Hi Heinz, "inside" this box is the formula written above of it, an equation for 2nd harmonics. Without it there is no distortion. The high order harmonics are definitely created by nfb, the box only creates 2nd harmonics, all verified.
John Curl, i have read the article, you were refering to his equation #17. I am aware that his equation is closed loop. I tried in sims with a similar simple opamp model, all i got was an increase in phasehift of ~0.01% for very low levels, no measurable increase for increased levels.
We have 2 options:
1, This is only valid for extremely poor circuits like the OPA with a (GBW of 1mhz
)
2, his equation is missing the increased error voltage by phasedifference between input and output or something like that.
I must admit that i am not able to really reproduce his math.
I will continue researching.
Mike
John Curl, i have read the article, you were refering to his equation #17. I am aware that his equation is closed loop. I tried in sims with a similar simple opamp model, all i got was an increase in phasehift of ~0.01% for very low levels, no measurable increase for increased levels.
We have 2 options:
1, This is only valid for extremely poor circuits like the OPA with a (GBW of 1mhz
)2, his equation is missing the increased error voltage by phasedifference between input and output or something like that.
I must admit that i am not able to really reproduce his math.
I will continue researching.
Mike
jcx said:
I agree completely with this, some of the feedback bad name in certain circles, is because is not used enough amount of negative feedback.
Is note an easy cake to design a amp with lots of negative feedback , because stability issues became harder to solve
So the easy road , the safer low feedback.
But when we achieve a amp with so much feedback, in wich all the error signal (distortion) , is at thermal noise level . That's the crux .
Hi, MikeB,
Sorry, its 2 Y's, Symasym
Is that John Curl+JCX discussing that there are distortions that actually generated by feedback, but cannot be eliminated by feedback itself, hence non-feedback amp will always sound different than feedback amp (no matter how the feedback amp is configured)?
Hi, JCX,
From the your link, is that generated bipolars are better than Jfet differential?
Sorry, its 2 Y's, Symasym
Is that John Curl+JCX discussing that there are distortions that actually generated by feedback, but cannot be eliminated by feedback itself, hence non-feedback amp will always sound different than feedback amp (no matter how the feedback amp is configured)?
Hi, JCX,
From the your link, is that generated bipolars are better than Jfet differential?
So much to learn about feedback from your posts
This is the problem. If you have a very high open loop linearity, then feedback isn't really necessary. If you have poor open loop linearity, then you can have added effects, such as phase modulation, perhaps making a worse subjective result, than simply some lower order harmonic or IM distortion. My experience when looking at the open loop linearity of many IC op amps is that they are not very linear, i.e. their transfer function does not map as a straight line. This must create phase modulation, if the open loop gain is being modulated by the nonlinearity of the op amp open loop transfer function. Feedback would be great, IF it were equal over the entire working bandwidth, but it almost always isn't.
Personally, I can't convey this concept as well as Barrie Gilbert can, so I give up trying to do so.
Personally, I can't convey this concept as well as Barrie Gilbert can, so I give up trying to do so.
john curl said:
John, Mike,
I think no one has issue with the non-linearity of the opamp open loop. I always try to remember that whatever you do to the loop, the opamp itself ALWAYS works open loop: if the OL gain is 100.000, and the Vout is 10V, the effective Vin = 100uV, and that 100uV is amplified 100.000 times with all the raw non-linearity in the OL transfer function.
The fact that initially, when increasing NFB from zero, the higher harmonics actually INCREASE, until you get beyond a certain point, and it all starts to decrease again, has also been clearly documented by Peter Baxandall in WW in the 70-ies.
I also don't think anyone has issue with the fact that PM would result from this. What I read in MikeB's posts is that he has identified it, but the numbers are quite small. So it appears to be a case of how significant this is? If indeed we can increase the effective FB such that ALL artifacts disappear below the noise level, have we then not got a "perfect amp"?
There appears to be a continuum in "quality" if you like from zero through moderate to high NFB, also shaped by the inherent (non)linearity of the main forward loop, that determines at what spot in the continuum you get best results.
So, I feel it is quite difficult to speak about these issues in an absolute sense: it depends on a particular topology and implementation whether any of them is significant enough to lose sleep over.
My 2 (euro)cents worth.
Jan Didden
This is difficult
If the mechanism of this one is like above, then I think it will not show up with a single or simple sinusoidal feed as input. This will show up more clearly if the input signal has many frequencies with many amplitudes = real music
Distortion = downscaled output - input.
How can we "null" an amp (to see the distortion artifacts) with real music as the input signal? Is there a measurement device/program that is clean and fast and accurate/carefull enough to do that?
.... and it has also been done copmputerized by digitizing a section of both input and output and convolving the two in the freq domain. That gives you the difference between in-and output with real-life misuc signals. Don't ask me how it works, as I haven't got a clue (well, a small clue anyway), but that's how it was called. It apparently is a standard technique in comms engineering. Forgot where I read it, I think is one of the AES journal issues.
Jan Didden
Jan Didden
Hi !
I made more simulations with my virtual amp...
Here are the results (Phaseshifts) for 2 different outputvoltages:
gain 2000,1%
1v: -2.360E-01
2v: -2.364E-01
change: 0.0004°, 0.17%
gain 20000,1%
1v: -2.437E-02
2v: -2.441E-02
change: 0.00004°, 0.16%
gain 2000,5%
1v: -2.394E-01
2v: -2.539E-01
change: 0.0145°, 6.1%
gain 20000,5%
1v: -2.473E-02
2v: -2.629E-02
change: 0.00156°, 6.31%
symasym: (10khz,gain ~20000,<1%)
12v: -1.484E+00
6v: -1.484E+00
change: below 4digit float...
The effect does exist, phasehift variated by amplitude of signal. (causing FM-distortion for complex signals)
I think it is caused by increased distortion with level, getting less compensation for phasehift.
But again, reasonable low openloop distortion and high gain will reduce the effect below noiselevel.
An amp with low gain and high openloop distortion will suffer from it.
The nfb itself does not create any dynamic phaseshifts, only in combination with level dependent distortion.
In symasym, the distortions are barely level dependent, and the input-LTP has too low swing to really distort.
Mike
I made more simulations with my virtual amp...
Here are the results (Phaseshifts) for 2 different outputvoltages:
gain 2000,1%
1v: -2.360E-01
2v: -2.364E-01
change: 0.0004°, 0.17%
gain 20000,1%
1v: -2.437E-02
2v: -2.441E-02
change: 0.00004°, 0.16%
gain 2000,5%
1v: -2.394E-01
2v: -2.539E-01
change: 0.0145°, 6.1%
gain 20000,5%
1v: -2.473E-02
2v: -2.629E-02
change: 0.00156°, 6.31%
symasym: (10khz,gain ~20000,<1%)
12v: -1.484E+00
6v: -1.484E+00
change: below 4digit float...
The effect does exist, phasehift variated by amplitude of signal. (causing FM-distortion for complex signals)
I think it is caused by increased distortion with level, getting less compensation for phasehift.
But again, reasonable low openloop distortion and high gain will reduce the effect below noiselevel.
An amp with low gain and high openloop distortion will suffer from it.
The nfb itself does not create any dynamic phaseshifts, only in combination with level dependent distortion.
In symasym, the distortions are barely level dependent, and the input-LTP has too low swing to really distort.
Mike
Funny thing some issues wax an wane endlessly. Nothing wrong, for it shoud be far worse to fall into complacency instead of being kept sharp reading once in a while thought provoking arguments.
First, we should acknowledge maths, and theory built on this tool, cannot be questioned per se, but at most missuse or missinterpretation of its results. If high gain, high global feedback supports the thesis of reduced distortion and improved reproduction accuracy, it cannot be falsified as a concept. Improper application / understanding is to be blamed instead.
Second, as much as there will never be a perfect amplifier (or reproduction chain to be more precise), it is also true from what we know and even from what we suspect will be learned from the auditory system (including cortical processing), there are thresholds below which disturbances simply are ignored. This includes all forms of noise and distortion, so it is in priniple possible to build amplifiers "perfect" in the sense the contributed arctifacts fall below this threshold.
