Dave
Your correct that I pay little to no attention to above 10 kHz its not a significant factor to sound quality. But I don't see how you can contend that any direct radiating speaker can be CD, low Fc or not. It can only be CD in the sense that it is always omni-directional if the sources are small enough, but this is just adding new problems. I had concluded long ago that adequite CD could only be achieved with directivity control via a waveguide (it does NOT have to be a compression driver, but the compression drivers do have other advantages.) You'd have to explain to me how CD is possible with direct radiators that is not omni-directional with extremely low efficiency and power handling.
The minimum phase (MP) issue is really such a Pandora's box. We mine as well say the frequency response doesn't apply either. The argument re MP is that the sound fields is 3-d so MP doesn't apply. The same statement can be made about frequency response (FR), or any other acoustic measurement. It is 3-d so how can a single measurement be meaningful? It takes a 3-d mapping of the FR. And that FR mapping include amplitude and phase. And that phase mapping may reduce, at each point, to MP or not. It will certainly vary point to point, of that we can be sure. It is not the application of MP that is in error or which does not apply, it is the assumption that any single measurement at a given position has any global significants to the sound of a system. A single measurement has significance to only one thing, how the direct sound at that position behaves.
So why do we worry about whether a driver is MP or not? Today we really don't need to. Our measurement systems are capable of measuring both amplitude and phase at our chosen design point and with that we can go about using our modern software to design a speaker and crossover and have confidence that the result of out predictions will match the measurements of the completed system. Whether or not the phase reduces to MP + excess delay at the design point is of littler concern. But go back a few years (or maybe decades). Things weren't so well developed. A frequency response measurement consisted of a frequency sweep yielding amplitude and little more. There was no phase data. Now I ask each of you to design a speaker given the amplitude response of each driver and little more. Forget impedance and impedance phase. Or maybe I'll give you impedance amplitude as well. It will soon become apparent why speaker design was considered an art. Then the light when on and it was realized that accounting the the phase of the acoustic output of the driver (and impedance phase) would have an impact on design of the crossover. It was further recognized (as we can easily show today) that relative to the design point the phase response of the driver would consist of a MP component plus a propagation delay. Procedures were developed to construct the minimum phase corresponding to the measured amplitude response and the rest is history. Crossover design was much simplified. If drivers have the same frequency response, relative to some measurement point, they also have the same phase, once the excess phase is removed.
Anyway, why I bring this up is because it equally applies to the statements about CSD. If two system truly have the same SPL FR response (amplitude and phase) then they will have the same decay and CSD as well. But this is, again, a measurement relative to a single point in space. Look at two systems which have the same amplitude response but different phase and you get different CSDs. Remember that FR and CSD are both obtained form the same impulse. Thus if there are differences in the CSD there must be corresponding differences in the FR. It's the same argument for burst testing revealing things that aren't in the FR. All these things are just different ways of looking at a system. They are all just post processing of an impulse response captured at some point in space. Looking at the burst response or CSD may make certain differences between systems more apparent but those differences are in the FR, though perhaps not so obvious. For example, it is difficult to gleam the decay characteristics of a system by looking at the phase response. But it is the phase response that is indicative of the decay.
Last edited:
John
I quite agree with your whole discussion. With DSP we can even exactly EQ non-minimum phase responses (I saw this incorrectly stated around here somewhere - not by you of course.)
Where ???
Didn't you realise this place is to spread rare nuggets from the cuff ?
You really shouldn't take that personally !
😀
Michael
Thanks a lot for pointing me there.
Well now I actually have three impulse responses from OS that
1.) don't match and
2.) no one seems to know (nor care) *how* thay were taken and
3.) *what* dut exactly has been measured
at least we know for sure *who* has taken the measurements - great ! - no ?
Michael
Last edited:
I just looked at the minimum phase thing for a horn in the other thread using EarlK's impulse response. The low end isn't perfectly minimum phase but it's close enough at the likely crossover frequency. The pics are ungated but gating the impulse didn't change things much. I think it was a pretty nearfield measurement and the results could of course change out in the far field.
http://www.diyaudio.com/forums/showthread.php?p=1935073#post1935073
Excess phase plot with the delay adjusted to flatten it as much as possible:
http://www.diyaudio.com/forums/showthread.php?p=1935073#post1935073
Excess phase plot with the delay adjusted to flatten it as much as possible:
Last edited:
Well, based on some of the tests that I've been doing, I can change the CSD significantly with nearly no change in the FR. I don't believe anyone has mathematically proven the relation between change in CSD and FR of real speaker driver measurements. Acoustics damping model in reality is much more complicated than current math models mainly due to the way energy is dissipated. Bear in mind that using different signal types to conduct measurements wil yield different FR results, this does not mean the performance differs.
