Class-AABB output stage. - diyAudio
Go Back   Home > Forums > Amplifiers > Solid State
Home Forums Rules Articles diyAudio Store Gallery Wiki Blogs Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Solid State Talk all about solid state amplification.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools Search this Thread
Old 19th February 2005, 09:49 PM   #1
Graham Maynard's Avatar
Join Date: Dec 2003
Location: UK
Default Class-AABB output stage.

Hi All,

I've been doing many little audio amplifier experiments recently.

Over a year ago I showed the circuit of my 25W class-A amplifier which was similar to a JLH class-A, but with a mirrored differential input stage in place of the original single transistor.

Someone suggested that the original must be better because its signal path contains fewer active devices. However that is not so, for mirroring reduces waveform distortion and much improves loudspeaker damping control without lengthening (slowing) the signal path.
I have since double-checked by directly comparing single transistor input with a four transistor mirrored differential: The latter is superior, especially on transients. It is also stable on sudden loud soundbursts where the single transistor allows 'cone breathing' due to imbalanced current draw. Nor is there a start-up thump, nor any bias offset to worry about.

However 25W can be found inadequate.

A couple of months back 'Lumanauw' asked about output stages that would never drop out of conduction in order for there to be no crossover distortion. I had already been working on this with the mirrored differential JLH, and this was why I suggested class-A plus class-AB plus class-B output biasing all working together.

As I would never publically release any circuit that I have not thoroughly auditioned, and health problems slow me down, there has been a delay in me releasing this first circuit. It is constructed as shown in the circuit.

My simulator stores a 200W-4 ohm version of this which uses no more than four additional BC547C with 50V power rails, but I have so enjoyed listening to this that I have been slowed even more. Slowly but surely this project is advancing nicely.

Nominally this No1 circuit is for a 50W-8 ohm amplifier, but it sounds even louder on 4 to 8 ohm loudspeakers, and it is not damaged by low impedance loading which can cause smoothed psu current draws of 7A from both rails. The strange thing is that my simulator shows the circuit as being incapable of producing even 40W, so if anyone is inspired to simulate it on their own computer I would be most interested in reading their findings.

The sound is clean - I can't write more, because that's it.
If the source signal is sweet - - - If the source wavefronts slam - - -

I equate the output as being similar to 2x KT88 in ultralinear AB1, which normally sound louder than 50W solid state anyway, though unlike a valve design the bass is completely unphased at LF.

Like the original JLH class-A and my 25W circuit, this one has a low first cycle distortion characteristic because there are no parallel path capacitors pulling the open loop bandwidth down at audio frequencies, something that can arise in most other NFB designs.

This tested circuit is not yet thoroughly optimised, but it works well and is presented here as yet another variation amongst the many already available.

Cheers ........... Graham.
Attached Images
File Type: jpg class-aabb. no1.jpg (57.5 KB, 3451 views)
  Reply With Quote
Old 20th February 2005, 03:15 AM   #2
diyAudio Moderator Emeritus
lumanauw's Avatar
Join Date: Oct 2002
Location: Bandung
Send a message via Yahoo to lumanauw
It's complex amp BUT really interesting. Although seems complex, it has minimal stages. I see it only has 2 stages (the VAS and output stage(s) are 1 stage?). The output stage also new to me. Class A, class AB, classB all together?
I hope AKSA will see it, cause it only use R for LTP and bootstrapped VAS. Usually this configuration sounds good (with careful matching offcourse).

What is the efficiency figure of this? How many watts audio and how many watts idle?
  Reply With Quote
Old 20th February 2005, 03:44 AM   #3
diyAudio Member
Ultima Thule's Avatar
Join Date: Jan 2004
Location: Koskenkorva Land
Hi Graham,

what is the purpose with those output transistors working in class B and the cascode transistor on the diff pair?

"If transistors are blueberries and FETs are strawberries, then tubes must be.. pears"
Michael 29th January 2010
  Reply With Quote
Old 20th February 2005, 03:44 AM   #4
diyAudio Member
apassgear's Avatar
Join Date: Nov 2002
Location: Viña del Mar, Torreon
Default Choice???

Hi Graham,
I've been following your very interesting work advancement on JLH circuitry but clearly I'm not the one to make any comments regarding the circuit operation. I will soon make a JLH and most likely will be one of your iterations that seem an excellent choice.

In the past three years I've been fooling around with Pass clones and I like the Zen with full range speakers (which is what I like to use now) but feel that JLH would be a much better amp and easy to build also.

My preferences seem to be amps with highish output impedance to drive low Qts speakers but would like some opinions regarding this statement.

If I may ask, which are your preferred speakers to drive with these JLH amps?
  Reply With Quote
Old 20th February 2005, 08:08 AM   #5
Graham Maynard's Avatar
Join Date: Dec 2003
Location: UK
Hi Lumenauw,

I don't see this as being complicated and when you see how few parts there are in the actual construction it really is not so. The output resistors are on the heatsink with the 2SCs.
The mirrored differential has one BC557C in cascode to match the differential collector voltage dynamics (improves hf) also one as a simple follower to unload the first transistor of output voltage variable VAS C.bc thereby improving hf current balance within the differential pair. The fixed Cs limit closed loop gain at rf without affecting af.

Yes the 2SC3421 is an integral part of the output stage, and in this version it does not even get warm without a heatsink.

The quiescent (quietescent) heat here is 80W and the operation never drops out of class-A, but the audio output genuinely sounds like that of a conventional 80W solid state design. My heatsink is cooled by a Papst ultra-silent fan and it does not get hot, only gently warm.

The diode at the current phase splitting driver transistor is not active. It is there to prevent excessive current flow by maintaining conduction in the event of a negative power rail failure. If the upper rail goes there is no bias. My main circuit has an additional RC filter to the differential tail resistor, but the protype is working well without it !

Hi Ultima Thule,

Clearly I could have had all of the ouput transistors driven by the second class-A driver; ie. as constant class-A with class-AB boosting. This does work, BUT, all the ABs would then switch in and out of conduction at the same instant, and this can become audible due to back emf acting against NFB loop control.when the load is a dynamic loudspeaker.

However by using two differently biased class-A stages with 1 and 10 ohm emitter resistors, these are made to operate within different current ranges, thus by driving and correcting differently their efforts overlap such that one always smooths the other. Also there is no single dominant hf phase change due to Darlington drive that would otherwise degrade stability.
The separate class-B provides additional conduction for momentary power into loudspeaker impedance dips, but there are no crossover artifacts because the overall linearity is controlled by the class-A devices running within their own narrow current ranges and thus much less into conventional gain non-linearities. The class-AB and class-B currents also crossover at different cyclical time instants, so that all the class-As need do is smooth the overall characteristic. Without that class-B a NFB loop controlled correction of loudspeaker back emf induction would introduce a bigger 'wiggle'. All simulatable (but not via conventional THD/resistor testing), and as also borne out in listening tests.

Hi Apassgear,

My preference is for full range loudspeaker drivers too, but you need small ones to cover the higher frequencies, and then lots of them to go low.

In order to obviate loudspeaker back EMF induced 'wiggles' some designers go for non-feedback circuitry. I am not saying that this is wrong, but I prefer a loudspeaker sited amplifier with low output impedance drive so that a voice coil is made to follow signal waveform and limit a dynamic driver's natural overexcursions. This though is where amplifiers can come unstuck, because using NFB to correct driver modified output potential can interact with input due to amplifier propagation delay. The bipolar type of class-A circuit published by JLH is particularly good in this regard.

Cheers ............. Graham.
  Reply With Quote
Old 20th February 2005, 08:38 AM   #6
AKSA is offline AKSA  Australia
diyAudio Member
Join Date: Sep 2001
Location: Melbourne, Australia
Ah, Yes Lumenauw,

I see it!!


