thoglette Posted October 21, 2014 Posted October 21, 2014 Agree to some degree that THD can be overly simple -- but the audible thresholds were established using common distortion patterns, so they are relevant and usable. The 2003 Geddes/Lee paper looks like a revolution at first glance (ignoring that fact that Shorter of BBC proposed almost identical harmonic weightings in 1950), Agree to this point. And then you lose me completely. The Gm score will be ignored for exactly the same reason Shorter was ignored - there's no money to be made from it. THD and THX - now those we can sell. Oh, and watts. However, your post underlines my point about the compromises we are willing to take (within our budgets) to get closest to our version of Nirvana - you have a completely different set of "hot buttons" to me, w.r.t. audio. eg. Clipping is a regular part of my diet. Avoiding it would take about two more zeros than my budget has.
Newman Posted October 21, 2014 Posted October 21, 2014 (edited) Where did I lose you? Where I said that Gedlee generated unnatural (to hifi components) harmonic distortion patterns to do the research? Where I said that we have no evidence that any real-life components have high Gm scores? Where I said that inaudible distortion is a good thing [edit: and an achievable thing], no matter what its 'pattern'? Those were my main points after the point where I lost you. Edited October 21, 2014 by Newman
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 Er, no, Floyd Toole's and Sean Olive's research (both of whom were employed by Harman after conducting there research at the NRC) show that a speaker system with a smoothly declining power response that mimics real acoustic sources is an important aspect in preferences for loudspeakers. Dipole speakers do not readily meat this criteria at all. I don't understand how a dipole doesn't meet this requirement. eg. http://www.linkwitzlab.com/frontiers_7.htm#C2 .... and indeed if you want to extend the "smooth decline in power response" ..... down to a lower frequency .... then I don't know how you can do this without a horn, or a dipole (cardioid), etc. A quite enlightening experience for me has been to setup a number of different speakers in the same 8m x 5m room (as difficult as that is without the room being a problem) Box with smoothly declining power response above 1khz (6" + horn) Box with wide flat(er) horisontal power response (tall ribbon tweeter) (acoustically narrow) Dipole which keeps relatively high and constant directivity to < 300hz, and a smoothly declining power response as frequency increases Horn with some coverage control down to ~300hz, and a smoothly declining power response as frequency increases I expected the differences I heard.... but I was not quite prepared to discover the large similarities between the two speakers with kept an even radiation pattern down low. The horn is 110dB single point of radiation .... the dipole is << 90dB, 1.2m high 4-way, and has the rearward radiation which the horn doesn't. They are quite different in a number of ways that people consider important .... but they sound very similar. ... but more importantly, they sound staggeringly good. They sound like someone has removed a blanket from over the speakers (yep, that ol cliche).... No, wait. 3 blankets. IME... This is what taking the research of Toole, Olive, others to extreme lengths does (and admittedly impractical lengths for most domestic hifi). LOL. I'm starting to sound like Geddes... but the more I experiment, the less I think he's crazy If you enjoy your stereo system, then that is absolutely great! Some people liken Hi Fi to the fashion industry - all a matter of personal taste. It's not something you can argue with, eh?! .... in the end it's ultimately true (there is no accounting for taste). Like I said earlier (anyone can be forgiven though, as it was buried inside a wall o text) .... people study this, and they find that punters really do agree that "better is better" .... Preference as such really only comes into it when we're talking about choosing between compromises. It really should make logical sense, that if I (really do) remove an error from the playback system ... then you're hearing more of the recording .... and this is good. It kinda goes to another pervasive misnomer IMHO .... "we don't want more of the recording, because some recordings are bad". I think this one actually gels right in with what I'm discussing here. "Modern recordings" ... you know, that bad sounding ones.... generally have a lot of energy in the lower mids and down. There's 4 guitars and a man screaming... and then in the mixing session we throw 20 other tracks over the top too. This is quite a workout for the 200-400hz range where no speaker other than the ones which go to extreme lengths to control their coverage pattern down to these low frequencies can get the direct to reflected balance right. IMHO all the research and experience which says that direct to reflected balance is the key to speech intelligibility .... is also the reason why modern (busy, dense, compressed) recordings sound as bad as they do.... because typical speakers have become omni right where they need not to be. I've found that the biggest improvement experimenting with speakers which control directivity to lower frequencies has been these densely packed modern recordings. They've gone from sounding like mud, to actually having some clarity and depth.
