0
\$\begingroup\$

I am simulating a unregulated push pull converter in LTspice (voltage isolator). There is the effect of growing winding currents during the simulation. Is this a flux walk (seen in the first image) ?

enter image description here

I have the duty cycle shown here (~90%)

enter image description here

And this circuit. (The low input/output capacitances are intentional). Is there a significant error somewhere i am missing, what can i do to fix this ? Explanation of the source of this would be helpful

enter image description here

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Run the simulation for more than 2ms to see if things stabilize. I think you are seeing the transient response here (initial conditions ). \$\endgroup\$ – Marla Nov 11 '16 at 13:54
4
\$\begingroup\$

You've got one MOSFET driven at about 60% and the other MOSFET driven about 40% duty cycle - that puts dc into your transformer and walks the flux.

If anyone reads this focusses on the 2nd picture you'll hopefully see what I mean. I've had to invert the colours and make yellow red to make it viewable more legibly: -

enter image description here

If you are driving a centre-tapped transformer like this you have to keep the average on-time for both transistors the same. Now this sounds like a "ban" on PWM control but it isn't; one way you can PWM the transformer is just reduce the on time for both transistor whilst keeping as short a gap between the two pulses as possible. So you charge current in one direction proportional to on-time, then discharge current (via the other transistor) for an equal length of time, then hold-off for some period until ready to start the re-charge process. This keeps the magnetization current from walking the core into saturation.

\$\endgroup\$
10
  • \$\begingroup\$ I've been curious about this and I like the answer. But I have a question, too. Because of part variations, drive variations, etc., it seems that there may still be slight, residual differences in the exact timing even if driving things as you say. Is it the case that the gap between operation of the transistors is how this slight difference is "made up for" in the end? That there must be enough of a gap to cover over this slight issue and allow an exact zero for the accumulation of volt-seconds? I keep thinking one must observe that the current reaches zero. \$\endgroup\$ – jonk Nov 11 '16 at 19:53
  • \$\begingroup\$ @jonk after the 2nd transistor turns off any residual current (aka energy) will dissipate in the snubber.. Even if going full pelt the small off time between successive transistor switching will remove a lot of that energy but it can be a problem without care to the detail. \$\endgroup\$ – Andy aka Nov 11 '16 at 20:24
  • \$\begingroup\$ Thanks. I felt that care would be required and you've confirmed it. This is a very tricky area (to me, when I've thought about it) and I've refrained from going this direction in part because I could handle my problems other ways and in part because I was worried about just how much learning curve ahead I might have in getting this done well and correctly. \$\endgroup\$ – jonk Nov 11 '16 at 20:31
  • \$\begingroup\$ To make things worse there's not much on the internet about driving transformers like this with pwm. What I have found doesn't appear to go into much detail about timing. I've chickened out in the past and put a really efficient buck regulator feeding the centretap with feedback control of it with the push pull transistors going full pelt with accurate 50-50 timing. \$\endgroup\$ – Andy aka Nov 11 '16 at 20:37
  • \$\begingroup\$ Well, you've confirmed by "fears" and as a hobbyist I'll probably stay clear of this. It's not as though I these are goals of mine, anyway. Just means to ends. I'll use techniques I know how to do and continue holding short of this one. I like the centertap idea. Thanks for the thoughts. You've settled this issue just a little further for me. \$\endgroup\$ – jonk Nov 11 '16 at 20:41
-1
\$\begingroup\$

My solution was to check the "Start external voltages at zero" checkbox. I just have a simple push-pull circuit simulated, and the gate waveforms were perfectly symmetrical... to the nanosecond. Damn thing pulled 6kA at the very first pulse rolls eyes. Indeed, it's flux walking, leading to core saturation. In real life, there are circuits to keep the net DC current through each winding the same. Yes, it's a little tricky. Yes, it's necessary...

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.