Well, years have past. My unit still works like the first day
The only thing needs to be replaced is the old little fan I recycled form old computer (although spins fine
).
I used 2 tips in total (today using the second and the first broke when it fell to the floor), ~4h/day use. Fast sleep function of the code I wrote really protects the tip despite fast heating (<4s).
7 years
And thank you mamalala for starting this project.
@hww How much did it cost you? If you can be detailed that'd be great! Also, How did you build it? any guides?
I have been working on this (one weekend per month or less) for about a year. At the end of this time I can only make an estimated bill.
Iron ~39 €
Iron tip ~26.26 €
Transformer ~21 €
Aluminum sheets ~20 €
PCB + coils ~30 €
The remaining items were not significant, probably less than 50 €
The guide is in this forum. If the PCB is assembled correctly the circuit does not require any adjustments or tuning. Just follow a few steps from the guide.
What is the difference of this DIY with all the products listed here:
[Sorry if the questions are dumb!]
The main difference: Those products you can buy, this project you can build. Also, some of those products use 450KHz while others use 13.5MHz, and as result they have different irons and tips
Have you managed to keep EMI at a reasonable level?
That would be one of my main concerns here.
Have you managed to keep EMI at a reasonable level?
That would be one of my main concerns here.
Please tune the impedance of your transistor/s and use an metal enclosure for the final assembled station and you'll be fine.
Dear Colleagues,
I have ordered a few sample PCBs from JLCpcb in order to replicate this project. The BOM is almost straightforward and clear, but the custom made inductors. Could somebody offer me a more human understandable description about them? Like the material of the core - and how could they be replaced with other types if possible - salvaged from other SMPSs?
Thanks in advance,
Tibi
Could somebody offer me a more human understandable description about them? Like the material of the core - and how could they be replaced with other types if possible - salvaged from other SMPSs?
The thread is long, though there have been various attempts to group the information. I would recommend using the cores specified and then it's just a matter of winding the right number of turns. From memory I think the winding instructions were in one of the word documents describing the assembly process. If you want to change cores then you need to know much more what you're doing, but if you must change to due limited availability then I would strongly recommend purchasing cores with a known specification. Working with salvaged cores is a problem because you don't know the material characteristics and you end up having to make a number of measurements on each core before you can decide if it's likely suitable or not.
Could somebody offer me a more human understandable description about them? Like the material of the core - and how could they be replaced with other types if possible - salvaged from other SMPSs?
The thread is long, though there have been various attempts to group the information. I would recommend using the cores specified and then it's just a matter of winding the right number of turns. From memory I think the winding instructions were in one of the word documents describing the assembly process. If you want to change cores then you need to know much more what you're doing, but if you must change to due limited availability then I would strongly recommend purchasing cores with a known specification. Working with salvaged cores is a problem because you don't know the material characteristics and you end up having to make a number of measurements on each core before you can decide if it's likely suitable or not.
Thank you for your answer. I wanted to salvage cores since i could not find nor on farnell or mouser the corresponding cores. Do you have any ideas about where i can source them from?
This thread seems to be winding down... sorry couldn't resist
I'm trying to re-use the toroids from a Metcal PS2E supply (similar to MX500) I bought. The PS2E had too many issues and this thread encouraged me to design one from scratch. I built a prototype loosely based on the circuits here. Draws a lot of current, it doesn't seem right. I've sim'd so many circuit topologies including output filter, and hoping to build another prototype. So, I'd like to know if anyone had an opinion on using the PS2E toroids which are T50-6. They seem pretty small, I wonder if the original Metcal was limited to 25W or so. I measured them with an LCR meter and they are 1uH, 0.88uH, 0.33uH. Also measured the transformer which reads 6.3uH for both primary/sec.
The topology that is interesting uses two xformers and is driven push-pull. See attached. The output filter is similar to Sergey's. Produces a large output, current draw from power supply is quite reasonable. Thinking about re-winding the toroids and hacking my proto board.
Thank you for your reply, Richard. Are T50 for the output inductors OK to use, or should T80's be used? They are apparently supposed to pass a couple amps of current, without saturating. I am able to calculate inductors, wind them, measure them. I'm curious if the 3 filter inductors ought to be 1uH 1uH 0.4uH or 0.2uH 0.4uH 0.5uH. When I simulate, the latter order seems to work better.
I built mine as per the schematic and notes, with 1uH, 1uH, 400n, 11 turns on a T80, 14 turns on a T50 and then 10 on the next T50. It works well and I've made no attempt to optimise it further.
There is a whole load of design documentation buried somewhere in the thread but I've not managed to find exactly where. I have local copies - PM me your email address and I'll send you what I have, there are schematics, BOM, and a build guide that includes the winding instructions.
