(Now with pictures!) I built Sergey's 80W metcal-compatible soldering station.

[rfmerrill]

Going by what I could see from SPICE simulations, it looks like the voltage across C75 is roughly (Vrms*Irms/31 A) * cos θ, which some of you may recognize as P/31A (real power). Doesn't quite gibe with what I measured on the real circuit though (I was seeing 1.3V with the iron in the stand and negative voltage with the iron under load). I'm sure there's something my simulation is missing.

Of course, I got the sign backwards, it's actually negative that XD

[rfmerrill]

I replaced L6, L7 and T2 with parts that seem to work a lot better

For L6 and L7 this EPCOS VHF choke works pretty well although it is a little tricky to bodge into place: https://www.digikey.com/en/products/detail/epcos-tdk-electronics/B82111B0000C022/4245895

For T2 I wound a new transformer with the same turns ratio on an ft-63a-61 core from Amidon. The specs on the core don't look ideal (complex permeability goes bumpy a little bit above the operating freq) but it seems to work and it's probably a lot better than the uxcell iron powder core.

[rfmerrill]

I should probably mention that I'm normally a software engineer and this is my first time doing a project with this much analog and really any RF in it 

[rfmerrill]

Looks similar to a Bruene coupler, sans the low side of the capacitive voltage divider.

Thanks for the lead on this! It looks like it is indeed similar.

I've done hours of more head-scratching and simulation work and the circuit is much more nonlinear than I expected
Click for big


So coming back to his schematic:


Here's how it works:

Current flows in the same direction on the schematic on primary and secondary side (same polarity voltage = opposite polarity current!)

The node between D16 and D17 is clamped to approximately +/-2V by the diode ring.

Because of the above, we can think of C47 as being from the output voltage to essentially ground, as the voltage on the other side of C47 is much greater than 2V in magnitude. Therefore, the current through C47 is approximately the same as it would be were it connected to ground--in quadrature with the output voltage and equal to approx Vout / 1173 ohm (either RMS or peak since it's a sinusoid)

When primary current is flowing left to right--i.e. towards the soldering iron tip--D16 and D17 are reverse-biased, and therefore no current flows through R40

When primary current is flowing right to left, the current induced in the secondary flows through D16 and D17, and the additional current injected by C47 has nowhere to go but into R40--until the voltage at that node gets high enough that either D21 or D22 is forward biased. Because the current injected by C47 is so high, this results in R40 effectively seeing a square wave current which you can see in the LTSpice simulation above.

So if we want to create a simplified model of how R40 is being driven, we can think of it like this:

• Not driven when primary current is flowing left to right
• Driven at +2V when primary current is flowing right to left and current is flowing down through C47
• Driven at -2V when primary current is flowing right to left and current is flowing up through C47

Because the current through C47 is 90 degrees out of phase with the output voltage we can then reach the conclusion that if the output current and voltage are in-phase, periods 2 and 3 are of equal time and thus the voltage at C1 is zero. Of course, that voltage is also being pulled towards 0.8V via its connection to the feedback loop of the DC-DC, but we'll ignore that for now.

If the output voltage starts to lead the output current, then period 2 will lengthen and period 3 will shorten, and the voltage on C47 will be pulled higher. Likewise if the current leads the voltage, the inverse will happen and the voltage on C47 will be pulled lower.

So in essence this is not as dependent on the magnitude of output current and voltage, and much more on the relative phase, which makes sense.

[rfmerrill]

So the above got me curious to see what the actual current waveforms look like on the iron in operation. I took some scope shots:

Iron in stand


Iron heating up


Iron at temperature


Channel 1 (yellow) is the voltage at across the output jack. Channel 2 (blue) is the voltage across an 0.1 ohm current shunt right before the output jack.

The current waveform not being a clean sinusoid kinda threw me, but it makes sense if we're operating around the saturation of the heating element. Still, I'm curious why the current is leading the voltage while power is being delivered to the tip. Capacitance? Intentional effect of the length of the handpiece lead? Something about inductive heating I don't quite understand?

