No announcement yet.

[OT?] Soldering Litz wire for high frequency tranformer (1500W DC-DC converter)

  • Filter
  • Time
  • Show
Clear All
new posts

  • [OT?] Soldering Litz wire for high frequency tranformer (1500W DC-DC converter)

    I am building a DC-DC converter to take 12, 24, or 48 VDC from lead-acid batteries, and provide nominal 140, 280, and 560 VDC for the DC link of a VFD. I'm using E55 ferrite cores, type N27, and I plan to use about 80 kHz square wave from a full bridge. I have some Litz wire to use for the windings: 105/40 (#20 AWG equivalent), and 7x3x21/40 (#14 AWG equivalent). I tried to wind the primary first, using four windings of 3 turns each of the #14 AWG, but I was unable to get enough heat to melt the insulation, and the wire was too large to solder to the terminals of the coilformer (bobbin). But I decided to bring the leads out for direct wiring to the PCB or a terminal strip. I was able to tin the leads to some extent by using liquid flux and the iron set to 480C, and I put two windings of 5 turns each for the primary:

    The #14 AWG wire should handle about 15 amps, and I was able to apply a 30 volt square wave to one winding with only 50 mA of magnetizing current, so it should be good for 24-28 volts from two batteries:

    I plan to add another pair of windings so it should work on 48V and 30 amps (1500W), 24V and 60 amps (1500W), or 12V and 60 amps (750 watts). However, I really need to tin the leads properly, to make sure the insulation is melted off all the strands, so I have ordered a solder pot:

    I also ordered about 2 pounds of 63/37 eutectic solder:

    But I may also get some lead free Sn96.5/Ag3/Cu0.5 bar solder for about $20:

    I found some information on liquid flux and this was recommended:

    Here is information on soldering and terminations for Litz wire:

    I was surprised at the influence of skin effect at 80 kHz. It is 10.6 mils, so any wire larger than 21 mils (#24 AWG) will have a significant increase of AC resistance. Here is a skin effect calculator:

    Any tips and techniques for using a solder pot and terminating magnet wire (especially Litz wire) would be appreciated.
    Paul , P S Technology, Inc. and MrTibbs
    USA Maryland 21030

  • #2
    I think you're on the right track with a solder pot. I read through some of your links and it seems that most of the information is there. The molten salt process is interesting- I wonder how thermally conductive molten salt is compared to solder- and I wonder also how much hotter it could be without having oxidation issues. Seems to me that the hotter it is, the better able it will be to remove the enamel from the wire. Importantly, how well will it be able to penetrate all the strands and dissolve away the enamel. From there, it will be important that the tinning process fully penetrates the wire bundle.

    The heat sink just above the soldering region is a good idea to keep the wire from becoming damaged further up, but it will also limit the penetration of the solder. Getting the job done quickly and thoroughly is good practice, as you know anyway.

    I did experiment some with an available chemical enamel removal compound (probably from MG Chemicals) and wasn't too impressed. It definitely had a noxious smell to it, and it's probable that whatever chemical actually does the work is in short supply in it, for safety reasons. With proper care, perhaps that chemical can be used in a stronger formulation. I'm not sure what it is, or where you could obtain it.

    What might work best is a multi-part process. First use this chemical, then perhaps go directly to the solder pot or a molten salt dip before the solder pot.

    When I used to work with multi-strand magnet wire I sometimes used a propane torch to burn off the insulation, but then it was difficult to solder. Perhaps if I had used a hot pot cleaner before soldering it would have worked better.

    At any rate, I would sometimes wrap some solid copper wire around the end of the litz, leaving both ends sticking out before soldering. That gave me a way of making the termination to a pc board with two wires instead of having to heat and solder the full diameter of the bundle at that point. Made it a lot easier to get a good connection to the board in a short period of time to avoid hurting the pc board.

    Another thing you could consider is separating the bundle into perhaps four individual smaller bundles and tinning each separately- then also making the pc board connections with the four smaller wires rather than one large one. Each would be about an 18 ga equivalent, which is easy enough to solder to the board. Obviously the traces would be made to suit.

