DIY Metcal 13.56 MHz RF Supply

[Paulinho19]

Today's result, 3 new FETs dead.

I installed the new coils, removed L6, improved soldering on the CAPs and the result is still the same, FET dies when voltage on the output on L2 goes over 15V. The voltage p-p on XFMR_Center is under 60V at the moment of death.

The only thing left to try is a different dummy load.  Next week I will have one to test.

Paulo

[mamalala]

Sorry to hear that. It's really strange.

Have you checked the VR and VF voltages to see if anything is reflected back from the dummy and/or connection to it? Thing is, when the output is matched, there should be a considerably lower Vpp at the drain at only 15 Volts at L2. I reach the 60 Volts only with a tip that has heated up, i.e. has a mismatch, but even then the voltage at L2 is around 18 Volts.

At a around 13 Volts i get a bit below 40Vpp at the drain, with the probe that reads a bit higher, and current consumption is 1.3 Ampere at 30 Volts supply input.

If everything else fails i can send you a set of coils and two or three of the FET's i use. While i highly doubt that my FET's here are some special magic version of the IRF510, who knows what your cores actually are. Look the same than the ones i have, but again, who knows....

Greetings,

Chris

[Paulinho19]

No I didn't get to check the VR and VF voltages, I went out of FETs before I read your Post.

This is really a mystery, I'm confidante that the problem is the dummy load, and there is not much left. Next week we will see.

Thanks for all your help Chris.

By the way did anyone else assemble this board yet?

Paulo Almeida

[mamalala]

BTW,

attached is an image of the FET i use. Does anyone recognize the logo in the bottom-left?

Greetings,

Chris

[Paulinho19]

Could be Vishay Siliconix, I'm not sure. 

www.vishay.com/docs/91015/sihf510.pdf

Paulo

[az113]

Paulinho19
With big interest look for this theme.
Great project! May be will do a module to replace MX500 power unit...

P.S. Why you use IRF510?
It has a 0.5ohm internal resistance, may be something similar to IRFb4019 with 0.008ohm be more efficient?

P.P.S. Sorry, IRF630/IRF640/IRF740 i mean. Still, it is a UHF modulator

[mamalala]

P.S. Why you use IRF510?
It has a 0.5ohm internal resistance, may be something similar to IRFb4019 with 0.008ohm be more efficient?

P.P.S. Sorry, IRF630/IRF640/IRF740 i mean. Still, it is a UHF modulator

Basically for two simple reasons:

- RF circuits are not really my field of expertise, so for the initial design i limited myself to the types that i had already used once in circuits operating at the same or higher frequencies, so i would know they should work in that application.

- Because i wanted to reduce the overall complexity of the circuit, i chose to use a readily available FET driver chip instead of the much more complex driver circuit used in the original Metcal supply. However, that chip is very tiny, and to avoid overheating i chose a FET with a really low gate charge and capacitance.

(Edit: Of course i might have been too paranoid with that. I just chose to stay on the safe side so i can get something to work correctly)

Of course that is not to say that only the IRF510 would work. But since i lack proper equipment to really evaluate other FET's, it would be up to others to see what else can be used there. The IRFB4019 that you mentioned looks interresting. It has even faster switching times, while still having a relatively low total gate charge. Might be worthwile to experiment with that one.

Greetings,

Chris

[az113]

Because yesterday i totally felt asleep, i mentioned a IRFB4019, but not think about frequencies, my primary job today is repairing DC-DC converters in notebooks and others equipment, here is a cause of recommendation.

Today i think more about FETs, and remember,  5 or 6 years ago, in CB power amplifiers we used   same IRF510/540, but we have a lot of counterfeit in our stocks.
May be you got fake FETs?  At the photo your IRFs looks slightly scratched (like scratches from sandpaper), it get me think, that is a fake.
Try to buy from a more trusted source? Or try IRF630/640, it works well in 17" and 19" monitors in  B+ cirquits, that is more than 100V there.

[mamalala]

May be you got fake FETs?  At the photo your IRFs looks slightly scratched (like scratches from sandpaper), it get me think, that is a fake.

Do you mean my image from this message:

https://www.eevblog.com/forum/projects/diy-metcal-13-56-mhz-rf-supply/msg192751/#msg192751

or Paulinho19's image? Mine appear scratched because they are stored in a simple bag, rubbing against each other. They looked just fine when i got them in a tube.

It is important to note that the FET is not used in a linear fashion here, but as "switcher", which makes the turn-on/off delays and rise/fall times so important. You can use an, for example, IRF530 just fine as a RF final in the 10/11 meter band (28/27 MHz, and quite some radios do indeed use the IRF530) when operating as linear amp, but it would very likely fail to work at the same frequency as a pure switcher, because the delay and rise/fall times are far too big for that frequency.

Greetings,

Chris

[Paulinho19]

Yet another bad day,

Today I replaced C14-C19 just in case, and connected a Arduino with a LCD shield to monitor VF and VR. Started everything ok and I got to about 2V on VF and 0V VR when the FET died. So it didn't look like a problem with the load. Tried it again with another FET, the same result, but this time I also killed the FET driver and Q2, I really don't know what is going on here....

Paulo

[Paulinho19]

I think the problem are the FETs, I ordered some new IRF510 from Farnell and I also got one IRF620, I will try that tomorrow, after I replace The driver and Q2.

Paulo

[mamalala]

That is extremely strange. Maybe we are looking at the wrong place here. What i find suspicious, and why i was thinking of a problem with the filter/dummy, is that you reach so high RF voltages at such rather low supply voltages at L2.

Have you observed the voltage at L2 with a scope while increasing the output power with R19? I am now starting to wonder if there might be a problem with the buck converter, that for some reason it no longer regulates down when reaching a certain current draw from it. Or that for some reason it starts to oscillate wildly when reaching a certain current draw. A normal multimeter might not be able to catch that, if that transient is short enough, and surely not if there is some oscillation. The FET however would fail in a very short time under such conditions.

What about the +12V supply for the driver? It's really strange that not only the driver is fried, but that it also took down Q2, which is merely to enable/disable the driver using the powerdown-input.

Something is really fishy there. As said, i can send you a set of coils and some FET's, and probably a XFMR as well. Do you have a driver left? Otherwise i can send you one as well, and maybe a regulator as well (not sure if i have any left). At least with the coils/transfomer, those would be tested and correctly working ones, and the FET's would be the same ones that i use and that work. However, it would take me a while until i can send the stuff, since i'm mostly in bed and on my sofa due to a massive flu i catched...

Greetings,

Chris

[mamalala]

I think the problem are the FETs, I ordered some new IRF510 from Farnell and I also got one IRF620, I will try that tomorrow, after I replace The driver and Q2.

Paulo

As i wrote in my other post, i'm not that sure anymore that this is the (only) problem. You reach way too high RF voltages for the supply voltages at L2, compared to my unit here. Before you burn another batch of FET's, please try the following:

- Turn down the input supply to 22 volts.
- Monitor the voltage at L2 with the scope.
- Now increase the voltage at L2 with R19 again until it can no longer be increased.

If my theory about the buck regulator is right, it should not blow the FET this time. Now increase the input supply voltage by one volt, readjust R19 again to reach the maximum at L2.

Repeat the 1V increase and R19 readjustment until you get around 120 Volts peak-peak at the output. That is assuming the FET won't have been blown at this point.

At all times monitor the voltage at L2 with the scope. Check if there is anything unusual there, like big spikes, oscillations, etc. I get a maximum ripple of 1V peak-peak there, as you can see in the first scope-screenshot in this post: https://www.eevblog.com/forum/projects/diy-metcal-13-56-mhz-rf-supply/msg192370/#msg192370

Also, please make notes of the current consumption of the circuit. It should steadily increase as you turn up R19, no big jumps or the like. As mentioned earlier, mine draws about 3 Ampere at 30 Volts, with a dummy load connected, when the voltage at L2 reached around 23 Volts or thereabouts.

