Ghetto 10MHz GPSILTAOALAGKSINS

[purpose]

Afternoon all.

I've been playing around with one of those 5 quid, fake Chinese ublox modules and thanks to theHWcave's video on these units, I now know that anything other than whole number divisions of the 48MHz on board oscillator produces jitter.
So I thought of dividing to 2MHz (which is rock solid, six zeros when locked) and multiplying X5 with an NB3N502 and then band pass filtering.

Would you?

Peter

[Gertjan]

NB3N502 data sheet says: "Output Frequency Range: Min. 14MHz, Max 190MHz".....
https://www.onsemi.com/pub/Collateral/NB3N502-D.PDF

[purpose]

Thanks Gertjan.... I went straight to the small print.

[purpose]

I present the ggpsdo... No distribution, but I've got nothing to distribute it to yet.
Perfect sine wave output.

[bingo600]

Seems like a fun experiment.

But why are you always writing mHz (milli Hertz) , when you mean MHz (Mega Hertz) 

/Bingo

[purpose]

Don't report me to the police... I beg you.

[justanothername]

I present the ggpsdo... No distribution, but I've got nothing to distribute it to yet.
Perfect sine wave output.

I'd really like to see a close up from above.

[purpose]

Best I can do, I'm afraid.

[CJay]

Afternoon all.

I've been playing around with one of those 5 quid, fake Chinese ublox modules and thanks to theHWcave's video on these units, I now know that anything other than whole number divisions of the 24MHz on board oscillator produces jitter.
So I thought of dividing to 2MHz (which is rock solid, six zeros when locked) and multiplying X5 with an NB3N502 and then band pass filtering.

Would you?

Peter

Even an integer division of the internal clock will result in jitter because the correction isn't a nice clean frequency shift, it drops pulses which is great if all that matters is you have the correct number of pulses per time interval, not so great if you're looking for low jitter and not using it to discipline an external 'clean up' oscillator.

[iMo]

I always wonder how the below 30ns RMS translates into the output signal quality (without disciplining an OCXO..)..

PS: you will get the NEO's output with "a not broken jittery signal" when 48MHz is divided by N, where N=4,5,6,7,8,...
The 30nS RMS is a different stuff, however, it comes from above..

[rfclown]

What is the reference on your counter?

[purpose]

I'll have to take your word for that, Clay.
The ublox has visible jitter at anything other than whole number divisions, so no visible is fine with me.
This is after all an experiment for someone (me) that barely knows the difference between resistors and raspberries.

[purpose]

I'm sure you're right imo, but just seeing a nice 10 meg sine come from a 2 meg square was reward enough for my tiny mind.

[purpose]

rfclown... do I have to open it up?

ACT 10.00MHz 9452

[rfclown]

rfclown... do I have to open it up?

ACT 10.00MHz 9452

No. The answer is the internal reference. Some counters allow for an external reference.

[purpose]

This one not.

[CJay]

I'll have to take your word for that, Clay.
The ublox has visible jitter at anything other than whole number divisions, so no visible is fine with me.
This is after all an experiment for someone (me) that barely knows the difference between resistors and raspberries.

They aren't terrible, it really depends what you're using the output for,  were you seeing jitter on the counter at 10MHz?

[purpose]

With the ublox set at 10MHz, was seeing jitter on the scope and the counter was certainly jumping around.
2Mhz was solid on both, so thought I'd isolate the 10Mhz harmonic, but was then seeing instability.
Went for the 5X multiplier and band pass. Rock solid on both scope and counter.

[CJay]

With the ublox set at 10MHz, was seeing jitter on the scope and the counter was certainly jumping around.
2Mhz was solid on both, so thought I'd isolate the 10Mhz harmonic, but was then seeing instability.
Went for the 5X multiplier and band pass. Rock solid on both scope and counter.

Yeah, that's my experience of them too, even on a known genuine U-Blox, integer divisors are a *lot* better and make them useable in a lot of cases.

