Crossover Distortion, the truth

OnAudio

Member
2011-10-13 3:36 am
There are three major classes of amplifiers, class A, A/B and D. Class A does not suffer from crossover distortion, nobody cares about it in class D, we actually put a dead time. Class A/B may suffer from it if not well biased. Intresting solutions to mimic class A include ibias by Krell. Attempts to shift the crossover point of a class A/B from zero such as class XD by Self, trying to mimic single ended class A. A good bias point does eliminate the problem in class A/B leading to better THD figures.
Electronic components do produce structural noises. Class A holds a steady current throughout the amplifier which is converted to heat, providing the much needed heating in winter. Many urban dwellers have nostalgic memories of winter nights spent huddled together around a class A master piece. This steady current also keeps the output transistors and rectifier capacitors structural noise in check. The birth of MOSFETs was what class A was waiting for.
Class A/B and class D share the relationship that the output devices are switched on/off , the difference is that class A/B is switched by the audio signal while in class D we switch the output devices on/off using a higher frequency than the audio frequency, keeping the devices cooler and shifting their structural noise, a side effect is bus pumping by the half bridge, creating an unintended boost converter. A class D amplifier will give you no warmth during winter, however a few cool LEDs might just fool you.
The choice between a switch mode power supply and linear power supply dictates the structural noise that the power supply produces. If your below the age of 25 years you may hear your computers switch mode power supply sing. Everything else around us hums at 50/60 Hz including the hum you see or hear emanating from your body when you touch an oscilloscopes probe or power amplifier input.
A linear power supply is bulky, heavy and expensive, but if well designed blends in with the environmental noise. A switch mode power supply is smaller, lighter and cheaper but sings at a different frequency from the environment probably necessitating EMI treatment.
The modern generation is more accustomed to class D sound from the modern electronic devices such as BT speakers, laptops, TVs, xphones, etc. It may be difficult to pickout a well designed class D amplifier from a class A amplifier. One might even argue that a young person is already conditioned to find class D sound more 'normal'. Class A/B amplifiers are the grail of simplicity and are often coupled with a high quality switch mode power supply with or without rail switching for cooler higher current operation. A popular setup for car audio and professional PA.



Feel free to share your expertise on this matter
 
Class A holds a steady current throughout the amplifier which is converted to heat
Class A average current is steady. In some cases the instantaneous current varies.

Electronic components do produce structural noises.
What is structural noise? Is it what everybody else calls excess noise or shot noise or partition noise or flicker noise?

A linear power supply is bulky, heavy and expensive, but if well designed blends in with the environmental noise.
What is environmental noise? Is it what everyone else calls hum or RFI?

You claim to offer us the truth about crossover distortion, then hardly mention it in your post.
 

OnAudio

Member
2011-10-13 3:36 am
What is structural noise? Is it what everybody else calls excess noise or shot noise or partition noise or flicker noise?
Not quite, caps sing, transformer laminations sing, everything sings
What is environmental noise? Is it what everyone else calls hum or RFI?
Both plus more
You claim to offer us the truth about crossover distortion, then hardly mention it in your post.
Its an open ended discussion for further interaction, theres mention of bias, most people with heat-sink real estate may prefer to bias their A/B amplifier where the 1st Watt is class A
 
OnAudio said:
Not quite, caps sing, transformer laminations sing, everything sings
Ok, so when you say "noise" you don't actually mean 'noise'. That knowledge may help future communication.

Its an open ended discussion for further interaction, theres mention of bias, most people with heat-sink real estate may prefer to bias their A/B amplifier where the 1st Watt is class A
You started the thread. Maybe you should tell us what you know about crossover distortion? Otherwise it could be misunderstood as being merely some sort of publicity stunt?
 
  • Like
Reactions: 1 user
I think that MOSFET output is a better choice for class AB Amp (D. Self calls it "Optimal Class B", that means this Bias current is relatively small, for example 15-30mA). Because BJT output needs precise "optimal bias current" regarding D. Self and B. Cordell. Some variations (+-15mA) introduce Higher crossover distortion. Even Higher current - if it is not A class - does not help here, because of the Gm doubling (when two devices conduct) and droop (when one device conducts) effect. Mosfets do not produce Gm doubling, so when it is more bias current, there is less crossover distortion. When bias current is relatively big (for example 300-500mA) a little Bias variations (+-15mA) because of the thermal compensation slowness does not produce more/noticeable distortion.
 

OnAudio

Member
2011-10-13 3:36 am
Self appeals to the audience that the increase of distortion indicates the presence of gm doubling, on the other hand Prof Leach https://leachlegacy.ece.gatech.edu/papers/classab.pdf
appeals otherwise. Distortion doesn't necessary indicate that one amplifier will sound better than another

I think that MOSFET output is a better choice for class AB Amp (D. Self calls it "Optimal Class B", that means this Bias current is relatively small, for example 15-30mA). Because BJT output needs precise "optimal bias current" regarding D. Self and B. Cordell. Some variations (+-15mA) introduce Higher crossover distortion. Even Higher current - if it is not A class - does not help here, because of the Gm doubling (when two devices conduct) and droop (when one device conducts) effect. Mosfets do not produce Gm doubling, so when it is more bias current, there is less crossover distortion. When bias current is relatively big (for example 300-500mA) a little Bias variations (+-15mA) because of the thermal compensation slowness does not produce more/noticeable distortion.
 
Last edited:

OnAudio

Member
2011-10-13 3:36 am
A theoretical 8Ohm woofer with a sensitivity of 85dB at 1watt/1 meter will require 33mV across 0.22 emitter resistors to operate in class A at 1 watt
V^2/R = W
V^2 = 9 (approximate)

I = V/R
I = 3/9 = 333mA

Voltage across emitter resistors for two pairs of output transistors
= .150 * 0.22 = 33mV

The deal with BJTs vs MOSFETs is thermal runaway
 
Last edited:
Distortion doesn't necessary indicate that one amplifier will sound better than another
If we look at distortion from one point. But if we analyze all possible factors of distortion components, then it should indicate.

The deal with BJTs vs MOSFETs is thermal runaway
Not only. There are many factors, for example Ft (BJT transition frequency) from 30Mhz drops to 0.5Mhz... drop starts when current increases more than 2-3A, in very good transistor "for audio" rated 230W.
www.onsemi.ru.com/pub/Collateral/MJW3281A-D.PDF
Worse things are in not so good BJT transistors.
It also drastically increases crossover distortion, at higher output. Compensation should be strictly overlooked, witch will degrade even more negative feedback at high frequencies causing less feedback, and more distortion.
 
No distortion=no sound.
Even your eardrum will distort.

Not to mention System 1 processes in the brain. What is heard may be distorted, memory of it distorted, explanations of it distorted, lots of distortion everywhere.

A thread is named Crossover Distortion, but appears to mostly be about other things, so the name is distorted, or the topic has become distorted.

What does the OP say? What did you want to talk about? Did you have a question, did you mean to offer some new insights, what?
 
Last edited: