1 pps signal. So a deviation of that 1 second signal by a nano second or two can cause the oscillator to move about depending on how often it updates and how the GPSDO software handles the 1 pps signal.

Leo Bodnar has a nice external gps antenna that may get a a nice external signal.

I have two gps units here that I have mostly built. I'm just having issues with the STM32 chip. As I would love to test the two against each other then change things on one to see how it effects the readings.

[Solder_Junkie]

Interesting stuff. There is also some useful information on antennas in the Fluke 910 GPS manual, they have an incredible spec and mention the benefit of a clear sky view, but their unit is probably 10 times more accurate than my budget units.

SJ

[edpalmer42]

From a time point of view, indoor vs outdoor GPS antennas are quoted as 500 uS vs 20 uS accuracy. I am trying to improve the short term stability of a GPS disciplined oscillator and am currently using the supplied magnetic mount antenna on a Window sill.

Am I likely to see an improvement in short term stability if I fit an outdoor “mushroom” type of antenna? I am seeing at least 8 satellites and am just using the GPS ones. Most antennas do not cover Galileo satellite frequencies, is it worth using them compared to GPS?

I am obtaining +/- 2 parts in 10^10 over a 5 minute period, maybe that’s as good as a low cost GPSDO can manage. I am not bothered about longer term stability.

SJ

I don't know about a 'mushroom' antenna, but an actual 'timing' antenna mounted outside with a clear view towards the equator is worth considering.

When you're using an indoor antenna, the signal from the satellites is coming through wood, glass, concrete, cinder blocks, asphalt shingles, etc.  I expect that each of these materials has different RF characteristics in terms of loss and, in particular, delay.  Since GPS determines position by triangulating on 4 or more satellites and measuring the time between signal reception from the various birds, anything that changes the delay will distort the resulting position.  As the satellites move across the sky, the delay will shift as different materials enter and leave the path.  The loss caused by the roof can also be a problem, not only by reducing the signal level, but potentially causing satellites to appear and disappear as the signal attenuation changes.  With an outdoor antenna, all this nuisance disappears.

Timing antennas typically have tight filtering to reject nearby interference signals.  Navigation antennas, like your magnetic mount one, need a wider bandwidth to ensure they remain locked in a moving vehicle.  It's hard to tell if that improves the results, but it seems like a 'nice to have'.

If Lady Heather can talk to your GPSDO, you can set it up to plot the apparent position over a period of hours.  The result will be a fuzzy blob.  If the size of the blob is a few meters across, you're doing pretty good.  If it's 10 meters across or more, that's not good and a better antenna should be considered.

Ed

[3isenhorn]

Hi,

Just some Remarks, Galileo transmit at the same frequencies as GPS at least for L1/E1, so all antenna can receive booth systems. However, the receiver need to be capable to decode Galileo or other GNSS systems.

Within the system there are different frequencies named (GPS/Galileo name), finde more details on the nice band overview from R&S: https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/premiumdownloads/premium_dl_brochures_and_datasheets/premium_dl_poster/Wireless_po_en_0758-1029-82_v1900.pdf

Of course having a clear view to the sky improves the performance. However in metrology also multi path needs to be considered! This means no nearby obstacles where the signal can be reflected. So in summery a good antenna supporting a good view to the sky improve accuracy. Also a more advanced receiver with more frequency and systems support should yield better results.

[GigaJoe]

i got this , for a single reason - 3-18V
but it was a while ago
https://www.aliexpress.com/item/32987989516.html

now it more options for a cheaper
https://www.aliexpress.com/store/group/GPS-Antenna/1100221127_40000003293711.html

some work on 5V , some just 3.3V and internal VR blow up if it higher ...

[JOFlaherty]

Why would a view toward the equator be important for a timing antenna?

[edpalmer42]

Why would a view toward the equator be important for a timing antenna?

GPS satellites have an orbit that goes between about 60 degrees north to 60 degrees south latitude.  A clear view toward the equator will give you the most visible satellites.  Your GPS receiver, which should be a timing receiver rather than a navigation receiver, will then pick the best satellites to use.  A clear view towards the east or west will probably be okay.  If your only view is toward the north, GPS isn't going to work well, if at all.

The other GNSS systems like Galileo, Glonass, and Beidou probably have different orbital limits.

It also occured to me that any antenna that can receive multiple GNSS systems likely won't have a narrow-band filter like a GPS L1-only timing antenna has.  I have a Panasonic VIC100 antenna that specs its bandwidth as 1575.42 MHz +-1.023 MHz.  I don't know that the filter actually improves the performance.  It likely depends on how noisy your location is.  YMMV.

Ed

[mendip_discovery]

Just a bit of bedtime reading for you.

https://www.researchgate.net/publication/320798359_Evaluating_the_Frequency_and_Time_Uncertainty_of_GPS_Disciplined_Oscillators_and_Clocks

I heard that Michael Lombardi is the guru of the Freq/Timing world. https://www.researchgate.net/profile/Michael-Lombardi-4

[Johnny B Good]

 That paper did make interesting "Bedtime" reading for me even though MEGO skipping past the mathematical formulae. The main highlights for me being the estimates of ionospheric and tropospheric delays.

 Its antiquity shows in their out of date regard of the use of multiband GNSS receivers being a rarity in GPSDOs and GPSDCs. Since I'll be using a uBlox ZED9T (purchased over 10 months ago now) in my MK III GPSDO which I'm in the middle of constructing, I suspect that most calibration labs will have been using such dual band enabled GPSDOs for the past 3 or 4 years by now, along with the necessary multi band GNSS/RTK antenna (typically featuring multi-path reduction measures such as choke rings and the like).

 Apropos the original question, the first step in upgrading from an active patch antenna resting on a window ledge being to move it to an outside location, preferably with as unobstructed an all round view of the horizon as possible. If there are any local obstructions or sources of multi-path reflections, raise the RX's minimum elevation angle from the 5 deg default (a uBlox default) to 25 to 35 degrees.

 If you're going to invest in a better antenna, consider a multi band GNSS/RTK type in anticipation of (eventually) upgrading to a GPSDO blessed with a dual frequency timing receiver module such as that aforementioned ZED9T. I found that even with just the M8T module in the MK II, upgrading to a multi band GNSS/RTK antenna https://www.aliexpress.com/item/32815205556.html (similar to that BEITIAN linked to earlier) showed a noticeable further improvement in phase stability compared against my RFS. In other words, don't waste your money on those marine grade navigation GPS antenna kits.

[mendip_discovery]

I have it on my list to go though the maths and try and build it into a calculator that people can play with and see where the numbers are coming from. I would like to pull updates from circular-T to have it constantly update. I have been trying to learn Python and Jupyter to do this but have yet to get my head around it....oooh butterfly...as I get easily distracted by the 99 other things I would like to do.

I was wondering if the multi system setup would make a difference. The various satellite setups get thier timing from different locations so if one of them is significantly off would it cause extra error? But I will admit I am not sure where each system gets it timing from.

[fourfathom]

All gps provides to the oscillator is a 1 pps signal. So a deviation of that 1 second signal by a nano second or two can cause the oscillator to move about depending on how often it updates and how the GPSDO software handles the 1 pps signal.

Many GPS receiver chips/modules can deliver a sync clock much faster than 1 pps.  I'm fairly sure that the Bodnar uses a much higher frequency, and I know that some others do.
The principle is correct, the GPS-provided clock will drift around a bit, either from GPS limitations or due to internal clock frequency differences, and this jitter / wander must be filtered out by the VCXO feedback loop.  But with a reference faster than 1 Hz at least the problem is easier to deal with.

[Solder_Junkie]

Thanks everyone for their input. I opted for a standard single frequency “mushroom” antenna and have it mounted on a short pole on the edge of my roof. It has a view to the horizon over around 270 degrees, and down to around an elevation of 30 degrees to the north west.

My current home made GPSDO is purely “analog” and is based on the James Miller simple GPSDO. Originally it used a Jupiter GPS module, sadly after several years service that module failed and has been replaced with a Chinese GPS board that came with a fake Ublox module, that worked OK and gave similar results to a Leo Bodnar GPSDO that I also have.

I swapped the fake GPS module for a genuine Ublox M8T, and immediately obtained an improvement in the short term “jitter”, that was now around half that of the Leo Bodnar unit. The M8T is used in fixed position timing mode, set for GPS and Galileo satellites.