Worth noting that multiplying it up is a neat trick but does have the disadvantage of multiplying the errors too so the better the signal you put in, the better the signal you get out

Having said all that, I like your project, nicely done

[purpose]

Cheers CJay,
I won't be winning any awards though.

[justanothername]

This actually seems useful enough for me to try to replicate it. I really like the simpleness. Whats the type of the PLL pcb? Can't read it from the picture.

[purpose]

Clock Multiplier Module Frequency Multiplier Module 2-50MHz SMA Interface

Hold on though... The spectrum was much better before I lumped them together, so if you do the same, maybe space them out more and keep the sma stuff.

[justanothername]

It looks cleanly done to me. I propably would omit the series resistor from the ublox module to the PLL board, or maybe use the resistor itself as jumper.
You may want to try to use different power supplies since most of the USB chargers are mostly switched and noisy. And maybe put everything in a metal case.
I don't expect it to be as good as my GPSDO with ovenized crystal, but it is promising, especially if you put everything together onto a cheap single layout this could be made super cheap and may be enough for half of the tasks for an GPSDO.

[CJay]

Cheers CJay,
I won't be winning any awards though.

Nope, you won't but nor would I, I tend to build things like that to get a job done and then chuck them in a box or tear them down.

SMA connectors between boards and some shielding may improve matters but I wonder how good the 2MHz input looks on that spectrum analyser and if it shows the same pattern 'divided by 5'?

If you want to explore and experiment more it's worth thinking about what you intend to achieve and how the different approaches work, all the various 'frequency standards' have advantages and disadvantages, some are great short term, some are pretty awful.

FWIW, it's not that much more difficult to 'discipline' a cheap OCXO and get a really clean signal.

It is however considerably more difficult to discipline it well

[tszaboo]

Is this a GPSDO?
Afaik, those Ublox modules have a TCXO in them, where the 24MHz is not disciplined ( derived/compensated ) with the GPS signal.
The 1 PPS signal is, and you need to use that with the usual controller to make it 10MHz.

[purpose]

Justanothername,

That resistor is going to the led. I lost the smd part in the carpet!

I just separated the pcbs, powering the multiplier with its own supply and things are looking as they were.

I'll work on a pcb layout and share if it improves things.

[purpose]

Nope, you won't but nor would I, I tend to build things like that to get a job done and then chuck them in a box or tear them down.

SMA connectors between boards and some shielding may improve matters but I wonder how good the 2MHz input looks on that spectrum analyser and if it shows the same pattern 'divided by 5'?

If you want to explore and experiment more it's worth thinking about what you intend to achieve and how the different approaches work, all the various 'frequency standards' have advantages and disadvantages, some are great short term, some are pretty awful.

FWIW, it's not that much more difficult to 'discipline' a cheap OCXO and get a really clean signal.

It is however considerably more difficult to discipline it well

CJay,
I'm just a cheap tinkering old man. I just love to mess around.
I'm going to reinstate all the connectors and start spectrumizing for your delectation... hold my beer.

I know the scope likes 2Mhz, but I never shoved the spectrum on it.
The scope hated the 5X signal until I loaded it with 50R, but I'll get busy with a forthcoming slide show.

Cheers

[purpose]

Is this a GPSDO?
Afaik, those Ublox modules have a TCXO in them, where the 24MHz is not disciplined ( derived/compensated ) with the GPS signal.
The 1 PPS signal is, and you need to use that with the usual controller to make it 10MHz.

I'm just farting around, Mr. NANDblog and you just made me poop. I may as well throw it in the dog.

[tszaboo]

Is this a GPSDO?
Afaik, those Ublox modules have a TCXO in them, where the 24MHz is not disciplined ( derived/compensated ) with the GPS signal.
The 1 PPS signal is, and you need to use that with the usual controller to make it 10MHz.

I'm just farting around, Mr. NANDblog and you just made me poop. I may as well throw it in the dog.

What's with the attitude? I think I was civil when asking the question.

[purpose]

It was an attempt at humour. Obv failed miserably.

[justanothername]

Justanothername,
I just separated the pcbs, powering the multiplier with its own supply and things are looking as they were.

