Deaf Acceptance
Certainly these are reasons for Compression Driver R & D as well, but they are not the only ones, and when the new drivers are deployed in a mission, they must sound good while surviving the perils of the venue.
The assertion I take issue with is that all compression drivers sound the same and thus should be treated as commodities.
Well, they do not, and the attribution that designates who has design control, is irrelevant.
When comparison is made between compression drivers of widely differing design, materials, and mission, each yields its own observable and measurable acoustic signature that is audible.
My definition of distortion is simple, it is anything that departs from the original signal of the instruments(s)/voices(s) being reproduced as reflected by the driver’s acoustic signature.
The claim that such is inaudible is no better than the study used to support it.
Here untrained listeners were used, and as we now find out, only three drivers of unknown design were evaluated ( DUT’s).
So, if the drivers are nearly the same in design and construction, then it is not unreasonable to suppose that they will sound nearly the same: the study finding.
The study simply does not prove the author’s claim.
Bottom line, some compression drivers sound better than others. And when we hear a nasty one, there is no doubt about the cause.
I call it distortion and for the purpose of this discussion, I could care less how it is measured or otherwise characterized.
Regards,
WHG
Certainly these are reasons for Compression Driver R & D as well, but they are not the only ones, and when the new drivers are deployed in a mission, they must sound good while surviving the perils of the venue.
The assertion I take issue with is that all compression drivers sound the same and thus should be treated as commodities.
Well, they do not, and the attribution that designates who has design control, is irrelevant.
When comparison is made between compression drivers of widely differing design, materials, and mission, each yields its own observable and measurable acoustic signature that is audible.
My definition of distortion is simple, it is anything that departs from the original signal of the instruments(s)/voices(s) being reproduced as reflected by the driver’s acoustic signature.
The claim that such is inaudible is no better than the study used to support it.
Here untrained listeners were used, and as we now find out, only three drivers of unknown design were evaluated ( DUT’s).
So, if the drivers are nearly the same in design and construction, then it is not unreasonable to suppose that they will sound nearly the same: the study finding.
The study simply does not prove the author’s claim.
Bottom line, some compression drivers sound better than others. And when we hear a nasty one, there is no doubt about the cause.
I call it distortion and for the purpose of this discussion, I could care less how it is measured or otherwise characterized.
Regards,
WHG
Last edited:
I never said it "proves" anything. It, and several other studies, support my claim, unlike yours which is offered up with no supporting evidence what-so-ever. Oh yea, YOU hear it so that makes it right. 🙄
If I read the above, it seem like you interpret Dr. Geddes statement as “just exchange one CD with another and listen again, and you hear no difference”.
This is of course not the case. The throats should match the waveguide used in the comparison, AND the system should be equalized for the same frequency response (because there are differences in frequency response between differenent CDs, but these can be easily equalized, so they do not make a difference).
So, when the drivers fits the waveguide AND when the system is equalized flat again (and in the case of a passive system, when the cross-over is adapted to correct for possible impedance differences between both CDs compared), in that case both drives will sound the same.
This is however not such an easy test to perform, and this is also why we should trust the available scientific data...
Very true,
By irrelevant you mean "irrelevant in the levels typically found". The same with phase distortions, typical levels are below audibility but if we had, say, 2 seconds shift between woofer and tweeter we might expect to hear it. We fall into the trap of saying "phase doesn't matter" when we can only really say it doesn't appear to matter at typical levels.
The same with THD. We would expect to hear it at some point but it might be low enough in general practice to not be a factor. I would certainly accept that your tests were carefully run and didn't show differences within the range tested.
Don Keele told me of a test he did once. He ran a signal through both inputs of an analog multiplier. That squared the input giving essentially pure second harmonic and intermod products for all inputs. This distortion was mixed in with the input signal at a variable level. At some point he said it was perceived as an extra "brightening" to the sound.
Wish I could remember at what level that was....
David
Hornless Evaluation
These comprsion drivers were evaluated using the signals coming from a microphone imbeded in A PWT, no horns were used here.
The blocked low pass signals were then recombined with that of the microphone and then reproduced through headphones.
