Heatsinking drivers, do most drivers get very hot?

[EWorkshop1708]

Sometimes we may want to mount our drivers on the PCB but away from or off the main outputs heatsink.

What would be some general things for a DIY'er to look for when figuring whether or not to heatsink the drivers for an output stage?

Do drivers themselves generally put out much heat in high power amps? I've always put drivers on the main heatsink, but now I want to individually heatsink some drivers on an upcoming amp

Some amps I've seen output as much as 200W RMS @ 2 ohm with TO-126 drivers and no heatsink, and they are hot. Probably OK, but for a DIY setup, I'd prefer much cooler drivers. I figure even a small heatsink may help, but I'd rather use a DIY example, rather than a consumer amp that runs too hot.

[Ian Finch]

Drivers for typical EF O/P stages in class AB have virtually constant dissipation, zero to max. output, so It won't get worse as the wick is turned up.
Though counter-intuitive and not the traditional method of simply sharing the main sink, only enough sink for the temp. at idle plus an allowance
for internal case heating may be fine, since there can't be a benefit from trying to get the drivers to track the output transistors.
It should be fairly obvious though, that higher voltage rails and multiple output pairs will result in higher dissipation in the drivers at any given input level.
See Self or Slone for reference.

If you want some visual clues, browse the clones on Epay. There are several examples there, often using just T0220 sinks on small - medium amps.

[Bonsai]

Can poetry's circuit of the output stage. We canteen calculate the dissipation, and from that some guidance on the output driver driver temperature rise over ambient.

[Ian Finch]

Hmm..that's some translator, there.

[AndrewT]

The quiescent dissipation of the driver is governed by the emitter resistor current and the current fed to the next stage.
Ic * Vrail gives the dissipation which for a typical 100W into 8ohm ClassAB amplifier is around 300mW to 600mW.
An unsinked To126 or To220 can survive this. The problem comes when the output stage demands a big current to satisfy a big transient into the reactive speaker load.
The already warm to hot device is asked to pass far higher current (likely far higher Vce) and can exceed it's short term high temperature SOA.

Adding a little 20C/W sink to each driver makes an enormous difference to the operating Tc of the drivers and allows much higher SOA to be used to meet those transient current demands.

However, there is an alternative to using low quiescent dissipation in the drivers.
Instead bias them so that they remain in ClassA till much higher output current levels are reached.
This increases the ~400mW Pq (50Vce & 8mA bias) to ~5W of Pq (50Vce & 100mA bias).
These need much more dissipation capability to keep some SOA reserve for transient output currents. Design the drivers' heatsink requirements just like you would do for an output stage taking account of "normal" loading and transient loading and duty cycle and ambient temperature etc.

Don't guess.
And this applies to the low Pq and high Pq methods of operation.

[EWorkshop1708]

Besides quiescent dissipation, I'm mainly concerned with dissipation under load.
Especially loud volumes for long periods of time.

For example with an amplifier stage +/-37V Rails 1 ohm load.........37A peak in the output stage.......
I calculate for worst case, that drivers will absorb a peak of 3A, because of beta droop in the outputs.

However, quiescent current of 65mA at idle for drivers (2.5W)

Scary part 37V x 3A = 111W of peak output from drivers. I'm afraid that will get them very hot, even using larger drivers (MJL0281/0302) may only help some.

[AndrewT]

Use the Bensen spreadsheet to evaluate the driver.
It does a similar job to evaluating an output device when stressed in continuous high temperature/high voltage/high current operating conditions and by interpretation can be used to estimate transient operating conditions.

Scary part 37V x 3A = 111W of peak output from drivers.

This is impossible.
Even a very reactive speaker load can never impose this condition on a driver. Pure 4uF as a speaker load cannot do that to a driver.

quiescent current of 65mA at idle for drivers (2.5W)

A pair of 15W To126 attached to a 5C/W common sink will operate at Tc~=25 + 5 * 5 + 2.5 * 5 ~=62.5degC, reducing the continuous maximum SOA to ~ [125-62.5] / [150 - 25] * 15 ~ 7.5W (at below the Vce knee in the SOA curve). This 65mA/2.5W of bias is stressing the drivers to ~33% of their rating.

That leaves a margin for transient SOA stresses.

[Ian Finch]

I think you miss the point that... "In all variations of the EF configuration, (driver) power dissipation varies little with output." see fig 15.1 - Audio Power Amplifier Design Handbook, 5th ed. chap.15 -Thermal compensation and Thermal Dynamics. This does not apply to CFP designs, but they are less likely DIY candidates.

I suggest then, this means the driver does not simply mirror the output transistor dissipation and vice-versa. It actually runs at fairly constant dissipation barring errors like clipping so is not a serious heat concern for current demands, only from external heat from the output sink and transformer heat inside the case. Thus the dissipation is all but steady and the sink can be rated based on quiescent conditions, with ~20% to admit the small variation of the common EFII type, before ambient temp. is factored in.
Yep, sounds unusual, but that is the science.

[Bonsai]

Originally Posted by Bonsai

Can poetry's circuit of the output stage. We canteen calculate the dissipation, and from that some guidance on the output driver driver temperature rise over ambient.

LoL!!

Holy crap! what the hell was that?


What was supposed to come out was:

"Please post up a schematic so we can look at it and from that estimate what the temperature rise above ambient will be"

[Ian Finch]

Originally Posted by Bonsai

LoL!!Holy crap! what the hell was that?

'Seems you have a third party in the emailer..