DSR and NPS
Radiation (by Renato Giussani)
The concept of audio frequencies "Spectral Distribution" is applicable both to DSR (Distributed Spectrum Radiation) and NPS (Natural Perspective System), following two different ways that we will simply call "Horizontal Distribution" and "Vertical Distribution".
P
Soundstage :
It
is the "acoustical volume" which surrounds the listener, mainly in
front of him. It is all the physical
space that one thinks is occupied by real, virtual (i.e. reflected sources) and
unreal sources (i.e. sources that are created by listener's brain, for example
the acoustical central source when a monophonic signal is played by a stereo
system, or every source which is created by complex psychoacoustical
effects).
- While listening to a live musical performance: the main acoustical sources are real.
- While listening to a stereo or Dolby Surround system: the main acoustical sources, considering to the most valuable audio
reproduction systems, are virtual and unreal sources.
Acoustical
image changes upon different playing systems, their shapes and dimensions,
their peculiar radiation pattern, their
impulse response, environmental conditions , listener, his spatial
position, and his experiences concerning real and reproduced music, his
subjective psychological or physical feelings while he is listening...
It
is stated that acoustical image has its wideness (that should be
"stable"), its tallness (that is often undefined) and its depth,
inside that one can easily or not, identify many "acoustical planes".
Horizontal DSR
In any stereophonic system
the maximum acoustical wideness is equal to the distance between loudspeakers.
Listener, occupying a spatial
position, perceives the acoustical field as a summation of direct acoustical
field (generated directly from loudspeaker) and reverberated
acoustical field which is due to all possible reflections in the room.
In typical rooms reverberated
field dominates direct field in low frequencies band while at high frequencies
direct field is stronger than reverberated field.
The
problem
If distances between listener
and both loudspeakers, are equal and if loudspeakers play the same signal, he
will perceive an unreal source placed in the middle point between loudspeakers.
Of course if listener is not exactly placed at the same distance between
loudspeakers, he will listen louder to the closer loudspeaker. It is important
to say (considering the total acoustic field) that only at high frequencies
(1000/2000 Hz), we
can assume this statement, while at low frequencies room modes can produce
either attenuation or boost.
When listening to traditional
constant directivity or omni-directional or wide radiating loudspeakers, in a
non axis-symmetrical position we can think that we will have two kind of
distortion:
- perspective,
it is due to wrong localization of unreal sources which will be closer to the
closer loudspeaker.
This is true for all kind of
sources except for virtual ones that are generated by “only left” or “only
right” channel so that the wideness of acoustical image does nor change but it
is distorted becoming more compressed on one side and more thinned out on the
other one;
Acoustic sources shift with listener's movements (after Bauer)
- timber, it is due to the lower pressure level coming from the far loudspeaker and a higher from the closer at high frequencies.
The solution
In a classic paper, Stevens
and Neumann (1) showed that to localize acoustical sources, human auditory
system uses both timing and amplitude information. For example if we have two acoustical
sources, unreal source position will be closer to the source which is playing
before or louder than the other one.
Experiments on subject showed
that timing information are more important at frequencies below 1500 Hz while
at frequencies above 3/4000 Hz, amplitude information are dominant.
Since at low frequencies room
modes are dominant and not negligible and reverberating field does not contain
any directional information, it seems interesting to get a
audio reproduction system that at mid-high frequencies can reproduce a right
localization sound field.
The innovative approach of
Horizontal DSR to solve problems linked to non equal distance between listener and
loudspeakers lies in these two claims:
1) In a typical room the localization
of unreal acoustical source is due to the amplitude difference between the left
and right channel at frequencies over 1000/2000 Hz;;
2) System must compensate perspective
distortion so that timbre is unchanging onto the area that is supposed will be
occupied by listeners.
As it was told, constant timbre reproduction at mid-high frequencies can be achieved by orienting loudspeaker maximum amplitude axis toward the opposite edge of the area that will be occupied by listeners. This is well explained in the following figure:
Fig.1
here
they are two loudspeakers, one unreal source "V", and listener
position 1 and 1' and vector which length is the acoustical pressure amplitude
due to the direct sounds emitted by each loudspeaker. We want that variation of
the amplitude of mid-high frequencies moving fro 1 to 1', will be compensated
by an opposite variation that will be a function of the emitting angle.
Orienting emitting lobes as in figure 1, produces exactly what we want and
rightly dimensioning parameters, can lead to compensate moving from 1 to 1' at
a given distance.
When the same signal drives
both loudspeakers, listener can ear acoustical signals, summation of direct and
reverberated field, coming from "L" and "R", almost equal
in a reasonable listening area: in this way we can achieve that perspective of
the acoustical image does not change as well as timbre, for each source.
