Amp idea for the adventurous/insane...

[earthtonesaudio]

Just an idea...



...Caution!  It won't work as shown!  Needs lots of other bits, such as input coupling capacitor, output DC isolation from speaker, regulation for switching supply, etc.

But the idea would be to use a LM386, LM380, or similar audio power op-amp for the input IC, and some power MOSFETs for the output and PSU drivers, for a simple, efficient, and powerful class-D guitar amp.

Oscillator frequency is the same for the PWM section and the power supply, to avoid beat frequency problems.

[edit]  ...I just realized that the PWM output is modulating the power supply output voltage with the audio signal.  That's probably bad.  Perhaps I'll re-think this.  The main thing I wanted to show was a class-D amp that uses the same oscillator frequency for a high-voltage power supply for the output transistors.  That part is okay, but otherwise not so much.   

[gez]

There's a good overview of class-D in one of Penfold's books.  Don't take this the wrong way, but you need to read up on the subject (based on what you've posted).  Penfold uses a simple class-D design based around a 555.  A good place to start.  IIR, I bridged the output (two square waves out of phase) and managed to drive headphones.  Not exactly hi-fi, but it worked.  Led me to experiment quite a bit with class-D and I ended up with a love-hate relationship towards them. 

A lot of the efficiency you gain from class D-designs is lost at the final stage as there's an inevitable voltage-drop across the filter network (used to block the carrier signal, to avoid excessive current flow at the output) just before the speaker(s).  As a result, you end up running them at higher supply voltages to get the calculated efficiency, which usually creates higher current consuption in previous stages.  Incredibly quiet amps, though. 

[gez]

PS  There are purpose-built Class-D chips that you can buy.  They use a lot less components than discrete designs and are pretty efficient.

[earthtonesaudio]

I hear you loud and clear with the "read more" recommendation.  I actually understand more than a little (I think) about class-D, the trade-offs involved with switching time and board layout, dead time in the output drivers, etc.  I left out a bunch of critical stuff both for clarity and to discourage someone from attempting to build such a thing without proper knowledge first.

Mainly I wanted to show a relatively simple approach that could be used (with the necessary parts added) when you consider the requirements for a guitar amp: moderate to low-fidelity, and moderate power output.  Most designs I've seen are either for hi-fi or super high-power like car subwoofers or PA systems.

Designing for guitar lets you strip the design to it's bare essentials and offers some other advantages: the switching frequency can be lower (don't need 20kHz bandwidth) so the board layout can be a little less critical.  I was thinking the power output could be less than 50W, keeping thermal issues simple.


The main idea I was wanting to show was the use of a single oscillator for both the PWM output signal and the switching power supply oscillator.

[gez]

The main idea I was wanting to show was the use of a single oscillator for both the PWM output signal and the switching power supply oscillator.

I don't think that can be done.  The signal will effectively modulate the supply to the output MOSFETs.  I should think that will cause distortion.

[earthtonesaudio]

I was unclear, sorry.  I meant this:



And by this I'm implying that the triangle waveform at the op-amp (-) pin would most likely be sent into a high input impedance comparator along with a DC reference voltage to create a usable duty cycle to switch the power supply transistor.

[gez]

I don't think it will make any difference.  No matter what, the duty cycle isn't fixed: the supply to the MOSFETs will be modulated by the signal.  Amplitude Modulation, as in AM radio.  As an incoming sine wave rises (during its +ve cycle), not only do you get a corresponding rise at the output due to higher mark:space ration (assuming things are in-phase), but I'm pretty sure you also also get a shift upwards in bias point (no expert on switched supplies, but I think that's what happens).  The output signal will, therefore, be 'stretched'.  Similar thing happens for -ve cycles. 

How linear that stretching is depends on your oscillator.  The one you've used (won't work as shown, by the way) doesn't knock out a proper triangle, so distortion is going to be worse than it would be if the supply were fixed.  Some primitive designs do use 'relaxation' oscillator triangles, and the distortion isn't that noticeable, aurally, but the circuit won't exactly be hi-fi.  For decent audio you need a high switching speed (similar principle to sampling rate in the digital world) and a very linear triangle wave. 

