snip))) So... is it possible that GedLee place higher preference on localization and SL on spaciousness?
snip))
I'll say this, my attempt to perform Dr. Geddes's 'system' resulted in the most spacious presentation I've ever heard from stereo to date. I think he's more looking at the full monte. My attempt at SLs was less spacious but made a few instruments sound great.
Dan
Last edited:
More than possible, probably highly likely, but its more complicated than that. For a given room there is a tradeoff between imaging and spaciousness, we have pretty well concluded that, but when the room IS not a given then this tradeoff is different. I have tried to make this point many times, but it never seems to get through.
The temporal aspects of spaciousness are not very great, in other words, I will get good spaciousness from numerous lateral reflections for early times as well as late times. But good imaging requires a clean very early time, it is not flexible at all as to when these reflections occur, they have to be delayed as much as possible.
The solution should thus be apparent. If I limit the early reflections in the room through the loudspeaker and room design, but then have a very lively room, I can achieve both good imaging and good spatiousness. Unfortunately this is exactly the opposite of what most people do and hence they find themselves in the spacious versus image tradeoff. All studies of this relationship have dealt with a fixed room and one of "normal" design, which is usually too dead to work as I suggest.
My rooms have only a single absorptive surface (behind the speakers) the rest are rock hard - in fact one wall is stone. There is a ceiling diffuser - no absorption - but a single 4" thick pad to null the floor bounce. Wide directivity speakers sound terrible in a room like this becasue it amplifies all of their flaws (poor polar response, many early reflections, etc.). But properly designed narrow directivity speakers have superb imaging with great spaciousness - no tradeoff.
So while I do not discount the results of the studies, nor the comment about the tradeoffs in "typical rooms" that situation is not necessary, only quite typical.
Last edited:
how does one tame the bass in such a room? I know under 100hz or so you can add in multiple spaced subs but the ~two octaves above that tend to be a mess in such rooms with monopoles unless you've got heavy bass trapping...
Or is it an evil you choose to live with given the other advantages and cost? Do you have any waterfalls of the room?
Last edited:
More details: my rooms are highly absorptive at low frequencies, just not at HF. The walls themselves - at least one of every pair - are on springs and mounted with Contrained Layer Damping (CLD). At LFs they are extremely dead, but this goes away at HFs. I am sure that you will recognize that this is almost the exact opposite of what most rooms do. Certainly what any typical living room does or rooms that use "bass traps" for example. None of these devices do much of anything at LFs, they are simply not big enough, but they do all too much at HFs. Only an entire wall has enough surface area to actually yield any significant damping of the very lowest modes. Then use multiple subs and you have the ideal situation.
Makes plenty of sense for a dedicated HT, should I ever build one. I wish I had a chance to hear one of your systems. Abbey/Summa's up there with the Nao Note and DEQX-based stuff as one of the few expensive audio pieces I'd want to own. I'm somewhat turned off by the idea that the orion needed a "listening-based tweak". It usually implies to me something is wrong with the speaker design.. kind of like the infamous "BBC Dip".
I don't know who said it, or where, but a person implied to me once that even the best speaker designers agree that you can build a speaker that measures well, but once you hear it you'll still have to invariably "tune by ear". I... found it royally offensive. Our ears can discover flawed logic evidenced in the measurements, but I'm a person who really wants to build a speaker designed by science, not "ear-tuning"...
A part of me wants to buy a harper kit to use a center channel in my broke-man's "won't be upgraded for a while" crappy HT 😀 But I think it's still way out of my budget.
I don't know who said it, or where, but a person implied to me once that even the best speaker designers agree that you can build a speaker that measures well, but once you hear it you'll still have to invariably "tune by ear". I... found it royally offensive. Our ears can discover flawed logic evidenced in the measurements, but I'm a person who really wants to build a speaker designed by science, not "ear-tuning"...
A part of me wants to buy a harper kit to use a center channel in my broke-man's "won't be upgraded for a while" crappy HT 😀 But I think it's still way out of my budget.
Dr, Geddes, my question would be "why only 1/2 of the wall pairs?" Are both sides worse? It takes a lot of space to kill the bass. To build the wall for diaphragmatic bass trapping seems ideal for all the typical room issues, but something that requires major renovation.
If just 1/2 is best, should it be the inside, or outside walls? More importantly, will less noise escape through the vibrating wall or the stationary?
