|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
70.91.122.97
In Reply to: As a professional speaker designer...... posted by Jon Risch on February 18, 2007 at 20:43:45:
All measurements show where a speaker deviates from theoretical perfection, although someone else's measurements can't include the effects of your own listening room, or account for your unique sound quality preferences.Large deviations are not something an audiophile should ignore when deciding which speakers to audition.
The real issue is to decide which speakers you want to consider for purchase -- there are FAR too many brands and models for one audiophile to choose from.
Measurements won't tell you how any speakers will sound at home in your room -- but even an audition in a store won't tell you that.
Measurements will guide you toward speakers that are MORE LIKELY to sound above average when you hear them at home.
I'm assuming a listener over many years has compared speakers whose sound quality he has experienced at home, with their published measurements, and HAS found some measurement(s) that generally correlates with speakers he likes and/or dislikes at home.
For one example: I have found over many years that step-response, as measured by John Atkinson at Stereophile, correlates well with my own subjective speaker opinions of speakers I have borrowed for an audition at home, at least for the near field speaker positions I use at home.
On axis anechoic frequency response shows the ability of a speaker to accurately reproduce the input signal.
Room reflections in a real room will make the frequency response at the listeners ears worse, compared with on-axis anechoic measurements ... but there is no reason to design speakers with a poor on-axis frequency response and hope the room reflections will make the sound more accurate by the time it reaches the listener's ears!
There may be debates over what amount of high freuency roll-off over 2kHz sounds natural, and over the need for bass ramp-up under 80Hz. (because a "flat" frequency response at frequency extremes doesn't sound flat to most listeners) ... but there is little debate over the desire for a flat frequency response in the 80-2000Hz, range.
Step response shows the ability of a speaker to accurately reproduce a transient sound. It correlates with sound quality better than flipping a coin.
While flat on-axis anechoic frequency response in the mid-range frequencies, and an excellent step response, don't guarantee you will like a speaker's sound quality in your room, the measurements do roughly correlate with speaker preferences for most listeners in most rooms.
Most important: Measurements may help an audiophile separate above-average speakers from below-average speakers, and that can really help an audiophile decide what speakers he will audition before his next purchase.
.
.
.
Richard BassNut Greene
Subjective Audiophile 2007
Follow Ups:
What is amazing is how ready you are to twist things around to seem more controversial.The other interesting thing is your use of 'incremental logic', where you start making statements that start out logical, but each one gets a little further from the facts of the matter at hand, until we end up where your convoluted train of logic deliver's us: YOUR opinion turned into some sort of psuedo-logical fact.
For instance, take this gem:
"Measurements will guide you toward speakers that are MORE LIKELY to sound above average when you hear them at home."Sounds eminently logical, but in reality, it sucks.
I stand by my post, those unfamilair with acoustic measurements are more likely to be confused or mislead tham aided by the typical batch of loudspeaker measurements.
You are, in essence, saying that the measurements can only help, that somehow, folks without the background and experience with acoustic measurements are going to be able to make useful judgements from simply looking at a frequency response curve, at a dispersion plot, at an impedance curve, etc.,
While this might be true to a limited extent if all we were going to study was whether or not a transistor radio speaker measured AND sounded worse than a Polk, that is a somewhat trivial pairing as an example. But once we get down to the very much more subtle measured differences between mid-fi speaker systems, even the professionals are going to be hard-pressed to make any solid sense out of the measurement data, and CERTAINLY, no professional I know would be so foolish as to try and judge a speaker, or try to determine whether or not any given speaker SOUNDED better than another, simply by looking at measurements taken by someone else.
And THAT is what you are saying, that folks can dismiss a given speaker strictly due to their inexperienced judgement based on a FR curve or a polar plot taken via who-knows-what-method.
Once we get above the level of a trivial pairing, NO WAY!
Yet this is exactly what you say:
" Most important: Measurements may help an audiophile separate above-average speakers from below-average speakers, and that can really help an audiophile decide what speakers he will audition before his next purchase. "I say again, NO WAY!
