Audio Engineering Society

Convention Paper
Presented at the AES 57th International Conference
2015 March 6–8 Hollywood, CA, USA

The papers at this Convention have been selected on the basis of a submitted abstract and extended precis that have been peer
reviewed by at least two qualified anonymous reviewers. This convention paper has been reproduced from the author's advance
manuscript, without editing, corrections, or consideration by the Review Board. The AES takes no responsibility for the contents.
Additional papers may be obtained by sending request and remittance to Audio Engineering Society, 60 East 42 nd Street, New
York, New York 10165-2520, USA; also see [Link]. All rights reserved. Reproduction of this paper, or any portion thereof,
is not permitted without direct permission from the Journal of the Audio Engineering Society.

EQUALIZING THE EFFECTS OF PERFORATED CINEMA SCREENS

Glenn Leembruggen1
1
Acoustic Directions, ICE Design Australia, and University of Sydney: Sydney Australia
glenn@[Link]

ABSTRACT

Perforated cinema screens are currently in widespread use in cinemas and dubbing stages due to their high light-
reflectivity. However, in the acoustical domain, such screens form a low pass filter which attenuates the high
frequency response of the cinema loudspeakers. Recent studies have shown that the X-curve, which has long been
adopted by the SMPTE as a standard response curve, reflects this low pass filter action. It is therefore of
significance to explore the ramifications of equalizing the effects of this low pass filter, in order to provide a flat
frequency response for listeners. This paper measures the spectral content of an action-based segment of a sci-fi
movie, and assesses the impacts on the headroom in the playback chain and the long-term power requirements of the
loudspeaker drivers that equalization of this type would impose with the spectral content of these movies.

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 1 of 11
Leembruggen EQUALIZING THE EFFECTS OF PERFORATED CINEMA SCREENS

1. MOTIVATION FOR THE WORK A set of parametric filters whose frequency response
had been matched to the X curve was then applied to
It is current practice to equalize a cinema sound
the signal chain so that the system’s response
system to the X curve specified by SMPTE standard
complied with the specified response
ST202:2010 (1). Although the X curve has been
used for at least two decades, there are now a range It was then a simple exercise to listen to cinema
of opinions about its exact origin. Some audio sound tracks from Dolby Control Prints and DVDs
practitioners consider that it was a preferred with and without the X curve in the signal chain. .
equalization based on subjective listening. Figure 1
My colleagues and I concluded that in every case, the
shows a reproduction of the specified X curve.
system sounded better without the X curve. Voices
The author recently calibrated two reference cinemas sounded like voices, a piano sounded like a piano,
in Australia which are used to promote cultural and the speech clarity was much more comfortable.
awareness of moving image productions. In broad
There is also a growing body of learned opinion that
terms, the calibration consisted of level adjustments
the use of the X curve as a desired frequency
and setting the frequency response of the systems to
response results in a substantial loss of fidelity,
match the X curve.
speech intelligibility and listener enjoyment.
Before implementing the X curve, the systems were
If the goal of cinema sound reproduction is true hi-fi,
equalized to produce a flat average response over the
then it is of interest to understand the implications of
listening area. The subjective result was essentially
removing the X curve would have for power-
hi-fi when listening to music replayed from a CD
handling capacity of cinema loudspeakers.
source.

Figure 1 X curve given in SMPTE ST202:2010

2. OVERVIEW
incidence) being directly related to the percentage of
Cinema screens generally consist of perforated vinyl
open area, hole size and screen thickness.
and are available with a small range of hole sizes and
percentage of open areas. As shown by Long et al in The good agreement between predictions and
(2), a perforated screen forms an acoustical low pass measurements described in (2) confirms that these
filter in front of the loudspeaker; with the resulting losses are entirely predictable for normally incident
loss of high frequency response (with normal sound.

