That s correct...
With a push pull , this work for both halves of the signal.
Forcibly, the single ended will thus be less efficient that
a push pull , whatever the output level..
With a single ended, there will be an extra current when
the active side is conducting, since this side will provide
the current for the load but also for the generally attached
current source..
When the current source supply the load, the active side
will absorb less current, balancing the extra losses that occur
during the other half of the signal...
Last edited:
So what are the drawbacks of Class AB?
Does the fallback mode mechanism (the switch from A to B) affect the sound quality whilst it is in Class A bias ?
Do we loose efficiency ? Or do we add distortion ?
What sort of distortion ? Phase shift? non-linearities ?
It must add to the circuit complexity.
Although I don't understand 100% of the bias discussion, its interesting and I like to know what I'm getting into, cause and effect etc.
At some point I'll have to bite the bullet. At the moment JLH 96 is in the lead, with F3 and F5 very close.
JLH 96 as I have the output bjt already, suitable transformer (460VA 4 x 23v) and the dumpster dive heatsinks.
The PCBs don't seem too difficult to get hold of, although I might try it PTP. I have some nice ceramic standoffs with solder tags.
The F3 uses not so easy to obtain jfets LU1014, IRFP240
The F5 IRF240 and IRD9240.
They don't seem to be available from the usual suspects like CPC, RS or Farnell.
There are plenty from eb, but I am wary of component sellers from HK/CN ...
But I'm still undecided, Class AB, with high Class A bias, say 20W is attractive too.
All these designs have higher distorsion than a competently
designed AB class amp....Isn t it wonderfull ?..
looking at a ClassA push pull amplifier.
If the output bias is set to 1A through Re = 0r1, then Vre=100mV.
This is effectively an overbiased ClassAB.
However if you never ask for more than 2Apk from the amplifier then it remains in ClassA for all outputs. As such it can never have crossover distortion and so the distortion will be less than ClassB and less than optimally biased ClassAB and less than overbiased ClassAB and less then underbiased ClassAB.
Now to what the supply rails are doing when the amp is quiescent.
They supply 1A to the +ve supply that flows through the upper half of the output pair.
The -ve supply rail sinks 1A from the lower half of the output pair.
At the junctione between the Re of the upper and lower outputs, the current coming in at the top exactly matches the current going out at the bottom.
The output current to the load is the difference between the +ve and -ve rail currents.
There are a couple of conditions. The NFB takes a tiny bit of current back to the -IN input and the Zobel takes a tiny bit of noise current to ground. For normal operation these currents are so small they are ignored for the sake of simplicity, but they do not go to zero.
Measure the output offset voltage. In perfect conditions it will be 0.0mVdc.
Connect an 8r0 load from output to ground. The output current will be 0.0mA because the voltage is zero mV. But, there is some noise voltage (uVac). again we choose to ignore this for simplicity.
Put a signal into the amp that delivers 1Vac to the 8r0 load.
The current in the load will be sinusoidal and will vary from -1.4Apk through zero Apeak to +1.4Apk.
To deliver than varying current to the load the supply rail currents will not match all the time.
They should match at output zero crossing.
But at all other output currents the difference in supply rail currents equals the output current.
Lets apply +200mVdc to the input of the DC coupled ClassA amp biased to 1A.
The gain is 40. The output voltage will be +8Vdc The output current will be 1Adc to the load and returning via audio ground.
The +ve supply rail at this instant will pass 1.5A to the output stage.
The -ve supply rail at this instant will pass 1.5A from the output stage. The difference is 1A and it all goes out to the 8r0 load.
Now apply +400mVdc to the input. The +ve supply current goes to 2A, the -ve supply current goes to 0A and the difference of 2A goes through the 8r0 load to audio ground. We are at the ClassA maximum current limit. The -ve supply rail is passing zero current. It cannot supply any less.
The +ve supply rail is sending the whole 2A to the Load. The load will have 16Vdc across it. The gain is still 40.
Now change the input voltage to -400mVdc. The +ve supply passes zero current, the -ve supply rail sinks 2A and all that 2A comes through the load from audio ground.
The AC signal waveform does much the same. At any instant the amplifier converts the input signal to current into the load such that the voltage across the load is [input * gain]. The supply rails pass a modulated current that alternately takes them from 1A to 2A back to 1A and then to zero A before returning for a second time to 1A.
That is not constant current.
The output to ground takes the current from the load and that ground current is the difference between the supply rail currents.
The 1A of output bias allows a maximum ClassA output of 2Apk.
