Some Conclusions . . .
Sound reproduction in domestic size living spaces has been
a life-long interest to me. I have learned much from observation and from my own
experimentation. Mostly they confirmed what others already had written about,
such as in the many publications of the AES, or the classic texts of Olson and
Beranek. But in sound reproduction there are still areas that have not been
studied to the point where clear prescriptions or limits can be given as, for
example, to sound distribution in acoustically small spaces, or to audibility
thresholds for non-linear distortion. I have always been fascinated by the
multi-disciplinary approach that has to be taken to the design of loudspeakers,
if the goal is higher than another typical consumer product. That approach must
include old and new understandings in the fields of mechanics, electronics,
acoustics, and psycho-acoustics, as well as extensive experience with test and
So here are some conclusions that I have come to. I will
write them in the form of statements. The background for many of them can be
found on the pages of this website. Other conclusions were drawn from
observations at CES, dealer show rooms and private listening venues. Many of
them are obvious, but I state them since their sonic consequences seem to
be underestimated. There are probably more and I might add them as I think of
- The best one can hope for with 2-channel sound
reproduction is the illusion of listening into the recording venue. Physics
does not allow the accurate reproduction of the original sound field with
only two speakers.
- Since sound reproduction is about creating an illusion
it becomes very important to avoid or minimize any clues that would detract
from the illusion. Such clues come from linear-distortions, such as
frequency and polar response, and from non-linear distortions with their
generation of tones and sounds that were not in the original.
- Linear distortion - frequency response, polar response,
resonance - affects primarily the timbre and clarity of a loudspeaker.
- Non-linear distortion - intermodulation, harmonic,
clipping - affects primarily the maximum tolerable sound pressure level.
- There is a level of non-linear distortion that is
"good enough" relative to other flaws in the loudspeaker. Further
reduction of this distortion brings no audible improvement.
- For accuracy it is necessary to reproduce sound at near
realistic SPL so that the ear generates the correct timbre due to its own
distortion. Loudness control or response shaping gives a poor approximation
to the Fletcher-Munson curves.
- I have heard at the AES Convention in 2000 a 6-channel
(2 front, 2 elevated side/front, 2 side speakers) Ambisonic microphone
recorded (by Chesky) surround demonstration which totally excited me. My
listening experiences since then with SACD and DVD-A commercial recordings
have left me cold and I happily return to two channels with the ORION.
- There are many different loudspeaker designs available
commercially. They all change electrical signals into acoustic signals. But
if the goal is to reproduce sounds accurately, then a speaker must be either
an acoustic point source (monopole, omni-directional) or an acoustically
small bi-directional source (dipole). "Small" means that the
physical dimensions are small compared to the wavelength being radiated or
that the shapes do not interfere with the polar response of the point
- I have not come to conclusions about a line source that
extends floor-to-ceiling, is infinitely long acoustically, and thus
generates a cylindrical wave. It seems that this could be an alternate
approach to illuminating a room uniformly at all frequencies.
- All accurate speakers will essentially sound the same
when listened to in a setup that is appropriate to their specific design.
- Since loudspeakers are listened to in closed spaces
there are fundamentally only two ways in which they should illuminate the room
sound-wise: Either omni-directionally or uniformly directional over the
whole range from low to high frequencies. This allows the delayed, reflected
sounds from the room boundaries to have the same spectral signature as the
The transition in polar radiation from 4p
to 2p (baffle step)
that is typical for the majority of loudspeakers guarantees non-uniform
illumination of the room.
- Reflected sounds are perceptually masked if their
initial delay is >6 ms and if the reflections are full spectrum copies of
the direct sound. This requires omni-directional or dipole loudspeakers that
are free standing in the room.
- Omni and dipole loudspeakers can sound nearly identical
in any given room when properly set up. The room must approach frequency
independent reflection-diffusion-absorption behavior above 100 Hz.
- Conventional box speakers are always omni-directional
at low frequencies and increasingly forward directional at high frequencies
and thus the room reflections color the sound.
- When designing a loudspeaker it is essential to perform
free-space measurements to see the effects of driver directivity and baffle
shape on the important polar response. This requires that any reflecting
surfaces and objects are at least 10' (3 m) away from the source, that the
distance to the microphone is greater than the largest dimension of the
baffle, and that the source is rotated around its acoustic center axis. This
setup can provide a reflection free 10 ms time record and frequency response
data down to 100 Hz. At lower frequencies it is more practical to use
boundary measurements, but their integration with the free-space data
requires thought and experimental verification.
- The 3D free-space response of tweeters and very small
loudspeakers can be measured in typical domestic rooms, if the required
microphone distance is small compared to the reflection path
- In an active loudspeaker system each driver has its own
power amplifier. This gives maximum control over the mechanical motion of
each driver and most efficient use of amplifier power. Drivers of different
sensitivities (SPL/W/m) are easily combined, while with passive crossovers
the driver of lowest sensitivity determines the loudspeaker's overall
sensitivity. Amplifier power has to be wasted in the process.
- The power amplifiers of an active loudspeaker system
see a benign load (resistive, slightly inductive) over their assigned
frequency range, unless it includes the mechanical resonance of the driver
(highly capacitive and inductive). The single and much larger power
amplifier that is required for a passive crossover loudspeaker has to drive
a complex load, which places more stringent requirements on its dynamic
stability and overall performance. Different amplifiers may sound different.
Open baffle loudspeakers
- Open baffle speakers are inefficient in terms of the
mechanical movement that is required to create a given level of sound. This
not only applies to speaker cones but also to panel vibrations.
- Open baffle loudspeakers reproduce bass with less room
interaction. It is more articulate than from box speakers.
- If dipole behavior covers the full frequency range,
then the room response becomes perceptually masked by the direct sound.
- The radiation from the rear of the cone must not be
absorbed, but the distance to the nearest reflecting/diffusing surface
should be at least 3' (1 m).
- An open baffle circumvents the box problems of delayed
radiation through cone and enclosure panels. They occur typically in the
mid-frequency range and are difficult to suppress.
- Large panel radiators or long line radiators suffer
from severe lobing at higher frequencies. It manifests in critical room and
- Even though a dipole requires a 6 dB/oct boost towards
low frequencies, it takes little power to drive it to maximum excursion at
its lowest bass frequencies. Amplifier power could be an issue as frequency
increases, where it requires higher cone acceleration to reach Xmax. Thus
SPL is limited by driver volume displacement at the very lowest frequencies
and becomes amplifier limited as frequency increases.
- Realistic bass levels can be obtained from dynamic
drivers in open baffles, not from panels. For extreme SPL requirements the
number of drivers could get very large and, therefore,
below 50 Hz they are more economically replaced by sealed box subwoofers.
- At frequencies where a 8" driver would become
directional it has wider frontal dispersion for an open baffle than if the
baffle were closed in the back.
- Open baffle speakers reach deeper into the room and are
less subject to the room response if their polar response is well behaved.
- ORION exemplifies open baffle loudspeaker design in
terms of polar response control and dynamic range. It circumvents the
limitations of large panel radiators and yields a small package.
- The low masses of the moving parts in an ESL, a planar
magnetic, or a ribbon driver are necessary to generate useful sound pressure
levels. The force generated by an electrostatic or planar magnetic motor is
weak. Since SPL is proportional to air volume acceleration, and moving parts
Acceleration is Force divided by Mass, the mass has to be lower if the force
is too weak to generate sufficient acceleration. Furthermore, since
excursion is limited with these drivers the radiating area has to be large
to move a sufficient air volume.. These relationships seem to be difficult
to grasp by audiophiles. Marketing departments and even some designers like
to tout low mass as an inherent benefit giving greater "speed" or
frequency response to their speaker, when it is only affecting sensitivity
- It is difficult to screw up an open baffle speaker
design to where it sounds worse than your typical box speaker.
- Small boxes have fewer problems with panel resonances,
cone re-radiation, and polar response than large boxes.
- Box panels can radiate more sound at certain frequencies
than coming from the cone.
- When building boxes from 3/4 inch thick wood, then the
un-braced areas should be less than 4 inch squares to obtain high stiffness
and to push panel resonances into the kHz region and where they can be
decoupled from the driver's structure and airborne vibration.
- The sound behind the driver cone should not come back
out through the cone.
- Typical box speakers have a generic sound due to their
polar response, panel resonances, re-radiation through the cone and vented
- Bass from box speakers has more "punch" than
from open baffle speakers, but is less airy.
- Vented bass speakers are resonant structures and store
energy which is released over time. For accuracy, bass must be reproduced
from sealed or open baffle speakers that are non-resonant.
- Closed box speakers are best listened to from very
close distance to minimize masking from an uneven room response.
- PLUTO is not merely another 2-way box speaker.
Secondary radiation from enclosure panels and through the driver membrane
from inside of the enclosure were eliminated. Being an active speaker, two
drivers of very different sensitivities could be combined in order to obtain
omni-directional radiation. Bass response was extended by equalization and
not by a resonant vent.
- The room is rarely at fault. If it is comfortable for
conversation and living in it, then it is also suited for sound
reproduction. The problem is usually the inadequate polar response of the
loudspeakers and their placement in the room.
- Loudspeakers should be positioned out in the room, at
least 3' (1 m) away from reflecting surfaces. The further the better.
- Speaker placement to the inch based on some room
acoustic calculation is nonsense.
- Rooms should have lots of diffusive elements and not
sound like a stuffed pillow if open baffle or omni speakers are used.
- Placing absorbers at reflection points is the wrong
approach. It only absorbs high frequencies and increases the difference
between the direct sound and the delayed room response. It works against
perceptually masking the room response as merely a copy of the direct sound.
- Equalization for a certain response at the listening
position is fraught with serious problems. DSP can do many things, but which
acoustic inputs to take, and how to process them, is still very much at a
research stage. It will change the sound you hear.
- When I hear an unfamiliar loudspeaker in an unfamiliar
room and it does not sound right, then I look for faults in the
loudspeaker's design and placement long before I blame the room.
- People listen differently. Performing musicians and
members of the audience are used to different perspectives and focus on
different aspects of the sound. Both are valuable for analyzing a
loudspeaker. People who only listen to loudspeakers and thus always compare
loudspeakers are poor judges of accuracy.
- Very few sales people of "high end audio"
ever listen to unamplified life sounds. They are highly susceptible to
marketing department suggestions.
- Unbiased listeners have no difficulty recognizing
accurate sound reproduction, even with hearing damage or with hearing aids.
- Unfortunately, marketing departments and dealers think
that bass and high frequencies need to be emphasized for products to
- Some listeners prefer euphonic loudspeakers. Accurate,
and thus neutral, loudspeakers are not that exciting unless the source
- I find it disappointing when loudspeaker manufacturers
run extensive double-blind listening tests with trained and untrained
listeners where they only compare loudspeakers to each other, but not to any
live source. These are strictly preference tests within a given paradigm.
- A loudspeaker can never do better than to accurately
convert electrical signals into acoustic signals. Thus the source material
determines ultimately how well an illusion can be created.
- Recording is still an art, not a science. Two
loudspeakers in a room cannot reproduce the original sound field. Surround
sound could be science based, but today is far from it and mostly pan-potted
- Lossy compressed recordings (e.g. MP3) lose too much
inner detail when encoded at less than 128 kbps to be perceptually accurate.
This shows up most easily on applause and the least on voice.
- The loudspeaker is by far the weakest link in the
reproduction chain. Unless you have really poorly designed associated
equipment you cannot get significant improvement in accuracy by going to
very expensive equipment. Marketing departments like to tell you otherwise.
Hearing a change is not an indication of greater accuracy. Some products are
designed to make an audible change so the customer will notice it. Other
products rely on the power of suggestion which works the better, the higher
the price tag.
- In Class A/B power amplifiers the time variant and
amplitude dependent crossover
distortion is more harmful than harmonic distortion, because of its
impulsive and thus wideband nature. It does not register in the typical high
signal level harmonic distortion specification. Crossover distortion
changes with bias conditions and is thus a function of the thermal control
loop of the amplifier. It must be tested dynamically as output power
switches from high to low levels and device temperatures change.
This is It