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 Dave, your definition is wrong. Fundamentally wrong. There are forms of distortion we are much more sensitive to than others. (eg. 5th, 7th, 9th , harmonics) and as our friends in MP3 land have found out, our ears "mask" other sounds out quite effectively. Geddes and Lee (2003) describe this and test it quite convincingly. Their work is just the latest in a long line of research which has been continuously and conveniently ignored since 1938 when, and I'll paraphase (as I don't have the quote with me), Masa noted that we should stop using THD as a measure of performance. I'll go further and join some dots - the various "all amps sound the same" challenges I've seen all insist that the amplifiers don't go anywhere near clipping. Ergo clipping behaviour is actually very important (see Crowhurst 1957 JEAS 5(4) or Hamm 1973 JEAS 21(4) ) Accuracy is important. But not all accuracy is equally important to our ears. I agree with everything you say here, except the first sentence. When I say distorting the input vs output .... I am largely talking about frequency response. On Geddes... I don't think you are giving the full story (unless I have slightly misunderstood you , which is a distinct possibility). 5 years ago we probably could have got him over here to comment, alas. Geddes position on non-linear distortion (ie. harmonic distortion) is this. When kept below certain levels it is completely irrelevant. The reason he suggests not using THD as a measure.... is because THD does not communicate "certain level" very well. 10% THD of mostly 2nd .... is a completely different proposition to 10% THD which is mostly 7th. If we go with Geddes (as we should IMHO) that non-linear distortion (harmonic, intermodulation) is a non issue (when kept to a certain level) ..... then there is only distortion of the frequency response left.... and that is what my ramblings are all about. Sample the sound coming from a typical speaker 60 degrees of to the side.... There is intense distortion of the frequency response. If we go back to the complexities of human hearing .... which say that the reflected sound is integrated with the direct sound, into a single "event" in your brain, based on its similarity to the direct sound ..... we can see that these reflections are not going to be integrated by our ears due to the enormous difference they have to the direct sound. Hah... See?!? All that wall-o-text and I finally made it back to the thread title. 1
hochopeper Posted October 21, 2014 Posted October 21, 2014 (edited) An interesting extract from David Griesinger's page: Fudamentally the ear/brain system needs to 1.separate one or more sounds of interest from a complex and noisy sound field, and 2. to identify the pitch, direction, distance, and timbre (and thus the meaning) of the information in each of the separated sound streams. Previous research into acoustic quality has mostly ignored the problem of sound stream separation - the fundamental process by which we can consciously or unconsciously select one or more of a potentially large number of people talking at the same time (the cocktail party effect) or multiple musical lines in a concert. In the absence of separation multiple talkers become babble. Music is more forgiving. Harmony and dynamics are preserved, but much of the complexity (and the ability to engage our interest) is lost. Previous acoustic research has focused on how we perceive a single sound source under various acoustic conditions. Previous research has also concentrated primarily on how sound decays in rooms - on how notes and syllables end. But sounds of interest to both humans and animals pack most of the information they contain in the onset of syllables or notes. It is as if we have been studying the tails of animals rather than their heads. He's talking about a measurement that allows us to characterise the quality of an acoustic space for performance, we certainly know which aspects are of importance but he's taking the next step from what I can tell. Driven by the desire to improve performance hall design. I think there is some application here for audio reproduction. His papers/presentations are of interest I reckon and supports Dave's assertion regarding importance of direct/reflected ratio and the envelope of the reflected energy rather than the magnitude of a single reflection. Edited October 21, 2014 by hochopeper 2
Newman Posted October 21, 2014 Posted October 21, 2014 Geddes position on non-linear distortion (ie. harmonic distortion) is this. When kept below certain levels it is completely irrelevant. A bit like my posts #25 and #27, incomprehensible though they be. 1