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 2 of 11
Figure 35 of (2) shows that when a loudspeaker with been selected for its strong high frequency
a flat on-axis response is located behind a “Mini- content.
Perf” screen with a hole diameter of 0.5 mm and a 2. Figure 2 shows the specified location of the
perforation ratio of 1.7%, the resulting frequency Reference Position in a cinema calibration. This
response shows considerable similarity to the X measured response includes the following factors:
curve.
 Loudspeaker and horn responses
If the high frequency response of the loudspeaker is  Low pass filter action of the screen
flat (presumably by design or equalization
recommended by the manufacturer) and the room  Attenuation due to air absorption, which in a
frequency response is equalized to the X curve, the typical cinema is only significant at
resulting response will be a combination of screen frequencies above 7 kHz.
loss due to low-pass filtering, air loss and any  A five minute sound track of an action
equalization applied in the cinema. movie, Transformers 4, was obtained in 5.1
format from the mastering house. The
While there is current discussion about what the ideal
segment contained an action sequence,
high frequency response of the cinema playback
which had been selected for its strong high
chain should be, it is a useful exercise to consider
frequency content.
what the additional power and headroom demands on
the system would be if the high frequency response
was equalized to be flat.
A mathematical model was used to explore the
changes to the statistical temporal parameters that an
equalizer with an inverse X-curve shape would
impose on the system with a typical movie sound
track. Changes to these statistical parameters will
directly affect the amplifier power requirements.
It is most probable that the final mixdown of the
movie would have been made in a dubbing stage that
had been equalized to the X curve. As the sound
track was intended for playback in public cinemas,
the author assumes that the mastering engineer had Reference position
optimized the sound of this movie for playback in a
reference theatre.

3. FOUNDATIONS OF THE MODELLING

Figure 2 Physical layout of the reference position
The modelling is based on a number of assumptions in a cinema.
and foundations:
1. The model assumes that the loudspeaker system is 3. The loudspeaker system is assumed to be
operating comfortably with the calibration operating comfortably with the calibration
settings and equalisation specified by the SMPTE. settings and equalization noted in 1 and 2 above.
The equalisation therefore imposes an increased 4. The goal is to equalize the system to produce a
stress on the loudspeaker from a starting point flat frequency response; i.e. an inverse X curve is
that is comfortably within the amplifier’s and to be applied.
loudspeaker’s power capacity. 5. The loudspeaker is a two way type, using an 8th
 The measured response of the system at the order Linkwitz Riley crossover with transition
Reference Position noted in ST202 matches frequencies of 1200 Hz or 2500 Hz between the
the nominal X curve, i.e. the solid plot in low and high frequency drivers.
Error! Reference source not found.. A five 6. The process ignores the distance dependence of
minute sound track of an action movie, loss due to air absorption and the desirability, or
Transformers 4, was obtained in 5.1 format otherwise, of compensating for these losses.
from the mastering house. The segment In a large room and assuming ideal loudspeaker
contained an action sequence, which had directivity, flattening the response at the reference

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 3 of 11
 position would result in a slightly boosted
response at positions near the screen. For
example, if the reference position is 15 m from
the screen, air absorption would introduce losses
of 1.6 dB at 10 kHz and 3.5 dB at 15 kHz, while
at 4 m from the screen, the losses are 0.4 dB at 10
kHz and 0.9 dB at 15 kHz.
In the context that the X curve is measured using
stationary pink noise, the desirability of restoring
a flat frequency response to the stationary sound
field is a matter of current debate. Recent work
undertaken by the B Chain Committee of the
SMPTE (3) examined the relationship between
direct field and steady state frequency responses.
Figure 3 Responses shown in (4) with different
Figures A6, B6, C6, D6 and E6 from the SMPTE
windows at reference position of a Center Channel.
B Chain Report (3) show that, for a typical
NB 10 ms and 50 ms responses are truncated
cinema, the differences in the sound pressure
according to the time/frequency limit.
level at frequencies above 1 kHz between the
direct and steady state fields are typically less
than 0.5 dB. 14
filter 1
Illustrating this finding, Figure 3 reproduces 12 filter 2
Figure B6 from (3) and shows the direct field filter 3
represented by the frequency responses with 10 10
combined
ms and 48 points per octave windows while the
Response dB

8 Inverse X curve
steady state is represented by a 2 second window.
This small difference between direct and steady- 6
state sound fields can be essentially ignored and 4
therefore equalization of the steady-state X curve
equates to a restoration of a flat direct-field 2
frequency response.
0

