This afternoon was the first time I tested my amp at +/-40V, I was actually surprised how hot it got.
Here are a few pictures of the heatsink I bought on ebay that measures 13" X 6" X 1 3/4". By all measure, it's huge. I am testing my own amp with 9 pairs of output transistor. I only use +/-40V rail and high bias of 0.8A just now. I had it on for 15mins with no air flow, just sitting in the room as shown. It got so hot I cannot hold on with my hands, not even close, maybe 1 sec max. From calculation, it's 0.8A X 2 X 40V= 64W of heat dissipation. Granded it is not a very thick heatsink but that size and got that hot with 64W. This gives a very good perceptive on power dissipation of those heatsink, gives a realistic perceptive of what heatsink can or cannot do.
This is the setup I test the heatsink
This is the back view of the heatsink with my OPS pcb.
This is the front view of the heatsink
You can see, the back of the heatsink is a little thin, but still it is 0.2 inches thick, it's by no means thin compare to a lot of other heat sinks. You'll be lucky to get a heat sink half the size of this inside a typical chassis particular with this kind of thickness. Also there is no air flow inside the chassis. I doubt any internal heatsink can dissipate over 25W. That is not comforting for an 80W or higher amp.
Also, I use 9 pairs of output transistors in TO-264 package to avoid beta droop, more surface to lower the thermal resistance from transistor to heatsink. I only run +/-40V and only 75W into 8ohm or 150W into 4ohm. I don't care about how many watts, I only care about the quality of the watts.
Of cause, TO-264 is smaller than the MT-200, but that will do!!!!
I first bought a smaller chassis with heatsink similar size to this, a little smaller, but 2" fins instead of 1 3/4" and 3/8" thick back. So the power dissipation is about the same. The more I calculate, the more uncomfortable I got, so I spent $350 buying a Krell clone 50W class A chassis. I always question whether I was wasting money. The result today make me feel a lot better, that it's a right move. Save the smaller chassis for a lower bias amp, maybe do lower voltage, make it a bridge monobloc that get back the power to use heatsink of both sides for a single channel.
The one on the left is the new Krell clone, the right one is the old smaller chassis.