Audio Acoustics 6 12 05.ppt 384KB Jun 23 2011 07:19:58 AM
Audio Acoustics
in Small Rooms
By Earl R. Geddes
GedLee LLC
www.gedlee.com
My Background
• My PhD. thesis was on the modal response
of small non-rectangular rooms
– Conclusions:
• The first mode was independent on room shape –
depended only on volume
• Concluded that room shape had little effect, except
in cases of extreme symmetry
• Distribution of absorption was far more important in
the more symmetrical shapes – shape did help to
distribute the damping evenly among the modes.
• Only damping can help the smoothness of the LF
response
Conclusions Con’t
•
•
•
With damping comes loss of LF energy
Above Fs, damping has no modal effect
These conclusions have significant
implications to LF in small rooms
1) substantial LF damping is required for good
LF response smoothness
2) the damping must be well distributed
3) There will be substantial energy loss to
make up for
The small room problem
•
Small rooms have two/three regions of
importance that need to be attended to.
1) The LF modal region
•
•
Modes are discrete
Free waves are not permitted
2) Above Fs (Schroeder frequency)
•
•
Modes are irrelevant
Waves propagate freely – geometrically
3) Transition region is possible
The small room problem
• There is no reason to believe that these three
regions will have the same characteristics,
problems or solutions
• What works in one region may be completely
wrong in another
• Hence, these regions must be approached
independently, finding those solutions that work
for that region and then seeing if the two sets of
solutions can be somehow melded together
The small room problem
• The goal is to produce at the listeners
position a perceived playback of a sound
source that is optimized for imaging,
timbre (coloration) and spaciousness
– Basically the same things that a large room is
designed for
• Since “perception” is involved some
psychoacoustics is relevant
Psychoacoustic Fundamentals
• The perception of image location is
dominated by frequencies in the range of
1kHz – 8 kHz in the Auditory System (AS),
with a greater emphasis on the middle of
this range (See Blauert)
• If good imaging is desired then this range
must be relatively free from frequency
response aberrations, diffraction and wall
reflections < about 10ms.
Psychoacoustic Fundamentals
• Recent work (to be published) has shown
that diffraction and very early reflections
(< 1 ms.) are far more perceptively
important than a spectral analysis of their
effects would indicate
• This is hypothesized to be due to the
masking of the ear being far poorer in the
time domain than the frequency domain
Psychoacoustic Fundamentals
• If this hypothesis is true (the data indicates
that it is) then it is also suspected that these
diffraction effects would be level dependent
• Hence, while the diffraction is linear in a
mathematical sense, its perception may be
highly nonlinear
• To a listener, this diffraction might appear to
be nonlinear distortion since it becomes
more audible at higher levels than lower
levels – yet it has a physically linear cause
Psychoacoustic Fundamentals
• This hypothesis is completely consistent
with another study (AES) that showed that
perceptually nonlinear distortion in
compression drivers is virtually nonexistent
• Yet it seems to be common knowledge
than horns sound worse at higher levels
than lower levels
• Horns add only very low orders of
nonlinearity, but virtually all horns have
diffraction effects in them
Implications
• It then appears that it is not only critical
that the room not have early reflections,
but it is just as important that the sources
not have any near field diffraction from the
cabinets, any waveguide devices, or
nearby structures
• Source diffraction of any sort must be held
as just as undesirable as early reflections
More Psychoacoustics
• As stated before, at low frequencies, we
need not be too concerned about
reflections, and probably diffraction and
we certainly need not be concerned about
these problems down into the modal
region - this region is dominated by the
room and has hardly anything to do with
the source
Why Not?
• A simple solution would seem to be to just
put sound absorbing material everywhere,
or better yet just move outside!
– A non-reflection room is usually not found to
be perceptually adequate – that’s because it
lacks a very important acoustic property
known as spaciousness.
– Spaciousness occupies a large part of the
study and design of spaces for the performing
arts – it is nearly as important in small room
acoustics.
Spaciousness
• To fully understand spaciousness we need to
understand the concepts of direct and
reverberant fields.
• The direct field (not to be confused with the near
field) is where the sound from the source
dominates over the reverberant sound.
– There is a 6 dB/octave falloff with distance
• The reverberant field is when the reverb
dominates
– There is no level dependence on location
100
10
1
0.1
1.0
10.0
Distance from source
100.0
Into the Recording
• The importance of spaciousness can be
easily demonstrated
– tonight if time permits
– By moving closer and closer to speakers that
are canted inward, the sound field becomes
more and more dominated by the direct field –
the direct to reverb ratio goes up.
– Moving back beyond a certain point has no
effect.
Into the Recording con’t
• Moving forward creates a subjective effect that I
call “in the recording”
• Backward - “in the room”
• The former gives the subjective impression of
“being there” – you are moved into the recorded
space
• The later gives the impression that the
musicians have been transported into the room
with you
• Some like the “in the recording” effect, but I find
it unnatural - precise imaging beyond reality, no
spaciousness, a kind of headphone effect
Spaciousness
• Clearly spaciousness does not just
happen, to have it or not have it requires
some design considerations
• Not paying attention to it will likely leave
an audio reproduction with poor imaging
and or a colored sound character along
with a lack of spaciousness
Sources and Spaciousness
• Clearly the first arrival sound should be nearly
flat (a subtle HF roll-off is usually preferred) but
definitely smooth
• What is not commonly attended to is that for
spaciousness to occur there must be a
substantial reverberant field component at the
seating location
• The reverberant field response in a reverberant
room is dependent on the sources power
response – not its anechoic response
Sources and Spaciousness
• Very few loudspeakers have both a flat
anechoic response and a flat power
response.
• That’s because it cannot be achieved with
piston sources - a piston source does not
have a flat power response when it has a
flat anechoic response – it beams at HF
• Pistons can be both, but only below ka=1,
and then only as omni-directional sources
Return to room acoustics
• Now lets consider the source placement in
the room along with the sources directivity
• All sources have negligible directivity at LF
and most have a directivity that varies with
frequency throughout its operating range
– This means that the anechoic response and
the power response cannot match
• Consider a omni-directional speaker and
one with 90° coverage
Sources in rooms
• The omni source will have a multitude of
early reflections while the directive source,
if properly aimed, will have only a single
reflection (horizontal plane), which arrives
at the ear opposite to the direct sound
from this source.
• The opposite ear effect is notable because
it is far less objectionable than a reflection
to the same ear.
Specification of Source
• The sources should have the following
characteristics:
– They should be directive at < 90°
– They need to have off-axis responses that are
flat as well as on-axis
– They need to have a flat power response for
low coloration in a lively room (required for
good spaciousness – to be discussed)
• This is because of a time-intensity tradeoff in the
AS – longer signal times yield louder perception
Specification of Source
• These criteria need not be carried to the very lowest
frequencies – i.e. below about 500 Hz – since imaging is
not influenced and coloration effects are low
• Some increase in the LF power response would help to
offset the well damped LF sound field.
• The loudspeaker therefore can, and should, widen in
directivity below about 500 Hz with no problems – this is
a “no brainer”
• 1000 Hz. is a more workable starting point for this
transition and will probably not affect the imaging if the
widening is down slowly.
• Above 1 kHz, especially 2-6 kHz, high directivity is
crucial, but it must be smooth and nearly flat at all points
within its coverage field
The Summa Loudspeaker
• The Gedlee Summa was specifically
designed to meet these small room
requirements
• It uses a waveguide for narrow directivity
with constant coverage, but also contains
an internal foam plug (patent pending) to
help to control internal reflections and
diffraction and higher order modes
Summa Con’t
• The cabinet edges and the waveguide
termination are all substantially rounded
for an absolute minimum of cabinet and
waveguide diffraction
• The freq. resp. is equalized (near) flat at
an off-axial location of 22.5° and is uniform
and smooth at virtually all points contained
within its coverage.
Summa Con’t
• The polar response transition to a piston
source is done precisely where the piston
and waveguide match polar patterns. The
waveguide is axi-symmetric to match the
woofer. Polar response through crossover
is flawless
• The cabinet is molded composite with very
high internal damping and is internally
braced to be very rigid
• It is extremely efficient: 97 dB/watt @ 1 m
Summa
Summa Cum Laude
90
Angle
60
30
0
100
1,000
Frequency
10,000
Summa Cum Laude (test1)
10
SPL(dB - normalized)
0
-10
-20
-30
-40
1,000
10,000
Frequency
Room Acoustics at LF
• The room dominates the LF situation
where the source has little effect
• At LF there are two things that will improve
the expected frequency and spatial
response.
– The first is to dampen the room as much as
possible
– And the second is to place multiple sources at
various locations around the room.
LF Damping
• Providing LF damping is a daunting task,
because we know that we want as little HF
damping as we can get to lower the
direct/reverb ratio for better spaciousness.
• Virtually all commercial sound absorbing
materials do exactly the wrong thing – lots
of HF absorption, negligible LF absorption
• The only effective solution is that the
absorption must be built into the structure
Construction
• The construction techniques are not
difficult and details can be found in my
book “Premium Home Theater”.
LF smoothness
• Once the maximum amount of LF
absorption has been utilized (and
hopefully the room is still live!) The only
other thing that can be done is to use
multiple subs.
• Since the Summa’s use 15” Pro
loudspeakers each one can handle lots of
LF energy – that’s three
• I use two more – one at the front of the
room and one in a back corner.
LF smoothness
• Studies have shown that the use of
multiple subs can substantially smooth out
the LF sound field both in frequency and in
space. The improvement goes as about
1/N, where N is the number of
independent sources.
• Others studies claim that particular
locations work best – I claim that random
locations work just as well (if not better!).
Final analysis
• Finally, the following slides are in-room
measurements of frequency responses
• Remember that these are in-situ
measurements and not gated (except for a
tapered 10 ms window)
10
Front 4 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
fro nt 10 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
left 1 0 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
cent er 10 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
right 10 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
Spat ial averaged 1/3 oct ave
Falloff due to
measurement antialiasing filter
dB(SPL) Arbitrary
20
30
40
.
50
60
100
3
1 10
Frequency
1 10
4
sound level at listeners position
right speaker
right speaker
left speaker
Median
line
toe'd-in directional
source
move left
left speaker
Omni-source
move right
in Small Rooms
By Earl R. Geddes
GedLee LLC
www.gedlee.com
My Background
• My PhD. thesis was on the modal response
of small non-rectangular rooms
– Conclusions:
• The first mode was independent on room shape –
depended only on volume
• Concluded that room shape had little effect, except
in cases of extreme symmetry
• Distribution of absorption was far more important in
the more symmetrical shapes – shape did help to
distribute the damping evenly among the modes.
• Only damping can help the smoothness of the LF
response
Conclusions Con’t
•
•
•
With damping comes loss of LF energy
Above Fs, damping has no modal effect
These conclusions have significant
implications to LF in small rooms
1) substantial LF damping is required for good
LF response smoothness
2) the damping must be well distributed
3) There will be substantial energy loss to
make up for
The small room problem
•
Small rooms have two/three regions of
importance that need to be attended to.
1) The LF modal region
•
•
Modes are discrete
Free waves are not permitted
2) Above Fs (Schroeder frequency)
•
•
Modes are irrelevant
Waves propagate freely – geometrically
3) Transition region is possible
The small room problem
• There is no reason to believe that these three
regions will have the same characteristics,
problems or solutions
• What works in one region may be completely
wrong in another
• Hence, these regions must be approached
independently, finding those solutions that work
for that region and then seeing if the two sets of
solutions can be somehow melded together
The small room problem
• The goal is to produce at the listeners
position a perceived playback of a sound
source that is optimized for imaging,
timbre (coloration) and spaciousness
– Basically the same things that a large room is
designed for
• Since “perception” is involved some
psychoacoustics is relevant
Psychoacoustic Fundamentals
• The perception of image location is
dominated by frequencies in the range of
1kHz – 8 kHz in the Auditory System (AS),
with a greater emphasis on the middle of
this range (See Blauert)
• If good imaging is desired then this range
must be relatively free from frequency
response aberrations, diffraction and wall
reflections < about 10ms.
Psychoacoustic Fundamentals
• Recent work (to be published) has shown
that diffraction and very early reflections
(< 1 ms.) are far more perceptively
important than a spectral analysis of their
effects would indicate
• This is hypothesized to be due to the
masking of the ear being far poorer in the
time domain than the frequency domain
Psychoacoustic Fundamentals
• If this hypothesis is true (the data indicates
that it is) then it is also suspected that these
diffraction effects would be level dependent
• Hence, while the diffraction is linear in a
mathematical sense, its perception may be
highly nonlinear
• To a listener, this diffraction might appear to
be nonlinear distortion since it becomes
more audible at higher levels than lower
levels – yet it has a physically linear cause
Psychoacoustic Fundamentals
• This hypothesis is completely consistent
with another study (AES) that showed that
perceptually nonlinear distortion in
compression drivers is virtually nonexistent
• Yet it seems to be common knowledge
than horns sound worse at higher levels
than lower levels
• Horns add only very low orders of
nonlinearity, but virtually all horns have
diffraction effects in them
Implications
• It then appears that it is not only critical
that the room not have early reflections,
but it is just as important that the sources
not have any near field diffraction from the
cabinets, any waveguide devices, or
nearby structures
• Source diffraction of any sort must be held
as just as undesirable as early reflections
More Psychoacoustics
• As stated before, at low frequencies, we
need not be too concerned about
reflections, and probably diffraction and
we certainly need not be concerned about
these problems down into the modal
region - this region is dominated by the
room and has hardly anything to do with
the source
Why Not?
• A simple solution would seem to be to just
put sound absorbing material everywhere,
or better yet just move outside!
– A non-reflection room is usually not found to
be perceptually adequate – that’s because it
lacks a very important acoustic property
known as spaciousness.
– Spaciousness occupies a large part of the
study and design of spaces for the performing
arts – it is nearly as important in small room
acoustics.
Spaciousness
• To fully understand spaciousness we need to
understand the concepts of direct and
reverberant fields.
• The direct field (not to be confused with the near
field) is where the sound from the source
dominates over the reverberant sound.
– There is a 6 dB/octave falloff with distance
• The reverberant field is when the reverb
dominates
– There is no level dependence on location
100
10
1
0.1
1.0
10.0
Distance from source
100.0
Into the Recording
• The importance of spaciousness can be
easily demonstrated
– tonight if time permits
– By moving closer and closer to speakers that
are canted inward, the sound field becomes
more and more dominated by the direct field –
the direct to reverb ratio goes up.
– Moving back beyond a certain point has no
effect.
Into the Recording con’t
• Moving forward creates a subjective effect that I
call “in the recording”
• Backward - “in the room”
• The former gives the subjective impression of
“being there” – you are moved into the recorded
space
• The later gives the impression that the
musicians have been transported into the room
with you
• Some like the “in the recording” effect, but I find
it unnatural - precise imaging beyond reality, no
spaciousness, a kind of headphone effect
Spaciousness
• Clearly spaciousness does not just
happen, to have it or not have it requires
some design considerations
• Not paying attention to it will likely leave
an audio reproduction with poor imaging
and or a colored sound character along
with a lack of spaciousness
Sources and Spaciousness
• Clearly the first arrival sound should be nearly
flat (a subtle HF roll-off is usually preferred) but
definitely smooth
• What is not commonly attended to is that for
spaciousness to occur there must be a
substantial reverberant field component at the
seating location
• The reverberant field response in a reverberant
room is dependent on the sources power
response – not its anechoic response
Sources and Spaciousness
• Very few loudspeakers have both a flat
anechoic response and a flat power
response.
• That’s because it cannot be achieved with
piston sources - a piston source does not
have a flat power response when it has a
flat anechoic response – it beams at HF
• Pistons can be both, but only below ka=1,
and then only as omni-directional sources
Return to room acoustics
• Now lets consider the source placement in
the room along with the sources directivity
• All sources have negligible directivity at LF
and most have a directivity that varies with
frequency throughout its operating range
– This means that the anechoic response and
the power response cannot match
• Consider a omni-directional speaker and
one with 90° coverage
Sources in rooms
• The omni source will have a multitude of
early reflections while the directive source,
if properly aimed, will have only a single
reflection (horizontal plane), which arrives
at the ear opposite to the direct sound
from this source.
• The opposite ear effect is notable because
it is far less objectionable than a reflection
to the same ear.
Specification of Source
• The sources should have the following
characteristics:
– They should be directive at < 90°
– They need to have off-axis responses that are
flat as well as on-axis
– They need to have a flat power response for
low coloration in a lively room (required for
good spaciousness – to be discussed)
• This is because of a time-intensity tradeoff in the
AS – longer signal times yield louder perception
Specification of Source
• These criteria need not be carried to the very lowest
frequencies – i.e. below about 500 Hz – since imaging is
not influenced and coloration effects are low
• Some increase in the LF power response would help to
offset the well damped LF sound field.
• The loudspeaker therefore can, and should, widen in
directivity below about 500 Hz with no problems – this is
a “no brainer”
• 1000 Hz. is a more workable starting point for this
transition and will probably not affect the imaging if the
widening is down slowly.
• Above 1 kHz, especially 2-6 kHz, high directivity is
crucial, but it must be smooth and nearly flat at all points
within its coverage field
The Summa Loudspeaker
• The Gedlee Summa was specifically
designed to meet these small room
requirements
• It uses a waveguide for narrow directivity
with constant coverage, but also contains
an internal foam plug (patent pending) to
help to control internal reflections and
diffraction and higher order modes
Summa Con’t
• The cabinet edges and the waveguide
termination are all substantially rounded
for an absolute minimum of cabinet and
waveguide diffraction
• The freq. resp. is equalized (near) flat at
an off-axial location of 22.5° and is uniform
and smooth at virtually all points contained
within its coverage.
Summa Con’t
• The polar response transition to a piston
source is done precisely where the piston
and waveguide match polar patterns. The
waveguide is axi-symmetric to match the
woofer. Polar response through crossover
is flawless
• The cabinet is molded composite with very
high internal damping and is internally
braced to be very rigid
• It is extremely efficient: 97 dB/watt @ 1 m
Summa
Summa Cum Laude
90
Angle
60
30
0
100
1,000
Frequency
10,000
Summa Cum Laude (test1)
10
SPL(dB - normalized)
0
-10
-20
-30
-40
1,000
10,000
Frequency
Room Acoustics at LF
• The room dominates the LF situation
where the source has little effect
• At LF there are two things that will improve
the expected frequency and spatial
response.
– The first is to dampen the room as much as
possible
– And the second is to place multiple sources at
various locations around the room.
LF Damping
• Providing LF damping is a daunting task,
because we know that we want as little HF
damping as we can get to lower the
direct/reverb ratio for better spaciousness.
• Virtually all commercial sound absorbing
materials do exactly the wrong thing – lots
of HF absorption, negligible LF absorption
• The only effective solution is that the
absorption must be built into the structure
Construction
• The construction techniques are not
difficult and details can be found in my
book “Premium Home Theater”.
LF smoothness
• Once the maximum amount of LF
absorption has been utilized (and
hopefully the room is still live!) The only
other thing that can be done is to use
multiple subs.
• Since the Summa’s use 15” Pro
loudspeakers each one can handle lots of
LF energy – that’s three
• I use two more – one at the front of the
room and one in a back corner.
LF smoothness
• Studies have shown that the use of
multiple subs can substantially smooth out
the LF sound field both in frequency and in
space. The improvement goes as about
1/N, where N is the number of
independent sources.
• Others studies claim that particular
locations work best – I claim that random
locations work just as well (if not better!).
Final analysis
• Finally, the following slides are in-room
measurements of frequency responses
• Remember that these are in-situ
measurements and not gated (except for a
tapered 10 ms window)
10
Front 4 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
fro nt 10 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
left 1 0 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
cent er 10 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
right 10 ms window
Falloff due to
measurement antialiasing filter
dB SPL (uncalibrated)
0
10
20
.
30
40
100
3
1 10
Frequency
4
1 10
10
Spat ial averaged 1/3 oct ave
Falloff due to
measurement antialiasing filter
dB(SPL) Arbitrary
20
30
40
.
50
60
100
3
1 10
Frequency
1 10
4
sound level at listeners position
right speaker
right speaker
left speaker
Median
line
toe'd-in directional
source
move left
left speaker
Omni-source
move right