# Need help understanding an Class D amplifier design.

Thread Starter

#### Chethan_Raj

Joined Aug 1, 2022
22
So I've been recently trying to build a simple class D amplifier, I've seen through the block diagrams and design considerations online, but recently came across an Amplifier design which doesn't seems to fit the majority of the circuits online. So in the input stage we use a comparator to modulate the input audio signal to PWM. But the circuit below doesn't seem to include any triangle wave generator or comparator circuitry. The circuit is from an Infineon tutorial, there isn't any explanation either. Here is the link to the PDF file which contains the circuitry. It would be helpful if someone explained how this is supposed to work in general, the idea behind these components. Thanks

#### Ian0

Joined Aug 7, 2020
9,835
It looks like it might be a self-oscillating design, rather than one based on a sawtooth and comparator.

#### Attachments

• 452.8 KB Views: 7

#### ronsimpson

Joined Oct 7, 2019
3,042
For those that don't want to take the time to read Iran0's paper, here is the first diagram, slightly modified.
The power stage is "digital" in that it has two states, +supply or -supply. The LC filter on the output, filters out the switching frequency.
The comparator looks at the input and the output (with filtering) and chooses if the power stage needs to pull up or down.

If the output is to be at 0V then the power stage will be at +supply 50% and at -supply 50%. The LC averages that to 0V.
There is no 'oscillator', but the whole thing is an oscillator.
That is a very simple explanation.

#### Ian0

Joined Aug 7, 2020
9,835
The feedback comes from the output side of the LC filter, so the LC filter provides a phase shift of almost-but-not-quite-180° at the filter frequency- not enough to make it oscillate. Now the last thing you want is for it to oscillate at the filter frequency, because the Q of the filter might be quite high depending on the load impedance at that frequency, so there is a phase-lead and a phase-lag circuit to push if over the edge into oscillation at a more suitable frequency.
The input voltage then shifts the average (this is an average over quite a short time, 50us at most) voltage of the oscillator, so that when the oscillator frequency is removed by the filter the output waveform is the input waveform amplified.

What I don't understand is why the Infineon application notes explains how the sawtooth-oscillator and comparator circuit operates and then goes on to present a self-oscillating design. Confusing, or what?

#### ronsimpson

Joined Oct 7, 2019
3,042
The feedback comes from the output side of the LC filter,
Not in your schematic.

#### Ian0

Joined Aug 7, 2020
9,835
Not in your schematic.
Not in the TS's circuit, but it does in the paper I posted. It would seem that R3/C1 does the extra phase shift in the TS's circuit. Needless to say, they are both self-oscillating circuits, but the manner of providing the 180° phase shift for the oscillator is different.
I would expect an oscillator that is timed by the output inductor would not be as frequency-stable as an RC phase shift circuit, as the inductance will vary with the load current.

#### ronsimpson

Joined Oct 7, 2019
3,042
but it does in the paper I posted.
This paper? from your pdf.

OK never mind....
When I build these I use an oscillator/ramp generator very much like above and take the feedback from the speaker voltage, because that is what I want to regulate. (I build power supplies so that is how I think)
If I wanted to make a self-oscillating amp I would use something like in post #3. Note there are RCs missing.

#### Ian0

Joined Aug 7, 2020
9,835
This paper? from your pdf.
View attachment 312026
OK never mind....
When I build these I use an oscillator/ramp generator very much like above and take the feedback from the speaker voltage, because that is what I want to regulate. (I build power supplies so that is how I think)
If I wanted to make a self-oscillating amp I would use something like in post #3. Note there are RCs missing.
That's from the TS's first post.
This is what I posted.