Tested and working feedback induction heater driver with current limiter

[radbloke]

Hi all!
I have developed a driver that will run an induction heater at its  resonant frequency. The reason for the driver is to enable resonant action of the induction heater tank circuit and to allow  IGBT switches to shut off on a falling or rising edge during overcurrent situations for ZVS operation. This driver can sample the tank voltage. The tank voltage is passed across a smaller capacitor say 10 to 100nF.  For example, if the tank is 4.0 uF connected to 100 nF capacitor some distance away, if there is a 500 A flowing in the primary tank, then the current flowing in the smaller capacitor would be (0.1/4.0) x 500 =12.5 A. If a small ferrite current transformer with 40 turns magnet wire is coupled with one of the legs of the smaller capacitor, then the total current flowing in the current transformer would be 12.5/40 equals 312.5 mA. If this is passed across a 56 Ohm burden resistor, then the voltage across the resistor would be 0.3125×56 equals 17.5 V. This voltage is used for feedback.

This is fed across opposing 9.1 V zeners and clamped to 5V with opposing 5.1 V zeners connected to ground. The signal then goes to an 74HC14 hex inverter and twice inverted.  A separate weakened signal above the tank resonance from a stable 555 is also passed to the same input of the hex inverter. This weakened signal causes low, current nonresident oscillations in the tank circuit, which allows enough current to enable resonance to start and once resonance kicks in the weak signal from the stable 555 is overridden by the stronger resonant signal which takes over resonant action. This is much like an SSTC. The reason for the 9.1 V opposing diodes is to prevent the attenuated startup signal from the startup astable 555 from shorting out through the current feedback transformer. Without the opposing 9.1 V (or higher) zeners all the weak startup signal is lost and spontaneous resonance won’t start. Both the weak startup detuned signal and resonant  signal go to the CLK of a D flip-flop. Once resonance starts only the resonant signal makes it to CLK on the Flip flop.  If CLR on the flip flop is low ( via 10 k pull down resistor) the resonance will continue normally. In an overcurrent event, the bistable 555 outputs high from pin 3 which makes CLR go from low to high.

This causes output on Q* to go from low to high. As UCC27424 enable pin is active low, this disables the UCC27424 and shuts off output for the  time period determined by the bistable 555.  This allows current in the tank to drop. Normal tank resonance can resume until the next over current event where the cycle will repeat

The output from the  ( UCC27424 or UCC27425) is 5 V. This drives high current gate drivers such as the 14A   IXDN and IXDI614 non-inverting and inverting driver chips set up in a push pull arrangement to run a gate drive transformer.  The signal can also run paired 40A FDD8424H  N and P channel power trench drivers in push pull mode for gate drive transformer operation if using UCC27425 which has 2 active high inputs and separate inverted and non inverted outputs to run the N and P modules as push pull.

The UCC27424 or 27425 enable pin can be activated (active low) during overcurrent situations. When there’s overcurrent, the voltage across a resistor divider from rectified tank voltage biases the gate of a NPN transistor which sinks pin 2 on bistable 555 to output high signal on pin3 of the 555. This  high signal persists for several seconds depending on the size of the resistor and the capacitor connected to the 555. This signal is used to disable the UCC27424 as above.

The driver works great and I have built 2 induction heaters with it driving SKM100 GB bricks in full bridge configuration.  These are larger induction heaters with series resonant tank and ferrite coupling transformer. These units operate at 50-70 kHz. And the current control works perfectly. I have the Gerber files if anyone is interested in building this.

The main reason for this post is the fact that the opposing 9.1 V Zeners have a small reverse leakage current ranging from 1 uA to 10 uA. Although this is small it is not a problem at lower frequencies. However as the attenuated startup signal goes above 70 kHz, this becomes a problem with loss of the signal in the current transformer and no available signal to start resonance. Any suggestions to alleviate this issue would be much appreciated. See fig 1 for the circuit diagram

[davekni]

I'm not understanding the issue.  How is 10uA leakage current making any significant signal on 28 ohm CT burden resistance?

