Apart from using a well designed Class A configuration for the audio power amplifier output stages, does any one know of a Class AB configuration that offers almost complete freedom from crossover distortion across the normal range of hearing from normal listening levels right down to near the lowest threshold of hearing? Clearly any such design must offer near zero disturbance in passage through the crossover regions, at all signal levels, but especially at the lowest signal levels.
Maybe something like this? http://www.diyaudio.com/forums/solid-state/182554-thermaltrak-tmc-amp-25.html#post4038551
That's a very wide question.
Better question is what you want to do, and is this an academic exploration?
A short answer is that there are a variety of methods by which engineers and designers attempt to optimize class AB designs, quite a large variety. There are examples that succeed in terms of different aspects. Quite a few have been thrown up the flagpole here.
Fwiw, it is essentially impossible to have a class AB amp without some sort of "disturbance" that can measured due to changing from class A to B and from the positive rail to the negative.
So, what's ur interest in this topic?
Better question is what you want to do, and is this an academic exploration?
A short answer is that there are a variety of methods by which engineers and designers attempt to optimize class AB designs, quite a large variety. There are examples that succeed in terms of different aspects. Quite a few have been thrown up the flagpole here.
Fwiw, it is essentially impossible to have a class AB amp without some sort of "disturbance" that can measured due to changing from class A to B and from the positive rail to the negative.
So, what's ur interest in this topic?
Over my years as a professional electronics engineer in telecommunications, and even longer, since my lifelong fascination with with audio and radio began in my childhood, I have followed with great interest (and explored myself when possible) the arguments between audiophiles about the various merits of power amplifiers in various configurations such as: Class A vs Class B, Valve vs Solid State, then Class AB, (that in most cases was no more than Class B with a higher quiescent current setting and a transfer function around the cross over region more like the hind leg of a donkey than anything linear), then the growing popularity of MOSFETs and more recently CLASS D which is, I believe best suited to Rock Bands and PA Systems but is NOT sited to the true pursuit of High Quality Sound Reproduction, as expounded so well by the likes of James Moir, in his 1958 book of the same name, and others such as John Lindsey Hood, (Arthur?) Williamson, Doug Self, Ted Jordan and (John?) Baxandale here in the UK - with articles and prototype designs for amps and simple but musically satisfying loud speakers, published in The Wireless World Magazine. I am now particularly interested in any recent work done that seeks to control the THD, TID and IMD at the lowest output power levels, in power amps in that magical 'quiescent region' of the typical, 'so-called' Class AB power amplifier since the human ear has a dynamic range of 10^12 which far exceeds the measurement range of most audio test equipment and yet deserves much more attention. It is probably the sound between the notes and the way the sound 'decays to nothing' that really defines the realism of reproduced music and speech and not whether it is realistically loud and fully ranged from 20 Hz to 30KHz - "just like being right there in front of the orchestra"!
Many thanks Dadod for this design information - it looks very interesting. Agood example of what we used to call a Darlington or Super Darlington configuration and is basically a series of two or more complementary emitter followers in the output stages. The THD figure at 1 watt seems to offer the possibility of sustained low THD at much lower output power levels - which was what I was seeking. However, the proof of the pudding is in the listening of course..
The simplest way would be to load the output with a current source, to move the crossover distortion away from the zero crossing.
Preferably, this would keep the NPN transistor side turned on.
There are a variety of approaches that you can find here.
You do know about "Hawksford" error correction?
Also, do you know about Halcro? (An amp mfr, but you should read the patents)
There are discussions of that amp here.
Search online and here for Syn08. He left this forum, but his website has an amp or two that might interest you also.
Then too there is an idea probably first started by the Japanese, where you float a pure class A amp between powerful class B sides. That might fit your bill?
Depending on the implementation, the class A amp might never be driven to saturation (bootstrapped).
All sorts of neat ideas. Someone could take all the compatible ones and design a super amp of some sort, I guess.
But, a whole lot on the listening side probably depends on the sensitivity of your speakers, unless ur of the club who feels that what goes on at and below the noise floor (acoustic noise floor in your room too!) matters much.
And what is your source going to be that does not add in noise of its own? And do we care if it is correlated or uncorrelated noise?
BEWARE - you have opened a large can of worms!
You do know about "Hawksford" error correction?
Also, do you know about Halcro? (An amp mfr, but you should read the patents)
There are discussions of that amp here.
Search online and here for Syn08. He left this forum, but his website has an amp or two that might interest you also.
Then too there is an idea probably first started by the Japanese, where you float a pure class A amp between powerful class B sides. That might fit your bill?