[rfmerrill]

Talking about the tip detect circuit located right near the output jack, I was a bit confused how it works but now I see that basically that it's basically measuring the DC resistance of the hot side of the RF circuit. With no iron tip connected it gets pulled up to 3.3 and otherwise gets shorted to ground.



I have smoked R37 a few times and I'm not 100% sure but I think this was just a result of accidentally shorting the 22V supply to the iron (probably through my panavise). One thing I'm confused about is why he used such a huge 22 microhenry inductor, especially since I have a lot of trouble finding axial inductors with a self-resonant frequency above 13.56 MHz (although having a self-resonant frequency close to that might be beneficial since it would have a much higher impedance). Seems to me R37 and R41 could easily both increase in value, and L12 seems like it could be replaced by a ferrite bead rather than a plain inductor?

[TimNJ]

OK, I might need help here.



For L6 and L7 I bought something that seemed close in AL, permeability and size to the weird Russian cores Sergey used. The only problem is I accidentally bought iron powder cores instead of ferrite and I'm pretty sure that's why L6 cooked itself. I swapped a new one in before I had any clue what was wrong, and it makes its way up to 90C in not very much time even with the iron in the stand.

Importantly, I'm pretty sure it's the core getting hot and not the winding.

So what should I use for those inductors instead? I'm looking at ferrite cores and they all have much higher AL and permeability than the core Sergey mentions in the blog article ("HF-ferrite 16x8x6mm ring cores with permeability of 50"). I think the max DC current it needs to take is 4A. Could I just buy an inductor instead of making my own? Or could I use a higher-permeability ferrite with way fewer turns? (like 3?)

Edit: A big part of why I'm unsure about this is that I don't know how much effect a change in that inductance would impact the circuit, or how to compensate for it, and hitting 9 uH on the dot is hard with high-permeability cores.

"HF-ferrite 16x8x6mm ring cores with permeability of 50". A typical ferrite material (that is, one made from MnZn) does not have a permeability of 50u. The minimum permeability for a MnZn material is about 1,000u, and more typical values are 3,000 to 5,000u.

If the core he used was 50u, this implies some sort of powder core, whether that's iron powder, sendust, high-flux etc.

You're probably looking for something like Micrometals T60-18, although I have not looked at the circuit to see exactly what's needed.

You can buy here: https://www.cwsbytemark.com/index.php?main_page=index&cPath=206_217&zenid=s7qsh136tlkgqjm4iuslnf8fk4

[rfmerrill]

"HF-ferrite 16x8x6mm ring cores with permeability of 50". A typical ferrite material (that is, one made from MnZn) does not have a permeability of 50u. The minimum permeability for a MnZn material is about 1,000u, and more typical values are 3,000 to 5,000u.

Yeah this gibes with what I could figure out on my own and why I was confused initially.

If the core he used was 50u, this implies some sort of powder core, whether that's iron powder, sendust, high-flux etc.

I honestly believe he used some kind of cold war era soviet ferrite made with much older tech. I googled some of the terms he used and get mostly russian results with some people talking about old russian radios. I went and looked just now and it does look like NiZn ferrite materials have permeability in that kind of range. Fair-Rite's 67 material is around 40 for example, and does say recommended freq up to 100 MHz

You're probably looking for something like Micrometals T60-18, although I have not looked at the circuit to see exactly what's needed.

It's a drain choke for something kinda-sorta like a class E amp but not really I guess? It needs to withstand DC current of around 4-5 amps, and around 100+V peak-to-peak at 13.56 MHz, and be around 9 uH

You can buy here: https://www.cwsbytemark.com/index.php?main_page=index&cPath=206_217&zenid=s7qsh136tlkgqjm4iuslnf8fk4

Thanks!

[james_s]

Seems like an RF guru could probably suggest some suitable cores for the task. Maybe try on a ham radio forum? Someone who builds/modifies shortwave transmitters and linear amplifiers is likely to have relevant experience, if anyone does that anymore.

