Mac said:
No apology necessary, since you were right. Thanks for looking it over. I
don't really know of any biasing arrangements for power MOSFET's, so I
just did what seemed sensible to me. I did want to avoid having a bias
adjust pot, but maybe I'll let that requirement drop.
I've seen two topologies for push-pull class AB (or B) MOSFET amps. In one,
like yours, the sources of the P- and N-channel FETs are joined and create
the output. In the other, it's the drains: so, the sources are connected to
the power rails, and it's the P-channel FET that's connected to the positive
rail.
This latter one obviously requires some trickier biasing, and a phase
inverter. And it's easy to make it go "bang". The reason for it is, I
believe, that the gate is always closer to ground than the supply rails.
With the sources-joined topology (like yours), you can't get the output any
closer to the supply rail than Vth (and then some, because output that high
implies a lot of Id, which means that the MOSFET is turned hard on, which
takes extra Vgs - see typical Id/Vgs curves). There might also be some
additional protection against inductive loads, but I haven't thought it
through.
So you either need to be willing to sacrifice a few hard-bought volts of
output (in the MOSFETs, as heat), or you need a high-v gate supply, or you
need the tricky topology with drains connected to output.
These are not really factors in your design; you're talking low power, so
"sacrifice a few volts" is a fine answer. With 15V supply rails, you'll get
11V out, and that's enough to put 10W rms (sine) into an 8 ohm speaker. In
your case, the biasing is not really any different than an ordinary old
BJT-based amp, except that the "Vbe" is a lot higher: instead of voltage
difference of about 1.5v between bases, you need around 4v or 8v or whatever
it is between gates.
By the way, I see people still using BJT's as the bias generator in MOSFET
amps. I don't know why that is; you'd sort of think that you'd want to use
another same-type MOSFET, so as to get comparable temp curves. (That is,
make a Vgs multiplier rather than a Vbe multiplier.) But differences in Id
mean you never get a perfect match; so maybe BJT's are just as good.