Thanks,
Dan
If just 1/2 is best, should it be the inside, or outside walls? More importantly, will less noise escape through the vibrating wall or the stationary?
Thanks,
Dan
... surely the first reflections will be correlated unless
the room is heavily armed with broadband diffusers and/or
a speaker is used, which has a noticeably decorrelated
off axis radiation.
The reason why usually many reflections in the room are
needed until sufficient decorrelaton is achieved, is diffusivity
introduced with the first few reflections being too low.
Btw. a speaker radiating off axis in a diffuse manner
does not only introduce considerably higher decorrelation
"from the start", even with the first reflections, it does
not even have serious problems with edge diffraction of
the cabinet or baffle.
Last edited:
Or that our understanding of how wiggly lines on a graph translate into perceived sound quality in a complex acoustic environment is far from complete. Or that we're not measuring the right things. Or both. 😉
I'll assume you're referring to listening based tweaks of on axis frequency response via crossover network / EQ, which on the Orion is what started this thread.
This thread is replete with reasons why Flat on-axis anechoic is a good starting place as a reference, but isn't exactly optimal for any given speaker design, and that there is also no "generic curve" away from flat that can be applied to any random speaker to give the best result.
Without repeating tons of stuff from this thread let me give just one example. Take several different speakers each of a different design topology, baffle size, number and diameter of drivers, crossover frequencies, driver heights (from the floor) and so on, EQ them until they all measure flat on axis anechoic.
Put them in a living room and listen to them. Not one of those speakers will sound the same, even just considering tonal balance, let alone other characteristics like distortion and dynamics.
Which one is correct ? Errr........... None of them ? 🙂 If they all measure flat on axis but all sound different in a normal acoustic space they all must be flawed, right ?
A lot of the difference is directivity vs frequency, but you also have proximity of each driver to the floor to consider which can't be ignored, yet isn't measured by an anechoic response.
By the way, the "BBC dip" at least as it's usually portrayed, is a bit of an urban legend. When you start digging into the background of it you find there was no deliberate design choice to introduce a dip at that frequency, nor did most of the BBC monitor speakers have such a dip.
You might find it offensive, but it's true, and I doubt many in this thread who has ever built a good speaker would disagree.
Why is it so surprising ? The whole point of designing a speaker is to create a device which provides a musically satisfying and realistic sounding reproduction (or illusion) of a recorded event for your ears, not to produce a device which when measured by a microphone in a very specific set of non-typical circumstances generates a flat line on a piece of paper or computer screen.
Looking at graphs of frequency response, phase, group delay, CSD, THD & IM, DI, and so on through your eyes is not very informative in telling you how all of those various factors mixed together will sound.
Your ears will tell you within a few minutes of listening everything you need to know about whether a speaker sounds good or not, looking at graphs will help unravel some of the reasons why it might sound good or bad, and some possible remedies, but will not on their own without hearing it predict whether it will sound "ok" or outstanding.
Each parameter can be measured and studied in isolation and thresholds of audibility determined (for example for harmonic distortion) but when you have a mass of different potential flaws of varying degrees present at once (eg frequency response, phase, group delay, CSD, THD & IM, DI mentioned above) which ones are more important.
Nobody would argue for example that higher distortion is better, clearly lower is better, but how important is it, how low does it have to be before it doesn't matter any more, how important is it compared to other parameters which will inevitably be sacrificed to optimize distortion beyond a certain point ? There are no black and white answers for some of these things, which is why there is such a huge variety in speaker designs.
If the design goal is solely an engineering exercise to produce a speaker which meets certain predefined measurements, that's fine, that's your prerogative, but don't expect to end up with a great sounding speaker.
In any real speaker optimizing a particular parameter beyond a certain point almost always degrades other parameters due to fundamental engineering tradeoffs, so a fixation on one parameter such as absolutely dead flat frequency response (measured how?) is bound to lead to a speaker which is deficient in other areas, and in doing so just doesn't sound good.
Only with a good balance between all the competing and conflicting parameters of the design, and some manual tweaking will a great sounding speaker result, and weighing up how important all those conflicting parameters are and how much each should be optimized is far harder than it appears, and everyone has a different opinion on the matter, or we wouldn't have such radically different, often polar opposite design approaches...
Last edited:
The question is what level of decorrelation is necessary and how would a speaker design need to look like to meet the goal? Bose 901?, dipole? From what I read on Linkwitz's site he's in the process of correlating first reflections (by adding a rear tweeter) instead of decorrelating them.