Aspen Amplifiers P/L (Australia)
  Reply With Quote
Old 20th February 2005, 06:16 PM   #7
diyAudio Member
Ultima Thule's Avatar
Join Date: Jan 2004
Location: Koskenkorva Land
Hi graham,

now I realize how your output works and have more questions.

I wonder if the dramatic current gain changes when some of the class AB or B output transistors "switch in" would change the current drive signal for the class A transistors, because as soon as the current demand at the output growes there will be an exponential current gain when those class AB and B comes into play, I feel the FB mechanism have to adopt to a nonlinear currentgain drive fo the class A transistors which acts both output as well driver transistors, yes/no?

Spreading out the the working point on several output transistors gives smaller changes at a time each one break in but I wonder if not the higher harmonics will show up more but the overall harmonics level will be lower because of each output transistor have a small contribution?
Well, this is not an easy question, just speculating here!

Regarding the casode transistor at the diff stage, I was puzzled what it should do beacause I couldn't guess if there where some "speciall issue" more than it would give a symmetrical/simmilar voltage change at the both diff-stage transistors collector.
But now when I read your answer to Lumanauw you said "(improves hf)", but I don't follow this as the voltage now at both collectors are actually counteracting against both the input voltage as well as the feedback input voltage, am I missing something here?

Regards Michael
"If transistors are blueberries and FETs are strawberries, then tubes must be.. pears"
Michael 29th January 2010
  Reply With Quote
Old 20th February 2005, 07:02 PM   #8
diyAudio Member
Join Date: May 2002
Location: Switzerland
To me this looks like a more refined (i.e. improved) version of the Edwin amplifier, isn't it ?

Regarding the "never switching off" amplifier there was once a circuit in the "circuit ideas" section of EW+WW that was called "low power class A" by Jeff MacAulay (spelling ?
). The published example had only very low power, best suited for headphones (and it was using OP-AMPs). This could very well be adapted for higher power using discretes. The main principle is that only one of the two NPN ouput transistors is within the voltage NFB loop. The other one is within a current control loop taking care that there is always a minimal common current (don't know a better expression, maybe someone else knows a better one, in German it would be called "Querstrom") flowing through both series connected transistors. IMO it would be best called single-ended class AB.
I once made simulations with a discrete version using old-fashioned MOSFET models as output devices. Simulated (!) THD was very low for low drive signals. THD is slowly rising with output swing, with 2nd order being dominant.


  Reply With Quote
Old 20th February 2005, 07:16 PM   #9
expert in tautology
diyAudio Member
bear's Avatar
Join Date: Apr 2002
Location: New York State USA

I have to admit I am unsure of how your output stage works... looks interesting...

I see that you are calling for 10 x 2SC3281 in the "AB" section per rail. I presume these are large high power output devices...

Is that the sum total of all the devices labeled A, B, AB? or just the AB section? or the B and AB sections, and then how is this distributed?

Then, I notice that the quiescent current is 1 amp per rail at 42 vdc...

Which part of the "output" stage is drawing what?
At only 100ma per "AB" device, that's 1 amp. Assuming 10...

So, where are the "switch over points" for each device?

In the actual unit, is the set zero pot sufficient to maintain DC offset stability and for output bias stability??

I guess the neat part of this design is that you don't need a critical voltage/current Vbe multiplier (or similar) to set up the bias point(s)? You can just set the values for a nominal "Class A" bias for the first "A" transistor and not worry too much about it's operating point since it runs out of steam anyhow, and the AB stage takes over?? The "B" device is a "B" device since it doesn't have enough bias... the second A device has the 10 ohm which makes the AB section have bias... Is that the idea?

Did you test the output impedance/DF?? does it vary with output level?

jpeg of the beast??