Greetings,

Chris

[Paulinho19]

I successfully killed my first board.

Testing the buck converter I applied 12V to pin 5 by mistake and fried a whole section around the Cristal. So tomorrow I will start assembling a new board, maybe this time everything will work. I will keep you updated.

Paulo

[az113]

Guys, i am not a guru in RF technics, but here is an idea: do you think about SWR in this case?
As far i remember from 10/11meter radios, simply recalculate a length of a coaxial wire to antenna and properly done weldings  give me additional power up from 9+5db to 9+10db and SWR was get lower from 1.25 to 1.15.
Paulinho19 What type of a coax you use?
Don't know surely, does Metcal's gun use a 50 ohm cable, or 75 ohm?
May me here is a problem?

[mamalala]

Guys, i am not a guru in RF technics, but here is an idea: do you think about SWR in this case?

Just a few posts above, in #159, Paulo checked the VR (reflected signal of the RF) and it is at 0 Volts, so there is no problem with the SWR.

Also, during evaluation of the first circuit i made, i used varying lengths of cable between the board and the F connector, between none (i.e. the connector directly to the board) and around 20 cm. I checked RG58 and RG59, that is, 50 Ohms and 75 Ohms coax. There was virtually no difference.

Considering the length of that short cable in relation to the wavelength of the signal (around 22 meters), it wouldn't have any notable effect anyways. Plus, a slightly increased SWR does not kill the final. I had situation with a VSWR of at least 2 at full power. While the FET heated up more than normal, it did not blow.

Greetings,

Chris

[Paulinho19]

Hi,

I'm using 90cm of RG58 coax.

Today I started assembling a new board, its almost finish, I tested the buck converter 30V 3A, no problems.

From the old board I only used 4 caps from C14-C19 and I will use also L3-L5. The XFMR I will do again just to be sure.

Tomorrow I'll have a 500W 50ohm dummy load and some new IRF510 from Farnell.

Paulo

[Paulinho19]

Hi,

All new, same old problem, FET dies before voltage on L2 gets to 21V.

I just tried the new Dummy load (1000W) and the new FET's,

http://pt.farnell.com/vishay-formerly-i-r/irl510pbf/mosfet-n-logic-to-220/dp/9102779?Ntt=9102779

with the new board I assembled yesterday and the result was exactly the same, the FET died at 20V on L2, VF 2V , VR 0V, output power around 30W.

This time I noticed that above 15V on L2 the voltage starts to rise without futher regulation on R19.

I have 4 more FETs to Kill, what should I do to try to find out were the problem is?

Paulo

[mamalala]

Just for information:

Paulo and me had a chat, and i urged him to stop any further testing until he has sorted out what capacitors he used in the buck converter circuit. The fact that the buck's output voltage increases by itself by 5 volts just when the load gets higher could be a strong hint at a problem there.

Right now it is unclear what caps he used, first he thought some Panasonic EFK (which are completely wrong for this application, only 350 mA ripple current for example). But it turned out that someone else ordered them, and so we have no idea what they actually are.

Chances are that they are simply unsuitable for this application, causing the buck regulator to go haywire once a certain load is reached, which in turn kills the output FET. After all, the caps are under quite some stress here, not only the 500 kHz ripple from the buck converter, but also a 13.56 MHz ripple from the RF output stage.

Greetings,

Chris

[Paulinho19]

Quick update,

the capacitors he used in the buck converter circuit were this:

http://pt.farnell.com/jsp/search/productdetail.jsp?sku=2069205

they are low ESR but not as low as the ones Chris used (340mOhm / 28mOhm) and they support also much less ripple current (670mA / 2000mA) :

http://pt.rs-online.com/web/p/condensadores-de-aluminio/7581272/

So I ordered some like Chris's today. Hopefully next Monday they are here and I can test the whole thing again.

