25.10.2018, 10:12
Lustigerweise hat Toole zu dem Thema vor wenigen Tagen in einem US Forum quasi eine schöne Zusammenfassung als Antwort zu einen "Transieten-Jünger" geschrieben:
Sorry for the delayed response, but life intervened. I’m glad that you have found some of my work useful.
I assume that you have not read my book or papers, because the answer to your “transient” and waveform issues are there, and in numerous papers by others referenced in the book. You are not the first person to believe that “transients” and the time domain, phase response, etc. are the “missing ingredients” to good sound.
After a bit of thought, I realized that this is a topic arises from time to time in forum discussions, so I decided to be more thorough than normal for a forum comment. A version of what follows will eventually make its way to the companion website to my book as an entry in a new FAQ category.
With respect to the importance of transients it is relevant that my PhD research topic was sound localization, involving countless experiments using binaural transients. Here are publications from my thesis years:
Sayers, B. McA. and F.E. Toole (1964). “Acoustical Image Lateralization Judgments with Binaural Transients”, J. Acoust. Soc. Amer., vol. 36, pp. 1199-1205.
Toole, F. E. and Sayers, B. McA. (1965a). “Laterization Judgments and the nature of Binaural Acoustic Images”, J. Acoust. Soc. Amer., vol. 37, 319-324.
Toole, F. E. and Sayers, B. McA. (1965b). “Inferences of Neural Activity Associated with Binaural Acoustic Images”, J. Acoust. Soc. Amer., vol. 37, 769-779.
It is very satisfying to me that this work is acknowledged, including a figure, in what is widely respected as the reference document on spatial perception:
Blauert, J. (1996). “Spatial Hearing: The Psychophysics of Human Sound Localization”, MIT Press, Cambridge, Mass.
The answer to your challenge about “transients” and waveforms is that we both are right. (You) Transients are important in sound localization. (Me) Waveforms can be altered by phase shifts without significant audible degradation. Small time differences are perceptible in interaural differences – binaural hearing - but that does not mean that those small time/phase/group delay differences are consequential in the sounds generating those binaural differences.
For binaural localization – in normal hearing, not stereo listening - at frequencies below about 2 kHz, both interaural time and amplitude differences are related to localization. Obviously, at very long wavelengths neither matter. At shorter wavelengths the onset of the signal matters, as well as the continuing portion. Both have the potential of providing confirming directional information, but both are corrupted by reflections in normal hearing and recordings. Nobody records or plays back music in anechoic spaces so reliability in this frequency range is low. Above about 2 kHz, the ears cannot resolve the “carrier” information in the time domain but instead attend to envelope information. In terms of sound quality – timbre – the spectral content does matter, however. In real-world reflective venues we rely on cues from these brief acoustical events to localize sound sources, which, once localized will continue to be perceived as being in those locations after the trustworthy sounds have passed – a kind of flywheel effect. It is dominant in venues like concert halls where pizzicato, and percussive sounds allow us to localize some instruments, while the sounds of others, lacking such transient elements, simply are broadly localized, sometimes seeming to fill the hall.
My early papers conclude that the ears perform what amounts to a Fourier transform on repetitive impulses. In those experiments, in headphone listening to 200 Hz impulses, listeners could simultaneously track an impulse or transient image, a “buzz”, and also the fundamental and up to four harmonics of the impulses which sounded like pure tones and which followed their own very different trajectories as a function of interaural time difference. The buzz image reached the full lateral position in 0.7 to 1.0 ms and stayed there, while at longer interaural delays the tonal harmonics oscillated from side to side in a manor appropriate to their periods. In elaborate experiments these could be individually manipulated. I remember doing Fourier calculations using a slide rule. Times have certainly changed! This discovery provided insight into how our binaural hearing system works. However, the dominant localization to a casual listener was the high frequency envelope information – the buzz. We have no knowledge of how or if the additional tonal harmonic information is used by the brain and for what purpose (spaciousness?).
As for one’s sensitivity to interaural differences, you are right that it is possible to demonstrate reactions to very small time-differences in the horizontal plane close to the median plane (0 or 180 deg). That sensitivity deteriorates for sources away from the forward/rear axis. This is sometimes thought to mean that such small differences are audible in the waveforms of musical sound. They aren’t.
All that is normal hearing. In stereo listening, only the sound from hard-panned L & R images are the only sounds to be delivered to a listener’s ears without being degraded by stereo itself – they have the highest sound quality of any image on the soundstage - one sound arriving at each ear. Any image perceived to be between the loudspeakers, including the featured artist at center, is a creation of amplitude or time panned double mono – the sound from both loudspeakers arrives at both ears, with amplitude and delay head-diffraction effects added. The result, as shown, in Figure 7.2 in my book, is that the acoustical interference at the ears generates a near 10 dB octave-wide frequency response dip around 2 kHz for a center image. Naturally phase is corrupted for any of the “soundstage” images other than hard left and right pans. This fact is obviously not known to fans of linear phase who, like the rest of us, listen to stereo. It is another good argument for a center channel.
The audibility of phase shift is discussed at some length in Section 4.8 of the book. There are several references to serious scientific efforts to determine the audibility of phase in monophonic signals (as would be the case in real life, as opposed to stereo listening). I quote my summary: “In every case it has been shown that, if it is audible, it is a subtle effect, most easily heard through headphones or in an anechoic chamber, using carefully chosen or contrived signals. There is quite general agreement that with music, reproduced through loudspeakers in normally reflective rooms, phase shift is substantially or completely inaudible. When it has been audible as a difference, when it is switched in and out, it is not clear that listeners had a preference.” Perhaps it is because in real life virtually everything we hear is amplitude and phase modified by reflections. They are normal.
My early loudspeaker evaluations published in 1986 show no evidence of phase being a factor in listener preferences. However, narrow-band glitches in phase responses associated with resonances were highly significant. See Figure 5.2 in my book or go to the source: Toole, F. E. (1986). “Loudspeaker measurements and their relationship to listener preferences”, J. Audio Eng. Soc., 34, pt.1, pp. 227-235, pt. 2, pp. 323-348.
As for the ability of humans to discern small differences in time-domain behavior, the evidence from studies of resonance detection thresholds indicates that ringing is not the dominant perceptual factor. The most reliable indicator of an audible resonance is the spectral bump in the frequency response. See Figure 4.10 in my book, and associated text. Or: Toole, F. E. and Olive, S.E. (1988). “The modification of timbre by resonances: perception and measurement”, J. Audio Eng. Soc., 36, pp. 122-142.
A waveform is completely described by the transfer function, amplitude and phase. For the integrity of waveforms both must be unaltered. We can easily demonstrate that humans are highly responsive to small amplitude changes, but not to even very large phase changes. This means that, in a fundamental sense, we do not “hear” waveforms.
You say: “Everyone should know that electrical capacitance and inductance kill transient information!” Actually, this kind of “murder” is not only legal but encouraged, as these crossover elements are used in combination with the complex impedances and complex acoustical outputs (complex meaning that they incorporate the equivalent effects of inductance and capacitance) of loudspeakers to ensure proper acoustical summation at all frequencies through the crossover regions.
You say: “Thus, besides spinoramas, loudspeakers should also specify their time-domain accuracy or resolution!” Loudspeaker transducers are minimum-phase devices, so when we see flat and smooth amplitude responses we know that the time domain is under control. An exception occurs in crossover regions, which can include non-minimum phase behavior and group delay because of driver properties and placement, but humans are notoriously insensitive to that (about 2 ms), as well as phase shift. However, we do respond to amplitude response aberrations caused by poor acoustical summation in the crossover regions – which are revealed clearly in the spinorama data, and to the ears when listening in normally reflective rooms.
The information content of the spinorama is such that Dr. Sean Olive was able to predict from anechoic data preference ratings that had a 0.86 correlation with the results of double-blind tests for 70 loudspeakers of all sizes and prices. For loudspeakers of similar size, and bass response, the correlation was 0.995 – perfection. These are predictions from anechoic measurements being compared to subjective evaluations done in an acoustically normal room. Whatever effects the room had they did not swamp the inherent performance attributes of the loudspeakers. See Section 5.7 in my book, or: Olive, S.E. (2004a). “A multiple regression model for predicting loudspeaker preference using objective measurements: part 1 – listening test results”, 116th Convention, Audio Eng. Soc., Preprint 6113. Olive, S.E. (2004b). “A multiple regression model for predicting loudspeaker preference using objective measurements: part 2 – development of the model”, 117th Convention, Audio Eng. Soc., Preprint 6190.
Conclusion: the linear distortion (i.e. amplitude and phase) information we need can be inferred from the spinorama. In stereo, the key factor is that the L & R loudspeakers deliver identical direct sounds to the listener. Phase perfection in those signals is not a requirement.
Finally, null testing. This metric is a waveform-comparison test, a comparison of input and output, and as such, it lumps all possible contributors to waveform change into a single “error” signal. As such it suffers from a common problem; it is a linear measure of performance and human listeners are not linear in their perceptions. The “error” signal will include evidence of differences between input and output waveforms that are caused by factors that are not audible.
In the previous discussion it is pointed out that phase shifts change waveforms, but phase shifts by themselves are not significantly audible. So, an “error” may be seen that is caused by something that is not audible. With loudspeakers such a test is impractical because in multi-way systems phase behavior is meaningful only in the direct sound, in the far-field of the source, at a point in anechoic space – no reflections.
Non-linear distortion has long been problematic because the numbers generated in harmonic and intermodulation distortion measurements don’t correlate well with perceptions – except for zero % . Simultaneous masking in human listeners means that all measured distortion products are not heard, meaning that the numbers are wrong. This is discussed in Section 4.9 in my book. So, a null test may show some “errors” that are not audible.
These are reasons to employ analytical metrics of performance, so that we can identify detection thresholds for individual factors, and apply them to the efficient and effective design of products.
Do you have references to scientific experiments supporting your opinions? It is good to have chewy discussions.
Enjoy your loudspeakers, whatever they are.
Wenn man es auf Deutsch lesen möchte kann ich https://www.deepl.com/translate sehr empfehlen.
Sorry for the delayed response, but life intervened. I’m glad that you have found some of my work useful.
I assume that you have not read my book or papers, because the answer to your “transient” and waveform issues are there, and in numerous papers by others referenced in the book. You are not the first person to believe that “transients” and the time domain, phase response, etc. are the “missing ingredients” to good sound.
After a bit of thought, I realized that this is a topic arises from time to time in forum discussions, so I decided to be more thorough than normal for a forum comment. A version of what follows will eventually make its way to the companion website to my book as an entry in a new FAQ category.
With respect to the importance of transients it is relevant that my PhD research topic was sound localization, involving countless experiments using binaural transients. Here are publications from my thesis years:
Sayers, B. McA. and F.E. Toole (1964). “Acoustical Image Lateralization Judgments with Binaural Transients”, J. Acoust. Soc. Amer., vol. 36, pp. 1199-1205.
Toole, F. E. and Sayers, B. McA. (1965a). “Laterization Judgments and the nature of Binaural Acoustic Images”, J. Acoust. Soc. Amer., vol. 37, 319-324.
Toole, F. E. and Sayers, B. McA. (1965b). “Inferences of Neural Activity Associated with Binaural Acoustic Images”, J. Acoust. Soc. Amer., vol. 37, 769-779.
It is very satisfying to me that this work is acknowledged, including a figure, in what is widely respected as the reference document on spatial perception:
Blauert, J. (1996). “Spatial Hearing: The Psychophysics of Human Sound Localization”, MIT Press, Cambridge, Mass.
The answer to your challenge about “transients” and waveforms is that we both are right. (You) Transients are important in sound localization. (Me) Waveforms can be altered by phase shifts without significant audible degradation. Small time differences are perceptible in interaural differences – binaural hearing - but that does not mean that those small time/phase/group delay differences are consequential in the sounds generating those binaural differences.
For binaural localization – in normal hearing, not stereo listening - at frequencies below about 2 kHz, both interaural time and amplitude differences are related to localization. Obviously, at very long wavelengths neither matter. At shorter wavelengths the onset of the signal matters, as well as the continuing portion. Both have the potential of providing confirming directional information, but both are corrupted by reflections in normal hearing and recordings. Nobody records or plays back music in anechoic spaces so reliability in this frequency range is low. Above about 2 kHz, the ears cannot resolve the “carrier” information in the time domain but instead attend to envelope information. In terms of sound quality – timbre – the spectral content does matter, however. In real-world reflective venues we rely on cues from these brief acoustical events to localize sound sources, which, once localized will continue to be perceived as being in those locations after the trustworthy sounds have passed – a kind of flywheel effect. It is dominant in venues like concert halls where pizzicato, and percussive sounds allow us to localize some instruments, while the sounds of others, lacking such transient elements, simply are broadly localized, sometimes seeming to fill the hall.
My early papers conclude that the ears perform what amounts to a Fourier transform on repetitive impulses. In those experiments, in headphone listening to 200 Hz impulses, listeners could simultaneously track an impulse or transient image, a “buzz”, and also the fundamental and up to four harmonics of the impulses which sounded like pure tones and which followed their own very different trajectories as a function of interaural time difference. The buzz image reached the full lateral position in 0.7 to 1.0 ms and stayed there, while at longer interaural delays the tonal harmonics oscillated from side to side in a manor appropriate to their periods. In elaborate experiments these could be individually manipulated. I remember doing Fourier calculations using a slide rule. Times have certainly changed! This discovery provided insight into how our binaural hearing system works. However, the dominant localization to a casual listener was the high frequency envelope information – the buzz. We have no knowledge of how or if the additional tonal harmonic information is used by the brain and for what purpose (spaciousness?).
As for one’s sensitivity to interaural differences, you are right that it is possible to demonstrate reactions to very small time-differences in the horizontal plane close to the median plane (0 or 180 deg). That sensitivity deteriorates for sources away from the forward/rear axis. This is sometimes thought to mean that such small differences are audible in the waveforms of musical sound. They aren’t.
All that is normal hearing. In stereo listening, only the sound from hard-panned L & R images are the only sounds to be delivered to a listener’s ears without being degraded by stereo itself – they have the highest sound quality of any image on the soundstage - one sound arriving at each ear. Any image perceived to be between the loudspeakers, including the featured artist at center, is a creation of amplitude or time panned double mono – the sound from both loudspeakers arrives at both ears, with amplitude and delay head-diffraction effects added. The result, as shown, in Figure 7.2 in my book, is that the acoustical interference at the ears generates a near 10 dB octave-wide frequency response dip around 2 kHz for a center image. Naturally phase is corrupted for any of the “soundstage” images other than hard left and right pans. This fact is obviously not known to fans of linear phase who, like the rest of us, listen to stereo. It is another good argument for a center channel.
The audibility of phase shift is discussed at some length in Section 4.8 of the book. There are several references to serious scientific efforts to determine the audibility of phase in monophonic signals (as would be the case in real life, as opposed to stereo listening). I quote my summary: “In every case it has been shown that, if it is audible, it is a subtle effect, most easily heard through headphones or in an anechoic chamber, using carefully chosen or contrived signals. There is quite general agreement that with music, reproduced through loudspeakers in normally reflective rooms, phase shift is substantially or completely inaudible. When it has been audible as a difference, when it is switched in and out, it is not clear that listeners had a preference.” Perhaps it is because in real life virtually everything we hear is amplitude and phase modified by reflections. They are normal.
My early loudspeaker evaluations published in 1986 show no evidence of phase being a factor in listener preferences. However, narrow-band glitches in phase responses associated with resonances were highly significant. See Figure 5.2 in my book or go to the source: Toole, F. E. (1986). “Loudspeaker measurements and their relationship to listener preferences”, J. Audio Eng. Soc., 34, pt.1, pp. 227-235, pt. 2, pp. 323-348.
As for the ability of humans to discern small differences in time-domain behavior, the evidence from studies of resonance detection thresholds indicates that ringing is not the dominant perceptual factor. The most reliable indicator of an audible resonance is the spectral bump in the frequency response. See Figure 4.10 in my book, and associated text. Or: Toole, F. E. and Olive, S.E. (1988). “The modification of timbre by resonances: perception and measurement”, J. Audio Eng. Soc., 36, pp. 122-142.
A waveform is completely described by the transfer function, amplitude and phase. For the integrity of waveforms both must be unaltered. We can easily demonstrate that humans are highly responsive to small amplitude changes, but not to even very large phase changes. This means that, in a fundamental sense, we do not “hear” waveforms.
You say: “Everyone should know that electrical capacitance and inductance kill transient information!” Actually, this kind of “murder” is not only legal but encouraged, as these crossover elements are used in combination with the complex impedances and complex acoustical outputs (complex meaning that they incorporate the equivalent effects of inductance and capacitance) of loudspeakers to ensure proper acoustical summation at all frequencies through the crossover regions.
You say: “Thus, besides spinoramas, loudspeakers should also specify their time-domain accuracy or resolution!” Loudspeaker transducers are minimum-phase devices, so when we see flat and smooth amplitude responses we know that the time domain is under control. An exception occurs in crossover regions, which can include non-minimum phase behavior and group delay because of driver properties and placement, but humans are notoriously insensitive to that (about 2 ms), as well as phase shift. However, we do respond to amplitude response aberrations caused by poor acoustical summation in the crossover regions – which are revealed clearly in the spinorama data, and to the ears when listening in normally reflective rooms.
The information content of the spinorama is such that Dr. Sean Olive was able to predict from anechoic data preference ratings that had a 0.86 correlation with the results of double-blind tests for 70 loudspeakers of all sizes and prices. For loudspeakers of similar size, and bass response, the correlation was 0.995 – perfection. These are predictions from anechoic measurements being compared to subjective evaluations done in an acoustically normal room. Whatever effects the room had they did not swamp the inherent performance attributes of the loudspeakers. See Section 5.7 in my book, or: Olive, S.E. (2004a). “A multiple regression model for predicting loudspeaker preference using objective measurements: part 1 – listening test results”, 116th Convention, Audio Eng. Soc., Preprint 6113. Olive, S.E. (2004b). “A multiple regression model for predicting loudspeaker preference using objective measurements: part 2 – development of the model”, 117th Convention, Audio Eng. Soc., Preprint 6190.
Conclusion: the linear distortion (i.e. amplitude and phase) information we need can be inferred from the spinorama. In stereo, the key factor is that the L & R loudspeakers deliver identical direct sounds to the listener. Phase perfection in those signals is not a requirement.
Finally, null testing. This metric is a waveform-comparison test, a comparison of input and output, and as such, it lumps all possible contributors to waveform change into a single “error” signal. As such it suffers from a common problem; it is a linear measure of performance and human listeners are not linear in their perceptions. The “error” signal will include evidence of differences between input and output waveforms that are caused by factors that are not audible.
In the previous discussion it is pointed out that phase shifts change waveforms, but phase shifts by themselves are not significantly audible. So, an “error” may be seen that is caused by something that is not audible. With loudspeakers such a test is impractical because in multi-way systems phase behavior is meaningful only in the direct sound, in the far-field of the source, at a point in anechoic space – no reflections.
Non-linear distortion has long been problematic because the numbers generated in harmonic and intermodulation distortion measurements don’t correlate well with perceptions – except for zero % . Simultaneous masking in human listeners means that all measured distortion products are not heard, meaning that the numbers are wrong. This is discussed in Section 4.9 in my book. So, a null test may show some “errors” that are not audible.
These are reasons to employ analytical metrics of performance, so that we can identify detection thresholds for individual factors, and apply them to the efficient and effective design of products.
Do you have references to scientific experiments supporting your opinions? It is good to have chewy discussions.
Enjoy your loudspeakers, whatever they are.
Wenn man es auf Deutsch lesen möchte kann ich https://www.deepl.com/translate sehr empfehlen.