ESB 7/06 loudspeakers have
the right orientation (34°) for a listening distance assuming to be 1.5 times
the distance between loudspeakers. Listening central point in front of woofer buffle which is 18° horizontally tilted. Such condition is
found in Kates (2) in tab n. 1 assuming Y/D1= 3 and
theoretically solution should have and amplitude of -3dB for lobe at 90°.
Since we took in account
reverberating field in common listening room and unchanging timbre condition
into the listening area, in ESB 7/06 lobe was modelled such as it varies from
110° at 2000Hz to 60° at 12,5kHz, reaching 90° at 4000Hz.
As final result we
build loudspeakers which, despite of conventional ones, have an oriented
maximum pressure axis, and a decreasing polar horizontal response to achieve
the aims.
This choice leads to two main
advantages:
1) acoustical horizontal image can be rightly perceived in all their
sources as virtual as real, in every listening position; 2) timbre
perception is coherent in every listening position.
7/06
loudspeakers are tall comparing with their length and their depth furthermore
speakers are placed far from each other. For example the distance between
woofer and midrange centres is 63 cm: a distance which is not easy to find in
other systems.
It is known that to achieve
the maximum radiation possible, one can place speakers to a distance which
should be not less than ½ wavelength at crossover
frequency. In DSR system, distance is choose to be
almost equal to the wavelength.
The reason for this choice is
based upon some considerations about resolution capability of human auditory
system and its influence regarding vertical angle and frequency (
Real acoustical sources are
placed in a three-dimensional geometrical space and they have three dimensions
too (they occupy a finite space volume). Human ear can condition signals coming
from sources and from horizontal or vertical directions and focusing to space
placement, one can pay his attention to a specific source separating it from
other ones (i.e. speaking to someone in a crowded room in a so called
"Cocktail Party Effect").
Regarding to a artificial acoustic source (in a loudspeakers system)
which plays all signals in a single spatial point, this signal treatment is not
possible anymore. If one spreads in vertical directions emitting sources (and
so without interfering with the horizontal stereo effect), so that at different
signals we have different emitting zones, we can give to human ear the
possibility to separate virtual sources working both with the spectral
composition and reception angle. Such situation is more realistic than the
assumption of the simple sphere radiation which collapses three-dimensional
emission into a single point.
At higher distance (up to
several meters) can result to the impossibility for the human ear to give a
spatial source coherence, spectral composition of signal will appear as emitted
by completely different sources, loosing the capability of a credible signal
reconstruction, it is important that spread sources must provide to the ear,
the perception of a single acoustical source in a well defined spatial
position. This condition must be achieved for the emitting range of each unreal
acoustical source, independently from the maximum vertical dimension and from
the tallness that the listener will think it will be placed.
Spreading sources in the
vertical direction and distributing audio spectrum as a reception angle
variable, leads to this advantages:
1) capability
to separate complex musical signal in several elementary signals;
2) acoustical
vertical image will have a more realistic vertical dimension;
3) emitting zones will have the right dimension as emitting
spectral content, is coherent with real situation. Choosing to spread audio
spectrum in both horizontal and vertical directions, one can obtain the result
to have a reproduction system which will produce an acoustical image that is a
three-dimensional and stable one, so that loudspeakers will "disappear",
giving to the listener the illusion to experience a real musical event.
Vertical NPS
In NPS, the
vertical dimension of acoustical image, is achieved by
making reproduce the different intervals in which the audio spectrum is divided
and reproduced by the several ways, from emitting zones that will have a
vertical dimension approximating its middle-band frequency wavelength. Emitting
zone centres of the intervals of the reproduced audio spectrum (physically they
are the ways of loudspeaker system, such as the specific speaker or group of
them for each way) are placed at a short distance, in a sequence so that
frequency of emitting spectrums increase as tallness from the floor.
NPS and acoustical depth control...
|
I
would like to focus to some details regarding horizontal DSR radiation which
justify the greater acoustical image depth when listening to NPS1000 versus
ESB 7/06. First of
all, let me remind you some basic elements about horizontal DSR. If we
place loudspeakers so that maximum sound pressure axis is oriented
orthogonally with the front wall (it's the wall just behind loudspeakers …) ,
it is trivial to see: when a
listener is in a spatial position such as his distances between the two
loudspeaker are equal (this point is in a segment which is perpendicular to
the line defined by the two loudspeakers), he will be able to listen to a
sound which spectrum will be given by the reverberating field and the direct
field, as and angle which will be function of the listening distance. If the
triangle formed by Left speaker/Right Speaker/Listening Point is equilateral,
this angle is 30° Now
let the listener move laterally. Overall
timbre response for all the sources (coming from left, right and
central) will be compromised due to an alteration of the high
frequencies level and more to some tilting due to the number of speakers in s
multi-way system. Let start
from a central position, if loudspeakers (we will assume they have a
rectangular base shape) are parallel to the frontal wall, when listener moves
laterally, he will perceive a direct sound field with a stronger acoustical
spectrum at high frequencies coming from the closer speaker and a weaker from
the other. And distances between him and loudspeaker will be different. But
it is common to place loudspeaker such as their axes are orientated exactly
toward the listener. In this way the result will be different than the
previous example. When listener moves, the high frequencies level will drop
down for both loudspeakers, causing a high frequencies lacking timbre
response at lateral positions for all sources, while level changing is the same as
changing distance. As a
result, we will have a timbre distortion and a perspective
distortion: unreal sources coming from all acoustical image,
will be coming from the listener side. Horizontal
DSR is aimed to correct both problems. Perspective for all sources and timbre
for almost central sources. This
result is achieved building up speakers systems that will have a more
coherent horizontal distribution on a broad frequencies band (such as changing
receiver direct field angle, frequency response will behave as straight line
based upon around 500 Hz) and with maximum level axes such as they intersect
in a central point that is placed forward the listener. In this way, when
listener moves laterally we will listen to a sound field that will have a
stronger high frequencies spectrum coming from the more far
loudspeaker. In this way it is possible to correct perspective distortion in
a similar way as the balance control (working only with the higher part of
audio spectrum) and correcting timbre distortion for central sources (often
the most important one), as its sound is always given by adding both channels
(one is more “rich” of frequencies while the other is more “poor”, but result
must remain the same as the one that we can perceive coming from
centre). Let’s
go a little further. Unreal
sources that are subject of perspective correction,
are the one whose spectrum keeps middle-high frequencies information which
are the ones that will have a variation distribution given by DSR. Correction
will be stronger or less due to the importance that have
these sources in the total spectrum. During a live recording, far real and
virtual sources will have a poor high frequencies spectrum than the closer ones, with DSR they will have a right position. If
you try to design a typical listening case as above and place two unreal
central sources (two violins?) behind the frontal wall, one closer (for example
3 meters in front of you) and the other one more far (10 meters), in a
central position you will listen to them as they are almost superimposed in
front of you, but it will be hard to argue the right distance, since the only
differences between the two signals are different amplitude levels and
different spectrum (but they could be generated by two different violins
played in different ways, for example one by one). Now let us
move slightly laterally. The closer source, thanks to DSR system, will be
stable in the centre of loudspeakers, while the far one will move in the
direction between the two loudspeakers, from the same side as you are moving.
If you draw two lines from your position to each source, as showed in the
figure above, you will see that the straight line from you to the far source,
will intersect the line between the loudspeaker, not at centre but at you
side. In this
way the brain can assume that the unreal source is farther in fact the source
has a narrow frequency spectrum and a lower level. If this effect is performed for all signals of
a big orchestra with a DSR that is less compensated as frequency decreases
such as the NPS of 1000 (thanks to small loudspeakers since 140Hz and a front
panel very narrow), the depth of the
acoustical scene is increased. And this is true al for monophonic signals. |
Let us add the “right” environmental frequency spectrum
The NPS
(Natural Perspective System) applied to the NPS 1000, so that the
listener can have a more natural listening sensation of deeper acoustical scene
by controlling directional information as frequency changes as showed above, adds
the capability to adjust radiation angles and cross frequencies of all
loudspeakers to achieve a global environmental response (the sum of the direct
filed and the reverberating field) that will have a bigger decreasing behaviour
with frequency than the classical one suggested by H. Møller.
Optimal frequency response for home hi-fi
systems as suggested in ’70 by H. Møller (B&K).
In this way, the listener,
who will think to perceive an acoustical scene with a well defined depth further
on the frontal wall thanks to the directional information given by NPS 1000, will be immersed in
an acoustical field that is congruent with the reproduced listening distance,
in other words the acoustical filed will have the same trend as the real one,
where the listener was placed at a distance equal to the one reproduced by the
NPS compared with the real acoustical ones. This peculiar working system makes
the NPS 1000 system very suitable when we want to reproduce listening
sensations very similar to the ones that we feel in a natural environment or in
an auditorium or a live rock concert, with the listener placed to the right
distance from music players.
Acoustical Pressure plotted as function of
the distance from an acoustical source with a directivity factor Q,
pressure is plotted for three different value of the room constant.
It easy to see that for small distances,
level decreases of 6 dB as distance doubles,
than it grows up asymptotically to the value of the reverberating field in
that environment.
This trend is similar for each frequency that makes the complex musical
signal, but one must know that increasing frequency, the reverberating field is
lower.
References
1) S.S.
Stevens and E,B. Newman, "The Localization of
Actual Sources of Sound", Am. J. Psych., vol. 48, pp. 297-300 (1936).
2)
J.N. Kates, "Optimum Loudspeaker Directional
Patterns", J. Audio
3)
4) W.B.
