# Bipolar class-AB amplifier analysis questions

I am analyzing the following amplifier schematic Bipolar amplifier

I'm wondering if their are any transistors that need to be cooled apart from the 4281A and 4302A output stage transistors?

What is the function of R22 and C7 and how do these work?

http://sound.whsites.net/project68.htm

• I do not see any c3856 and a1492 – G36 Dec 30 '18 at 10:14
• They are different from my printed schematic, I will adapt, thank you – J. Joly Dec 30 '18 at 10:15
• D2/D3 should be thermally coupled with Q5/Q6 and Q4. – G36 Dec 30 '18 at 10:32

And the purpose of the Zobel network is to ensure that the amplifier output stage sees some "real"(resistive) load (($$\10\Omega\$$) at very high frequencies. And they are here to prevent high-frequency oscillations.

You will have thermal distortion in the NPN diffpairs Q2 and Q3, and in the wide-voltage PNP Q4. Q4 is particularly vulnerable, because there is no emitter resistor to provide local-current-stabilizing feedback.

Extremely tight thermal coupling between Q2 and Q3 may reduce this distortion; plastic cases cause poor heat transfer and thus poor tracking.

Consider using cascodes on each of those 3 transistors.

Single tone sinusoidal distortion tests will not notice this.

Impose an Intermodulation Test, say 20Hz and 10,000 Hertz tones provided by separate function generators; use 3dB or 10dB pads (attenuators) in the 2 paths before you resistively combine the 2 tones.

You will find the 10,000 Hz output to now have 20Hz sidebands.

Also, at high enough voltages, and high enough collector currents (about 50 volts and 10mA) your Q4 will exhibit thermal runaway. A cascade should prevent that.

You might find the Safe Operating Area curves to be interesting. The MJE15033 (Q4) datasheet, Figure 1, shows the Thermal Resistance down to 100 microSeconds pulse width. Why? Because the heat has no time to exit the silicon die, so the apparent thermal resistance is "excellent"; another interpretation is that the heat generated ---- almost totally in the collector bulk region ---- is free to move up and rapidly heat the crucial emitter-base junction, setting up the transistor for positive-feedback thermal self destruction.

================================

If you wish to improve on the thermal distortion, then use matched-pair NPNs, located near each other on the SAME DIE. The CA 3046, with 5 devices, two as diffpair and others to use in cascading, is useful except for the voltage ratings: only 20 volts.

On the other hand, if the transistors are thermally nearby (1micron = 11.4 nanosecond tau; 10micron = 1.14microsecond tau; 100 micron = 114microsecond tau), your low-frequency mismatching should be reduced and so should your vulnerability to up-conversion (AM up-shifting).