_-_-bear :Pawprint"
_-_-bear -- Btw, I don't actually know anything, FYI -- every once in a while I say something that makes sense... ]
  Reply With Quote
Old 20th February 2005, 09:23 PM   #10
Graham Maynard's Avatar
Join Date: Dec 2003
Location: UK
Hi Michael,

Yes there is an increase in current gain, and that increase is very much related to load current demand, but what this means is that there is less of a decrease in NFB loop controlled linearity when compared to a normal scenario.
Distortion increases less with increasing load current than is normally the case, it also tends to be second harmonic in nature.

I show a quiescent of 1A, with half of this passing through the first class-A driver, and 1/6th each between the second class-A driver and the AB outputs, but the amplifier still sounds good when idling at 500mA when the AB outputs are passing but a few milliamps, there is then less of a maximum drive.

The higher frequency harmonics do not appear to come up, and I put this down to the principle of operation for this kind of output stage - - - the drive to one half cannot limit or affect ongoing drive to the other half, as can happen with a common VAS driven complementary output stage.
The half that is conducting most strongly tends to be NFB loop error corrected by the half that is working less hard !!!

The differential stage works better (whether there is a mirror or not) when both collectors have very similar potentials - alternating as well as dc. This must have something to do with internal activity within the second differential transistor with its low impedance connections, and is good for a 3dB improvement in loudspeaker damping without incurring phase shift penalties, as well as isolating the mirror from second transistor C.bc, which improves hf performance. Maybe this slight improvement does not arise when the differential has emmitter resistors - I don't use this arrangement and don't know.

I tried this amplifier without any input capacitor, without the 22pF capacitor plus 22 ohm resistor, and without the 4,700uF NFB capacitor. It ran exactly the same and had a frequency response from DC to RF. I actually tried it out driving an AM radio and it did amplify beyond 1.6MHz without instability. However, without the 22pF and 22ohm resistor it was capable of amplifying house wiring interference, whilst what may have been transistor flicker noise kept the output potential from ever accurately settling to zero.

Hi Charles,

I am not aware of the Edwin amplifier by name. I have seen many circuits, but to me they are like the faces of people I meet - I'm not good with names.

There have been several 'never switching off' amplifier circuits that purport to maintain low level class-A operation, but most of them rely upon output stage current or voltage sensing. Few designers appear to consider what happens when dynamically induced loudspeaker back EMF shifts the resultant output stage current to become out of phase with the incoming audio waveform. If the Jeff Macaulay circuit you mention is the one that had a 1 ohm sensing resistor in one output half, then I would agree with your IMO; the output stage had not been tested with a dynamic load, and for that reason did not give it further thought. It has only been mentioned as a 'circuit' since then, not as an audio amplifier.

Both halves of my circuit remain actively waveform controlled (to the limits of their natural capabilities) no matter what angle the output current might be.

Hi Bear,

The total number of output devices is 10, they may be whatever is to hand. I had 10x 2SC3281.
The 'Set zero' will cover many input impedances and is quite stable.
Yes I think you've got the idea, but the class-A does not run out of steam - it remains in control.
The output impedance still changes with load. Guess I would need to double-check that to answer your question.
This is one of the earlier versions.

Cheers ............. Graham.
Attached Images
File Type: jpg gm.200-4. view..jpg (23.3 KB, 2199 views)
  Reply With Quote


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off

Similar Threads
Thread Thread Starter Forum Replies Last Post
CCS for PP Class A output stage? AudioGeek Tubes / Valves 5 24th January 2004 02:58 PM
Class A with AB Output-stage ManUtd Solid State 14 1st December 2003 05:49 PM
Class A output stage for CD Player PTSOUNDLAB Digital Source 0 30th December 2001 12:04 PM
Class h or g output stage? Opie Solid State 1 24th October 2001 09:30 PM

New To Site? Need Help?

All times are GMT. The time now is 12:19 PM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio

Content Relevant URLs by vBSEO 3.3.2