Hi Steven,
thanks for the links, so it looks like this is feed forward.
In the first link I think there is an error in fig. 3, the correcting amp must be inverting.
The current dumper of the second link does not seem promising for me.
My goal is to build an near distortion free amp, something like susy, that's why there are two classA output stages.
99% of forum members seem to be looking for the perfect distortion free amp, why not try this ? 🙄
Bernhard
thanks for the links, so it looks like this is feed forward.
In the first link I think there is an error in fig. 3, the correcting amp must be inverting.
The current dumper of the second link does not seem promising for me.
My goal is to build an near distortion free amp, something like susy, that's why there are two classA output stages.
99% of forum members seem to be looking for the perfect distortion free amp, why not try this ? 🙄
Bernhard
You're right; G2 in that link should be inverting or the + and - of the summing node should be reversed. This is also clear since the next figure 4 shows the same polarities but the summing node is now on the same side of the loudspeaker/load.
I wonder whether the distortion improvement outweights the additional complexity/cost of a complete second amplifier for error correction.
Maybe error feedback is more worthwile?
Steven
I wonder whether the distortion improvement outweights the additional complexity/cost of a complete second amplifier for error correction.
Maybe error feedback is more worthwile?
Steven
Bernhard said:
That logic is very flawed!
Anyway I like what you are doing. Do what you like, explore crazy idea put them up for scrutiny. It may never give you the ultimate amp but you 'll be learning electronics in the process. 😉
"Distortion Free"
There are distorting elements throughout the circuit. The input distorts, the output distorts, the summing junctions distort. Feedback corrects what it can distingush between signal and error. The circuit will correct what appears like an error. Some of the error will appear like it is part of the signal.
Having complex circuitry to correct distortion just adds more distorting elements that something has to correct. But the correcting part of the circuit is not perfect. It also distorts. You can acheive low distortion, but not zero distortion.
What part of the circuit does not distort? The resistors? You can not make an amplifier with only resistors. The rest of the circuit does not exactly know what it is doing. The rest of the circuit can not fix itself. And the resistors can not fix the errors.
JF
There are distorting elements throughout the circuit. The input distorts, the output distorts, the summing junctions distort. Feedback corrects what it can distingush between signal and error. The circuit will correct what appears like an error. Some of the error will appear like it is part of the signal.
Having complex circuitry to correct distortion just adds more distorting elements that something has to correct. But the correcting part of the circuit is not perfect. It also distorts. You can acheive low distortion, but not zero distortion.
What part of the circuit does not distort? The resistors? You can not make an amplifier with only resistors. The rest of the circuit does not exactly know what it is doing. The rest of the circuit can not fix itself. And the resistors can not fix the errors.
JF
John,
I never said that this amp could ever be free from distortion.
But hopefully very low.
Bernhard
I never said that this amp could ever be free from distortion.
But hopefully very low.
Bernhard
efficiency of this circuit in cancelling distortion
Some calculations:
My calculations say that the distortion of this amp will be zero
Ok, ok... 🙄
I assume that the % distortion of an output stage will not rise or fall with output voltage.
The new circuit has a right side output of 0,25mVpp for 10Vpp left side sine output.
The correction signal is 92dB below the output signal.
So the error part of the output signal of the output stage is -92dB. In the sim world of course.
The distortion part is 0,000025 x output signal.
If the output signal of the right side amp has to be 0,25mV, the distortion part newly introduced will be 0,000025 x 0,25mV = 6,25x10 -9 V
If the error correction is 100% and it should be like that, except of course for the new introduced distortion, the total distortion of the amp will be -184dB.
I do not know if the calculation is right, I am very weak in math, but if both output stages are of the same type, the total distortion -dB of the whole amp will be double.
So because the ideal matched parts in the sim world do not exist in the real world, lets say the real world distortion of one output stage is -60dB, the resulting real world amp distortion will be -120dB.
Conclusion: I believe that the circuit doubles the signal to distortion ratio in dB
P.S. The op amp used is NE5532...
Some calculations:
My calculations say that the distortion of this amp will be zero
Ok, ok... 🙄
I assume that the % distortion of an output stage will not rise or fall with output voltage.
The new circuit has a right side output of 0,25mVpp for 10Vpp left side sine output.
The correction signal is 92dB below the output signal.
So the error part of the output signal of the output stage is -92dB. In the sim world of course.
The distortion part is 0,000025 x output signal.
