RF 'echoes'

For the design and construction details of ion guns, necessary for more advanced designs and lower vacuums.
Linda Haile
Posts: 230
Joined: Mon Jan 04, 2010 9:28 pm
Real name:

RF 'echoes'

Post by Linda Haile » Sat Jan 23, 2010 10:03 pm

I hope I've posted this in the correct forum. Can anyone throw any light on the mechanism by which so called automatch units virtually eliminate RF echoes by tricking the RF supply into thinking there is a 50 ohm impedance in the circuit and reducing the RF echo to between two and four percent of the signal? This is something I've come accross during my research into ion sources but I've not been able to find out much about it. Thanks.

User avatar
Doug Coulter
Posts: 1312
Joined: Sun May 27, 2007 7:18 pm
Real name: Doug Coulter
Location: Floyd, VA, USA
Contact:

Re: RF 'echoes'

Post by Doug Coulter » Sat Jan 23, 2010 10:42 pm

The real trouble is that I can easily say more than this forum can hold on that subject. So I will try for brief here.

Most of these can look at forward and reflected power via a mechanism you can find in most older ARRL handbooks -- that part is easy if you know the range of power and frequency you are going to be operating in. Some fancier ones can see phase of volts vs current as well. Most pro ion source RF units or auto tuners have this in some form.

They also have what amounts to a variable transformer, only it's usually done the same as in a ham transmitter, as a Pi network. This usually has a variable cap to ground on the input, one on the output, and a coil in series. By adjusting the caps you can get to a place where you have it all resonant at the desired frequency, while having a given ratio of the two capacitors, which is what determines the impedance ratio input to output. They do this with some opamps looking at the power sensors, and little motors on the caps to change them. I have here a unit designed for 3kw at 13.59 mhz, it's fairly impressive. It also looks at the final tube plate current so it can see the dip there at resonance.

This tech is from WWII in figher planes and such, a lot of it was done by Collins, then translated to the less advanced physics world for guys who don't "get" RF. In some of the fancier avionics, they used a motor driven rotary inductor to get resonance, as the basic impedance ratios were well known to match the amp to the antenna from first principles, so on the inductor needed tuning when the frequency changed. You can find this kind of thing at many hamfests if you know what to look for.
Why guess when you can know? Measure!

Linda Haile
Posts: 230
Joined: Mon Jan 04, 2010 9:28 pm
Real name:

Re: RF 'echoes'

Post by Linda Haile » Sat Jan 23, 2010 10:53 pm

I can follow the part about tuning the two variable caps to minimize the current, but where do you start when designing an ion source? Assuming a fixed frequency.

User avatar
Doug Coulter
Posts: 1312
Joined: Sun May 27, 2007 7:18 pm
Real name: Doug Coulter
Location: Floyd, VA, USA
Contact:

Re: RF 'echoes'

Post by Doug Coulter » Sat Jan 23, 2010 11:08 pm

Well, in this case I have a bit of an axe to grind. This thread details what *I* did myself, and I'm still very happy with it.

viewtopic.php?f=12&t=5022#p32330

If you want to use more power and lower frequency for some reason, I think Carl had something nice worked up and maybe he'll chime in with a link to that version.

Due some very careful use of overdesign, my thing doesn't care about the issues of impedance matching so much, it just works if duplicated as described. So there's a start, all you have to do is make that one, or Carl's. Or for that matter, the one Andrew Seltzman did, which I've not tried here yet, as mine gave no indication I needed improvement , but it looks good (and nicely simple if you're a master machinist) too.
Why guess when you can know? Measure!

Linda Haile
Posts: 230
Joined: Mon Jan 04, 2010 9:28 pm
Real name:

Re: RF 'echoes'

Post by Linda Haile » Sat Jan 23, 2010 11:19 pm

I have read those threads but found no mention of eliminating echoes. I was just thinking that in the interests of improving efficiency it may be advantageous to address this issue.

