Help needed to design ~1Ghz high power Oscillator. 2

[ElectLect]

I am new to this forum, but was suggested it for my question.

I am trying to design an oscillator capable of generating ~1Ghz fixed frequency signal. It needs to be a very high current capability, but will not be modulated with any signal (ie fixed pitch and amplitude). It doesn't even need to be a perfect sinewave, as long as it has only odd harmonics.

Don't worry about RF contamination, the circuit will drive a completely enclosed system, so no RF will be emitted. Most likely the impedance will also be very low. I am doing this as part of a Post Grad research project into materials property testing, so am not sure who to ask.

Like everything in the (underfunded) research establishment, it has to be developed on a shoestring. Any of you more experianced electronics guys know how I should go about it?

Graham

 

[logbook]

An oscillator is an amplifier with a selective feedback path to define the frequency and an amplitude limiting stage to define the amplitude. You therefore need an active device with power gain at 1GHz.

A "very high current capability" is a meaningless statement. Gives us a number. In power terms, do you want 1mW, 1W or 1MW? Then what is the load impedance? One would typically have a load of 50 ohms but you are saying it will be "very low". This impedance is going to be crucial to the current output since you will probably need some sort of reactive matching network.

RF devices will be characterised for 50 ohm operation, so it is probably going to be easier to design the oscillator to drive a defined 50 ohm load then put on a matching network to change the impedance to match the load. Given that the system is running at a fixed frequency, the bandwidth of the matching network should not be a problem.

If you can say you want 100mA into 1 ohm, for example, that gives a good starting point.

 

[ElectLect]

Well, the application is the subject of research and may lead to a patent for industrial application. It is a non-destructive testing technique for various types of non-porous weld. As I say, it will be an enclosed system, so impedance is likely to be very low. I am interested in inducing very high currents in the test piece, so the oscillator will only be used to sustain the currents long enough to map the field.

The current required, and this is the hard part, may well need to be as high as a kiloAmp. Since the impedance will be very low (although may be reactive), there is no power requirement as such. Ideally I would like the system to be efficient, so have already ruled out Cavitron type devices.
I have some familiarity with Power Mosfets, and hoped I might be able to use a number of these in parallel.

Since the frequency will be fixed, resonant techniques are fine. I did have some papers by Peter Baxandall, who had similar ideas using bipolar transistors for UHF oscillators, but seem to have misplaced them. I realise this is a difficult challenge, but my topic is as much to determine the feasibilty of the system as to develop it.

At this stage I am as open to scathing rebuttals as I am helpful suggestions...

 

[logbook]

I think the power is going to rather higher than you might like.

Let’s take 1000A and 1milliohm. That’s 1kW.
1kW at 1GHz sounds like a magnetron to me.

… but wait a minute. 1GHz currents are not going to penetrate a metal very far.

9mm at 50Hz
0.002mm (2 microns) at 1GHz.

… assuming non-ferrous metal. And I am not sure if anyone has a good handle on the relative permeability of ferrous materials at GHz frequencies.

I was going to say that eddy current NDT was done at lower frequencies (100Hz to 100kHz) but I just found a paper (dated 1981) where the guys (Auld & Winslow) were using 1GHz frequencies and a ferromagnetic resonance probe to look at surface cracks.

 

[ElectLect]

Thanks for your feedback so far, logbook.

Agreed about the 1kW estimate. I was figuring about that. Eventually i will need higher still! The thing i don't like about magnetrons is that they are only 50% efficient. I was hoping to use a transistor in switch mode to minimise heat dissapation in the unit itself. Seeing as it is not modulated, i hoped this would help - doesn't even need to be a perfect sine wave.

This is really the crux of my question. Is there any way to design a solid state device capable of oscillating at these power levels and frequencies? This is a very tall order, but I can't believe that there isn't some way of doing it. Maybe there is, at least, a cutoff frequency that perhaps I could go down to to improve my chances.

Another related question is how can i minimise skin effect in a conductor at these freqs? I was thinking along the lines of a cross section of concentric copper rings, laminated in such a way as to insulate each concentric layer. The idea being to stop the charge migration to the outer layer. Each could be 2um foil - although it could be a manufacturing pain. There would be a high hall effect voltage across each layer, but the whole section would still be conducting.

Any ideas?