Meanwhile I will finish my design on the Arduino based Controller board.

Paulo

[amspire]

Hi,

All new, same old problem, FET dies before voltage on L2 gets to 21V.

I haven't read through the thread, but I did see a circuit with a IRF510 going to a transformer. Is that the one blowing?

I don't have Eagle installed so I cannot open any Eagle schematics.

If that is the circuit, there are two things I noticed.

First, all transformers have leakage inductance and you usually need a snubber on the drain - otherwise the leakage inductance will cause the peak drain voltage rise to the point where it is limited by the FET's avalanche breakdown.

The second point is you have 10 ohm resistors from the mosfet driver IC to the IRF510 - why?  Here is the problem. The reverse capacitance to gate capacitance ratio is probably something like 1:15. This means if you are switching any more then 30V, then when the FET is turning off, the reverse capacitance will turn the FET back on again during the low to high drain transition. This causes a momentary high dissipation in the MOSFET and I have seen this blow mosfets. Some mosfets handle this stress well - others do not. The two 10 ohm resistors may not low enough to stop this if you are switching more then 0.5A. Using one of the drivers without any resistor would probably do a much better job. It is often necessary to drive the gates of MOSFETS negative to stop the device turning on again during low to high drain voltage transitions. Often the gate resistance is not low enough to ensure the mosfet stays off even if the gate is clamped to 0V, and it sometimes needs more like -10V at high currents to really switch the mosfet off during the low to high drain voltage transition. At the frequency you are operating at, this drain power spike is happening at a fast rate. If you were switching at 50KHz, it would be no problem, but you are not.

If you do get the mosfet switching hard off, there will be less transformer leakage inductance energy absorbed, so you will need a snubber circuit (usually a high speed diode, capacitor and resistor) even more.

Richard.

 

[mamalala]

Hello Richard,

all good points, but i don't think that those are what cause the problem here. I had and have the cirsuit running here without problems using two different brands of IRF510. During development i used gate resistors larger then 10 ohms without problems as well (22 ohms, to be exact). In every case it worked. Also, testing at high output powers i drove the supply above 25 Volts, with the voltage at the drain/xfmr rising above 80 or so volts, still no problem. If i had FET's dying in a similar way during that period, i would be inclined to agree with all/some of your points. But i never had. Again, with two different brands, so it couldn't really have been some fluke due to some "out of spec" variant of the IRF510 either. The ones i killed died due to other reasons: shorting stuff in the filter section by accident during testing, and non-working RF power feedback while load mismatches occurred, both of which are to be expected to kill the output final.

Also, the fact that in his ciruict the buck converters output voltage rises all by itself once a certain level is reached (and thus a certain current draw) to me strongly indicates that there is something wrong with it. It's just that this rise previously wasn't mentioned, so i assumed the supply was stable. It simply should not rise from 15V to 20V all by itself. And that is "only" what was observerd, could well be that it rises way beyond that very quickly. And in the end, there _is_ a limit as to how much supply voltage the RF stage can take. The FET it good for 100V max. "only", after all.

Greetings,

Chris

[amspire]

Chris,

Why are the 10 ohms resistors there at all? You have a driver that has a rise fill time under 2nS and 3.5A capacity, and that is great. But then you slow it down with the 10 ohm resistors. I do not get it. I understand putting 10 ohms gate resistors in a linear circuit to prevent oscillations, but in this case, the only purpose is to make the mosfet switch slower and get hotter. Are you slowing down the mosfet so it acts as the snubber for the transformer?