Spurred on by the improvements, I am in the process of building another GPSDO based on the design by Lars Walenius, and further refined by Murray Greenman. My measurements of the performance is done against my rubidium standard with a TinyPFA and TimeLab software.

Links to the above:
https://www.tinydevices.org/wiki/pmwiki.php?n=TinyPFA.HomePage

https://www.qsl.net/zl1bpu/PROJ/NGPSDO/New%20GPSDO.htm

http://www.jrmiller.online/projects/ministd/manual.pdf

SJ

[edpalmer42]

All gps provides to the oscillator is a 1 pps signal. So a deviation of that 1 second signal by a nano second or two can cause the oscillator to move about depending on how often it updates and how the GPSDO software handles the 1 pps signal.

Many GPS receiver chips/modules can deliver a sync clock much faster than 1 pps.  I'm fairly sure that the Bodnar uses a much higher frequency, and I know that some others do.
The principle is correct, the GPS-provided clock will drift around a bit, either from GPS limitations or due to internal clock frequency differences, and this jitter / wander must be filtered out by the VCXO feedback loop.  But with a reference faster than 1 Hz at least the problem is easier to deal with.

GPSDOs should always use a TIMING receiver rather than a NAVIGATION receiver.  Most navigation receivers throw in the 1PPS or other output signals whenever they have a few spare processor cycles.  The result is high jitter.  A timing receiver uses most of its processing power to make sure that only the best satellite signals are used to output the lowest possible jitter, regardless of frequency.  Since they don't have to frequently recalculate the location of a moving unit, they also have enough processing power to provide good location data.

While the VCXO can filter out noise and jitter, you obviously want to start with the best possible signal.

Ed

[Johnny B Good]

I have it on my list to go though the maths and try and build it into a calculator that people can play with and see where the numbers are coming from. I would like to pull updates from circular-T to have it constantly update. I have been trying to learn Python and Jupyter to do this but have yet to get my head around it....oooh butterfly...as I get easily distracted by the 99 other things I would like to do.

I was wondering if the multi system setup would make a difference. The various satellite setups get thier timing from different locations so if one of them is significantly off would it cause extra error? But I will admit I am not sure where each system gets it timing from.

 That's an interesting question. I've experimented with combining GPS with GLONAS and Galileo constellations in the past. some two or three years ago, and concluded that it was best to stick with a single constellation. The worst combination being GPS and GLONAS, second best being GPS and Galileo (only slightly poorer than GPS alone).

 At the time, the problem with Galileo being that it had fewer space vehicles (SVs) aloft. Assuming there is now a full population of Galileo SVs in service, I'd expect there'll be little to nothing to choose between either the Galileo or the GPS constellation regarding timing performance.

 Personally, I wouldn't trust GLONAS any further than I could throw one of their heavyweight SVs. I monitor the phase difference simply by using the infinite persist feature in the SDS1202X-E (effectively a graphical alternative to a TDC) and from this I got the distinct impression that the timing difference between GPS and Galileo was just a couple of nanoseconds whereas between GLONAS and GPS, it looked more like an 8 nanosecond discrepancy.

 Since other members had noted similar discrepancies using actual TDC measurements, I took the advice that it would be best to choose just a single constellation and disable SBAS (preferably the GPS one since it was the longest established and developed system at the time). I expect the advice today (assuming the Galileo system is now fully provisioned) would be to pick either GPS or Galileo and give GLONAS a wide berth. In any case, choosing GPS  on its own is a safe bet.

 As for the "...oooh butterfly..." effect you mentioned, that plagues me too In this case my presence here being a form of distraction from assembling my MK III GPSDO project. I'm still deciding where to mount the DIP chips on the single sided copper clad board I'm using.

 The construction method is a cross between "dead bug" and through hole mounting. I drill the holes to accept the DIP chips (mounted onto the copper clad side), chamfering those holes where I don't want the pins to be grounded and use point to point wiring on the underside. It's a neat alternative to getting a double sided PCB made up with a ground plane on one side and circuit traces on the other. Pretty well all the benefit of a ground plane with the advantage of prototyping flexibility (without a single "bodge wire" in sight  ).

 Although this MK III is essentially a respin of the MK II (based on the RUH design BTW), I'm still agonising over getting an improved component layout. The MK II design had no holdover to speak of so I'm looking to address that despite the absence of an mcu so whilst the design is essentially the same, there are some modifications I'd like to add (and some features I no longer require that need to be removed).

 I'd originally thought I could get away with it by working from the original circuit schematics and just "wing it" as far as the modifications were concerned. However, it finally dawned on me that I really need to redraw the schematics, meaning I'd have to put the time in to come up with the extra circuitry to provide at least a token effort at endowing it with a modicum of holdover along with the other changes.

 I'm mulling the problem over and trawling these topic threads to see if I can garner any hints that would help me optimise the final design, an apparent lack of activity that can so easily be misinterpreted as "procrastination".

[Johnny B Good]

All gps provides to the oscillator is a 1 pps signal. So a deviation of that 1 second signal by a nano second or two can cause the oscillator to move about depending on how often it updates and how the GPSDO software handles the 1 pps signal.

Many GPS receiver chips/modules can deliver a sync clock much faster than 1 pps.  I'm fairly sure that the Bodnar uses a much higher frequency, and I know that some others do.
The principle is correct, the GPS-provided clock will drift around a bit, either from GPS limitations or due to internal clock frequency differences, and this jitter / wander must be filtered out by the VCXO feedback loop.  But with a reference faster than 1 Hz at least the problem is easier to deal with.

 The original RUH design took advantage of the Jupiter-T's 10KHz locked to the PPS rate feature to reduce the divider chip count down from the otherwise required four to phase lock to a 1PPS to just two to lock to the Jupiter-T's 10KHz output.

 The u-Blox M7 and M8 series offer a programmable version of the Jupiter-T's 10KHz that could be programmed in 1Hz increments from 1PPS all the way to 16MPPS (or was it 24MPPS?) allowing the use of just a single 74CH390 dual decade ripple counter chip by configuring a 100KPPS output.

 You still see the 48MHz tcxo induced jitter in the 100KPPS output but the jitter induced corrections can occur as soon as the gps receiver detects the need which can vary from 5 times or so a second to maybe as few as one or two per minute, depending on how far the tcxo has drifted off its nominal frequency between its temperature compensation corrections. In the case of generating a frequency standard, this is of little significance over the minutes long averaging times typically used to discipline an OCXO.

 For single frequency receivers (both navigation and timing types) there are far worse imperfections in the GNSS system to contend with such as the minute to minute phase wander due to the limitations of ionospheric correction data packets issued every 12 (or is it 15? - I forget which) minutes.

 The dual/multi frequency receivers such as the ZED9T can remove reliance upon such correction data and calculate the required ionospheric corrections dynamically for each and every SV they're locked onto. There are, of course, other sources of timing/navigation errors but it's the effect of the varying ionosphere that is by far the greatest source of error which a ZED9T can reduce by an order of magnitude or so.

[Solder_Junkie]

Interesting info Johnny

Incidentally, I sent you a PM.

For anyone looking to use a Ublox M6, 7 or 8 (inc the M8T), there is a UK vendor on eBay selling cheap boards with fake Ublox modules that are pin compatible with a genuine Neo M8T. Removing the old module needs care, but I managed it.

The eBay item is 195486735732 and includes a USB configuration socket and SMA antenna socket. You can use the boards with the original module, just take the module label with a large pinch of salt (the label on mine showed Neo-M8N, U-Center showed it looked like a 7).

While a ZED9T sounds nice, obtaining a board with one ready mounted, plus a dual band “survey” antenna, looks a bit much for my needs. The ZED9T is not pin compatible with the Neo 6/7/8 series.

SJ

[MIS42N]

You do not say what this GPSDO is used for. 2 parts in 10^-10 is better than most hobby needs. If it was the basis of a down converter for amateur radio then 2Hz error in a 10GHz signal would bring smiles to most operators. Also it is a phase error of less than 100ns or less than 1 cycle (there are 3 x 10^9 cycles of 10MHz in 5 minutes - 300 seconds)

My GPSDO design is different to all others (that I've seen) in that the control voltage is changed infrequently (currently 1024 seconds). This means the short term stability is entirely up to the oscillator and not influenced at all by short term GPS variations. The amount of change is indicative of the frequency error in the last 1024 second period. So it is self monitoring.

I am using a patch antenna on a window sill and a NEO-7N, in general the control voltage changes are parts in 10^-11 (most under 2, 90% less than 4). So with sufficient averaging you can get quite good results with cheap gear. I have also used an M8T in the same location with some improvement. However, the survey in had an error of 10 metres indicating some of the signals were being reflected (no surprise there). I have not yet tried a well positioned outdoor antenna (currently in a frenzy of renovation).

A worthwhile design choice I made was to have separate power supplies for the oscillator and the rest of the circuitry. The control voltage is isolated as it is presented as a PWM signal through a buffer and filter so is independent of the rest of the circuit. Not so easy in an analog design. This design change was based on a previous design that showed variations from temperature fluctuations.

I have considered ways to improve this. The oscillator and other circuit share a ground plane, they should have separate planes joined at a single point. And some metal shielding as the circuit is sensitive to the computer monitor being switched on (only a few cm between them). So the theoretical ideal at this stage is a roof mounted helix antenna, decent low loss coax to the receiver, M8T is adequate when used with the long averaging times, a high end oscillator with low EFC sensitivity (e.g an OSC5A2B02 is 1V/ppm so quite sensitive to small EFC change. A Morion MV89 is 10V/ppm so much less affected by external influences). And metal enclosure with internal shielding. Maybe next year the way things are going.

[Johnny B Good]

Interesting info Johnny

Incidentally, I sent you a PM.

For anyone looking to use a Ublox M6, 7 or 8 (inc the M8T), there is a UK vendor on eBay selling cheap boards with fake Ublox modules that are pin compatible with a genuine Neo M8T. Removing the old module needs care, but I managed it.

The eBay item is 195486735732 and includes a USB configuration socket and SMA antenna socket. You can use the boards with the original module, just take the module label with a large pinch of salt (the label on mine showed Neo-M8N, U-Center showed it looked like a 7).

While a ZED9T sounds nice, obtaining a board with one ready mounted, plus a dual band “survey” antenna, looks a bit much for my needs. The ZED9T is not pin compatible with the Neo 6/7/8 series.

SJ

 I saw your PM but rather than answer it via PM, I think it would be better to reply here since the PM system doesn't support attachments and the reply, though peripheral to the topic thread, will nevertheless be of some interest to anyone thinking of DIYing their own cheap GPSDO setup.

 For starters, just in case you missed Gyro's original post, have a read of this one page topic thread (it was the diagram he'd posted at the bottom of that page which had inspired me to create my first prototyped  on a breadboard GPSDO from a genuine M8N module I'd bought several months earlier and a 13MHz AEL ocxo that I'd picked up for a mere 4 quid at the blackpool NARSA radioham rally just a couple of months after that)

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

 I didn't have a suitable RRO 5v cmos opamp in my parts collection at the time so I simply joined the LPF out direct to the EFC pin on the ocxo. Surprisngly at the time it worked quite well, considering the use of solderless breadboard to cobble the bits together.

 Just this week, I discovered why the lack of the opamp hadn't upset the circuit - it turns out that the 10MHz sine output versions' EFC pin have an input impedance in excess of 10GΩ! I'm assuming (untested as yet) that this was also the case with the original 13MHz sq wave version, which I'd purchased at the rally, that I just happened to be using at the time.

 So, it seems in retrospect that I could have gone one better (by design if I'd realised that the ocxo had possessed such a high EFC impedance) in reducing the IC count to just two (but I didn't have a benchmeter with a 10GΩ input option to test this at the time)! The fact that I had managed to reduce the IC count without compromising the LPF had been rather  lost on me (I was only trying this as a temporary bodge just to try and get some (or any!) sort of result to be getting on with. Of course, since I'd like to retain the EFC voltage monitoring feature, I've no intention of removing the RRO opamp from BoM.

 Using a 13MHz ocxo to create a 10MHz gpsdo frequency standard is far from ideal even if not too technical a challenge to achieve (see the attached images of the hand drawn cct diagrams for the gory details).

 I managed to touch lucky with a UK ebay seller who was selling the 10MHz sinewave versions of these AEL ocxos (supposedly NOS but in reality very carefully recovered pre-aged specimens). They were so cheap (£4.99 each - just a pound more than I'd paid for the 13MHz sample) that I took a punt on three. After testing them I bought another four. Only much later did I discover that they were indeed a high grade product by seeing exactly the same model mounted on the mainboard inside of a rack mount Symmetricom time and frequency standard in a YT teardown video.

