You have to sum the upper and lower currents, not just the upper as done here. The sum of the two currents should be a constant for a class-A amp. This is a simple class-A amp where the sum of the two currents is constant:
That's only true for the dc state, whereas the ac is a modulation upon it. Then the sum is not constant anymore, or there is no current driven into the load.
The autobias scheme I am describing here is not affected by the signal delivered to the load, as long as the outputs stay fully in the class A mode. The average sensed current is THE SAME as the idle current. This is in the context of a push-pull class A design.
I have verified this by turning the input signal off and examining the idle current. It just returns to its designed set point once the signal has zero amplitude.
In addition to the requirement of always running in class A, the current sensing circuitry also cannot clip due to the max/min current excursion. This also would cause the idle current control loop's set point to shift. I dumped some of my designs due to this issue.
The latter issue affects the JLH symmetrical headphone amplifier, but it has a reputation of producing very good audio. Perhaps it simply has a "tasty" balance of harmonics?
Another issue the JLH design has, and one I don't like much, is that its autobias scheme produces a pretty substantial spike in idle current at turn-on, because the capacitors in the IC feedback loop have to charge up before the autobias circuitry begins to work -- and in the meantime, the output transistors are pretty much ON full-bore. I know this because I built one and observed the current spike, and simulations confirmed the cause. I mentioned that kind of problem earlier in this thread.
Thanks to everyone for a very lively discussion of the subject!
I have verified this by turning the input signal off and examining the idle current. It just returns to its designed set point once the signal has zero amplitude.
In addition to the requirement of always running in class A, the current sensing circuitry also cannot clip due to the max/min current excursion. This also would cause the idle current control loop's set point to shift. I dumped some of my designs due to this issue.
The latter issue affects the JLH symmetrical headphone amplifier, but it has a reputation of producing very good audio. Perhaps it simply has a "tasty" balance of harmonics?
Another issue the JLH design has, and one I don't like much, is that its autobias scheme produces a pretty substantial spike in idle current at turn-on, because the capacitors in the IC feedback loop have to charge up before the autobias circuitry begins to work -- and in the meantime, the output transistors are pretty much ON full-bore. I know this because I built one and observed the current spike, and simulations confirmed the cause. I mentioned that kind of problem earlier in this thread.
Thanks to everyone for a very lively discussion of the subject!
Here's a simulation showing the current flowing in R11, one of the .25 ohm output resistors. Output data for plotting started at 100mS so the autobias circuit had pretty much settled down to its stable operating point. In this simulation the IC control loop's LPF capacitor is in the feedback leg of the opamp, which also helps to reduce the likelihood that the idle-current control signal will clip.
So may be way off base here but just wondering if you tried to use a PID controller with all of its built in variability within its control schemes? Would be able to ip.iment changes to the control scheme with simple parameterization of the controller. For example if you were to Google the PIDE function that gives you complete control and variation to adjust. Out of the industrial controls playback. I would bet there are some modules that could be had for a descent price.
I understand where you're coming from but I believe a full-blown PID controller is overkill. There also are some issues that will complicate its application in this case. We have a desired DC component -- the idle current -- hiding inside an AC signal, the current delivered to the load. There are digital-filtering approaches to extract that DC component, but on a cost-effectiveness basis it's pretty hard to beat a simple RC lowpass filter. Using it we get the "P" and "I" portions of a PID so......is there really a need to buy a PID?
After the (approximately) DC signal is extracted, it might be possible to employ a PID to really nail things down but, gee, what I've got going now seems to work pretty well and just needs a few filter/gain components to achieve pretty good results. As a DIYer I really like simplicity, as long as it delivers the performance I want. I am aware that the opamps and current sensor(s) have a lot going on under the hood: but as long as they do what I want, they're just another component as far as I'm concerned.
After the (approximately) DC signal is extracted, it might be possible to employ a PID to really nail things down but, gee, what I've got going now seems to work pretty well and just needs a few filter/gain components to achieve pretty good results. As a DIYer I really like simplicity, as long as it delivers the performance I want. I am aware that the opamps and current sensor(s) have a lot going on under the hood: but as long as they do what I want, they're just another component as far as I'm concerned.
Here's a version that has been altered to make it suitable for a headphone amplifier (but it does drive an 8 ohm load OK). The idle current is about 1 amp.
The optoisolators are wired a bit differently, more like the way I'd want in a real-world design. The input resistors to the LEDs are different, to account for small differences between the + and - halves of the amplifier.
Harmonic distortion results this good should be taken with a very large grain of salt. I used a maximum stepsize of 1us to generate the FFT.
It doesn't look too bad for an amp that uses one opamp and 4 transistors in the signal path. It doesn't hurt that the opamp is well-suited for this kind of application.
The optoisolators are wired a bit differently, more like the way I'd want in a real-world design. The input resistors to the LEDs are different, to account for small differences between the + and - halves of the amplifier.
Harmonic distortion results this good should be taken with a very large grain of salt. I used a maximum stepsize of 1us to generate the FFT.
It doesn't look too bad for an amp that uses one opamp and 4 transistors in the signal path. It doesn't hurt that the opamp is well-suited for this kind of application.
The performance @10KHz is similar in terms of THD.
One thing to note: it appears that most opto-isolator output transistors have a maximum VCE specification around 30V. So the above design will be (just barely ) OK, but higher-power amplifiers will either need a couple of cascode transistors in there, or use a different scheme, placing the opto-iso transistors across the 600 ohm resistors. The former approach will introduce a bit more of a limit on the peak output voltage (w/o a bootstrap scheme, anyway). The latter will accommodate just about any amplifier topology that uses a Vbe multiplier.
One thing to note: it appears that most opto-isolator output transistors have a maximum VCE specification around 30V. So the above design will be (just barely ) OK, but higher-power amplifiers will either need a couple of cascode transistors in there, or use a different scheme, placing the opto-iso transistors across the 600 ohm resistors. The former approach will introduce a bit more of a limit on the peak output voltage (w/o a bootstrap scheme, anyway). The latter will accommodate just about any amplifier topology that uses a Vbe multiplier.