DIY Metcal 13.56 MHz RF Supply

[rfmerrill]

The document says to use magnet wire between 0.7 and 0.8 mm. If I follow that strictly the only AWG that fits is 21 which is not easy to buy. I already have 22 AWG which is ~0.64mm diameter, think that will be ok?

[richard.cs]

It'll probably be fine, it may run a little warmer. I don't remember what size I used on mine, but it was not especially difficult to fit all the windings on so a little larger may also be OK.

[rfmerrill]

Since this thread is more active than mine about sergey's design, I thought I'd ask a question that sort of applies to both:

What is a good way to measure the output filter to verify that you've built it right? I've tried using a NanoVNA and func gen sweeps but I'm not familiar enough with RF to know if that is the right way to do it.

Also, I know to some extent you can compensate for variation in the inductor value by increasing or decreasing the capacitance in inverse proportion, but does anything bad happen if that goes too far in either way?

[rfmerrill]

Sorry for double reply, but unrelated question: Is there a big problem using RG-179 instead of RG-58? The unit I built so far works fine with it, and I figure the length of the run is short enough that it probably doesn't make a huge diff.

[rfmerrill]

I think I just realized why the higher-power design uses the current transformer: Regulating simply for constant RF current limits the amount of power you can deliver without burning up the iron tip once it reaches the curie point. If you assume actually constant RF current magnitude, then there is an approximately 5:1 ratio between your maximum delivered power and the power delivered to an iron above the curie point going by the impedance measurements given by richard.cs

[rfmerrill]

So I got the board made and built it up. It *was* working and then I think the DC-DC converter burned out. I replaced it and changed my thermal coupling assuming it overheated

But now I'm in a weird position where with a 50 ohm dummy load connected I can't make the buck converter output voltage or the power draw change no matter how much I turn the two pots. I've checked the pots and they are definitely doing their job right.

No matter what, with a 50 ohm load connected, it draws 0.7-0.8 A from a 32V supply, and the voltage at L2 is I think 11.7V or so.

For comparison purposes, could someone with a working unit tell me what they measure at RF_DET and VFB with a 50 ohm dummy load connected?

[rfmerrill]

I just realized the voltage at the FB pin of the buck regulator was only 0.8V, which is wrong, but then while probing it I *also* discovered that shorting SW to Vin momentarily murders the device immediately, argh!

Also, I was reading mike's comment to use the -ADJ version of the buck regulator, should I actually be using the -5 version instead? The datasheet says either can be used for voltages >5V

[richard.cs]

Also, I was reading mike's comment to use the -ADJ version of the buck regulator, should I actually be using the -5 version instead? The datasheet says either can be used for voltages >5V

I am assuming you mean the buck on the RF board, checking my Farnell order history I used LM22676TJE-ADJ

I would expect a higher voltage with a 50 Ohm load connected, though I have not checked. It's hard to see how the pots could fail to change the voltage, but yes observing what happens on the FB pin is a good way of debugging it.

[rfmerrill]

Unfortunately I killed the buck regulator, but if the problem is still there after I replace it, the only hypothesis I have is that the FET failed and is dragging the rail down.

[rfmerrill]

I figured it out. D6 had blown short, probably due to me enabling the output with no load connected.

Here's a question: What kind of efficiency are people seeing? I'm seeing around 75 watts in for 38 watts out, which is not deplorably bad but that's a lot of heat dissipation.

That's on par with my actual MX-500 (not apples to apples as the input power was measured at mains with the MX-500 but at 32V with the clone), so that explains the big heatsink

On the other hand Sergey's clone seems to do a bit better, consuming only 56 watts from the wall to deliver 38 watts out. The efficiency gets worse at higher power, consuming 88 watts to deliver 51.5, but from my observations you rarely exceed 40 watts in typical use. Measurements here are from the wall and include the display and losses in the AC-DC and DC-DC converters (note that I did not use Sergey's AC-DC circuit as I live in a 120V country).

The 80W design (Sergey's) seems like it might be the best DIY choice if it could be revised to take universal mains and easier to source parts.

[richard.cs]

I've never measured efficiency or input power on mine. Out of interest how are you measuring output power? Into a fixed 50R load?