Just got mine working today. The main FET was the problem all along. The transistor must have low Ciss and Qg. I had FETs that were much too capacitive and caused supply to draw amps of current. My proto board uses a 13.56MHz xtal, 1GX04, ADP3654 MOSFET driver, IRFR110 Nch (DPAK) to drive the coupled inductor/transformer. I took out the last LC section and have only two sections. Haven't yet hooked up rf-detect from 2nd tap or feedback path, just running it 'wide open'. Proto board is 2.5" x 3". Inductors pulled from broken Metcal PS2E supply.
Here's what I find. My switching reg is disabled, and I'm using a bench supply (xtal osc and driver regulated to 5V). The Metcal wand holder is magnetic and when solder hand piece/tip are placed in the holder, supply current goes to 220mA (Vs=14V). When hand piece is removed, current jumps immediately to 570mA and tip starts to heat up. To melt solder, Vs needs to be 16V or higher. At Vs=20V, current is 810mA. Wow this is quite reasonable, I can easily solder anything and the power is only 16W! STTC-126 tip. Surprisingly low current, low power. What I also find interesting is that after the tip heats up, it sort of self regulates and doesn't thermally run away - at least the supply current is quite stable. The FET (DPAK) doesn't have a heatsink yet but it gets hot when I remove wand and Vs up to 18V. Case is 90C. At idle (Vs=14V), case temp is low 50s. Might just attach a small heatsink with fins.
I had planned to use a 24V 2A brick supply, plug into a DC jack. That's still the plan. I had also planned to use a small plastic case from Amazon. 3" x 2.75" from Eightwood.
Yes, I remember that various people tried substitutions for the main FET and several types that looked promising on paper turned out to not work that well. I think part of it was that the selected device actually exceeds its datasheet specs by quite a bit.
The design of the filter is intended to make the output approximate a current source by taking feedback from the voltage at a particular node. I just took it for granted and never actually analysed how that part works. What I can tell you is that the current source behaviour is required because the tip impedance when hot is such that if it were fed from a conventional 50 Ohm source the power would only be halved. I made some tip measurements which are here:
http://randomfunprojects.co.uk/metcal.html which shows S11 from 50 Ohms is about -20 dB when cold but -3 dB when hot.
Thank you Chris for such an exciting project. I am planning to make this RF generator. However I'd like to incorporate all the advancements onto the PCB. And probably create GitHub repo with all the project history for easier tracking and further development.
The thing I have concerns about is the RF output part. Could somebody give me a hint about designing such waveguides (I assume it is a waveguide? Or is it a part of matching network?). Any good articles or books about purpose, and calculating those parts? I'd better approach PCB redesign/improvement when I understand how all parts work. Other than this RF out trace all parts are relatively easy to understand and check for relevance according to main components datasheets and overall theory of operation. But RF is a bit of a dark area for me in some contexts.
EDIT: OK, I have checked the idea that this wavy line on the RF output of PCB is a waveguide and came with the results of a small research, controlled by online calculators, that this is a grounded coplanar waveguide with an approximate impedance of 50 Ohms, if PCB is manufactured on 1.6mm FR-4 epoxy. What is left unknown is the question of the length of the waveguide. According to theory, the inductance of the waveguide is almost constant with the length, however the capacitance is accumulating with the length. But is there any significant need in this capacitance? Any other ideas?
EDIT: OK, I was looking at unpoured polygons in eagle and missed, that there are special traces along the waveguide and these combined together make up an SWR bridge. Now it totally makes sense and I understand all the parts of the PCB.
I have also found few improvements at russian forum VRTP, that help to keep power FET alive and some tweaks of parameters. Will include them also in the combined design.
-Den
Hi!
I'm very interested myself in this D.I.Y. 12.56 MHz Metcal Soldering Tool R.F. Supply and would definitely like to try one myself in the New Year!
I am writing a very large book on Test & Measurement Equipment Repair, does any of the Members who've designed this project have any objections to me including a set of plans, which would be my own format drawings, PCB layouts and assembly documents, based on those provided, as a Chapter in my T & M Theory & Repair Book?
An acknowledgement of sources will be on the front cover of the Chapter!
(The Book is being written and made available for free download under the CC–SA–BY 3.0 Creative Commons Licence, so there'll be absolutely no question of monetary gain from it!)
Chris Williams
I haven't had time to read this entire thread but: I'm curious why people have been waiting to pool PCB orders. Is this not a design you could simply send to OSH park to get built? Does it require something more advanced than what standard PCB manufacturers tend to provide?
I'm reading the only schematic PDF of the RF board I could find, and I'm curious if there has been an updated one (this one is dated 2012) and also to some extent how it works. Sergey's design uses a current transformer and something like a ring mixer to generate a feedback signal to the DC-DC that depends on both the magnitude of the output voltage and the phase between voltage and current.
This one appears to purely use the magnitude of a voltage sampled just before the last filter stage. I'm kind of a novice when it comes to RF so I'm wondering if the two are basically equivalent or if there are advantages.
And then I also wonder if the SWR bridge is just for testing purposes? what I read is that the RF board can function just fine without it.
I have also been working on a test load that can be used in place of a soldering iron tip to stress test or tune either power supply:
(click for big)
I am also thinking about adding a sampling port there to look at the output power (without fancy RF power circuits).
I haven't had time to read this entire thread but: I'm curious why people have been waiting to pool PCB orders. Is this not a design you could simply send to OSH park to get built? Does it require something more advanced than what standard PCB manufacturers tend to provide?
No particular reason or PCB complexity, it's just still cheaper per-board to pool orders, though perhaps not as much now as in 2012.
I'm reading the only schematic PDF of the RF board I could find, and I'm curious if there has been an updated one (this one is dated 2012) and also to some extent how it works. Sergey's design uses a current transformer and something like a ring mixer to generate a feedback signal to the DC-DC that depends on both the magnitude of the output voltage and the phase between voltage and current.
This one appears to purely use the magnitude of a voltage sampled just before the last filter stage. I'm kind of a novice when it comes to RF so I'm wondering if the two are basically equivalent or if there are advantages.
The 2012 one is current, there have been a couple of mods (e.g. my undervoltage lockout circuit) but the 2012 schematic essentially works. Mamalala's design regulates for peak voltage at a filter node, which to my understanding corresponds approximately to constant-RF-current. Temperature regulation is then by the varying power absorption form the tip reaching an equilibrium temperature very close to the Curie point. The official Metcal ones also use constant RF current but the feedback approach is different. Sergey's design works in a more complex way - by regulating for 45 degrees phase angle he is essentially servoing the tip to a particular temperature. In principle this could give tighter temperature control but I suspect it does not make any significant difference in practice as the transition is very sharp.
And then I also wonder if the SWR bridge is just for testing purposes? what I read is that the RF board can function just fine without it.
The controller board makes use of it for "smart features", primarily:
- Plotting a pretty bargraph of real-time forward power on the display
- Detecting when the power is constant and low, and in doing so identifying that the iron is in the stand. Sleep mode (reduced power, lower temperature for long tip life) is then enabled after a configurable time.
- Detecting when the iron is removed from the stand, just wafting it though the air is detectable as a small increase in forward power (like a hot-wire anemometer). This is used to exist sleep mode, and optionally enabled a timed power boost. The end result is that by the time the tip touches the work it is back up to operating temperature.
I have also been working on a test load that can be used in place of a soldering iron tip to stress test or tune either power supply:
I did all my testing with a similar-looking load, but without reactive components.
If you are still keen on increasing output power, note that the FET runs relatively cool in the current design, and adjusting transformer to present a lower impedance to the FET may be all that is needed, for a modest output increase. Two paralleled FETs of the same type and a further impedance decrease might then be the next sensible step - there's plenty of gate drive. I very rarely see mine hit maximum power with the tips I use.
Thank you very much for the answers! If it's not too expensive I might just build this design up too.
Mamalala's design regulates for peak voltage at a filter node, which to my understanding corresponds approximately to constant-RF-current. Temperature regulation is then by the varying power absorption form the tip reaching an equilibrium temperature very close to the Curie point. The official Metcal ones also use constant RF current but the feedback approach is different. Sergey's design works in a more complex way - by regulating for 45 degrees phase angle he is essentially servoing the tip to a particular temperature. In principle this could give tighter temperature control but I suspect it does not make any significant difference in practice as the transition is very sharp.
I would love to ask a bunch of questions to Sergey to know why he did some things
He says he copied the feedback circuit from the higher-end 80W metcal supply, and it's possible that Metcal switched to that circuit simply because it was needed to handle the simultaneous output capability.
For anyone who wants to use OSH Park to buy these: They don't like the gerbers, but they will take the .brd just fine.
For anyone who wants to use OSH Park to buy these: They don't like the gerbers, but they will take the .brd just fine.
Huh? No way, I've always sent Gerbers to OSHPark and gotten beautiful results. If they are rejecting your Gerbers, then you're violating a design rule or the file format is wrong.
They may have some automatic system that tries to fix the ".BRD" files to pass DRC. However, your sources are still wrong and you shouldn't depend on tools fix it the correct way (or just cover over it).
Huh? No way, I've always sent Gerbers to OSHPark and gotten beautiful results. If they are rejecting your Gerbers, then you're violating a design rule or the file format is wrong.
I'm using the google drive link posted earlier in this thread. When I got Sergey's board made I used his gerbers and they worked fine. With the ones posted earlier in this thread (only two search results for "gerber") it complains it can't find a board outline or drill file. The previews from the .brd look ok though.
I think it's just about how the gerbers were exported more than the design itself. I would look at the brd file if I could open them in kicad but the eagle files are too old.