[rfmerrill]

Seems like an RF guru could probably suggest some suitable cores for the task. Maybe try on a ham radio forum? Someone who builds/modifies shortwave transmitters and linear amplifiers is likely to have relevant experience, if anyone does that anymore.

Was my thought, yeah. I joined a ham radio chat but there didn't seem to be much discussion of radio circuit details.

So I've somehow ended up stuck in a state where Q2 is cooking and R37 goes out with a bang any time I disconnect the handpiece. The only major changes I can think of are that I switched to a different core for the current transformer and the RF jack has a shorter lead to the PCB... Q2 is getting hot enough to melt the plastic insulator bushing, even the ones I just bought that are polyphenylene sulfide!

One time I did seem to blow Q2 by pulling the cartridge, which I ascribed to the TVS not doing its job. I'm now wondering why there's no TVS on the output of the filter. I also tried a few different inductors for L12 because I realized that jellybean axial inductors are probably past self-resonance at 13.56 MHz, but that might have accidentally made it worse!

This is all so frustrating because I can't even scope these signals before the components blow up or cook themselves. I'm glad I bought extras, but those STP19NF20s aren't cheap!

[rfmerrill]

OK, I think I might have a good idea of what is going on. Here's what I've observed:

• Thermocouple on Q2 heatsink shows it slowly rise to about 40-60C before staying there, unless you give it a heavy load
• Everything seems totally fine until you pull the cartridge out of the handpiece
• once you do that, R37 will enthusiastically fail, and Q2 heatsink temperature starts to climb rapidly.
• I have not directly confirmed, but it seems like D19 (or something near it) is also getting hot during this time
• Importantly, this doesn't happen if you start the system with the tip disconnected.

So this made me scratch my head because the microcontroller does properly display "TIP FAILED" immediately after you pull it out, so it should not be driving the signal reference into the amplifier. I believe what is happening is that the inductive components I'm using are significantly lower loss/higher saturation current than what Sergey used, and this is allowing the amplifier to start to self-oscillate once D19 opens up.

I did however find a surprisingly simple workaround for the short term: Connecting the hot side of C63 to the EN pin on U2 will shut down the DC-DC converter when the tip is disconnected. This is so simple and obvious that I'm starting to question if there's a really bad downside to it that caused the original designer not to do it.

[rfmerrill]

As I suspected, the drain voltage is coupling throught Q2 to its gate, causing self-resonanceoscillation. I'm now scratching my head wondering how this could have worked for Sergey, maybe the saturation and hysteresis loss in the inductors he used dampens it? Or maybe I just have it tuned wrong.

[rfmerrill]

One thing I was scratching my head about, I think I've solved.



I was wondering what the point of D15 was, as it's not possible for current to flow the other direction anyway (as it would instead flow through the body diode of Q2 well before exceeding the clamping voltage of the TVS).

I did a quick google and confirmed one of my hypotheses: This is a neat trick you can use to reduce the effect of parasitic capacitance from a Zener or TVS diode. It seems like there are a few different factors that accomplish this:

• The parasitic capacitances of the two diodes form a capacitive divider and an overall lower capacitance
• The DC voltage of the middle node will quickly settle on a value such that the rectifier diode stays reverse biased (the positive peaks of either side line up, and since the middle node has a smaller AC voltage on it, the top node will then stay <= the middle node).
• With a DC bias on the TVS diode, its parasitic capacitance decreases

[rfmerrill]

Side note: Am I breaking any etiquette by making update replies to this? I don't want to annoy anyone.

My weird Russian toroids finally arrived, only to find out that they're the wrong weird Russian toroids

Thanks to a friend I met in a Telegram chat who is familiar with RF, speaks Russian and lives in eastern Europe, I have learned more about the toroids he used for the current transformer and the two drain chokes: In the original Russian version of Sergey's blog post, he in fact mentions the cores are M50BH, but for some reason the translation replaces that with just their permittivity.

[KE5FX]

Side note: Am I breaking any etiquette by making update replies to this? I don't want to annoy anyone.

Not at all, it's good to see where the proverbial bodies were buried at Metcal.  There's a lot of interesting stuff in there.

