symmetric amplifier

[Rambi]

Dear all,

Over the last few months I've built several prototypes of power amplifers. I started with OpAmp driven circuits (which sounded OK), but decided to try a SS design as well, just to be 'sure'. The first (working ) protoype sounded so much better I decided to stick with that (this is personal preference of course, besides I can only compare it to an 8 yr old Phony(TM) audio set).
From various sources combined (mostly Elektor) I ended up with this design for a symmetric (class AB) power amplifier.
The only difference between this one and my last working prototype is the Vbe multiplier (prototype has trim-pot between C and E) and the end transistors (BD911/912 instead of Sanken types since BD91x is somewhat cheaper to experiment with, although I did use the Sankens in a version without the cascodes in the LTPs successfully).

I would welcome any thoughts or comments you might have on this design.

Thanks!

Remco

P.S. One thought already: R2 & R102 are probably superfluous, the reason I kept them in was because I like to stick as much maybe's in a PCB design as possible, it's always easier to leave components out and bridge them than to insert new components in an exisiting PCB

[Rambi]

Ooops...
Comment about the Vbe multiplier should be:
prototype has trim-pot between C and B (not C and E).
Sorry about the zipped PDF, but it's the only way I could get a decent picture


Remco

[vbd]

My 2 cents (of Euro, of course...):Q3,4,and 6 (and 103..etc..) dont have to be BC546-556, as they are cascoded and see only small voltage. It is better for gain and offset to use high gain transistors BC550-560, for instance...

[dimitri]

It is bad idea to shunt Zener (LED) by a capacitor. The capacitor will not filter unwanted noise because of the low dynamic impedance of the Zener (LED). For reducing hum you can split R7(R107) as 5.1k + 5.1k, and connect C4 (C104) (better 100u) between supply rail and the common point of 5.1k resistors. If you aware about hf LED noise, you can use R-C filter, i.e. put an extra resistor between D1 and base of Q5 and reconnect C3 between supply rail and base of Q5.

The voltage swing on the Q8(Q108) base will be limited by (30-UD2-0.7)=25V. The excess voltage drop on the output device will be 10V at maximum current, would you like to warm up your room? Your 30V supply should be 40V.

Why do you use emitter bypassing in Q6(Q106)?

[dimitri]

oops, R56 should be ten times lower

[halojoy]

Originally posted by Rambi
Dear all,

Over the last few months I've built several prototypes of power amplifers. I started with OpAmp driven circuits (which sounded OK), but decided to try a SS design as well, just to be 'sure'. The first (working ) protoype sounded so much better I decided to stick with that (this is personal preference of course, besides I can only compare it to an 8 yr old Phony(TM) audio set).
From various sources combined (mostly Elektor) I ended up with this design for a symmetric (class AB) power amplifier.
The only difference between this one and my last working prototype is the Vbe multiplier (prototype has trim-pot between C and E) and the end transistors (BD911/912 instead of Sanken types since BD91x is somewhat cheaper to experiment with, although I did use the Sankens in a version without the cascodes in the LTPs successfully).

I would welcome any thoughts or comments you might have on this design.

Thanks!

Remco

P.S. One thought already: R2 & R102 are probably superfluous, the reason I kept them in was because I like to stick as much maybe's in a PCB design as possible, it's always easier to leave components out and bridge them than to insert new components in an exisiting PCB

I had a look at your amp.
I like it.
BD911 are good working horse devices.
Can take a lot of heat for being TO220
A choice I also like.
Regarding the details you will get more opinions
by others.
Enjoy your project.
It is yours!

[traderbam]

You have done your homework!
My only suggestion for now is that you try this and see what it sounds like: Remove R56 and C51. Add 100 ohm resitors between Q9 base and emitter and Q109 base and emitter.
I believe you should have decoupling across your CCS LEDs - it's nothing to do with LED noise as you probably already know.
BAM

[Rambi]

Originally posted by dimitri
It is bad idea to shunt Zener (LED) by a capacitor. The capacitor will not filter unwanted noise because of the low dynamic impedance of the Zener (LED). For reducing hum you can split R7(R107) as 5.1k + 5.1k, and connect C4 (C104) (better 100u) between supply rail and the common point of 5.1k resistors. If you aware about hf LED noise, you can use R-C filter, i.e. put an extra resistor between D1 and base of Q5 and reconnect C3 between supply rail and base of Q5.

The voltage swing on the Q8(Q108) base will be limited by (30-UD2-0.7)=25V. The excess voltage drop on the output device will be 10V at maximum current, would you like to warm up your room? Your 30V supply should be 40V.

Why do you use emitter bypassing in Q6(Q106)?

C6/C106 bypasses R8/R108 for audio (but not for DC), seems to give it more 'oomph'. Since my latest PCB has Q7/Q107 mounted directly to the heatsink I could probably just as well lower R8/108 and ditch the capacitors.

Using a cascode will not change the voltage drop over, or power disipation by, the output devices, that's determined by supply and output voltage (and of course the kind of load). It does limit the maximum output swing by a few volt, but I figured that disadvantage was compensated because it allows me to use BC5x devices in the VAS.

As for the supply: I use a 2*30V toroid, after rectifying etc the output voltage is more like 40V (slightly higher even).

R56 can be a factor 10 lower, I'll try it and see (hear) what it does to the sound.

Thanks for your comments!

Remco

[Rambi]

Originally posted by traderbam
You have done your homework!
My only suggestion for now is that you try this and see what it sounds like: Remove R56 and C51. Add 100 ohm resitors between Q9 base and emitter and Q109 base and emitter.
I believe you should have decoupling across your CCS LEDs - it's nothing to do with LED noise as you probably already know.
BAM

I'll try a lower value for R56 (appr. 100 Ohm instead of 1200) first, since that's easier to do. Using two resistors and connecting the 'middle' to the output (like you describe) is one of the 'options' I did not include in the PCB design (but should have...). Anyone out there ever tried and compared these two variations? (fig.13 A and B from http://www.dself.demon.co.uk/dipa.htm#5 )

Thanks!

Remco

[dimitri]

Anyone out there ever tried and compared these two variations?
-----------------------
Yes, I’d tried and compared these configurations many many years ago… I can address you to my article "Building Better Buffers," Electronics World + Wireless World, 1992, Nov., pp.931-934.

Yes, Mr.Self is right, that his Fig.13(b) gives much better turn-off, but another drawback of Fig.13(a) is that the driver transistors will also switch off, after the output ones, one more singularity! In Fig.13(b) the driver transistors will work with nearly constant collector current (class A) during the whole cycle.