 

[logbook]

I don’t see how saturated switching is going to be possible at these frequencies. Typically RF devices quote ft and not turn on & off times. The older saturated switching transistors were of the order of several nanoseconds as I recall. SiGe would be the best but the currently available parts are too weedy (<200mW from Infineon). In Silicon the biggest Infineon offers is 2W. That would take an awful lot to get up to the power levels you are talking about, even if they were able to switch fully on and off, which I doubt. Given the low ft (5GHz) the base current requirement would not be trivial.

People do make kW Rf transmitters using semiconductors rather than valves, but the frequency is much much lower. The efficiency is good though.

I would think your best bet using semiconductors would be an emitter coupled switch with push-pull transformer coupling from both collectors. This should be fast and more efficient than a Class A stage.

Can you force a current to flow down the inner part of a coaxial cylinder if you laminate coaxially with coaxial insulators? I wouldn’t like to say definitively no, but I certainly think the answer is no. One of the earliest experiments on skin effect was passing a current through a gun barrel with a wire running down the middle of the gun barrel. "No current" went down the wire. It was a crude experiment and open to criticism but there have been other similar ones. I think it has been done with wire cages, producing a similar result.

 

[ElectLect]

Thanks for your help logbook - you really are making a big dent in my thought processes!

"People do make kW Rf transmitters using semiconductors rather than valves, but the frequency is much much lower. The efficiency is good though."

Maybe this would be a better starting point for me. I had a Baxandall circuit for an emitter coupled switch with push-pull transformer coupling from both collectors, but I have lost the paper. I'm quite used to machining my own inductor parts so I like this approach. Recommend any books/sites on the subject? - Circuit description and theory ideally. I know I'm going to have to do a lot of reading and experimentation...

"One of the earliest experiments on skin effect was passing a current through a gun barrel with a wire running down the middle of the gun barrel. "No current" went down the wire."

With a travelling wave I can definately see this, since current will escape to outside wherever the inner and outer conductors contact. I was thinking more along the lines of having the transistors feed each coaxial cylinder independantly, so no chance existed for hall effect to push current to outside conductor. Alternately if the line had a standing wave then as long as inner and outer conductors contacted at voltage nodes only (ie no current) then again hall effect voltage would not exist, so no skin effect current.

I appreciate that this is all a brute force approach to persuading the electrons to go where I want them. What can I say, I'm a mean guy to electrons! ;-)

 

[VEBill]

Can you use 2.45 GHz instead of 1.0 GHz ? The advantages include easier licensing (ISM frequency) and readily available and incredibly cheap 1kw sources (microwave ovens).

 

[logbook]

Well ElectLect, you are making my brain creak! So we now have two isolated circuits driving individual isolated coaxial conductors. What happens? I couldn't think of this in terms of solving Maxwell's equations, but it occurs to me that we now have a coaxial cable. Whether this is an inapplicable quasi-static approximation to the problem is not yet clear.

I believe your association of the Hall Effect and the skin effect to be entirely spurious. The Hall coefficient in copper is 5 orders of magnitude smaller than in Silicon, for example, and the Hall effect in Silicon is already pretty weak. The Hall Effect also does not increase with the square root of frequency, if at all.

In a coaxial cable the inner is coupled very tightly to the outer by the mutual inductance, the mutual inductance being equal to the self inductance of the inner conductor. If I first energise the inner circuit with an AC signal the resulting voltage drop from end to end will depend on the product of current and impedance. At high enough frequencies the reactive portion must ‘win’. I think the self inductance is relatively constant so the reactive drop increases linearly with frequency. The resistive drop is only increasing as the square root of frequency due to the skin effect.

Most of the drop is in the self inductance, but this couples to the outer conductor due to the mutual inductance acting as a 1:1 transformer. There is therefore the same volt drop in the outer conductor, both magnitude and phase. When I now connect up my second circuit, having a generator with the same magnitude and phase as the first generator, which circuit wins? Well the outer conductor has less self inductance and less resistance, although the difference is relatively small.

I started off with a current creating a voltage, now I prefer a voltage creating a current. The current that would be created in the outer is greater than the current that would be created in the inner if the circuits were only present one at a time. If the outer current flows, the volt drop created across the mutual inductance is then greater than that which would have been created by the inner current, and I have already argued that most of the volt drop is reactive. I think the inner circuit impedance is therefore bootstrapped, increasing the inner impedance by orders of magnitude. No current will therefor be allowed down the inner conductor; it has a built-in generator backing off any applied signal.

 

[zeitghost]

The easiest option must be the magnetron.

