This from a few days ago but probably should be responded to.
As a working speaker designer who, for over 30 years, made a living pushing many speaker designs into production, I can say that it isn't as simple as you describe.
I don't know of any successful designer of well regarded products who doesn't do a combination of listening and measuring and allow his listening tests to modify the system, even if that is slightly at odds with best measurements.
The extent of this thread is evidence that there is uncertainty about the ideal measurements for a speaker. I am more of an objectionist than subjectivist. I also feel like I have a good idea of how an ideal speaker should measure, but I would never design a system in an anechoic chamber and call it done. For one thing if you assume that from 200 Hz up anechoic flat performance is the model, you still must look at real room interaction to define the performance for the bottom 1/3rd of the spectrum.
Secondly, everything I read and believe points to the direct or early sound as being preeminant in setting perceived balance, but few studies have shown that power response is totally inconsequential. This lets in the posibility that systems with the same direct response and different power response will have secondary response differences. If power response has a secondary influence, nobody has come up with guidlines of how to find the right compromise between the meaurements of the two.
When I worked at KEF there was a strong belief (lets say arrogance) that their measurements were so accurate that systems shouldn't be listened to, that listening would lead to confusion and tempt you to tweak things. Unfortunately, KEF designed a lot of products that were considered "close but not quite". They were always runners up to B&W in the regard of the reviewers (and yes there were other issues involved). I believe that some subtle tweaking here and there to the final response curve would have elevated the overal performance of their products.
My approach has always been to use measurements to get the best possible blending of drivers and the flattest and smoothest starting point, and then to use comparative listening tests to adjust the final balance to what I believe was the most accurate final response. It is possible that this is a flawed approach. I might be adjusting to a particular inaccurate balance that I like. I might be adjusting to the particular room effects of that room and that placement, rather than to a universal improvement. I might have been adjusting to the software played that day. I'm sure that the speakers I was comparing to had an influence on the outcome.
I would love to have a measuring system that was beyond reproach. This is something I have been searching for through my career and is evidenced by my contributiont to this thread and the papers I continuously quote. Still in the end, I feel the products I've developed were always stronger when designed with this combination of measuring and listening.
I really wonder who is being arrogant here?
David S.
speaker dave I have to agree with you. Last Sunday I spent a few hours setting my tweeters flat within a dB or two window, at the level of the woofers.
After this, they sounded a little bright and I couldn't quite get used to it. So I dropped the tweeters by a dB each and it sounds much better blended.
Dr Geddes stance has me feeling that if it doesn't make sense then I must be missing something or have something wrong. Your stance is more practical and for me this makes it a more realistic one.
So although I wont trust my ears to detect all problems, I counter-intuitively trust them at the business end to tell me simply whether they are happy or not.
After this, they sounded a little bright and I couldn't quite get used to it. So I dropped the tweeters by a dB each and it sounds much better blended.
Dr Geddes stance has me feeling that if it doesn't make sense then I must be missing something or have something wrong. Your stance is more practical and for me this makes it a more realistic one.
So although I wont trust my ears to detect all problems, I counter-intuitively trust them at the business end to tell me simply whether they are happy or not.
I find very agreeable what you a are proposing.
John Kreskovsky statet in the beginning, that his
current approach is to keep the on axis response flat as
possible and do the "room adaption" or "voicing"
(my words) due to listening at R>>D distances by
adapting polar dispersion.
This seems to be a valid approach to me, since it may
yield a speaker which works in the nearfield as well as
in the far field, which to me is a very desirable property
of a speaker.
Maybe the highest possible goal to achieve.
Nevertheless also this approach rises the question, if
there is a "universal frequency dependent dispersion pattern"
that works for every room, yielding subjectively balanced
response for R>>D distances as well.
Since rooms differ vastly (home or professional) my expected
answer is 'not really'.
Interesting idea ... maybe we will find individual styles
of our "skilled persons" how to compromise power
response and on axis FR, dependent on room
characteristics and dispersion pattern of the speaker.
If we are looking for the "most universal speaker", the
one needing the least adjustment over different listening
rooms and over all "skilled modifying designers"
- having tweaked the on axis FR in your proposed study-
is the winner, right ?
If that imaginary speaker comes with tone controls
- dispersion controls would be favourable, but hard
to implement - resembling the "most common tweaks"
for representative room settings, we are done.
Btw. i think early decay time over frequency being more
adequate for describing the different listening rooms,
than RT30.
My current favourites would be speakers having
broadband high directivity above Schroeder frequency
of the room and/or speakers able to mitigate the effect
of early reflections on loudness by decorrelating them.
But i have been told, especially the latter is not what
'we want'. So maybe i will work out that approach -
including real or apparent problems concerning the
direct sound radiation - in a different thread sometime.
