Hallo Steve,

may be I should correct myself: Audio power amplifiers do NOT always contain freewheel diodes at output. And may be, that I only put those schematics into my 'archive', which DO contain them...
In the mean time I have also made a little recherche and found, that those designs where big currents in combination with certain SOAR protection have to be handled, do contain them, or should contain them.
The reason why Douglas Self, a person which I do highly adore by the way, does not use freewheel diodes, is, that his circuits do not contain SOAR protection circuitry! (This would not make any sense, because he wants the least amount of harmonic distortion. This totally forbidds the use SOAR protection circuitry, because of well known introduce of high unlinearity, already at rather low output power levels.) But if his circuits do not contain SOAR protection, then there's nothing what could cause a high change rate of current through loudspeaker, and by this there's no need for freewheel diodes!

SOAR protection provides the shut-down of output transistors whenever the product of voltage drop across collector emitter path times collector current goes above a certain limit. But what looks nice in theory makes trouble in reality: If load current and output voltage show a certain phase shift, SOAR protection circuitry begins to 'protect' the power transistors, although there is no actual need for this. And this can happen at rather low output power levels and is the cause of heavy unlinearity.
On the other hand, SOAR protection IS needed for professional public address amplifiers! What can be avoided in living room applications, namely sudden and unwanted short circuits directly at output, happens nearly every evening 'on stage', when cables are plugged in, etc.. So, for home applications existence of SOAR protection is not really crucial, but for PA applications it IS.

There's a nice application note of 'National Semiconductors' about 'LM12'

http://www.national.com/an/AN/AN-446.pdf

which demonstrates, what happens, if SOAR protection begins to act, while a loudspeaker is connected to output. They state:





And in datasheet of' LM12' they recommend the following circuit:




There are many many other datasheets of audio amplifier chips, which state, that there are output protection diodes present (already built-in) or must be added. E.g. for TDA2006, TDA2030A, LM3886, etc.

http://www.st.com/stonline/books/pdf/docs/1452.pdf
http://www.unisonic.com.tw/spec/UTC%20TDA2030A.pdf
http://www.ben.cz/download/121098/lm3886.pdf

Many power amplifier modules do also have protection diodes at output, although they are not explicitely shown in datasheet. E.g. STK 4038XI through STK 4048XI are such ones, where the diodes can be detected by the help of a digital multimeter.

Here is a link to a professional audio power amplifier, which also shows protection diodes at output:

http://www.bryston.ca/pdfs/38bschem.pdf

Concluding, for ME, all these examples demonstrate, that whenever the current, which flows through a loudspeaker undergoes a fast change rate, a freewheel diode is needed to handle the overvoltage spike caused by the relevant inductivity. So, in the schematic I recommended to Aman, I would add this 1N4148.
May be, freewheel diode can be omitted, when there's no 'really fast' change rate. But how to decide, what's 'really fast'? Wouldn't it be better and simplier, just always to add this freewheel diode?

Regards,
Kai