Guest Posted October 21, 2014 Posted October 21, 2014 (edited) I don't understand how a dipole doesn't meet this requirement. eg. http://www.linkwitzlab.com/frontiers_7.htm#C2 I would say that the power response discontinuities in the L-07 are clearly evident, being ripples of ~3dB: Almost all loudspeakers have these power response ripples and it is one reason why almost all loudspeakers reveal their Gestalt, to use Linkwitz's term. (The ripples are in fact a function of the physics of high order crossovers - it is not possible to derive a crossover solution that exhibits flat amplitude response and flat power response simultaneously for crossover orders of 2 and above, and that's before adding real world drivers into the equation.) So I probably haven't answered the question properly. But I remain convinced that a dipole is not a good mimic for real world acoustics sources (i.e. voices, instruments, etc) that tend to have a power response that matches more closely to that which would be exhibited by an idealised monopole. (For example, most singers do not have a hole in the back of their neck to radiate sibilance towards the wall behind them.) There are monopole designs that exhibit much flatter power responses than Linkwitz's examples, but they are very rare, though not very exotic, Duntech Classic models, almost all Dynaudio consumer models, but not Dynaudio professional monitors, for example. How do they achieve it? Broadband drivers that have well defined response a couple of octaves either side of the intended passband and first order crossovers are mandatory. It is interesting that Linwitz has played with KEF 107s - they would have to have one of the "horrible-est" room responses for a product purporting to be high fidelity that I have experienced. If Linkwitz has these as a reference, I am not surprised by his views that dipoles are better! It reminds me of a subwoofer that was developed locally. The manufacturer asked for, and was supplied with, a couple of the "typical" subwoofers from a friendly retail store and went to market confident that he had produced a better product, oblivious to the fact that the random samples they were compared to were probably the worst performing products at their price point at the time! LFE-O Dear Me! Edited October 21, 2014 by Guest
hochopeper Posted October 21, 2014 Posted October 21, 2014 I got lost when you made the assumption that the reasoning for even power response in speakers is to mimic the power response of a instrument/voice/sound. That doesn't compute.
ophool Posted October 21, 2014 Posted October 21, 2014 Don't sweat it Chris, my whole house of cards crashed to the floor when I read that listening tests and subjective comparisons are basically useless. Seems I have been using the wrong methodology to improve my listening experience and gradually "upgrade" my system. I had thought that it sounded much better than it did 7 years ago, but now how would I know ? Oh the horror and shame of self delusion - to say nothing of the needless expense. :ph34r:
Guest Posted October 21, 2014 Posted October 21, 2014 (edited) I got lost when you made the assumption that the reasoning for even power response in speakers is to mimic the power response of a instrument/voice/sound. That doesn't compute. I haven't made that assumption. You have read into my statement something unintended. Generally all natural acoustical sources have a power response that declines with frequency. For a loudspeaker to do a half decent job of mimicking real acoustical sources it has to have a power response that falls somewhere inside the range of natural acoustical sources. Edited October 21, 2014 by Guest
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 For example, most singers do not have a hole in the back of their neck to radiate sibilance towards the wall behind them. Quite a pertinent line of discussion for this thread. The rear wave (assuming the speaker is placed far enough away from boundaries) is delayed far enough that the sound is outside the window for integration by our hearing.... meaning it doesn't add to the perception of the "sound event" .... only to the actual perception of the 'reverberant' sound. This means that that aside from affecting the later reverberant sound (which all environments have) .... it's only effect is how it manifests on the frequency response. I think it's extremely telling that a dipole and horn both sound extremely different to other speakers.... and extremely similar (in many ways) to each other. There are monopole designs that exhibit much flatter power responses than Linkwitz's examples, but they are very rare, though not very exotic, Duntech Classic models, almost all Dynaudio consumer models, but not Dynaudio professional monitors, for example. How do they achieve it? Broadband drivers that have well defined response a couple of octaves either side of the intended passband and first order crossovers are mandatory. That is what the 6" + waveguide in my example is.... 1st order crossover ..... although the crossover transitions quickly to >4th order outside the passbands. Similar to this: http://customanalogue.com/elsinore/elsinore_17.htm ... but it doesn't sound anything like the speakers which control their pattern down to a much lower frequency. I would say that the power response discontinuities in the L-07 are clearly evident, being ripples of ~3dB Sure. In hindsight not the best choice. I plan to try and take some better polars of mine this summer (in the lower midrange which is difficult). They use LR4 filters right now, but can very potentially use lower orders.
