Unstable Gunn oscillator

Au contraire, Rene. ISO only requires that you do what you say you are going to do. So the junk pile provides junk. As long as that is properly documented, you are in fine shape !

Some possibilities:

-corrosion or poor contact in tuning mechanism ?

-I've run across Gunn oscillators with absorber bits glued in the cavity...if the absorber comes loose lots of things can happen.

-some oscillators do seem to have odd mode-jumping in which case a small change in supply voltage might help for a given fixed frequency.

But most of my experience has been with cheap 10.5 GHz alarm Gunns...I assume you're dealing with better stuff at 35 GHz.

Steve

Are you trying to operate close to the Gunn's tuning range edges? I too only have 10.5 ghz experience, but have often had two identical units that would not tune the same. Try rotating the Gunn in its mount, look for loose cavity parts, change the tuning screw to a different metal or file its tip a bit, or try a nylon tuning screw if your close to the lower end of the range. It often times helps to look at the choke structure that the Gunn is mounted in, and the components attached. Also is anything performing current limit on the diode? Is any part of teh waveguide downstream forming a high Q cavity close to where your operating at? Steve Roberts

Both of those figures sound fairly bad at first blush.

For fun, I tried phase-locking a high-quality (relatively speaking) 10 GHz Gunn transmitter awhile back, using a relatively-narrow PLL to get an idea of the unit's broadband noise profile. I saw about -107 dBc/Hz at a 100-kHz offset (see graphs at

At 200 kHz it was perhaps 6-7 dB worse, around -100 dBc/Hz.

I'd expect these readings to degrade by about 6 dB/octave as you go up in frequency, so I'm thinking the 200 kHz reading on your 35 GHz part should be around -90 dBc/Hz rather than -70.

Do you have a large capacitor across the Gunn diode's terminals? Large electrolytics can do funny things there, due to the diode's negative- resistance properties. Easy to make an audio oscillator. I'd say try playing with different bypassing schemes, and maybe take the diode out and clean its contacts with ProGold or something like that.

-- jm

------------------------------------------------------

Note: My E-mail address has been altered to avoid spam

------------------------------------------------------

With a 2Ghz varactor tuneable Q band (34-36GHz) GUNN from the junk pile we encountered extremely high phase noise. Instead of being as stable as 100kHz with phasenoise down 70dB at 200kHz as measured with a previous one, this one jumps around 2MHz and the phase noise in no more than 50dB down farther out. The power supply is stable to below a few 10uV and the varactor supply is also stable to below 100uV. Replacing them didn't change anything. The temperature is also stabilized. Is it thinkable that an unfortunate mounting of the subparts creates unstable modes ? It could also have been dropped. Or that it would become more stable with more or less drive current ?

{ It slowly becomes apparent that the junk pile has no ISO 9000 quality control. }

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net

Thanks for the reference to your page. I wasn't aware the Gunns were that good in the X band, At 35GHz they are a bit noisier :

With the newer controller and Gunn : There are only about 220uF parallel to the Gunn, but back in the powersupply. The negative resistance applies to the microwave only, not to the DC I guess. I'm feeding 7.1V at 750mA or so and get 60mW out of it. This particular Gunn doesn't have a mechanical adjustment screw, just a varactor. and yet it does 2GHz between 5 and 22V. Yes, that scales down to 130kHz/mV. I wouldn't want to open it, the waveguide is a WR28 with just 4 by

10mm crossection

Rene

Rene:

The negative resistance applies at other frequencies. Think about it...under normal operating conditions the DC current drops as the DC voltage is increased. So John Miles' suggestion of a bypass cap right at the Gunn might help - it is common practice to use one.

If you are comparing the phase noise of a 2 GHz tuning range oscillator with the 100 MHz tuning range unit you describe on your website, part of the difference might just be due to degradation of cavity Q by the more tightly coupled varactor. You could also be getting external RF signals coupled onto the more sensitive varactor.

Steve

Looks like you figured out that I left an 'l' off the end :) -- sorry. It should have been

.

Like Steve pointed out, that's not correct. Try running your diode from a variable supply with an ammeter, and watch what happens to the current as you raise the voltage. Be careful -- you can exceed the diode's current spec by running it under its rated voltage!

Yep... add a frequency discriminator and you have a neat parametric amp!

-- jm

------------------------------------------------------

Note: My E-mail address has been altered to avoid spam

------------------------------------------------------

I wasn't aware of this. Possibly the reson we blew one. Took us 2000$ and 12 weeks.

Yes, the Q is somewhat lower but that doesn't mean there must be competing modes as in a laser. The current setup involves only two isolators, an attenuator, a PLL locking counter and a spectrum analyzer.

I'll try the suggestions, thanks.

Rene

Yes, typically there is a maximum output bias setting, usually near the rated voltage. This is normally a substantially higher voltage than the minimum required for negative resistance. It is rather hard to find experimentally as the optimum matching varies with supply voltage (partly because the frequency changes, too).

Steve

Meanwhile I reinserted the previously installed 1.2 Ohm series resistor, to no avail. I also tried a low ESR 10000uF cap, to no avail.

Does anyone has any hints as to what happens with the output power when the voltage increases and the current decreases ? There should be an optimum efficiency somewhere. Not necessarily where the noise is lowest.I guess I have to do some measurements.

Rene

Oh, the matching changes too. Could this mean that off the optimum point, the Q is getting smaller ? The specification said as powersupply range 6 to 8V, and for this particular one suggested 7.5V, while I for some reason did choose 7.1V. The matching is somewhat difficult to measure as we have two isolators in series after the Gunn.

Being confident that we do have the better measurement equipment than the junkyard : Could the optimum power supply voltage be found with a network analyzer ?

Rene

Getting a bit beyond my direct experience here, Rene, but I'll give you my gut feel, which is that it probably makes a small difference, but likely less than tuning the Gunn over the whole varactor range.

It would probably be worthwhile trying the specified 7.5V. This is likely the maximum output voltage derived by the manufacturer by testing. Though with such a wide tuning range device it might also be just some sort of compromise that ensures that the Gunn keeps oscillating over the full varactor voltage and temperature ranges. I would be surprised if the phase noise would be vastly improved by this small voltage change, but I don't have direct experience with this.

I can't see any way that it could. The Gunn must be on since its active impedance is part of the matching scenario and its output will mess up or even damage the network analyzer, I think.

In practice, to find the optimum (max. power output) bias I would start by measuring the output power with a power meter (or spectrum analyzer if power meter unavailable) while varying the supply voltage over the specified range, keeping the isolators in place. The supply current should be monitored as well, so as not to exceed the desirable power dissipation at the low end of the voltage range. You might find that you could improve the power output a bit at any single frequency by adding tuning screws or an E-H tuner in the waveguide, though I suspect that at 35 GHz the loss of the tuning device might be more than the potential increase in Gunn output available from optimizing the load impedance !

Steve