If the output signal of the right side amp has to be 0,25mV, the distortion part newly introduced will be 0,000025 x 0,25mV = 6,25x10 -9 V
If the error correction is 100% and it should be like that, except of course for the new introduced distortion, the total distortion of the amp will be -184dB.
I do not know if the calculation is right, I am very weak in math, but if both output stages are of the same type, the total distortion -dB of the whole amp will be double.
So because the ideal matched parts in the sim world do not exist in the real world, lets say the real world distortion of one output stage is -60dB, the resulting real world amp distortion will be -120dB.
Conclusion: I believe that the circuit doubles the signal to distortion ratio in dB
P.S. The op amp used is NE5532...
Bernhard,
At least one flaw in your reasoning:
Although the voltage output of your correction amp will be low, the current output (or capability) will be as high as the current output of the main amplifier. So, I think it is a mistake to assume that the distortion of the correction amplifier wil be very low. Let's assume for the moment that the main amplifier is perfect and has no distortion at all. The correction amplifier should then have zero voltage output (since there is nothing to correct) but how capable will that correction amplifier be to keep its output dead quiet while sinking and sourcing the complete load current?
A second source of error is all the noise the correction amplifier will add to the output. OK, this will be small, as it is from the main amplifier, and because the noise of both amplifiers is uncorrelated it will only add partially. But still it will keep you from signal to noise or distortion ratios anywhere near -184dB. But this is just academic. Distortion will be far higher than noise in a power amplifier.
Steven
At least one flaw in your reasoning:
Although the voltage output of your correction amp will be low, the current output (or capability) will be as high as the current output of the main amplifier. So, I think it is a mistake to assume that the distortion of the correction amplifier wil be very low. Let's assume for the moment that the main amplifier is perfect and has no distortion at all. The correction amplifier should then have zero voltage output (since there is nothing to correct) but how capable will that correction amplifier be to keep its output dead quiet while sinking and sourcing the complete load current?
A second source of error is all the noise the correction amplifier will add to the output. OK, this will be small, as it is from the main amplifier, and because the noise of both amplifiers is uncorrelated it will only add partially. But still it will keep you from signal to noise or distortion ratios anywhere near -184dB. But this is just academic. Distortion will be far higher than noise in a power amplifier.
Steven
Hi Bernhard,
It seems to me that you are simulating with Micro-Cap.
Why don’t you make distortion measurements
with thd(harm(v(R8))) or the like?
Ciao
Federico
It seems to me that you are simulating with Micro-Cap.
Why don’t you make distortion measurements
with thd(harm(v(R8))) or the like?
Ciao
Federico
Steven's point on current capability of the error amp. is
important and interesting. I think that is also somewhat
related to another issue I think has not been brought up
so far (sorry if I have just missed it). We have two separate
amplifiers, each with their own control (ie. feedback) loop. I am
no expert on control theory, but I do know that separate
contol loops can cause nasty phenomena unless they are
somehow synchronized or one is a master loop of the other.
True, we have the same situation for bridged amps, for instance,
but at least we usually have two identical systems with identical
control loops there. I am not saying this won't work, just that it
might be quite difficult to analyze the stability of such a combined
system as this one. What happens if you throw in some
parasitic capacitance in various places? How do the poles of
the two amplifiers behave together?
important and interesting. I think that is also somewhat
related to another issue I think has not been brought up
so far (sorry if I have just missed it). We have two separate
amplifiers, each with their own control (ie. feedback) loop. I am
no expert on control theory, but I do know that separate
contol loops can cause nasty phenomena unless they are
somehow synchronized or one is a master loop of the other.
True, we have the same situation for bridged amps, for instance,
but at least we usually have two identical systems with identical
control loops there. I am not saying this won't work, just that it
might be quite difficult to analyze the stability of such a combined
system as this one. What happens if you throw in some
parasitic capacitance in various places? How do the poles of
the two amplifiers behave together?
Steven, Christer,
the cleaner really carries lots of current but little voltage, that could rise distortion, the question is if the distortion of the cleaner is above or below the distortion of the worker.
If it is below, there should be a benefit for distortion of the whole amp.
The cleaner needs the same idle current like the worker.
Good question if the whole thing is stable...
How could I synchronise feedback loops, may be X them ?
There was often problems with my recent X circuits, the sim just stopped for many values of resistors like the magic resistor.
This design shows absolutely no problems like this.
Anyway I have to build and see/listen.
This schematic not complete... The worker will have 7V rails and will have added an A/B amp with 18V rails which will drive it's floating ground.
