What causes BJT based amplifier distortion?

[edbarx]

The obvious answer is active device non-linearity, but this is too generic and abstruct. BJTs have an exponential input characteristic, which means, the more current passes through the base, the easier it is to pass more. The output characteristic, that is, the collector current at a constant base current, is not a horizontal line. It is neither a line with a positive gradient, but a curve albeit with small positive gradients. As if these imperfections were not enough, there is also a voltage dependent capacitance formed by both depletion layers at the emitter-base interface and at the base-collector interface. Since, a transistor silicon chip is thin, the emitter-base depletion layer and the base-collector depletion layers are in close proximity, and the electric field at the latter affects the electric field at the former. All these unwanted characteristics of NPN and PNP doped silicon crystals contribute to a transistor operation that is not simply linear, but depends on device voltages and applied frequencies.

Since, this forum is intended for enthusiasts to build their own amplifiers, I would like those who have primarily the understanding of the concepts to comment. The question is how does amplifier topology circumvent these unwanted transistor properties? I would like the discussion to first consider simple three stage amplifiers, differential inputs, VAS and the power stage including its driver.

Please note, I am not asking to repair my amplifier. That, thanks to this forum, specifically, Mooly and mjona and other contributors, still works and is daily filling my home with beautiful music.

[indra1]

Perhaps Dr. Arto Kolinummi book can shed some light for you. Audio Power Amplfiers - towards inherently linear amplifiers.

[nigelwright7557]

A transistor doesn't need to be linear as feedback forces it to be linear.
A typical LTP front end just matches output signal to input signal.

The only place this can cause a problem is crossover distortion.
And this has been solved for many years with a bias circuit.

[edbarx]

Thank you for the link. I read the book preview and it is clear the book has been properly proof read before it was sent for printing.

[edbarx]

Originally Posted by nigelwright7557

A transistor doesn't need to be linear as feedback forces it to be linear.
A typical LTP front end just matches output signal to input signal.

Mathematically, for global negative feedback to reduce the difference between the non-inverting and inverting input to zero, the open loop gain must be infinite. Besides, an infinite gain can never be achieved, if such a gain were possible, it would send the amplifier into self oscillation. Self oscillation is why an amplifier employing global negative feedback, has to have its open loop gain reduced, using Miller capacitor compensation, notwithstanding real amplifiers do not have an infinite open loop gain. The capacitor shunts the main audio signal to the VAS's collector, and this shunting becomes more and more pronounced with increasing frequency. So, an amplifier employing global negative feedback behaves differently for all frequencies in the audio range and beyond. At the lowest extreem, negative feedback compensation is most effective because of more open loop gain. However, this effectiveness, decays with increasing frequency, forcing an even bigger voltage difference between the inverting and non-inverting inputs for the same output voltage. This means, the distortion components, that is, the non-linear components that should be added to the distortionless signal to get the voltage difference signal, increase with increasing frequency.

Global negative feedback as used in audio amplifiers is a sword with two edges: it simplifies amplifier design and it introduces problems that would not be present without it.

[Mark Tillotson]

Bipolar transistors are fairly linear if used in common-base or emitter-follower configuration though - and not too bad if current-driven in common-emitter. Its only voltage drive that exposes the exponential characteristic, and good amp designs seek to minimize that.

For instance the standard 3-stage BJT amp uses a differential pair, then a current-driven VAS, then emitter-follower output devices.

The differential pair linearizes the characteristic to arctangent, and global feedback then reduces the signal swing over that arctangent region to a very low voltage indeed.

The VAS is linearized at higher frequencies by a capacitor, which also serves to swamp the non-linear capacitance. The capacitor means the global feedback decreases at higher frequencies though, but then there is less VAS distortion for it to cure due to the capacitor.

Basically clever design reduces non-linearities.


There are also some clever front-end designs using the translinear loop principle, where exponentials are used to advantage.

[N101N]

You can`t in any way circumvent the bipolar structure comprising two strongly distortive junction diodes.

[stanislav1957]

How, then, do they achieve ultra-low distortion on these transistors ?

[N101N]

They don`t.

[stanislav1957]

Originally Posted by N101N

They don`t.

Yes, they seem to reach it.