Richard stated:
" For one example: I have found over many years that step-response, as measured by John Atkinson at Stereophile, correlates well with my own subjective speaker opinions of speakers I have borrowed for an audition at home, at least for the near field speaker positions I use at home. "Good for you Richard, but what about the rest of the world, with THEIR particular system and components? How many people listen near-field? How has your subjective opinion been shown to correlate with anyone else's?
Richard stated:
" On axis anechoic frequency response shows the ability of a speaker to accurately reproduce the input signal. "Sez who? This is so laughably inaccurate a statement, that I can't even begin to explain. On axis FR may tell us something, it can even tell us more than one thing if done a certain way, but a FR plot BY ITSELF actually can fool us into thinking that a speaker is doing more/better than it really is.
All kinds of things can 'hide' in a typical 1/3 octave averaged FR plot, including speaker component resonances, cabinet or vent resonances, distortion, cone or diaphragm break-up, poor transient response, baffle diffraction, etc., etc.
So much for "accurately reproducing the input signal".Richard stated:
" Room reflections in a real room will make the frequency response at the listeners ears worse, compared with on-axis anechoic measurements ... but there is no reason to design speakers with a poor on-axis frequency response and hope the room reflections will make the sound more accurate by the time it reaches the listener's ears! 'A red herring. A so-called poor FR just might be a real and honest FR curve, showing better than average phase behavior, transient response and low distortion, all the while looking much worse than a smoothed, homgenized FR curve, one with poor phase behavior, lousy transient response, and high levels of higher order distortions.
Richard stated:
" There may be debates over what amount of high freuency roll-off over 2kHz sounds natural, and over the need for bass ramp-up under 80Hz. (because a "flat" frequency response at frequency extremes doesn't sound flat to most listeners) ... but there is little debate over the desire for a flat frequency response in the 80-2000Hz, range. "Actually, you have this one wrong too, as a FR that measured flat from 80 Hz to 2 kHz in an anechoic chamber would sound pretty thin, boomy and all around terrible. But then, everyone knows this, right Richard? They would also know that one should not be designing in a FR roll-off above 2 kHz, as this means that they are seriously confused about the difference between flat on-axis direct sound and the power response of a loudspeaker in a room.
Richard stated:
" Step response shows the ability of a speaker to accurately reproduce a transient sound. It correlates with sound quality better than flipping a coin. "This one is also wrong. Technically, one would use an impulse, not a step function, to check transient response. In fact, most people do not know how to properly interpret a band-limited step fucntion, INCLUDING a great many loudspeaker designer's!
As for your cute little comment:
" It correlates with sound quality better than flipping a coin. "
I doubt that you, or a great many folks, would know what to look for in a step function, how to interpret one, and whether or not one particular step function was 'better' than another. In that case, flipping a coin would be a more sensible thing to do.You can portray my experience and knowledge as being condecending, but I am merely talking about reality, not some ideal dream world where _everyone_ has been educated and trained about acoustics and filter theory, and has had the sheer bench time looking at loudspeaker measurements and data just like someone who does it year after year for a living, 5 days a week.
Jon Risch
Ok Jon, all that being said, which measurements do YOU consider to be of great importance to a speaker design? Is power response a more reasonable measurement than on-axis FR response. Cone and cabinet resonances? The onset of thermal and/or dynamic compression, phase shift, harmonic distortion? Having designed several pairs of speakers on an amateur level, I have some idea what these measurements mean and how they interact. As a professional what should we be looking for in measurements and which ones are not being shown that would be important?One of my biggest disappointments with box speakers (and why I have exclusively planar speakers now) is the colorations from the drivers and especially cabinets. IMO, these outweigh FR (unless it is simply a horrid response) because they are not constants so the brain has no chance to filter these things out. Its there, then its not, depending on whether the resonance is excited. At least FR is pretty constant (with in a reasonable volume level).
Another one that gets me is high order filters with multiple drivers all of different materials. Talk about hearing each driver!! The worst case of this for me was the B&W N802. You have Carbon fiber woofers, Kevlar mid, and aluminum tweeter. All have breakup modes that are audible and very sharp filters so transitions of instruments between drivers are painfully obvious.
I get to work with some very advanced technology, and so, have the luxury of being able to 'bend the rules' as far as conventional wisdom is concerned for home playback loudspeaker systems.
To respond to your questions:As for what I think is the most important, I will answwer that one last, so look down below.
I am of the 'direct frequency response' school, as opposed to flat power response. I also am not very enamored of omni type systems, they are too much at the mercy of the room, even one that is well treated acoustically. I also have found that speakers that control the radiation pattern (not necesarily horns/waveguides) seem to do better in the vast majority of home environments than ones with extra wide dispersion.
In that sense, dipolar speakers, such as your planars, have a form of controlled dispersion, which extends down into the bass, which places them in a completely different category as the vast majority of omni bass speaker systems. However, being one who values deep bass response (as long as it is tight and accurate), I have found that an enclosed woofer can be made to be fairly transparent and accurate, as long as heroic efforts are undertaken to minimize the box talk, internal box reflections, and driver resonances.
As for power compression and dynamic changes, by using Pro grade drivers, I avoid a great many of the worst of those issues and problems, as normal home playback levels do not even begin to tax a 95 dB sensitivity speaker system whose components can normally handle nearly a thousand watts in a typical pro sound loudspeaker system configuration.
I have developed a special topology that allows such Pro sound drivers to be used in a very simple and pure mode, for a taste of what I am talking about, see:
http://www.geocities.com/jonrisch/LBIseries.htm
and subsequent pages, actually, the last page at:
http://www.geocities.com/jonrisch/LBIseries4.htmThe distortion for this sytem at nominal home playback levels is in the tenths of a percent of distortion range.
Now, all of that aside, what measurements do I consider important?
The most important 'measurement' I can make, is to listen to a speaker using familar components in a familiar environment using program material I am familiar with and enjoy or like.
I know, I know, you want the so-called objective measurement side of things. OK. The thing I prize the most, and strive for in my designs, is to achieve good clarity of sound playback. Unfortunately, there are no single dimension measurements for this factor. What do I mean by clarity? Some call it articulation, others definition or detail, and yet others may call it presence or call it 'low distortion'. What I want to achieve most of all, is to provide the clarity and cleanliness that is irretrievable via any processor I am aware of. You can't EQ clarity back into a fuzzy ill-defined speaker system, nor can you DSP it back in, it has to be preserved, via superior drivers, superior execution of the system as a whole, superior filter networks, a whole host of intangibles that are not as in your face as the all-mightly and one-dimensional FR plot.Yes, the FR does need to be reasonably flat, but whose idea of 'flat' is always the main sticking point. Flat as a ruler, ah yes, but what is the scale and range of the FR graph? Was 1/3 octave averaging (or even greater averaging) used or not? What is the inherent resolution of the measurement at any given frequency? And so on, and so forth.
Then there is allowing one's self to get hung up on a specific number or frequency, say, requiring that the FR be flat down to 50 Hz or some such criteria. Hey, that's what subwoofers are for, eh? Anyway, what matters more than reaching 50 Hz dead flat is, what is the roll-off rate below that, and how consistent is it with drive level? What is the distortion level at 50 Hz? 40 Hz?, and on and on.
When evaluating drivers or designing systems, I tend to use a holistic approach, and try not to get hung up on any single metric, and I certainly do not let the microphone rule my goals and define my milestones. What it listens like is the single most important thing.
Come on Jon -- I'm retired now but still don't have enough time to read such a long post. More concise please.This is about deciding which speakers to audition.
There are way too many brands and models to hear them all.
So how do you narrow down the candidates for auditions?
I say measurements work better than reviews (text).
The reviews are so often positive that I'm not sure you could eliminate any speakers just from the text!
Even when the review finds many faults, the review conclusion often says you should audition them anyway!
The measurements are usually average, or don't reveal much to the layman, but sometimes there are unusually bad or unusually good measurements.
Avoiding speakers with unusually bad measurements, and seeking speakers with unusually good measurements, makes sense.
More sense than flipping a coin.
Example 1:
A BassNut who strongly prefers full range sound will be able to glance at an on-axis anechoic frequency response curve and get a good idea whether the speakers will have enough bass output under 50Hz. to satisfy him.Reading the text of a speaker review will too often reveal an overexcited positive opinion by a reviewer who can live with bass roll-off under 50Hz., because he likes so much else about the speakers, or doesn't notice the weak bass so much because he places the speakers in the room corners of his walk-in-closet-sized room!
Over time an audiophile may find that a certain measurement(s) done by a certain reviewer, or by Atkinson, works fairly well to identify speakers worthy of an audition. You could miss good sounding speakers that happen to measure poorly, but you are more likely to find good speakers that measure well ... no matter what you say.
Please point out one speaker model whose anechoic on-axis frequency response is poor in the 80-2000Hz. "midrange" (let's say worse than +/-3dB with one-third octave smoothing)... but most listeners still think it's a great sounding speaker? You'd have us believe poor on-axis response has no correlation with subjective auditions at all.
.
.
.
I'm getting scared though. When i agree with Richard over Jon, it makes me wonder what's going on. Sean
>
I think Jon's point was that the final result needs to actually sound good even if it measures well. Jon's first point was that it is difficult to trust somebody else's measurements. One subtle point that I got out of it was that there can be very subtle changes which can make all the difference yet still have the graphs, etc, look rather similar. Translation; You need to listen. That's what the customers do.I think Richard is correct in that measurements DO matter, but they are not the only thing. I have been building speakers for 29 years but couldn't really match what I considered decent high end speakers until I got my own measuring and test equipment. Richard is correct when he talks about room treatments and speaker setup too. It makes a huge difference... bigger than component choices and almost as big as the speaker choice.
I think we have two rather intelligent folks here yet one is a very cautious even-keeled personality where the other is gregarious and over the top and may sometimes forget to take his medication. You can guess which one cracks me up sometimes. ;)
They both give good advice but Jon's is confidently understated while Richard jams his down your throat like giving a dog a pill. He did that to me once and I went off and experimented and found that Richard was right. I do have to wonder if Richard took Jon's post as if Jon was saying that measurements don't matter. My take was that although they do matter, one must listen too. It's funny that most of the folks who state this are the ones whose bread-and-butter is a function of how good their speakers are. Follow the money...
Bill
Next time I'll wrap the pill in bacon.To use the word "Jon" (Risch) and "understated" in the same sentence seems unusual because he claims to have such incredible hearing ability that he can hear differences among one dozen different wire insulation materials under double-blind conditions with high scores and can rank-order those materials by sound quality.
I, on the other hand, tried to replicate his alleged wire experiment but could never find speaker cables that were identical except for their insulation materials.
I did get to compare a $1000 Tara Labs speaker cable with $10 of 14AWG Radio Shack speaker wire and heard no difference, along with almost one-dozen other audio club members.
I didn't write a white paper about the test, get it peer-reviewed and published in an audio journal ... but only because a pitcher of beer was spilled on the data sheets.
.
.
.
NT
If I've got enough watts (and I currently do), for the job. Then all I care about is what my heart is doing. Nothing more. Certainly any gross abherrations are audible, but I don't think anyone is claiming they are also musical? My (more than a few years now) journey through fullrange drivers/speaks, has entirely redirected my view about what is actually required in/for music reproduction. And what is possible, what can be achieved. With next to nothing.For me it has become merely doing as litle damage to the music as possible. Setting it free. Letting it soar. Neither measuring it, quantifying it, comparing it, flattening it, or damping it to death. Music has become the only issue, and my heart the only arbiter.
Sing it with me now, sing it loud, here we go .... y'alls mileage may vary :-)
This post is made possible by the generous support of people like you and our sponsors: