What is a Gain Clone exactly?

[Morse]

Hi Ron;

Good to see you here too!

All the best,
Morse

[PRR]

> Do you think that a 4"x6" sheet of 1/8" thick aluminium will suffice as a heatsink for a (approx) 12wpc chipamp?

VERY ROUGH rules of thumb:

A surface runs about 100°C per watt per square inch.

So 4"x6" with both sides exposed is about 100/(2*4*6)= 2°C per watt.

A class B amp at full sine power dissipates about half its output power (and much less with unclipped speech/music signals).

Two 12W amps is 2*(12/2)= about 12 watts. I would round-up since chip-amps sometimes have 4-7 volt losses, so working at lower voltage they are not near theoretical efficiency. 20 watts dissipation is surely conservative.

20W*(2°C/W)= 40°C temperature rise. Is that OK?

The old habit for glass-seal transistors in commercial class B music service was to keep temperature rise down to 50°C rise, because the constant hot/warm/hot/warm cycling with the musical beat would fatigue the seals even though the continuous-duty rating was 100°C or so. Seals are better now, and the epoxy seals seem to beat the old glass-seal TO-3, and you are probably not going to abuse your amp as hard as a commercial sound system. 40° rise is quite safe. (And the National chips have a thermal-cutout to control temperature cycling, and IIRC it is set much higher than 50°C. I'm sure they thought about this and tested their seals, so my old 50°C rule is old-dog trivia.)

If the heat has to flow a long distance, it needs a fat conductor. The thickness to length ratio of heatsinks should be held to 1:10 or maybe 1:20. 1:10 is nice for parallel-flow as in fins; 1:20 is OK with radial-flow as for a small chip in the middle of a large area. With 0.125" plate, you would like the longest heat-path to be 20*0.125= 2.5".

Optimal chip location and heat-flow distance: To a rough approximation, you locate two chips as-if you had sawn the plate in half and centered each chip on its plate. So draw a line down the center of the 6" dimension (2" from each long edge), then divide it in quarters. Put a chip at the 1/4 and 3/4 mark, or 1.5" from each short edge. Each chip is then 2" from a long edge and 1.5" from a short edge, and 3" from a corner. We see that the corners are less effective because they are more than 20*0.125=2.5" from the chip, but not much more, and a lot of area is inside the 1:20 thickness/length rule of thumb, which is only a crude guide to begin with. (And if other layout factors suggest off-centered location, it will still be pretty good.)

I think you could get away with half this area. But if you have the plate and it fits your box, use it. It is always better to over-cool than under-cool.

[fmak]

[QUOTE]Originally posted by peranders
[B]The advantage of LM3875, 3886 is the simplicity and all protection schemes. I think this is the main advantage.

Eventually I will make pcb's but so far I haven't heard a LM3886 live yet but I have no doubts that it's quite good.
--------------------------------------------------------------------------------
I have just built the LM3875 with dual 22 V dc supply. Clips at about 11V into 8 ohms and rings on a sqaure wave into resistive load. Distortion at 4V is only about -63 dB thd+N. I will make the other channel and listen.

What I don't like is dc shift with ramp change in input signal level, about 60mV on a digital meter with quite slow response. This evenyually settles to about 5mV.

[Morse]

Thank you very much PRR!!

That is a lot of REALLY useful info - just what I was looking for, in fact!

To maximise cooling along the length of the heat sink, I could mount a separate sheet of .125" stock underneath (frankly I've got enough plate stock left to double 'em up to .25" for the whole top deck, if that's desireable...).

Do you think there'd be a problem with using a mica spacer between the chip and the heat sink (i.e. will it change the numbers on heat transfer significantly)? I was planning on using some Radio Shack heat sink compound.

Thanks again! Your help is greatly appreciated!
All the best,
Morse