I guess the PLL can work with 3V3, so a low noise LDO like the TPS79333 may be used. A small pcb layout would be a nice thing to share! If you want to put something in just because you can, then go for a small battery charger to make it a portable device. The MCP733811T is an easy part to put in. VBAT is the input for the regulators then.

[shakalnokturn]

Best I can do, I'm afraid.

Ohhh...!
Ali-patch-art-work!

[purpose]

C'est vrai.

[purpose]

justanothername,
Battery power was my first thought. Leaving off the ftdi and putting on a programming header for it, was another, but my ftdi has just given up, so might have to get another type in to try.

cheers

[justanothername]

I'd say put it in, you can always choose not to solder it in. Looking forward to the result.
PM me if you need schematic design snippets for the ublox receiver with supercap instead of the battery.

[purpose]

Will do.

[TheSteve]

Is this a GPSDO?
Afaik, those Ublox modules have a TCXO in them, where the 24MHz is not disciplined ( derived/compensated ) with the GPS signal.
The 1 PPS signal is, and you need to use that with the usual controller to make it 10MHz.

Lady Heather would not be impressed, we all know she loves to discipline
Maybe this is more of a "GPS Derived Output" device.

purpose - btw, not knocking it all, I love DIY projects. It might be "Ghetto" as you call it for now, but overtime you can grow/improve it into very nice device.

[purpose]

Very encouraging, Steve. I'll have to get busy.

[rfclown]

Very encouraging, Steve. I'll have to get busy.

ot not. I would never discourage you from turning this into some super duper mind bending oscillator, but... I like the fact that this is simple and cheap. I've got a GPSDO, and I'm even waiting on a second one (eBay, China, don't hold my breath). But for most applications of mine, I don't need that combination of long term and short term accuracy. I like the idea of a relatively cheap reference that I know has good absolute long term accuracy.  I'd be interested in characterizing what the phase looks like in time. Does it jump occastionally to keep it in line with the 1pps?

[purpose]

Does it jump occasionally to keep it in line with the 1pps?

Can't say I've noticed it, but how long would I have to stare at the scope to discover such behaviour? Cough.
It does however have a 20mV, or so amplitude modulation that I'd like to iron out.

[justanothername]

I like the fact that this is simple and cheap.

+1 for simplicity and portability. Make the barrier low for others to reproduce your nice little creation, then they can help you classify it thoroughly.

[purpose]

Simple (and cheap) was the goal, justy.
WRT battery...18650, or bigish li-po?
On board patch antenna, or sma in?

I love these design by committee projects.

[justanothername]

WRT battery...18650, or bigish li-po?

I'd say just a JST connector, you can connect a li-po or a li-ion, the charger won't care.

On board patch antenna, or sma in?

SMA with bias if you ask me. Patch antennas are a little difficult to get good performance out of them, you need a big enough ground plane, had some serious troubles with them.
Furthermore, you mostly want to use it inside a room with the antenna outside. I'll send you a schematic screenshot via PM for a proven solution.

I love these design by committee projects.

[purpose]

CJay's slideshow.

2MHz out of ublox into 50R. Ignore Rigol's counter. Trust me.


Same on the spectrum.


2MHz X5 no load.


2MHz X5 into 50R


10MHz sine post BPF

[purpose]

Mr. Justy,
I'd say just a JST connector
So off board? I thought a li-po under the pcb would have been the trick, but a JST makes sense.

I'll send you a schematic screenshot via PM for a proven solution

Send away.

[geggi1]

These UBLOX gps modules have a clock that runs on 48Mhz.
The frequency on the puls output must be a even number of 48 to get the cleanest square puls out.
Frequencies like (Mhz) 24 , 16, 12 , 8 , 6 , 4 , 3 & 2 are the cleanest frequencies.

[purpose]

CJay,
The one thing you asked for (I think), I omitted. So ere ya go.
Post 5X multiplier, no band pass.


Your hint of filtering pre multiplier has just sunk in.

[rfclown]

I tried this... sort of.