Last edited:
About the debate over correlated and uncorrelated between ddl and Earl, these really are on a continuum and what brains do with the echo cue likewise is on a continuum. No echo can be entirely random, just verging in that direction and harder to use to create a stable perception. Likewise, some phase shifts might be harder for the brain to work with - perhaps like Griesinger's 90 degree shift that improves the sound.*
Earl often trots out masking to explain things. For sure, the senses need masking (which as PhotoShop users know is a form of sharpening) because the input from all the senses is blurry (for example, your eyes are in constant motion). But masking as an explanatory tool suffers from being an infinite regress and even when it bears some relevance, there are limits.
*"Correlated" is bad in that it means you are able to image the sound as coming from the speakers - which you don't want to do at home or in sound reinforcement settings.
Earl often trots out masking to explain things. For sure, the senses need masking (which as PhotoShop users know is a form of sharpening) because the input from all the senses is blurry (for example, your eyes are in constant motion). But masking as an explanatory tool suffers from being an infinite regress and even when it bears some relevance, there are limits.
*"Correlated" is bad in that it means you are able to image the sound as coming from the speakers - which you don't want to do at home or in sound reinforcement settings.
Last edited:
Great post which sums up the position of many of us very succinctly, but I would add one additional remark - even if it were possible to measure so perfectly and accurately in a way that corresponds 100% with our hearing perception, only those with access to very expensive laboratory grade calibrated microphones, test equipment, and large near ideal anechoic chambers could ever make such measurements to a degree of accuracy that measurements alone could conceivably be relied upon without listening tests.
For those of us in the DIY world with non-perfect, more often than not un-calibrated microphones, PC sound cards and software, (although available software is getting very good now) and a distinct lack of anechoic chambers, there will always be enough measurement uncertainty and error despite our best efforts to make a small amount of tweaking by ear a necessity, even if the "big boys" didn't have to do this. (Which from what you say, they still do have to, if they want the very best possible result from a design)
So in the context of DIY with less than perfect measuring equipment and measuring environments, a little bit of tweaking becomes a must if you want to get that last few percent of performance, and all other things being equal (technical competence etc) those that have a good ear for judging what is right and wrong about a speaker, and more importantly have the right insights about what to do to remedy a problem, are more likely to end up with a better result, even if their arsenal of test equipment is somewhat limited.
Flat measured how though ?
If you took a steady state room measurement at the listening position and adjusted it to be flat from treble down to bass it would indeed sound too bright. MUCH too bright in my opinion.
As has been discussed in this thread (who knows how many pages back now 😀 ) a room response of a speaker that measures flat on axis will generally have a roll off in the treble due to typical power response, but more importantly will appear to have a large amount of boost below about 200Hz (several dB) - measured in room, to actually sound like it has a balanced bass response.
When you think about brightness in the treble you have to know what frequency range to compare it to - the relative level between treble and bass <100Hz is not what defines brightness.
It's the relative balance between the treble and the region approximately around 150-400Hz. The best way I can describe it is that this low midrange / upper bass junction is the "anchor" for the perceived spectral balance, and the balance of other frequencies relative to this range is what gives us characteristics like "brightness", (more upper mid to treble) "forwardness" (more middle midrange) warmth, (less 400Hz through to treble) etc.
So what you want to be checking is whether the on axis response (windowed, anechoic etc) is flat between treble down to about 200Hz, and then see whether an additional 1dB reduction in the treble sounds right to you. It's not clear from your post, but it sounds like you've changed the attenuators for your tweeters ?
For what it's worth in the whole "rolling off tweeters" debate running through this thread, I don't deliberately drop the level of my ribbon tweeters below on axis flat, either as a shelf, or by adding variable slope roll off, and they sound just right to me, both in close up listening (~2 metres) and at the far end of a big room.
On the other hand, wave-guide loaded ribbons have considerably different directivity (more, and a very different profile) than the typical dome tweeters on which a lot of the "rules of thumb" about rolling off the top end are formulated based on. Perhaps this is more evidence that the problem of needing to drop the treble slightly stems from the directivity profile of many tweeters.
Last edited:
Soooo, we have a number of highly respected people, speaker dave, ddl, and even on a good day, Earl, not to mention word from KEF and B&W which conflict with one another.
I think we should keep sight that the goal IS to create physical measurements which do relate to human judgments, even if there is dispute about whether we are there now or not.
Hope it is not re-stating the obvious, but physical measurements are easier, quicker, cheaper, subject to less artifact, etc. than human judgments. They do not require the resources of Harman International to conduct, even if the work of Toole and Olive are truly foundational. Understanding the basis for physical measurements advances our understanding of music reproduction and human hearing.
