GPSDO project

[throbscottle]

The Marconi arrived today. Frequency counting works ok, counting works ok, haven't been able to test the ratio function. Period measurement not working AFAICT though it should be easy to troubleshoot since I found a schematic (on mods.dk) and that particular function is down to a switch and some gates. A co-ax connector had fallen off it's mate inside the case, didn't work till I put that back on.
Shame it's in the 19" rack option, I really wanted the standard width. But then it occurred to me, since there is an empty bay in the rack case, that would be a good plac to put the GPSDO!
I had been thinking of cutting the case in half since it has an internal divider. Ugly but I gets me approximately what I want, since I don't have room for a 19" case. Guess I have to think again. Time to put another shelf up in the shed I think.
This thing is OLD! I found a manual dated 1979... But I checked it against the OCXO, only 3.7Hz out. Don't know if that's the oscillator that's off or the counter.

 

[dr pepper]

'79 and still going, I wonder if a newun would still be working in 50 years.
Cut your bench in half, and add a bit in to make room for the 19" rack.

 

[throbscottle]

I doubt if it's actually that old though. I wonder how long they made them for?
Cut my bench in half! Ha ha ha ha ha ha!
My "workshop" is a garden shed, 8' x 8', with one corner chamfered off where the door fits it and opens on to a diagonal, and my bench is made out of the door of the old shed that stood there before it, and also does duty for woodworking and metalworking projects. Between the bench all the stuff stacked up there is room to stand or sit on a stool. On the actual bench there's about half of it I can actually do anything on, and quite a lot of that is taken with various bits and pieces.
It all desperately needs sorting out and I've been meaning to do it all year, just haven't found the time. And now it's going to rain until March so that limits the sorting out because I can't stick stuff out on the grass whilst I do it.
So at the moment a shelf is the thing, I'll have to move the tool board perhaps, or just have a slightly high shelf. At least I can make some room on the bench that way.

 

[throbscottle]

I just tried the circuit. Got the /5 wrong! The reset pin needs to be high, not low, and it won't act as a divider as I've shown it, the outputs just stay low so it doesn't get the high it needs to set off the division. Tried putting in an inverter in the feedback, works but divides by twice what it would have done if it worked without, so my /5 is /10, completely useless. Grrr. So I went to try putting input A on the clock and input B in the loop I had, but suddenly I'm getting no output from the oscillator. Grrrr. Hope it's just a flakey connection or I will be getting in touch with the seller after all!

 

[dr pepper]

The problem with circuits that use frequency and phase is that you have all kinds of counters & dividers.
No idea how you managed to double division ratio with an inverter, sounds like a 'what the' moment.
If I get in a situation where everything is going weird I tidy & sort out all the little bodges & bad connections etc & try again, you can get yourself in a situation where you have something lashed up & end up chasing yer tail will iffy joints & components not making contact in breadboard etc.

 

[throbscottle]

Well it was more of a "whaaaa??" but yes. I built a model of the '164 in Logisim, inverter definitely needed.
After many hours of head scratching - even built the circuit on solderless bb with pushbutton input and led output, I realised what is going on. I had interpreted 5 input cycles with either 5 1's or 5 0's on the output as divide by 5. IMA dummass.
So now 1 MHz it is. Still an integer division of 48MHz so that's ok. Now I'm curious as to how to divide by odd numbers.

 

[dr pepper]

Me too no doubt.
My Msf standard divides its 10mc ocxo down to 60kc by first dividing by 10 to get 1mc, then multiplies by 3 using a Pll and a 4017, then divides by 50 to get 60kc using a 'hc390.
You can divide or multiply by an odd number by using a 4017 or 4017 and a 4046, theres also a cmos series chip that has a built in counter osc that has 'jam' inputs and can be used to divide by 0 to 255, plus prescalers and maybe even postscalers.
You can tell why you have phase noise with a non integer division ratios, there are probably other means to do it however a Pll is one of the common ones, various noise introduces oscillator drift.

 

[throbscottle]

I started looking it up, you can do it by dividing down the clock by the same amount creating 2 even-division phases with 90° shift, and xor them together. I saw an example for /3, I presume the same principle applies for other amounts. You get a jitter free output that way.

 

[dr pepper]

So said jitter would be repetitive and could be removed by a loop filter.
You can do something similarish by looking at rising edges of 2 waveforms.

 

[throbscottle]

Oven ready detector works reasonably well, except sensitive to changes in supply voltage. Given it it's own zener.
Found at least one transformer I can use. I can lose a couple of 1n4001's as split windings are available, yay!
Getting ready to make PCB's - which means it will probably take me another month!

 

[dr pepper]

Learn from my mistake keep power & transformer magnetic fields away from yer control circuit, if you have a torroidal trans use it.
A wall wart would be good here.
I spose if you were inclined make a reasonably accurate oven ready Led circuit, integrate the current consumption of the ocxo, and if stable for longer than it tkes the integrator to hit the rail light an Led.

 

[throbscottle]

Most of what I've read for these things say use an internal linear supply and avoid wall warts because of switching noise, so I've gone with the same route. The plan is to divide up the box with shields anyway, although I admit I'd got obsessed with a more and more compact unit and forgotten about it!
Not sure about making a more accurate "oven ready" circuit. I suppose just an R and C would do the integration.
If the PLL is missing it's sigin input from the GPS, I was under the impression it detects this and stops adjusting the output, but apparently not, it just adjusts to the lowest level. So I need to make a detector that can turn off the PLL if there's no GPS. Hmmm. Could also warn if the GPS has lost its settings by flashing a PPS LED. I'm going to need a bigger PCB. OTOH, I've already got a few components on the other side and was thinking about flipping more, so maybe that's the way to go.

 

[dr pepper]

Yes of course I meant a linear wall wart, digital roar from an smps is going to drown your signals.
There isnt a precision way to really get an oven ready signal, I look at the control voltage on mine, I can tell when its all warmed up & ready to go.
If your not using a micro you could detect missing 1pps pulses using a 555, and another can flash the led, though this does kinda call for a micro doesnt it, then your into digital algorithms and everything goes really complicated.
I think an analogue Pll is really hard to beat for synchronizing a Pll, even with really noisy signals, you can improve on this with even an arduino but you'd be in for a lot of headache.
The Z box diver sent me is still working well, the 10mc o/p reports acuracy of a few nS.

 

[throbscottle]

I was thinking to use a diode detector like for AM except it just outputs DC. It could drive an LED via a tranny. Off = no GPS, on = signal good, flashing = re-program needed because it's now at the default 1PPS instead of 1MHz. Though I guess I might need an extra filter for that. The GPS module has a LED built in to flash when it gets the 1PPS signal - don't know the effect it would have on the 1MHz that's replacing it - so it's possible it will do the job all by itself. If it does I could run a light-pipe from it to the front. Or use a photo-transistor to pick it up. Oh but I'd still need the detector to turn the PLL off.
The oven-ready signal is more a case of bells and whistles. I know it needs x amount of time to settle anyway. The only thing I could think of to make it more accurate is a thermistor to compensate for the ambient temp, but then I realised ambient will likely stabilise at a similar level regardless inside the box anyway, especially with the oven's regulator chucking out about 3W anyway.
The only linear wall-warts I've got won't supply enough current!

 

[throbscottle]

This works in LTSpice. I have to wait for new GPS module to test it for real.

 

[dr pepper]

Looks Ok.
However I seem to remember theres something goofy with the enable i/p on the 4046, certain manufacturers version that is, Pc2 doesnt go high impedance when the chip is disabled on some devices.
You'll have to play with it & see, preferably before you etch a board.
I have some 6a 12v wall warts, Ok probably not warts they are inline, mains cable in Dc cable out, oddly enough one is called a 'power brick'.

 

[throbscottle]

Oh ok, thanks for the heads up on that! Is that just the original or are the 74 versions affected as well? I'll have a play anyway. It may be a significant stumbling block because I'm publishing the project as an Instructable, I wouldn't want other people to end up with something that has a hard-to-spot fault. I suppose I could switch the OCXO output to both inputs to achieve the effect or pull them both one way or the other.
Power bricks. I thought it was only me that said that (or just "brick")

(Edit) Oh wait - I've got the PCII output going back to the VCO input anyway, it uses DEMout to drive the oscillator's control pin. Inhibit definitely controls that one! I'm assuming the output goes hi-z because it's there to minimise power consumption. I just never thought to check! The oscillator's control pin does have a voltage on it, about 0.5v in this case, so I guess it will tolerate some resistance there, just don't know how much.

 

[throbscottle]

Well I did the experiment - a while ago now. Once the inhibit pin is high you can see the voltage going back to whatever the oscillator is it at on the output. Replacement GPS module arrived a couple of days ago - taken long enough! So I can test properly now. Also got an antenna to stick in the shed roof. I'm going to assume the thin roofing boards won't have much effect on it (don't trust it being outside). Been trying to figure out how to get the "survey in" function to work (for fixed location devices). Just sits there not getting any coordinates. It's one job where a micro would come into it's own, running routines like that. Not that I'm going to bother else I'll never get it done. Well, maybe one day.
There was one other thing I thought to add to it, and that's a tuning control for if it's got no GPS signal, so it's still got some calibration. Not sure if it's just a case of adding a trim-pot to the oscillator's tuning pin or if there needs to be more to it. It's not "holdover" but it's better than nothing.
Had a good sort out in the shed so now the counter is sitting in the spot where the box of awkwardly-sized bits of wood used to be. (Too small for one storage space, too big for the other). Pretty sure I'm going to put this GPSDO in the empty half of its case. Now I have a ton of board pulls to sort out!

 

[dr pepper]

Use ucentre the pc software for the neo series, you can do all the setup stuff with it, you can even use the ublox as a programmable frequency source.
This is assuming you have a ublox receiver.
Mine works indoors through cement tiles & wood, all depends if said wood has a metal foil or something.
Yes one way of a simple 'holdover' would be to run it for a couple of days, make note of the tuning voltage then set a trim pot to that voltage and switch it through to the oscillators voltage control pin using a 4066 or relay when the gps signal fails.

 

[throbscottle]

I think you forgot, I already said I'm using a Neo7 and the U-Center to program the Neo as a frequency source! 4066, hmmm, not sure if I have one. But already have lots of small relays, hmmmmm...