I amended my idea a few days ago in measuring the Re. For class A, I still think you can simply measuring the average and current.
For class AB where conduction is less than 360deg, you can use a second processor to detect breaks in the music and inform the floating processor to do calibration. You have breaks in between songs, do the calibration at that time.
I am just talk for the sake of talking, not that I believe in this whole thing. Amps have been working reliably already.
I really don't like controlling the bias spreader, I don't like to hang all the crap on this sensitive note.
My class A sx-Amp has an OPS standing current of 1.4 Amps. The peak current into the load with both output halves still conducting is a shade under 2x 1.4A or 2.8A. It's good for 28 watts class A into 8 Ohms. The 15 W rating is actually the 1.4A rating so quite underrated from that perspective - details and associated plots are in the write-up on my website.
Sure it varies, but does it affect performance..??? I would think the shaping of the cross over point (gradual turn on/off) is much more of a factor, than a few % varying Idle.
And only half that for single ended class A. SE is inefficient, but it does force some honesty onto the marketing dept. Linear output ends when class A runs out in SE.
--
With a class push-pull AB circuit, you could detect when the drop across the upper device's resistor equals the drop across the lower one's and trigger a sample-hold circuit to measure the level then (Jan's zero-cross equivalent). But that might never happen if you have a DC output offset and no input signal, so you'd have to work around that possibility.
Just use an analog computer
Yes, see: Auto Bias part II
Notice that this circuit provides a sliding bias class AB.
Cheers, E.
Yes, see: Auto Bias part II
Notice that this circuit provides a sliding bias class AB.
Cheers, E.
Well if you would accept 10% accuracy on say 50mA bias, 5mA is equivalent to 40mV offset with an 8 ohms load. I think that would be manageable.
BTW I also used a sample&hold to measure Ibias.
Jan
So the final question would be even with a workable sliding bias scheme how audible would this be, does it even make a real difference or is it just a nice exercise in control theory?
It might be a nice for manufacturing or for long-term stability. No manual bias adjustment to make.
Yeah that's why Peter Walker came up with current dumping, and see what that got him 😎
Jan
I am thinking of Marcel van de Gevel's "Audio power with a new loop"
Electronics World, February 1996, pp140-143,
which describes a power amp where
"quiescent current is stabilized via feedback, using a non-linear common mode loop".
Yes I know that article - the key word is 'non-linear'. IIRC it mimics the log V/I curve of a Vbe, no?
Jan
Basically, my circuit is based on Marcel's auto-bias circuit, though with a few twists.
But my point is that it is way easier to create a nonlinear common mode control loop with a few trannies instead of using a uP plus a precision ADC plus a lot of other stuff.
Cheers, E.
E Stuart why has no one built your circuits? I have admired what seems to be lots of hard and good work on your part, but it seems as if no one wants to take the task of layout and assembly, then test...Complex but DOable from my perspective...
Hi Jan,
The primary objective is to keep the product of currents of the "upper" respectively the "lower" output devices constant (well, more or less), same as "Class i" and the LT1166 chip. This way the OP devices are never completely turned off, a necessary condition to monitor and control the bias.
Cheers, E
I think monitor ONLY the current of the Re is much better way to do it because you totally eliminate the temperature term in the whole equation. Temperature dependent is not reliable, you have delay, problem of accurate sensing. That's another reason I don't want to touch the bias spreader for compensation.
Measuring the emitter resistor Re does not depend on temperature, you monitor and adjust the current, which is the ULTIMATE goal of this whole thing. Who cares about temperature?
There are different ways of controlling the current using Re, using a second uP to look for dead spot is one of the many ways. The point is to totally get out of the temperature dependency. It will work best to TOTALLY ELIMINATE the conventional bias spreader and use a SHUNT VOLTAGE REGULATOR as spreader. Then controlling the voltage drop across the base stop resistor to adjust the bias like what I posted a few days ago. Then you can put the shunt regulator away from the heat sink and all as it's only job is to provide a temperature independent voltage to get the bias into the ball park and let the closed loop servo to set the final current.
I am only talking, I have no intention to build any of this. But I am always interested in thinking, it's like a jigsaw puzzle for me.
Last edited: