Hello,
I have experienced that if you have big feedback and therefor low distortion you have lost good soundstage or imaging reproduction. I think the keyword is transients. I think feedback destroys the ability for transients.
The german test magazine Stereoplay I think current issue July 2011 shows an ampflifier without feedback but with another trick to lower distortion.
Does somebody know more about this amp?
What do you think, guys?
I have experienced that if you have big feedback and therefor low distortion you have lost good soundstage or imaging reproduction. I think the keyword is transients. I think feedback destroys the ability for transients.
The german test magazine Stereoplay I think current issue July 2011 shows an ampflifier without feedback but with another trick to lower distortion.
Does somebody know more about this amp?
What do you think, guys?
I think the keyword is transients. I think feedback destroys the ability for transients.
The transient response is what happens between two stable states.
If feedback destroys the ability to reproduce transients, it would be preferable to try demonstrate it (it would be quite easy) than to stay at a stable belief state.
The transient response is what happens between two stable states.
If feedback destroys the ability to reproduce transients, it would be preferable to try demonstrate it (it would be quite easy) than to stay at a stable belief state.
The transient (if really did) should affect high freq response, and less in bass, but most differences of amplifiers sound quality is also in low (bass).
Often, oscillation start when I connect independent VAS to output buffer (with no global feedback) just because the input of buffer disturb the VAS output.
Stability is very crucial here, but the key isn't just transient, but something else, may be don't know, natural perhaps.
Often, oscillation start when I connect independent VAS to output buffer (with no global feedback) just because the input of buffer disturb the VAS output.
Stability is very crucial here, but the key isn't just transient, but something else, may be don't know, natural perhaps.
Hi,
A 100% correct observation, that´s due to time errors and disrupted phase relationships.
Large amplitude-high frequency-short duration signals are not transients, just ordinary music signals. Transients are distortion of oscillatory nature, often as a result of insufficient circuit response, originated from various mechanisms, some of them intimately associated with global feedback, which is the main cause of instability, reduced bandwidth, compressed dynamic range, degraded signal handling capability, slew-limiting, decreased dynamic margins, increased susceptibility to overloading, overdrive, overshoots, clipping, voltage peaks, spikes, surges and so on...giving the proper hard, cold, lifeless, unmusical low-THD-sound.
I've seen the same thing, but more with hollow state designs. I did a project using PP 807s for the final. Without any NFB connected, the 807s made lots of nasty higher order harmonics that became difficult to listen to for any length of time. However, the soundstage was wide open. Connecting both local and global NFB fixed the nastiness, but at the cost of soundstage. So it looks like a design trade-off here.
No idea if that's right or not, and not guessing. There's definitely more going on than the simplistic answer of just throwing as much NFB at the distortion problem as you can manage.
Which kind of feedback we talk about? - go to
http://www.eecs.berkeley.edu/~bora/publications/ISCAS98-Feedback.pdf
And how much gain stages are in the negative feedback (NFB) loop?
Read also the paper about
http://www.eecs.berkeley.edu/~bora/publications/ISCAS98-Feedback.pdf
and post #2097 about
http://www.diyaudio.com/forums/solid-state/171159-bob-cordells-power-amplifier-book-210.html
Last edited:
Quote
>I have experienced that if you have big feedback
>and therefor low distortion you have lost
>good soundstage or imaging reproduction
A 100% correct observation, that´s due
to time errors and disrupted phase relationships.
>I think the keyword is transients.
>I think feedback destroys the ability for transients.
Large amplitude-high frequency-short duration signals are not transients, just ordinary music signals. Transients are distortion of oscillatory nature, often as a result of insufficient circuit response, originated from various mechanisms, some of them intimately associated with global feedback, which is the main cause of instability, reduced bandwidth, compressed dynamic range, degraded signal handling capability, slew-limiting, decreased dynamic margins, increased susceptibility to overloading, overdrive, overshoots, clipping, voltage peaks, spikes, surges and so on...giving the proper hard, cold, lifeless, unmusical low-THD-sound.
As at the recording stage, the music signal already passes through some amp-ops having high global NFB, must not we understand that the soundstage is lost for ever ?
All what is underlined above is related to observable facts which could be seen even with a modest apparatus. Can you provide some convincing pictures showing them ? If not, the good soundstage and image reproduction in the absence of global negative feedback must be correlated to something else, most probably, the presence of high distorsions of all kinds.
>I have experienced that if you have big feedback
>and therefor low distortion you have lost
>good soundstage or imaging reproduction
A 100% correct observation, that´s due
to time errors and disrupted phase relationships.
>I think the keyword is transients.
>I think feedback destroys the ability for transients.
Large amplitude-high frequency-short duration signals are not transients, just ordinary music signals. Transients are distortion of oscillatory nature, often as a result of insufficient circuit response, originated from various mechanisms, some of them intimately associated with global feedback, which is the main cause of instability, reduced bandwidth, compressed dynamic range, degraded signal handling capability, slew-limiting, decreased dynamic margins, increased susceptibility to overloading, overdrive, overshoots, clipping, voltage peaks, spikes, surges and so on...giving the proper hard, cold, lifeless, unmusical low-THD-sound.
As at the recording stage, the music signal already passes through some amp-ops having high global NFB, must not we understand that the soundstage is lost for ever ?
All what is underlined above is related to observable facts which could be seen even with a modest apparatus. Can you provide some convincing pictures showing them ? If not, the good soundstage and image reproduction in the absence of global negative feedback must be correlated to something else, most probably, the presence of high distorsions of all kinds.
Last edited:
This is just a bucketful of silly generalizations, a blind repitition of what many mis-guided souls have said and repeated over and over again. As I have pointed out in my book "Designing Audio Power Amplifiers", negative feedback has gotten an undeserved bad rap over the years. Much of this is from guilt by association - with early solid-state amplifiers that were pretty bad in many respects, but which happened to typically have more negative feedback than tube amplifiers.
It is important to recognize that there are very bad amplifier examples of every genre. People who do not know how to apply feedback properly will indeed be fully capable of producing a poor-sounding amplifier. People who start with a junk design and then expect to make it sound good by applying negative feedback will also produce a poor-sounding amplifier.
There are many things about the sound of amplifiers that we do not fully understand or are not usually measured. However, most of the things mentioned above are readily measurable and those measurements on a good feedback amplifier show that these statements are just generalizations.
Cheers,
Bob
but we don't "understand" any such thing about PIM/TIM audibility - there are no DBT studies of audibility threshold for amplifier level of ns to ps
the nearest would be Doppler distortion from broad band audio loudspeaker drivers, also short cantilver geometric variations in phonocartridge setup and modulation with tracking angles
even the "Gold Standard" of anaolg master tapes have "FM" errors from mechanical scrape, stick/slip, bearing noise ect. that are many orders of magnitude higher than well designed negative feedback amplifer TIM
the nearest would be Doppler distortion from broad band audio loudspeaker drivers, also short cantilver geometric variations in phonocartridge setup and modulation with tracking angles
even the "Gold Standard" of anaolg master tapes have "FM" errors from mechanical scrape, stick/slip, bearing noise ect. that are many orders of magnitude higher than well designed negative feedback amplifer TIM
Last edited:
Hi all,
My experience is that the quality of the mains has a BIG influence in imaging and transients quality, regardless the amp topology. l always believed that putting more elco's in a analog PS would gain imaging and transients, but i was wrong. Most amplifiers with traditional analog power supply plays dull, because of a lazy current supply out of the mains power.
My experience is that the quality of the mains has a BIG influence in imaging and transients quality, regardless the amp topology. l always believed that putting more elco's in a analog PS would gain imaging and transients, but i was wrong. Most amplifiers with traditional analog power supply plays dull, because of a lazy current supply out of the mains power.
How exactly does that make it "play dull"?
With your typical capacitor input power supply, the mains power isn't even in the circuit the majority of the time. The majority of the time the amplifier is being powered from the reservoir capacitance.
So how does that translate into "playing dull"?
se
Having recently read some of your other posts and admiring your knowledge and apparent competence, I am puzzled that you have made such statements.
I am sorry but your definition of "transient" is incorrect. Please look up "transient response" in any good book on linear analog system theory. After everyone agreed on that, they also generalized the word transient, in terms of signals, to mean almost anything that is not "steady state". So, yes, a voltage step, a Dirac delta function (impulse), almost any relatively sudden change in a signal is typically referred-to as "a transient". "Ordinary music signals", whether distorted or not, usually contain many, many transient signals.
An amplifier's response to the transients is _related_ to the formal "transient response" of the amplifier, as discussed in all analog system theory textbooks, which, if I recall correctly after over 30 years, is a system's response to a unit STEP function, which is an input with amplitude that changes from 0 to 1 in zero time (usually at time=0) and then stays = 1 forever, having been 0 forever before time = 0. The formal "impulse response" is similar, but uses a mathematical impulse function instead of a step function.
----------
Your statement that global feedback is the main cause of instability and overshoots could probably not be farther away from what I have learned about and experienced with feedback systems.
One of the most-fundamental ideas and goals in feedback control system theory is system stability and probably the most-fundamental goal is control, as in perfect tracking with "no overshoot", or with a known overshoot taken as a trade-off.
The desire to be able to have known control characteristics and known stability margins are the two main reasons that feedback control theory was even invented!
Can feedback be done "wrong", with bad results? Yes. Can a high-performance sports car be driven "wrong" so as to perform poorly? Yes. Same difference!
First, one would need to learn system theory (time-domain analysis with differential equations and then frequency-domain signal and system analysis with Fourier and Laplace transforms). Then they should be able to learn single-variable feedback control system theory.
You might be on to something, regarding problems with "time errors and disrupted phase relationships". But feedback should be the most-reliable and best-performing way to be able to ensure that those are NOT problems.
If you don't happen to prefer the "over-regulated" sound of negative feedback that's working "too effectively", which is only what I might guess your complaints stem from, then maybe you should try changing the feedback path's gain or phase response, or increasing the forward (non-feedback) gain of the system (or maybe other things I'm not remembering), to make the system LESS stable, and less accurate. Many people prefer the sound of a little more overshoot and a little more distortion. I, on the other hand, prefer the most-accurate reproduction of the source, EVEN if that means that I must listen to "hard, cold, lifeless, unmusical low-THD-sound".
Interestingly, it does not sound hard, cold, lifeless, nor unmusical, at all, to me; quite the contrary! And my soundstage is exquisite, and glorious!
Cheers,
Tom
P.S. What IS IT with "feedback", among some in the audio crowd?! Can you imagine a similar cult in the weapon systems crowd?
"Heck no! We don't want your eeevil feedback constraining our beautiful, oxygen-free air-to-air missile!"
"Did the missile hit the target?"
"Well no, but just look at it go!!!"
Last edited:
Seems like that would just mean that you need to pay more attention to using proper decoupling capacitors at the point of load. (It's practically an art, involving finding the right number of caps with the right values and the right ESR specs, and low instrinsic inductances, and getting them as close as possible to the power and ground pins.)
The idea of a "lazy current supply" from the main filter caps actually can be (and often is) a real problem, especially for the fast-transient portions of the current demands at the power devices' power supply input pins.
There can be problems due to the inductance of the supply conductors between the main caps and the point of load. You just can't get current through an inductance to start or stop faster than the inductance will allow, which is controlled by the laws of physics. So you need the proper decoupling capacitors RIGHT AT the point of load, to supply the fast transient portions of the currents that the power device wants to let flow at a particular moment.
As a nice side effect, the decoupling caps help prevent voltage ripple on the supply rails. If they weren't there, then when the device tried to pull fast-changing current transients through the inductance of the supply rail conductor, the usual differential equation for the voltage across an inductance would come into play, with the voltage being the inductance times the rate-of-change of the current.
Interestingly, that means that even small-amplitude but fast current changes can generate large voltages, since the amplitude of the resulting voltage depends only on the RATE-OF-CHANGE of the current, NOT on the current's amplitude, as would be the case with a resistance and Ohm's Law.
That's one reason why the power rails and ground returns in digital circuits are often full of voltage spikes. It's because the currents being pulled into and out of the chips' power pins are often pulse-shaped, usually with edge rise and fall times that are very short, i.e. with a large "rate-of-change", and therefore a large spike of voltage forms across the rail inductance whenever a current-pulse edge occurs at a chip's power or ground pin.
Sorry to have blathered-on for so long, about all of that.
Cheers,
Tom
Last edited:
Some of you seem to believe that music reproduction is better without feedback built into the system.
If you believe that about music reproduction, then surely you would also have to believe that music _production_ would be much better, if feedback were not involved, correct?
If feedback is "a bad thing", then you who believe that would have to also believe that the BEST-sounding, say, violin music, would be produced by a violinist with no feedback.
That would mean that the violinist would have to have perfect earplugs so he could not hear himself play, at all, and would have to be blindfolded so he could not see himself play, and would have to have his fingers, hands, arms, and body made completely numb with something like many lidocaine anesthetic injections, so he could not feel himself play.
Voila! Zero global feedback! Bravissimo!
The music might not be perfectly produced, but it would "sound better", right?
<grin>
My 2 cents in this respect
NFB is very good in theory, there is nothing to object against it.
But, in real practice, it is really hard to make it operating fast (with slew rate lets say 500V/usec in the FB loop), and definite compromises arise due to it.
Maybe, OpAmps are the most suitable field for correct using of NFB, since FB parts can be located most closely.
PCB design for deep NFB amp is really a tricky thing, stands at the top of the present days possibilities. Maybe some designers and companies are more lucky in that, like for instance MBL with MBL9007 amp.
Disigning without NFB is much more forgiving, and offers very much higher probability of getting goog sound (since effects of imperfectly operating NFB are most unpleasant).
Good PSRR, due to deep NFB, of an amp, also gives no warranties of good sound, due to the same reason (finite NFB slew rate and not perfect transient responce).
NFB is very good in theory, there is nothing to object against it.
But, in real practice, it is really hard to make it operating fast (with slew rate lets say 500V/usec in the FB loop), and definite compromises arise due to it.
Maybe, OpAmps are the most suitable field for correct using of NFB, since FB parts can be located most closely.
PCB design for deep NFB amp is really a tricky thing, stands at the top of the present days possibilities. Maybe some designers and companies are more lucky in that, like for instance MBL with MBL9007 amp.
Disigning without NFB is much more forgiving, and offers very much higher probability of getting goog sound (since effects of imperfectly operating NFB are most unpleasant).
Good PSRR, due to deep NFB, of an amp, also gives no warranties of good sound, due to the same reason (finite NFB slew rate and not perfect transient responce).
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.