Evaas said:
Coming from a digital circuits background, I had a hard time trying to understand NFB. How does it work? we're half-way to the next clock tick by the time the signal comes around, and the train already left the station.
clearly I overestimated how well it worked, hence my confusion.
It is all a matter of perspective. A nfb loop has a delay measured in the fractions of uSec. Lets take an example of 0.5uS (and many amps are mcuh faster). A worst-case high-frequency wave of 20kHz advances 3.6 degrees in 0.5uS. Thus, there is a phase shift beween the signal and the nfb of a few degrees. That means that the nfb doesn't correct an error to 100%, perhaps to 98%.
Just like when you drive your car. If nfb came to late, you wouldn't be able to follow the road. But your sensing that the car sways from the correct path and your correction (the nfb) is much, much faster than the car going astray, so you can keep the car straight. Of course, there still remains a small deviation, and if you would look in the small details of the car trajectory you would see a wiggly line. This is equivalent to the remaining distortion you would see in a nfb amp.
As we can all see in our daily lives, nfb does a great job. As it does in amps.
Jan Didden
GRollins said:
Not really and who says you can't use music signals as test signals?
Of course it does.. many instruments have a slow build up and as long as you play the note it is a sine with some other sines added ontop of it. I'll try to post a graph I took of a flute I was playing.
No it's very useful since you can see if the harmonics present in a signal stays at the correct levels (=frequency response) and also if some spectral components is added or subtracted from the signal.
Absolutely not but of course it make sense to use non static signals as well.
True!
At radio frequencies possibly but I think it's safe to say that modern electroncs with multi MHz bandwith is fast enough for slow audiosignals.
That would entirely depend upon the type of music but still a device can be judged by it's steady state performance and it's transient performance and since modern electronics can pass audio signals without audible degradation what you say has no practical relevance.
Cheers!
/Peter
Attachments
okapi said:
The higher OL gain and the more feedback you apply the lower the distortion. Good intrinsic linearity is obviously a good thing for starters.
NFB has its bad rep from low OL gain circuits and slow circuits.
With low Ol gain you can only add a small amount of feedback and that is often a sure way to decreased low order harmonics but increased high order.
The lowest distortion devices there is have good intrinsic linearity (may I propose calss A) high OL gain and plenty feedback.
Some prefer to add color to their audio though and that is not against the law..
/Peter
GRollins said:
This irony used time and again I can not understand.
Can you show a piece of gear that actually had 1ppm performance or are you only passing on that old myth?
And even if some of the '70s manufacturers put out some box with that numbers, at what frequency and level? Surely not sum of all distortions at all levels and all audio frequencies if I may guess.
You mentioned that gear with same specs can sound different. Well maybe time for another test round on the lab bench then maybe? Or remove the bias from knowing what gear is in the chain while playing.. ;-)
/Peter
janneman said:
yes, and of course I was exaggerating. digital circuits run in the GHz range these days, and even audio feedback circuits are as you say, much faster than the audio signal. but are they faster than we can perceive? it moves at a higher frequency than we can hear, but does that mean the signal it makes is also out of our hearing range? i'm trying to think up some way of having a lower fundamental.. in any case, 98% doesn't sound that good to me. I guess 10kHz would be 99% corrected. not perfect and very much within hearing range (granted you probably picked a slow example). But there's the small (stable) oscillation you mentioned. to how big it is? dunno, but the input is always changing and maybe the previous adjustments are working against the new signal. just for that half microsecond. then it has to overcorrect. not enough to go unstable, but enough to make it sound bad. i need to study this more though..