But we should also never forget that human judgments are the basic criteria and physical measurements are just playing catch-up.
I think we should keep sight that the goal IS to create physical measurements which do relate to human judgments, even if there is dispute about whether we are there now or not.
Hope it is not re-stating the obvious, but physical measurements are easier, quicker, cheaper, subject to less artifact, etc. than human judgments. They do not require the resources of Harman International to conduct, even if the work of Toole and Olive are truly foundational. Understanding the basis for physical measurements advances our understanding of music reproduction and human hearing.
But we should also never forget that human judgments are the basic criteria and physical measurements are just playing catch-up.
Last edited:
but You have to be unbiased - only unbiased listeners have no difficulty recognizing accurate sound reproduction, even with hearing damage or with hearing aids...
Member
Joined 2003
That is probably what many of us really do, at least I've found it to be a very practical approach. I use short and long gate measurements, on and off-axis, to guide changes, while listening to dozens of tracks, usually over a period of weeks.
For me, no single response fits all speakers in all rooms. The result depends on the speaker, room acoustics, speaker position in the room, listening position, etc etc etc. My personal preference is usually somewhat elevated deep bass with an overall spectral down-tilt...though not quite as tilted as the preferred response in the Harmon study cited earlier in this thread.
Posting from my phone so bear with me. Dave, your post implies that I am not successful designer of highly regarded speakers. You need to reconsiderl that position.
Historically listening was required. And I agree that there are not measurement systems that do a good enough job. That's why mine are all custom.
Historically listening was required. And I agree that there are not measurement systems that do a good enough job. That's why mine are all custom.
but we all have biases, Q.E.D. no one can recognize accutate sound.
Oh gawd!
Yes everybody is biased and influenced by irrelevant, transitory, and/or artifactual influences. The people who think they are least influenced are the biggest fools of all.
BUT, there are reliable methods of collecting opinions in a trustworthy manner whether from just yourself (hard to do double-blind, of course) and from groups. If you didn't take Psych 101, you may be pretty naive about what these methods are and how to apply them - esp. for your own personal auditioning, let alone group opinion when designing speakers.
The fact a person is clueless about proper test methods (and hence skeptical about their value) doesn't mean other people are as clueless.
So as not to seem too negative this morning, here's one tiny methodological note (pompous word, "methodological" intended for comic effect): I always play the same music for testing. I have recordings of stuff I've used since 1956 on more systems, rooms, mods, and states of altered consciousness than I can ever remember. No kidding. I ain't saying that's brilliant, just mentioning a good procedure.
Last edited:
Works for me.
Thing about computers is, they don't actualy know what music is.
Just a load of numbers to them.
Simon
So... inquiring minds would like to know;
- How does the RT30 of your room measure?
- What is the polar response of your speakers?
- How much elevation in the bottom end have you settled on?
- How much overall down-tilt (gated measurement on axis to avoid room influence) have you finally settled at?
- What, if any, deviations from a flat line (however tilted) have you incorporated?
When I finaly choose a driver for my 18" + waveguide project,
The first thing I plan to do is to listen to music on it for a couple of weeks.
In my listening room, not a dead chamber.
Also I won't be listening 1 meter from the speaker.
And I won't be listening at weird off axis angles.
I may even play a bit of pink noise and a few sweeps through it.
The last thing I would consider doing is to measure it.
I rarely listen to pink noise for pleasure, just a bit of Pink Floyd.
Simon
The first thing I plan to do is to listen to music on it for a couple of weeks.
In my listening room, not a dead chamber.
Also I won't be listening 1 meter from the speaker.
And I won't be listening at weird off axis angles.
I may even play a bit of pink noise and a few sweeps through it.
The last thing I would consider doing is to measure it.
I rarely listen to pink noise for pleasure, just a bit of Pink Floyd.
Simon
You have amazing powers with your ears....memories that lies years back you can recall...to tiniest detail...
Don't have to be unbiased....But to hear if things have pace rhythm and times the music.. for that you may have to be musical..
The thing is we all like the same..Has nothing to to do with flatness..it simply doesn't matter how flat or non flat the system is...if it's wrong and unmusical from the beginning you can adjust yourself to death.. and you'll never get there anyway...
I mean does a violin come with tone controls...?? should it..??
Don't have to be unbiased....But to hear if things have pace rhythm and times the music.. for that you may have to be musical..
The thing is we all like the same..Has nothing to to do with flatness..it simply doesn't matter how flat or non flat the system is...if it's wrong and unmusical from the beginning you can adjust yourself to death.. and you'll never get there anyway...
I mean does a violin come with tone controls...?? should it..??
- Status
- Not open for further replies.