50% efficient or not, they are pretty robust, a lot more so than any semiconductor you are likely to find.

To obtain the equivalent amount of power from semiconductors, it appears to be necessary to use many low power (100W?) stages and use combiners to sum the outputs.

By the time you take the losses in the combiners into account, I'll bet my left leg that there isn't much difference in efficiency compared with the magnetron.

Also 2.4GHz magnetrons are cheap, which is not something you can accuse power RF silicon of being.

See if you can find anything on the web about semiconductor radar transmitters.

rgds
Zeit.

 

[ElectLect]

"Can you use 2.45 GHz instead of 1.0 GHz?"

I don't see why not, but my concern here would be the safety implication of using microwave frequencies. Is anything in the GHz range dangerous or does 2.45GHz require extra precations?

How do antannae work at these frequencies? I have never been quite sure how microwaves are launched. I may need to study Maxwells equations in greater detail.

"I believe your association of the Hall Effect and the skin effect to be entirely spurious...
...No current will therefor be allowed down the inner conductor..."

Well explained, i see your point. So any conductor may as well be hollow then. To force current down a thin central wire requires infinite voltage. This is bad news, I may have to consider superconductors too - yuk! Is there any way to maximise the conducting area? This could well force me down to lower frequencies.

OK, howabout oscillators? I'm still interested in the emitter coupled switch with push-pull transformer coupling from both collectors, you mentioned. Got any circuit details for this? I would still like to read up more on this topic. This too will limit the frequencies I can play with. Maybe I need to think about lower frequencies first.

 

[ElectLect]

Thinking a little more about this, I may be able to use magnetrons. They are the most cost effective option. I'll need to do some serious reading on design an application. In particular I need to understand in some detail how the waves are launched into any kind of conduit.

Any suggestions? Books etc...

 

[VEBill]

There isn't much difference in danger between 1 GHz and 2.45 GHz - both are equally dangerous at 1kw level. Do a Google search on "Safety Code 6" to find an excellent reference on human safety w.r.t. RF.

For this sort of application you'll likely be forced to stay on ISM frequencies. Even with very good shielding, a kilowatt would have the potential to cause plenty of illegal interference to other communications users. It is a bad assumption that you don't have to worry about frequency assignments.

You shouldn't have to go back to Maxwell's Equations to independantly derive antenna design. I think that you'd benefit from a more recent and more practical reference. Search on ARRL and RSGB (for examples) and have a look at their quite practical guides to microwave.

Here's another good read on the subject of patents:

http://www.tinaja.com/glib/casagpat.pdf

 

[GOTWW]

Another option not discussed is a resonant loop of waveguide using circulators and a low powered source. The low power CW source circulates in the ring, accumlating power. I have seen this for testing components at power levels not available, the actual numbers were quite impressive. It has been a while and I do not remember the details, but it is real, no bull.

 

[ElectLect]

Agree about the patent thing VE1BLL. Thanks for the refs.

Interesting GOTWW. Got any papers links, books etc?

 

[GOTWW]

I started to look at the IEEE periodicals site, but apparently my sub has expired, I know there is some ref's out there.

The one I saw was (millimeterwave) many wavelengths around. (I believe) It was feed from one port of a circulator, with the other two ports connected to the loop. The trick was I believe that the source port was coupled to the circulator input using a small iris. therefore when the signal tried to exit this port, the majority of it bounced off and continued going around the loop. I never had a chance to analyze it, and am alittle confused about the the true effect of the iris. The fact that the system is resonant must have play a part. I have seen a paper, so I know one exist. Hope someone can clear this up so we will both know.

 

[GOTWW]

http://www.lns.cornell.edu/public/ERL/2003/ERL03-15/ERL03-152.pdf

http://www.astec.ac.uk/rf/resonant ring poster.pdf

basic google search

 

[ElectLect]

Thanks GOTWW - Interesting read.

I'm very interested to know what EM simulation packages are recommended. Seems to be the norm for this sort of research now, and has to make life a lot easier than using the Maxwell equations directly - even though my vect calc is ok...

 

[GOTWW]

Ansoft and HP HFSS is pretty much the industry standards for waveguide systems, but the license and price are a bit steep. There are others, but am really only familiar with HFSS

 

[Comcokid]

By choosing 2.45 GHz, you have made your application very similar to many other high power industrial microwave applications. Do a search on "Industrial Microwave Heating". You will find many companies that do development of high power microwave systems for non-communication purposes.