Well, with a deep cone driver with high cone damping the HF mainly comes from the center/dustcap while the LF is radiated by all of the cone surface, that makes the LF arriving a bit ahead of the HF, so there is a small tilt (tiny amount of excess phase). Acoustic center being frequency dependant. But it's a practical non-issue.
Last edited:
I've seen this said, but never proven. It has to be untrue at some point - more delay, more delayed signals - because we know that the impulse response form a room is not minimum phase. So when does this happen? One reflection, the second one? This is what troubles me about this claim is that it cannot be true in general, there have to be some limitations/assumptions to it.
Kstr, export from Bagby sim, remove headers and space at beginning of each line, import into Holm or Rephase, rest you know🙂
In any study of anything psychoacoustic, one must be very careful with the design so as to be varying only those factors being studied and that no other variables are changing. This is extremely hard to do and results in extremely simplified studies in order to achieve this goal. There was nothing "realistic" about the signals that were used as far as them occurring in real loudspeakers. Such a thing would not have been possible or at the very least extremely difficult to do.
The signals were idealized to the extreme allowing for manipulation of only the simplest variables. The 2 dB, to 6 dB were the level changes to the signal after the addition of the diffraction signal. So 6 dB would be equal direct and diffracted signals, a virtual impossibility but also the extreme case. 2 dB would be more likely.
And yes, in a real baffle there would be a plurality of delays and the entire diffracted signal would be smeared in time. How would one then specify the "time delay"?
One can always pick apart these kinds of tests, like using impulses or sine waves as much of Toole's work was done, and one can always say that it is not realistic or it should have been done this way ... I simply respond to each and every one of these critiques as: "Well then do it yourself however you want it done!"
This is precisely why so little is know about critical issues because it is so difficult to study "real situations" and there is always questions of applicability in the grossly simplified (but doable) studies.
What one has to do is to look at every situation and see it what is believed is consistent with what is known from statistically valid studies. You cannot say from our study if Box A will be better or worse than box B, but you can say that the more diffraction that there is the "more likely" it is to be audible and that the more delay that there is the "more likely" that it is to be audible and, most importantly, the higher the SPL the more it will be audible. This last aspect being very clearly shown in our study.
Finally, my interest in this study was the audibility of a waveguides internal diffraction and the results were completely consistent with what I was finding in their development - reducing HOMs improved the sound quality and that the HOMs became more audible with SPL making them appear to be nonlinear distortion, which we had already shown was not audible. I was perfectly happy with those results (and remember, I was somewhat disturbed to find that nonlinearity was NOT audible, I had not expected that, but I did accept it.)
Eventually I bought that driver and found the same thing. JohnK was right. 🙂. (yes I know, "duality", but I mean in the sense that it was an issue within the driver itself).
Holm works fine on my Win7 laptop, for whatever that's worth. I don't use it much, but I have actually done exactly what you're trying to do on that computer. Maybe something about your FRD is making it crash?
Self contained operation works for me, but .WAV import always failed. Have to admit I haven't tried text file imports. VMWare still is a good idea, have a ton of old and new (self-coded) DOS executables...
Oh, I don't know, never tried to import a WAV, just importing FRD and exporting WAV, like you were talking about on the last page.
Holm works on every machine that I have tried from XP through Win 7 and now on all of my Win 8 machines. I have never had even the slightest trouble with it and I use it a lot!
At this point, I would just ask that we stay at least somewhere in the vicinity of the original topic. Diffraction from enclosures was getting a little wide of the mark, but measurements are just too far off.
Last edited:
KSTR, it was crashing for me also well while importing the files from Bagby's software. Then I found that removing the header and the space in front of every line in the exported file made Holm happy.
Yes, the diffraction may be getting off topic, especially, the measurements and listening test. I'll start a new thread. But it is a feature of your waveguides and speaker design concepts.
1. When you say minimum phase, it must be in comparison against known source signal to determine phase and response. Perhaps you can elaborate how you can do that in real practice?
2. Actually because it will vary at different points, you will not consider it minimum phase, and need to measure multiple locations before you decide how to eq so that you are equing the minimum phase part of the response.
Minimum phase system has to be point source, measured by point receiver. At higher frequencies different points of driver differ in magnitude and phase.