[rfmerrill]

I've never measured efficiency or input power on mine. Out of interest how are you measuring output power? Into a fixed 50R load?

I made a test circuit that allows me to vary the reactance in front of an RF dummy load. It includes a peak detect circuit across the dummy load terminals and since the output is close to pure sine you can get output power as (Vpeak)^2/(2*50 ohm). (The factor of 2 is the peak to RMS conversion)

In use:

Inside:

Peak detect circuit:


full size images: https://imgur.com/a/rKOOebT

The circuit just has a variable capacitance in series with the 50 ohm load to let me test the regulation of the power supply -- the 40W design should have fairly constant power into the dummy load while the 80W design should vary when you turn the shaft. The range of the series capacitance is around 500-900 pF

[rfmerrill]

Edit: I found the peak detect output here to be not as linear as I'd hoped. I got the diode wrong in the schematic--I was actually using 1N914, not 1N4148. I modified it to use three 1N914s in series which gives me an output that's much easier to translate to the true peak. Also, the capacitor is 100nF X7R with a 33nF C0G in parallel, and the discharge resistor is 1M

Here's a rough schematic for the test circuit. I'm not sure about the discharge resistor and cap values on the peak detect circuit, but it's a fairly standard circuit and you can probably re-derive them

[t0m]

Drawing substantially on the previous efforts of @NavyBOFH, I've attempted to collate the information in this thread into a github repo: https://github.com/t0mpr1c3/mamalala-Metcal-MX-500-remix

@mamalala: kudos for this project. Could you suggest a license under which to publish the repository?

[rfmerrill]

Repeating the thanks 

One thing I can flag up if I haven't already: I used SOD-123 diodes (MMSD4148) that I already had for the SergeyMax/chbrr design in place of some of the MELF diodes (LL4148) and they seem to work fine. MELFs are a pain, Most End up Lying on the Floor.

Also I ended up using thru hole electrolytics because I happened to have some that were the right spec and happened to fit nicely on the pads with the legs folded out (again, from building Sergey's design)

One thing a coworker of mine pointed out (who's experienced with dc-dc layout): There's no ceramic on the output of the main switcher, which might smooth out its output a bit.

[t0m]

Looking over the documents, there are a few pieces of information that are lacking or inconsistent.

1. What is the value of R18? The schematic says 82K, a nearby annotation says 80K. The build description says 120K. I believe @zayance used 160K for the resistor, which was labelled R26.

2. What is the value of R23? The schematic says 22K, the build description says 2K2.

3. What are the values of R33 and R34? The schematic says 100R, the build description says 2K2.

4. What is the value of R24? The schematic says 100R, a nearby annotation says 200R. I believe @zayance used 200R for the resistor, which was labelled R41.

5. What are the values of C20/C21? The schematic says 22pF, the build description says 56pF.

6. Which components are omitted during installation? As far as I can gather, the omitted components are C20, C30, C40, and R25.

In my most recent commit, which edits the BOM and guide, I assume that the answers are 160K, 22K, 100R/100R, 200R, and 22pF. Please feel free to correct me.

[t0m]

There is an additional consistency about whether R32 is 2K (schematic) or 1K (guide). I have gone with 2K.

[quadtech]

@gman76 , it would be interesting to see your protoboard implementation - do post some pics - TIA

[rfmerrill]

Went back and edited to update:

Edit: I found the peak detect output here to be not as linear as I'd hoped. I got the diode wrong in the schematic--I was actually using 1N914, not 1N4148. I modified it to use three 1N914s in series which gives me an output that's much easier to translate to the true peak. Also, the capacitor is 100nF X7R with a 33nF C0G in parallel, and the discharge resistor is 1M

Here's a rough schematic for the test circuit. I'm not sure about the discharge resistor and cap values on the peak detect circuit, but it's a fairly standard circuit and you can probably re-derive them

[rfmerrill]

After calibrating my peak detect circuit I decided to go back and run it on this design.

With both the SergeyMax and mamalala designs I used the same bench power supply outputting 32 V.

The SergeyMax design gets about 72-80% gross efficiency from 25 to 70 watts output with the average around 75%. This is including DC-DC converter overhead, the microcontroller and the display. Excluding the microcontroller, display and pre-amp circuit (everything powered by the 10V rail) that bumps the percentage up to 77-83% efficiency with the average around 79%.