My reference was to smoothly varying off-axis, not CD. It's a given that Dunlavy's were not CD nor can any direct radiators achieve that. It is possible, however, to have a relatively smoothly varying off-axis response and to that end, diffraction control is required if the drivers are direct radiators. Low Fc also enters the picture to some degree here as well due to directionality issues with direct radiators. It takes more than two drivers, however, with the attendant drawbacks.
Dave
With "direct radiators" as diaphragms flush mounted radiatiing into 2Pi or 4Pi space - as I use the term "direct radiators" - there ain't many degrees of freedom left to control diffraction ( and subsequently directivity) - non, to be more precisely - beside the baflle size and shape your direct radiator is mounted in - given you don't wannna go with "lenses" or slit diffraction (no *direct* radiator anyway, in my understanding).
See also
http://www.diyaudio.com/forums/showthread.php?p=1891535
Anything else regarding "serious diffraction alignment" - in the sense of achieving swiss cheese defects free sound field - is either due to "round over" or "horn contour" right away...
Michael
Last edited:
I'm not speaking of controlling in the sense of designing an OS waveguide, I'm referring to selecting a set of drivers and taking their inherent directivity into account in a system to get a relatively smoothly varying horizontal off-axis response. It does not require roundover or horn contour, there are other means to that end. The vertical off-axis of a multi-way system is an issue, but then so it is with any non-coincident system. Earl's approach with an OS waveguide and low Fc provides improved vertical off-axis response and multi-way systems of the WMTMW are another approach, though neither is perfect. But then even the direct on-axis for his design is not flat, he is taking advantage of this to provide a wider listening window in that it is not fully CD when one considers the direct on-axis. It does, however, complement and support his desired goal of wider listening window. My efforts are in a similar vein, only my goal is the optimal "sweet spot" response, so CD is not necessarily a requirement, a smoothly varying off-axis horizontally is, within a specific window. That is possible with direct radiators, but diffraction control is a requirement, at least for mono-poles.
Since I am not so interested in a wide lateral window as is Earl, so I can focus on that "sweet spot". With that as a goal, there are more options. The vertical off-axis is not necessarily an over-riding concern and if I recall correctly, Earl also is somewhat less concerned with that in his 2-way design than he is the horizontal, again, if I recall correctly. If I'm in error, please correct this. That becomes more a concern of power response and muddies the issue to some degree.
I don't see CD as a panacea for all ills of speaker design. If you do, that's fine. Earl's tradeoffs perfectly suit his goals. My goals are a bit different, some of those same tradeoffs do not suit mine. Diffraction control, as in baffle and adjacent driver diffraction, is one of them.
Dave
There is nothing to prove. As I said, given an impulse you can do what you want with it. If two systems have the same impulse, IM, they have the same behavior. There are no restrictions or assumptions. They behave identically. How you process the impulse to see what is in it (FR, CSD, Burst, Log (IM^2) ) may well show different things on different scales. Of course there are things in the CSD that will be observed much clearer than in the FR, but that is a different issue. You can question whether the FFt of an impulse is truly frequency response (i.e. question that the system is linear, time invariant) but you can not question that if the impulse for two systems, IM, is the same, then the FFt(IM) or any part of IM will be identical identical. It is a property of the FFT, if A(t) = B(t) , FFt(a) = FFt(B), etc.
Yes, if you throw out CD then there are a lot more "options". And yes I am far less concerned with the vertical directivity than the horizontal. Dealing with the vertical problems in a room has other options, but the horizontal does not. To do what I do however - extreme toe-in in live rooms - requires narrow CD which direct radiators cannot achieve (some CD that people refer to here is not "narrow" in my sense.) Although I would not call my approach a "wide lateral window", because it is in fact very narrow by most standards. It's as narrow as I can get it in a reasonable sized system.
John
One point that Soongsc might be trying to make, that can be true, is that the same magnitude of FR does not mean that there has to be identical CSD. They can have different phases and still have the same FR, but different IR and hence different CSD. Otherwise you are quite correct again.
I would offer that such a situation is going to be rare to nonexistant in real loudspeakers, so I have to discount that it actually happened that way.
An individual driver and its (analog) crossover will be close to minimum phase. However, once you combine two or more drivers and their crossovers, the sum is no longer minimum phase. You could have two speakers with identical (flat) combined SPL and very different combined phase if they used different crossover slopes or frequencies.
Last edited:
Member
Joined 2003
No I have not. I have the hardest part done though and that is the basic mold. But there is still a lot to be done. I am simply swamped here with things to do. I have recently hired some help and this has been a big help, but we are mostly just catching up at the present time.
Maybe the next couple of months.
Earl,
The link in your sig for the summa is missing an underscore.
Neat looking speakers btw!
The link in your sig for the summa is missing an underscore.
Neat looking speakers btw!
I don't think that any of the URL links have an underscore until your mouse pointer passes over them.