New idea for low dissipation Class A amplifier.

I had an idea on this. You could use a single volume pot and ac couple it to the pre and power stages then run dc through it and conect the wiper through a low pass filter to comparators that switch on different bais circutry. when they reach their triger voltage. I was thinking of a single ended common source mosfet amp where you could switch on different bias resistors with mosfets controled by the comparators. You could get three currents with two resistors by switching on one the other or both.
 
just as an example of what could be achieved in the way of power (energy) saving, lets start with a 100W 2.6A bias ClassA amplifier (similar to a Krell KSA100)

volume control setting -0dB to -14dB PSU +-52V Bias 2.6A, dissipation 260W
volume control setting -15dB to -24dB PSU +-30V Bias 2.6A, dissipation 156W
volume control setting -25dB to -34dB PSU +-30V Bias 1A, dissipation 60W
volume control setting -35dB to -44dB PSU +-20V Bias 1A, dissipation 40W
volume control setting -45dB to -80dB PSU +-20V Bias 0.4A, dissipation 16W

This requires 3 voltage settings and 3 bias settings brought in staggered as volume control is advanced/retarded.

The lowest setting of 0.4A bias +-20Vdc allows an 8ohm 95dB/W @ 1m pair of speakers to produce ~85dB maximum (or about 65dB average) at a 2.5m listening distance from ~0.3W of pure ClassA power and staying in class A for ~ three times the nominal output current demand of an 8r0 load. If an unexpectedly high transient came through then the push/pull nature of the output stage would still allow 14W of maximum output power producing a maximum SPL at the listening position of ~101dB and this at just 16W of dissipation/channel.

The energy (heat) saving potential is enormous.
Why has this never been promoted by the ClassA manufacturers?
 
Last edited:
AX tech editor
Joined 2002
Paid Member
just as an example of what could be achieved in the way of power (energy) saving, lets start with a 100W 2.6A bias ClassA amplifier (similar to a Krell KSA100)

volume control setting -0dB to -14dB PSU +-52V Bias 2.6A, dissipation 260W
volume control setting -15dB to -24dB PSU +-30V Bias 2.6A, dissipation 156W
volume control setting -25dB to -34dB PSU +-30V Bias 1A, dissipation 60W
volume control setting -35dB to -44dB PSU +-20V Bias 1A, dissipation 40W
volume control setting -45dB to -80dB PSU +-20V Bias 0.4A, dissipation 16W

This requires 3 voltage settings and 3 bias settings brought in staggered as volume control is advanced/retarded.

The lowest setting of 0.4A bias +-20Vdc allows an 8ohm 95dB/W @ 1m pair of speakers to produce ~85dB maximum (or about 65dB average) at a 2.5m listening distance from ~0.3W of pure ClassA power and staying in class A for ~ three times the nominal output current demand of an 8r0 load. If an unexpectedly high transient came through then the push/pull nature of the output stage would still allow 14W of maximum output power producing a maximum SPL at the listening position of ~101dB and this at just 16W of dissipation/channel.

The energy (heat) saving potential is enormous.
Why has this never been promoted by the ClassA manufacturers?

Hmmm. Nice idea. The variable power supply would probably not offer a huge power savings in practise (unless it is SMPS), but the bias current would. You could couple a separate pot to the volume pot to set the bias.
It does require an integrated amp rather than a preamp-power amp set of course. Also, one would need to take a hard look at the dynamic range in each setting to find the best/optimum bias point in each case. But interesting it is.

jd
 
Also, one would need to take a hard look at the dynamic range in each setting to find the best/optimum bias point in each case.
that was my starting point for each setting of the volume control. I asked myself "what signal voltage and what worst case peak transient current would be required at each setting?"
Then calculated the bias and PSU voltage requirements accordingly. I based all these on achieving 110dB @ 2.5m using a pair of 95dB/W @ 1m speakers with the volume control @ -10dB.
 
Administrator
Joined 2007
Paid Member
Andrew,
That's all nicely presented, and it really does put the idea into perspective too...

The use of different PSU rails is another "addition" to my basic idea... certainly for a manufacturer using a SMPS it should present few challenges, might be difficult to implement at a DIY level. A linear regulated supply wouldn't be energy efficient, and however any of it were implemented (bias, supplies) it would have to operate seamlessly with no clicks or noises etc.
 
