Haha
clever people you French are ! – it's a nice and multi-facet saying – depending a lot on where you put yourself
On the other hand I would't see it that “absolute”:
CMP concept will bring current "FR concept paradigm" to an end, thats pretty sure – but – at least we'll gain precise knowledge about the borders FR concept is valid in only and may be stimulated to look at certain effects in audio and room acoustics from a pretty different point of view.
This IMO already is quite something, once CMP concept gets "intuitive / common knowledge" 😉 among audio geeks in all its implications.
It even may give an extraordinary impulse to the paralysed audio sceen, staring at FR specs mostly and also may open doors in other fields too.
Michael
There is a fundamental difference in understanding / looking a t the topic when searching for "relevant (and non compromised) time windows" versus "accepting that FR is a generally flawed concept" as soon as we get to "real world" listening
🙂
Michael
Yeah and on top of that :
"perspective is everything"
Replace "perspective" with the "concepts we have in mind" and you know where I'm heading
Michael
Yes, yes, and cure the common cold.
I fail to see how this is a revolutionary concept or in any way replaces frequency response when measured in an appropriate way.
The question remains "what form of measurement correlates with what we hear?" Much research points in the same direction, that time windowed frequency response is fundamental. It also encompasses Olive's findings in that early reflections will impact response smoothness that he finds a factor, late reflections will not.
David S.
You need a speaker with the same polar behaviour from20Hz to 20kHz. So with the increasing distance spl stays constant for the hole FR-band.
Which would add flat power response to flat frequency response. Problem is, flat power response is perceived as "too bright".
Climbing directivity is typical and desirable.
David S.
In the infamous "RIBM" comparison (Toole Sound reproduction p.394) speaker R wins. It is only second in frequency response linearity, but first in linearity of early reflections and constant directivity. Of course speaker M is a clear example that good directivity doesn't help if frequency response is a mess.
Rudolf
"R and I are in a statistical tie for first place" says Toole just below.
Go back to his earlier larger studies and it is clear that power response is not an indicator. For example in his comparison of 20 speakers both the 10th place and 18th place systems have far nicer power response than 6 of the first 8 speakers in the rankings.
We fall back on the Toole analysis frequently because it is the primary study that looks at a broad range of system mesurements and correlates them with controlled blind listening tests. The conclusions are pretty clear. To do well in the listening tests you need smooth and flat on-axis and near on-axis response. You need wide bandwidth (bass extension).
You do not need any particular polar pattern. You do not need any particular directivity index shape. A d.i. climbing to about 6dB is the norm of the better ranked speakers. You don't need a particular power response shape other than a smooth power response (free from resonances). Dips in power response at crossover points are acceptable. No particular phase response is required.
Kates, Salmi, and Lipshitz and Vanderkooy point in the same direction: direct and early sound is key. Early sound should measure flat and smooth. Flat power response or a flat in-room curves are undesirable.
David S.
To add more trouble, I would like to take one of you as an experimentation subject :😱
with the same system, the same filter slopes, the same EQ, the same levels, the same source, the speakers perfectly disposed, in a symmetric room.
I could make you listening two settings on the DCX (instant comparison then). One is perceived as louder, and not 0,5 dB, you could bet it's roughly 3dB more.
Measuring it, both channels on action (of course, the trick is here) gated or not, the result shows that the levels over 500 Hz are the same, and worse, the 100/500 band is 2 dB less, this on the setting perceived as "louder".
How comes ? Just by some tiny time and phase adjustments between each channel smearing the cross talk.
By this anecdote, I wanted to remind you that we absolutely don't react as a mike.
The target is biologic, think of it 😀.
with the same system, the same filter slopes, the same EQ, the same levels, the same source, the speakers perfectly disposed, in a symmetric room.
I could make you listening two settings on the DCX (instant comparison then). One is perceived as louder, and not 0,5 dB, you could bet it's roughly 3dB more.
Measuring it, both channels on action (of course, the trick is here) gated or not, the result shows that the levels over 500 Hz are the same, and worse, the 100/500 band is 2 dB less, this on the setting perceived as "louder".
How comes ? Just by some tiny time and phase adjustments between each channel smearing the cross talk.
By this anecdote, I wanted to remind you that we absolutely don't react as a mike.
The target is biologic, think of it 😀.
Last edited:
David,
I took my time to merge some diagrams from the 20 speaker studie - place 1-4 and 9-12. Place 10 ist not what I would consider a decent CD. A power response which is just a statistical mean of awkwardly different responses under certain angles is in fact almost useless.
I certainly haven't read much of Tooles work. But from what I know, he never did an earnest, even less a decided comparison of traditional versus CD speakers of comparable quality. The speakers in figure 18.8 (p.381) would be nice CD in todays standard.
In the end the findings of Toole et al have to be taken as what they are: measurements and conclusions that can only apply to the class of speakers that has been tested. I certainly don't deny that a smooth on-axis response (possibly linear, but not necessarily flat) and wide bandwidth are first requirements for a good speaker. But good CD (regardless of pattern or index number) is an added advantage, which is easily recognisable.
Rudolf
Attachments
I haven't looked into that – but - I will have a look what CMP possibly can do for your cold
😀
meanwhile – may I recommend you some herbal tee – preferable from Hollander and Lime-tree blossom - and having a good overnight sweating in your bed ?
Michael
Last edited:
Hi Rudolf,
There are many things we can optimize in speaker design including frequency response, power response, harmonic distortion levels, phase response, etc. It never hurts to improve most of these factors. The question when designing for a competitive market is what are the necessary and sufficient parameters to optimize. That is, what must be on the list as a priority and what might we exclude from the list.
As one of the first to design and write about the use of CD horns at JBL in the 80s, I don't want to be in the position of arguing against them. But many have taken a leap from constant directivity is good, to flat power response, or linear power response is essential. This is not correct.
First off CD is an unfortunate term. Multicell horns have constant directivity but very poor performance and great variability of response. However highly uniform or smooth polar performance is a good thing because we can equalize a speaker to have flat axial response and see smaller variation near the design axis. This will give useful latitude of listener position and maybe allow a friend to enjoy the experience with us!
Although smooth power response seems like a great thing, tests don't confirm it is a necessary criterion. All of the better ranked speakers in the Toole tests were conventional 2 and 3 way systems with power response dips at every crossover point. It didn't prevent them being ranked well. Lipshitz and Vanderkooy did an interesting test with a side firing dipole on top of a convenional forward firing system (Quad ESL 63 on a KEF 104.2). With this settup they could independently manipulate axial response and power response. They found that holes in the power response were innocuous (peaks were not).
From this you can't say that smoother power response would increase the ranking of a system, even if Floyd's test don't show a system as good as you think could currently be made. Clearly if you look at power response curves for the rank ordered systems there are none with flat power and there is no correlation between power response shape and ranking . The two power response curves I mentioned were better than those of all the better ranked systems but since the 0 and 15 degree curves were poor these systems suffered in the ranking.
Flat power response, that some people advocate, and that would come along with true constant directivity, is always shot down in scientific studies. The L&V study I mentioned found that achieving flat power by equalizing the axial response to compensate for rising directivity, sounded bright and colored. Achieving flat power response by equalizing the side firing dipole (therefor not corrupting the axial response) still sounded too bright.
Necessary and sufficient criterion: flat and smooth axial response, smooth power response free from evidence of resonances, sufficient bass extension.
David S.
There are many things we can optimize in speaker design including frequency response, power response, harmonic distortion levels, phase response, etc. It never hurts to improve most of these factors. The question when designing for a competitive market is what are the necessary and sufficient parameters to optimize. That is, what must be on the list as a priority and what might we exclude from the list.
As one of the first to design and write about the use of CD horns at JBL in the 80s, I don't want to be in the position of arguing against them. But many have taken a leap from constant directivity is good, to flat power response, or linear power response is essential. This is not correct.
First off CD is an unfortunate term. Multicell horns have constant directivity but very poor performance and great variability of response. However highly uniform or smooth polar performance is a good thing because we can equalize a speaker to have flat axial response and see smaller variation near the design axis. This will give useful latitude of listener position and maybe allow a friend to enjoy the experience with us!
