Induction Heater schematic modification

[ritaismyconscience]

I found this schematic online:

(There are a few errors, like one of the diodes is backwards and one of the optocouplers is missing +15V.

Because I am a poor student and I don't want to wait for parts, I made this schematic which uses only parts that I have. Basically, I got rid of the optocouplers, which I think act as level shifters. I also added a negative voltage generator because I want to run it off a single 15V supply. 


I don't have a lot of experience, so I would appreciate if someone checked my schematic before I turn it into a PCB. Thanks in advance.
 

[T3sl4co1l]

Don't use slow rectifiers for D5/6, use 1N4148.

D4 isn't really doing much.  It will draw most of the current through R10, that would otherwise flow entirely into U3's ESD diodes (and in turn from +15V*), but a schottky (e.g. BAT85) will do better.  Can also place a series resistor from D4 into U3 (say 1k) to ensure current takes the intended path.

*Little known fact about CMOS ESD diodes: they act like BJT emitters, with the base tied to VSS and the collector to VDD.  A fraction of the ESD current is drawn through the base, but the remainder is "cascoded" straight to VDD.  This can greatly increase the power dissipation, of a chip that otherwise shouldn't really be dissipating anything at all.  Activation of ESD diodes can also interfere with internal functions; you'd have to test the CD4046B (or whichever version you're using) to see if anything else happens.  In short, ESD diodes can be used on a repetitive basis but should be used sparingly and at low currents (< mA).

D3 is clamping the tank input to about -0.6V.  Is this intended?  Why not symmetrical? (i.e., to -V, with a TVS diode on that supply to ensure it doesn't get cranked way down.)  Also, if there's no shunt resistance, just the series resistors (which I suppose are for current limiting), then does U5 really need gain at all (it's set for G = 101)?  Seems like it could be an LM311 instead, or not used at all (just cap-couple to U3)?

The output driver is fine, but you may want an enable signal into it, to turn the inverter on/off and provide an input for fault protection (which in turn can be something like peak current detect).

The big problem though, is that it's open loop.

1. Why use a 4046 if you're not using it as a PLL?
2. Why drive the 4046 VCO from a -- very slow -- PWMDAC?
3. Why read the 4046 PC from a -- very slow -- ADC?
4. Why not do it entirely digitally?  Granted, this may not be very easy in the Arduino framework, but an ATmega is more than capable of generating the signals used here, and can respond to them much faster (and can be reprogrammed much faster than the wiring on a PCB).
5. Why not measure PC with a timer routine?  (Again, I don't know how difficult this is in Arduinoland.  I've got to imagine someone's written a routine or library for a function this basic.  Incorporating that function into the overall program, I don't know.)
6. Why not any voltage or current measurement, or controls?  (Okay, you want to keep it simple for starters.  But you also want to be able to expand later, and expanding in very likely directions is worth some thought beforehand.)
7. Why not any PWM?

The last one is actually a two-parter:
a. You need some dead time between gate drives, otherwise the inverter transistors may be on simultaneously, and that's a problem (assuming the usual, voltage-sourcing inverter, and depending on what exact gate drive circuit is used).  Normally, a small delay is introduced to one of the switching edges, effectively reducing the duty cycle of each channel, and this sets dead time.  Alternately, you can use a PWM generator like TL494, which has a maximum duty cycle (per channel) of 45% or thereabouts.

TL494 doesn't have a PLL.  It is still a VCO.  As you aren't using a PLL right now, it would be viable from where you are now.

b. There's actually a good reason to run at maximum PWM% all the time, depending on the type of inverter used.  In short, it's not actually very feasible to control PWM with a half or full bridge, and simple on/off (per cycle) PWM.  The problem is this: when the inverter is turned off from all sides, the output impedance is high, and that changes what the load is doing (namely, it allows it to ring down, discharging some of its energy into the supply rails, and shifting phase angle with respect to where your oscillator is still running at).  You actually have to use a three-level inverter (shorting the output to 0V instead of leaving it open circuit), or a full bridge with phase-shift PWM (which is hard to generate); either way you need four transistors to do it.

So, even as a professional, I tend not to do that, and just stick with full-wave operation, leaving PWM at max (i.e., whatever is needed for dead time, 45% or whatever).  I would rather vary power output by varying driven frequency (which is presumably what you're going to do here), or varying supply voltage (which requires a completely separate buck converter stage, but it does offer big wins for controllability -- the frequency-shift method is actually quite tricky to control).

