Paralleled Class A with Class D

[darkfenriz]

Good work!

I think LT1160 is not a very good choice, the manufacturer does not advice to use it over 100kHz as far as I remember. Go for gate drivers from IR, like IRS2011 for example.

Also, you may like to get to know the SMALA concept, which works basically the same as yours except for it senses the current draw from supplies, so it is generally targeted to classAB.

P.s. 100uH and 10nF for LC filter? Its impedance is far from 8ohms! Why such values?

[stocktrader200]

Are you using this over the full 20 - 20khz range ?
the sine on the scope looks good.
I cranked mine up to 1.35 Mhz and used a 15uH coil and .22uf cap as the output filter to achieve full range class A performance out of a class D amp.

[QED047]

Originally posted by darkfenriz
Good work!

I think LT1160 is not a very good choice, the manufacturer does not advice to use it over 100kHz as far as I remember. Go for gate drivers from IR, like IRS2011 for example.

Also, you may like to get to know the SMALA concept, which works basically the same as yours except for it senses the current draw from supplies, so it is generally targeted to classAB.

P.s. 100uH and 10nF for LC filter? Its impedance is far from 8ohms! Why such values?

Thanks for the steer away from the LT chip. After a bit of digging around I decided on using the IRS20955 to drive a pair of IRF540Z's for my next attempt. Attached photo of PCB below. I'm trying to split the whole thing into modules like this as the design lends itself to modularization - and it will allow me to audition different sections along the way. BTW, the schematics did have one or two bum values after I (ahem) tidied them up before posting. Calculated values will appear in due course.

Originally posted by stocktrader200
Are you using this over the full 20 - 20khz range ?
the sine on the scope looks good.
I cranked mine up to 1.35 Mhz and used a 15uH coil and .22uf cap as the output filter to achieve full range class A performance out of a class D amp.

Excellent! Have you posted the project on these forums? I couldn't find any other examples of actual paralleled amps. As for range, yes I want all I can get!

[QED047]

I was having some serious trouble with the IRS20955 - I killed four of them! But I finally got on top of the problem (documented in this thread: Why am I blowing IRS20955's?) and now things are moving on nicely. I'm finding that the IRS20955 is very useful because it allows the PWM input to be referenced to mid-rail where the comparator is. I think has definitely been worth persevering!

I now have the "parallel amp" connected to a minimalistic transistor/resistor source follower: An IRFP240 with a 100R source load and thanks to the magic of Class D the O/P is driving about 75W into 4Ohms. I have adjusted the switch frequency to be 600kHz at idle and the Loudspeaker signal has approx. 100mV ripple on it. The sound is unmistakably "Pure Class A" and I'm about excited as a lottery winner!

Perhaps someone should try bringing me down to the ground by pointing out that a Class A preamp for a standard UcD ought to yield the same results... but I can drop the enable pin on the IRS20955 and still hear the music (albeit at lower input levels to prevent problems with -ve swing). Also, when looking at the O/P of the Class D when it's not "in parallel" it's clear how much the Class A damps down the carrier - even without any feedback.

And it's the elimination of feedback that I've been striving for all along: the reactance of the Loudspeaker seems to be of little concern to this particular circuit.

[QED047]

The experimental design is now boxed-up and running without any more tweaking. I've been comparing it with an old TRIPATH amp that I've got working again (see TRIPATH EB-TA0104 revisited ) and it's very much holding its own in terms of sound quality.

What I find seems to be the acid test for amps like these is playing around with an old analogue signal generator at high levels (only when the rest of the house is deserted!). At between 10KHz and my upper limit (13K and falling) fine-tuning the frequency reveals all the aliasing that happens as the carrier shifts with amplitude. Even at low amplitudes this has quite a nasty effect, but it's much more pronounced with the TRIPATH.

In the photos you can see an old ATX PSU case special dedicated to the trafo supply (I find these make great enclosures for quick-build projects like this) and the extruded case for the amp. It's a bit out of focus but the 100R 25 watt resistor that loads the follower is visible at the far end. Where the signal comes in there's a switch that controls the enable pin on the MOSFET driver chip and throwing this allows you to listen in pure "first watt" mode or with "Class D assistance". Unfortunately the only appreciable difference is a slight gain when Class D is used - presumably from the change in transconductance of the "unloaded" follower. Slightly louder always comes across as "better" but being critical it's not masking anything bad. I can't see why this approach isn't more popular!

[phase_accurate]

I once did simulations with paralleled amps. The circuit was some sort of first-order delta-sigma. It was mostly free of supply pumping and there was no switching at all at low signal levels. And only the switch for the side where power was drawn was operated when needed. Such a thing could of course also be built in a self-oscillating fashion and there is even an (Australian ?) patent for something like this.

It all looked promising but a pure class-d is still a lot more interesting to me.

Regards

Charles

[QED047]

Originally Posted by phase_accurate

It all looked promising but a pure class-d is still a lot more interesting to me.

That's cool - but I'm curious to know what you mean by "pure" class-d ~ there are many significantly different approaches to getting the switching signal. Do you mean comparing the signal with a sawtooth to get PWM?

[phase_accurate]

By pure class-d I mean one with not linear output devices - neither in parallel nor serial (like tracking PSU).

Regards

Charles

[d to the g]

Do you have a schematic of the final version?