DIY Metcal 13.56 MHz RF Supply

[mamalala]

Here's a pic of an old (slightly hacked) US STSS-002. If you want to borrow the inductors to test/compare you're welcome - not sure if I have suitable kit to measure here.
Main output device is a TO-3 IRF130 on the back.
Also a thermal image running - the hottest coil is only hitting about 35 deg.C

That's interresting. I have a STSS-PS2V-02 here, and it has a quite different PCB layout, see attached image (it's quickly stitched together from three separate pics, so ignore the distortion and seams...).

The toroids have a different color. Seems that Metcal was also trying to figure out which ones work best.

The output stage in my circuit is slightly different, since i don't use a pre-driver circuit here but a cheap FET driver chip instead. Saves a bunch of parts and another inductor. Plus, the IRF510 that i use works just well so far. Also, i'm wondering, looking at the datasheet. The IRF130 has much longer turn on/off and rise/fall times than the IRF510.

Anyways, FT82-61 for the xformer works well, T80-2 or -6 for the first filter coil, and T50-2 or -6 for the remaining filter coils. They don't get hot anymore, only a bit warmer. Thing is i have to work with the parts that i can get here. Makes no sense for me to design using some exotic parts that are hard to get. Amidon cores are widely available. It's already bad enough that a 13.56 MHz is so uncommon. Which is quite a surprise, given the fact that this is a common frequency (RFID uses that, for example).

Fun fact: I use the AVX - K50-HC1CSE13.5600MR. Go to Farnell Belgium. There it costs 4,28 Euro. On Farnell in Germany it costs 9,57 Euro. On Newark (basically the US Farnell) it is 3,17 US$.

That's craziness...


Have you taken the thermal images after an hour or so of the station idling? Because it takes some time for the inductors to "warm up".

Greetings,

Chris

[mamalala]

Also just calculated the output power, it nicely dumps over 30 watts into a 50 ohms dummy (i have around 115 volts Vpp at the dummy), so that part is OK.

Might be interesting to try measuring the actual impedance of a Metcal tip to optimise matching & make an accurate dummy load for testing.

Well, given what little info can be found on the net, it seems that during heat-up there is also about 100odd volts at the output. Heatup isn't that much of a concern for me right now anyways. It's takes the same time as with the original. The interesting part starts when it reached temperature. I still only have the 11.0something MHz osc in the circuit. So after heat-up the voltage drops considerably at the RF output. This is where i am pretty sure that the inductance of the coil in the tip itself plays an important role. After all, that is what keeps it warm then, it's not that it completely "switches off" power consumption. One i reached temp it behaves _almost_ the same as the original when soldering on a massive groundplane, it's just _slightly_ worse with thermal recovery. Using the proper frequency should remedy that little difference then, i think.

Dumping more power into the tip is not a big problem for the circuitry. The question is just how the tip cartridges will handle that in the long term. And then, i see no real need to do that anyways. I can live with 6 seconds heat-up time. Bringing it down to 5 or 4 serves no real useful purpose, IMHO. What is a real gain is the fact that you can now change the tips without having to power-cycle the supply. Just pull the old one out and insert the new one, that's it.

Greetings,

Chris

Greetings,

Chris

[mamalala]

I have an antenna analyzer , but would have to make up an adapter to the meter .
But again , I would assume [ we know what that spells ] that Metcal would make the system resonant on the one frequency it uses ?

Hehe, should have checked the new posts. My previous one basically goes into the same direction.

Greetings,

Chris

[mamalala]

Oh, and another thing. Obviously i am trying to come up with a circuit of my own. Not trying to copy the original circuit here, where would be the fun in that... So it's only natural that a few things are different, which is just another reason why knowing the exact specs of the original toroids is not really that important. In the end it's just a simple step-up transformer followed by a low-pass filter.

My focus is mainly on making the whole thing a bit simpler, so that others can build one as DIY project. If in the end it is 90% as good as the original, i'm already happy. I'm hoping that then others will follow up and do the final tweaks to get it to 100% or even better

Greetings,

Chris

[mamalala]

I've uploaded some hi-res PCB pics of the MX5000 here

Now _that_ is very interresting. It seems that in this unit they also used T80-6 toroids. Yellow, with gray side, and the size seems to be that of the T80 (0.8"). Also an IRF510, but not as the final i think, but as a driver. More SMD capacitors instead of all-THT-mica.

Greetings,

Chris

[mamalala]

Here's a pic of an old (slightly hacked) US STSS-002. If you want to borrow the inductors to test/compare you're welcome - not sure if I have suitable kit to measure here.