 As a result of this good fortune, I have a rather jaundiced view of these CTI ocxos so I'd recommend you use the NEC ones you mentioned in your PM and put the CTI one aside as a "consumable/disposable" test component. BTW, what are the specs on these NEC units? Presumably much better than the CTI, surely?

 Since the ocxo / docxo / rubidium oscillator is the key component in any type of gpsdo design it's best not to skimp on its quality if you can afford it. The only good thing to be said about the CTI is that it's a cost effective step up from a vcxo. It was no accident as to why the G3RUH unit performed so well against commercial gpsdos since this was down to the deliberate choice of a high grade AXTAL ocxo.

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

BTW, I'm using the Sparkfun ZED9 T adapter board rather than a bare ZED9 T module. A more expensive option I know but less likely to end up in the "Valuable Parts" drawer unused. The Sparkfun adapter board is much easier to neatly integrate into a DIY GPSDO project (I've already got it mounted onto the mainboard).

[EDIT] I forgot to comment on those cheap Ebay modules you linked to. Aside from the fact that they're fake uBlox modules, there's also the possibility that the groundplane, onto which those patch antennas are stuck, may have active circuit traces carrying high speed signals which can render the attached patch antenna a useless ornament due to the unfiltered LNA chip being driven into saturation.

 This had finally proved to be real cause of the mystery failure of the patch antenna on my original genuine M8N module. If you're buying them to extract the fake modules to solder into a custom GPSDO board, that doesn't matter except to anyone wanting to use them as is.

 [EDIT] As per usual, I managed to get hold of the wrong end of the stick regarding your purchasing those cheap break out boards to use as carriers for your collection of genuine u-Blox modules. Hopefully, I've got the right end of the stick this time.   Either way, the issue regarding a patch antenna groundplane being riddled with circuit traces to batter the unfiltered LNA into submission remains a valid concern.

[Solder_Junkie]

Mis42 and Johnny, thanks again for the info.

My usage is non critical, 1 part in 10^9 is more than enough for my amateur radio purposes (cal/stability no higher than 1296 MHz). However, it’s keeping my brain active trying to improve on the equipment I have.

Attached is an image showing the comparison between a cheap eBay GPS module fitted with a fake Neo-6M module (green trace) and another of the same modules where the fake has been replaced with a genuine Neo-M8T (blue trace), In both cases the same outdoor “mushroom” roof mounted antenna was used.

The measurements were done with a TinyPFA (phase frequency analyser) and a rubidium reference. The software is TimeLab. The GPSDO is the James Miller Simple GPSDO fitted with an NEC Toyocom TCO-6703N oven. The rubidium is roughly 2 parts in 10^11 off frequency by my estimation, not that it matters for the comparison of two GPS modules.

By way of further comparison, almost identical results to the green Neo-6M trace are obtained using a Leo Bodnar dual BNC GPSDO, again on the outdoor antenna. The magenta and blue trace image is a comparison between my Simple GPSDO (Neo-M8T) and the Leo Bodnar, both on the outdoor antenna.

From what I can gather, a lot of traces published on line, mostly using TimeLab, are averaged over several hours… I’m not sure they prove much as taken over a long enough timescale they will match the satellite reference standard.

SJ

[edpalmer42]

What does the Allan Deviation of those fake module units look like?  That usually shows how good or bad a GPSDO is.

[Solder_Junkie]

What does the Allan Deviation of those fake module units look like?  That usually shows how good or bad a GPSDO is.

See attached. Unfortunately, as mentioned in an earlier post Allan Deviation over a longer time scale tends to show a good result from any GPSDO. I am more interested in the short term stability over perhaps no more than 10 seconds as the GPSDO is used to calibrate test equipment and to hold UHF radio equipment on frequency.

The setup for the Simple GPSDO by James Miller (G3RUH) is with an external "mushroom" antenna with a fairly clear sky view, Toyocom TCO-6703N oven osc, cheap eBay (£7.90) GPS module with a fake Ublox Neo-6M module. The reference is a rubidium oscillator which is off frequency by 2 parts in 10^11, as shown in the long term averaged image. A TinyPFA was used as the method of connecting the 10 MHz reference signals to TimeLab.

SJ