What is this amplifier doing, exactly?

Well, actually, not "quite" a Class A design. Class A is intended to bias the transistor
into its linear region of operation, but you'll never get that with the ckt as shown.

It looks like the person who designed it had the intent of Class A, but you would need
a resistor in the emitter lead, so the 100Ks on the base have something to bias in
a stable fashion. As it is, they simply turn on the transistor, and whether or not the
collector sits at Vcc/2 is strictly a factor of the hFE [beta] dc current-gain of the
transistor.

But there is a much more serious problem with this design, namely that there will be
a constant dc-bias on the speaker and a non-trivial amount of dc-current continually
running through it. Not good, the speaker can burn up. Instead, speakers should be
AC-coupled through a large capacitor with a ckt like this - ie, attempted Class A. And
unfortunately, that cap needs to be a large value, eg 220 uF, to get proper low-
frequency response, since the speaker impedance is so small.

Also, with a proper Class B, as Krupski mentioned, the positive and negative drive
ckts will be adjusted so the DC current through the speaker is 0, so that takes care of
the problem of continuous current burning up the speaker.

Advice - throw this ckt away, use something better.

Thanks for the comments everyone!

oric_dan:
But there is a much more serious problem with this design, namely that there will be
a constant dc-bias on the speaker and a non-trivial amount of dc-current continually
running through it.

Why is there that current (which incidentally the scope trace appears to agree with)?

If the transistor is "off" it shouldn't be conducting, should it? Or does the biasing keep it on? I guess it must be that, because on the scope it looks like when the input signal is off, the output signal is around 2.2V.

Yes, that's it. If you remove the 100K pullup, and keep the 100K pulldown - to keep
the transistor turned off, and then capacitively couple the input signal, as others have
mentioned, then the transistor will stay off when no signal is present. Then, you also
don't need to capacitor-couple the speaker.

However, you'll only ever get square-wave outputs [ie, rasty sounding things] from
your speaker.

It is possible to get sweeter sounding audio using the Arduino. Would require going to
a true Class A amplifier ckt [or more complex Class B, etc, as Krupski mentioned], and
then using high-frequency PWM, modulated at a lower audio rate, and then using
low-pass filters ahead of the amplifier to smooth out and anti-alias the PWM.

Interesting. In this particular case I am feeding in square waves anyway, so that's no great loss.

I mean, you can get amplifier chips for a couple of dollars if you want proper amplification, but I was leaning towards getting the square wave tones out of the thing to be loud enough to hear, and not damage the output pin.

So your suggestion of adding the capacitor, and removing the resistor, could well achieve that with minimal effort.

BTW, if I remove the resistor between collector and base, wouldn't the capacitor need to have the positive side (if it had one) to the Arduino output pin, and not the transistor base, as the output pin would be more positive?

Nick,
Wire it up like my Mosfet example. The cap keeps the DC out of the speaker, and the sound is nice.

Here's the heart of the code:

// info on alarm sound
#include "pitches.h"

// notes in the melody:
int thisNote = 0;
int noteDuration = 0;
int pauseBetweenNotes = 0;
int melody[] = {
  NOTE_C6, NOTE_A5, NOTE_C6, NOTE_A5, NOTE_C6, NOTE_A5, NOTE_C6};
// note durations: 4 = quarter note, 8 = eighth note, etc.:
int noteDurations[] = {
  12,12,12,12,12,12,4};
  // create a warble once
  for (thisNote = 0; thisNote < 8; thisNote++) 
  {
    // to calculate the note duration, take one second 
    // divided by the note type.
    //e.g. quarter note = 1000 / 4, eighth note = 1000/8, etc.
    noteDuration = 1000/noteDurations[thisNote];
    noTone(6);       //apparent known bug - need this for the tone to play next.
    tone(6, melody[thisNote],noteDuration);
    // to distinguish the notes, set a minimum time between them.
    // using the note's duration + 10%:
    pauseBetweenNotes = noteDuration * 1.10;
    delay(pauseBetweenNotes);
    // stop the tone playing:
    //  noTone(6); 
  }

pitches.h is tab in my sketch:

/*************************************************
 * Public Constants
 *************************************************/

#define NOTE_B0  31
#define NOTE_C1  33
#define NOTE_CS1 35
#define NOTE_D1  37
#define NOTE_DS1 39
#define NOTE_E1  41
#define NOTE_F1  44
#define NOTE_FS1 46
#define NOTE_G1  49
#define NOTE_GS1 52
#define NOTE_A1  55
#define NOTE_AS1 58
#define NOTE_B1  62
#define NOTE_C2  65
#define NOTE_CS2 69
#define NOTE_D2  73
#define NOTE_DS2 78
#define NOTE_E2  82
#define NOTE_F2  87
#define NOTE_FS2 93
#define NOTE_G2  98
#define NOTE_GS2 104
#define NOTE_A2  110
#define NOTE_AS2 117
#define NOTE_B2  123
#define NOTE_C3  131
#define NOTE_CS3 139
#define NOTE_D3  147
#define NOTE_DS3 156
#define NOTE_E3  165
#define NOTE_F3  175
#define NOTE_FS3 185
#define NOTE_G3  196
#define NOTE_GS3 208
#define NOTE_A3  220
#define NOTE_AS3 233
#define NOTE_B3  247
#define NOTE_C4  262
#define NOTE_CS4 277
#define NOTE_D4  294
#define NOTE_DS4 311
#define NOTE_E4  330
#define NOTE_F4  349
#define NOTE_FS4 370
#define NOTE_G4  392
#define NOTE_GS4 415
#define NOTE_A4  440
#define NOTE_AS4 466
#define NOTE_B4  494
#define NOTE_C5  523
#define NOTE_CS5 554
#define NOTE_D5  587
#define NOTE_DS5 622
#define NOTE_E5  659
#define NOTE_F5  698
#define NOTE_FS5 740
#define NOTE_G5  784
#define NOTE_GS5 831
#define NOTE_A5  880
#define NOTE_AS5 932
#define NOTE_B5  988
#define NOTE_C6  1047
#define NOTE_CS6 1109
#define NOTE_D6  1175
#define NOTE_DS6 1245
#define NOTE_E6  1319
#define NOTE_F6  1397
#define NOTE_FS6 1480
#define NOTE_G6  1568
#define NOTE_GS6 1661
#define NOTE_A6  1760
#define NOTE_AS6 1865
#define NOTE_B6  1976
#define NOTE_C7  2093
#define NOTE_CS7 2217
#define NOTE_D7  2349
#define NOTE_DS7 2489
#define NOTE_E7  2637
#define NOTE_F7  2794
#define NOTE_FS7 2960
#define NOTE_G7  3136
#define NOTE_GS7 3322
#define NOTE_A7  3520
#define NOTE_AS7 3729
#define NOTE_B7  3951
#define NOTE_C8  4186
#define NOTE_CS8 4435
#define NOTE_D8  4699
#define NOTE_DS8 4978

Erdin:
So I would remove both 100k, and perhaps add a flyback diode over the speaker.

First experiment ...

Without flyback diode:


With flyback diode:

Nice pictures! Even I didn't expect a normal loudspeaker to cause that spike.

Next experiment:

Krupski:
I would try this for fun: Remove the 2.2K resistor between the Arduino and the transistor base and replace it with a 1 uF (not critical) capacitor (positive side to the transistor base).

0.33 uF capacitor added in series with the 2.2K resistor:


0.33 uF capacitor added instead of the 2.2K resistor:

The sound is a bit thin with this one.

Erdin:
So I would remove both 100k, and perhaps add a flyback diode over the speaker.

Back to original circuit, (no capacitor), however the 100K resistor from collector to base removed (flyback diode still in place):

Now both 100K resistors removed (looks much the same to me):

The DC bias on the speaker will push (or pull) the diaphragm well away from centre and thus
can give strong even harmonics or worst-case damage the mechanical suspension - not good.

For a novelty speaker driver try a MOSFET driver chip like the MIC4422 which takes logic level
in and can run from 5 to 18V and deliver up to 9A!! (at 18V). Definitely want to use an output
capacitor to protect the chip and speaker, and check it doesn't overheat, but it is also plenty
fast enough to run as a class-D amplifier very nicely - a fast PWM signal for instance will get
you 8-bit audio of sorts.

This is a variation of what CrossRoads suggested:

I kept the transistor, put 100 ohm and a 0.33 uF in series with the signal to the base.

No resistor from collector to base, and 1K from base to emitter.

Sounds a bit thin still. Maybe I'm not doing it right.

Maybe depends on the speaker too. Mine definitely do not sound thin.

Wiring it like this (I think):

I get this output:


Change the 2.2K base resistor to 100 ohms (is that good for the output pin?) and I get:

That last one sounds pretty good (and loud!).

I was starting to doubt that the thing I had in my hand really was a capacitor (it was just sitting around on a breadboard) so I swapped it out for a 1 uF electrolytic. Got much the same results as with the other capacitor, with the last circuit above (with the 100 ohm resistor) and a 1 uF capacitor in series with that:

I just don't "get" the shape of the blue line. I obviously need to learn more about capacitors and the way they work in circuits like this.

throw this ckt away, use something better.

Bah. It attempts to be an audio amplifier, but the arduino isn't feeding it "audio" anyway, so there is little point.
Throw it away and replace it with something simpler that ISN'T anything close to an audio amp. Like the traditional "higher power" transistor switch...

The circuit in reply #24 is close to that isn't it, Westfw?

The cap is doing what caps do, it's trying to prevent a change in voltage by supplying a large amount of current when it first switches. Notice how the upper trace (yellow) pin voltage sags when it switches. I think the 100 ohm resistor performance shows that the transistor is too small since driving it so hard gets it to produce a much better looking wave.

Try a 10 or 22 uF cap, instead of 0.33 uF. Maybe also use a 100 ohm R from emitter
to ground. This should greatly improve the low-frequency response. Those rapid decays
in the waveforms are due to the input time-constant being way too small.

The large overshoot without the snubbing diode is the typical inductive kickback that
occurs when you open the current to the speaker.

I found a 220 uF cap lying around and that gives this:

I presume I put the + side on the Arduino side? That's the more positive side of it, right?


As for the 100 ohm between emitter and ground, that made it much softer:

Did you mean to replace the 1K with 100 ohm between base and ground?

The 220 uF looks much better than 0.33 uF. The quick decay is gone, and you'll
not have any dc-currents through the speaker.

I thought the 100R in the emitter would help improve the low-freq response, but
it also kills the gain too much. So you might go back to tying the emitter to gnd,
and use a 10K in series with the 220 uF cap on the base. I might also use a larger R
on the base to gnd, eg go back to the 100K.

I'm not sure why you're getting the overshoot on the leading edge. Do you have the
scope probe ground lead tied close to the same point as the speaker gnd, or right
at the emitter?