Forgot to mention: If you could post a scope trace from the waveform at the drain of the final (IRF130) that would be very helpful. That way i can check if mine matches somewhat.

Greetings,

Chris

[mikeselectricstuff]

It's already bad enough that a 13.56 MHz is so uncommon. Which is quite a surprise, given the fact that this is a common frequency (RFID uses that, for example).

13.56MHz crystals are much more readily available and cheaper than oscillator modules (eg. 14 vs. 1 at farnell,cost below GBP0.50) -using an xtal with an HC04 or whatever inverter will be much cheaper and not take much space.

Have you taken the thermal images after an hour or so of the station idling? Because it takes some time for the inductors to "warm up".

No - was in a hurry so only a few mins. May try again later

Now _that_ is very interresting. It seems that in this unit they also used T80-6 toroids. Yellow, with gray side, and the size seems to be that of the T80 (0.8"). Also an IRF510, but not as the final i think, but as a driver. More SMD capacitors instead of all-THT-mica.

Looks like that http://uk.farnell.com/jsp/displayProduct.jsp?sku=1347752&action=view&CMP=GRHB-FINDCHIPS1-1007025]ixys 1xfh12N50F[/url] part is the main output - Ixys parts are popular for solid-state tesla coils. 
Interesting to note The SMD caps are white instead of the usual brown - oddball dielectric perhaps? Also interesting that they are using quite  a few different values in parallel - I wonder if maybe this is to reduce self-resonant effects.

The question is just how the tip cartridges will handle that in the long term. And then, i see no real need to do that anyways. I can live with 6 seconds heat-up time. Bringing it down to 5 or 4 serves no real useful purpose, IMHO. What is a real gain is the fact that you can now change the tips without having to power-cycle the supply. Just pull the old one out and insert the new one, that's it.

Reducing peak draw from cold may also be useful for SMPSU sizing and possible battery operation. I also wonder if teher is scope for fairly slow PWM control instead of analogue to simplify control. I'd think something like 50-100Hz wouldn't produce much in the way of sidebands etc, while still retaining fast response.

Although the 5000 auto-resets I still tend to turn off out of habit. They could have added auto-reset quite easily to the older units.

It would be interesting to see the insides of a Thermaltronics PSU  - the Metcal ones don't seem to have changed much over the years. 

[mikeselectricstuff]

The toroids have a different color. Seems that Metcal was also trying to figure out which ones work best.

Or maybe just a different supplier

[mamalala]

13.56MHz crystals are much more readily available and cheaper than oscillator modules (eg. 14 vs. 1 at farnell,cost below GBP0.50) -using an xtal with an HC04 or whatever inverter will be much cheaper and not take much space.

True. Maybe i go for that, if board space allows. Right now it's rather crammed, but i should be able to make some space for that.
 

Interesting to note The SMD caps are white instead of the usual brown - oddball dielectric perhaps? Also interesting that they are using quite  a few different values in parallel - I wonder if maybe this is to reduce self-resonant effects.

Well, either that or to get the proper value out of a combination, and also for current handling capacity i think.

The white/gray dielectric is usually NP0, which is preferred to be used in such places anyways due to low dielectric losses and low drift.

Reducing peak draw from cold may also be useful for SMPSU sizing and possible battery operation. I also wonder if teher is scope for fairly slow PWM control instead of analogue to simplify control. I'd think something like 50-100Hz wouldn't produce much in the way of sidebands etc, while still retaining fast response.

Right now it draws up to 2.3 amperes during heat up for a few seconds, at 26 volts supply voltage. It drops to about 700something mA once the tip reached it's temperature.

You can't really PWM regulate that thing, i think. What it does is to pick up the peak voltage right before the last output filter coil. It then tries to keep the voltage at that point constant, by feeding that back into the buck regulators feedback pin. The principle here is that if the load matches (i.e. cold tip), the voltage before and after that coil are roundabout the same. The hotter the tip gets, the more of the RF power is reflected back. That means that after that coil the voltage drops, while before it will rise. But since that is in the feedback loop, the buck will lower the output voltage, and thus reduces power fed into the cartridge. And, of course, it keeps the final from going up in smoke: that reflected power would cause a voltage increase there as well, effectively destroying it if there was to be no regulation.

Greetings,

Chris

[SeanB]

Probably a low loss ceramic material, possibly pure aluminia,  as they do have a very high RF current in them The parallel units will also help to increase the cooling as the current gets spread over a larger area of capacitor

[mikeselectricstuff]

13.56MHz crystals are much more readily available and cheaper than oscillator modules (eg. 14 vs. 1 at farnell,cost below GBP0.50) -using an xtal with an HC04 or whatever inverter will be much cheaper and not take much space.