Snow, U.S, Patent No. 2, 137, 032 (November 1938).
5) Beranek L., "Acoustics",
6) A.W.Mills, "On the Minimum Audible Angle", J. Acoust. Soc. Am., vol. 30, pp. 237-246 (1958).
7)
B.B. Bauer, "Broadening the Area of Stereophonic Perception",
J. Audio
8) J.Enock, "Loudspeakers for Stereo", Hi-Fi News p.597 (1964 Jan.); and "Stabilising Stereo lmages", Hi-Fi
News,
p. 159 (1967 July).
9)
R.C. Heyser, "Acoustical Measurements by Time
Delay Spectrometry", J. Audio
10)
S.K. Roffler and R.A. Butler, "Factors that lnfluence the Localization of Sound in the Vertical
Plane", J. Acoust.
Soc. Am., vol. 43, pp. 1.255-1.259 (1968).
11) Eli
Osman, "Correlation Model of Binaural Detection:
lnteraural Amplitude Ratio and Phase Variation for
Signal", J. of Acoust. Soc. Am.,vol. 54, pp.386-389 (1973 No. 2).
12)
H. Staffeldt, "Correlation between Subjective
and Data for Quality Loudspeakers", 47th Convention Audio Eng. Soc., 26-29
marzo 1974.
13)
Allison R. F., "Influence of Room Boundaries on Loudspeaker Power
Output", J. Audio
14)
J.M. Kates,
15)
H.D, Harwood, "Some Factors in Loudspeaker Quality", BBC Research
Department, Wireless Worid (1976 may).
16)
Martin Colloms, "High Performance
Loudspeaker", Pentech Press Limited,
17) D.
Queen, "The Effect of Loudspeaker Radiation Patterns on Stereo lmaging and Clarity", J. Audio Eng. vol. 27, pp.
368-379 (1979 May).
18) J. Crabbe, "Broadening the Stereo Seat", Hi-Fi New & Record
Review, pp. 65-68 (1979 June).
19) R. Giussani,
"Proposta di Sistema Stereofonico di Diffusione Sonora Alta Fedeltà di
Nuova Concezione", Comun. Uff.
Ricerca ESB, 2 agosto 1979.
20) R. Giussani,
"Sistemi Stereofonici e loro Limiti nella Ricostruzione della Scena
Acustica Soggettiva.
Proposta di Miglioramento", 1° Seminario di Elettroacustica
e Alta Fedeltà, 3-7 settembre 1981, Milano.
21)
O.K. Petterson and U.R. Kristiansen, "Describing
Acoustic Energy Flow in Two Dimensions by the use of lntensity
Vectors", lnternational Congress on recent
developments in acoustic intensity measurements. 30 Sept. - 2
Oct., Senlis (France).
22)
Duane H. Cooper, "Calculator Program for Head-Related Transfer
Function", J. Audio Eng. Soc., vol. 30, pp. 34-38 (1982 Jan.).
To go deeper inside, see
also:
Formules (in italian)
The cocktail party effect is an interesting phenomena that tells us a lot about how attention can effect how perceptual stimuli are processed. During a conversation at a party, where there are a lot of other conversations occurring, and music nearby, we somehow manage to tune into the voice of the person that we are talking to. All of the other noise is filtered out and largely ignored. This generally happens in all perception: some of the stimulus is filtered out for conscious analysis. This enables us to filter out the rest of the conversation at a party and concentrate on only one person's voice. The 'figure-ground' phenomenon is the separation of the auditory input into the components of figure (the attended signal) and ground (everything else, in the background). However, an interesting point is that if someone over the other side of the room suddenly sees us and calls out our name, we generally notice quite quickly. This suggests that some processing of the other information does occur, enough to often enable to pick up on bits of it in certain situations, for example if it is a familiar voice.
Cherry (1953) discovered that it is based upon characteristics of the speech that we are attending to, and its differences from other sounds that are present. In the case of separating one voice from many in a room, the ability to do this depends on characteristics of the speech that in turn depend on the gender of the speaker, the intensity of the voice and the location of the speaker. Cherry discovered that if a subject is presented different messages in each ear through a pair of headphones, at the same time, if the voice that is used is the same then there is much more difficulty in separating the two messages on the basis of their meaning alone, which is the only cue left. Cherry also discovered that, if one of the messages that the subject was hearing was shadowed, that is, the subject had to repeat what was said in one of the messages out loud, then information from the other message was very rarely extracted. In fact, even when the unattended message was changed somehow, such as changing to a foreign language or being reversed, subjects very rarely noticed. However, if the unattended signal had a non-speech sound suddenly added to it, subjects almost always notice (after Eysenck and Keane, 1994). This explains how we can detect a sound such as tyres screeching when we cross the road, as explained on the timbre page (contrast with previous sounds).
Home