Third, for the case of "how much OL gain and linearity" vs. "how much feedback", it must be remembered NFB provides **error attenuation**, that is, for a given OL gain and given level of distortions, NFB corrects as much as the difference between the OL gain and the desired system gain. If you increase OL gain somehow keeping OL distortion constant, final distortion will be reduced by way of increased feedback to keep system gain. Conversely, reducing OL distortion while keeping OL gain will also reduce final distortion (keeping the same feedback correction factor). None of these features (OL gain and distortion, and amount of feedback) should be taken separately or as different concepts, but interrelated.
One of the first issues rised here had to do with the necessary bandwith for the sake of good reproduction. Again, acoustic bandwith from the standpoint of hearing should not be confused with required bandwith in particular taking into consideration NFB based topologies.
If we want a constant correction factor provided by NFB up to 20 KHz, and assume for simplicity a 6 dB/Oct. rolloff, then it is clear a 6 dB feedback margin implies automatically the same flattening but up to 40 KHz for the whole system. You guess the rest. And I am not implying here the desirable requisite of minimum phase shift up to 20 KHz, a wellcome byproduct.
Some more considerations are condensed in a pdf I posted here for those interested.
In the end, no matter technollogy or topology, there is not at present times excuse for not designing and building amplifiers capable of dumping undesirable reproduction byproducts essentially below an average threshold of listening perception. This is what the smartest solid state analog designers are doing banking on large resources at TI, National etc.
Rodolfo
First, we should acknowledge maths, and theory built on this tool, cannot be questioned per se, but at most missuse or missinterpretation of its results. If high gain, high global feedback supports the thesis of reduced distortion and improved reproduction accuracy, it cannot be falsified as a concept. Improper application / understanding is to be blamed instead.
Second, as much as there will never be a perfect amplifier (or reproduction chain to be more precise), it is also true from what we know and even from what we suspect will be learned from the auditory system (including cortical processing), there are thresholds below which disturbances simply are ignored. This includes all forms of noise and distortion, so it is in priniple possible to build amplifiers "perfect" in the sense the contributed arctifacts fall below this threshold.
Third, for the case of "how much OL gain and linearity" vs. "how much feedback", it must be remembered NFB provides **error attenuation**, that is, for a given OL gain and given level of distortions, NFB corrects as much as the difference between the OL gain and the desired system gain. If you increase OL gain somehow keeping OL distortion constant, final distortion will be reduced by way of increased feedback to keep system gain. Conversely, reducing OL distortion while keeping OL gain will also reduce final distortion (keeping the same feedback correction factor). None of these features (OL gain and distortion, and amount of feedback) should be taken separately or as different concepts, but interrelated.
One of the first issues rised here had to do with the necessary bandwith for the sake of good reproduction. Again, acoustic bandwith from the standpoint of hearing should not be confused with required bandwith in particular taking into consideration NFB based topologies.
If we want a constant correction factor provided by NFB up to 20 KHz, and assume for simplicity a 6 dB/Oct. rolloff, then it is clear a 6 dB feedback margin implies automatically the same flattening but up to 40 KHz for the whole system. You guess the rest. And I am not implying here the desirable requisite of minimum phase shift up to 20 KHz, a wellcome byproduct.
Some more considerations are condensed in a pdf I posted here for those interested.
In the end, no matter technollogy or topology, there is not at present times excuse for not designing and building amplifiers capable of dumping undesirable reproduction byproducts essentially below an average threshold of listening perception. This is what the smartest solid state analog designers are doing banking on large resources at TI, National etc.
Rodolfo
I hope that some of you have learned something new, here. Personally, I have found that removing global feedback can make a better preamp, all else being close to equal My CTC preamp does not use global feedback, but it is difficult to make properly, as well.
The Parasound JC-2 preamp that is soon to be released, has lots of loop feedback, even though it uses exactly the same toplogy on the input stage. Parasound cannot afford to have any sort of 'marginal' specs, if it is to sell also in the mid-fi marketplace. I doubt that it will sound quite as good, as the CTC, because of previous experience with three other similar preamps that use almost the same topology.
My associate, Charles Hansen, does not use global feedback at all, even in his power amps. He doesn't do this for nothing. He knows that there is a difference in sound quality, when he uses global negative feedback.
I think the true test will be with direct subjective experiences between Halcro, Parasound, and Ayre. Each of these components are first rate, but there will be serious sonic differences to their ears.
