"zvs" driver with active feedback, how to start the oscillation? / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

"zvs" driver with active feedback, how to start the oscillation?

Move Thread

LAN_403

Dr. Dark Current

Sat Nov 28 2009, 12:58PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Hi, I'm experimenting with active feedback for the "mazilli" or "zvs" driver, and I'm having problems with reliably starting the oscillation.
If I use a feedback winding on the transformer (or on the primary if its a SSTC) and connect this to the input of inverting and non-inverting gate drivers, one of the transistors will stay ON and the oscillation won't start upon turning the supply on and something goes pop.
The oscillations start if I use a start-up oscillator, but it interferes with the feedback and there is a lot of noise and "randomity" of the switching transitions if that makes any sense.
I can't use a PLL because the feedback needs to lock instantly.
Any ideas?

Antonio

Sat Nov 28 2009, 01:03PM

Registered Member #834 Joined: Tue Jun 12 2007, 10:57PM
Location: Brazil
Posts: 644

With the inductance of the series inductor around the same value of the inductance of one of the primary coils it should start by itself, but it drains more current than normal for a few cycles. Using slightly different resistors from the power supply to the zeners helps in the startup.

Steve Conner

Sat Nov 28 2009, 02:15PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

If it won't start reliably, it's because your feedback system doesn't have enough gain for small signals. To start, it needs to be able to amplify its own internal noise with a loop gain of more than 1.

One way of fixing this is to increase the gain. If there's some comparator- or Schmitt-trigger-like thing in there, it has a gain of zero for signals too small to trip it, by definition. Bias it closer to its threshold, so that smaller signals will trip it.

An alternative is to give the circuit a hefty startup kick (for example, with a diac and capacitor, as used in CFLs, switched mode halogen transformers, etc) so that it never has to worry about small signals.

If it has loop gain less than 1 for large signals too, then it won't keep running once started, but this is a different problem.

Dr. Dark Current

Sat Nov 28 2009, 02:32PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Hi Steve, well the problem is that whenever the oscillation stops for whatever reason, the start up circuit must "kick" it back on fast enough so that the transistor which remained conducting doesn't get fried.
Like you say, this might probably be solved by some gate driver circuits without the hysteresis on the input, and setting the DC feedback voltage so that it falls into the linear region of the circuits, but I can't think of any such circuit which would be fast enough...

Steve Conner

Sat Nov 28 2009, 02:41PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

In that case, why not design it so that both transistors turn off when the oscillation stops?

For instance, use two gate drive chips of the same polarity, and hang a centre-tapped feedback winding between them. Bias the centre tap with a DC voltage so that both drivers are only just off in the absence of a signal.

The downside of this is that it introduces deadtime, whereas the ZVS actually needs both FETs on at once during the zero crossing to work properly! Being a current-source inverter, it needs anti-deadtime.

The best way to protect a current-source inverter is to have a current-limited power supply and just let both FETs wedge on if the oscillation stops. If you turn them both off, the stored energy in the DC link choke will kick back and explode them. This is the dual of shoot-through in an ordinary voltage-source inverter.

Dr. Dark Current

Sat Nov 28 2009, 02:50PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Steve McConner wrote ...

The downside of this is that it introduces deadtime, whereas the ZVS actually needs both FETs on at once during the zero crossing to work properly! Being a current-source inverter, it needs anti-deadtime.

This could be fixed by putting small caps (in relative to the resonant one) across the D-S of both FETs. But MOSFETs should non-destructively avalanche anyway...

Steve Conner

Sat Nov 28 2009, 03:04PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Maybe you can get away with it in practice, but my point is, it's theoretically wrong to turn both FETs off at once, for the same reasons (up to duality) that it's wrong to turn both ON at once in a voltage-source inverter.

You wouldn't cure shoot-through in a VSI by putting a small inductor in series with each MOSFET, but that's the dual of what you're proposing.

Proud Mary

Sat Nov 28 2009, 04:35PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

At the risk of seeming dull, might I suggest that you Google "conditions of oscillation" whereupon you will find most of the plausible oscillatory conditions that underlie your own dreams, and those of others.

It has always been a maxim of 4HV that we shall not needlessly repeat here what is common knowledge elsewhere, or a quick fix solution that might be better addressed by Google or the Wiki -

love,

Stella X

Dr. Dark Current

Sun Nov 29 2009, 01:36PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

It works! I've put a lowpass filter on the output of the start-up oscillator so the output is now close to a triangle wave (not the square as previously), so no sharp edges, and the feedback from the transformer easily overrides it (higher dv/dt).

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob