First, welcome to the forum.
In my quick read of that paper, it is about phase differential between two channels of a stereo system. That is a different matter than what it means with respect to drivers in one speaker, i.e. mono, that we are discussing here. Small timing changes can indeed shift sound images in stereo reproduction.
This is where the engineers begin to take over the asylum, and their view of everything in terms of the frequency domain - including the human ear and brain - colours everything they think and do.
He didn't do that because that would be silly. Let me explain because this is a technical topic and its message seems to be completely lost.
What is an all-pass filter? It is a filter where the amplitude and hence frequency response of what goes in, is the same as what comes out. What changes however is the phase response. The phase changes proportional with frequency. Here is a good picture that shows that:
The phenomenon Mitchco was talking about was the time alignment of drivers in a multiway speaker. This is not just a question of phase shift, but actual physical delays. Hint: if I swap a speaker's cables over, what phase shift have I achieved? If I "cascade" 100 such swaps, what phase shift have I achieved? And how much delay?The first set of research data I post in the peer reviewed journal of Audio Engineering Society was authored (in part) by Professors Lipshitz and Vanderkooy. As their title indicates, they teach at the University (Waterloo). They are luminaries in audio industry especially when it comes to matters related to signal processing. These are the names we need to know and not throw them in some random pile of "engineers."
https://uwaterloo.ca/audio-research-group/people-profiles/john-vanderkooy
John Vanderkooy
BEng, PhD (McMaster) – Distinguished Professor Emeritus
PHY 205 Ext: 32223 Email: [email protected]
Research Topics: Acoustics, Audio and Electroacoustics, Digital Signal Processing as it applies to audio; Diffraction and Radiation of sound by loudspeakers; Electro-acoustic Measurement Techniques.
Present Research Activities: Active acoustic absorbers, Noise measurements of Wind Turbines, the Acoustics of the trumpet. There is always an interest in measurement techniques
Measurement systems of various types normally work satisfactory when the system under test is truly linear, but nonlinearities can show up surprising differences. Maximum-length sequence (MLS), swept-tone and pulse systems are under study for the characterizations of transducers and acoustic spaces. As well as experiments, computer simulations are used which allow better control of the nonlinearities.
The diffraction of sound by loudspeaker cabinets has been studied with a view to producing simple predictive methods for the effect of a box. A computer algorithm has been applied to solve the Helmholtz sound equation for various shapes, and theoretical analyses show some asymptotic results that are in general agreement with experiment. Currently the diffraction from cabinet edges is under study. The geometrical theory of diffraction is easily modelled on a computer and is efficient enough that second-order diffraction effects can be included. Low-frequency asymptotic results are being used to correct near-field measurements.
Digital Signal Processing is done in our group with an express emphasis on acoustics and audio. Projects include band splitting for multi-way transducers, active noise cancellation, and measurement systems.
Some significant past papers
U.P. Svensson, R.I. Anderson, J. Vanderkooy. Analytical Time-Domain Model of Edge Diffraction, in Proceedings of NAM98, Nordic Acoustical Meeting, Stockholm, Sweden, September 7-9, 269-272 (1998).
Nonlinearities in Loudspeaker Ports, Presented at 104th AES Convention, Amsterdam, May 16-19, 1998. Preprint 4748, 32 pages.
U.P. Svensson, R.I. Anderson, J. Vanderkooy. A Time-Domain Model of Edge Diffraction Based on the Exact Biot-Tolstoy Solution, in the Technical Report of the Institute of Electronics, Information and Communication Engineers, EA97-39, Tokyo, Japan, September 26, 9-16 (1997).
J. Vanderkooy. Aspects of MLS Measuring Systems. J. Audio Eng. Soc. 42, 219-231. (1994)
S.P. Lipshitz, R.A. Wannamaker and J. Vanderkooy. Quantization and Dither: A Theoretical Survey. J. Audio Eng. Soc. 40, 355-375. (1992)
S.P. Lipshitz, J. Vanderkooy and R.A. Wannamaker. Minimally-audible noise shaping. J. Audio Eng. Soc. 39, 836-852. J. (1991)
Vanderkooy. 1991. A simple theory of cabinet edge diffraction. J. Audio Eng. Soc. 39, 923- 933.
J. Vanderkooy and S.P. Lipshitz. 1990. Uses and abuses of the energy-time curve. J. Audio Eng. Soc. 38, 819-836.
----
https://uwaterloo.ca/audio-research-group/people-profiles/stanley-p-lipshitz
Stanley P. Lipshitz
BSc (Natal), MSc (South Africa), PhD (Witwatersrand) – Professor Emeritus
Lab: PHY 205 Ext: 33755 Email: [email protected]
Research Topics : Audio and electro-acoustics; Digital signal processing for audio; Diffraction and radiation of sound by loudspeakers; Electro-acoustic measurement techniques, Dither in image processing.
Present Research Activities: Both theoretical and experimental investigations are being made into the low-frequency radiation and diffraction of sound by loudspeakers. These are based on the use of the Helmholtz boundary integral formulation, and approximations thereto such as the Rubinowicz diffraction integral. It is hoped that this will lead to improved low-frequency measurement procedures in reverberant rooms.
The use of analogue as well as digital dithering techniques in digital audio, and digital signal processing in general, is being investigated both theoretically and experimentally. Dither can eliminate errors due to data quantization. Spectrally shaped dither signals and the design of dithered noise-shaping quantizers are specific current interests.
The application of digital signal processing techniques to audio promises to revolutionize the field in the next decade. We are pursuing investigations in such areas as room deconvolution, adaptive equalization, and surround- sound reproduction.
Some significant past papers
S.P. Lipshitz and J. Vanderkooy. 1983. A family of linear-phase crossover networks of high slope derived by time delay. J. Audio Eng. Soc. 31, 2-20.
J. Vanderkooy and S.P. Lipshitz. 1987. Dither in digital audio. J. Audio Eng. Soc. 35, 966-975.
J. Vanderkooy and S.P. Lipshitz. 1989. Digital dither: Signal processing with resolution far below the least significant bit. Proc. AES 7th International Conference "Audio in Digital Times", Toronto, Canada, May 14-17, 1989.
J. Vanderkooy and S.P. Lipshitz. 1990. Uses and abuses of the energy-time curve. J. Audio Eng. Soc. 38, 819-836.
S.P. Lipshitz, R.A. Wannamaker, J. Vanderkooy and J.N. Wright. 1991. Non-subtractive dither. Proc. 1991 IEEE ASSP Workshop on Applications of Signal Processing to Audio and Acoustics, New Paltz, NY, October 20-23, 1991.
S.P. Lipshitz, J. Vanderkooy and R.A. Wannamaker. 1991. Minimally audible noise shaping. J. Audio Eng. Soc. 39, 836-852.
S.P. Lipshitz, R.A. Wannamaker and J. Vanderkooy. 1992. Quantization and dither: A theoretical survey. J. Audio Eng. Soc. 40, 355-375.
----
Their work on dither was so significant that chances are really good that the products you use to listen to music has the results of their work in it! (TPDF dither)
Now, it doesn't mean you automatically agree with both of them. You can dispute what they say but you need to rise up to their level of knowledge. The papers are full of mathematics and deep level of analysis. So this is not a task anyone should take lightly.
So let's be done with "what do engineers know" type of comment. Do a bit of work on who these authors are, what their work has been, etc. If that is too hard to do, just ask and I will be happy to explain.
And that is what the research investigates. I have tried to explain it multiple times. If it is still not understood, then chasing this goal is unwise.
I can cite other luminaries who take the opposite view, but that doesn't particularly interest me. I enjoy the idea that they could all be wrong and we could think it through, and come up with an interesting take on the subject. Engineers and scientists are prone to groupthink, and at the end of the day none of them can tell you why we enjoy listening to music, nor how our brains process it. Why are we sometimes 'on fire' when listening to music, and sometimes not interested? And so on.
It doesn't interest me unless their views has this type of evidence prior to the paragraph I posted:
Sorry, that's exactly what the field of Auditory Scene Analysis, ASA, is doing - and making good headway in gaining understanding of the process. Why we're "on fire" at times is that we have an internal model, "schema" is the jargon, of how the sound should be, and when the match up is excellent then the pleasure factor goes sky high - and, the converse is true ...
First, welcome to the forum.
On your comment here, I am not sure what you are saying. Any time difference can be described in terms of amount of phase shift. The two are one in the same.
Here is Clark's experiment on this showing the interchangeable terminology: “Measuring Audible Effects of Time Delays in Listening Rooms,” Clark, David, AES Convention: 74 (October 1983)
It is possible to create a speaker system that will give you good coherency between drivers but not reproduce a steady state waveform fed into it without phase distortion.The best I have heard it put is that, “Time-alignment is absolute where as phase-alignment is frequency dependent. In other words, phase-alignment applies to a specific frequency or band of frequencies, where as time-alignment is applied across the frequency spectrum.” Do you agree with that? Please don’t quote me something, let’s hear your critical thinking
Their data is a deeper dive of what I have been posting. To wit, they used a headphone and synthetic signals to arrive at audibility of phase distortion: "The listening test was performed in a quiet listening room using Sennheiser HD650 headphones." I have shared already that this audibility does exist with headphones and especially with synthesized signals. So this research confirms what we know, and does not dispute it.
Thanks for the post. I have to run and don't have time to post a proper response. But let me say that I now know what you are arguing, even though you continue to state it in improper technical terms.Thanks. Consider these two configurations of the same sound reproduction system.
Config A: 3-way passive minimum phase XO. Each of the drivers acoustic centers are offset in the z- axis. Relative to the tweeter’s measured acoustic center, the midrange acoustic center is 21 samples behind, at 48 kHz sample rate is 1 sample = 0.3” x 21 = 6.3” behind. The woofers acoustic center relative to the tweeters are 2 samples or 0.6”behind. Note that this system is not time aligned, meaning those z-offsets are not corrected, just measured from the linear phase XO in Config B.
Config B: 3-way digital linear phase XO. I take the bass and midrange digital XO, open them up in Acourate and rotate both signals by the number of samples relative to the tweeter sample position from the measurements referenced above. You were once reviewing my eBook, under the advanced chapter is a chapter on Time Alignment – has the process and screen shots of the acoustical peaks being time aligned. In the digital domain, I am moving the impulse peak of the woofer and midrange to line up with the tweeter impulse peak in the time domain.
A couple of important differences, one can observe the mathematical differences between a minimum phase crossover, including delay, versus linear phase crossover in this document. http://files.computeraudiophile.com/2013/1202/XOWhitePaper.pdf Nothing new here, but good foundation review and anyone using a digital FIR filter design software kit will obtain same results.
Note one can use https://sourceforge.net/projects/rephase/ to adjust the amplitude and phase responses of the filter independently for one's loudspeakers to make them phase coherent, but not time aligned. See this article and Example 1: https://www.minidsp.com/applications/advanced-tools/rephase-fir-tool I encourage folks to try it out and see if you can hear a difference. On my passive XO speakers and linearizing the phase, I could not reliably hear a difference. Which agrees with most of the science you have quoted. However, that’s not what I am talking about.
It is with both time alignment and phase coherence that I notice a change in imaging. Using a linear phase crossover represents the phase coherent part, and the aligning the impulse peaks of the drivers is the time alignment or time coherence part. The best I have heard it put is that, “Time-alignment is absolute where as phase-alignment is frequency dependent. In other words, phase-alignment applies to a specific frequency or band of frequencies, where as time-alignment is applied across the frequency spectrum.” Do you agree with that? Please don’t quote me something, let’s hear your critical thinking
As Michael says, maybe we could create a repeatable experiment to validate this…
When I get a chance I will state the points and my response to them.