Greetings, Bernhard
the cleaner really carries lots of current but little voltage, that could rise distortion, the question is if the distortion of the cleaner is above or below the distortion of the worker.
If it is below, there should be a benefit for distortion of the whole amp.
The cleaner needs the same idle current like the worker.
Good question if the whole thing is stable...
How could I synchronise feedback loops, may be X them ?
There was often problems with my recent X circuits, the sim just stopped for many values of resistors like the magic resistor.
This design shows absolutely no problems like this.
Anyway I have to build and see/listen.
This schematic not complete... The worker will have 7V rails and will have added an A/B amp with 18V rails which will drive it's floating ground.
Greetings, Bernhard
Hi Federico,
what does that tell ? Why is it so high in the positive halfwave ?
It is without clipping. With clipping, max. is 82m
A SE classA X circuit shows max. 60m and my SE classA X circuit with floating ground driven by A/B amp shows max. 120m.
All during positive halfwave.
Greetings, Bernhard
what does that tell ? Why is it so high in the positive halfwave ?
It is without clipping. With clipping, max. is 82m
An externally hosted image should be here but it was not working when we last tested it.
A SE classA X circuit shows max. 60m and my SE classA X circuit with floating ground driven by A/B amp shows max. 120m.
All during positive halfwave.
Greetings, Bernhard
O.K.
Try follow these instructions, I hope to make no mistakes
1) select 1K Hz in the signal generator freq.
2) go to transient analysis
3) run option =normal,State var=Zero, check operating point
4) in the limits form put time range= 5m
5) maximum time step = 0.001/1024 (a good choice for accurate
results)
6) in “X-expression” type “F” ( means freq.), in Y- expr. type
thd(harm(v(R8)))
7)in X- range type 10k( or more), in Y range type 50,1e-5
8) click on the button for logarithmic Y plot ( second from the left)
9) click on the “transient “ menu (top of the screen), go to DSP
parameters
7) set upper time limit= 5m, lower time limit = 4m, number of
point= 1024; status= ON
(in this way the analysis is from 4m to 5m cutting out transient)
then run the simulation
the result are as a percentage value. If you put the cursor on the right you have total thd%
( 10 harmonics) if you put on the 2k line you have 2nd harm, on 3k you have 2nd plus 3rd harm. Etc
tell me if it works
PS.
Instead, If you try only harm(v(r8)) you’ll get standard harmonic analysis, every line is the amplitude of the relative harmonic.
No cumulative percentage values.
Federico
Try follow these instructions, I hope to make no mistakes
1) select 1K Hz in the signal generator freq.
2) go to transient analysis
3) run option =normal,State var=Zero, check operating point
4) in the limits form put time range= 5m
5) maximum time step = 0.001/1024 (a good choice for accurate
results)
6) in “X-expression” type “F” ( means freq.), in Y- expr. type
thd(harm(v(R8)))
7)in X- range type 10k( or more), in Y range type 50,1e-5
8) click on the button for logarithmic Y plot ( second from the left)
9) click on the “transient “ menu (top of the screen), go to DSP
parameters
7) set upper time limit= 5m, lower time limit = 4m, number of
point= 1024; status= ON
(in this way the analysis is from 4m to 5m cutting out transient)
then run the simulation
the result are as a percentage value. If you put the cursor on the right you have total thd%
( 10 harmonics) if you put on the 2k line you have 2nd harm, on 3k you have 2nd plus 3rd harm. Etc
tell me if it works
PS.
Instead, If you try only harm(v(r8)) you’ll get standard harmonic analysis, every line is the amplitude of the relative harmonic.
No cumulative percentage values.
Federico
The strange thing is that left output is 60m, right output is 400m and across load is again 60m.
It seem that it works only on clipping, because left and right go tovery high values but load is only 80m.
I will try to follow instructions...
It seem that it works only on clipping, because left and right go tovery high values but load is only 80m.
I will try to follow instructions...
Bernhard,
You are optimizing the circuit to mimimize certain type(s) of distortion (particularly the simulator's distortion). With real circuitry, there are many types of distortion. Your circuit will not correct all types of distortion. When you optimize to correct some types of distortion, other types are likely to get worse. And the simulator is not showing that.
For example, I don't know if the simulator takes into consideration the distortion caused by the op-amps (NE5532). How does that distortion cancel?
The real test will be to build the circuit and use a distortion analyzer.
Now, I may "sound like a broken record", but I'm learning about this too as I go along.
JF
You are optimizing the circuit to mimimize certain type(s) of distortion (particularly the simulator's distortion). With real circuitry, there are many types of distortion. Your circuit will not correct all types of distortion. When you optimize to correct some types of distortion, other types are likely to get worse. And the simulator is not showing that.
For example, I don't know if the simulator takes into consideration the distortion caused by the op-amps (NE5532). How does that distortion cancel?
The real test will be to build the circuit and use a distortion analyzer.
Now, I may "sound like a broken record", but I'm learning about this too as I go along.
JF
Bernhard,
Feed forward techniques work but they are difficult to tune
also in simulation.
I made (simulated) some tube preamp and I obtained
good performance but not amazing.
I expect real things will go only slightly better with feed forward
(probably distortion will halve)
Too much effort for a technique that SANSUI promoted ( do you remember
the so called Super feed forward) but that people always said to have a bad
sound.
Federico
.
Feed forward techniques work but they are difficult to tune
also in simulation.
I made (simulated) some tube preamp and I obtained
good performance but not amazing.
I expect real things will go only slightly better with feed forward
(probably distortion will halve)
Too much effort for a technique that SANSUI promoted ( do you remember
the so called Super feed forward) but that people always said to have a bad
sound.
Federico
.
problem...
Federico,
after trying the THD(HARM(v(X))) with some amps and bare op amps, I got some terrible suspect...
Did it with a bare sine source: Looks exactly the same 🙁
That does not work, the amp got the same thd like the sine source itself
Could you prove that ? F=1k A=5
So there is hope again 🙂
Thanks, Bernhard
I will delete the thd picture as it is not valid...
Federico,
after trying the THD(HARM(v(X))) with some amps and bare op amps, I got some terrible suspect...
Did it with a bare sine source: Looks exactly the same 🙁
That does not work, the amp got the same thd like the sine source itself
Could you prove that ? F=1k A=5
So there is hope again 🙂
Thanks, Bernhard
I will delete the thd picture as it is not valid...
I usually plot a transient analysis before doing anything else. If it doesn't look like a sine wave, no mater how small its THD is, it is NOT a sine wave.
this looks ugly
Want a transient analysys ?
Here it is:
I connected a 5V sine source between left output and load to simulate heavy distortion/modulation of the left output stage.
The sine source is not inside the feedback loop of the output stage.
Black: Error output, blue: left output stage, green: Voltage across load.
5V square,
The gaps of the spikes in the green curve (voltage across load) are 0,63µs.
Seems to work rather good, eye ?
The question for real life is:
Will the distortion of the error amp depend on output voltage or output current ?
Will the distortion of the error amp be smaller than that of the left side ?
How much will distortion be reduced ?
😕
millwood said:
Want a transient analysys ?
Here it is:
I connected a 5V sine source between left output and load to simulate heavy distortion/modulation of the left output stage.
The sine source is not inside the feedback loop of the output stage.
Black: Error output, blue: left output stage, green: Voltage across load.
5V square,
The gaps of the spikes in the green curve (voltage across load) are 0,63µs.
Seems to work rather good, eye ?
The question for real life is:
Will the distortion of the error amp depend on output voltage or output current ?
Will the distortion of the error amp be smaller than that of the left side ?
How much will distortion be reduced ?
😕
new idea !!!!!!!!!!
The reference for the error amp is the voltage of the left output stage against ground.
As mentioned, there could eventaually be a problem for the error amp to keep the floating ground for the speaker stable @ 0 Volts.
Assumed that the ouput stage on the left side provides a 100% clean signal, there is nothing to do than keep the floating ground @ zero.
It could improve performance, if the voltage across the speaker is referenced for the error amp.
Or will it be the same 😕
I think it will be the same 🙄
The reference for the error amp is the voltage of the left output stage against ground.
As mentioned, there could eventaually be a problem for the error amp to keep the floating ground for the speaker stable @ 0 Volts.
Assumed that the ouput stage on the left side provides a 100% clean signal, there is nothing to do than keep the floating ground @ zero.
It could improve performance, if the voltage across the speaker is referenced for the error amp.
Or will it be the same 😕
I think it will be the same 🙄
I have decided that Bernard is an AI, kin to Gibson's Wintermute
or Continuity. He has taken it unto himself to confuse us with
conundrums and Escher schematics.
or Continuity. He has taken it unto himself to confuse us with
conundrums and Escher schematics.
- Status
- Not open for further replies.