The maximum power available from the ClassA amplifier is 2^2 * Load /2 = 16W. The supply rail voltage to allow all this output to happen into an 8r0 load will be approximately +-20Vdc to +-22Vdc. The heat dissipated in the output stage will be ~42V * 1A ~ 42W. That will be hot. This average heat output will not change when zero output current flows. So at 0W output the Pq=40W This is seen as wasteful. Some use this argument to tempt the buyer to settle for a lower heat wasting amplifier. There are dozens that can give you 16W into 8r0 that waste much less than 40W continuously. It's choices.
If the output bias is set to 1A through Re = 0r1, then Vre=100mV.
This is effectively an overbiased ClassAB.
However if you never ask for more than 2Apk from the amplifier then it remains in ClassA for all outputs. As such it can never have crossover distortion and so the distortion will be less than ClassB and less than optimally biased ClassAB and less than overbiased ClassAB and less then underbiased ClassAB.
Now to what the supply rails are doing when the amp is quiescent.
They supply 1A to the +ve supply that flows through the upper half of the output pair.
The -ve supply rail sinks 1A from the lower half of the output pair.
At the junctione between the Re of the upper and lower outputs, the current coming in at the top exactly matches the current going out at the bottom.
The output current to the load is the difference between the +ve and -ve rail currents.
There are a couple of conditions. The NFB takes a tiny bit of current back to the -IN input and the Zobel takes a tiny bit of noise current to ground. For normal operation these currents are so small they are ignored for the sake of simplicity, but they do not go to zero.
Measure the output offset voltage. In perfect conditions it will be 0.0mVdc.
Connect an 8r0 load from output to ground. The output current will be 0.0mA because the voltage is zero mV. But, there is some noise voltage (uVac). again we choose to ignore this for simplicity.
Put a signal into the amp that delivers 1Vac to the 8r0 load.
The current in the load will be sinusoidal and will vary from -1.4Apk through zero Apeak to +1.4Apk.
To deliver than varying current to the load the supply rail currents will not match all the time.
They should match at output zero crossing.
But at all other output currents the difference in supply rail currents equals the output current.
Lets apply +200mVdc to the input of the DC coupled ClassA amp biased to 1A.
The gain is 40. The output voltage will be +8Vdc The output current will be 1Adc to the load and returning via audio ground.
The +ve supply rail at this instant will pass 1.5A to the output stage.
The -ve supply rail at this instant will pass 1.5A from the output stage. The difference is 1A and it all goes out to the 8r0 load.
Now apply +400mVdc to the input. The +ve supply current goes to 2A, the -ve supply current goes to 0A and the difference of 2A goes through the 8r0 load to audio ground. We are at the ClassA maximum current limit. The -ve supply rail is passing zero current. It cannot supply any less.
The +ve supply rail is sending the whole 2A to the Load. The load will have 16Vdc across it. The gain is still 40.
Now change the input voltage to -400mVdc. The +ve supply passes zero current, the -ve supply rail sinks 2A and all that 2A comes through the load from audio ground.
The AC signal waveform does much the same. At any instant the amplifier converts the input signal to current into the load such that the voltage across the load is [input * gain]. The supply rails pass a modulated current that alternately takes them from 1A to 2A back to 1A and then to zero A before returning for a second time to 1A.
That is not constant current.
The output to ground takes the current from the load and that ground current is the difference between the supply rail currents.
The 1A of output bias allows a maximum ClassA output of 2Apk.
The maximum power available from the ClassA amplifier is 2^2 * Load /2 = 16W. The supply rail voltage to allow all this output to happen into an 8r0 load will be approximately +-20Vdc to +-22Vdc. The heat dissipated in the output stage will be ~42V * 1A ~ 42W. That will be hot. This average heat output will not change when zero output current flows. So at 0W output the Pq=40W This is seen as wasteful. Some use this argument to tempt the buyer to settle for a lower heat wasting amplifier. There are dozens that can give you 16W into 8r0 that waste much less than 40W continuously. It's choices.
Last edited:
That bit I understand, the rest I'll have to sit down some where quiet with a pencil and some paper.
Environmental aspect
I don't see Class A amps as being wasteful for 75% of the time. I live in a country with a temperate climate. If I had to expend energy (money) running air conditioning to be able to be comfortable, then I would consider the heat dissipated as wasted. All that happens is that I burn less natural gas heating my house up.
I have the same view about CFL lights, which I will try to avoid, as most seem to be too cold and not very comfortable on the eyes.
If heat was an issue, I would concentrate on Class D and it's varieties.
As far as distortion is concerned, I like listening to music, if it makes it sound better, then I'm all for it. A little extra bit of even harmonics ? Why not ? I look upon it as some salt and pepper, it helps brings out the flavour. If zero distortion was the name of the game we'd all be listening to and enjoying lab instrument amplifers.
Theres a modified JLH by Nelson Pass, the PLH, he suggests making the current balance between the PP output stage adjustable. It also changes the feedback somewhat. In effect making it a SE and PP amplifier on demand and anything in between. That is something I'd like to incorporate in my (PP) amp.
Another idea would be to make the bias variable. So that you could turn the bias down to get a cooler running amplifier and more power. Not sure how you would reconcile the different B+ requirements though.
I don't see Class A amps as being wasteful for 75% of the time. I live in a country with a temperate climate. If I had to expend energy (money) running air conditioning to be able to be comfortable, then I would consider the heat dissipated as wasted. All that happens is that I burn less natural gas heating my house up.
I have the same view about CFL lights, which I will try to avoid, as most seem to be too cold and not very comfortable on the eyes.
If heat was an issue, I would concentrate on Class D and it's varieties.
As far as distortion is concerned, I like listening to music, if it makes it sound better, then I'm all for it. A little extra bit of even harmonics ? Why not ? I look upon it as some salt and pepper, it helps brings out the flavour. If zero distortion was the name of the game we'd all be listening to and enjoying lab instrument amplifers.
Theres a modified JLH by Nelson Pass, the PLH, he suggests making the current balance between the PP output stage adjustable. It also changes the feedback somewhat. In effect making it a SE and PP amplifier on demand and anything in between. That is something I'd like to incorporate in my (PP) amp.
Another idea would be to make the bias variable. So that you could turn the bias down to get a cooler running amplifier and more power. Not sure how you would reconcile the different B+ requirements though.
or switchable bias.
The first Krell Klone PCB has a switchable bias. It could be set up as either optimally biased ClassAB or as ClassA. effectively summer/autumn+winter+spring settings.
D.Self also did a Tri-Modal version of the Blameless with a switchable summer/winter
I was having a dig at the new technologies that claim perfection in sound and claim zero power wastage and oddly, thousands of watts of output through a 100mm diam speaker with a 20mm Voice coil and supplied from a 1A mains fuse.
Andrew - you have just described my perfect amplifier and speaker system.
Where can I get one ?
The fallback mechanism doesn't affect anything and is less of a mechanism and more of a mode of operation. You only incur this penalty if you exceed the amount of power the amplifier can deliver as class A.
The fallback mechanism also doesn't require a more complex circuit nor does it introduce any further penalties.
Put simply here's an amplifier, you have a potentiometer to adjust the bias.
Before turning the amplifier on for the very first time I make sure the bias is turned down to the minimum value.
After it has been turned on, I slowly increase the bias point until I reach the optimum bias for a class AB amplifier. This is where you should leave things if you're not interested in class A.
If I keep increasing the bias, the amplifier will now start working in class A. With a small increase in bias, it will be working at say 1 watt into 8 ohms. If I keep increasing it, I will end up with say 10 watts into 8 ohms and so on. Obviously I cannot exceed the safe operating area of the output transistors otherwise it will explode.
Class A amplifiers usually have a fixed bias point that cannot be adjusted.
Douglas Self's trimodal amplifier uses a voltage reference to define the class A biasing point and a standard Vbe multiplier + potentiometer to set up the optimum bias for non class A operation.
I highly recommend buying a good quality PCB if you can find them. This reduces user error by a significant factor and makes fault finding a hell of a lot easier if something goes wrong.
You can buy high quality PCBs for Selfs trimodal amplifier from here.
You can buy from Mouser btw, they have a much larger stock then Farnell. I've purchased a couple of times from them and had good success. If you order £50 worth of goods you get free P&P. Yes you still have to pay V.A.T when it comes into the country and you have to pay import tax too, but it isn't so expensive. I've tried both the shipping method where they pay (and charge you) for the VAT and import duty and where you pay for them on delivery. The first method worked out to be a bit cheaper I think.
If I was going to build a class A design, this is the route that I would want to take. Same number of class A watts as a standard class A amp, but has the ability to deliver significantly more oomph if it's required. Also if you find your constantly jumping into the overbiased class AB region, you can always lower the bias setting to make it optimally biased.
5th element thanks for that, much clearer now.
Can any AB amplifier be over biased so that the A mode is significant? Assuming the output stage can handle the extra current and heat ?
I'll consider PTP for NP F3/F5, but D Self Trimodal is too complicated for that.
The JHL will should be ok too, the driver and output stage will be fine on the standoffs and the dc servo could be made debug style, though it might get a bit ugly
Didn't know about Mouser, I'll check them out. I've been trawling the threads on F3/F5 construction, a place called deepsurplus.com sell tested LU1014, so that part will be taken care of.
