New idea for low dissipation Class A amplifier.

Stuart and others.
What is the output of the amplifier when fed with a 2Vac signal from a digital source and the volume control is at maximum.

Now turn down the volume control to -20dB.
The maximum input signal has dropped to 200mVac. From this one can determine the maximum output voltage and the maximum current to drive the attached speaker load. Design the stepping relays and comparator that is controlling them to at least meet this lower maximum output conditon.

Now turn down the volume control to -40dB.
Design the steps to suit a maximum input signal of 20mVac.

No sliding bias, no automatic adjust, no distortion detection. no change in heat output (and thus dynamic change in output conditions). It's a fixed bias scheme that is set in steps to suit the volume control setting.
 
Of course there is a distortion detector, a very bad one: the user, when he hears distortion he turns the amp "up", generating more heat and the output conditions change dynamically ;-) After a while he realizes he is getting too hot and turns it down again, or perhaps he doesn't and wastes power.

Seriously I guess it all depends on what you are striving for. If simplicity is a goal, the manual solution may be the best answer. If saving power is it then I think an automatic scheme will work better. Assuming the same small selection of operating points are chosen there are no sonic qualitative differences between the two, since all other operating parameters for the chosen range of bias points is identical.

I've done this with the rail voltage on KSA50s by using a variac. For the widest possible range of voltages you need the front end constant current source, but nothing fancier: I watched using a scope, the output offset voltage didn't noticeably twitch.

A non-sequitur: The few channels of KSA50 I've built, adjusted and listened to suggest that from a sonic perspective, getting the drivers to remain in a heavily biased state may be as important as the output transistors class A state. Whatever scheme is chosen to change the output bias and rail levels I'd work hard at keeping the drivers in their sweet spot.

As I said; it's a great idea, the exact means by which control of the levels is achieved seems almost irrelevant compared to the initial design work needed to select among a small number of known good bias and voltage pairs.

There are some different options that suggest themselves: for instance at lower voltages is it better to have a single pair of optimally biased outputs rather than 4 pair of barely biased outputs? What little experimentation I've done tells me a single pair of mj15003/4 transistors sounds better than two etc...

Stuart
 
hi.I have 600 VA torodidal transformer and it has 4x18 volts output.I will connect in series and have 2x36volts ac.I am planning to build a low bias Class AB krell ksa100 or ksa50 version amp with this transformer..How many output transistor should i use? I have MJL1302/3281 and NJW132/3281 transistors.and which of them are more suitable?thank you
 
Last edited:
Stuart,

I think ur missing the idea a little bit.

The original idea is to adjust the "bias point" as a function of the position of the volume control - which in turn is set incrementally per dB down from "0" (the max level available from a known source). The volume control having a dual function now, volume and bias.

So, IF you play your system with the "volume knob" set "halfway" the full bias is never applied to the amp, but sufficient bias is supplied to keep the amp in Class A for the calaculated maximum input signal level. Beyond that max level, presumably the amp merely goes into AB.

Think of the volume knob as the "heat knob"??

It is not the same as the Krell sliding plateau bias at all. Afaik no sliding bias is fast enough to catch the leading edge of a sudden onset of a high level signal. Although I suppose one could possibly design an ultra fast attack, slow decay circuit the problem then is that the output stage is being modulated in two directions at the same time, and while I don't know for sure I certainly suspect that this makes for all sorts of non-linearities to deal with.

Myself, I would prefer other solutions than a massive Class A amp that you have to turn down for "normal listening". But the idea seems to have some merit for a custom system that would actually benefit from being run with a massive class A amp, and where the power and heat removal is a significant penalty to be paid...

umut1001 why not check out the several Krell clone design/build threads?
Fwiw the krell KSA series appears to be basically the same as early Bryston amps but with more output devices... dunno if that is good or not, you decide.

_-_-bear
 
Bear,

Regarding my missing the idea: Perhaps you are right; I read the original idea as varying the output stage voltage and current through carefully controlled levels, thus leading to quadratic power savings with little or no sonic degradation. Varying the current alone is not to be sneered at from the power saving perspective, but is a minor elaboration of the hi-lo bias components on the original Klone50 board.

IIUC Krell's assumption is that the audibility of distortion of a single (partial?) cycle of a waveform is minimal (perhaps debatable), and that the long hysteresis of the circuitry makes the aggregate 'leading edge' distortion contribution minimal. There are discrete levels, rather than a continuous range for the bias points, so the overall output stage modulation is again only through carefully chosen operating points. Power savings notwithstanding, the original reviews of the amps did not imply the changes were all better sonically.

Stuart
 
any news ?
my favorite approach for low dissipation class A are only one class A power amplifier (3-15W) for the high frequency area as part of a multi-amping system (bi-amping for 2-way). When an active crossover network in this class-A power amp is implemented, the maximal output power could be still more less.
Check out
https://www.diyaudio.com/community/...-op-amp-for-ultimate-sounding-phl1230.154703/
A disadvantage is, that for single speaker systems (only one high power full range driver without sub woofer and tweeter) this isn't an appropriate idea.
 
"It is not the same as the Krell sliding plateau bias at all. Afaik no sliding bias is fast enough to catch the leading edge of a sudden onset of a high level signal. Although I suppose one could possibly design an ultra fast attack, slow decay circuit the problem then is that the output stage is being modulated in two directions at the same time, and while I don't know for sure I certainly suspect that this makes for all sorts of non-linearities to deal with".

--bear

I think this hits the core of the problem. You can't underestimate the concerns of implementing designs with high speed bias settings. Most basic and fool proof would be to use an extra wiper on a potentiometer. That's 4 on a stereo amp, that's old tech.

I love the idea's and solutions you all bring to the table. But from a home user perspective, the average maximum power you would need for a low to normal listening volume, is about 4 watts max. When you want to get to higher volumes, it goes up substantially. So much that the difference between a 50 watt and a 150 watts amp is not even that big. One doesn't make 150 watts class A amplifier. 50 watts class A is already a lot, at least I consider that not an standard class A design or common goal.
But, for the sake of argument, let's set that power as a max setting.
For a minimum, there's alrrady about 10 to 20 watts losses for the linear transformers, diodes etc.
So going well below that in minimal class A setting also isn't very sensible.

Is it achievable to set the minimum at around 25 watts dissipation, let's see, from the top of my head:
50 watts 8 ohm is around 28 volts ac, let's say 4 amps idle= 244 watts.
Minimum idling for class A 4 watts at same PSU voltage is 1A, power dissipation=56 watts.

In theory, switching voltage to 6 volts×1.4=8.4, make it 10 volts is possible, to do 20 watts.

So, going from usual listening levels (2 to 4 watts max) to loud, you most likely need all the juice you can, which means there's hardly a class A level in between that 4 watts and the 50 max that you really need switching, or sliding through.
If this holds true, correct me if I'm wrong, then all you need is a Class A amp that has a switch to set in lower class A (current and/or voltage) mode and since you probably know when you're gonna play loud, that could even be a physical switch on the front.
I admit, way less sexy than all these idea's, but it would also be applicable to all current class A design without breaking a sweat and still gain 220 watts for 95% of the listening levels.
YMMV of course.
Btw, this is by no means me trying to pull you away from making the most advanced switching/slide class A mechanisms in the world, it is just here to portrait the fact that the idea is great and even the simplest of solutions really can be simple and very effective.

P.S. number are from the top of my head, can be off!