What's wrong with variable bias?

Lots of people seem to dismiss the Sony TA-A1ES as "not really class A" because it implements variable bias. Based on my lay understanding, while a sufficient deviation in bias when using a fixed quiescent current will result in distortion (clipping / saturation), a proportional deviation in voltage which keeps the q-point on the dc load line of the transistor will in fact result in linear operation. In other words, I don't see what's going on that makes it "not really class A."

Showing 1 response by kijanki

lostrak, In class A both output transistors operate all the time. Current thru one transistor is equal bias current plus half of the load current, while current thru other transistor is equal bias current minus half of the load current. Because of that linear range of output current is from -2x to +2x bias current (-2IB < IL < +2IB). Adjusting bias current lower will reduce power losses, but also will limit output current. Predicting ahead of time magnitude of incoming signal might be difficult while transistors will be forced to operate in less linear region.

Class AB with adjustable bias is not the best either. In class AB feedback has to be much deeper in comparison to class A (to reduce nonlinearities), so the damage is already done (TIM), while switching between classes switches also amplifier’s transconductance (input voltage to output current gain) since in class A two output transistors work and in class B only one. This will also add distortions.

My Benchmark AHB2 is a class AB amp with non-recursive feedback (separate error amplifier) and adjustable voltage rails (class G and H). They obtained AAA (Achromatic Audio Amplifier) patent from THX Ltd. (George Lucas). This itself reveals how difficult is to design such thing.