Hi,
window comparators (monitoring the third track of the pot) controlling relays would do it.
The variable bias just needs extra parallel resistors just like the first Krell Klone PCB had on board.
The variable PSU voltage just needs three sets of tappings on the secondaries.
ClassH? has something similar, it's only the output stage that needs the variable, high current, supply.
Finally a DC servo to keep the output offset near zero and of course DC detect and speaker isolation if something goes wrong.

Just noticed that a 0-15,25,35Vac dual secondary ~800VA would do this example admirably.
 
Last edited:
to implement an idea like that ( which actually sounds very nice )will be an easy thing to do even in diy level ...

reading the thread, but also hate too many digital things i thought about a simple vu meter led made 3915 style conected in the out for example with some capacitor in the input to create delay

so you ask ammount of volume A so you get 3 leds on and since a capacitor exists in the input leds will not play up and down but stay lited untill you adjust volume higer or lower again ....

so then take this information use it how ever you like through opotocouplers and/or other to adjust any bias you like ...

but on the other hand why use 3915 it can be done also with juts one transitor or two as long as you mange to create this kind of delay so the info you have is stable .... something like an AGC with delay ....sounds fairly simply
 
Administrator
Joined 2007
Paid Member
Did my silly noobie suggestion stop the flow of this conversation?

Not at all :) Just can't see how you would achieve it transparently and how would you switch the outputs in and out of circuit ? You say "add a switch"... can you eleborate on that, relays ? solid state etc ? :)

To vary the bias as I proposed at the start means it can all be implemented at a "low signal level" with no power switching.
 
Not at all :) Just can't see how you would achieve it transparently and how would you switch the outputs in and out of circuit ? You say "add a switch"... can you eleborate on that, relays ? solid state etc ? :)

To vary the bias as I proposed at the start means it can all be implemented at a "low signal level" with no power switching.

Look at it from the other side - Having the Power MOSFETs or BJTs idle at Class A even if you are not using is not going to be very efficient. Thus, it's impossible to call it Low Dissipation even if switch the signal to low level. It's not the input signal, it's about the power Class A devices consume without any signal (sitting idle).

If you truly want efficient low power Class A, I think it's possible for one to tap into the last stage of pure Class A Pre-driver stage and switch off all the idleing Power output devices. I don't know even if it's possible with POWER MOSFETs but I think some BJT amps will allow you to that without much of any problem.
 
What about switching on/off out put pairs such as 2 devices always on and then a switch to add additional pairs on demand? The bias can stay fixed for all devices.
This is an interesting idea. I don't see why it shouldn't work. I'm building an Aleph 4 and if I understand the circuit at all, I should be able to cut the maximum output power by simply switching the gates of some of the FETs over to their sources (obviously switching off equally many from the bias current source as from the amplifying FETs). I think I'll try that when I get to that point in construction.
 
Administrator
Joined 2007
Paid Member
I can see you all have your own ideas :)

Switching the gates of a fet is possible as long as the "switch" if it is solid state can interface to and withstand the voltages it might see on "each side" of the switch.

Doing the same with bjt's may be more of a problem as the base current is significant and the abrubt loading of switching them in and out will cause loading on the vbe multiplyer causing it too change it's characteristics.

My original idea was for it to operate seamlessly on an amp, whether that amp used a single or multiple pair output configuration.
 
Hi. I have just read through this thread. Not my usual topic but, hey, I was bored !

My immediate thought, on reading post 28, was this sounds a little like QUAD current dumping but class A.

A small linear class A amplifier in use most of the time with a class A "power booster" (ugh!) for peaks in demand.

The bias could be set to a low level for normal listening with sliding automatic bias adjustment ( VU-style, fast attack, slower decay ) for higher listening levels.

DC servo control for offset and "quad-style feed-forward error correction"


Andy
 
Administrator
Joined 2007
Paid Member
Look at it from the other side - Having the Power MOSFETs or BJTs idle at Class A even if you are not using is not going to be very efficient. Thus, it's impossible to call it Low Dissipation even if switch the signal to low level. It's not the input signal, it's about the power Class A devices consume without any signal (sitting idle).

