Heatsink Placement

[Jrp27]

P.S.to #16.. Note that the bias temperature sensing on the original PCB at least, is a long way from the output stage and uses the the case to conduct heat toward the sense transistors, so it actually settles a bit faster than older models. Any independent heatsink will most likely impede this thermal link and could result in a loss of bias control. Even the typical (and tedious) 30 min settling time for a stable bias with Naim's old products may never happen at all.

NAP and many other quasi-complementary designs are fairly forgiving of bias setting but if there is no effective bias servo action, you may also find yourself in trouble with thermal runaway, even with a perfectly working fan and heatsink.

I believe 2N5089 transistors are used as sense transistors. They are mounted below the PCB.

Now that I won't be using the chasis as extended heatsink so perhaps I could clamp these transistors to the heatsink.

What do you think?

 

[Ian Finch]

I've thought about this approach (as applies to EF design output stages) many times but it results in overcompensation of quasi-complementary. designs. i.e. the bias rises initially but levels out and then falls too much. It is more usual to mount the sense transistors close to, but not on the heatsink. I would try fitting insulated leads, twisting them together a few times and sleeving with heatshrink or similar heat resistant tubing to keep them neatly together.

Fasten the transistors with small and very thin metal clips covered in thin heatshrink for insulation if necessary, at the edge of the slot where shown. This will keep them out of harm's way and avoid any need to refit them when you need to remove the heatsink. If the response is too slow, or bias won't hold as the amplier warms up, then you could still move the sense transitors a little closer to the the output transistors. Note that with loose thermal coupling like any NAP model, you should not do a final adjustment until the amplifier is fully warm.

 

[Jrp27]

Right. From what I understand. Coupling the transistors to the heatsink will result in over-compensation.

How about this...I attach the sense transistors to the heatsink with a thin layer of polyester cloth sandwiched between mica sheets.That should limit the heat transfer to acceptable levels.

What are your thoughts on this?

 

[Ian Finch]

I wouldn't just increase the thermal resistance that way, with the temperature reference being only that of the output transistors. You may not appreciate that we don't really need to raise thermal resistance but we do need a wider range of thermal conditions such that the CFP driver temperature and air temperature of the output stage, also figure in the bias regulation characteristic. It's quite tricky to do this in dynamic conditions.

In a practical quasi amplifier, bias is determined in part from output transistor temperature, part internal air temperature and part driver transistor temperature. Usually, that provides acceptable bias control - at least within the range needed here for consistent high quality audio without disasters.

However, if you plan to locate the amplifier in a deep case, it likely will never warm up enough internally to reach equilibrium and even then, not track as well as it could. It just adds another problem. The controller will always be a compromise between dissimilar thermal characteristics in either EF or CFP half but optimal location of the sensor seems to be the easiest and most reliable way to establish and maintain good bias - part from the output transistors, part driver temperature and part case air temperature. I think it will be problematic as you are suggesting though.