Heat sink sizing for power amps

[prairiemystic]

I'm trying to ballpark heatsink requirements for Class B power amplifier (output transistors). I started with the calculations here: Elliot Sound Amplifier Efficiency

Theoretically, a rail-to-rail amp with +/20V rails, puts 25W RMS into an 8ohm load. This is Iout=1.768A RMS with Voutput transistor=one rail leading to... the output stage dissipating 10.36W

There is some argument over using RMS power (0.707), average power (0.636) or music power (?) in these calcs. An amp used in a dance club at full bore would dissipate less than max. sinewave on a bench. Extending the same calculation to a Leach low-TIM amp:
With a 80VCT transformer, rails at +/-58VDC and assuming 5 volts/rail loss in the driver/output stages/emitter resistors... =about +/-53VDC output swing into 8 ohms is 175W, with the output transistors dissipating around 96W.

Leach used Wakefield 423K thermal resistance of 0.94 °C/W or a 47°C rise above ambient at 50W in convection. That's about 110°C/230°F on the bench at full power sinewave.
But to what to use for calcs in music?

[mnemneth]

Originally Posted by prairiemystic

Theoretically, a rail-to-rail amp with +/20V rails, puts 25W RMS into an 8ohm load. This is Iout=1.768A RMS with Voutput transistor=one rail leading to... the output stage dissipating 10.36W

There is some argument over using RMS power (0.707), average power (0.636) or music power (?) in these calcs. An amp used in a dance club at full bore would dissipate less than max. sinewave on a bench. Extending the same calculation to a Leach low-TIM amp:

No argument at all.

RMS power is a myth and "Music power" is a gimmick put forward by amplifier makers to make their product seem better than it is and should be regarded with the utter contempt is deserves.

Average power rules. Simples.

What you would need to find is the maximum power your amp could produce in the most arduous of circumstances and I would say that would be a square wave at half your rail to rail voltage into your minimum load value.

This would pre-suppose output into a resistive load, of course.

Sandy

[flg]

A square wave, if I remember right, would be 1.33% greater rms voltage than a sine wave of equal peak voltage.
Did I get that right?

[prairiemystic]

I looked at National Semi's chip-amp app note AN-1192.pdf and they calculate Pdiss max=(Vcc tot^2)/(2pi^2)*RL. I have no idea where the pi squared came from. Using their eqn with my previous Leach amp calc, I get 71W instead of the 96W for Pdiss.

Also, they mention that using a real "audio" application (avg. music power) doesn't work well, as unregulated power supplies have 15-35% higher no-load voltage "causing the overall maximum power dissipation to be higher than expected", and the two cancel out.

[nigelwright7557]

In class AB the worst case of power into the heatsink comes at 2/3 power and that is 40% of total power is dissipated in the heatsink and 60% into the load.

Class D heatsinks are usually much smaller for the same power.