If you truly want efficient low power Class A, I think it's possible for one to tap into the last stage of pure Class A Pre-driver stage and switch off all the idleing Power output devices. I don't know even if it's possible with POWER MOSFETs but I think some BJT amps will allow you to that without much of any problem.

Your statement shows you misunderstand the original idea.

I have never mentioned reducing signal level. The idea is to have the outputs conducting enough that any input signal for a given volume setting could never take it outside Class A.

For example with the volume at say 30% and knowing that a CD player can never put more than 2 volts RMS into the system we have known conditions to work too. We know absolutely how much power a given system can put into the load with the volume at any given point for a given signal :)

When you talk of "tapping" into the Class A driver stages... if I understand you correctly... then I forsee these problems and many others.
1. How would you switch a speaker into the output of the driver stage.
2. If it's DC coupled the driver "output" is not at zero volts.
3. The linearity of such a stage would be ruined by such a low impedance load.

Maybe I misunderstand what you mean.
 
Your statement shows you misunderstand the original idea.

I have never mentioned reducing signal level. The idea is to have the outputs conducting enough that any input signal for a given volume setting could never take it outside Class A.

For example with the volume at say 30% and knowing that a CD player can never put more than 2 volts RMS into the system we have known conditions to work too. We know absolutely how much power a given system can put into the load with the volume at any given point for a given signal :)

I think I got your original idea in the first time. Without starting from scratch and some implementation idea of what you want to accomplish, it seems nearly impossible for Class A with low dissipation. When you said you want to put a switch to limit biasing current / voltage for the output devices, they still needs minimum current / voltage requirement to be in full conduction region based on the output device used.

That's why I said these two ideas are like parallel lines that will not cross path anywhere. At least, we need more info as the basic design of such amp.



When you talk of "tapping" into the Class A driver stages... if I understand you correctly... then I forsee these problems and many others.
1. How would you switch a speaker into the output of the driver stage.
2. If it's DC coupled the driver "output" is not at zero volts.
3. The linearity of such a stage would be ruined by such a low impedance load.

Maybe I misunderstand what you mean.

I may probably misunderstand a bit. I thought the reason you wanted to make your Class A amp more efficient is for using in low power requirements to drive headphone or efficient speakers. Tapping into Class A driver stage will not work for the demanding speakers.

Tapping into pre-driver stage needs to be done at the design level and give more consideration as Safe Operating Region of your Class A pre driver stage transistors and used parallel devices in the pre driver stage.

For DC coupled amplifiers, what difference would it make for DC coupling at the output devices and in the last stage of the Class A Driver Stage?

If you are using Class A Output conduction even at 1/3 of the Bias, it'll still dissipate a lot of heat for an amp with many parallel output devices. I don't know if it's acceptable as to call "Low Dissipation in Pure Class A" compared to some other Class or other Class A designs.
 
If you are using Class A Output conduction even at 1/3 of the Bias, it'll still dissipate a lot of heat for an amp with many parallel output devices. I don't know if it's acceptable as to call "Low Dissipation in Pure Class A" compared to some other Class or other Class A designs.
Hi,
a ClassAB Roender has a bias of ~555mA. Re=0r1 and Vre~18.5mV for optimum ClassAB bias.
A ClassA Krell KSA50 has a bias of 1.9A.
Bias the Krell to 1/3rd and you have Ib=633mA. That's only 14% more than the Roender.
I would call that a very significant saving in quiescent power and reduced temperatures.
633mA allows for 6W of ClassA and that is just 9dB down from maximum power.
If one knows that the volume control is set to more than 9dB below maximum output power then the underbiased ClassA stage will not leave ClassA when driving 8r0.
 
Hi,
a ClassAB Roender has a bias of ~555mA. Re=0r1 and Vre~18.5mV for optimum ClassAB bias.
A ClassA Krell KSA50 has a bias of 1.9A.
Bias the Krell to 1/3rd and you have Ib=633mA. That's only 14% more than the Roender.
I would call that a very significant saving in quiescent power and reduced temperatures.
633mA allows for 6W of ClassA and that is just 9dB down from maximum power.
If one knows that the volume control is set to more than 9dB below maximum output power then the underbiased ClassA stage will not leave ClassA when driving 8r0.

I agrees in these two particular amps. Do you have a link I can read about Roender Class AB amp? I tried looking for it but not sure which one we are talking about.