HP8444A cavity oscillator

[G0HZU_JMR]

To show what I mean (in practice) I put together a basic oscillator this evening using a small coaxial resonator. I have a fair bit of experience with this type of oscillator. I designed it for about 1500MHz and it ended up at 1490MHz. I used a fairly exotic transmission line to make the resonator and it has a solid centre that is silver plated like a mirror and the dielectric is PTFE. I used ATC porcelain caps to make up the rest of the resonator.

I used a BFR91 BJT because that's what I'm used to and I have good models for it. I would have preferred to have used a physically fatter resonator down at 1500MHz but this resonator type has worked well up to several GHz in the past.

During the design phase I experimented with the impact of having shunt capacitance at the common node near the BJT collector on both the simulator and the real oscillator. Even 0.2pF really degraded the performance and it only took a 0.7pF ATC SMD cap here to totally prevent the oscillator being controlled by the lightly coupled 1500MHz resonator.

The 0.7pF and the BFR91 then formed the new resonator up near 2GHz and the coaxial resonator was effectively redundant.

See below for a video showing how stable it is over two minutes.

https://www.youtube.com/watch?v=u-PhCMxB8_I

There is very little drift across the 2 minutes of the video. The short term stability is very good and the phase noise was excellent. When tested with tighter coupling to the pickup probe the phase noise looks to be better than -110dBc/Hz at 10kHz offset which is similar to a high end spectrum analyser at this offset. It's not in a box and it's run from a fairly decent PSU but some of the wobble is microphony from me moving and possibly some is from the PSU drift. It's still a good result although I think I can improve on this with a better resonator.

The HP cavity oscillator should be better than this, especially if tested across a temperature gradient. However, I found that adding shunt capacitance at the common node near the collector of the BFR91 caused the oscillator to oscillate way off frequency with a new mode of oscillation.

[G0HZU_JMR]

If you have access to various SMD caps in the range 0.3pF to 0.8pF you could also try reducing the emitter capacitor. This should achieve two things. It should increase the self resonance frequency of the BJT and it should also reduce the amount of negative resistance produced by the BJT.

What I found when I added the unwelcome shunt capacitance was that I had to increase the coupling capacitance slightly to the resonator (which spoils the Q) and I also found that I had to start adding capacitance values up near 1pF at the emitter just like HP have done. Otherwise, the 1500MHz oscillation would stop. If I kept on adding shunt capacitance at the collector up to 0.7pF then the oscillator then came back to life but up at just under 2GHz.

It could well be that the 1pF cap at the emitter is a select on test part where the technician tweaks a few things to get the oscillator to start up reliably at 1500MHz.

[Alan Bain]

Thanks to all the amazing helpful information here (huge thanks to G0HZU!) it works again. I had a visit to a hospital for some unfortunate human diagnostics which meant it took me longer to report.

I rigged up a variable PSU and I could see it switching oscillation modes as the voltage increased from one around the cavity at supply below 12V jumping up to the 1.84GHz one (which was very powerful) at positive rail > 12.6V (using the 8558 to observe!), presumably as the transistor generated more negative impedance at higher bias currents.

Back to the Smith chart (lacking the excellent looking simulation software shown in posts above) I realised I had transformed transistor CE S params into CB ones assuming no base inductance to ground. I wondered if the combination of C1 and the base inductance could form a series resonant circuit, so I moved the ground end of C1 right onto the base lead to reduce the inductance and low and behold it worked properly at 1.55GHz. [I should add that I had previously replaced as I found, but as found it was dead - and potentially someone had been in there!]

Power is about 3db low on the manual spec at around +5dbm - but that could just mean my HP432 and 8478B are out since the latter doesn't even have the little calibration factor graph and given the pile of adaptors I used to connect SMC into N!

[G0HZU_JMR]

Wow that's a good result! I'm glad it worked out OK. I was beginning to worry that something bad had happened to part of the cavity assembly.

I saw similar behaviour on my 1.49GHz tline oscillator if I added an extra shunt path at the collector using a small SMD cap. If the shunt cap value was quite small I could get the oscillator to swap modes between 1.49GHz and just under 2GHz by playing with the collector voltage just like you found. If the shunt capacitance path at the collector was increased it preferred to stay up at 2GHz.

It does look to be a very critical circuit in terms of component placement after repair. Well done for getting it going again!