I think I disagree with you on that point. I don't see too many examples where it works well. I'd bet the cruise control on your car is some form of PID, not NFB. I don't know if it's a perfect fit, but I think of the proportional part of PID to be analagous to NFB. heck, voice coils provide some nice damping too, but they aren't very adjustable. as for the integral part, any amp analogies? the point is, getting those constants right is an art form. in an amp, maybe some loudmouths
think it doesn't sound right. if it's a robotic arm (I was playing with PID and other stuff on this arm last year) you'd better duck when you get it wrong. hmm, i guess that'd be the case of magic smoke escaping. still, on a big slow system you can see smaller problems even though it looks fine from far away. push it the wrong way and it starts shaking like mad for a while. small, but present. I got a lot of insight into it from that experience. as for driving, I'm not sure, but I'd wager there's a lot more going on than simple feedback. I've seen a lot of literature on robotic steering mechanisms, autonomous cars, etc, and IIRC most robust systems do a lot more than simple feedback for steering angle.
I've heard good evidence that feedback can't be used for many components of normal walking (because it's too slow) here's one relevant study . Ever been walking down the stairs and couldn't see your feet? maybe carrying a big box of audio parts? how stable are you when you expect more steps but instead discover you're at the bottom? i may not always fall, but i need some feed-forward there to look graceful.
ok, i don't know if i'm making any sense. I've been awake for 24 hours.
Pan said:
Peter,
i own two preamplifiers that use the same input devices, exact same output devices, both have shunt regulated powersupplies, both are balanced designs and both have high Class A level biased output stages.
One is an ultralinear NFB design, the other open loop.
Guess what, the differences are subtle and i seriously doubt that more than 100ppm of the world population can tell them apart.
But Hey, in this world one has to be either one or the other.
Evaas said:
yes, and of course I was exaggerating. digital circuits run in the GHz range these days, and even audio feedback circuits are as you say, much faster than the audio signal. but are they faster than we can perceive? it moves at a higher frequency than we can hear, but does that mean the signal it makes is also out of our hearing range? i'm trying to think up some way of having a lower fundamental.. in any case, 98% doesn't sound that good to me. I guess 10kHz would be 99% corrected. not perfect and very much within hearing range (granted you probably picked a slow example). But there's the small (stable) oscillation you mentioned. to how big it is? dunno, but the input is always changing and maybe the previous adjustments are working against the new signal. just for that half microsecond. then it has to overcorrect. not enough to go unstable, but enough to make it sound bad. i need to study this more though..
I think I disagree with you on that point. I don't see too many examples where it works well. I'd bet the cruise control on your car is some form of PID, not NFB. I don't know if it's a perfect fit, but I think of the proportional part of PID to be analagous to NFB. heck, voice coils provide some nice damping too, but they aren't very adjustable. as for the integral part, any amp analogies? the point is, getting those constants right is an art form. in an amp, maybe some loudmouths
as for driving, I'm not sure, but I'd wager there's a lot more going on than simple feedback. I've seen a lot of literature on robotic steering mechanisms, autonomous cars, etc, and IIRC most robust systems do a lot more than simple feedback for steering angle.
I've heard good evidence that feedback can't be used for many components of normal walking (because it's too slow) here's one relevant study . Ever been walking down the stairs and couldn't see your feet? maybe carrying a big box of audio parts? how stable are you when you expect more steps but instead discover you're at the bottom? i may not always fall, but i need some feed-forward there to look graceful.
ok, i don't know if i'm making any sense. I've been awake for 24 hours.
Lots of good points. I should have known better than to get mixed up in this
Of course there's more involved in driving, but in keeping the car on a specific track, its a nfb loop: you sense that the car deviates (either because you see it, feel it, both and maybe you wife tells you 'stay on the road'). You correct it by turning the steering wheel in the opposite direction.
There is probably some smart processing in your brain that determines how far or fast you are correcting, but the concept is simple. Robots of course have a lot of complex processing to sense the environment and then decide if they want to change direction and how far, but we're lucky in audio because we can tell the fb when to chance in which direction with a simple resistive divider.
I just brought up this example to show that the very general 'the feedback comes too late to be effective' doesn't hold water.
I like your analog of walking down the stair carrying a box. When you can't see the steps for the box, your feedback loop is effectively broken and you are acting open loop. Which, as I'm sure you agree, is scary
Jan Didden
janneman said:
I like your analog of walking down the stair carrying a box. When you can't see the steps for the box, your feedback loop is effectively broken and you are acting open loop. Which, as I'm sure you agree, is scary
Jan Didden
You're right, the body does work on feedback. Part of this feedback is a function of what we've learned from past experience. And the walking down the stairs example, which I agree is scary, is possible due past experience. Some activities we perform on a regular basis we don't even need to think about, because we've formed Conditioned Reflexes for these activities. It all happen in the subconscious. When you practice activities like in sports these reflexes get formed. I know just a little bit about neurophysiology, not much.
tnargs said:
--I'm "anti-DBT" because it doesn't work. Reread what I wrote above. It takes time to sort out subtle effects and DBT tests do not and cannot run over months. That one flaw alone is enough to kill double blind tests as a realistic test regime.
--No one measures system distortion "at the listener's ear." It's always piece by piece, because that's the way gear is sold. Anytime I've mentioned testing cumulative distortion through the entire playback system, I've been jeered at. Honestly, I'm not sure it would be all that useful in the real world. What are you going to do...run a feedback loop all the way from the speaker cone (i.e. a servo) all the way back into the phono/digital playback amplification? I don't think so.
Egad...think of the delay involved!
--Clearly you were either too young or not into audio in the late '70s during the Distortion Wars. Nikko is/was my favorite example (awesome specs...ghastly sound), but all the mid-fi manufacturers (e.g. Yamaha) and some of the high end companies were hammering the THD spec. As a technical exercise, it was marvelous. As an audio experience, it left much to be desired. In hindsight, they got a lot of other things wrong (and THD proponents attempt to explain away the awful sound solely in terms of those other factors--no criticism of THD allowed--ever), but it was an interesting time to live through.
--I am always interested in anything that might lead to rational explanations for what we hear. That has never changed.
The mindset that says, "If we can't explain it, it either doesn't exist or must be forced into an existing model whether it fits or not," is the same attitude that leads people to say, "It's God's will..." when in fact there is a more prosaic explanation for what's going on. Lunar eclipses occur because the Moon moves into the Earth's shadow--not because a monster is eating the Moon's physical substance. Guess which viewpoint prevailed throughout most of mans' existence (and persists to this day in some corners of the world). Sometimes a little patience is required. Just because the answer isn't known today does not mean that it won't be known fifty years from now. In the interim it would be downright silly to deny that eclipses--or any other natural phenomenon--occur, just because you can't explain them.
janneman said:
Jan, I don't think your car analogy holds up. My biggest objection being that we, being sighted creatures, see the road ahead and make predictive correction to the steering before we arrive in the curve itself. (Yes, NFB is involved in driving...I'm not trying to claim it isn't.) Perhaps you could recast your analogy as an argument for feed-forward correction, though...
However...
My larger, and more subtle, point about negative feedback (and by extension, double blind testing, THD, etc.) is well demonstrated by peoples' reaction to your post.
Several posters chimed in with words to the effect of "Good point," whilst failing to perceive the flaw in the analogy. This is exactly the point I was attempting to make earlier. Those who 'believe in' (this is what I refer to as "Science as Religion") negative feedback will hasten to agree with anything that appears to support their position. There's zero evidence that the average audiophile has ever really given NFB, THD, etc. any critical thought. Even the 'above-average' audiophile tends to take it as an article of faith that such things work as advertised. Why? Because the explanation advanced seems so compelling on the face of it. I, myself, believed in it for years.
But what if there's more to it?
I would like to ask that you not take what I'm saying as condescending, critical, or 'snarky.' In fact, I am indebted to you for providing an opportunity to examine the subjective belief that NFB is a panacea for all ills. To the extent possible, try to look objectively at your post (pretend that someone else advanced the car analogy?) then at the enthusiastic and uncritical acceptance evidenced by some of the posts that followed. Compare that to the uncritical-listeners-following-the-audio-guru complaint of the 'objective' measuring crowd. See the correlation? (Congratulations! You have achieved Guru status!) Uncritical acceptance is not good, regardless of which side of the debate you're on. Unfortunately, it is rife on the measurement side of the argument.
Perhaps this will illuminate my position towards conventional measurements as well--perhaps even better--than anything I could say in isolation.
Pan said:
...who says you can't use music signals as test signals?
.. many instruments have a slow build up and as long as you play the note it is a sine with some other sines added ontop of it. I'll try to post a graph I took of a flute I was playing.
--If you can come up with a good test using real music, then we will all be indebted to you. It's not as though the concept is new. Many have tried, some have claimed success (Hafler had a cancellation test that they pushed for a while), then faded away when it was realized that they weren't really advancing our understanding of what's going on.
--There are really very few instruments such as you describe. Flute happens to be one of them, but it is an atypical instrument. Synthesizers are another, depending on the settings. Organs are frequently cited, but are not a good example, due to the complexity of the waveform.
In any event, most people do not listen only to flute, and a distortion correction mechanism that only worked with one instrument would be a poor mechanism, indeed.
Pan said:
--Old myth? Hardly. See above.
--You contradict yourself in two consecutive posts. You can't have it both ways. Either negative feedback can handle all levels and all frequencies simultaneously...or it cannot. I submit that it cannot.
--Or perhaps it's time to abandon the bias that negative feedback and THD specs are sufficient descriptors of the sound of audio gear. Might I suggest that you spend a little less time typing and a little more time listening? Not to the gear you already own, but to gear you aren't familiar with. For most people, this will involve a lot of friends with good equipment or a lot of travel to decent audio shops. Preferably both. Yes, I am well aware of the hassle and effort involved. I was in a fortunate position in that I lived in a relatively affluent area for a while--one with many audio shops carrying many kinds of gear--then worked in the industry for a time. I've heard literally hundreds of pieces of audio gear, not five or ten, and I can assure you it was a revelation. I used to 'believe' that measurements told us anything and everything that was relevant. I got over that attitude. Had to. Reality came knocking.
Try expanding your listening experience before you make indefensible claims about distortion, NFB, etc.
Grey
Hello Grey,
I am well aware that you are anti-nfb. That's OK, it's free country. But often these personal opinions start a life of their own and before you know it people think that everyone has that opinion.
The fact that many thousands of people select nfb amps for their system should put the absoluteness of such statements in perspective, unless you wish to maintain that all those who listen with pleasure to nfb amps are all dilusioned, which I think we won't want to do.
Of course, the car analogy is just that, an analogy. Feedforward *may* be involved in steering a car around the bend that you see coming. Still, even then, you are constantly trying to keep the vehicle on the intended path, and the only way to do that is to react in negative phase to deviations, i.e. nfb.
Jan Didden
I am well aware that you are anti-nfb. That's OK, it's free country. But often these personal opinions start a life of their own and before you know it people think that everyone has that opinion.
The fact that many thousands of people select nfb amps for their system should put the absoluteness of such statements in perspective, unless you wish to maintain that all those who listen with pleasure to nfb amps are all dilusioned, which I think we won't want to do.
Of course, the car analogy is just that, an analogy. Feedforward *may* be involved in steering a car around the bend that you see coming. Still, even then, you are constantly trying to keep the vehicle on the intended path, and the only way to do that is to react in negative phase to deviations, i.e. nfb.
Jan Didden
i think that you can capture everything you need to know about an amplifier in a distortion measurement if you accept a few simple suppositions.
1. the purpose of an amplifier is to linearly duplicate the input signal (a voltage waveform) with some gain.
2. distortion is simply the difference between the input waveform and the output waveform of the amplifier. In other words, take the input waveform scale it by a factor to give the best fit to the output and subtract. what you are left with is the Total distortion (this includes amplitude and phase distortion). If there is no distortion the subtraction will equal zero.
so there you have it. a simple distortion measure that captures all aspects of an amplifier.
If you made an amp that could produce a difference waveform not significantly different from zero, you could listen to this amplifier, but you would not need to to know that it is doing exactly what it is designed to do - amplify a signal linearly.
although it is somewhat off topic, i did want to confirm for those interested that the nervous system uses feedback everywhere. i cannot think of a single neural circuit where this is not the case. a great example of feedback that some of you may be familiar with is when a chicken manages to run around after it's head is cut off. the ability to run is entirely based on feedback circuits located in the spinal column. i would also like to point out that there is no way you could know that your brain is using feedback just by listening and theorizing. you would need to go in and do an experiment that involved amplifiers.
1. the purpose of an amplifier is to linearly duplicate the input signal (a voltage waveform) with some gain.
2. distortion is simply the difference between the input waveform and the output waveform of the amplifier. In other words, take the input waveform scale it by a factor to give the best fit to the output and subtract. what you are left with is the Total distortion (this includes amplitude and phase distortion). If there is no distortion the subtraction will equal zero.
so there you have it. a simple distortion measure that captures all aspects of an amplifier.
If you made an amp that could produce a difference waveform not significantly different from zero, you could listen to this amplifier, but you would not need to to know that it is doing exactly what it is designed to do - amplify a signal linearly.
although it is somewhat off topic, i did want to confirm for those interested that the nervous system uses feedback everywhere. i cannot think of a single neural circuit where this is not the case. a great example of feedback that some of you may be familiar with is when a chicken manages to run around after it's head is cut off. the ability to run is entirely based on feedback circuits located in the spinal column. i would also like to point out that there is no way you could know that your brain is using feedback just by listening and theorizing. you would need to go in and do an experiment that involved amplifiers.
GRollins said:
That is your opinion. I know there are a lots of people that uses null tests and bypass test with success. So no, it has not faded away and yes, it is helping us in understanding what is going on.
Flute, voice, the ringing notes of a solo string instruments, I'd say most instruments has a relaively basic multiple sinelike waveform during significant periods of time, and static measurements tell us all about the behaviour on such signals.
Sure that doesn't tell us the whole story but a lot.
What makes you feel an organ note is a complex waveform? It's a big flute in my book.
Of course and how lucky we are that a good feedback circuit gives us perfect fidelity even on complex music and transients in the audio bandwith.
Wow wow wow, now we're going fast!
Nope I do not contradict myself. I never said that anyone can take any circuit and apply feedback and all of a sudden it's a perfect piece of gear, that is not what I said/wrote or meant.
A piece of gear do not become perfect just becasue of feedback, but a piece of gear may use feedback to get SOTA performance and achieve HiFi and transparency. I hope you realize the difference?
Hehe, actually I was leaning towards the beliefs and dogma of no feedback circuits for many years and my current main set up use no feedback loops. ;-) Got ya´!
BUT I have always been observant and a hungry student and now I know better. Feedback circuits are just fine and gives better results if transarency is the goal. So I often plug in my soundcards wich gives better results and I'm building the ExtremA. I want not only good sound, I want amazing sound and that can only be achieved with feedback circuits.
But please you don't know how much time I have spent and spend listening.
Hehe you mean I should not listen to the no-NFB gear I have at home... ok!
But please who do you think you are? Or rather, who do you think I am? Do you think I was born yesterday? I stopped visiting audio shops years ago. I have higher standards than that. I have lots of friends with high performance rigs.
Oh wow, and I only listened to ten?? Get real..!
Try to get a grip before you let your fingers type such nonsense. I'm surprised you can spit out so much rubbish without knowing me and my background.
I'm not angry but I'm really suprised anyone would wanna look as bad as you make yourself do right now.
Have a nice day!
/Peter
For those that are REALLY interested in the difference test method, check out AudioDiffmaker from Bill Waszlo (the guy behind Liberty Instruments that makes the Praxis measurement system and others). Now you can do those difference tests at leisure, without tedious adjustments, to incredible sensitivity. I've seen/heard it demonstrated. Two audio tracks that on the face of it were identical. Run them through Diffmaker and we heard a complete sousa orchester, buried in one of the tracks below 100db or so. Can you imagine, a whole sousa orchester, and we didn't hear it. But it was brought out.
And it's free.
Jan Didden
And it's free.
Jan Didden