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 Don't sweat it Chris, my whole house of cards crashed to the floor when I read that listening tests and subjective comparisons are basically useless. I wish it wasn't so ..... but the more you try to rely on your ears (and work with what they tell you), the more you will find they are not very good tools. When we are faced with two compromises in a playback system .... and we need to pick one. Then most certainly 'preference' comes into it. However if you are attempting to "quantify" something .... perhaps to determine what compromises exist in a system .... perhaps to determine if what you just did actually helped or not ..... Then our ears and brains are useless. They just say "oh that sounds nice" .... "oh that doesn't agree with me".... This isn't actually helpful when you're trying to work something out .... or to get something that will apply over all program material.
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 An interesting extract from David Griesinger's page: He's talking about a measurement that allows us to characterise the quality of an acoustic space for performance, we certainly know which aspects are of importance but he's taking the next step from what I can tell. Driven by the desire to improve performance hall design. I think there is some application here for audio reproduction. His papers/presentations are of interest I reckon and supports Dave's assertion regarding importance of direct/reflected ratio and the envelope of the reflected energy rather than the magnitude of a single reflection. Very much so. Reflected sound after a certain time (late) is not particularly important as long as it is below a certain level (this is easy to achieve unless you're in a small very reflective room, and/or you play very loud) Reflected sound before a certain time needs to be both below a certain level (which seems to be fairly low, so quite difficult for wide coverage speaker in small rooms) ... and needs to have the same 'shape' (envelope) as the direct sound --- otherwise it isn't taken as part of the original event
Newman Posted October 21, 2014 Posted October 21, 2014 my whole house of cards crashed to the floor when I read that listening tests and subjective comparisons are basically useless. Well, you didn't read it in this thread..... In fact, the research that has been mentioned in this thread relies mainly on 'listening tests and subjective comparisons' to advance our knowledge of audio. So it's the right thing to do. If done right.
LHC Posted October 21, 2014 Posted October 21, 2014 I wish it wasn't so ..... but the more you try to rely on your ears (and work with what they tell you), the more you will find they are not very good tools. When we are faced with two compromises in a playback system .... and we need to pick one. Then most certainly 'preference' comes into it. However if you are attempting to "quantify" something .... perhaps to determine what compromises exist in a system .... perhaps to determine if what you just did actually helped or not ..... Then our ears and brains are useless. They just say "oh that sounds nice" .... "oh that doesn't agree with me".... This isn't actually helpful when you're trying to work something out .... or to get something that will apply over all program material. Well, you didn't read it in this thread..... In fact, the research that has been mentioned in this thread relies mainly on 'listening tests and subjective comparisons' to advance our knowledge of audio. So it's the right thing to do. If done right. One is saying our ears and brains are useless for determining what helps a system; the other said subjective listening will advance our knowledge of audio. Hum, something doesn't compute here. The reconciliation is that 'done right' refers to a carefully selected group of test listeners, those that don't have hearing losses, and are trained to listen out for certain aspects in the test. That is what they've used to achieved a controlled testing regime giving consistent and repeatable results. It also means that the results are not necessarily applicable to people who are not qualified to be tested.
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 Well, you didn't read it in this thread..... In fact, the research that has been mentioned in this thread relies mainly on 'listening tests and subjective comparisons' to advance our knowledge of audio. So it's the right thing to do. If done right. There's two points there people might miss. They're difficult (and more so tedious) to do 'right' ... and unless you have a way to quantify what your listening test was examining (ie. you've measured the speaker) .... then you can get very lost. 3
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 The reconciliation is that 'done right' refers to a carefully selected group of test listeners, those that don't have hearing losses, and are trained to listen out for certain aspects in the test. Done right is more: Same room Same listener location Same speaker location Very short time between different speakers Listening test are just fine if you want to answer the question "do I like this?" ..... If you want to draw some conclusions you need to control variables -- cos the ones above have a huge effect. 1