Electronic Projects for Experimenters by R A Penfold will give you a nice little intro to the world Class-D (thank you GeWhiz for recommending that one).  Not much else in the book of interesting to us (unless you have a insatiable urge to detect breezes), but it costs little, especially if you buy it secondhand using Amazon/eBay. 

[sean k]

Detect breezes... thats what wind midi controllers need to have... to detect and adjust volume! Good Ol' R.A.

[gez]

Here's the schematic!

[Zben3129]

or the 100W version

[brett]

+1 for class D chips if you want this sort of thing.  There's a 50W chip around (forget the part no., but it's in lots of 50W kits).
But for for musicality and efficiency together, give me class B anyday.  As simple as a pair of TIPs and a transformer (e.g. those splendid little Pignose amps).
cheers

[earthtonesaudio]

Thanks for pointing out the error of my assumption about the oscillator, Gez.  I just assumed the frequency stayed the same, but alas, no.  I guess the way to go would be a master oscillator for the audio PWM and switching power supply, but separate from both. 

Brett, I've looked at several class-D chips and while they do seem like a really great solution for DIY, I have this obsession with doing things the hard way...

[gez]

Thanks for pointing out the error of my assumption about the oscillator, Gez.  I just assumed the frequency stayed the same, but alas, no. 

The frequency does remain the same.  However, if the switching duty-cycle varies then the voltage available to the MOSFETs varies.  So, if you use a PW modulated switching signal, you end up amplitude modulating the supply. 

As you seem to want to go discrete, I'd recommend that Penfold book as a good place to start (plenty of practical advice and the circuits work).  The first design he comes up with is easy to implement as it's based around a 555 timer.  Not exactly hi-fi, but it could possibly be made more linear with the addition of an op-amp for the oscillator.  Plus bridging the output will allow you to drive a heavier load (you can just about get a headphone amp out of this with a 12V supply).

[earthtonesaudio]

The square wave generator with the control voltage modulating the duty cycle has been done before with a LM386, basically as I've drawn it.  The inputs are ground-referenced so it looks like some biasing rules have been violated, but it has been reported to work.  I might try the same basic thing with a LM380 for a little more power.  Bridging the output seems problematic, but not impossible. 

The frequency does remain the same.  However, if the switching duty-cycle varies then the voltage available to the MOSFETs varies.  So, if you use a PW modulated switching signal, you end up amplitude modulating the supply. 

As you seem to want to go discrete, I'd recommend that Penfold book as a good place to start (plenty of practical advice and the circuits work).  The first design he comes up with is easy to implement as it's based around a 555 timer.  Not exactly hi-fi, but it could possibly be made more linear with the addition of an op-amp for the oscillator.  Plus bridging the output will allow you to drive a heavier load (you can just about get a headphone amp out of this with a 12V supply).

I'll have to try to find that book.  I like Penfold's stuff.  The county library here has one of his books (which I've memorized cover to cover  ), so I'll see if they have the one you suggested.

I am planning on bridging the output anyway, mainly to remove the necessity of an output coupling cap, but also to help drive my 16 ohm speaker cab. 

If I'm going to go all-out and build a "proper" class-D amp, then I might try Crown Amps' implementation, which varies duty cycle to each of the output FETs in opposition (upper switch increases duty cycle while lower switch decreases duty cycle, rather than just inverting one).  They call this "interleaved" or "class I" but it's really just another way of doing class D.  Seems cool, and a nice way of avoiding shoot-through currents in the output section. 

[earthtonesaudio]

In my first post I edited it to say that the signal modulating the supply to the output devices was probably a bad thing.  But recently I was reading about efficient amplifier classes (E/F/G/H/Dogherty) and stumbled across the phrase "envelope tracking amplifier" and it all clicked into place.  Modulating the output devices' supply using the signal envelope is way more efficient!

The first schematic I posted might actually work as shown, as a hybrid class D/envelope tracking/class H amplifier.  However, I can imagine the frequency response of the 386 being a problem, and there would probably need to be some phase delay on the signal so the power supply could keep up.  At any rate, it's fun to think that, once again, I've "invented" something that's actually been around for decades. 

It will be interesting to explore this very simple topology to see what kind of power outputs can be obtained at low supply voltages.