Tim

[ritaismyconscience]

From my understanding (and from the original schematic comments), I think the tank voltage goes through the op amp (I should use the 311 instead probably), then goes through the 4046, then the 4046 finds the phase difference and feeds it into the arduino, which does a few calculations and sends a new voltage back into the VCO. So isn't this closed loop?

[T3sl4co1l]

It may be closed loop eventually, but as mentioned it will be very slow.  Conditions on the tank can change much more rapidly than that (say if the coil gets shorted out).  Or if the CPU crashes.

Which is another matter: with care, this can all be done in software, but one tiny slip-up and the inverter explodes.  I wouldn't recommend an Arduino beginner to do that, at least unless they want to spend the money and time to replace all those transistors!

Tim

[ritaismyconscience]

I don't think the original author intended for it to respond to shorts, I think the goal of the arduino is to slowly change the frequency as the metal heats up.

[petespaco]

I think the goal of the arduino is to slowly change the frequency as the metal heats up.

From what, to what, and why?

[ritaismyconscience]

Well the resonant frequency changes as the metal heats up, so the arduino changes the frequency to match it. There's probably a version which doesn't use an arduino.

[petespaco]

Okay, please allow me to restate my question(s):

What is the actual  benefit that you see of changing the frequency as the metal heats up?

What is the  default frequency that is applied to the tank when the heating first begins?

Does the frequency increase or decrease as the metal heats up?
And, how much change is there?

As an example:   Let's assume that the frequency starts out at   50 kHz.
Does the frequency increase to just 70 kHz or so, or does it increase to 200 kHz or so?

[T3sl4co1l]

Depends on the load.  The most extreme possible case would probably be around a 2:1 change, for a steel (~half freq) or copper (~double freq) pipe fitting tightly inside a solenoid coil.  The change under operation would be due to, say, melting a slot in said pipe, breaking the circuit.

Which would also be a fairly sharp transition when it breaks.

Frequency is normally starting high, but this circuit has no enable or startup condition so it reduces to a power-on case, which is around zero since the filter capacitors are ground referenced.  Which is, I forget what the 4046 does, is that minimum frequency?

Tim

[ritaismyconscience]

Here is my new circuit, which should still lock on the resonant frequency without the arduino. I mostly copied it from uzzors2k. What do you all think?

Edit: Here's the original

[petespaco]

Sorry to be bothering you guys again----
    I don't get very large changes in tank frequency as the metal heats up with the ZVS heaters that I have.  For example, starting out at about 40kHz, the frequency may increase by 2kHz or even less when a significant chunk of iron bearing metal is inserted into the coil, cold.  Then I see only a frequency increase of a few hundred Hz or so as the metal gets toward the curie point.
  Okay--- 
So it seems that you guys are saying that the circuit under discussion "listens" to see what the frequency of the tank WANTS to be and then changes the driving oscillator to match?  If that is what is desired, why bother, if all you are doing is to match what the present  frequency already is?

Also, I'd like to understand what the use of this circuit will be.  If I understand correctly, (and please correct me if I am wrong), this thing has a 15 volt power supply.   If so, it doesn't sound like it would be very powerful.  So, how much power are you expecting, and  what will this induction heater be used for?

[ritaismyconscience]

There are two terminals for driving the GDT. I will be using this to drive a full bridge of IGBTs on mains voltage. Also I assume that the frequency should change more as the metal melts. See pic below for more information



Edit: I found out that the PNP transistors are backwards. I also made a PCB of this schematic.



Since this is my first time making a PCB there might be a few mistakes I made.

[T3sl4co1l]

Hah, I forget that Uzzors used some of my bitmap schematic symbols from back in the day.

Yeah, that's about right, though again it doesn't have an enable, so it's always flat out and you don't have any way to stop it if something goes pear-shaped.  Like if the inverter draws destructive current or such.

Tim

[ritaismyconscience]

Update: Improved layout so it would fit on my blank pcb:


I'll probably make the pcb tomorrow.

[johnf]

In your output stage --tracks too long try to shorten them up
gates /bases ok just  a via in a normal pcb is 2nH so your tracks are many tens of nH all of this comes back to bite as ringing

[ritaismyconscience]

Looks like there's a mistake with my circuit. Turning the pot changes the frequency instead of the phase shift. I also found out that the op amp was not needed.

This one should hopefully be correct: