Oscillators -waveguide

Hello Guys,
I am considering building some high frequency oscillators.
The plan is to use waveguide cavity and a negative resistance device, probably a Gunn.

The frequency I have in mind is somewhere in the W-band ie (75 to 110 GHz), so fairly high.
I have started working out waveguide sizes and impedance transformations to match the device
to the waveguide, the initial frequencies are 90 GHz and 100 GHz. These would form the basis of a terahertz chain.

Has anybody got some advice on practical consideration ie plating and machining or material for the waveguide cavity ? There is very little information around ! Any tips advice appreciated.

It seems to me that your knowledge and experience is not enough to start such demanding task.
1. You should know that above ~30 GHz, all Gunns operate as harmonic oscillators, so at W-band, 3rd harmonic is used.
2. If you intend to generate THz signals, using a Gunn is not good for stable frequency. A better approach is to multiply from say 20-30 GHz using active frequency multipliers.
3. Instead of starting to learn how to build a W-band Gunn oscillator, please consider to purchase either such Gunn or the active multiplier which can give you a stable and professional THz source.

Please look at "www.spaceklabs.com". Here I worked for 18 years and we developed similar devices for up to 110 GHz that are now operated as base generators for THz users, including wideband spectroscopy. Spacek Labs.Inc. can offer you a TEchnical proposal to your specifications, with a quote.

Hello Jiripolivka,
Thank you for the reply and information.
I am aware of spacek, millitec, quinstar, wisewave etc.

1)You are correct about harmonic operation , however this is typically above 50 Ghz from what I have researched in the literature, particularly for 90 & 100 GHz generation.
2) again you are correct if stability is an issue, however noise for multiplied source will be higher 20Log(N).
also the module become more complex ,expensive and greater opportunity for problems.
3) I have looked into purchasing Gunn oscillator & multiplier , but these are a few thousand S, so expensive .

Currently I have built a 100 GHz module , providing in excess of +13dBm, using commercial Gunn device, I used a aluminum cavity to keep costs down. The stability is not to bad. My intention is now to incorporate a varactor to provide rapid
frequency adjustment,which could be used to track the frequency drift or I may attach a heater.
The initial vco is providing ~ +12dBm at 100 GHz with 1.5 GHz bandwidth, but I have seen 3.5 GHz with +11dBm

The whole process has taken about 6 months & a lot of effort.
I was trying to seek some advice , perhaps I may not have as much experience as others in this field including the guys at Spacek (George), however there is progress although slow but positive.

Thanks

Okay, so you have some good experience, congratulations! In my opinion, for W-band it is quite difficult to find a good theory. MOst Gunn designers use their experience to get good results. The cost then corresponds to the effort spent to make a good design.
I have not seen a wide-tuned W-band oscillator using a varactor. Such varactors are rare and expensive and power usually varies a lot when tuning. Mechanical tuners are better.

I have designed varactor multipliers for one frequency or a narrow band. It was tough varactors had to be selected from a dozen units. Now everybody makes wideband amplified frequency multipliers.
Oscillator noise is a problem, too. I was quite surprised by Elmika BWO for W-band which was noiseless even compared to a Gunn.

To generate THz signals by multipliers, noise will remain a problem even if you started with a W-band Gunn. The other design by IR laser beating on LiNBo etc. promises some power and recently I saw papers promising a low noise, too.

well, gunn diodes are tricky beasts. first, they like to run at ONE temperature, not a range of temperatures, so consider heating the whole structure up. Otherwise, they might jump modes.

Last I looked, gunn diode itself (doping, cacitance,etc) roughly determines the operating frequency, and the cavity just tunes it a little. So you really need to pick the gunn carefully. Ideally, the gunn diode manufacturer will have a "test cavity" or reference design u can copy/modify. Basically u need something that is 1) resonant at the right frequency, and 2) has the ability to vary the coupling of cavity to outside world.

you might get a lot of gunns that will simply not work in a cavity. you would want access to a different lot of gunns in that case.

hello Biff44,
Yes holding the temperature of the cavity steady is a good idea. e.g. heater at +55C or so.

What you are suggesting about the device parameters is interesting , there is also the parasitic and packaging. A colleague suggested that most devices are assembled semi-automatic so subtle differences. Consequently different behaviors in a cavity.

I have purchased a few devices , however the supplier does not provide detail of cavity, just DC conditions & RF performance. So I am having to establish the circuit detail to 'match' the device, or rather taming the device.
As suggested by you I have estimated cavity dimensions and installed devices and generally they are now behaving reasonably well. Coupling to the external world or external Qext is proving a little difficult to implement, having said that since one is operating at 100 GHz the fundamental mode is heavily decoupled so Qext is high, the frequency is not "pulling"
by load variations ? The down side is it is a little more difficult to adjust the frequency to take the device variation, I think
As mentioned previously , I now have the circuit working with about +10dBm/20mW using a waveguide power sensor( expensive). From the 5 devices I have , all have worked around 100 GHz +/- 300 GHz , with careful adjustment of matching elemnents I can get all to work at 100 GHz with power 18 to 25mW, power is difficult to predict & control.

Recently I installed a varactor device , a tricky process, I managed to get 1.5 to 3.5 GHz with reasonable power10 to 18 mW. One arrangement gave me about 8 GHz with about 9 mW. I am now investigating the coupling mechanism & controlling it to get a specific electronic bandwidth, again power is difficult to predict so frustrating.
I think the power levels are sufficient to drive a amplifier multiplier( x3) chain to get to say 300 GHz, but I could be wrong.

The noise around 100 GHz seems to be about -75dBc/Hz at 100 KHz, not sure if this is good enough but how can I improve it? maybe I change waveguide dimensions to provide more volume so higher Q whilst ensuring the cut of frequency is ok to trap the fundamental

All interesting but confusing and frustrating, however when these little demon devices work they work well.

If i gold plated the cavities would I reduce losses and in return get more power ?
plating is expensive though..... all confusing.

Concerning the "expensive" power meter, try a PIR sensor used in alarms, or a low-cost IR thermometer with such sensor!
I tested my Omega and other pocket thermometers: above 80 GHz (PIR win dow) at +6...+13 dBm was nicely indicated as a temperature increase to 40-80 deg.C.

Gold plating is actually worse than polished copper or aluminum surface. Gunn output power is a compromise of Gunn itself running at 3rd harmonic,and output coupling. A high-pass (wr-8) can reject the 2nd harmonic if it is a problem.

Also beware, I tested six power heads for W-band against one excellent BWO (Elmika generator, Russian BWO), the differences were >6 dB mainly over 100 GHz.

Phase noise of a Gunn usually has no effect on low-noise mixers, also measuring it is difficult. Multipliers are often not worse even by scaling from ~20 GHz but as their amplifiers are wideband, thermal noise grows and filtering is needed.

Consult Virginia Diodes, they make multipliers to THz range and can offer some experience.