Can any AB amplifier be over biased so that the A mode is significant? Assuming the output stage can handle the extra current and heat ?
I'll consider PTP for NP F3/F5, but D Self Trimodal is too complicated for that.
The JHL will should be ok too, the driver and output stage will be fine on the standoffs and the dc servo could be made debug style, though it might get a bit ugly
Didn't know about Mouser, I'll check them out. I've been trawling the threads on F3/F5 construction, a place called deepsurplus.com sell tested LU1014, so that part will be taken care of.
Yes, but might not be any advantage, on the contrary
The few amps I have built have ALL had a certain optimal bias point
maybe adjusting a few other components could have changed that, I dont know
either its built for classA or its not
To me, very high biased AB have mostly been a sales trick
but as said, it may be design related
but I doubt it have been
anyway, its not just a matter of turning a bias pot
maybe it relates to "the sweet spot" by Nelson Pass
it may be your ticket to heavenly sound, if you can find that magical sweet spot
but there are ways to get very high ClassA currents from an optimally biased ClassAB amp.
Some are in production.
Fit the lowest Re that keeps the output device thermally stable (Cordell's formula).
Fit many parallel pairs.
Set bias by Vre~18 to 25mVre
eg.
10pair To220 stage Re=0r08 Vre=18mVre.
Single pair bias = 225mA
10pr bias = 2.25A
Max ClassA current = 4.5A
That is enough to send 80W of ClassA to an 8r0 load.
For 80W, supply rails of +-40Vdc would be about right since they will be loaded to 180W in quiescent operation. 9W/To220.
This is a pure ClassA amplifier delivering 80W into 8ohms and yet is biased to optimal ClassAB. This means that if extra current is drawn then the transition beyond 4.5A is at the lowest distortion available. I would expect this amp to go to 150W into 4r0 and easily capable of driving a 4ohm speaker.
Some are in production.
Fit the lowest Re that keeps the output device thermally stable (Cordell's formula).
Fit many parallel pairs.
Set bias by Vre~18 to 25mVre
eg.
10pair To220 stage Re=0r08 Vre=18mVre.
Single pair bias = 225mA
10pr bias = 2.25A
Max ClassA current = 4.5A
That is enough to send 80W of ClassA to an 8r0 load.
For 80W, supply rails of +-40Vdc would be about right since they will be loaded to 180W in quiescent operation. 9W/To220.
This is a pure ClassA amplifier delivering 80W into 8ohms and yet is biased to optimal ClassAB. This means that if extra current is drawn then the transition beyond 4.5A is at the lowest distortion available. I would expect this amp to go to 150W into 4r0 and easily capable of driving a 4ohm speaker.
This is very interesting I must say. This would be an EF output stage I take it? You're giving me ideas Andrew
I don't think it is limited to only EF, but I guess EF is the safe option.This is very interesting I must say. This would be an EF output stage I take it? You're giving me ideas Andrew
Quick update
I've ordered Bob Cordell's book and Doug Self's Audio Amp Design Handbook.
Haven't found what CFP stands for ... yet.
I've also acquired another lump of aluminium ...
The nice thing about this one is the hardware for mounting 10 x TO3 devices are already there !
I have a bunch of MJ15003/4 from onsemi that'll fit in that postion nicely. The emitter resistors are 0r33, so thats good too.
May have to change the fan to something a low noise one, its a 230v AC mains one at the moment.
I've ordered Bob Cordell's book and Doug Self's Audio Amp Design Handbook.
Haven't found what CFP stands for ... yet.
I've also acquired another lump of aluminium ...
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
The nice thing about this one is the hardware for mounting 10 x TO3 devices are already there !
I have a bunch of MJ15003/4 from onsemi that'll fit in that postion nicely. The emitter resistors are 0r33, so thats good too.
May have to change the fan to something a low noise one, its a 230v AC mains one at the moment.
Not a lot to understand. Those heatsinks with the fan "choked" (7v or PWM controller) would be the cat's meow with mongrel GX's and SNG001's PB120 powerboard. 55-0-55DC (38-0-38Vac) trafo , the most common HT receiver trafo and some caps , you could do the poor man's DIY for $100 or less.
My genesis heatsinks are considerably larger and do 100 degree (39C) days with ease ,so 2 pairs would be optimal for BJT AB and you would still have a nice 120 watts to play with.
My big one's (200w GX's), are better than the blameless or any other amp I've thrown together from this forum.
PS - I could even do a 4 X TO-3 power board
OS
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.