My first Class D from scratch...

[thefish]

As some of you may know I have been working on a building a Class D amp for a while, well I finally finished it but it is having some very strange problems.

When the Mosfets are not powered and the input is grounded everything looks great. As soon as the Mosfets are powered I get a ton of ringing in the system. The output also looks terrible, the surprising thing is that it does not sound to terrible (about like a cheap boom box). Everything is being run from 13.8V power supply. Notice the ringing in the high side gate is almost 100V.



Triangle Generator

High Side Gate

Low Side Gate

Output (only half the bridge signal).

Schematic

What is the best way to measure the output with the oscilloscope? I was just going to attach the probe to one side and the ground clip to the other but the manual says to never attach the ground clip to anything except true ground.

[ChocoHolic]

-Step1: Put SMD ceramic caps at your 50V rails.
Place 1 piece of these caps (everything between 0.1uF...1uF is OK) directly at each half bridge.
Ceramic material should be X7R or X5R.

-Step2: Put a RCL (series) snubber at each MosFet from drain to source. L is formed by the unavoidable inductance of your snubber geometry. A good starting point for C would be 220pF and for the R 10 Ohms.

Good luck !
Markus

[ChocoHolic]

Furtheron:
The very strong magnitude of ringing is indicating that
your circuit is dealing with quite some current peaks.
And this seems to happen already without load.
High Current Peaks without load do only happen if your dead time is to short, or let's better say: negative dead time ==> shoot through current peak.

From your circuit it also looks like almost no dead time at the input of the level shifter. ...resulting in some overlap at the gate drive voltages.
To increase your dead time you might add some 100pF...470pF from J1 to GND and J2 to GND. Place these caps as close as possible to the IR 2011. X7R SMD types should be fine.

[thefish]

I added 470pf ceramic cap from J1 and J2 to ground, it helped clean up the pre IR2011 switching form, as well as the triangle generator, but it did not do anything for the gates.

Triangle Generator

Pre IR2011 Switch

Just Clipping

I also placed a 1uF cap at each half bridge and it made no visible difference. I have not yet tried the RC snubber.

[thefish]

Here is a Scope image of the J1 and J2 complementary signals along with thier addition, it looks like there is a spike (shoot through) just before each dead time.

J1 and J2 with Addition

[ChocoHolic]

The 1uF at the rails must be very close to the transistors and it really has to be a lossy ceramic like
X7R or X5R.
A MKT or MKP will not introduce enough damping.

Your 470pf should bring you several tens of ns more deadtime. But with 2us/grid you cannot see this.
It seems like you have quite a long shoot through.
With this adjustment you will probably have a lot of
heat in your MosFets, even if you manage to quite down the ringing, which is triggered by your shoot through peak.
470pF is large enough. To increase the dead time further you might increase the 100 Ohms. Probably you need several hundrets of Ohms.

Your J1 & J2 signals are clearly showing quite some overlap.
It does not look like you would have any real dead time
The idea of the dead time is to have a short time during which both MOSFETS are off. This means both Ugs must be below the treshhold voltage.

Zoom into your gate drives (i.e. 100ns/grid) and you will probably see a certain time were both gates are higher than the MosFet treshhold.

[thefish]

I swapped out the 100ohm resistors with 330ohm resistors, there is still the ringing though the magnitude is a bit less. More imporantly the mosfets are now running cool to the touch even after letting it run for 1 hour. Its still not perfect but it is getting there. I will upload a scope image a bit later, but I think it could use a bit more dead time still. Thanks for the help!

[ChocoHolic]

This is sounding reasonable.
You should get adjusted both: Ringing and shoot through.
The ringing is triggered by the high di/dt of the current peak. So there will be probably a point of dead time adjustment where the ringing becomes suddenly much less.

But don't trust this calm no load scenario.
As soon as you go to some higher output current you will get the ringing back, because then you will have to handle the hard switching condition of the body diode.

To handle this you probably have to settle a proper snubber.
Also it might be helpful to use a 10...47 Ohm gate resistor for turn ON, with a parallel diode for turn OFF.
The gate resistor will help to reduce the turn ON di/dt and by this help to reduce the ringing.

[thefish]

Originally posted by ChocoHolic
But don't trust this calm no load scenario.
As soon as you go to some higher output current you will get the ringing back, because then you will have to handle the hard switching condition of the body diode.

To handle this you probably have to settle a proper snubber.
Also it might be helpful to use a 10...47 Ohm gate resistor for turn ON, with a parallel diode for turn OFF.
The gate resistor will help to reduce the turn ON di/dt and by this help to reduce the ringing.

I let it play music for 2 hours with a 8 ohm speaker and still everything stayed cool. I am still holding the rails at 13.8V so when I up it to the designed 50V things will probably change... oops I forgot the power was on and shorted out one of the mosfets, it blew a couple of parts...

[speakerguy79]

The above advice is good. Snubbers are a must, I would use both RC and a Schottky if possible. Your layout is very large, have you tried using an AM or FM radio in its vicinity?