The beginning of an audio amplifier

[Joel Kolstad]

Hi guys,

I was wondering if you might take a look at this schematic here:

http://oregonstate.edu/~kolstadj/amp.gif

It's the beginning of a class AB audio amplifier (ostensibly for around 50
watts), and I had a few questions...

-- Q2/Q3 and Q1/Q4 effectively make up 'compound' transistors, but this is a
technique that I hadn't seen until recently. Are there any rules of thumb
to choose the resistor sizes (R8 and R9)? I'm finding that making them
smaller can significnatly reduce the quiessent current through the power
transistors -- this occurs before a smaller resistors raises Vbe of Q3 or Q4
to more closely match the biasing set up by D4 and D2. Hmm... perhaps I
should choose R8/R9 to have the same current through them as D2/D4 do?

-- I did start out just using two transistors in a Darlington configuration
rather than the compound transistor approach shown. The drawback seemed to
be the additional ~0.7 volts lost in 'head room.' Does this configuration
have any additional benefits?

-- How would you go about deciding the quiessent current through the
R2/D4/D2/Q5/R3 leg? In classes I've taken on circuit design, usually it's
dictated by frequency response, which doesn't seem to be much of an issue in
an audio amplifier. I did discover some of the tradeoffs between that leg's
gain and the output swing -- as gain is reduced (R3 becomes closer to R2),
the negative-going output will saturate more quickly since there's a larger
voltage drop across R3 and therefore the base of Q4 approaches Vee sooner.
On the other hand, higher gains with smaller R3's make biasing Q5 more
sensitive.

-- Any of suggestions would be appreciated. Thanks!

---Joel Kolstad

 

[Winfield Hill]

Joel Kolstad wrote...

-- How would you go about deciding the quiessent current through the
R2/D4/D2/Q5/R3 leg? In classes I've taken on circuit design...

You should worry about the Q5 current arising from the pullup resistor
R2, which varies from a maxiumu for an output near Vee to nearly zero
for an output near Vcc. The creates serious distortion and slew-rate
problems that must be corrested by changing the design. At minimum R2
must be replaced by a current source or by a bootstrap resistive tap.

Thanks,
- Win

 

[Genome]

Really... hmm... doesn't this circuit have the same problem? -->
http://oregonstate.edu/~kolstadj/amp2.gif ... it's from the book, "Analog
Electronics, an integrated PSpice approach."

I guess I'll have to go grab my copy of The Art of Electronics at home this
evening.

Thanks for the help,
---Joel Kolstad

Sure, last time I was presented with a similar question on my HNC I ended up
kicking my fucking copy of AoE around the fucking kitchen in disgust. I
kicked the fucker in half. Then I re-wrote the fucking question and gave a
proper fucking answer to the question. I'm still a stupid fucker, but I'm
almost half less stupid than the fuckup who asked the question in the first
place.

And, Win, stop answering questions while you're pissed. I'm totally fucking
pissed and still manage to retain a level of keyboard control that is
unbefuckinglievable.

DNA

 

[Genome]

Oh.... Pissed means drunk. But you probably knew that.

DNA

 

[nospam]

I was wondering if you might take a look at this schematic here:

http://oregonstate.edu/~kolstadj/amp.gif

-- How would you go about deciding the quiessent current through the
R2/D4/D2/Q5/R3 leg? In classes I've taken on circuit design, usually it's
dictated by frequency response, which doesn't seem to be much of an issue in
an audio amplifier. I did discover some of the tradeoffs between that leg's
gain and the output swing -- as gain is reduced (R3 becomes closer to R2),
the negative-going output will saturate more quickly since there's a larger
voltage drop across R3 and therefore the base of Q4 approaches Vee sooner.
On the other hand, higher gains with smaller R3's make biasing Q5 more
sensitive.

You might try adding some gain in the output stage. A resistor from Q2
collector to Q3 emitter and one from Q3 emitter to ground (and mirror to
the -ve half).

A gain of 2 (equal value resistors) will halve the voltage swing required
from the input stage. Iooking like an additional load on the output it does
waste some power.

 

[Joel Kolstad]

Alright, I went home for dinner and picked up my copy of The Art of
Electronics...

You should worry about the Q5 current arising from the pullup resistor
R2, which varies from a maxiumu for an output near Vee to nearly zero
for an output near Vcc.

I thought I had worried about it by including Q5's degeneration resistor,
R3? The distortoin is due to gm (of Q5) being a function of its collector
current, as discussed on page 83, correct? Page 98 suggests that a current
source is a better approach to fixing this problem than bootstrapping.

It appears that what I'm calling the 'composite transistor' of, e.g., Q2 and
Q3 is known as a Szikiai connection, discussed on page 95. Resistor R8,
then, in my schematic is just the 'recommended' transistor to improve
turn-off time and the value seems to be not too critical.

Finally, I'm not following the example in the last paragraph on page 93: The
temperature difference is 30 deg., so Vbe changes by 63mV. Great. But --
assuming R3 and R4 have 50mA through them (preceeding paragraph) and
therefore 50mV across them, won't we now have another 63mV->63mA across R3
and R4, giving 113mA total, a 126% increase? (The book says 20%) -- What am
I missing here?

Thanks,
---Joel

 

[Walter Harley]

Hi guys,

I was wondering if you might take a look at this schematic here:

http://oregonstate.edu/~kolstadj/amp.gif

It's the beginning of a class AB audio amplifier (ostensibly for around 50
watts), and I had a few questions...

In addition to your wise choice of AoE, let me also recommend Douglas Self's
"Audio Power Amplifier Design Handbook". It addresses the practical design
issues of typical class-AB audio power amplifiers in excellent detail.

 

[Ted Wilson]

Hi guys,

I was wondering if you might take a look at this schematic here:

http://oregonstate.edu/~kolstadj/amp.gif

It's the beginning of a class AB audio amplifier (ostensibly for around 50
watts), and I had a few questions...

-- Q2/Q3 and Q1/Q4 effectively make up 'compound' transistors, but this is a
technique that I hadn't seen until recently. Are there any rules of thumb
to choose the resistor sizes (R8 and R9)? I'm finding that making them
smaller can significnatly reduce the quiessent current through the power
transistors -- this occurs before a smaller resistors raises Vbe of Q3 or Q4
to more closely match the biasing set up by D4 and D2. Hmm... perhaps I
should choose R8/R9 to have the same current through them as D2/D4 do?

-- I did start out just using two transistors in a Darlington configuration
rather than the compound transistor approach shown. The drawback seemed to
be the additional ~0.7 volts lost in 'head room.' Does this configuration
have any additional benefits?

-- How would you go about deciding the quiessent current through the
R2/D4/D2/Q5/R3 leg? In classes I've taken on circuit design, usually it's
dictated by frequency response, which doesn't seem to be much of an issue in
an audio amplifier. I did discover some of the tradeoffs between that leg's
gain and the output swing -- as gain is reduced (R3 becomes closer to R2),
the negative-going output will saturate more quickly since there's a larger
voltage drop across R3 and therefore the base of Q4 approaches Vee sooner.
On the other hand, higher gains with smaller R3's make biasing Q5 more
sensitive.

-- Any of suggestions would be appreciated. Thanks!

---Joel Kolstad

Hi

One of the major benefits of this type of Darlington arrangement is
that it eliminates two Vb-e drops from the quiescent current
definition and that the Vb-es removed are those of the transistors
doing all the hard work and therefore subject to the greater
temperature variations.

Basically, you are attempting to define a voltage across R4 and R5 in
series, which in turn defines the quiescent current in Q1/Q2. This is
done by generating a voltage across D4 and D2 and applying this
between the bases of Q3 and Q4. What is left across R4 and R5 is this
voltage minus the Vb-e drops of Q3 and Q4.

If you recall what you were taught about the accuracy of a small
difference between two large numbers, you will see that significantly
better than 2:1 improvement in quiescent current definition is
achieved by reducing the number of Vb-es in the loop from four to two.

Personally, I would substitute D4 and D2 with a Vb-e multiplier, using
either a BD139 or a BD140, and thermally link this to Q3 and Q4.

As Win points out, you need to define a constant through D4/D2, (or
the Vb-e multiplier), so that, apart from dynamic considerations
already mentioned, the voltage across them doesn't vary with output
voltage, otherwise quiescent current will vary wildly with output
voltage.

Regards

Ted Wilson

 

[Ian Buckner]

In addition to your wise choice of AoE, let me also recommend Douglas Self's
"Audio Power Amplifier Design Handbook". It addresses the practical design
issues of typical class-AB audio power amplifiers in excellent detail.

You can also look at some of Doglas Self's work on his home site:

http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm

(dipa = Distortion In Power Amplifiers.)

Regards
Ian

 

[Genome]

Genome wrote...

Hey, DNA, the horrible circuit referenced above isn't from our
book, thanks be to the gods (and ignoring figure 2.63), instead
we suggest a proper bootstrap circuit, see figure 2.66, and we
explain why it's necessary. opefully Joel will have read and
learned.

Thanks,
- Win

Whoaaa Neddy,

The reason why I kicked your book around the kitchen was because the
original question involved dogend like the circuit the OP presented.

Some lazy arsed electronics lecturer asked us to design a piece of
non-functional dirt single transistor amplifier with a gain of ten working
off a 15V rail. Then he wanted us to add another gain of ten stage.... with
an input signal of 1V pk-pk..... words such as.... 'on the basis of the
transistors Hfe calculate the required base resistor to bias the transistor
into its linear region and achieve a Gm of bloop bloop bloop.'

I was lucky (?). I'd bought and read the AoE.

First off I was annoyed at being asked to design a non-functional piece of
dirt. Second I was seriously pissed (annoyed) because everyone else couldn't
do it because the guys lectures were crap. Third I knew how to do a proper
job because I'd read the AoE.

What's the point of that? What a waste of time. I don't know why you and Mr
Horrowitz bothered writing it. Arrggghhhhhh..... Kick kick kick.

I gave it to one of the other people on the course and bought another one.
I've been through another four copies since then. I'm a dickhead, I buy them
and then give them away. Every time I end up working with a new keen
engineer/technician they get a copy.

Hope that clears up any misunderstanding...

DNA

 

[Winfield Hill]

Genome wrote...

Hope that clears up any misunderstanding...

DNA

Whew! Thanks, I feel beytter already.

Thanks,
- Win

 

[George R. Gonzalez]

Hi guys,

I was wondering if you might take a look at this schematic here:

http://oregonstate.edu/~kolstadj/amp.gif

Nice start, but

Have you looked at the DC stability of this circuit?

With no output capacitor, it's important to keep the output near zero volts
DC.

But you're doing this by dead-reconing. The bias point for the first
transistor is determined by R3 R6 and R7. If they're the regular 5%
resistors, your output could easily be 5% off, which is a bit much DC. Also
the bias point is a bit power-supply sensitive. Most amplifiers run off
less than perfectly regulated supplies, so this might be a concern. And oh,
a 2N3904 isnt specd at 60 volts is it?

You might cheat and look at a 1964 GE transistor manual. There they gave
examples of power amps made from germanium power transistors. Any bias
circuit that keeps those transistors in a reasonable bias has GOT to work
awfully darn well for silicons....