GPSDO project

[throbscottle]

Aha! I found this highly useful document: https://www.idt.com/document/apn/849-loop-filter-component-selection-vcxo-based-plls
Just need to change a bit to account for 4046 not having settable charge pump current. I finally see the reasoning behind this loop filter: https://www.simonsdialogs.com/2014/12/dcf77-frequency-reference-a-resonably-accurate-10-mhz-source/

 

[dr pepper]

Interesting.
Good to see you learning.

 

[throbscottle]

Update
It's coming along nicely. I made my first ever plated-through-hole pcb for it. There were some mistakes, like deciding this was the time to experiment with a longer pcb exposure time - leading to bridged tracks and spread pads (on the msop chips - ouch!), relay footprint is mirrored so had to put the relay on the bottom of the board (and very glad the PTH worked on /those/ holes!) got the inverter connected the wrong way on the divide-by-10 - don't know how that happened but it's annoying I must have looked at it dozens of times and not noticed!
Seeing some noises on the tops and bottoms of the square wave outputs but it could just be an artefact of my flakey test gear.
I'd changed the diode orientation on the lock detector because it worked better on the perf. board, but think I need to change it back again.
Changed the LPF at the input of the distribution amps for a ferrite bead. It's dropping more signal than I hoped. Have to try some others. Might have another go at the LPF, need to wind a 1uH coil for it, is a pita.
So anyway, need to bash some metal next, chassis to go in the case.

 

[dr pepper]

Excellent.
It took me a while to get exposure right, 4 mins for small boards and 4.30 for larger ones, on my maplin box.
Did you tune your 'scope probe using x10 and the o/p on your 'scope?
Have a play with Ltspice for the lpf, I refused to use it for years, then suddenly wondered why when I saw how handy it is.
Post some pics along the way.
Divers Hp reference is still working well, I did some mods now it locks faster.
My references still work too, I use the Radio4 and Msf one because they are ready in a minute at good accuracy.
I found one handy thing to do is pre charge the loop filter cap using a 4066 switch close to the correct running freq, lock time is seconds.

 

[throbscottle]

I used the Iowa Hills RF filter designer. (download link at bottom of page) Neat little program. But since it only needs to block any stray harmonics from the digital side I'm sure a ferrite bead is all it really needs.
it's a cheap probe I can't quite set right, and the 'scope is ancient and I'm not sure it's response is flat any more, following the last round of repairs. Need to check it with the 'scope's supplied probe (though the tip is huge!) My next project will be getting the 'scope set up right again.
I think my UV box is over-powered. 23 or 24 elephants exposure is all I need!

 

[dr pepper]

Ok I'll have a look at that, yep a bead might do.
There are certain kinds of coax that they say degrades, so its possible your olde probe has had its day, though I have to say I've never experienced co ax degradation, not sure what happens to it.

 

[throbscottle]

No no, the old probe is fine, it's the new (cheap) ones that are dodgy!
I see commercial equipment cables that get degraded by repeated flexing all the time though (particularly bar-code scanners). Maybe that's what is meant, the conductors break and the insulation wears away.

 

[dr pepper]

Yes that sounds about right.
Old stuff often outlasts new stuff.

 

[throbscottle]

Seems the pigtails I bought are pretty crappy, signal-wise. I think the rest may just be down to board layout. Probe not so bad after all, it appears.

There were supposed to be areas clear of copper under the chips but I lost the copper-keep-out areas from 2 of them and didn't notice until I'd etched the board. I'll try and carve them out now I've seen what's going on. Looks like I have a slight de-coupling problem too, so maybe chuck in some different sorts of caps and see what happens.

So what I'm seeing on the TTL outputs is over-shoot and a bit of ringing on the rising edge, slight ripple on the top of the square, and some rounding off of the corners of the falling edges. The 10MHz output has very strong over-shoot, but if I connect it through the 'scope's 50ohm input it goes quite rounded instead, but I might be seeing a limitation of my 40 or 50 year old 40MHz 'scope there, in need of adjustment as it is. Maybe I should put some 1k resistors across the outputs and it'll come out right.
So is there a straightforward fix for the peaking and rounding?

 

[throbscottle]

Photo of construction so far. The PLL chip is on the underside of the board with a few other bits.