The Parasound JC-2 preamp that is soon to be released, has lots of loop feedback, even though it uses exactly the same toplogy on the input stage. Parasound cannot afford to have any sort of 'marginal' specs, if it is to sell also in the mid-fi marketplace. I doubt that it will sound quite as good, as the CTC, because of previous experience with three other similar preamps that use almost the same topology.
My associate, Charles Hansen, does not use global feedback at all, even in his power amps. He doesn't do this for nothing. He knows that there is a difference in sound quality, when he uses global negative feedback.
I think the true test will be with direct subjective experiences between Halcro, Parasound, and Ayre. Each of these components are first rate, but there will be serious sonic differences to their ears.
Hi, Mr. John Curl,
There is another gentlemen saying about the same thing as you describe (I think),but he said it in "Tube vs Transistor" theme. http://www.diyaudio.com/forums/showthread.php?postid=829887#post829887
If one of the cause is inherent of the natural properties of transistor (varying internal C due to varying Voltage and Current), would cascoding helps reducing this "feedback-non-aidable" distortion?
There is another gentlemen saying about the same thing as you describe (I think
),but he said it in "Tube vs Transistor" theme. http://www.diyaudio.com/forums/showthread.php?postid=829887#post829887If one of the cause is inherent of the natural properties of transistor (varying internal C due to varying Voltage and Current), would cascoding helps reducing this "feedback-non-aidable" distortion?
Mike, Rodolfo,
This is a very educational thread, between MikeB, providing the technical background and insight, and Rodolfo's last post, tying it all together conceptually. This gives a consistent view of NFB and what it can and cannot do, and the interrelationship between OL gain, OL distortion, feedback gain and distortion reduction. It really shows that you HAVE to treat those related factors together in assessing an amplifier.
I hope the thread continues in that vein.
Jan Didden
This is a very educational thread, between MikeB, providing the technical background and insight, and Rodolfo's last post, tying it all together conceptually. This gives a consistent view of NFB and what it can and cannot do, and the interrelationship between OL gain, OL distortion, feedback gain and distortion reduction. It really shows that you HAVE to treat those related factors together in assessing an amplifier.
I hope the thread continues in that vein.
Jan Didden
Yes, Jan
but still there is no simple answer why 'blameless-styled' high feedback amplifiers are not prefered by audiophiles. Candidates are:
1)FM, PIM... if they have the same nature that doppler distortion of moving membrane and much lower extend, much doubts...
2)clipping behavious... yes, bad, but why should it clip at normal levels
3)HD at low levels and high frequency- still not very high
all Mike's point are absolutely true sound-wise, but there are still no test signals to uncover this hidden distortions...
regards
but still there is no simple answer why 'blameless-styled' high feedback amplifiers are not prefered by audiophiles. Candidates are:
1)FM, PIM... if they have the same nature that doppler distortion of moving membrane and much lower extend, much doubts...
2)clipping behavious... yes, bad, but why should it clip at normal levels
3)HD at low levels and high frequency- still not very high
all Mike's point are absolutely true sound-wise, but there are still no test signals to uncover this hidden distortions...
regards
This may sound a bit harsh, but I think that audiophiles may be easily fooled into thinking that they are listening to whatever type of amplifier that you want by just swapping enclosures. They will describe the sound relying on the kind of amplifier they have been told they are listening to, despite whatever other thing you have put inside the box
Sometimes I'm tempted to mount a small SMPS and a LM3875 hidden inside the chasis of some working tube monoblocks, just for fun...
Sometimes I'm tempted to mount a small SMPS and a LM3875 hidden inside the chasis of some working tube monoblocks, just for fun...
darkfenriz said:
Hi darkfreniz,
Your post implies that audiophiles don't like the blameless amp (and I assume that this is so, just for the sake of discussion) because of technical reasons. That is like saying that many europeans don't like japanese cars for technical reasons. In reality it has to do with image and often wrong perceptions and hear-say; objective technical data shows that japanese cars are at least the equal on technical terms to any other car.
Possibly, a similar effect is going on in audio, making it more difficult even to establish if the technical reasons have any influence at all.
Jan Didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.