let's talk Half Bridge Flux Walking

[sphinx]

Noob to SMPS - be gentle

I'm conflicted about dealing with flux walk in a half bridge SMPS. Easily dealt with using a capacitor in series with the primary winding however most of the schematics I find in the controller manufacturer's data sheets and app notes do not include this balancing capacitor. Also easily dealt with using current mode control but I'm not interested in that.

I've read that the flux balancing cap really isn't required. I've also read that the flux balance can never be perfect so a method of balancing it is absolutely required.

Just for fun, I simulated a HB circuit in spice and I definitely need the flux balancing cap. So if that's true, why have I run across many designs (non-current mode control) that do not use it?

[luka]

you could exclude it, but does it really hurt to use one and but 100% sure it can't do wrong?

[sphinx]

Originally Posted by luka

you could exclude it, but does it really hurt to use one and but 100% sure it can't do wrong?

I don't believe it hurts and my prototype will likely use one but I would just like to understand how a half bridge supply can allegedly operate without it. My spice simulation confirms that it can only function properly with the series capacitor

[luka]

well it can, but it can get out of control pretty fast once something goes wrong + elements are not perfect, nor are windings and for that matter anything else
So it could work, but will work, but then again it can that easy stop too

[richie00boy]

How have you arrived at the value for the cap?

[Eva]

If the duty cycle is fixed or controlled by a single voltage loop, the capacitor is always required. This capacitor may be actually two capacitors, one to each rail, but this approach is more prone to parasitic resonances.

Nowadays most control ICs use peak current mode, these are probably the datasheets that you saw. This control scheme does not require any capacitor. In fact, the way to get rid of the capacitor is just to control current rather than duty cycle. Furthermore, in my experience, both peak current mode and average current mode actually don't work properly with the capacitor in the circuit.

When the power supply is idling with no load, things may appear to work properly sometimes, but if a load step is applied and the waveforms captured, the erratic behaviour involving brief transformer saturation is easily seen (if the SMPS has not yet failed after repeating this a few times).

[sphinx]

Thanks for the confirmation Eva. What really through me off was a chat I had with an IR field app who didn't know why I used a series cap. He didn't think I needed it and thought it would result in quasi-resonant operation. I looked back through some of the HB schematics I've collected and I now see that they use two capacitors connected to the rails instead of the single series cap. Indeed the only circuits without any of these caps are the current mode controlled power supplies.

richie00boy, my simulated circuit functions properly with a series capacitor => 0.33uF although 1uF seems to be a popular value.

[Eva]

Of course the capacitor results in series resonance, and it has to be sized so that the resonance either happens at the desired frequency or happens far from the operating frequency range (big value). I find little use for that kind of resonance in regulated SMPS.

[dtproff]

There is currently quite a trend among the major power supply manufacturers to use LLC resonance (both voltage domain and frequency domain). They offer powers from 90W adapters up through bi and tr-phase multi-kilowatt supplies for servers. Resonant Half-bridge (and full) designs with sync rectifier secondaries are now starting to dominate the newer designs due to the inherent efficiencies the topology yields.

Tony

[sphinx]

Originally Posted by dtproff

There is currently quite a trend among the major power supply manufacturers to use LLC resonance (both voltage domain and frequency domain). They offer powers from 90W adapters up through bi and tr-phase multi-kilowatt supplies for servers. Resonant Half-bridge (and full) designs with sync rectifier secondaries are now starting to dominate the newer designs due to the inherent efficiencies the topology yields.

Tony

OT but interesting! Is it really worth adding sync rectification to SMPSs with output voltages > 20V? Seems like a disproportionate increase in cost and complexity for a marginal gain in efficiency unless it's a 3.3V supply?