GPSDO project

Well your memory goes when you start getting on a bit.
If you use a relay you'll need a lpf on the control line for the vcxo.

I set up the Neo7 to only output a 1MHz signal when it has satellite lock, looked for a minute as though I couldn't do that. Had to do an odd configuration of 0% duty cycle when it doesn't have a signal. So the on air/off air indication works really well, and it flashes at the default 1pps if it forgets its settings. I connected another transistor to that which drives a relay to change to static voltage control when the signal is gone, and then found the OCXO's reference to be really intolerant of any loading. Even a 100k resistor pulled it down half a volt. So don't know what's going on there, maybe it got damaged when I roasted it (ran on 12v not 5v since it was sold as a 12v unit). Just a preset from ref to gnd isn't going to cut it anyhow. So maybe that's an "extra" I will do without, if I start farting about I lose the advantage of having a reference built into the can.
Something strange is going on with the pll's lock indicator. It flashes away when the signals are a bit different, flashes quicker and eventually lights solidly. But I found it can be lit solidly if there is quite a big difference in frequency too, even with the gps 1MHz signal set to 100KHz, still got that light lit up. Ho Hum.
Everything else looks to be working well.

Any suggestions as to a good op-amp (or other device) to buffer the reference?

I think the Led connection is open drain.
I connected mine direct to a 74hc4046 input, the one that can do analogue, and used 2 voltage dividing resistors to bias the input 1/2 Vcc and connected the ouput from the sat rx through a 100nF cap, not sure if that'll work with your setup.

Ok I'm not seeing what you are referring to, dp.
I'm thinking with the strange behaviour I'm seeing it is because every so many cycles the signals coincide so I get an output at pcp. So I just need to make the LED circuit less sensitive.

If you connect a 100k from gorund to the 4046 input, then another 100k from the 4046 input to Vcc, then the imput will float at 1/2 Vcc, then connect a 100nf cap from the output of the Gps Led output to the 4046 input.
The 4046 will then be biased at 1/2 supply, only the slightest amount of current will pull the 4046's input above & below the logic thresholds, the 100n just isolates the Dc on the 4046 input from getting to the Gps Led output.

It's internally biased anyway! Already connected via a cap.
I think we are at cross purposes. My fault, switching subject in one post.
I was referring to the voltage reference from the OCXO needing to be buffered, but then also to the PPS output from the 74HC4046 behaving inconsistently. No problem with the inputs at all!
Here's how it's looking now. Everything works AFAICT apart from the PLL lock indicator, and I haven't got round to adding the buffer to the oscillator's voltage reference output yet, and I haven't added the zener to the "oven ready" indicator yet (it is sensitive to changes in supply voltage). And I haven't built the sine output amp yet either, but now I have half a dual op-amp spare I might use that instead. Also seen some better discrete designs to try too.

Looks like we both got crossed wires!
If you have a spare op amp you can use that to buffer the reference, the op amp will be subject to offset drift with temperature, if your not bothered about a couple of mV though its still useable, otherwise you could use an ultra low offset drift op amp like a Op07, or a chopper stabilised amp, they have silly low offset.
Thinking about it the output from the Gps is 1pps as you say, that might not be so great for the analogue ip of the 4046, too slow, you might need to put a schmitt trigger or transistor circuit between the Gps and the 4046 to square the signal up a bit, another thing with being slow is the 4046 will warm up as its input buffer wont like being operating in its linear region long.

Neo 7 programmed to 1 MHz on the pps output remember? Only reverts to 1pps if it's forgotten it's settings, which is useful as an indication if this has happened. ICL make some very nice choppers. But is it worth it since it's feeding into a multi-turn preset anyway, as a standby in case the gps fails for some reason.
Maybe the OP07 or similar is the thing. I hadn't thought about offset and drift in terms of the amp being connected as a buffer.

Nope I dont remember, Alzheimers.
Ok now I remember that makes sense, mines the same.
My neo doesnt remember its settings anymore when powered down, the batteries dont seem right good.
Yep offset drift can make a mess of voltage references, I made a milliohmeter with low test current a while back I used an amp that has digitally selectable gain in decades, just controlled by a front panel switch, its a chopper and has less than a uV drift per deg C.
If you could squish the op amp in the oven with the crystal its drift accuracy would be much better.

I'm mounting the OA right under the oscillator anyway (on bottom of board) so it should be pretty good. The main enemy it seems is voltage swing, the supply is 5v, the reference is about 2.7v, the cheaper end of low drift OA's just barely have enough headroom. And single supply, low drift, r-r ones are quite expensive! Ok doesn't need full r-r but that's what's available... OP1177 won't break the bank... There are some nice ones from Microchip too but have to buy at least 5.
The whole board will be enclosed in a can with the oscillator so the temp should be pretty stable anyway.

You'll need some headroom on the op amp so it can drive the output under load, give it a try see what happens.
Keeping it nice & warm with the osc is a good idea.

LMC7101 is reasonable price and more than adequate. Ordering some from CPC.
I realised what is going on with the PCP output now so just need to come up with a more appropriate filter. Strange how copying the one from the 7046 doesn't work since everything internally up to the PCP output looks identical.

I cant remember now (remember Alzheimers) however the way the Pcp works changes depending which Pc you use, the Pcp output is part of Pc2, did that make any sense?, I think its getting late.
Beware diffo manufact chips can and do work differently, esp with cmos logic.
I have the fortunate circumstance of living a mile from Cpc and going past most days.

I think I'll go to bed now, night night.

I gave up on trying to design a discreet buffer for the sine output. So I've ordered some Exar CLC2007 dual op-amps from RS (tiny MSOP packages, eek!). Plenty of bandwidth and output drive.

Since I found I can get BNC sockets pre-wired with SMA terminated pigtails cheaper than solder connection sockets alone, I've increased the number of outputs. So with these op-amps there will now be 4 sine outputs.

I couldn't find anything I wanted from CPC to make up my order to avoid the handling charge, so I didn't order the LM7101's eventually, gone back to LM358. I realised stability isn't an issue since the control voltage wanders anyway and the op-amp will be at a stable temperature.

Having the re-do the PCB layout since a few things have changed now.

SMA edge mounted sockets for the board connections, now. Starting to look quite smart

Agreed, an op amp is going to be better than a quick designed discreet option, exar will have spent a load of time getting it right.
The control voltage is meant to change with temp anyways, the '358's will be fine, the osc will have a low pass on its i/p anyway.
Sounds like it'll be nice when finished.

Built psu board, just built one channel of distribution amp to test, will try it tomorrow. Very fiddly soldering bits of magnet wire to MSOP 8 package!
Trying to properly get my head round the loop filter calculations now. Uhhg.... The maths is easy enough, it's figuring out the parameters, especially when not using the on-chip vco.

Download Ltspice & simulate it, I owuldnt even bother with the calcs, it hardly ever comes out as calculated.

Usually my default method but I didn't get round to finding a model for the 4046. Our very own alec_t has supplied one however so I'll play with that later. Going to be slowwww...

Yeah, much too slow. Quicker to start changing parts around on the real thing!
The amp works at least, although the LPF is dropping more signal from the oscillator than it should. I might just put in a ferrite bead instead.
Almost ready to make the main PCB now. It will be my first attempt at through hole plating!