    You say that 14 ga can handle 15 amps- is that at 50% duty cycle or 100%? I'm sure you know, but in this application you are probably good to assume 30 amps for that wire- and the winding is short enough to keep losses to a minimum. Leaving the wiring exposed as much as possible will help to dissipate whatever heat is generated. You could in fact begin separating the bundle into smaller bundles right at the point where it leaves the core, and that will get a little more of it exposed to air.

    When I used a solder pot I tried to keep some flux on the surface of the solder. It smoked quite a bit, so I also tried to keep it covered as best as possible when not actually dipping wire into it. Perhaps a hinged lid with a safe handle to make it more or less foolproof to open and close would help.

    Just some ideas.
    I seldom do anything within the scope of logical reason and calculated cost/benefit, etc- I'm following my passion-


    • #3
      I know this wire lives in induction cookers, how is the induction coil terminated, core or strand by strand or as a bunch?, solder pots are fairly easy to use, just dip, however when I used one to tin a load of cables I did do half a dozen test wires at different times, I used a lab bench top timer (clockwork!) 3s, etc etc the wire was ptfe insulated fly leads on photomultiplier arrays I was changing on a spectrometers goniometer ( one of the most beautifully engineered things I've ever seen that belonged in a sci fi movie)
      I just followed instructions btw as electronics at that level is partly witchcraft,
      I get what your building and am in awe of your knowledge of electronics, we had a guy I used to work with who had a PhD in magnetic effects of silicon steel as we were making the sheet steel for transformers, high silicon 3% and get rid of the carbon as much as possible as it hurts the steels properties so really it wasn't steel it was silicon iron fascinating


      • #4
        I have made a spreadsheet with various size wire and ampacities based on 0.0037 A/sq mil, and also on 0.24 W/sq in surface area. The amps per cross-section area figure is often used for transformers, and is very conservative but takes into account that the wires may be in a thick bundle, so heat must travel through many layers of insulation before being dissipated by convection cooling to the surrounding air. In fact, I found a reference that was based on 700 circular mils/amp:

        The amps per square inch surface area is intended for a bare wire in free air, which allows for a much larger current.

        According to my chart:

        #14 AWG is good for 12 amps based on cross-section area, and 15.2 amps based on surface area. The 15A matches the NEC rating. It also matches the NEC rating of 20 amps for #12 and 30 amps for #10.

        #20 AWG is good for 2.98 amps based on cross-section area, and 5.34 amps based on surface area.

        I made a tight coil of the #20 AWG Litz wire and I have 5 amps flowing through it. It is barely warm, and it drops only about 93 mV, so that's less than 1/2 watt.

        Here is the test jig I made to observe the performance of the transformer:

        There are actually four MOSFETs in the upper right corner, and they were able to drive the transformer at 30 VDC. I forget what they actually are, probably IRLM0030, which are 30V 5A in an SOT-23 package. The tester was originally used for a 2W 12V-12V DC-DC converter I made for another project. The IRS2453 is a self-oscillating full bridge driver that uses a capacitor and resistor to determine frequency, and it has its own bootstrap circuit so it can be used up to 600V. My 1500W DC-DC converter will use a PIC and higher current MOSFET drivers, and will have current feedback to shut down on overcurrent.

        One nice thing about using a square wave is that the full-wave rectified output needs very little capacitance to achieve smooth DC with little ripple. And it's possible to use two circuits with drive signals shifted by 90 degrees so the outputs overlap, and no filtering will be required (except maybe some small capacitors for high frequency bypass. But it will have to work into the capacitors of a VFD, so I may need to do some testing. I had an aluminum electrolytic and ordinary silicon diodes in my first 2W DC-DC, and they got quite hot at 80 kHz. So I used Schottkys and a small ceramic capacitor. They only got hot under load.

        Thanks for the ideas. This circuit could be useful for powering various devices from low voltage batteries. Many electronic devices use switching supplies that work well on DC as well as AC.
        Paul , P S Technology, Inc. and MrTibbs
        USA Maryland 21030


        • #5
          What you really want for this kind of soldering is one of those big old soldering irons you often see at flea markets because nobody wants them. They have a massive copper head whose purpose is to store a large amount of heat, but not at a high temperature. When you touch it to the connection it instantly gets hot and melts the solder but the heat doesn't have time to travel to the delicate parts.