[rfmerrill]

Well, to be fair, I'm only reverse engineering sergey's design, which was based on him reverse-engineering Metcal's design XD

[rfmerrill]

Just did some more experimenting and measurements, and I've concluded that the number of turns specified on the three iron powder toroids is low enough that you can't calculate the inductance using the AL value

Since I have micrometals T130-6 instead of the cheap aliexpress ones that Sergey used, I've had to change the turn count and I've determined it's fairly consistent:

L9 - 3 turns, ~240 nH
L10 - 5 turns, ~440 nH
L11 - 6 turns, ~540 nH

L9 is the only one that you can't get reasonably close to the specified values. To compensate for that, one 100pF cap is replaced with 47pF in the first parallel array, resulting in a very similar response to Sergey's values at the two critical frequencies (13.56MHz and its first harmonic 27.12 MHz).

[rfmerrill]

Thanks to my Ukranian friend, I found this component library that contains everything you need to run Sergey's LTSPICE models: http://bordodynov.ltwiki.org/


With those simulations and some modeling calculations I have done, I am pretty confident that the drain choke inductance can be increased up to double with very little change in the behavior of the circuit:

(click for big)


Basically the coupling capacitor impedance needs to be much less than the load impedance and the choke impedance much greater. You only really need to calculate impedance at the fundamental because the condition gets easier at each harmonic. If that is satisfied:

• The current through the choke can be treated as approximately DC, its AC fluctuation is small compared to its DC value
• The voltage across the coupling capacitor is also essentially DC.
• The peak to peak value of the current through the coupling capacitor is approximately proportional to the current through the choke for a given load. The load can pull some additional current backwards through the FET if it's reactive enough, but this is small compared to the contribution from the supply.
• Since all of the power that is dissipated or used in the circuit comes from the supply, the current through the choke times the supply voltage must equal the average total power (mostly in the load and the FET).
  

I'm not 100% to completely understanding this but I'm psyched that I've made this jump.

[Free_WiFi]

Look at here :

[rfmerrill]

Look at here :

Yeah that could totally be a problem, but it seems to work for the most part anyhow. Are you saying that is what is causing the self-oscillation?

If it were up to me I would have not placed all of that so close together.

[rfmerrill]

From Sergey's LTSpice models (specifically rf4.asc) I finally found a simple simulated load circuit that works.


(click)


The resistor can be a 50 ohm RF dummy load off of a coax if you want. The inductor can go higher, it is just there to provide a DC short to fool the tip detect circuit.

Now if anyone knows why an inductive soldering handpiece has an impedance that looks like a series RC circuit (which gibes with the scope traces I took) that would be interesting to know. Update 2021-03-01 -- this was answered in the mamalala thread a while back: There's an LC matching network in the handpiece assembly.

[rfmerrill]

Look at here :
*snip*

You know, I just realized you might be on to something because the VHF chokes I substituted are solenoid inductors and thus have an open magnetic circuit, so it's likely there IS coupling going on there, which Sergey would not have experienced because he was using toroidal core inductors with a closed magnetic circuit.

[richard.cs]

Have you come across this thread?
https://www.eevblog.com/forum/projects/diy-metcal-13-56-mhz-rf-supply/

It's a different design to do the same thing, but quite a few people on eevblog (including myself) made one. My write-up is here: http://randomfunprojects.co.uk/metcal.html

[rfmerrill]

Have you come across this thread?
https://www.eevblog.com/forum/projects/diy-metcal-13-56-mhz-rf-supply/

It's a different design to do the same thing, but quite a few people on eevblog (including myself) made one. My write-up is here: http://randomfunprojects.co.uk/metcal.html

Huh... kinda wish I knew about it before I got so far into building this one! XD

It looks like sergey may have referenced that design a little uncredited, or maybe not.

Does it reach 80W?

[richard.cs]

Does it reach 80W?

From memory I had something like 100W out of it into a 50 Ohm dummy load when I was building it. In any case it works well with all the tips I have tried it with, and is not noticeably slower to heat than the genuine Metcals I have used at work.