A new idea for lower dissipation Class A... now all the usual arguments sliding bias etc are not Class A... that's been flogged to death and Class A to my way of thinking means the "old" definition, that the output devices remain in conduction for the full cycle.
So a different approach. What if you mostly listen at lowish volume, but want the full Class A performance at high levels for serious listening. In that case the volume control setting has determined the max power that can be output to the load. Put simply, apply say 2 volts RMS from a CD player (max possible output) and the max output of the whole system has been defined. If the bias could be set to match, then the amp would always be in Class A at that volume setting no matter what the music did.
Turn the volume up and the bias is low, Class AB, so some way of doing that automatically would have to be devised.
Two approaches came to mind... the electrically easy one, have another gang on the volume control to vary the bias as the volume is altered.
That's messy.
Another idea. We all love our remote volume controls (I couldn't live without one now anyway) so lets say you have a high quality motorised Alps pot and all.
Idea... the motor signal when it appears enables an oscillator (say 50 khz... way above audio anyway) that injects a low level sine of this frequency into the audio before the volume control. That signal is monitored at the output of the amp. Lets set the levels say that volume on full gives 3 or 4 volts output, so no tweeter damage etc.
That signal is picked off wih a filter and rectified and a peak value derived that then is used to set the bias on the amp.
Result, an amp whose bias is always correct for Class A but that doesn't generate excess power loss at low levels. Interfacing say a 0 to 10 volt DC voltage to a corresponding bias current would be a challenge... perhaps opto isolators could be used.
You could even apply the same idea to a manual non motorised pot and just have a push button to enable the osc and inject the HF signal manually on demand to set the value that way.
So there we are, just one of those ideas we all get from time to time.
Remember you read it here first
So a different approach. What if you mostly listen at lowish volume, but want the full Class A performance at high levels for serious listening. In that case the volume control setting has determined the max power that can be output to the load. Put simply, apply say 2 volts RMS from a CD player (max possible output) and the max output of the whole system has been defined. If the bias could be set to match, then the amp would always be in Class A at that volume setting no matter what the music did.
Turn the volume up and the bias is low, Class AB, so some way of doing that automatically would have to be devised.
Two approaches came to mind... the electrically easy one, have another gang on the volume control to vary the bias as the volume is altered.
That's messy.
Another idea. We all love our remote volume controls (I couldn't live without one now anyway) so lets say you have a high quality motorised Alps pot and all.
Idea... the motor signal when it appears enables an oscillator (say 50 khz... way above audio anyway) that injects a low level sine of this frequency into the audio before the volume control. That signal is monitored at the output of the amp. Lets set the levels say that volume on full gives 3 or 4 volts output, so no tweeter damage etc.
That signal is picked off wih a filter and rectified and a peak value derived that then is used to set the bias on the amp.
Result, an amp whose bias is always correct for Class A but that doesn't generate excess power loss at low levels. Interfacing say a 0 to 10 volt DC voltage to a corresponding bias current would be a challenge... perhaps opto isolators could be used.
You could even apply the same idea to a manual non motorised pot and just have a push button to enable the osc and inject the HF signal manually on demand to set the value that way.
So there we are, just one of those ideas we all get from time to time.
Remember you read it here first
Hi,
as you change the bias current, the Vbe of the devices will change. If output offset is to be avoided then the change in Vbe of the +ve half must EXACTLY match the change in Vbe of the -ve half. If this cannot be guaranteed then the output offset will vary with bias current.
A DC servo could attenuate this variable offset to manageable levels.
as you change the bias current, the Vbe of the devices will change. If output offset is to be avoided then the change in Vbe of the +ve half must EXACTLY match the change in Vbe of the -ve half. If this cannot be guaranteed then the output offset will vary with bias current.
A DC servo could attenuate this variable offset to manageable levels.
Have you considered a way around, that mostly the distortion rises, as the bias goes down?
I did actually have a bit of the same idea, a few years ago. I used a basic Pass SE design as base, and made a few switchable bias settings. It was very short lived, to say the least.
All the low settings, which actually are the ones that I wanted to have, sounded downright bad, compared to the higher bias settings.
As I recall, I had some 5 settings, ranging from 200mA (maybe 100mA, can't remember that for sure) to 3A bias. I seem to recall, things went better around 1A though.
Magura
I did actually have a bit of the same idea, a few years ago. I used a basic Pass SE design as base, and made a few switchable bias settings. It was very short lived, to say the least.
All the low settings, which actually are the ones that I wanted to have, sounded downright bad, compared to the higher bias settings.
As I recall, I had some 5 settings, ranging from 200mA (maybe 100mA, can't remember that for sure) to 3A bias. I seem to recall, things went better around 1A though.
Magura
As I recall, I had some 5 settings, ranging from 200mA (maybe 100mA, can't remember that for sure) to 3A bias. I seem to recall, things went better around 1A though.
Magura
That kind of makes sense if the amp used HEXFETS.
It would be intresting to try with something like a JLH topology design.
similar idea I had some years ago. But I prefer to change the supply voltage instead of the quiescent current/idle current (bias) for the output stage. But then one needs separate power supplies for the 0V offset adjustment potentiometer (and for the voltage gain stages and driver stages, if such are present).
In a "Parvel Macura SE buffer I realized this idea. I chose 3 different DC voltages for the output stage: 8V, 16V and 32V (+ /-4V, + /-8V and + /-16V) and for all voltages 3.5 A quiescent current through the single ended CFP output. The repositioned code plug changed voltage (auto transformer primary windings) and also the resistors for the each matching offset voltage. Depending on how loud music is heard, the user must appropriate repositioned code plug.
If you like remote control, you must develop a control unit, that operates with relais, but I cannot say, how it is to avoid overshooting effects.
I would be prefer a switch over by triacs between the auto trafo windings while operating. But therefore I have too low experience, because there are inductive load switching and the risk of damaged triac.
In general, it is always the royal way to work with bi- or multi amping. Pure Class-A is to use only at higher frequencies, e. g. as tweeter-amp (if I use an active crossover, I need only around 4 watts at arround 20W power dissipation). The other speakers can be drived with good class AB amplifiers.
In a "Parvel Macura SE buffer I realized this idea. I chose 3 different DC voltages for the output stage: 8V, 16V and 32V (+ /-4V, + /-8V and + /-16V) and for all voltages 3.5 A quiescent current through the single ended CFP output. The repositioned code plug changed voltage (auto transformer primary windings) and also the resistors for the each matching offset voltage. Depending on how loud music is heard, the user must appropriate repositioned code plug.
If you like remote control, you must develop a control unit, that operates with relais, but I cannot say, how it is to avoid overshooting effects.
I would be prefer a switch over by triacs between the auto trafo windings while operating. But therefore I have too low experience, because there are inductive load switching and the risk of damaged triac.
In general, it is always the royal way to work with bi- or multi amping. Pure Class-A is to use only at higher frequencies, e. g. as tweeter-amp (if I use an active crossover, I need only around 4 watts at arround 20W power dissipation). The other speakers can be drived with good class AB amplifiers.
I am talking about pure Class A in the accepted sense of the definition.
The bias current does not vary with the music as in sliding bias etc, it is determined by the max possible output for a given volume setting.
A class A amplifier rated at 100mw or 100watts correctly meets the definition... the bias is fixed.
Magura,
Nice link
Your comments earlier on the subjective quality deterioration you found when turning the bias current down... that's a whole new problem. That a given amplifier design sounds "different" at different bias currents while still meeting all the requirements of Class A for the expected output level. Perhaps that applies differently to different topologies/designs.
Maybe there is an ideal "sweet spot" of bias for many designs, above and below this and performance not so good.
That would require some serious double blind listening tests I think
Nice link
Your comments earlier on the subjective quality deterioration you found when turning the bias current down... that's a whole new problem. That a given amplifier design sounds "different" at different bias currents while still meeting all the requirements of Class A for the expected output level. Perhaps that applies differently to different topologies/designs.
Maybe there is an ideal "sweet spot" of bias for many designs, above and below this and performance not so good.
That would require some serious double blind listening tests I think
Hi Steve
You know what I mean more than most.
An output pair in pure class B (npn,pnp... bases tied together) plays music... look no bias
You know what I mean more than most.An output pair in pure class B (npn,pnp... bases tied together) plays music... look no bias
Then you will want to drop
"my way of thinking means the "old" definition, that the output devices remain in conduction for the full cycle"
I think I got the drift what you wanted to implement. I think you wanted to make 100W pure class A amp but you won't likely be needing 100W most of the time. When you are using it as headphone amp, dissipating 200W as heat for 100mW output is such a waste, right? But by definition, Class A amps requires output devices to be in full conduction region. Thus, high power dissipation. Let's say you wanted to put a switch to limit biasing current of output devices, how can you be sure that all output devices are in Class A conduction? Your choice is to use smaller power amps for such applications.
But putting the output devices in Class A (full conduction region) even for low power amps (such as 25W) will not make an amp low dissipation. It will still dissipates quite a lot as heat compared to what it'll output as an audio power.
Of coz, higher powered amps with many parallel output devices put in low biased setting may not work as Class A. You'll need minimum constant biasing current to put transistors in full conduction.
What kind of output devices you are trying to use? (Power BJT or Power MOSFETs)?
Hi megasat16
Easy one first... it's all just a theoretical idea, no design exists or is in the process of becoming a reality. Just an idea to discuss.
I am meaning one amp, that performs as well at 1 watt with the corresponding bias for that output, as it would at 100 watts (your example) and the bias set for that output level. The only thing that alters is the (no signal) quiescent current in the output stage, and that alters in proportion to the volume controls physical position, not the signal. Knowing the max signal that could ever be applied (2 volts rms say from a CD player) puts absolute known limits on it all.
And 1 watt Class A or 10 watt Class A is very very much cooler than 100 watt Class A.
Easy one first... it's all just a theoretical idea, no design exists or is in the process of becoming a reality. Just an idea to discuss.
I am meaning one amp, that performs as well at 1 watt with the corresponding bias for that output, as it would at 100 watts (your example) and the bias set for that output level. The only thing that alters is the (no signal) quiescent current in the output stage, and that alters in proportion to the volume controls physical position, not the signal. Knowing the max signal that could ever be applied (2 volts rms say from a CD player) puts absolute known limits on it all.
And 1 watt Class A or 10 watt Class A is very very much cooler than 100 watt Class A.
Nelson and Steve, I can see I'm not going to win on this one am I ? lol
You both know what I mean.
I think you mean "the sum of the bias currents in the out pair(s) does not vary". It does go up and down for each of the pair as music or other signal is output into a load. At least that is my understanding.