          I learned about this doing high voltage splicing where we applied a very light gauze shielding over the taped splice. You had to solder the drain shield wires to the gauze without melting the tape under it. A small iron did nothing but damage.

          Wire ampacities have a lot more to do with length than people realize. In looking for a way to detect ground wires that were nearly broken off without opening up a hand power tool my dad and I tried a 3kw load at 120 volts on a ground wire that had all but one strand cut with a knife. The single strand handled the current just fine, getting slightly warm. To reliably detect this condition we had to use a dead short across the circuit breaker which could supply hundreds or possibly thousands of amps for an instant. That would burn off a single strand.


          • #6
            I use a megger insulation tester, seems to work fairly well for me


            • #7
              The thing about Litz wire is that all the strands are woven so they appear on the surface at regular intervals. That's pretty much the definition. Usually the intervals are fairly small, a few diameters or so, maximum.

              The result of that is that you really do not need to soak the end of the wire in solder, and have it wick up. That is not harmful if it happens, but is not required. If you get the insulation off the outside of the wire over a half inch or a bit more, you should have contact with every strand when soldered, for most wire. You can examine it closely and see.

              Of course, I don't know, and I don't think you mentioned, whether the wire is the self stripping type that can just be dipped. I assume not, since otherwise there would not be an issue.

              I have no idea how you plan to wind the transformer. But for that power level, I suggest that you wind the secondary first, down close to the core, to minimize the leakage inductance of the secondary. Any primary side leakage inductance will return its energy to the source, but secondary side leakage will cause the output voltage to rise for any lightly loaded secondary.

              You cannot effectively snub that power level, and the rectifier plus filter and load ARE a snubber, hard to do better than that. You will need to provide a minimum load to keep it snubbed, which can run to some significant power.


              You might consider changing from Litz primary to copper strip. It occupies a lot less winding window, and lowers the leakage. If you make the strip as thick as you can vs the skin depth, you can get very decent area, and good heat dissipation. Plus lots less trouble terminating the "wire".
              Last edited by J Tiers; 09-26-2016, 10:42 AM.
              CNC machines only go through the motions


              • #8
                This is supposed to be the solderable type, probably polyurethane. The flux seems to turn the strands dark red, indicating some chemical action even at low temperature. Using the iron at 480C, bubbles of black stuff can be seen coming from inside the bundle, and can be scraped off, showing a pretty good wetting at least on the surface to some depth. Cutting the tinned end shows a variable amount of unwetted strands, after soldering with the iron. I think the solder pot should do the job nicely.

                I considered using copper sheet or tape for the primary, but the adhesive foil tape commonly available is only 1 oz (1.2 mil) copper so a 1" wide winding would be equivalent to about #18 or about 5 amps. I know it is available from specialty sources, and I have a friend who makes small transformers, and he has given me some small pieces, but I think the Litz wire will be OK for my purposes.

                Leakage inductance will probably not be a problem. It will cause poorer regulation, as you say, but that also provides some current limiting, which may be helpful. Snubbing does not seem to be necessary, even for an unloaded secondary. This waveform shows only a tiny bit of ringing at turn-off. You can also see the 1 uSec dead-time:

                The same basic circuit, for my much smaller transformer, shows a lot of ringing at 35 kHz, but it cleared up at 67 KHz:

                Paul , P S Technology, Inc. and MrTibbs
                USA Maryland 21030


                • #9
                  Originally posted by PStechPaul View Post
                  Leakage inductance will probably not be a problem. It will cause poorer regulation, as you say, but that also provides some current limiting, which may be helpful. Snubbing does not seem to be necessary, even for an unloaded secondary. ...
                  I expect that to change significantly when you get more wire on there for the high voltage outputs. More turns will give lots more inductance, and I would not expect you to do better than about 1% for leakage inductance. BTDT, at a somewhat higher power level, of just over 2500 watts.

                  If you wind the secondary over the primary, the leakage will be larger than if you wind the primary over the secondary. it's the secondary that has the leakage problem, because the leakage energy pumps up the output voltage. I would generally expect 110% to 150% voltage at very low load, depending on leakage.