True. Maybe i go for that, if board space allows. Right now it's rather crammed, but i should be able to make some space for that.

or use the osc on a PIC12 or ATTiny - smaller than a  HC04, and you get some simple control functions for free (auto-reset, powerdown timeout, startup current limit.)
 

Interesting to note The SMD caps are white instead of the usual brown - oddball dielectric perhaps? Also interesting that they are using quite  a few different values in parallel - I wonder if maybe this is to reduce self-resonant effects.

Well, either that or to get the proper value out of a combination, and also for current handling capacity i think.

The number will be to get current handling - don't have it in front of me but I think there were 3 or 4 different values. Could be to balance qties used in the BOM but considering other cost-inefficiencies (e.g.RTC and use of TH where SMD could have been used) it seems unlikely

Reducing peak draw from cold may also be useful for SMPSU sizing and possible battery operation. I also wonder if teher is scope for fairly slow PWM control instead of analogue to simplify control. I'd think something like 50-100Hz wouldn't produce much in the way of sidebands etc, while still retaining fast response.

Right now it draws up to 2.3 amperes during heat up for a few seconds, at 26 volts supply voltage. It drops to about 700something mA once the tip reached it's temperature.

I suppose worst-case when you hit a heavy tip to a groundplane is the most significant limit - if this is a lot lower than cold then it may be worth looking at some sort of startup current limit.

You can't really PWM regulate that thing, i think. What it does is to pick up the peak voltage right before the last output filter coil. It then tries to keep the voltage at that point constant, by feeding that back into the buck regulators feedback pin. The principle here is that if the load matches (i.e. cold tip), the voltage before and after that coil are roundabout the same. The hotter the tip gets, the more of the RF power is reflected back. That means that after that coil the voltage drops, while before it will rise. But since that is in the feedback loop, the buck will lower the output voltage, and thus reduces power fed into the cartridge. And, of course, it keeps the final from going up in smoke: that reflected power would cause a voltage increase there as well, effectively destroying it if there was to be no regulation.

I didn't appreciate the voltage issue though  - was thinking in terms of heat only.
It might be viable if you took more control of the loop - would maybe need to take the buck reg into a microcontroller. I supppose you could integrate some current limiting into the buck. 

[mamalala]

or use the osc on a PIC12 or ATTiny - smaller than a  HC04, and you get some simple control functions for free (auto-reset, powerdown timeout, startup current limit.)

Well, i would prefer to have the main RF board working without any µC. It should be purely optional to have µC in there.

However, the SN74LVC1GX04 looks like exactly the right part for this. Now i have to check where to get it. I only have an account at RS, and they don't have it. Digikey is too expensive wrt. shipping costs for just a few parts. Farnell has it, but i don't have an account there.

I didn't appreciate the voltage issue though  - was thinking in terms of heat only.
It might be viable if you took more control of the loop - would maybe need to take the buck reg into a microcontroller. I supppose you could integrate some current limiting into the buck.

Well, it's pretty easy. The max. current draw is a function of the voltage supplied by the buck converter. Reducing that will reduce the max. current it draws. It would be easy to inject some voltage at the feedback input during startup, through a diode, to limit that. A simple R/C network plus a small FET will do the trick. Of course that also means it will take a lot longer for the tip to reach it's temp. I can route the feedback input of the buck to a pin on the header as well, also placing a 1N4148 in series there.

Greetings,

Chris

[mikeselectricstuff]

This is the drain waveform on that old STSS unit starting from cold. Interesting how it slowly climbs for a few secs before peaking.

Regulator output goes 20V up to 24V then down to 18.5V at idle. Current draw at  30Vin goes 2.0 to 2.8 to 0.9A

Ignore 'incident' at end due to stuff shorting out on bench.... it survived!
http://youtu.be/Qslmsz4THvA

[mikeselectricstuff]

Here are the MX5000 waveforms :
Yellow - IXFH12N50F drain, 50V/div  Green = IRF510 drain, 10V/div    Blue=regulator out 5V/div



BTW the parallel caps on the 5000, they use following combinations :
150p,33p,82p  /  6x150p, 33p,39p  /   150px2, 33px2 39p

[mamalala]

Wonderful, thank you very much, this was indeed really helpful. Turns out that i'm not that far away Attached is a photo of the waveform i have at the drain of the IRF510 in my circuit. The "ringing" at the bottom can be reduced or increased by adding/removing turns from the transformer. Plus, i have a really lousy probe as well, that just shows crap at the 10:1 setting, so the voltages there are not correct at all. Anyways, finetuning of the xformer/filters makes sense only once i get the 13.56 MHz source.

The changing waveform on the STSS is correct, i have the very same effect in my circuit. It seems that a cold cartridge reflects more RF than when it has a little temperature. Then it reflects less, until to start reflect much more once it reached its temperature.

Again, thanks a lot for the help. That just confirmed that i'm pretty much on track with what i am doing here.

Oh, that the MX-5000 is a bit different there is because it seems not to have the xformer circuitry that the STSS/RFG/MX500 units have. Instead it seems that it just generates a high voltage "square" wave which is then simply filtered to a sine-like shape. Because the older units drive the center tap of an auto-transformer, you naturally get more ringing and stuff there.

BTW, i changed the circuit and layout here already to use bigger toroids for the xformer and the first filter coil, replaced the osc with the SN74.. and crystal, and also added the diode thingy to allow external limiting of the buck's voltage.

I will look and think about the current layout a bit during the next week, then i'll get a batch of boards produced in China. I guess you are interested in some boards as well?

Greetings,

Chris

[mikeselectricstuff]

Do you have an enclosure in mind? It's always easier to make a PCB fit an enclosure than find an enclosure to fit an arbitarily designed PCB.
As people may want to do different things with PSUs etc, maybe a good approach would be to find a small off-the-shelf well-shielded metal box, which forms a module that can then be put into whatever less shielding-critical case along with PSU.
This would have the RF connector  (these are usually long enough to go through multiple panel walls), and maybe some RF feedthroughs for DC-in and feedback/control.
Seems it would be simple to keep the noisy RF stuff contained in a small module rather than try to stop it coming out of a large box.

e.g. this sort of thing :
http://www.perancea.com/prod-fr.htm
or maybe a small diecast box.

Or at least include holes to allow for an off-the-shelf can to go over the PCB (might you need to add clearance to avoid eddy-current losses?)
e.g. http://www.perancea.com/prod-fr.htm

Getting good heatsink coupling with good shielding may make this slightly tricky.

Another option may be to make a set of shielding panels out of PCB that can be slotted together and seam-soldered.
 
 

[mamalala]

Do you have an enclosure in mind? It's always easier to make a PCB fit an enclosure than find an enclosure to fit an arbitarily designed PCB.

Ahh, well, the dreaded issue of enclosures. In the past decades i found that it is close to useless to design DIY stuff for some specific enclosure. The markets are way too different to even come close to a common denominator here. For example, in the US market you see Hammond very often. In Europe they are more of a curiosity, however. In Germany you see things like Bopla or OKW very often (BTW, check out OKW. They make _really_ nice enclosures!), but overseas they are basically unheard of.

The current PCB is exactly 91mm x 91mm. It has four mounting holes in the corners. The FET final sits in a cut-out inside the PCB. The reasons for that are various. For one, there are 100mm x 100mm heatsinks available. Or any other heatsink in a plain square format with roundabout the same dimensions. So the PCB would be mounted directly onto that heatsink, allowing the FET to be mounted on there as well. Since the heatsink is massive aluminium, it acts as a shield as well in that direction.

Then, all the connections are done as solder pads, and the header is also placed on the board in a way that does not require any "sideways" stuff. So the controlling/supply PCB can be stacked directly onto it. Using hex standoffs, they can act as the screws to mount the RF PCB on the heatsink, while allowing the control/supply PCB screwed on top of that. Like a sandwich.

What goes around is completely up to the folks building that thing. I would go for sheet aluminium and 90° aluminium profiles, all stuff that is readily available. A heatsink usually has a thick enough base to allow tapping in of threads on the sides.

Or the whole thing can be screwed into some die-cast metal box. There are many options, actually. In any case, i'm definitely not going to plan the PCB for a specific enclosure. Too much trouble down that route....

Oh, since 100x1600mm for example is a standard format, at least here in Europe, for PCB's (heck, that format is even called "Euro-Format"), there are plenty of enclosures available where it will fit into. For example that stuff that is basically extruded aluminium profiles with sheet aluminium as the front/rear panels (and sometimes top/bottom). So, with the PCB size i have chosen there should be little problems to find a suitable enclosure.

Greetings,

Chris

[mamalala]

Speaking of heatsink, this is the one i had in mind:

http://uk.farnell.com/fischer-elektronik/sk-508-100-sa/led-heatsink-standard-extruded/dp/1850032

Greetings,

Chris

[mikeselectricstuff]

Ahh, well, the dreaded issue of enclosures.

tell me about it - I've spend way too many hours of my life finding just the right one...