BTW, why not just make the HC14 self-oscillating with a resistor from pin 2 to pin 1 (or to bottom of R14)?  There are several posts about self-oscillating versions of SSTC and DRSSTC drivers.  This would replace U9.

[radbloke]

Using the 74HC14 oscillator instead of the 555 for startup oscillations is a great idea- I think would work great! Here’s is the modded circuit- what do you think?

With the original posted setup I have no problem with the startup signal being lost at higher startup frequencies when the Feedback transformer is disconnected from the circuit. I think it is related to the high capacitance of the larger 5W opposing 9.1V Zeners.  Will try and add a series resistor after the burden resistor so that I can use smaller 1 W Zener.

[Anders Mikkelsen]

Nice work and documentation! I've been following this project via your youtube videos.

For the zeners causing trouble with your current sense signal, this could also be related to their capacitance. When the current is too low to forward bias these diodes, still they will pass some current through their stray capacitance. This is in the ballpark of several milliamps, terminated by the 5k1 output resistance in series with your 555 output. This will already give a signal strong enough to impact the switching, but also with a phase shift due to the RC nature of the circuit. Once the voltage is high enough to forward bias the input diodes of the 74HC14, it will be terminated by the 100 ohm series resistor, giving a different phase shift again, so this could potentially cause some strange behavior during startup. My recommendation is to not have series zeners, but to use the current sense signal directly into a self oscillating circuit like Dave mentions, to avoid these potential issues.  I'd recommend putting a resistor in series with the signal from the CT + burden resistor, then you can use smaller diodes with less capacitance to clamp it to the 5 V supply rail. Something in the 1k ballpark would be my recommendation.

How many turns do you have on your CT? It could potentially be saturation of the transformer messing up the current sense waveform, which could be checked by looking at the waveform across the current sense resistor.

For self oscillation around the 74HC14, have a look at this https://4hv.org/e107_plugins/forum/forum_viewtopic.php?p=1&id=36475 induction heater project. It's an LCLR but the principle is the same. Using a smaller cap across the tank cap for current sensing is also mentioned here on page 4 for the first time.

I would recommend reducing the stray inductance in the wiring between the tank and the sense cap, and also in the CT secondary circuit. Both for reasons of avoiding stray resonances and to minimize noise pickup, based on my own experience. This can be done by twisting the wires and keeping the loop area as small as practical.

[davekni]

Here’s is the modded circuit- what do you think?

No, that has two HC14 outputs coupled together.  The upper one through 5.1k will generate insignificant voltage on other output.  Also, the idea is to make the existing HC14 receiving CT signal self-oscillating, not injecting a signal from another HC14.  Here are a few posts for self-oscillating SSTC circuits:
    https://highvoltageforum.net/index.php?topic=2338.msg18312#msg18312
    https://highvoltageforum.net/index.php?topic=840.msg19316#msg19316
    https://highvoltageforum.net/index.php?topic=2026.msg18969#msg18969

Separately, your 5V zener diodes are clamping input to about +-5.6V.  That's 11.2V peak to peak.  If intention is to avoid HC14 input body diode current, input needs to be close to 0-5V (5V peak to peak).  Clamping is more often implemented as diodes to ground and +5V on the HC14 side of coupling capacitor.  Schottky diodes can be used to reduce clamping voltage.  Or normal silicon signal diodes as in above schematics combined with resistor from diodes to HC14 input to minimize body diode current.
Alternately, sometimes anti-parallel silicon diodes are used as in UD2.7.  Results in much lower clamping voltage, fine for comparitor-input circuits, but might need multiple diode pairs (or LEDs or other higher-voltage diodes) to get closer to 5Vpp.

Coupling capacitor is much smaller in order to achieve desired self-oscillation frequency.  Coupling capacitor combined with resistor from HC14 output forms RC timing pair.  However, small coupling capacitor combined with clamping after capacitor results in phase shift depending on amplitude (on amount of clamping).  So either reduce burden resistor to minimize need for clamping or keep clamping where it is and reduce voltage.

