I have gone around in circles trying to find the problem.
Tried lots of extra decoupling.
Tried a capacitor/resistor on VS.
Tried diodes from VS to the rails to clip any overshoot.
Tried diodes from output to rails to clip any overshoot.
Tried different mosfets with small changes to the resetting.
Its so annoying that I am almost there, the resetting happens at the onset of clipping. Had it not worked at all then there is obviously a serious problem. I am getting around 500 watts out before resetting occurrs.
I have ordered some more pcbs with a powerplane. I dont hold out much hope but you never know.
VS with new inductor.
An externally hosted image should be here but it was not working when we last tested it.
freq drops a lot for you?
and for next picture, zoom in the waveform, you only need to show 1/2 of period, just to or bottom part, there is where information is
Coz this could look like only mosfet industance/ output capacitance ringing and so on, but then again I can't see if it is only that
Is that under load and what load was it, when it was taken (picture)?
and for next picture, zoom in the waveform, you only need to show 1/2 of period, just to or bottom part, there is where information is
Coz this could look like only mosfet industance/ output capacitance ringing and so on, but then again I can't see if it is only that
Is that under load and what load was it, when it was taken (picture)?
Last edited:
Hm, even if 10MHz appears to be a low frequency for specific half bridge ringing... May be you are suffering from half bridge ringing.
Did you also try a snubber at each MosFet?
A series connection of 470pF+10 Ohms placed directly at each MosFet?
I usually place such a snubber directly from the drain copper plate towards
the source leg. 470pF+10 Ohms is just a starting point, which has to be optimized for each application. Geometry of this snubber can have a massive influence, because in reality it is not an RC-snubber, but an RCL snubber. Best results can be achieved when the resonance frequency of this snubber is matched to the undesired ringing frequency, because then the snubber acts like a purely resistive damper.
Note: You may build the 10 Ohms by 5 pieces of 47 Ohms or 56 Ohms in parallel, because of heat. The resistor has to handle P=0.5*C*Ub²*f, whereas Ub is full difference from pos rail to neg rail.
P.S.
.. got this hint from Eva 3-4 years ago and wonder why she didn't propose it here, but probably because 10MHz seems pretty low for this topic ....
All very Interesting... (Minor Thread Hijack)
I have been following this thread with interest. I'm a few stages back with my 2092 project. Have done an initial through hole PCB design. It is based on the example schematic in the Application note. Have followed all the layout rules including ground planes and star point grounds. Kept the FET switching lines as short as possible and managed to fit the whole lot into a 100mm x 70mm PCB have spent ages trying to squeeze everything together.
Now, having read this thread I have picked up some more hints and design features. Seems to me that power rail decoupling is absolutely critical for all audio circuits eg. LM3886, Op amp, class d. I propose on my version to use 10uf electrolytic + 1uF of film to decouple.
I assume i'll have all the same ringing problems etc that the OP has. Looking at these and the need for snubbers, I think i'll add these to my PCB going from Drain Plate to power rails. Only question is how many watts are going to be dissapated? I know it's dependant on frequency but what would be a sensible value to go for while in lay out stage?
Sorry for the hi jack but this discussion seems relevant to my efforts as I expect to have the same problems.
I have been following this thread with interest. I'm a few stages back with my 2092 project. Have done an initial through hole PCB design. It is based on the example schematic in the Application note. Have followed all the layout rules including ground planes and star point grounds. Kept the FET switching lines as short as possible and managed to fit the whole lot into a 100mm x 70mm PCB have spent ages trying to squeeze everything together.
Now, having read this thread I have picked up some more hints and design features. Seems to me that power rail decoupling is absolutely critical for all audio circuits eg. LM3886, Op amp, class d. I propose on my version to use 10uf electrolytic + 1uF of film to decouple.
I assume i'll have all the same ringing problems etc that the OP has. Looking at these and the need for snubbers, I think i'll add these to my PCB going from Drain Plate to power rails. Only question is how many watts are going to be dissapated? I know it's dependant on frequency but what would be a sensible value to go for while in lay out stage?
Sorry for the hi jack but this discussion seems relevant to my efforts as I expect to have the same problems.
There seems to be mixed experiences with this chip.
Some people say it either just worked or it worked with a bit of extra work.
I have found at least one other person who had trouble with resetting and didnt found a solution. They too had resetting on clipping.
I have been round and around with it trying everything I can think of and it still resets on clipping and seriously resets into hard clipping.
I have even disabled over current protect and it still resets.
Even tried altering the class d base frequency and that made no difference.
I even put a battery on VB to make sure VB wasnt dropping out on clipping and it wasnt.
I would be very interested to hear how you get with especially on clipping and hard clipping.
I had some success tonight stopping resetting.
If I add 470uF across the power supply capacitors, use the new inductor and disable over current detect then the resetting goes away.
Previously disabling over current had no effect. I can only guess the 470uf's have quietened down the supply rails, although they didnt look bad before.
If I add 470uF across the power supply capacitors, use the new inductor and disable over current detect then the resetting goes away.
Previously disabling over current had no effect. I can only guess the 470uf's have quietened down the supply rails, although they didnt look bad before.
The large smoothing capacitors have a series resistance at higher frequencies so adding 470uf cancels out any high frequencies on the power rails. My test setup power supply is on 6 inch flying leads with crococdile clips on them so this could be complicating things, but the pcb does have smoothing on it. I found I was getting HF on the pcb but it was due to the long power leads. If I earthed my scope probe on the pcb there was no HF.
I need to do a bit more work to see if I can get away with smaller capacitors than 470uf.
I also need to work on the over current sensing so it still works with shorts on the output but doesnt reset before it should.
It was originally at the datasheet example settings.
I thought that was well inside what the mosfets were taking but I was mistaken. It could there is noise getting in somewhere.
Finally got the amplifier working fine after sorting out the decoupling, inductor and over current protect levels.
Many thanks to those who have contributed to helping me getting this amp going.
Its been a lot of hard work but also a great learning experience despite having been in electronics for 35 years.
Many thanks to those who have contributed to helping me getting this amp going.
Its been a lot of hard work but also a great learning experience despite having been in electronics for 35 years.
Not perfect but acceptable as the rining stops before 450ns when the over current trip samples the waveform.
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.