I used u-center to change the 1pps output of a ublox 7M module to 2MHz. Do you know the commands to do this without u-center? I downloaded the document, but am too lazy to decipher.

I have the 10MHz outputs of my two eBay GPSDOs going to a scope. They are solid compared to each other. If I trigger off the 2MHz signal, I see that it is jittering compared to the others. You see the same thing when you look at the GPSDO 1pps output compared to the 10MHz output. The 1pps output has good long term stability, but it has jitter. You can't see it until you compare it to a signal that doesn't. I am using persistence to show the jitter in a picture.

[rfclown]

Setup. No persistence.

[purpose]

rfclown,
Ah, the old trigger deception. I was hoodwinked.
In my defence, I've got nothing between my ears.

If you haven't torn it down yet, can you in TP5, set both frequencies to 2MHz and note any difference?

Do you know the commands to do this without u-center?

I do not.

[rfclown]

If you haven't torn it down yet, can you in TP5, set both frequencies to 2MHz and note any difference?

Do you mean the two different frequency entries in the TP5 window? I didn't know what the difference was between the two. I was just trying things until I saw 2 MHz come out. I think I made the second entry 2 MHz and not the first. I'll try it when I get home from work. it's not "torn down", the rat's nest mess is as I left it.

[purpose]

Ha... my bench is far worse.

Do you mean the two different frequency entries in the TP5 window?

Tis exactly what I mean.

[rfclown]

Ha... my bench is far worse.

I doubt it. You're only seeing a few inches of space around the scope.

I've never had a neat workbench, home or work. Back in the 1990's someone came around at work taking pictures:
http://www.saolson.com/pics/bench/moto.jpg

[purpose]

Neatness is overrated.
They say.. tidy space, tidy mind. I like a messy mind.

[purpose]

Just put a quick and dirty low pass before the multiplier and cleaned it up a little more.

[rfclown]

If you haven't torn it down yet, can you in TP5, set both frequencies to 2MHz and note any difference?

No difference. Changed all the different settings, and no change.

[purpose]

She'll just have to be a bit jittery.
Ta

[rfclown]

Different views of the 2 MHz signal. I measured with a E4406A which can show IQ (down converted to DC) signal. This will show instantaneous phase versus time.

The first view is polar. The signal is not drifting in frequency, but the phase is varying several degrees. If the signal was perfect, the plot would be a dot on the graph. If the signal was drifting in frequency, the data would be constantly rotating around a circle (which it isn't). It's jittering back and forth a small degree. This is the same as the spread seen in the oscilloscope persitence picture of the time waveform.

In the polar plot, I is the x axis, and Q is the y axis. The amplitude is SQR(I^2+Q^2). The second view is I and Q versus time. The amplitude is staying constant, but the phase makes a jump about every 2.476msec. Looking at the next view, it shows that occasionally there is a double jump, which I found highly unexpected. The double jumps aren't all the same.

A spectral plot show something at 403.909Hz. this is the phase jump frequency. Why this frequency (or time) I didn't look at other output frequencies to see if this changed. I might later.

Anyway... That's what my ublox 7M does when programmed for 2 MHz output.

I'm using one of my GPSDOs as the reference for the E4406A. If I measure the other GPSDO, it shows a pretty much a dot on the polar axis. It isn't perfect. There's a small about of phase noise. See views of GPSDO 10 MHz waveform (referenced to the other GPSDO). I measured the GPSDO first to see if it was a suitable reference for this test. If I had seen jitter, I would have had to use a different kind of reference (like OCXO) which centainly would have a frequency offset, but I still could have seen the phase jumps. That's the thing with measuring this kind of stuff. You've got to have a reference to base it against. That's why time nuts can go crazy. You're always wondering if your best reference is good (because you don't have a better one to measure against). At least with GPS the long term stability is based on a bunch of expensive government hardware that wasn't bought on eBay.

I still find your project interesting, but we should find a different acronyn for it. When people asked if it's a GPSDO... well... it is sort of, but the time constant is like 2.5msec instead of 1000sec. (only 5 orders of magnitude different).