[radbloke]

Thanks a bunch. That is really helpful. Actually, I did see the error in my posted hand sketched circuit and have redone it and just tested it with the set up as shown in the diagram. This seems to work. I can now do away with the astable 555 pretty much. Will do that on a future build. Here is what I have. The opposing 9.1V Zeners ( anything greater then 5 V) do seem to be important to prevent loss of the startup signal at higher frequency startup. When I took them out, there is loss of the startup signal. I’m waiting on the smaller value 1 W as these seem lest apt to lose the startup. The 330 nF coupling cap value is from trial and error at the 50kHz to 90KHz resonant frequency range. Do you think I could do away with this since it can result in phase shift, or should I ichange its valuei? The second circuit below on the lower part of the same sheet proposed and not tested.

Thanks

Brian

[davekni]

The opposing 9.1V Zeners ( anything greater then 5 V) do seem to be important to prevent loss of the startup signal at higher frequency startup. When I took them out, there is loss of the startup signal.

Yes, with the circuit you have, feedback signal needs to be isolated at low amplitude.
When feedback gets to high amplitude (>+-13V or so), HC14 input current to internal body diodes will be significant.  Limited by excess voltage divided by 100+56 ohms.

The second circuit below on the lower part of the same sheet proposed and not tested.

I'd expect it to work.  HC14 input current will be much lower.  1N5819 diodes add some capacitance, so might require reducing 7.5k resistor value some to avoid attenuating startup signal too much at high frequencies.

The 330 nF coupling cap value is from trial and error at the 50kHz to 90KHz resonant frequency range. Do you think I could do away with this since it can result in phase shift, or should I ichange its valuei?

Don't remove that cap.  There is a 2.5Vdc offset between average voltage on into and out of cap.
330nF is likely large enough, depending on how low frequency you need to run.
Only reason for a small coupling capacitor is to avoid 9.1V zener diodes.  But small capacitor requires clamping below +-2.8V or so before capacitor (across CT output) to avoid phase shift.

Circuits I linked to in above reply use single section of HC14 for both startup oscillation and receiving CT feedback.  That avoids any jitter due to startup oscillator continuing to run after current is high enough.  They should work with either small coupling capacitor (and pre-capacitor clamping to <+-2.8V) or with zener diodes in series.  In the case of zener diodes, those diodes become the startup oscillator capacitor.

[radbloke]

Thanks David- that is very helpful advice. I’m doing a rebuild and will keep you posted. The 74HC14 oscillator works well but as you mentioned, the modified approach will remove some of the startup jitter and reduce IC components!!!

[radbloke]

I have modded the schematic using the hex inverter as an oscillator for startup. Here is how the schematic looks. I have removed one of the 555s. Wanted to see if you could spot any potential issues- thanks for helping with this project and for some great ideas.

Brian

Please see attachment for revised schematic.

[davekni]

The 74HC14 oscillator works well but as you mentioned, the modified approach will remove some of the startup jitter and reduce IC components!!!

I have modded the schematic using the hex inverter as an oscillator for startup. Here is how the schematic looks. I have removed one of the 555s. Wanted to see if you could spot any potential issues

Were you intending to use "the modified approach" to reduce jitter?  Posted schematic still has a separate HC14 section for startup oscillator.  Continues to run after startup.  Looks like a working schematic, but doesn't reduce jitter.
One more detail:  There are several unconnected logic inputs.  Best practice is to connect all logic inputs to a stable voltage, 0V or 5V or a logic output.  (Of course, don't connect an inverter input to its own output.)

thanks for helping with this project and for some great ideas.

You are certainly welcome.  Glad to be of use.

