Guru
Joined: 09/08/2007
Location: AustraliaPosts: 4406
| Posted: 12:46pm 28 Mar 2017 |
Just trying to shine a bit of light into a rather dark place, but another thing to think seriously about is driving capacitive loads.
The large magic toroid is a wonderful thing, it reflects back anything connected to the secondary straight back into the primary. And that includes any capacitance across the secondary winding, either deliberately added, or what natural capacitance is normally there between the actual secondary turns and the steel toroid.
The natural self resonant frequency of the secondary is very likely to be well below 20 Khz, the bridge will then be driving a highly capacitive load, even if there is nothing else connected to the secondary. Its also fairly common to connect a large capacitor directly across the secondary to filter out the 20 Khz pwm.
That produces a lovely clean sine wave, but it is really going to stress out your mosfets every time they switch. To make the problem even worse, fitting a bunch of large electrolytics across the incoming dc supply to the bridge only increases the current spike during each switching cycle.
There absolutely must be sufficient inductance somewhere to limit the rate of current rise during switching. The ferrite core in series with the primary is an excellent way to limit the initial rate of current rise under no load, but it will very quickly saturate.
There is really no substitute for having a decent oscilloscope and monitoring all the voltage and current waveforms. The bridge itself may be working fine with reasonable speed and sufficient dead time, but what that bridge has to drive can have a significant effect on the switching stress.
Cheers, Tony.