Yikes, i almost forgot about all this. Damn real life and work stuff always getting in the way of the fun things
Anyways, to (re-)clarify a few things:
- I'm not an RF engineer. It all was mostly trial-and-error, as well as simulations. So there sure is some room for improvement in that section.
- I never spent much time figuring out why the output FET dies sometimes for some people. It usually only happens when powering off the system. However, i found that not all IRF510's are the same. They vary between manufacturers.
- Yes, the RF stage isn't incredibly powerfull, using large Talon tips means slower heatup. However, it shouldn't be noticably slower than on a RFG-30 supply. At least it never was for me.
- The whole design was done to be as simple as possible, with mostly cheap off-the-shelf parts. Especially the driver chip is a big point that could be improved, i think. It just can't handle really large FET's. More powerful drivers usually have slower rise/fall times. I guess the only way to reliably get more power with a larger output FET would be to build a discrete driver stage instead of using a simple driver IC.
- This thing was never meant to be a "final" version of anything. It was just meant to get things started, to get to a usable stage, so that other can jump in and improve on that. As said, i'm not a RF guy, so i'm sure that a lot can be improved there.
- Yes, the cores in the RF transfomer and filters are important. Different manufacturers seem to have wildly different real-world parameters, although the spec sheets suggest otherwise. At least that is what i observed back then.
- The IRF510 is pretty much at it's limit in this application. However, i think it would be somewhat trivial to increase the power with a little re-design of the circuit. Right now the primary side of the RF transformer is single ended, since there is only one FET driving it. One could make that a "real" transformer with a center-tapped primary instead, the center at the +24V supply. Then just use a second driver chip and IRF510, connected to the second leg of the primary, and drive it with an inverted signal. Unless i'm thinking wrong, that should increase the overall power output. This would also mean that the main supply voltage to the RF stage can be a bit lower, decreasing the stress on the FETs. The following output filtering stages (caps and inductors) probably need some beefing-up then as well to handle the increased power. Using proper RF litz wire may be useful as well, instead of solid core copper wire.
- Regarding the RF transformer and filtering stages in general: Those were also mostly done by trial-and-error (again, because i'm not a RF guy). I tweaked things so that i got the most output power into a 50 ohms RF dummy load. That obviously does not mean that everything is tuned as good as it can. It's possible that at some point there is a big mismatch and/or power loss, just covered up by having a better tuning at a different stage. I simply lack the tools required to make useful meassurements when it comes to RF power and matching.
Having said that, it's something that someone else has to pick up. As far as i am concerned, i got that initial experiment to work, that is, to come up with something that works and is usable, and that others can take as a starting point for further development. Sorry that i couldn't solve the issue with the output FET sometimes blowing up when switching the unit off hard. Since i only ever encountered that when powering it off by switching off the supply, and never had that issue when first doing a soft "power off" through the user interface, i would guess that this could be fixed in firmware (or hardware).
Take a small recitifier, hook it up to the AC input. Or "isolate" the DC by a big diode before it goes into the smoothing caps. The idea is to get a voltage that goes away as soon as the input voltage is powered off, and not being held up by the filtering caps. Feed that though a resistor+zener combination to give a +5V signal. Put that into a free pin on the microcontroller. Then all thats left is to just hack the firmware so that it disables the driver as soon as that signal goes away (or use some logic gate to combine that with the RF_EN signal). The basic idea being that RF_EN gets de-asserted as soon as the input supply voltage is turned off, even though the input filtering caps are still charged.
Oh, and i won't be making any new batches of boards. If anyone wants boards for the current design, they have to get to Seed, iTead, or similar, and have a batch made for themselves, and then hand out the excess boards to whoever wants some.
Greetings,
Chris