As far as I know even so called AB output works in class A under light load. So what is "class AB" in general? Is it an amplifier which CAN go into class AB under higher loads? And how "class AB" could be "properly biased" if load can change?
And how can I simulate this kind of amplifier to find an optimal solution? Should I step three parameters (bias voltage, emitter resistors and load) the same time? Can I fix at least one of them by sort of rule of thumb?
Should I assume that my design will work in class A most of the time and go into AB in extreme conditions only?
And how can I simulate this kind of amplifier to find an optimal solution? Should I step three parameters (bias voltage, emitter resistors and load) the same time? Can I fix at least one of them by sort of rule of thumb?
Should I assume that my design will work in class A most of the time and go into AB in extreme conditions only?
A = all output transistors are conducting throughout the entire extent of the voltage swing.
B = all output transistors are conducting for exactly half ...
AB = greater than 50%, less than 100%
Due to transistor behavior, "properly biased" class B stages are indeed very very slightly AB in order to make sure that there is no dead-spot at the zero-crossing. By most books, that's still qualified as "class B", whereas AB suggests there's a substantial amount more standing current through the output stage.
B = all output transistors are conducting for exactly half ...
AB = greater than 50%, less than 100%
Due to transistor behavior, "properly biased" class B stages are indeed very very slightly AB in order to make sure that there is no dead-spot at the zero-crossing. By most books, that's still qualified as "class B", whereas AB suggests there's a substantial amount more standing current through the output stage.
I assume we are talking about situation when we have some amount of quiescent current. When output current swing increases transistors goes into cutoff region (NPN on negative swing, PNP on positive). So how can we talk about classes and biasing without taking into account the current swing?
Class B is usually with cross over distortion.
Class AB is without cross over distortion.
Some amps come between this where the class B encroaches slightly into class AB using diodes.
Then there is the argument how much class AB should be biased.
I have found as little as 10mA is enough to get rid of crossover distortion.
Yet some people are convinced it needs 100mA or more.
Class AB is without cross over distortion.
Some amps come between this where the class B encroaches slightly into class AB using diodes.
Then there is the argument how much class AB should be biased.
I have found as little as 10mA is enough to get rid of crossover distortion.
Yet some people are convinced it needs 100mA or more.
But it should depend on output power. Driving 32 Ohm headphones, for example, your 100mA biased amp will be kept in pure class A all the time. So my question is when it should go out of class A. In relation to desired maximum output power.
See e.g. Douglas Self on optimum biasing. In general, a "fat biased" output stage will exhibit worse distortion once it crosses over into AB when compared to an optimum biased AB one. (Note that optimum AB bias is load impedance dependent.) This may be a worthwile tradeoff to make for something that is usually supposed to run in pure Class A. When should it transition into AB? Well, that depends entirely on what your definition of "untypically high levels into typical loads" is. Headphone sensitivity and impedance, for example, tend to vary a fair bit. Then there are thermal and power considerations, too.
BTW, there have been some suggestions that popular usage of class B in particular may be wrong. Typical so-called "class B" output stages (e.g. npn and pnp with bases tied together) tend to have a massive dead zone, arguably those would better be classified as "class C". The real class B would be something like a very lean-biased AB stage on the onset of bias current, though the exact point would be very hard to pin down due to the exponential control law involved.
BTW, there have been some suggestions that popular usage of class B in particular may be wrong. Typical so-called "class B" output stages (e.g. npn and pnp with bases tied together) tend to have a massive dead zone, arguably those would better be classified as "class C". The real class B would be something like a very lean-biased AB stage on the onset of bias current, though the exact point would be very hard to pin down due to the exponential control law involved.
I have read Self's book very inattentively 🙂 I know Self's definition of "proper class B" but I think I missed the key distinction of "proper B": it doesn't have class A region at all. Am I right?
There has been some discussion about what happens when a device is cut off completely... but ignoring that I'd bias it where it sounds best, IF you have the option of biasing in Class A or AB, bias where you think it sounds best - if you hear no differences, then maybe save power and bias low... chances are that A may sound better overall.
I suggest you get the book by Bob Cordell also. His book explain in detail on optimal biasing for Class AB. I like both books they complement each other.
- Status
- Not open for further replies.