Depending on the implementation, the class A amp might never be driven to saturation (bootstrapped).
All sorts of neat ideas. Someone could take all the compatible ones and design a super amp of some sort, I guess.
But, a whole lot on the listening side probably depends on the sensitivity of your speakers, unless ur of the club who feels that what goes on at and below the noise floor (acoustic noise floor in your room too!) matters much.
And what is your source going to be that does not add in noise of its own? And do we care if it is correlated or uncorrelated noise?
BEWARE - you have opened a large can of worms!
Member
Joined 2009
Paid Member
I remember reinventing that idea - I was informed that Graham Maynard developed something simple along these lines which he called the GEM amplifier - worth a search perhaps.
Actually, now that I re-read ur post, there are two separate issues.
The first is ultra low distortion and noise.
The second is effects at the "zero crossing".
The Gm "doubling" or "halving" depending on how you wish to view it happens in class AB or B output stages. But it does not happen in Class A.
So you could decide on a Class A design to eliminate that issue.
However that says little about distortion or noise, it merely eliminates one source of possible distortion and noise.
Peeling the pesky onion anyone?
The first is ultra low distortion and noise.
The second is effects at the "zero crossing".
The Gm "doubling" or "halving" depending on how you wish to view it happens in class AB or B output stages. But it does not happen in Class A.
So you could decide on a Class A design to eliminate that issue.
However that says little about distortion or noise, it merely eliminates one source of possible distortion and noise.
Peeling the pesky onion anyone?
the dynamic range is a bit unrealistic
human hearing would be better described as having ~60 dB dynamic range and 120 dB "rangebility" - you need many minutes of silence in anechoic chamber conditions to have the ear's protective and amplifying mechanisms to adapt to hear at our ultimate noise limit
this never happens will listening to music, created by live performers in real rooms, captured by real microphones, played back in even noiser home listening room at real world event SPL
human hearing would be better described as having ~60 dB dynamic range and 120 dB "rangebility" - you need many minutes of silence in anechoic chamber conditions to have the ear's protective and amplifying mechanisms to adapt to hear at our ultimate noise limit
this never happens will listening to music, created by live performers in real rooms, captured by real microphones, played back in even noiser home listening room at real world event SPL
Yes, there is a small idle current, so that for small signals when both devices amplify, the gain doubles. This is in class AB operation. Often, class B also has a small idle current.
Last edited:
"Class B" has different interpretations - literally 0 common conduction angle is a hard point to hit, maintain - so most allow for few mA bias, tiny overlap
I assume any here should at least know of Distortion In Power Amplifiers - output stage bias discussion ~1/2 way down
I assume any here should at least know of Distortion In Power Amplifiers - output stage bias discussion ~1/2 way down
Hi,
GM doubling does not occur in properly biased class aB.
It does occur in high bias class AB e.g. 1/4 class A (rest B).
It does occur by default in some class A designs, e.g push
pull biased to full Class A for a given output load.
Simple answer to the question is no, nothing has
come along recently that changes things much.
FWIW Self has shown a properly biased EFP
aB output has lower distortion than a CFP
aB output below 1 watt, and that intrinsic
distortion carries on reducing with signal
level level reducing, though THD+N does
not as intrinsic noise begins to dominate.
rgds, sreten.
GM doubling does not occur in properly biased class aB.
It does occur in high bias class AB e.g. 1/4 class A (rest B).
It does occur by default in some class A designs, e.g push
pull biased to full Class A for a given output load.
Simple answer to the question is no, nothing has
come along recently that changes things much.
FWIW Self has shown a properly biased EFP
aB output has lower distortion than a CFP
aB output below 1 watt, and that intrinsic
distortion carries on reducing with signal
level level reducing, though THD+N does
not as intrinsic noise begins to dominate.
rgds, sreten.
Last edited:
Care to back up that Gm doubling assertion?
Nothing I have read seems to support that.
The basic mechanism apparently is "both on" or "one on", which will happen in AB regardless of the bias level, it seems?
Regardless I also seem to recall seeing a whole lot of graphs that show the gain non-linearity that results from various bias points and schemes... Ben Duncan (iirc) published some maybe 20 years back now. I think.
Nothing I have read seems to support that.
The basic mechanism apparently is "both on" or "one on", which will happen in AB regardless of the bias level, it seems?
Regardless I also seem to recall seeing a whole lot of graphs that show the gain non-linearity that results from various bias points and schemes... Ben Duncan (iirc) published some maybe 20 years back now. I think.
Here's a very good three part article about this from Linear Audio that appeared in EDN.
Designing a low-distortion audio output stage - Part 1: Introduction, the problem with push-pull
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.