I've been doing electronics since 1977 but never seen this solution published anywhere before, - so . . .
Class B (bipolar) amplifiers typically have crossover distortion at low signal levels. Class AB attempts to reduce this distortion by keeping both transistors slighly biased on. It's still a balancing act.
Simple solution: Put a constant current sink or source between the output node (ie between the transistor emitters) and a power rail. Either rail is fine. Scale the current to dissipate just a couple of watts.
Result: One of the output transistors is biased 'on' at zero signal, while the other is off. The crossover point is now offset to where the distortion is virtually unmeasurable and certainly inaudible. Supply noise rejection is excellent because the constant current thing is Hi-Impedance.
Any questions, email me at....
email address removed.
Class B (bipolar) amplifiers typically have crossover distortion at low signal levels. Class AB attempts to reduce this distortion by keeping both transistors slighly biased on. It's still a balancing act.
Simple solution: Put a constant current sink or source between the output node (ie between the transistor emitters) and a power rail. Either rail is fine. Scale the current to dissipate just a couple of watts.
Result: One of the output transistors is biased 'on' at zero signal, while the other is off. The crossover point is now offset to where the distortion is virtually unmeasurable and certainly inaudible. Supply noise rejection is excellent because the constant current thing is Hi-Impedance.
Any questions, email me at....
email address removed.
Last edited by a moderator:
https://en.wikipedia.org/wiki/Class-XD_Amplifier
BTW putting your email address in the open internet may not be a good idea...
BTW putting your email address in the open internet may not be a good idea...
Yes, not recommend to put your email in a public place. The spam bots will love it! I can remove if you like.
Member
Joined 2009
Paid Member
The current source scheme actually increases the crossover distortion. True, it moves it away from the zero crossing, but tends to increase the level of crossover distortion once the crossover is actually reached. At the higher output level where it does occur, it is easier to mask. In the class A region, it is free of crossover distortion. When this is done with op amps and such, the current source is made large enough to supply the full signal load current so it STAYS in class A. At several k ohms load thats not too hard. Even an LM358 is listenable this way (provided you don’t slew limit).
Hi Pano, Yes, please remove my email address - and thank you!
Also thank you to all for the feedback - very interesting..
Thank you again for the replies. Thought I'd come up with something new and useful but the education is well received 🙂
IIRC, D. Self has proposed a similar scheme, and it has even been used commercially (is it Cambridge Audio, I am not sure). Anyway, trying to impose an external, fixed bias as a "keep alive" measure has never been effective, and often proved counter-productive.
If you want to do it the right way, you need to wrestle with the real problems: using an effective autobias, class I, or anything better, but simple, semi-passive measures do not actually work.
I have rejected innumerable numbers of "never off" designs because they brought more artefacts than the ones they were supposed to solve. I have found -a limited- number of solutions, but most were not simple.
There are a few exceptions, and you can find them by making a search on my threads, but the simple ones do not offer really high performance: they are a convenient substitute for conventional solutions, without the hassle of bias problems.
An elegant solution is Walker's current dumping, although higher order effects tend to degrade it, unless you are really prepared to tweak it to its optimum
If you want to do it the right way, you need to wrestle with the real problems: using an effective autobias, class I, or anything better, but simple, semi-passive measures do not actually work.
I have rejected innumerable numbers of "never off" designs because they brought more artefacts than the ones they were supposed to solve. I have found -a limited- number of solutions, but most were not simple.
There are a few exceptions, and you can find them by making a search on my threads, but the simple ones do not offer really high performance: they are a convenient substitute for conventional solutions, without the hassle of bias problems.
An elegant solution is Walker's current dumping, although higher order effects tend to degrade it, unless you are really prepared to tweak it to its optimum
With opamps it’s very easy to bootstrap the output into class A if you have a discrete buffer at the output - I use this technique in the X-Altra HPA-1 Headphone amp. Since the drive current into the class A discrete buffer is relatively low, the opamp class A current only has to be a mA or two. You could easily cuss this approach in a power amplifier as well.
The distortion doesn't go away, it just comes back as the level rises to the displacement point - you still have the same shape wing-spread graph. Doug Self calls this technique XD (crossover displacement), and it does allow distortion to be reduced when the signal is low level, but he makes the point its not hard to keep crossover distortion low enough to be completely inadible using existing techniques. Assuming, that is, that the bias point is stable.
The technique does lend itself to fairly simple automated ABX testing if you make the current sink switchable, which means you can evaluate its real audibility for yourself double-blind more simply than many things.
Check out Doug Self's "Audio power amplifier design" book for prior art on this, he's also got performance data for B v. A v. XD. Chapter 18 in the 6th edition is dedicated to this technique.