About stress caused by letting the FET turn back on during switch off, I have seen cases where a circuit works fine with IR parts, but blows with an equivalent Harris, ST or several other brands. All of the devices that blew were 100% healthy and fully met the published specs. Admittedly it was a more extreme case - IRF450s and 460s switching 350V. In these cases, if you build enough circuits, you find that the failures with the IR devices is unacceptably high. My suspicion is that when the gate drive is marginal, the heat when the mosfet switches on again during turn off gets dumped into a small percentage of the mosfet cells in the device, and it is these cells that fail. you are using a device in linear mode that is only designed for switching mode. With a good gate drive, this situation does not occur.

Let us say this turn off conduction only occurs for 5ns. Since you are running at 13MHz, this means that for over 65mS per second, the mosfet is still conducting as the drain voltage is increasing. You also get a slower turn on, so that is more stress. You end up with a small part of the mosfet exceeding the SOA even though the device is not even very hot.

About my first point, are you sure that the drain voltage never spikes up near the device limit? It can be a very fast spike. I do not know the parameters of the circuit, and it is possible that the amount of leakage inductance can be absorbed by the drain capacitance of the mosfet limiting the peak voltage on the drain.

Also, testing at high output powers i drove the supply above 25 Volts, with the voltage at the drain/xfmr rising above 80 or so volts, still no problem.

That is getting up towards maximum. What stops it from getting higher then 80V particularly under fault conditions? Are you getting ringing? (Which often radiates noise like crazy).  I still think a snubber may be needed - or  a 200V device.

Is there anything to limit the peak transformer current so that the transformer cannot saturate under fault conditions?

[Paulinho19]

Hi Richard,

so in or opinion we should replace the two 10 Ohm resistors with two 0 Ohm resistors and also replace the FET with a IRF620 for example, correct?

Paulo

[amspire]

Hi Richard,

so in or opinion we should replace the two 10 Ohm resistors with two 0 Ohm resistors and also replace the FET with a IRF620 for example, correct?

Paulo

First start with a low supplied voltage to the mosfet, and gradually increase. Record some peak drain voltages with a known load.

Replace 10 ohm with a short, leave the other open. One driver probably has the power, and they may not be designed to work in direct parallel.

Look at the drain voltage - see if it is higher. If it is, then the mosfet was definitely absorbing extra power during the transitions, and as I tried to explain, conduction in the linear mode is the most difficult power for the mosfet to handle safely.

If it is higher, you could use a higher voltage mosfet, but the better solution is to get the peak drain voltage under control if possible. I do not know the transformer ration, the voltages, and the circuit on the transformer output, so it is hard for me to suggest something. Based on the transformer ratios and peak expected flyback voltage of the transformer when the mosfet turns off, what would the calculated peak voltage be (excluding and leakage inductance - assume the transformer is ideal)?

If the mosfet is allowed to go into avalanche breakdown, it will mean that for a very short time perhaps half the peak drain current at 100V plus is flowing into the mosfet for probably a few nanoseconds, but 13 million times a second. If most of this current is going though a few cells in the mosfet, those are the cells that will fail first which will destroy the whole mosfet.

A simple drain snubber circuit is a capacitor and resistor in parallel connected to the mosfets + supply rail. The other end is connected to a switching diode (it will have to be very fast - perhaps a 100V schottky diode) from the mosfet drain (anode) to the capacitor-resistor (cathode).  The voltage across the capacitor will rise to a point when the average current from the leakage inductance equals the average current through the resistor. The bigger the resistor, the higher the voltage but the less power is probably being wasted. The resistor is picked so that under worse case conditions, the drain voltage never exceeds the maximum.

There are more sohpisticated snubbers that are more efficient, but start simple. It could even be that a capacitor and series resistor without the diode is sufficient, especially is the leakage inductance of the transformer is very low. (Probably is at 13MHz).

If it does not look like you can keep the peak drain voltage is under 100V under all load conditions, you may have to go to a 200V mosfet.

The IRF620 has substantially more capacitance. Can you use a IRF610? (I do not know your drain current). The greater capacitance is not necessarily a bad thing - the greater capacitance may slow down ringing on the drain.