Although smooth power response seems like a great thing, tests don't confirm it is a necessary criterion. All of the better ranked speakers in the Toole tests were conventional 2 and 3 way systems with power response dips at every crossover point. It didn't prevent them being ranked well. Lipshitz and Vanderkooy did an interesting test with a side firing dipole on top of a convenional forward firing system (Quad ESL 63 on a KEF 104.2). With this settup they could independently manipulate axial response and power response. They found that holes in the power response were innocuous (peaks were not).
From this you can't say that smoother power response would increase the ranking of a system, even if Floyd's test don't show a system as good as you think could currently be made. Clearly if you look at power response curves for the rank ordered systems there are none with flat power and there is no correlation between power response shape and ranking . The two power response curves I mentioned were better than those of all the better ranked systems but since the 0 and 15 degree curves were poor these systems suffered in the ranking.
Flat power response, that some people advocate, and that would come along with true constant directivity, is always shot down in scientific studies. The L&V study I mentioned found that achieving flat power by equalizing the axial response to compensate for rising directivity, sounded bright and colored. Achieving flat power response by equalizing the side firing dipole (therefor not corrupting the axial response) still sounded too bright.
Necessary and sufficient criterion: flat and smooth axial response, smooth power response free from evidence of resonances, sufficient bass extension.
David S.
MLSSA supports this type of windowing. It's interesting, but can't say I personally have been able to correlate its measured outcomes with my perceptions, but I didn't try a rigorous, blind experiment to do so. Still, its a useful design tool.
MLSSA Adaptive Window
I havent read books about the subject and I tell from my own experience.
I like to have in crease of spl from 10kHz and up. I like the sound better and I can measure that it is better at listening position.
I can not second designs that decrease high frequency above 10kHz.
And still I want a explanation why flat would be to bright It is just not logical. And due it short wave length spl decreases in stereo measurement.
I can find me in the raise of low frequency because it radiates in to full-space.
I have experienced this myself. Flat response in the high frequencies sounds bright. It needs to droop a little to sound right. That said, there was some discussion about pitch center which occurs at about 632 Hz (correct me if I'm wrong) and that the two halves separated by the pitch center should have equal energy. Pano, you may be able to describe this better?
That however could also mean, that flat power is not wrong but there are other effects at play, e.g. SL's "finding" that flat on axis FR is not correct for stereo and those effects weigh heavier.
My understanding is that the direct response should be flat or the anechoic axial response of the system should be flat (same thing) but the in-room curve will typically fall at high frequencies. This is well known also in commercial sound where large sound systems are always EQed to some falling steady state response curve.
Lets be clear on the difference. The direct sound, whether in a live or dead room has a response independent of any room effects. But when you measure the steady state response (as with an RTA), then you are looking at the sum of the direct sound and the reverberent field. The reverberent field is a picture of the total radiated power (omnidirectional radiation) of the system, with additional rolloff from the surface absorption of walls, floors and ceilings, plus even HF absorption due to the air. With any typical speaker and typical acoustic environment, if you are some distance from the system the direct sound has very little impact on the measured response curve. The response will roll of considerably. If you flatten out this room curve the direct sound will tilt up at HF and the system will sound way too bright.
Look up the "Cinema X" curve or typical PA "house curve" and you will see recomendations for up to 10dB or more rolloff at 10kHz to make the system sound natural. My belief has always been that these are roundabout ways of EQing the reverberent field to keep the direct sound flat. In the home listing room the effect is smaller, perhaps 2-3dB for most listeners, but the cause and effect is the same.
We want the direct sound to be flat, we want the reverberent field to roll off, if required, to keep the direct sound flat. In fact, if we could have both the direct and reverberent fields flat we would find that unatural and bright.
I remember talking to the Mirage designer of the Omnisat speaker line. He found that since they were nearly omnidirectional that he had to tilt down the response 3-4 dB for a flat sounding balance. Same thing again.
David S.
Last edited:
- Status
- Not open for further replies.