Hi,
I know of at least 1 other forum member who posted some thing similar, but I guess various versions of your circuit have posted here. I'm sorry that it isn't that ground-breaking new but that doesn't mean it doesn't work.
I do see some practical issues.. like both sides of your circuits have to be capable of high current? So this is almost like building 2 amplifiers. But you could make it a bridge circuit (LM4780 anyone ;-)); (maybe even use cross-couple of distortion across both sides (X-topology anyone ;-)) . And the R13/R14 are gonna have to be adjustable for precision scaling of the signal, especially if the output stage has no feedback (=relative high output impedance) , it needs to be adjusted to speaker impednace?
I am afraid that practical thingies will prohibit this topology from really being usefull.
Best regards,
Thijs
I know of at least 1 other forum member who posted some thing similar, but I guess various versions of your circuit have posted here. I'm sorry that it isn't that ground-breaking new but that doesn't mean it doesn't work.
I do see some practical issues.. like both sides of your circuits have to be capable of high current? So this is almost like building 2 amplifiers. But you could make it a bridge circuit (LM4780 anyone ;-)); (maybe even use cross-couple of distortion across both sides (X-topology anyone ;-)) . And the R13/R14 are gonna have to be adjustable for precision scaling of the signal, especially if the output stage has no feedback (=relative high output impedance) , it needs to be adjusted to speaker impednace?
I am afraid that practical thingies will prohibit this topology from really being usefull.
Best regards,
Thijs
tschrama said:
Do you have a link ? I am very interested.
Yes, its like two amps, but the right side can run from low voltage rails.
I got my problem with cross-coupling because I see no way to figure out how much coupling it has to be.
A pot between R13/R14 will do the job with scaling, but I leave it out for the sim. Theese are details...
Again: The output stage has to be inside the feedback loop for real world and normal condition sim.
Bernhard said:
One of the distortion mechanisms in FET input opamps is that input capacitance varies with voltage. If the + and - inputs are the same impedance this tends to cancel. Okay, now with your circuit, if there is some type of distortion at +X3 there will also be similar distortion at -X5. This will appear to both opamps like the signal. So to the extent that the distortion at +X3 and -X5 are the same, the distortion will not cancel. (If it cancels, then signal cancels too.) Transistor input opamps will have their own distortion mechanism.
JF
Same curve like before, but with output stage inside feedback loop. Still while clipping to see the effect.
Black curve is output of right side output stage.
Black and blue add each other to green.
Black curve is output of right side output stage.
Black and blue add each other to green.
hehehe...
I've abandoned my cross-coupling idea for a single-ended output. I think if it worked everyone would be doing it now. Nelson Pass's SuperSymmetry is probably as close as it gets.
JF
I've abandoned my cross-coupling idea for a single-ended output. I think if it worked everyone would be doing it now. Nelson Pass's SuperSymmetry is probably as close as it gets.
JF
Hi,
Sorry, but I don't have a link.. I wish I had.. we're gonna have to search the forum for that.. or actually I hope the orgininal auteur will read this thread and reply...
Regards,
Thijs
Sorry, but I don't have a link.. I wish I had.. we're gonna have to search the forum for that.. or actually I hope the orgininal auteur will read this thread and reply...
Regards,
Thijs
johnferrier said:
Huhu, to where did you cross-couple if it is single-ended ?
By the way, do you mean "not bridged/symmetric" ?
Aleph-X is single ended.
What we need is a circuit that doesn't distort in the first place, so that it doesn't need to be cancelled. People have patents attempting this. That is why I thought that if you post your schematic, it likely would not cancel distortion (otherwise you would seek a patent first).
JF
JF
Bernhard said:
I was going to do something like the cross-coupled circuit on the AD8610 datasheet (differential outputs). I'm not going to do that now. I'm going with a standard single-ended design. It was kind of sad for me to come to the realization that the cross-coupled idea would not work as I'd hoped. I guess I decided that it would create as much distortion as it cancelled.
As far as I know, the SuperSymmetry idea is kind of a cross-coupled design. I not very familiar with that design though.
*EDIT*
Thanks for sharing your ideas. Post more information when you have it.
