Negative feedback Amp=more faithful reproduction?
OK, so I will go out on a limb and agree with Kijanki that educated application of NF = more faithful reproduction, in a pure technical sense, though as we all know, no solution is perfect and there are always potential drawbacks to deal with. In the case of NF, it would seem to be potentially higher IM distortion that just so happens to come into play in the most sensitive region of the human hearing spectrum where it matters most. That would up the ante to get it all done right! Now I will go back and do more listening to determine further for myself who has done it right and who has not. |
"Mapman: The answer to your question is to do away with reproduction altogether. Sell all your equipment and go to stadiums, concert halls, jazz clubs, auditoriums, amphitheaters, subways, bathrooms, etc. Go live!" Agreed! Live is the reference standard! Everything else is a reproduction all of which are flawed in some way. Choose your poison! Abstract art is regarded by many as the most enjoyable! Its also one of the most highly distorted forms of art compared to other styles. The art is represented by the distortions. A photo would work best to reduce the distortions. A three dimensional hi res CGI model perhaps could do the best! Which of these would you prefer? Does it matter? As long as it affects you, I'd say it realy does not matter. I like variety! I'll take one of each! Still the hi res CGI model is probably the most faithful technically, though still surely noticably imperfect. Give it time. Technology will continue to advance to the point where the remaining distortions are below the threshold of human perception. Sound reproduction is not as data intensive and will probably get there faster. I tend to think it may already be getting there in some cases these days. Unfortunately, audiophiles may always still be futilely sweating the details that really no longer matter, though I suspect the #s that obsess on doing such things will continue to dwindle significantly. |
Edwyun, Yes, we have to realize that when they talk about zero feedback they mean global feedback. Amplifier can have many local feedbacks and they might still call it ZNFB. Why? Because ZNFB became catch phrase and sells amplifiers. You stated the you would buy amplifier with zero output impedance, infinite bandwidth zero distortions etc. but I'm not sure everybody would. Some people would complain that it doesn't have good bass with their speakers (zero output impedance), others would call sound analytical and sterile (lack of distortion and zero noise) while some would call it bright (infinite bandwidth) since their system has tendency for it. In addition many people like warm sound (enhanced even harmonics) and would not be happy with this amplifier. Shallow global feedback, properly used, can reduce distortions, widen bandwidth and lower output impedance WITHOUT introducing Transient Intermodulation. As for getting rid of the equipment, I think you're right. For the money some people constantly spend on audio they could hire Symphony Orchestra to play for them. |
NFB is a subtractive process, although as proved by GE long ago, it increases the perceived odd order harmonic distortion. NFB removes some of the signal, and not in real time, either,due to propagation delay within the amp. Although I prefer to trust my ears, and could not care less about the theories involved, theory does match what I hear: with NFB it sounds less "alive". Deadened. Something essential has been deleted from the signal along with the improvement in distortion specs. There does sound like slightly less distortion, but the baby is thrown out with the bathwater. I personally think those who cite the "improved" specs from NFB are missing the point, but to each his own. Not so different from those who prefer SS over tubes, who often cite the "less distortion" specs. No, I don't enjoy listening to distortion, just unwilling to sacrifice aural immediacy to lower it. |
"Mapman, the question here is if NFB makes reproduction MORE faithful. Some answered that there is no faithful reproduction, which doesn't answer this question, while others stated they prefer sound of ZNFB which doesn't answer this question either. The truth is, that there is no SS amp (and very few tube amps) with zero feedback. SS amp by nature is a voltage source, and as such needs some form of negative feedback, at least for the output stage. The question is how much and that might depend on the load, your preference etc. It is undisputed fact, that even small amount of NFB widens bandwidth, lowers output impedance and reduces THD/IMD distortions. Does it make reproduction more faithful on average? I think it might. What sound you like has nothing to do with original question." Kijanki, you seem to be the only one who understand the intent of my original post. Thank you for putting into words better than I. I find it amusing some found my OP to be an attack on their equipment and taste in music, which was not my intent at all. If some of you were offended, sorry. |
Dracule1, What I'm getting from this is that some amount of imperfection can sound to many people more natural. Difference between natural, pleasing and faithful is fluid. Article was very interesting and thank you for posting it. I would only disagree with notion that deep feedback has no negative effects if amplifier is stable and input limited to slew rates equivalent to 20kHz. In ideal world it might be true but bandwidth never ends at perfectly 20kHz (TT often go to 50kHz) while high gain amplifiers, including op-amps have, (in spite of compensation) tendency to ring when presented with capacitive loads (speaker). It is big temptation for designer when he can improve everything 10-fold just by introducing more gain. I know that top designers like Nelson Pass or Jeff Rowland don't take shortcuts and that's why I bought Rowland class D amp. I know that modules were designed and made by B&O but Jeff Rowland name is enough for me to get interested. Some amplifiers are very expensive but you get what you pay for (Brand name). My amazing Hyperion HPS-938 speakers were a bargain but company doesn't exist anymore (so it seems). Now I worry what to do if something happens to these exotic drivers since nobody else makes them. Next time I'll pay more. |
Edwyun, this particular design (Icepower) uses Nielsen Karsten multiple feedback. One feedback comes from modulator while the other is from the speaker output. Class D is relatively linear to start with, while output impedance is inherently low hence it doesn't require a lot of feedback. In addition it is only one stage (modulator) making for small signal delay - less chances of TIM. Early class D consisted of primitive sawtooth modulator while this resembles sigma-delta A/D converter's modulator. There is nothing wrong with class D (SACD,DSD are class D) but carrier frequencies are still too low (Mosfets are getting faster every year). For the money it is bargain IMHO. |
Edwyun, this particular design (Icepower) uses Nielsen Karsten multiple feedback. One feedback comes from modulator while the other is from the speaker output. Class D is relatively linear to start with, while output impedance is inherently low hence it doesn't require a lot of feedback. In addition it is only one stage (modulator) making for small signal delay - less chances of TIM. Early class D consisted of primitive sawtooth modulator while this resembles sigma-delta A/D converter's modulator. There is nothing wrong with class D (SACD,DSD are class D) but carrier frequencies are still too low (Mosfets are getting faster every year). For the money it is bargain IMHO. |
There seem to be several things not yet discussed on this thread that seem IMO to be important. The first is that the ear interprets harmonic distortion as tonality. For example, the 2nd harmonic of some tube amps is why they sound 'warm' or 'lush'. Another example is the trace amount of odd ordered harmonics typical of a solid state amp with lots (+20db) of feedback; they will sound 'hard' or 'bright'. (BTW, the lower orders, 2nd, 3rd and 4th, are proven to not be irritating to the human ear. But not irritating simply means they are musical; IOW that is not to say that they are inaudible!) The second is that negative feedback is used to control the speaker in two ways: by altering the power output of the amp depending on the speaker impedance at that frequency, and by providing increased damping. So here are the problems with adding feedback to any amplifier. I did note BTW that this seems to be ignored in the article in the OP. In fact I've seen a lot of engineers hide behind theorems (Shannon, Nyquist, et. al.) to explain why this is not a problem, and that problem is the delay time between input and output of the amps. The reason the theorems don't work is they don't apply. They apply in a theoretical world where propagation delay in the amplifier circuit does not exist. But almost any oscilloscope can measure this delay time in almost any amplifier made. If we could build amps that had no delay (propagation) delay, then feedback would not cause the problems it does. However we have to live in the real world, and experience has shown that being pragmatic about that fact reaps big rewards. In this case all amps have propagation delay so the application of loop negative feedback will have the benefits that Kijanki has described, except for a tiny bit of odd-ordered harmonic that is considered 'negligible' by degreed engineers. This 'tiny bit' of distortion can be modelled as a 'ringing' effect in the amplifier. Chaos Theory has something to say about this, I will go into it if people are interested. It is the mark of a good engineer to know when to use the word 'negligible' correctly! In the case of 'negligible odd ordered harmonics added' the word is misapplied. The reason is the human ear is more sensitive to the presence of odd ordered harmonics than it is to almost anything else. This is because the ear uses odd ordered harmonics (present in any sound) to figure out how loud that sound is. If the odd orders are slightly louder due to electronics, the resulting sound will not only sound louder than it really is, it will also sound brighter. Its funny that an engineer might know how to design the feedback loop but be completely uninformed about how the ear will perceive the result! Really the only way to get around this is to get rid of the distortion. But as Nelson Pass points out in a distortion article on his site https://passlabs.com/articles/audio-distortion-and-feedback you can add more feedback to reduce that distortion but you need to add more gain (adding to the distortion) to do it. You are chasing a carrot on a stick. With regards to damping factor: Some speakers require high damping factor (up to 40:1) and some speakers want very little (0.1:1 and yes you read that right). No speaker made needs in excess of 40:1. The chief engineer of Electro Voice wrote a two-part article on this some decades ago and the facts of that article have not changed which is how it usually works with facts. I will try to dig up the article (its online) if anyone is interested. For me there has always been a bottom line, which is that if your speaker requires that the amp have feedback you won't be able to get it to sound like real music because the amp is violating one of the most fundamental rules of human hearing. How do you get around this if you don't want the system to have colorations (audiophiles talk about these colorations all the time; warm, rich, soft highs, wooly bass, etc)?? The fact of the matter is you have to do everything in the book that you can to reduce distortion without using feedback, and stick to speakers that don't expect a low output impedance from the amp (BTW the crossover design rules can be quite different!). A fully symmetrical design will, for example, cancel the even ordered harmonics (otherwise the 2nd will be predominant in any design, tube or transistor). The even orders will thus cancel not only in the output of the amp but in every proceeding stage. That leaves the odd orders, and if you are careful they will not be generated. There are plenty of other techniques BTW. In addition any ZNFB amp, IMO should have a very low IM distortion figure if its done right, because IM is very audible. That is very possible in a ZNFB design- we get numbers that are as low as many amps with feedback. IM distortion (or lack thereof) is a function of linearity. |
Kijanki, They are not bad looking. What I mean by very nice is that they appear to be well designed to really deliver the music well at lifelike volumes and with meat on the bones for very reasonable cost compared to many tower or monitor type designs common today in the same price range. Its been said that teh main reason to have an expensive stereo is the ability to go lifelike loud and clear as needed. I agree with that! THose speakers look up to the task! |
Mapman, Woofers have only one invention - lack of suspension (spiderweb) compensated by large diameter voice coil mounted to flat disc in place of dustcap. It makes membranes lighter, faster and stiffer. Midrange (6.5") uses, in addition ferrofluid as a suspension. Midrange is absolutely breath taking. Tweeter is so well integrated that I cannot really tell transition from midrange. http://hyperionsound.com/Images/HPS-938%20review.pdf |
Correct! IMO the damping thing is way over-blown. I found that article: http://paulspeltz.com/tomcik/index.html |
Atmasphere, I've seen you write about how bad odd order harmonics can sound (ie, subjectively bright and hard) even at very low levels. Can you please specify at what level are odd order harmonics perceived as unpleasant? I don't recall you ever giving an actual number. I have owned your M60 amp which I think uses a 2 dB of NFB and compared to a pentode, class AB, tube amp with 20 dB of NFB. Everything else in my system was constant. The volume level was as close to even as I can get them with SPL meter, and the volume ranged between 85 to 90 dB max during the comparison. The speaker had minimum narrow impedance band of 5.8 ohms and was 90 dB efficient. By your own words, 20 dB of NFB is considered excessive. However, I did not hear any excessive hardness or brightness in either amp. Both were very smooth in the midrange and treble regions. How do you explain that? |
Atmasphere, could you please clarify a statement you made on another related thread? "Marqmike, unfortunately all amplifiers make odd ordered harmonics. It is the 5th, 7th and 9th that we are concerned about (the 3rd is considered musical to the human ear). When you add negative feedback to any amplifier design, the result will always be an enhancement of odd orders, even if the feedback is successful in removing most of the open loop distortions ('open loop' means the amp operating without feedback). The reason this is so has to do with the fixed amount of time it takes a signal to move through a circuit. This time is called 'propagation delay'. The issue is that a low frequencies feedback works pretty good, but as frequency is increased, the propagation delay means that the feedback signal is arriving later and later to correct the circuit. In fact a designer has to be careful with feedback because this effect means that at some frequency the feedback as actually positive rather than negative. This is entirely due to the propagation delay in the amp." I find your statement misleading. You state that NFB "enhances" odd order harmonics. According to the article I cited, NFB does not enhance (ie, increase) or add odd order harmonics. NFB reveals the odd order harmonics already produced by the amplifier before NFB is applied; this is a result of NFB greatly reducing other distortion components that mask the odd order harmonics. And NFB can reduce those higher order odd harmonics to below 110 dB from the fundamental. Can one hear distortion that low? You've claimed that we are very sensitive to odd order harmonic distortion, but could you provide a study that shows we can detect distortion this low? As for propagation delay, an amplifier operated in its linear range will have delay in nanosecond range as far as audio frequency is concerned, including frequencies at the limit of human hearing. This is also explained in the link. Nano second range delay seems pretty negligible to me. I, obviously, do not have more technical knowledge in amplifier design than you, but your claims seem to be directly at odds with the article I cited. |