User avatar
Doug Coulter
Posts: 1312
Joined: Sun May 27, 2007 7:18 pm
Real name: Doug Coulter
Location: Floyd, VA, USA
Contact:

Re: RF 'echoes'

Post by Doug Coulter » Sat Jan 23, 2010 11:36 pm

It's not so much efficiency in this part of things, it's making the power tube or coax and connectors live, and things not arc over where you don't want them to. This is all very basic transmission line concept stuff, in all basic books that exist that mention transmission lines (tons of them but check ARRL handbooks from the earlier days for a lot of detail, or any "radio engineer's handbook" whatever).

So, if you want to know more, do a little bit of homework, this stuff is very easy to find cheap or free, but you do have to go for it at all (hint, find used bookstores). You will almost certainly find something of interest here:

http://www.pmillett.com/tecnical_books_online.htm

(and so should a lot of other people here, and that's just one link of many).

Here's one on standing wave ratio, the standard term for what you're calling echoes. Both are actually correct, depending on what you are concentrating about thinking on.

http://en.wikipedia.org/wiki/Standing_wave_ratio


I prefer Fredrick Termans Radio Engineer's hanbook, 1943, over all others so far, but the ARRL stuff is also good if you get one from the years when hams were into building things themselves.
We just found a couple of those Termans by searching online, and got another one for about $15, and it's a piece of solid gold for anyone needing to learn the basics and even some pretty advanced things. I would pay far more for one and be very happy.

For example, it covers particle beam optics (electrons, but you can do the scaling for fusor ions), along with "what is an inductor", and so forth. From how much resistance is there on some size wire, to parasitic capacity to getting good figure of merit in a video amp, to complex filter design basics, it's all there.

The reason some lower frequency ion sources (particularly for sputtering at decent power levels) use an auto tuner rather than just one manual adjustment is partly because the load changes during use as the thing can be lit up, not lit, or the gas pressure etc may change so the load isn't constant.
The other reason is that most people out there can't actually handle two interacting adjustments if their life depended on it, and will fry things as they spend a lot of time in the wrong settings and the power tube, already near limits is burning up. That's why a lot of ham transmitters have a "tune" setting that reduces the power while someone not too good at it learns how.

The reason I don't mention SWR in my design, is that with a tube designed for 600 watts plus output, running at 20 watts DC input into high voltage hard-line coax, well, even if all the unused energy reflects back into the driver, no big deal at all -- the tube gets very slightly warmer (after all the filament alone is 30w and is pure heat before any DC comes in) -- and anyway, I have low SWR because I designed it right with that in mind, the coax is a 1/4 wave matching stub -- it's all taken care of already. The only impedance matching issue in that one is not having too high an impedance in the cavity feed point, so things don't arc in there at the desired power input levels and eventually melt solder joints in the cavity. I think Andrew's version takes the cake in low power in vs ions out, not so sure it makes the best ratio in monatomic vs diatomic ions though -- because mine worked fine, and I had other things to move on to at that point, problems all solved there.
Why guess when you can know? Measure!

Linda Haile
Posts: 230
Joined: Mon Jan 04, 2010 9:28 pm
Real name:

Re: RF 'echoes'

Post by Linda Haile » Sun Jan 24, 2010 9:52 am

I've come accross SWR meters. It's starting to make sense now. Surely the reflected energy is wasted, though? Thanks for the link, it looks very interesting from what I've seen so far.