The mamalala design in comparison gives me around 42-60% gross efficiency (including regulator losses and the LED I added) at a fairly consistent 35-40W real power output. The efficiency of the RF circuit itself is probably slightly more--there's no microcontroller or display in my build.

Because the mamalala design dissipates most of its power in one component (the power transistor) this means using a real heatsink is pretty much obligatory and you can't just sink the transistors to the housing like you can on the sergeymax design.

Interestingly I got lower efficiency numbers after adding heatsinking, but that might just be a fluke

You can see my data here: https://docs.google.com/spreadsheets/d/1Hmx8uLHZ7QaSIQNnxcVqKeQ27fBu1_gVMlza4V-48yg/edit?usp=sharing
(the "Efficiency/New Circuit" tab is the only one relevant, the other tabs were debugging my peak detect circuit).


(Fwiw I'm not trying to say people should drop this design and use the sergeymax one instead--on the contrary that one has a lot of work to be done before it's an easy build. If anything I'd say it'd be nice to come up with a new design that combines the best aspects of both. In the meantime, this design is totally fine if you provide enough heatsinking and don't mind the wasted power. It'll still likely be on the same order of size/weight/power as the old-school MX-500 supply).

[parasole]

The 80W design (Sergey's) seems like it might be the best DIY choice if it could be revised to take universal mains and easier to source parts.

I went through both threads, started by chance with rfmerrill, and ended up here, interesting stuff 

Since the core of the system is an RF amplifier, I think the effort should be diverted to design a good and suitable one, and after some search, I found that this frequency is in large use and there are quite a number of RF power sources available for implementation.
The most promising candidate for this particular application seems to be a class E amplifier, and I found few practical papers which I think could be relatively easily implemented (attached), one is a low-power example for the very same induction heating application, while the other looks to be quite versatile as might pump out up to 400W with the use of cheap IRFP450 transistor.
Anyway, the important aspect here is the design approach for such devices, and these papers possible may answer our questions...
 

[quadtech]

1. From what little I have read, Class E amps are tricky to make work consistently
2. For this application, even 20W seems quite adequate. The advantage of IRF510 is it's dirt cheap
and available everywhere , which is an important consideration for hobby use.
If you look at the schematic of the MX500, it uses garden variety through hole components - not even any exotic driver IC.
For me, in India, availability of parts in the local market is a big constraint, so I appreciate the MX500 design,
as it is robust , cheap and easy to build.

[rfmerrill]

There's value to all of them for sure. I think you could mix and match too--use the single-transistor power amp but with the SergeyMax-type feedback circuit. In fact, I'm curious if you could regulate using mamalala's SWR bridge and not even need much change to the design.

The mosfet driver chip that board is designed for doesn't seem like it's super critical and you could maybe swap it for the one used in the mamalala design.

For me the reason why I like the SergeyMax design boils down to:

• Less waste heat, and specifically less waste heat in a single component. No big heatsink required, just using the die-cast housing is enough.
• Immediate power cut to ~2-5W when the iron is placed in the stand, no need for a timed sleep circuit
• higher peak output power means much faster heating up of the iron when you remove it from the stand or swap tips. A caveat here is that I'm not sure which hurts the lifetime of the tip more: cycling or staying hot

The older design is a perfectly good replacement for the old MX500 power supplies, as its efficiency, need for cooling and size are on par or better. It does require more thought in the mechanical design because you do need a real heatsink.

[t0m]

The 470KHz SP200 design looks even simpler. No DC-DC converter, just a PWM controller, a power transistor heat sinked to the case, and an op amp. Only two big chokes. The handle is cheaper, too.

On the downside, the range of tips that are available is very small.

[rfmerrill]

On the downside, the range of tips that are available is very small.

I think all of the non-MX500 compatible products suffer from the same problem: The market share the MX500 family has is largely full of people that don't really want anything different, and the plentiful second-hand market means that new units from their budget line compete with old units from their flagship line.

It's very hard to find side by side comparison of the size of the handpieces. This makes me suspect that the 470K handpieces are significantly thicker, as the cartridges certainly are (going by the diameter of the cartridge compared to the tip).