I guess SL's goal with introducing the rear tweeter
is having uniform dispersion, avoiding discontinouus
directivity with frequency and thereby harmonizing
inroom power response and on axis response.
Correlation/decorrelation is a different kind of thing.
You can create two versions of a signal, which sound
virtually the same, have the same spectral fingerprint
but are nevertheless decorrelated.
A panel form bending wave transducer has rather
broadband uniform dispersion, nevertheless the off
axis radiation is decorrelated with the on axis radiation.
That kind of phase decorrelation is not achievable using
pistonic acting drivers.
If you have e.g. 2 uncorrelated sound sources , they do
not interfere with each other in the way correlated
sources do.
In that same way, interference (combing) with walls
in proximity to the speaker is reduced with a well
made panel form bending wave loudspeaker.
Pictures 11 and 12 illustrate that effect in the following paper.
http://www.wvier.de/texte/NXT_Tonmeistertagung.pdf
On pages 14 f. a setting for interference measurement of
a speaker with a wall in proximity is shown (picture 11).
Picture 12 (top) shows the interference effects of the
arrangement in using a pistonic speaker, and the bottom graph
shows the effects for a distributed mode loudspeaker.
I never found an english version of that paper containing
comparable pictures, nevertheless the sketches and
measurements are quite self explaining.
Last edited:
I know but making reflections "look" the same automatically correlates them. Probably not a good idea.
That sounds pretty blurry to me. When reading through the Kendall papers, decorrelation needs to have certain very well defined properties to be effective. How is this achieved with a "panel form bending wave transducer"?
My setup looks more or less exactly the same but there's one caveat. Any strong single reflection grows in importance. For example, my speakers are toed-in by 45° which creates 2 strong contralateral reflections at around 10 and 12ms. These reflections have a pronounced effect on localization although they are about 10dB down.
Sorry Markus, i cannot avoid that "blurryness" without
referring to a couple of "starter papers".
Since i do not want to hijack this thread i may send
you some links via PM if you are interested ... just
say a word.
Basically a panel of high modal density has a virtually
random distribution of local velocity. Only the average
velocity of such a speaker's membrane is virtually
constant with frequency. The higher the modal density
of the membrane is, the more precise that 'constant
average velocity over area with frequency' will be
approached (and the more 'random' velocity distribution
over area will be).
But to go back to the thread's title:
If a speaker has directivity only smoothly varying with
frequency, the room's walls have no pronounced
selective absorption and combing can me mitigated
by introducing diffusivity - however it may be achieved -
i have not found a clue why 'flat' should not be correct ...
If a pronounced deviation from flat is desirable from
audible impression, there might be a serious flaw in
the mentioned prerequisites IMO.
Sure, thanks.
Harman's "preference listening tests" suggests just that.
I wondered who would ask that question. In one of my books I did a calculation of the damping coefficient due to an absorptive wall. At LFs, this damping was independent of how much absorption was on which wall, it was only the sum that mattered. This means that a lot of absorption on one wall is indistinguishable from half this same amount on two opposing walls. Since making these walls, and particularly the ceiling (you can't even do the floor) is very difficult it is easiest just to do one of each opposing set.
I'm sorry, but this is completely incorrect. The off-axis radiation is perfectly correlated with the on-axis, and so your comment about the diffraction is also incorrect.
I, for one, do not tweak any of my designs by "listening". It's an absurd practice that requires a real arrogance on the part of the designer to believe that his ears are better than his measurements. This would only be true for designers who were not competent at making the measurements.
The high modal density of a panel bending wave device does cause the radiation to be less correlated than a piston source. But this is NOT what we want. We want a highly correlated and coherent direct sound, just not a correalted reverb field. When the direct field is decorrelated between two sources (say stereo), then all imaging information in the recording is lost. There is no image possible.
The wavefront from a waveguide is extremely coherent across its entire extent. This is why there can be such a pronounced dip on axis. With other horn designs this "hole" disappears simply because the wavefront is not highly coherent. Axial cancellation requires a very precise timing of the diffraction and this can only occur when the wavefront is highly coherent without any inherent phase or amplitude anomalies.
But had they not been contra-lateral, but co-lateral would they have not been significantly worse? You cannot get rid of all early reflections in a small room without making it dead, but you can optimize the reflections to minimum effect. I think that we both do that.
- Status
- Not open for further replies.