 

[dr pepper]

Did you setup your 'scope probe with the 'scopes squarewave test o/p first?, if not you maybe chasing ghosts.
As you approach the bandwidth of the 'scope a square will gradually turn into a sinewave.
If you still have issues then you have power supply or trace/wiring stray inductance issues, one way to deal with that is to forget them on the board and use a line driver chip, not very practical now you've built the board.
Could also be the o/p impedance isnt 50 ohms too, might be able to improve that with a impedance equalizer, first though is it really going to be an issue, my hp o/p's a fairly nasty squarewave with cheap leads, but it works fine for ls and hc logic.
I put togther a frequency divider circuit with 74hc chips, it worked, I blew on the hc's so I put in a 74F I think it was, it worked but everything on my bench went screwball from the sharp edges it produced compared to the hc, perfectly square edges might nit do you any favours anyway.
Nice board by the way.

 

[throbscottle]

I did, but it wouldn't go quite right. I did eventually reveal a certain amount of ghosts, the overshoot and ringing aren't nearly as bad as I thought they were. I think you are right about pushing the 'scopes limits too. It's a 40Meg CDU150 so the fast parts of a 10M square wave will be beyond it's limits.
Outputs are just TTL except the 4 analogue amps which are back-terminated.
To be honest as long as they're clean I'm not terribly bothered if the square waves aren't as square as they could be, whatever those outputs get connected to will sort it out anyway.
I'm re-designing the board anyway to make it more RF-friendly, correct some mistakes and put in a couple of changes. And to see if I can better THP!

 

[dr pepper]

If you cant calibrate the lead then either its not compatible with your 'scope or one of them is faulty, but if your able to get what you want from it fair enough.
A 10 mc square will probably never 'look' square on a 40mc 'scope, as the edges will contain some much higher harmonics.
Nice to see a nice professional job.

 

[throbscottle]

Thanks for the complement. I think because of it's age the 'scope is set up to use the leads that came with it to some extent. I can get a nice square test signal using them.
I've been reading up, and found my problem with overshoot and ringing is due to bad board design. I thought I was doing the right thing by removing ground plane under the digital output traces to reduce capacitance, turns out it has the effect of increasing inductance! Finding some other changes I need to make too. Just down to inexperience and having to learn as I go along.
Main thing now is to make sure it does actually lock in properly now it's on a PCB instead of perf-board. Flakey wiring is no longer an excuse!
I keep having the re-program the GPS module because of leaving the battery disconnected. What a dummy...

 

[dr pepper]

Just like I was saying in #111, stray inductance is tricky to deal with, it gets you when you make transformers too.
I have to keep reprogging my gps rx too, because it has a silly supercap that lasts about 2 hours.

 

[throbscottle]

I think you are doing well to get 2 hours out of it!
I put a battery in my PSU to supply the GPS's memory. In theory it should maintain it for a year when fully charged but it lost it's memory after a few weeks (between testing and this build) so something is wrong somewhere.
I saw one project someone had done where they'd avoided needing to supply backup power but they didn't show how they'd done it.
Someone suggested it might be better to burn the settings to a micro which can re-program the GPS every time it's started up. Tempting but I'm not sure I can be bothered!

 

[dr pepper]

I'd considered burning the rx every time too, realterm can record a block of data so it could be done, like you too lazy, plus I odnt use my gps standard, I use the Hp.
I wondered about combing the Radio4 standard with the Msf one, as one is north of me and the other south, so atmospherics would be compensated for, cant be bothered with that either, I just dont use it at sundown, no need to worry about sunup I'm either here at work or snoring.

 

[throbscottle]

Well, I ran it all day today, still didn't stabilise.. Need to change the loop filter. Seems the lock indicator LED is affecting the control voltage when it goes on and off too.

 

[dr pepper]

Yes I had troubles with the loop filter, by far its the most difficult bit.
I pinched some ideas from existing frequency standard projects on the net in the end.
Really you should have 2 power supplies, one for the osc & Pll, & one for eveything else, you'll probably find too that when you connect a load to the o/p it too will send the frequency on a bit of a wobble.
I odnt think you need to go as far as optically isolating everything, however a high value resistor driving a tranny on a seperate rail ought to improve things, start grounding with heavy traces/wires is also a good idea.

I just got myself a cheap hifi tuner off ebay, I'm toying with the idea of using the stereo pilot tone from the bbc as a reference for a super portable standard, its a constant umodulated tone, and is as accurate as radio4's carrier.

 

[throbscottle]

Hmm, few more mods for the next version then.
At the moment the osc has it's own PSU with regulator, the other 5v supply is split in 3 using 10 ohm resistors each with it's own filter and bulk cap so the digital chips and PLL share a branch, the GPS has a branch and the analogue amps have a branch.

I can see the sense in putting the PLL on the same supply as the osc though. I never thought of that before.

Wow. I find old tuners from time to time. Never thought of actually using one for anything!