Starfire
Posts: 1482
Joined: Wed Oct 24, 2001 6:14 pm
Real name:

Re: RF 'echoes'

Post by Starfire » Sun Jan 24, 2010 2:31 pm

Great link Doug - thanks

User avatar
Doug Coulter
Posts: 1312
Joined: Sun May 27, 2007 7:18 pm
Real name: Doug Coulter
Location: Floyd, VA, USA
Contact:

Re: RF 'echoes'

Post by Doug Coulter » Sun Jan 24, 2010 5:18 pm

The reflected energy is often wasted, but not always. In "echo thinking" it's used to help switch the lines in all PCI bus applications, actually -- most of us own computers that have that bus, and I will refrain from editorializing too much on other aspects of that which I don't like so much. The bus logic levels are only reached after the reflection off the unterminated end comes back and adds to the original drive....a fairly cool idea that saves power in the drivers. What I don't like is the latency of the design for master sharing and burst initiation, and the patent/licensing lockout that keeps small companies from being able to innovate and make products for this bus that are competitive in price. Big step backwards from ISA on that one. Put a company of mine out of business (would have more than tripled the product parts and license costs to convert the design), so I guess I do have the right to whine a bit.

The echo model is mostly used for broadband signals like pulses and sq waves, and is the best way to think usually when those are what you're working with. Motorola's old ECL handbook has a ton of good stuff on that, as most digital engineers then and now don't understand this set of phenomena.
That one is worth picking up if you find one and are interested in more, the explanations in there are very clear.

When you're thinking in impedance mismatch (SWR) kinds of terms you are most often working with a single frequency (narrow line on a spectra or FFT). As some of those links show, this will result in peaks and dips of voltage and current on a transmission line at lengths corresponding to 1/4 or 1/2 wavelengths of the frequency on that line (which will usually have a velocity factor less than C).
The peaks in voltage in a standing wave can cause arcing and increased dielectric losses. The peaks in current can cause additional I^2 * R losses. This can be a curse or a blessing, depending. For example if reflected energy is a problem for your source, and you have plenty source power, you might just want some loss in a long piece of coax to attenuate that reflected energy.

If tuned for proper match and resonance, reflected power to an RF source need not be wasted, though it will add a tiny amount due to the things mentioned above. The key here is that an arbitrary length of say 50 ohm line that has a not-50 ohm load on one end, will not present an impedance of 50 ohms to the other end either, but whether it's higher or lower than nominal depends on line length, as well as the full complex impedance presented to both ends of the line, not just the resistive component of the source and load.

For example, a 1/4 wave line shorted at one end has infinite impedance at the other, and this kind of thing is the basis for many cavity resonators and waveguides. A 1/2 wave line shows the same impedance at both ends no matter what impedance the coax is! (This is all disregarding coax losses, which can indeed be so low as to not matter). Multiples of 1/2 wave lines have this property as well, but will have peaks and nodes along the length were either current or voltage peak, which can cause more loss or arcing. Odd multiples of a 1/4 wave line act like a 1/4 wave line -- think of it as a quarter wave line in series with a half wave line (or multiple), and it's obvious.

Best of all, if a quarter wave line has a characteristic impedance that is the square root of the multiplied input and output impedances, the match is then perfect -- it's a transformer, not reactive at that one frequency. Sweet trick for narrowband applications.
This is what I did in my uWave ion source to match the tube output to the cavity input, and that's why I took some pains to publish the right dimensions for all that -- small changes affect this a lot.

Lines slightly shorter or longer than 1/4 wave with one end shorted act like high Q impedances either inductive or capacitive, depending on which direction the error is in. For a line a little on the short side, a variable capacitor can be tied across the high impedance end to tune it to a frequency desired, which is much easier to do than change the length in some construction methods. This effect is also used in my uWave ion source in the main cavity. In that, the tubing with gas in it acts like a capacitive load due to the dielectric constant of the tubing itself, with a resistor (a non linear one) due to the energy pumped into the gas to ionize it. So in that case, the load seen there is a capacitor in parallel with a resistor that changes with gas conditions, and in some cases this can require re tuning for both cavity resonance and gas loading. This particular design is overkill enough to not need that, though.

The big risk in bad impedance matching is having it happen with a source that is operating within an inch of its life when matched, seeing higher voltages or currents due to the mismatch, and getting fried. The answer there is of course to go ahead and match things up. Rarely is it a good idea to brute force by having a larger output device in the source as regards efficiency. The larger device will be harder to drive (more input drive power is a waste) will be slower (transit time admittances), and will usually draw more power for the same output RF.

I get away with this in my source design because I already have a magnetron designed for high power dissipation, but in CW mode have to run it well below its normal ratings anyway, for other reasons such as space charge accumulation in a tube designed only for pulsing. So there's plenty of ways to get rid of any excess heat, it was designed to dissipate far more than it does in that design even with 100% reflection of power when there's no load.

This power level is in the noise for a conventional fusor which is using 400w of DC HV input power anyway, and only would become significant when a much higher Q mode is happening than that.
Eg, it's smarter to work on the 400w of almost total losses first, then worry that last couple tens of watts, don't you think?

In the new pulse mode I'm running in, where the HV power is in the 5 watt region, yes, the ion source is using more power than anything else is, for now. I am also ditching about 99% or more of the resulting ions by not having enough ion extraction field to pull them out of the source against the curve of the magnetic field lines used to get ECR effects in it -- because I'm right now more interested in that high Q pulse mode and improving that further than in the total system power (which would include vacuum pumps and all the rest as well, a bigger factor yet, especially for those using mech and diffusion pumps running full out all the time, vs a 20w average turbo system that only runs the fore-pump intermittently). It is kind of like having a 500 hp car to go to church with -- you just don't push on the gas as hard in that mode, vs on a racetrack, and accept that it's nice to have that there is you need it, wasting a little gas to get to church. Once I establish that this vehicle is only going to church, I can substitute my ~5 hp go kart and save some gasoline, but I can't take that on the track or on the roads (legally ;~) for long distance travel either. If you could own just one car, well, the choice is fairly obvious, I think.

Now when making ions as the main task, like for sputtering or something like that, the efficiency becomes more important to be sure, but for a fusor, it's not that big a deal. For now, the main worries are not melting things, and you can tolerate some losses. You can always optimize post facto once the other parts are nailed down.

As for figuring out how to make an RF ion source -- the trouble there is that the effective load at the ionizer is not often calculated until after the fact -- easier to just lash it up and measure that. Then you can do matching all day long with some confidence to tweak things. And even know how to size the RF source correctly, which will improve efficiency further in most cases.

Premature optimization is often a waste of time on things like this (this is a famous saying in software).
At first, you don't even know what's going to be most important, and wasting a little so as to be able to evaluate the system as a whole first is most often a wiser move.

Hope I don't get too much crap about posting this link and picture here....

Go kart:

http://www.coultersmithing.com/kart.html

and see pic below for when I'm not going to church, though it gets used for that too.
Attachments
coolplates.jpg
coolplates.jpg (68.91 KiB) Viewed 2824 times
Why guess when you can know? Measure!

Linda Haile
Posts: 230
Joined: Mon Jan 04, 2010 9:28 pm
Real name:

Re: RF 'echoes'

Post by Linda Haile » Sun Jan 24, 2010 6:32 pm

Some interesting background there, Doug. You are correct in that it was while reading up on the type of ion sources used for sputtering/deposition that I came accross the 'echoes', thanks for pointing out that this is the same as SWR losses. While I appreciate your comments relating to power levels associated with normal fusor use (and ion sources are not usually required here), I have been considering ion sources for high gas feed/extraction rates as a way to keep the ratio of fusable ions/neutrals at a maximum and also maximize the ratio of monatomic ions/diatomic ions. Again, it may be simpler/more effective to just increase the power input but there comes a point where this becomes less practical than increasing efficiency.

Continuing your analogy, if you were to travel to church in a 1000HP car, you may have problems controlling all that power, especially in the wet, and could wind up wiping out some of the congregation that walk to church, which would not be a very desirable outcome, to say the least.

Post Reply