JF
I think the clipping test is good to see if it works.
If it cancels a big clipping problem, it should correct the small amount of distortion caused by a SE classA amp without effort. 🙂
Without clipping, a sinewave of 5V on the left side gives a sinewave of O.5mV on the right side.
That means the distortion of the left side alone is -80dB.
Not much work for the right side.
Distortion @ 0.5mV should be very small...
If it cancels a big clipping problem, it should correct the small amount of distortion caused by a SE classA amp without effort. 🙂
Without clipping, a sinewave of 5V on the left side gives a sinewave of O.5mV on the right side.
That means the distortion of the left side alone is -80dB.
Not much work for the right side.
Distortion @ 0.5mV should be very small...
Yehh, but aren't your resistor of the voltage dividing network are gonna have to be ultra-precise to detect the right amout of distortion?
tschrama said:
As mentioned before. that can be done with a pot. Also I have pairs of 600ohm resistors matched to better than 0.05%.
0.01% is also affordable in small quantities.
Bernhard said:
FWIW, 0.01% bulk metal foil resistors run about 8$US. Not bad, if you only need a few. They do add up fast, if you require more than a few...
JF
the worker and the cleaner
New improved circuit with less parts count. 🙂
X4 is a differential amp.
Performance in the sim is perfect now,
red input curve is absolutely hidden under the green output curve even on clipping of left side output stage 😎
Left stage is "worker" and right stage is "cleaner".
New improved circuit with less parts count. 🙂
X4 is a differential amp.
Performance in the sim is perfect now,
red input curve is absolutely hidden under the green output curve even on clipping of left side output stage 😎
Left stage is "worker" and right stage is "cleaner".
X3 could be made a differential amp too, providing balanced input.
Note:
Before this topologie, I still had the imagination of subtracting the input signal from the output signal to isolate distortion.
But I was focused to feeding that error signal back to the input somehow.
That is may be possible too, but can not cancel clipping.
Note:
Before this topologie, I still had the imagination of subtracting the input signal from the output signal to isolate distortion.
But I was focused to feeding that error signal back to the input somehow.
That is may be possible too, but can not cancel clipping.
Hi Bernhard,
Your amplifier proposal seems to be just an example of an error feedforward topology. You measure the difference between the output signal and the input signal (with scaling if required, depending on gain). That difference is added to the output with the right polarity.
There are more examples of this kind of principle. In most cases the designer tries to avoid the use of floating loudspeakers, i.e. one side of the loudspeaker remains connected to ground. Because of that, most of these error feedforard amplifiers look different from yours. Examples are the Quad current dumping amplifier (high accuracy class A amplifier adds the error part to the dumb class C power dumper), or the Sandman class S amplifier.
Please have a look at:
http://digilander.libero.it/paeng/feedforward_concepts.htm
http://www.quadesl.org/Family_Album/Interviews/Quad405WirelessW1975.doc
Although error feedforward is a good idea in itself, it is generally used to correct the errors of a cheap and dirty power amp by adding a small high quality correction amp. Your design uses an expensive single ended class A power amp and your correction amp is of the same class. Bit of waste?
Steven
Your amplifier proposal seems to be just an example of an error feedforward topology. You measure the difference between the output signal and the input signal (with scaling if required, depending on gain). That difference is added to the output with the right polarity.
There are more examples of this kind of principle. In most cases the designer tries to avoid the use of floating loudspeakers, i.e. one side of the loudspeaker remains connected to ground. Because of that, most of these error feedforard amplifiers look different from yours. Examples are the Quad current dumping amplifier (high accuracy class A amplifier adds the error part to the dumb class C power dumper), or the Sandman class S amplifier.
Please have a look at:
http://digilander.libero.it/paeng/feedforward_concepts.htm
http://www.quadesl.org/Family_Album/Interviews/Quad405WirelessW1975.doc
Although error feedforward is a good idea in itself, it is generally used to correct the errors of a cheap and dirty power amp by adding a small high quality correction amp. Your design uses an expensive single ended class A power amp and your correction amp is of the same class. Bit of waste?
Steven
- Status
- Not open for further replies.