Maitlynd Posted October 21, 2014 Posted October 21, 2014 The reconciliation is that 'done right' refers to a carefully selected group of test listeners, those that don't have hearing losses, and are trained to listen out for certain aspects in the test. That is what they've used to achieved a controlled testing regime giving consistent and repeatable results. It also means that the results are not necessarily applicable to people who are not qualified to be tested. What you are hinting at is untrue, that the tests are 'rigged'. The rationale for using trained listeners is very pragmatic, it saves a lot of time and therefore money and effort. They have crosschecked the results between trained an untrained listeners, and they are very close. Therefore, it is much quicker and easier to use trained listeners that produce the same results in a fraction of the time, simply because they are trained. THAT the two groups produce the same results is further confidence that the results found are applicable to the wider group, *us*. WHY exactly, are you as someone who has not participated or not seem to know much about it, so seemingly desperate to dismiss the findings? It is pretty simple, and (to my mind at least) very understandable, that we all converge towards agreement as the system approaches **perfection**. Is that so wacky? A lot of people rabbit on about 'you can never reproduce the original performance' (ie a criticism), yet when the science shows we do in fact prefer the system that better reproduces the original performance it is all of a sudden incomprehensible? 1
LHC Posted October 21, 2014 Posted October 21, 2014 What you are hinting at is untrue, that the tests are 'rigged'. The rationale for using trained listeners is very pragmatic, it saves a lot of time and therefore money and effort. They have crosschecked the results between trained an untrained listeners, and they are very close. Therefore, it is much quicker and easier to use trained listeners that produce the same results in a fraction of the time, simply because they are trained. THAT the two groups produce the same results is further confidence that the results found are applicable to the wider group, *us*. WHY exactly, are you as someone who has not participated or not seem to know much about it, so seemingly desperate to dismiss the findings? It is pretty simple, and (to my mind at least) very understandable, that we all converge towards agreement as the system approaches **perfection**. Is that so wacky? A lot of people rabbit on about 'you can never reproduce the original performance' (ie a criticism), yet when the science shows we do in fact prefer the system that better reproduces the original performance it is all of a sudden incomprehensible? I did not suggest the result is rigged or dismiss them. You are drawing the wrong conclusion my friend. What I am suggesting is that they are trading off wide applicability for improved rigour. Yes, you get better testing results, but the range of people that fits those tighter criteria are reduced. If you look at the results reported in that paper, there is significant differences in the result from the trained and untrained listeners. Also I don't recall in the research that suggests that they asked the listeners whether what they have heard resembles the original performance. Presumably none of the test subjects would have been there at the recording to know precisely what the original performance sounded like. They were asked if they subjectively liked one option over another, it was a comparative testing.
LHC Posted October 21, 2014 Posted October 21, 2014 I strongly suggests people do read fully the article posted by John in #5, it is very interesting and informative. It is completely wrong to think that I am against the specifics of the reported science and its findings, I am not. My issue is with other aspects of the article that I have written about earlier and I stand by those. I am also a bit concerned by selective references to the article that (in isolation) have the potential to be misleading. So in the interests of good science communication I urge everyone to read the whole article and the references cited within.
ophool Posted October 21, 2014 Posted October 21, 2014 (edited) Well, you didn't read it in this thread..... refer to post # 3 It is the fundamental reason why "listening tests" or "subjective comparison" is basically useless. Edited October 21, 2014 by ophool
Guest Muon Posted October 21, 2014 Posted October 21, 2014 (edited) refer to post # 3 Some of us will stick with using ears Others can use a mic..........(Que Seinfeld voice) Not that there's anything wrong with that. Edit: I need to proof read sooner than two days later Edited October 22, 2014 by ortofun
Guest Posted October 21, 2014 Posted October 21, 2014 (edited) johnmath, on 21 Oct 2014 - 4:57 PM, said: For example, most singers do not have a hole in the back of their neck to radiate sibilance towards the wall behind them. Quite a pertinent line of discussion for this thread. The rear wave (assuming the speaker is placed far enough away from boundaries) is delayed far enough that the sound is outside the window for integration by our hearing.... meaning it doesn't add to the perception of the "sound event" .... only to the actual perception of the 'reverberant' sound. This means that that aside from affecting the later reverberant sound (which all environments have) .... it's only effect is how it manifests on the frequency response. Just a point - in a "normal" domestic listening environment the first order and often second and even higher order reflections all fall within the window for integration, so the power response / directivity of a speaker has a huge impact on the "direct" sound, plus of course the speaker's power response is completely superimposed/evident in the room's reverberation signature. Edited October 21, 2014 by Guest
davewantsmoore Posted October 21, 2014 Posted October 21, 2014 Yes - why the dipole need to place far enough away from the wall behind it .... or are you proposing that this (for reasonable distances) is still too short? For sure it can be a problem.
Recommended Posts