In the past decades i found that it is close to useless to design DIY stuff for some specific enclosure. The markets are way too different to even come close to a common denominator here. For example, in the US market you see Hammond very often. In Europe they are more of a curiosity, however. In Germany you see things like Bopla or OKW very often (BTW, check out OKW. They make _really_ nice enclosures!), but overseas they are basically unheard of.

Nonsense - There are no significant national borders these days for that sort of stuff.  All of the above are widely available pretty much anywhere. I think they all have UK offices, but as most are stocked by Farnell/E14, Digikey etc. they're no problem (at least no more than any other component) to get pretty much anywhere.
OKW and Bopla are nice but often rather expensive. Hammond are usually very good value and well made - usually my first choice when looking for plastic enclosures.

Incidentally have you done any measurements on the actual dissipated power/heatsink requirement?

[mamalala]

Nonsense - There are no significant national borders these days for that sort of stuff.  All of the above are widely available pretty much anywhere. I think they all have UK offices, but as most are stocked by Farnell/E14, Digikey etc. they're no problem (at least no more than any other component) to get pretty much anywhere.

Well, yes and no. True, it is nonsense in so far that many distributors offer them. But then, in the EU they have the habit of not wanting private people as customers but only want to sell to businesses. And the shops that sell to everyone usually have a very limited range.

Sure, it's somewhat of a moot point anyways, since this circuit already uses parts that are not available in many hobbyist shops anyways. but then, it's kind of a habit of mine to ignore the issue of specific enclosures. It's too much of a waste of time for some generic DIY project that is intended to be modified or have different "extras". It's not like i'm doing this for some customer, where i can decide what has to be used and what not.

Talking about DigiKey. I find their shipping costs absurd, at least last time i checked. Oh, and have you ever ordered samples from Texas Instruments? Compare the paperwork with that of a DK shipping

Incidentally have you done any measurements on the actual dissipated power/heatsink requirement?

Not yet, no. I use a pretty inefficient heatsinking right now, since i'm still in development of that thing. This is what i have for testing:

http://www.pollin.de/shop/dt/MzI5OTY1OTk-/Bauelemente_Bauteile/Mechanische_Bauelemente/Kuehlkoerper/Kuehlkoerper.html

As you can see in the first images i posted, it actually is _under_ the PCB, fins down. Plus, it is really small. All i can say is that after 1 hour, my dummy load that uses 2 of these heatsinks gets hotter than the one on the FET.

Greetings,

Chris

[mikeselectricstuff]

Talking about DigiKey. I find their shipping costs absurd, at least last time i checked.

They ship free to UK for orders >GBP50  - I think it's similar in EU

Oh, and have you ever ordered samples from Texas Instruments?

Yes, but not recently - usually very quick & efficient - only issue I've had is once needing more then their max qty, so got someone else to get some as well. This was for parts that weren't in stock anywhere.

[mamalala]

Alright, just placed the order for a bunch of PCB's, 10pcs. of the RF stage PCB and 10 pcs. of the Controller/Supply PCB. Should arrive between mid and end of next week. Hope i didn't introduce some nasty mistakes, but after rechecking several times i found nothing suspicious.

Once they arrive i will assemble a unit to determine the final values of a few parts, and then release the design files plus some docu on how to assemble it, winding the toroids, how to adjust, etc.

Greetings,

Chris

[mamalala]

Small update: The PCB's arrived today, after the delivery guy messed it up yesterday. First he claimed my name was nowhere to be found. Then wrote a notice for the delivery attempt using a different name (from a neighbour) and put it in her mailbox. Also claimed my name was nowhere on any of the mailboxes here. DHL Express, my ass ;(

Greetings,

Chris

[SeanB]

What do you expect from the lowest paid worker? Here I have difficulty getting the post on a postbox, often the wrong box bgets my mail and vice versa. I often lean in through the box and put it in the next box where it should have been.

[mamalala]

What do you expect from the lowest paid worker? Here I have difficulty getting the post on a postbox, often the wrong box bgets my mail and vice versa. I often lean in through the box and put it in the next box where it should have been.

Well, i expect from them to be able to read what is written on the mailbox and the doorbell-knob. The latter being a really big one. I expect that especially from some alleged premium service. And i don't like to be lied to in my face. Because that guy today claimed that my name was not on any doorbell at the front, nor on any mailbox in the house. Which is just wrong. Not only is it there, for years, it is printed out in big letters. Funny enough, i usually get everything without any problems. It's only around holiday seasons that things go south, when they get a bunch of reserve people because the regular ones are in holidays.

I understand that they get little pay and have a heavy workload. But that still isn't a reason to be that sloppy, let alone to lie straight into my face.

Greetings,

Chris