[radbloke]

I just did another run with a circuit that will get rid of the startup jitter and it seems like it worked great. So I’m gonna modify it yet again. This time there’s only one input. Check out the attachment below and I’ve also attached a video where I’m adding in “feedback signal” using a signal generator. Let me know what you think. I had to mess around with different resistances because I was not sure of the exact inherent capacitance of the  1N5819. The start up isolations are approximately 95 kHz– 100 kHz using a 120 K resistor across pin 1 & pin 2 of the hex.

Here is a video of the test:
Jitter test hex inv oscillator startup

/>

[davekni]

I just did another run with a circuit that will get rid of the startup jitter and it seems like it worked great. So I’m gonna modify it yet again. This time there’s only one input. Check out the attachment below and I’ve also attached a video where I’m adding in “feedback signal” using a signal generator. Let me know what you think. I had to mess around with different resistances because I was not sure of the exact inherent capacitance of the  1N5819. The start up isolations are approximately 95 kHz– 100 kHz using a 120 K resistor across pin 1 & pin 2 of the hex.

Yes, this is exactly the circuit I was suggesting.  If you want to make it a bit less sensitive to diode capacitance, you could add an additional capacitor in parallel (ie. HC14 pin1 to ground) and reduce feedback resistor value correspondingly.

[radbloke]

Thanks David. I happy it’s the same circuit you were thinking of. Thanks for getting me thinking in the right direction. This approach will widen the range of startup frequencies to include higher frequency induction heaters. The jittering made strange sounds in the tank circuit so this should do away with that. The added cap across the shottky should make startup tuning easier.

Cheers,

Brian

[radbloke]

This is the latest update after I finally found some time away from my regular  job.

I have updated the original posted induction heater feedback current control schematic and now have a version ( ver3) that is basically a final version. I did away with the startup 555 as suggested by davekni as well as the 5.1k attenuating resistor and went with using the hex inverter as the primary start up oscillator and with feedback signal fed to pin  1 of the hex for secondary feedback from the induction heater tank circuit.  I will show what I have in this posting. After getting a PCB made I am running a test here. So far it has worked well at lower power levels and I plan to run the test at higher power mains input at a later time. As this idea for jitterless startup was suggested by Davekni, all credit goes to him.  This method also provides a much more stable startup frequency and smoother transition to resonance.

There is also a significant improvement in startup freq range. Here is a test demo of the circuit in action:

/>
Check out the circuit. I found the test setup does not need an extra series resistor between the output of the small current transformer/burden resistor and the feedback input. There is no heating of the opposing 9.1V zeners or the 1N5819 clamp resistors at least at lower test power levels.
IH Current Control Ver 3.pdf


Ok here is an update. I finally tested melting about a couple of pounds of copper scrap, and also tested it melting a lb of iron  iron scrap, bringing the iron scrap beyond the Curie point. With copper scrap, the current draw was 13 A at 180 V DC. The opposing zeners hit  50°C with this kind of current draw. The 1N6819 clamps hit 40 C. That was with no resistor between the current transformer/100 ohm burde resistor combo and the current sense input. I then put in a 100ohm resistor in series between the CT/ 100 ohm burden-  the same situation melting copper with 13 A draw resulted in heating of the opposing 9.1 V Zeners  to 36°C. There was no heating of the 1N5819 clamp resistors. I then switched out the 100 Ohm series resistor for. 470 ohm series resistor which fixed the heating issue.The opposing IN5819 diode clamp  did not heat up under this situation. After the melted copper solidified and became cool, I removed it and then placed into the coil approximately 1 pound of iron scrap and brought it beyond its Curie temperature so that it fused together as one big lump. The maximum current draw during this run was around 27 A. I checked the temperature of the opposing zeners and they were around 35°C well within their specs. So I think I have a working set up. The only difference between what I did with the higher current run and the previous low current test run is adding a 470 ohm resistor between the 20 turn frrrite coupling transformer with 100 ohm burden and the opposing zeners on the attached circuit diagram.

If anyone has any suggestions for further improvement, please let me know. Also, if anyone is interested in experimenting with this design, I can post a link to the Gerber files.

Thanks

Brian