Class D: Triangle or Sawtooth?

[Osvaldo de Banfield]

Hi all class D fans. Is there any audibly difference between PWM circuits using triangle or sawtooth PWM generators? Which is best between them? I made my self design and use triangle because it easiest to generate. But what about sounding results?

Best regards. Thanks in advance.

[gmarsh]

Triangle will produce less distortion; sawtooth creates an amplitude dependent time shift. There's really no reason why you'd ever want to use sawtooth.

[Osvaldo de Banfield]

Yes, I agree. But I see somewhere in the web some pdf´s that uses the sawtooth as PWM input to the comparator.

[Pafi]

I'd say sawtooth is simpler, since a single IC, TL494 contains oscillator, comparator, error amplifier (and differential outputs).

[J.R.Freeman]

Originally Posted by Osvaldo de Banfield

Yes, I agree. But I see somewhere in the web some pdf´s that uses the sawtooth as PWM input to the comparator.

Hmm... that's odd, I wonder why?

For an accurate PWM signal you'd want as symmetrical and pure a triangle wave as you could get. Feeding a nice clean square wave into an opamp integrator will produce a good triangle wave if you chose your components carefully (low temp drift caps and resistors, good layout and the rest).

Check this out, this is kind of neat - it's how to make a triangle wave with 2 opamps and some passives. It also talks about integrators:

Generating Triangle Waves

Jim

[krish2487]

I suppose the comparator of the Triangle/Sawtooth and the input signal is where the trouble is likely to be.

A triangle wave will give you a symmetrical center aligned PWM, for any given instant in the signal.Controlling the deadtimes becomes easier. Similarly checking for cross- conduction and even splitting up the waveform for a full/half bridge topology.

But with ramp the instant when the wave reaches its peak and goes to 0 is critical.
If the ramp is too slow you have cross conduction in a mosfets if it is too fast the comparator wont switch states to signify the end of the pulse.

It can be done, but takes too much effort to design it safe.

[nigelwright7557]

In a sawtooth there will be some delay as the waveform peaks then goes back to zero, during this time you will get distortion on the output.

A triangle waveform is better.

[Osvaldo de Banfield]

In a circuit of my own, I use an HCF4047 as a free running 500KHz oscillator, and take the output from pin 10 & 11 (complementary), and feed them an integrator using a CMOS opamp like TS274 used as double integrator, this means, integrating not only inverting input, but also non inverting, to ground. It gives a nice triangle wave.

The trouble in my two early projects is the same. I used the principles exposed in fig 5 of:

http://users.ece.gatech.edu/mleach/e...01/ClassD2.pdf

But sometimes, it doesn´t start because of "CF" gets some charge greater than (positive or negative related to ground) the triangle wave and the comparator becomes blocked. Had anybody tried that system?

[gearheadgene]

those self-oscillating circuits suffer from thermal drift - maybe it's not important to your design.

I chose to drive mine with an oscillator - stable always.

Watch out for DC drift on the opamp integrators. Also, depending on the frequency you run it at, the opamps require high slew-rate. When the triangle changes from ramp up to ramp down, a slow opamp will do some really bizarre stuff. So, either use a better part, or setup your comparator that stays away from the junky transitional spot.

[nigelwright7557]

I also found some op-amps flip the output if you go to near the +ve rail.
I used a tl082 and that was fine.
As said earlier you need a fast op-amp.