In our quest to eliminate crossover distortion, we often venture into dangerous territory with transistor physics that result in a phenomenon called thermal runaway. As transistor and diode junctions heat up, their forward voltage drop… um, drops. In a Class B or Class AB amplifier, if the bias generator is fixed, or if it does not have strong enough thermal compensation, then the output stage can become overbiased, causing excessive quiescent current to flow, thus increasing the resting power dissipation of the amp, heating it further. This cycle can very easily destroy Class AB amplifiers that might be stable at room temperature.
To combat this problem, and in order to design Class B amplifiers that cannot operate into Class A mode, I devised this method of testing a power amplifier biasing scheme. I used it to develop my Trimless Class B amplifier, and can say that it is very effective, if a bit conservative.
The first thing you will notice is that there is the voltage regulator with its output shorted to ground. This is intentional. We’ll get to that in a minute.
R3, R4, and Vbias will be your bias stage under test.
Q1 and Q2 represent your output stage as is will be constructed in your final amp. If you are using a Darlington configuration, then include your driver transistors, on or off of a heatsink, your output devices, on a small heatsink, and any driver load resistors that may be needed.
The 1K load resistors are there to provide a very small load to your output stage, without having to worry about it becoming over-biased and blowing up in your face. We don’t really care what this thing does under heavy loads, because thermal runaway happens when the amp is in a quiescent state.
The most important part of putting this whole thing together is putting any devices that will be on the main heatsink in your final amp on a small heatsink together with the shorted voltage regulator. If your driver transistors will be on the main heatsink with your main output devices, then bolt them on too.
How it Works:
We will be taking advantage of both of the built-in safety features of voltage regulators. First, the current limiting will keep the device from actually shorting out your power supply. At the same time, it will be heating up very quickly. So quickly, in fact, that the device’s thermal protection will kick in after a couple minutes and keep it around a toasty 125°C.
Because you bolted the regulator and the output devices to a small common heatsink, this will give you a cheap way to watch how your biasing scheme works from ambient temperature all the way up to OH SHIT THAT’S HOT! OWOWOWOWOWOWOW!
The forward voltage drop of your output devices will fall as they heat up, so watch your voltmeter as they do. Your biasing scheme is in the danger zone when your voltmeter is displaying a positive number. For a Class AB amp you will actually want a positive number here, and this is where you have to do some math.
Interpreting the Results
Carefully watch your voltmeter from the moment you switch on the power, until the small heatsink temperature stabilizes. Take note of the highest (most positive) voltage that you see over the whole temperature range, which will probably occur when the output devices are at their hottest.
For Class B, you want the voltage to be negative or very low (less than +10mV) across the whole temperature range. If it goes solidly above zero, you are entering Class AB mode.
For Class AB, calculate what the resistance will be between your output devices in your final build. If you use a single pair of output transistors and .1Ω emitter resistors, this number is .2Ω. If you have two pairs of output transistors, and .47Ω emitter resistors, then .47Ω.
Divide your highest voltage number by your resistance number to get your worst-case quiescent current.
Multiply the quiescent current by your total power supply voltage. This is your total quiescent, or Class A, power dissipation.
Now comes the guesswork. Try to estimate the total thermal resistance from your output devices to the heatsink. Unfortunately, this is way beyond the scope of this article. Google it. Several days and pots of coffee later, you should come up with a °C/W number.
Multiply this number by your Class A power dissipation. This is how much hotter your amp will get as a result of operating in Class A mode.
Let’s Work Through an Example
You have an amplifier with these specs:
- ±35V power supply
- A Darlington output stage with two main output devices per side
- Driver transistors are lightly loaded (they won’t get hot), not mounted on a heatsink
- .1Ω emitter resistors
- Bias tester gives a maximum of +100mV between the output emitters
- estimated 1°C/W total junction to air thermal resistance combined for the whole amp
There are two pairs of emitter resistors effectively in series-parallel, so the resistance between each half of the output stage will be .1Ω. That means that (100mV / 100mΩ) = 100mA is the worst case quiescent current.
100mA x 70V = 7W power dissipation due to quiescent current flow.
7W x 1°C/W = 7°C rise in heatsink temperature due to quiescent current. Not an insignificant amount. To make sure that this amp will never be in danger of thermal runaway, you need to subtract 7°C from all of your safe operating area calculations.
If you were building this amp, you might consider using a larger heatsink than you originally anticipated, or perhaps adding some thermal compensation to the bias generation.