Need feedback on DIY AB power amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
If you want to make it simpler at the expense of somewhat higher distortion (why on earth does one need sub 0.0005% THD?), especially so that to prevent the turn-on thump, you'd need a turn-on mute circuit with an output relay that will spoil the magnificent distortion figures at a very unpredictable way. (Unless you use an ultra low distortion solid state relay of course).

So here are my suggestions, many of them would somewhat increase the distortion, but some of them aim at better stability. I think the THD would still be below 0.001% so who cares? It's still at least 10 times lower THD than what the most sophisticated human ear can distinguish...

3 stage output stage is not needed, there is always more risk of oscillation with a CFP. Two emitter followers are the most stable config.
R26 and R27 probably not needed.
The enhanced current mirror is not needed.
R11 and C5 is fine, but the positive rail PSRR of this circuit is so bad that the beneficial effect of R11 and C5 on distortion will be hampered by PS ripple noise coming in from the positive rail (unless you use a regulated power supply). So if you plan to use a conventional 50-Hz transformer, rectifier bridge and filter capacitors I would heavily RC filter the positive rail to the small current stages. It will make the turn-on transient worse, but much better positive rail PSRR can be achieved. (And you need an output relay anyway to suppress the turn-on thump.) Then R11 and C5 would make more sense, although they are not strictly necessary.
The serial C7 and C10 18pF x 68 pF makes about 14 pF which is comparable to nonlinear transistor capacitances and perhaps to parasitic capacitances between PCB traces. It can work but I think it would be safer to double C7 and C10 and at the same time halve R20 and R21 to keep the same unity loop gain frequency. Halving R20 and R21 also decreases noise a bit.

What I would definitely change:
Beefier driver and output transistors, because loudspeakers are not a resistive load, for a 4 ohm speaker higher peak output currents can be expected than for a simple 4 ohm resistor.
KSC3503 and KSA1381 are only 100mA devices, 100mA is their absolute maximum rating at 25 Celsius. It's too low for drivers in a 20 Watt amplifier. Use e.g. MJE340 and MJE350 instead.
MJE15030 and MJE15031 are 8A devices but their hFE rolls off above 1A. Good coupling to heatsink would be needed or use bigger transistors in TO-247 package.
In the real world amp circuit 100uF || 100nF bypass capacitors will be needed on the supply rails, especially if the cables from the power supply are long.

Other things to consider:
C3 can be made higher (e.g. double) for better RF imunity.
Base resistors contribute to the effective emitter resistance by Rb/hFE. If Rb is 10, and hFE is 100, then this virtual increase in emitter resistance will be about 0.1 ohms. To limit this effect, I would consider using 4.7 ohms instead of 10 ohms.
But be prepared also that too small emitter resistance in the output transistors together with high output stage bias current increase the risk of thermal runaway.
 
Last edited:
If you want to make it simpler at the expense of somewhat higher distortion (why on earth does one need sub 0.0005% THD?), especially so that to prevent the turn-on thump, you'd need a turn-on mute circuit with an output relay that will spoil the magnificent distortion figures at a very unpredictable way. (Unless you use an ultra low distortion solid state relay of course).

So here are my suggestions, many of them would somewhat increase the distortion, but some of them aim at better stability. I think the THD would still be below 0.001% so who cares? It's still at least 10 times lower THD than what the most sophisticated human ear can distinguish...

3 stage output stage is not needed, there is always more risk of oscillation with a CFP. Two emitter followers are the most stable config.
R26 and R27 probably not needed.
The enhanced current mirror is not needed.
R11 and C5 is fine, but the positive rail PSRR of this circuit is so bad that the beneficial effect of R11 and C5 on distortion will be hampered by PS ripple noise coming in from the positive rail (unless you use a regulated power supply). So if you plan to use a conventional 50-Hz transformer, rectifier bridge and filter capacitors I would heavily RC filter the positive rail to the small current stages. It will make the turn-on transient worse, but much better positive rail PSRR can be achieved. (And you need an output relay anyway to suppress the turn-on thump.) Then R11 and C5 would make more sense, although they are not strictly necessary.
The serial C7 and C10 18pF x 68 pF makes about 14 pF which is comparable to nonlinear transistor capacitances and perhaps to parasitic capacitances between PCB traces. It can work but I think it would be safer to double C7 and C10 and at the same time halve R20 and R21 to keep the same unity loop gain frequency. Halving R20 and R21 also decreases noise a bit.

What I would definitely change:
Beefier driver and output transistors, because loudspeakers are not a resistive load, for a 4 ohm speaker higher peak output currents can be expected than for a simple 4 ohm resistor.
KSC3503 and KSA1381 are only 100mA devices, 100mA is their absolute maximum rating at 25 Celsius. It's too low for drivers in a 20 Watt amplifier. Use e.g. MJE340 and MJE350 instead.
MJE15030 and MJE15031 are 8A devices but their hFE rolls off above 1A. Good coupling to heatsink would be needed or use bigger transistors in TO-247 package.
In the real world amp circuit 100uF || 100nF bypass capacitors will be needed on the supply rails, especially if the cables from the power supply are long.

Other things to consider:
C3 can be made higher (e.g. double) for better RF imunity.
Base resistors contribute to the effective emitter resistance by Rb/hFE. If Rb is 10, and hFE is 100, then this virtual increase in emitter resistance will be about 0.1 ohms. To limit this effect, I would consider using 4.7 ohms instead of 10 ohms.
But be prepared also that too small emitter resistance in the output transistors together with high output stage bias current increase the risk of thermal runaway.

Thanks for the pointers man, I will work through them.
Why you would go against the enchanced current mirror in the input pair?
The output devices will be beefier (as stated in first post in the thread) - i just didn't have their model. The drivers will be changed to MJE340/350.
For now I had a lot to do at my work and was working on a software synth side project.
I have a through hole board designed but im not satisfied with the design, will probably do it again with surface mount everything but the outputs, Vbe multiplier and drivers - im going to use dual transistors for the LTP and current mirror, might resign from the enchancing transistor as you said.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.