RFDO (really fast disciplined oscillator)
JDO (jitter disciplined oscillator)
ILTAOALAGKSINS (inexpensive long term accuracy oscillator as long as the government keeps cesium in the sky)

[2N3055]

I would call it GPS Corrected Oscillator - GPSCO

[MIS42N]

A spectral plot show something at 403.909Hz. this is the phase jump frequency. Why this frequency (or time) I didn't look at other output frequencies to see if this changed. I might later.

My guess a beat between the not quite accurate 48(?)MHz clock in the GPO module, and the accurate GPS signal.
Please don't call it Budget GPSDO, I want to use that for my currently almost developed project [see https://www.eevblog.com/forum/projects/lars-diy-gpsdo-with-arduino-and-1ns-resolution-tic/msg3334250/#msg3334250]

[purpose]

rfclown,
That is what I would call thorough. Thank you very much.
There really was no need to travel back to 1969 in the process, but many thanks for dusting off the DeLorean.

I shall henceforth rename it the ITBYG (into the bin you go).

Now I'm wondering if I should get hold of a T version, for a play.

Cheers

[purpose]

Please don't call it Budget GPSDO

Wouldn't dream of it, MIS42N.
OK if I call it Marjorie?

Ta

[purpose]

Well, well, well.
I plugged it into something other than a scope, or a spectrum analyser and it works a treat.
Pulled my Siglent sig gen into line with the gps (1Hz difference) and I couldn't be happier.
 

[MIS42N]

Please don't call it Budget GPSDO

Wouldn't dream of it, MIS42N.
OK if I call it Marjorie?

Ta

I was thinking pedantically, where is the oscillator being disciplined? It's more a GPS disciplined frequency synthesiser. I'm thinking the u-blox creates the 2MHz output (are you still using 2MHz - talk of 4MHz * 5 / 2 earlier) as a form of PWM using its 48MHz and inserting or deleting an extra bit now and then to make a more accurate 2MHz from a not so accurate 48MHz. Then it goes through a PLL (which I guess is really a disciplined oscillator - the VFO) to make 10MHz.

Maybe its a GPSDFSAO - a GPS disciplined frequency synthesiser and oscillator called Marjorie.

And you are happy with +-1Hz, job done. I have programmed my SRCOIB (semi random collection of inexpensive bits) to store the control voltage of the OCXO in non volatile memory and use it until there's GPS. If I turn it on without GPS and leave it for 15 minutes, it settles to 10MHz+-0.1Hz (and if it is about the same temperature as the day it was calibrated, +-0.01Hz). I'm still trying for better (why? because it's interesting. I have no use for a GPSDO at all).

[purpose]

MIS42N,
Maybe its a GPSDFSAO - a GPS disciplined frequency synthesiser and oscillator called Marjorie.
Chuckle.

Where is the oscillator being disciplined?
From others' analysis, it sounds like an occasional kick from the heavens.

Are you still using 2MHz?
I am. Into a 2MHz low pass- into 5X multiplier- into 10MHz band pass.

And you are happy with +-1Hz?
Considering I didn't know you could turn a square wave into a sine wave, let alone multiply it... I'm delighted.
Of course your contraption sounds amazing and one day I'd love to check it out.

why? because it's interesting. I have no use for a GPSDO at all
Same... I just love to tinker.

Cheers.

[purpose]

You lot thought you had got rid of me, but no.
Having found that crystal filters are a thing and availed myself of a NanoVNA (what a fantastic little tool), I made myself my first crystal filter.
-3dB 400Hz bandwidth 10MHz bandpass...


Feeding 10MHz from the gps module into the filter gives me this...


Not quite there, but almost.

Peter

[rfclown]

I just got a NanoVNA S-A-A-2 in the mail today. What a coincidence. I've actually got HP VNAs, but wanted to see what these things are capable of.

[purpose]

I just got a NanoVNA S-A-A-2 in the mail today. What a coincidence. I've actually got HP VNAs, but wanted to see what these things are capable of.

I'm sure you're going to love it.