Sound energy propagates from voice coil through driver as speed of sound for membrane material.
Driver response from measurement point can be equalized to appear minimum phase; but movement to second measurement point results in different response because system isn't minimum phase to start with.
When driver bandwidth is restricted to wavelengths much greater than driver diameter and baffle system, a monopole with minimum phase behavior results.
Sound energy propagates from voice coil through driver as speed of sound for membrane material.
Driver response from measurement point can be equalized to appear minimum phase; but movement to second measurement point results in different response because system isn't minimum phase to start with.
When driver bandwidth is restricted to wavelengths much greater than driver diameter and baffle system, a monopole with minimum phase behavior results.
So basically, it is meaningless to even discuss whether diffraction is minimum phase or not from speaker design point of view. I seen no further need to address this.
I think you need to look up the definition of minimum phase, because its not what you seem to think it is. Minimum phase does not require comparison to an "original" signal source, any given single impulse measurement can be examined and determined to be minimum phase or not through the right calculations.
Minimum phase means that all of the measured phase shift is due to the amplitude response variations as calculated by a Hilbert transform, and that there is no additional or "excess" phase present.
For example a simple analogue low pass filter will cause inevitable phase shift which is inextricably linked to the change in amplitude response. Such a filter is minimum phase because there is no additional phase shift above that demanded by the amplitude response.
How do you measure it in practice ? Most software (including ARTA, which I'm most familiar with) does it like this:
An amplitude response is derived from the impulse response. The amplitude response is passed through a Hilbert transform to calculate the resulting minimum phase phase response. Actual phase is also directly derived from the impulse response. Excess phase is calculated by subtracting Hilbert transform derived minimum phase from actual measured phase.
Of course signal windowing and processing can introduce some errors and uncertainty in the process (especially trying to determine the phase at low frequencies) but ARTA seems to do a pretty good job of it.
In ARTA on the phase response screen you have the display options "minimum phase", (Hilbert transform calculated phase is displayed) "phase" (actual measured phase is displayed) and "excess phase". (Actual measured phase minus Hilbert transform calculated phase is displayed)
Although you can look directly at an excess phase graph to see if something is minimum phase its visually hard to spot it as you're looking for an absense of changes in slope (kinks) on an already sloping line, which is only a straight line on a linear frequency scale.
Looking at excess group delay (essentially the first derivative of excess phase) is far easier - you can use a normal log frequency scale and if the line is flat and horizontal (regardless of vertical position) there is no excess phase and the response is minimum phase. If the line is not flat and horizontal its not minimum phase, and you can see at what frequencies it is not.
On a minimum phase system the excess group delay line will still be flat regardless of any peaks and dips in the amplitude response, allowing you to "see through" any imperfections in driver response etc...
Try the same measurement on a multi driver crossed over speaker and you will clearly see both a peak in excess group delay near the crossover frequencies and an offset in the height of the line in the middle of the different drivers passbands which will show you in milliseconds precisely how far out the accoustic centre alignment of the different drivers is at the listening point...
The frequency and phase response will be different at every measuring point, but that has no bearing on whether its minimum phase or not.
A single driver is usually very close to minimum phase (but not some "special" types like whizzer cone drivers that are really two mechanically crossed over cones - they are not minimum phase) while any multiway system is nearly always not minimum phase with a few special exceptions.
This is determined by the drivers and crossover design, but not affected by diffraction.
Try some excess phase / excess group delay measurements on a variety of speaker systems and it can be quite revealing! On a multiway system you can get a good idea of crossover frequencies and slopes, and driver accoustic alignment all from a single measurement without disabling or close miking any drivers.
On a whizzer cone driver you can see the cones actual mechanical crossover frequency and slope, which is difficult to determine any other way.
Last edited:
DBMandrak, I think you ignore the fact of the fundamental assumptions of those calculations. Since whether or not diffraction is a minimum phase phenomena or not really will not lead to improvement of audio reproduction, I do not feel it is meaningful to spend more time on the issue. The only importance is how it effects audio reproduction. Understanding the phenomena helps more efficient solutions to be developed.
"whether or not diffraction is a minimum phase phenomena or not" IS "Understanding the phenomena".