Ideal operating temperature for transistors?

[starbase218]

Two days ago, I told my uncle about the amp I've build (ESP DoZ), and how I'm tweaking it. I told him my heatsinks get about 44C when the amp is on.

Then he said to me that transistors have an "ideal" operating temperature, at which they perform best. He said that, for most transistors, this would be around 75C. He also mentioned a Mark Levinson amp, which he said includes special electronics for keeping all output transistors at the same (constant) temperature.

I had actually observed that the sound improved as the amp was heating up, but I thought it was the result of the quiscent current increasing, rather than the temperature delta. So my question is, is there such an ideal operating temperature for transistors? I can't find this anywhere in the datasheets for the MJ15003 (the BJTs I use).

[Jonathan Bright]

Interesting point. I've ocasionally wondered about this myself. I don't know enough about the finer points of the internals of semi's so I can't speak with confidence but a couple of things come to mind. Ben Duncan ran two series of excessively detailed articles on opamp based pre-amps in an English magazine (Hi-Fi News and record Review,1984 and then some 5/6 years later) He always maintained that he liked to run his opamps at a high temp. His designs were aimed at the top end of the market and all factors seemed to be optimised for sound quality. (I'm not debating the merits of op-amps verses discretes at this point.) Again, I am not confident but I recall that I've seen suggestions that high current/high temp elevates the "averaging" effect of the electrons in the junction and hence produces a smoother wave form. I think that Stan Curtiss and Malcolm Hawksford were parties to this approach but it is many years since I saw the references and the physics were starting to get beyond me so again I'm not up for an argument on this as I can't advance the debate on these issues. Also I think that they were concerned with the potential effects of extremely small currents in low noise designs. Is "Brownian motion" relevant here?

[I_Forgot]

Some people may argue about "optimum" temperature for "optimum" sound, but the inescapable fact of semiconductors is that the higher their operating temperature, the shorter their lives will be.

This is why we mount semiconductors on heatsinks. The heatsinks are generally sized to be a compromise between cost/size and transistor operating temperature/life. Without a heat sink, the semiconductor lives would be measured in minutes or seconds. With a heatsink, they last longer. Ever see an amp that claims to be a "no-compromise" design? If the heatsinks exhibit a temperature rise at all, a compromise has been made.

This is also one reason why many transistor power amplifiers are designed for class AB or B operation. The lower bias current means lower heat and longer life on a smaller, cheaper heatsink.

A case or heatsink temperature is quite a bit lower than the die/junction temperature due to the thermal resistance of the bulk semiconductor material, the mount, and the case. If a heatsink is too warm to touch you are likely to be cooking your semiconductors. Don't expect them to last very long.

Run your transistors hot if you dare, but think about what they may take with them when they eventually burn up...

I_F

[davidsrsb]

Heat sink temperature only matters for health and safety if the user touches it. The die temperature is what matters and I like to keep below 100C for reliability.

Opamps run hot will be noisier than cool. Cmos and jfet types will have much higher input leakage when hot. They are tested at room temerature so that seems best.

[Pjotr]

Originally posted by I_Forgot
Some people may argue about "optimum" temperature for "optimum" sound, but the inescapable fact of semiconductors is that the higher their operating temperature, the shorter their lives will be.

Hi,

Besides that there is also the temperature cycling between on and off of the apparatus: mechanical stress. But it is not the heat sink temperature that matters, it is the junction temperature of the chip itself. Even if the heat sink is at 40 deg C, the junction temp can be 100 deg C. In general room temperature gives longest life but depending on the quality passivation of the chip, 70 - 100 deg C will be ok for the junction temperature in the long term.

Some parameters move with temperature, so the design of the circuit depends on that. In that way you can say there is an optimum operating temperature. But that can be at room temperature as well as at elevated temperature or at –60 deg C if noise is a concern.

Cheers

[starbase218]

I use 2x 2 BJT's in paralel in my DoZ, so each one will be dissipating about 13 watts (35 V at 1.5 A current). According to the datasheet, the maximum Rth j-c of the MJ15003 is 0.7 K/W, so the increase would be about 10C. I believe the BJT's don't get much hotter than the heatsinks (when I can touch the heatsink I can usually touch the BJTs as well). I used mica washers and thermal compound to mount them.

Is it possible that the amp might sound better with only 2 BJTs (because the junction would get hotter)?

[AndrewT]

Hi,
you should add Rth jc + cs to find delta T above sink temp. About 20Cdeg in your 13W case.

If you half your devices and double the device dissipation then Tj will be about 40Cdeg above Tsink. Now it's getting hot in there.