While this process eliminates horn artifacts, it introduces a new set of its own.
Regards,
WHG
These comprsion drivers were evaluated using the signals coming from a microphone imbeded in A PWT, no horns were used here.
The blocked low pass signals were then recombined with that of the microphone and then reproduced through headphones.
While this process eliminates horn artifacts, it introduces a new set of its own.
Regards,
WHG
Yes, of course. I can always break something to the point that it is audible. My point is that if the nonlinear distortion is audible then the device is broken, because it is always possible to design it such that the distortion is not audible. Over driving the device is the usual cause.
I once wrote an algorithm that would eliminate clipping in a wave file. You would be surprised how often this occurs in practice. I had people evaluate the clipped and not clipped files. Most preferred the clipped ones. "Brighter" was a common comment.
Such a test was done and reported on although it was never published. There were 16 audio club listeners and several speakers. Two were identical Summas except with different drivers, but EQ'd to be the same. The two Summas were statistically the same in a blind listening test. The drivers were B&C versus TAD.
Earl,
Not saying that you did this but I could see a test like the B&C against the TAD where one device has a wider bandwidth than the other, let's just say for now that the TAD was good to 24Khz and the B&C falls on its face at 16khz. So to equalize the two drivers you cut anything above 16Khz from the TAD and eq them the same. Wouldn't that meet the test goals and get the result you would want of a null test? It would be a true result for listening but would be a false test really in my eyes. As I said I didn't say you did that but that tests results must be looked at in context or they are easily skewed to meet the testers desired results. This is the game played with statistics all the time, if you know how to play the numbers you get the results you desire. Learned that in one of those horrid statistics classes in college!
Not saying that you did this but I could see a test like the B&C against the TAD where one device has a wider bandwidth than the other, let's just say for now that the TAD was good to 24Khz and the B&C falls on its face at 16khz. So to equalize the two drivers you cut anything above 16Khz from the TAD and eq them the same. Wouldn't that meet the test goals and get the result you would want of a null test? It would be a true result for listening but would be a false test really in my eyes. As I said I didn't say you did that but that tests results must be looked at in context or they are easily skewed to meet the testers desired results. This is the game played with statistics all the time, if you know how to play the numbers you get the results you desire. Learned that in one of those horrid statistics classes in college!
Argumentum Ad Hominem
That dog will not hunt well for you here.
My ears are not unique. If I can hear the differences, so can others.
There are a sufficient number of studies that document the audibility of compression driver distortion; so I will provide one of them here for all to read.
Speaker Auralization – Subjective Evaluation of Nonlinear Distortion
http://www.klippel.de/uploads/media/Speaker_Auralization-Subjective_Evaluation_of_nonlin_Dist_01.pdf
In any event, trying to prove a negative is sheer folly to begin with. Now you are saying that you are making the claim without the submission of proof.
Key information is missing from your study paper and when asked to provide it, you can't/won't.
You made a blanketing claim that constitutes a leap from the evaluation specifics of three unknown compression drivers to a generalization encompassing an entire class of them. Simply put, that is statistical nonsense.
The fact remains, that all loudspeakers have sonic signatures that are audible departures from the signal feed to them. Depending on hearing acruity, some will and some will not hear the differences in the reproduced signals. And, the quantities of auditors falling on either side of that ledger will depend a lot on training, program content, intensity level, and method of presentation.
WHG
That dog will not hunt well for you here.
My ears are not unique. If I can hear the differences, so can others.
There are a sufficient number of studies that document the audibility of compression driver distortion; so I will provide one of them here for all to read.
Speaker Auralization – Subjective Evaluation of Nonlinear Distortion
http://www.klippel.de/uploads/media/Speaker_Auralization-Subjective_Evaluation_of_nonlin_Dist_01.pdf
In any event, trying to prove a negative is sheer folly to begin with. Now you are saying that you are making the claim without the submission of proof.
Key information is missing from your study paper and when asked to provide it, you can't/won't.
You made a blanketing claim that constitutes a leap from the evaluation specifics of three unknown compression drivers to a generalization encompassing an entire class of them. Simply put, that is statistical nonsense.
The fact remains, that all loudspeakers have sonic signatures that are audible departures from the signal feed to them. Depending on hearing acruity, some will and some will not hear the differences in the reproduced signals. And, the quantities of auditors falling on either side of that ledger will depend a lot on training, program content, intensity level, and method of presentation.
WHG
Thank you to all the contributors to this thread! I have read most of it, and understand some of it 😉 I have a question for Dr. Geddes and others. From the beginning of this thread you stress the importance of pattern control. Is there any advantage to taking steps to control the pattern below 1k hz? For example, if you placed the woofer from your Summa speakers in a horn or waveguide with a 90 degree pattern, but with a throat much too big to actually horn load the driver, would this be an "improvement". Since you don't actually do this, I'm guessing not. And I suppose that if you go to all the trouble of horn loading the woofer, you might as well use a proper horn, but then you would have reduced bandwidth and need a third driver. So, waveguides for woofers, yum or yuck?
Would it be possible for an unethical person to cheat on the tests? Of course it would. You either trust the integrity of the researcher or you don't.
In a word: Yuck.
As Griesinger shows our hearing is dominated by the signal from 700 - 7000 Hz. Hence below 700 Hz the importance of directivity diminishes until it reaches the modal range where it has no relevance at all. So directivity below 700-1000 Hz is not a bad thing if the "costs" are low (i.e. no tradeoffs), but with low frequency waveguides this is simply not the case. They need to be enormous - an 18" waveguide just barely has directivity control below 1000 Hz. To get down to 500 Hz the waveguide would need to be about three feet across and more than three feet long, all this for a margin improvement in the subjective importance of polar response. And now you still need another driver from 500 Hz down to the modal region. The incremental "costs" are staggering for a benefit that is marginal.
Adding to this... are high order modes only present and/or audible above a certain frequency or would you have to take this into account using a waveguide on the woofer? If so, Foam probably wouldn't be enough to suppress them or you'd have to spend stupid money to get enough, right?
edit: just saw you answered his question, but if you were crazy enough to put in this giant waveguide and had a giant room, would we have to take HOMs into account?
That paper contains no objective to subjective scaling. It has nothing to do with the discussion.
Earl,
I don't think that the person doing the testing needs to be unethical or even intentionally skew a test to give a particular result, that could just be a technical failure or an attempt to equalize a test regime. I think the major point should be that you need to look at the actual test setup and understand the testing to give a test credibility and not just blindly take the conclusions for granted. I don't think that some of those physicists who come up with some really incredible results that can not be duplicated are unethical many times, just that the test conditions were flawed. After others attempt to duplicate the results the flaws become evident in the test protocols. It is all in the details I would say and the objective conclusions can have a misapplied constraint. Subjectivity is somewhat like trying to prove a negative, to subjective for some to objectively accept.
I don't think that the person doing the testing needs to be unethical or even intentionally skew a test to give a particular result, that could just be a technical failure or an attempt to equalize a test regime. I think the major point should be that you need to look at the actual test setup and understand the testing to give a test credibility and not just blindly take the conclusions for granted. I don't think that some of those physicists who come up with some really incredible results that can not be duplicated are unethical many times, just that the test conditions were flawed. After others attempt to duplicate the results the flaws become evident in the test protocols. It is all in the details I would say and the objective conclusions can have a misapplied constraint. Subjectivity is somewhat like trying to prove a negative, to subjective for some to objectively accept.
Not sure. I have never built or used a properly design LF waveguide so I don't know what would happen. The gross downsides are kind of obvious, the more subtle downsides are not. The gross ones were enough to steer me away.
HOMs do "cut-in" at some frequency which depends on the throat radius. So a larger waveguide would have much lower cut-in to the HOMs. Actual numbers I don't have except that a 1" throat cuts-in at something like 4 kHz for the lowest mode, about 7 kHz for the lowest symmetric mode. These numbers should scale linearly with throat radius.
You can accept the studies as they stand or take a wild-***-guess which is all you can do otherwise.
How are you going to get your hands on one? From what I hear, the only ones out there are part of JBL's home flagship speakers.