                  We'll see, just trying to save you some trouble.
                  CNC machines only go through the motions


                  • #10
                    This is largely a learning exercise, so I think I will proceed as originally conceived, with secondary over primary. I found some information that suggests sandwiching the secondary between an inner and outer primary, but that can lead to unequal impedance of the primary windings, which could cause problems if connected in parallel. Not so much in series:


                    According to that, ringing transients and overshoot increase with load, so my open circuit waveforms may not be representative of those that may be present under load.

                    Some other references:
           (mentions primary over secondary)


                    Does it make a difference if the transformer is step-up (like mine), or step-down (as are most switching supplies)?


                    BTW, I resoldered the ends of the Litz wire and it looks like the solder covers most of the strands:

                    (sorry for the blurry image)
                    Paul , P S Technology, Inc. and MrTibbs
                    USA Maryland 21030


                    • #11
                      The solder pot is the way to go. I used to work with this stuff pretty frequently in applications similar to what you are doing. A bit of RMA flux can help as well, but most magnet or Litz wire that has a solderable (is that a word???) film insulation has a film that is formulated to act as a flux at soldering temperatures.


                      • #12
                        Yes, if you think about it, the energy stored is related to the current, at 1/2(L*I)^2. So with more energy, you get more effect.

                        What happens with rectifiers is that you are drawing little AVERAGE current at low load, but the SPIKE current tends to be high during the short conduction period. Can be worse with a square wave, which applies more volt-seconds per second than a sine. The energy stored is related to the peak current as conduction ends, and that is what gives the problem.

                        With enough load, the extra energy just goes to the load, but with a low load, the spike current into the capacitors loads up the inductance, which then needs to dump the energy. Because it is LEAKAGE inductance, by definition it is not LINKED to any other winding, so you have to get rid of it right there, where it is.

                        The relation of the energy transferred to the load, and the energy in the leakage determines what voltage the load gets driven to. At some point there is a balance, and the voltage tends to hold there, at an elevated but limited level.
                        CNC machines only go through the motions


                        • #13
                          I measured the transformer as it is, and the primaries are just about 128 uH, while leakage inductance is about 2.7 uH, or 2%. No surprise there. The core is specified as 5.5 uH for a single turn, so 5 turns would be 137 uH. Close enough. The waveform shown above is for open circuit. So, the square wave of 30 volts into 128 uH should have a peak current of 1.46 amps after 6.25 uSec (80 kHz). This energy is 137 uJ. The energy in the leakage inductance would be about 3 uJ.

                          Here is a simulation of the circuit, although at 50 kHz:

                          Note that the peak current is about 3.4 amps, which is close to what is expected for 48 volts and 50 kHz. Also you might be able to see the slight blip on top of the waveform during the transition, which is also present on the actual waveform. This might be due to the leakage inductance. I did not model the parallel capacitance, but I doubt it is more than a couple hundred pF for 5 turns.

                          Now I added a FWB and a 500 uF capacitor with a minimal 10k load:

                          Note that the "blip" is wider, but still not really significant. There is some transient oscillation at the zero current point of the common current in R3, and also at the peaks. With a 10 ohm load, this is the waveform:

                          The current waveform is unbalanced, but basically as expected. There is definitely some ringing on the output during the dead-time. Here is a zoomed look at it:

                          [edit] I found that I had the on times of two of the gate drives longer than the others, which accounts for the asymmetry. I had done so previously to see the effects of an imbalance in the PWM waveforms. Making them all 9.5 uSec virtually eliminated the oscillations on the transitions of the output waveform.
                          Last edited by PStechPaul; 09-27-2016, 01:03 AM.
                          Paul , P S Technology, Inc. and MrTibbs
                          USA Maryland 21030


                          • #14
                            This is the waveform with balanced drives:

                            Paul , P S Technology, Inc. and MrTibbs
                            USA Maryland 21030


                            • #15
                              Your kicker is going to be the secondary inductance. With 48V in, and 280V out, you have about 34x the inductance, and thus 34x the leakage inductance.
                              CNC machines only go through the motions