-2
4. MODELLING METHODOLOGY 100 1000 10000
frequency Hz
The modelling was undertaken using Matlab
software. The following steps were used: Figure 4 Responses of peaking IIR filters and inverse
1. An equalization filter was designed to match the X curve.
inverse of the X curve, and consists of three IIR
parametric peaking filters. The responses of the
filters and the inverse of the X curve are shown in 5. The RMS, statistical exceedance levels and
Figure 4. histogram were computed for the following signal
types:
2. The digital audio file was divided into 30 seconds
segments at 48 kHz sampling rate with 32 bit.  raw broadband content
The entire file of 5.5 minutes duration was also  equalized broadband content
used.  unequalized low frequency content
3. Each raw broadband audio file was read into
 unequalized high frequency content.
Matlab and filtered with the inverse X curve
equalization.  equalized high frequency content.
4. Both the raw and equalized audio files were then All computations are relative to 0 dBFS.
low and high pass filtered with the crossover type
stated above.

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 4 of 11
6. The statistical exceedance levels are shown in the of the high frequency chain with and without
form of Ln, which is used by acousticians when equalization. For clarity, only the range of 0% to
referring to environmental noise levels. The Ln 20% is shown.
level is level that is exceeded for n percent of the
The largest differences in each percentile with and
measurement duration. In other words, for n
without equalization taken over all the file segments
percent of the time, the fluctuating sound pressure
and the complete file were computed and are shown
levels are higher than the Ln level. For example,
in Figure 12 for the 1200 Hz crossover and Figure 13
the 90th percentile of the signal level becomes the
for the 2500 Hz crossover.
L10 exceedance level.
In terms of the effect of equalization on power Histograms of the high frequency center channel
handling and system headroom, the smaller chains with the complete audio file are shown in
exceedance levels will have most impact and, in Figure 14 with and without equalization for the 1200
this context, the use of exceedance levels assists Hz crossover.
consideration of the impacts by the use of small The long term power spectra of the three screen
numbers. channels are shown in Figure 15.
7. When forming the histograms for each type, all
samples less than -40 dBFS were discarded for
clarity. The histograms were then found in half-
dB steps ranging from -40 dBFS to +5 dBFS. 6. DISCUSSION
The power spectral density of the five signal types The center channel has the highest RMS levels of the
was computed using the Welch method (4). For three screen channels.
this computation, a Hanning window was used Table 1 shows the differences between the L1 and L0
with a 16384 point FFT size and a window values of the high frequency chain with and without
overlap of 1,024 samples. The responses were equalization; they are unexpectedly high for both
then smoothed over a sliding 1/12th octave cases. The results suggest that if dynamic range
bandwidth. issues were important, a substantial decrease in
dynamic range could be achieved by reducing the
level of comparatively few samples.
5. RESULTS
The temporal parameters were computed for each Crossover point
audio file for the following five different signal types Parameter
or chains: 1200 Hz 2500 Hz
 original audio file - broadband (bb) L1 without eq dB re FS -13.0 -18.6
 original audio file with inverse x curve L1 with eq dB re FS -13.4 -17.4
equalization (bb eq) L0 without eq dB re FS -2.3 -4.0
 low frequency chain (lf) L0 with eq dB re FS +1* 0.05*
 high frequency chain (hf)
Table Exceedance levels of high frequency chain *re
 high frequency chain with inverse X curve dBFS of unequalized signal (NB L0 is the peak
equalization (hf eq) level)
Figure 5 shows the RMS values with the 1200 Hz Table 1 discusses the change in RMS and exceedance
crossover of the five spectral types of eleven levels with equalization for the two crossover points.
segments of the center channel along with the RMS
level of the complete audio file. Figure 6 shows the
same data for the 2500 Hz crossover.
Figure 7 compares the RMS values of the complete
audio files of the Centre, Left and Right channels
with the 1200 Hz crossover.
Figure 8 and Figure 9 show the exceedance levels of
the broadband content with and without equalization.
Figure 10 and Figure 11 show the exceedance levels

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 5 of 11
Leembruggen EQUALIZING THE EFFECTS OF PERFORATED CINEMA SCREENS

Figure 5 Computed RMS values for five signal chains of the Centre Channel with 1200 Hz crossover.

-10
RMS levels CC
-15

BB_rms
RMS level dB re FS

-20
BBeq_rms
-25 LF_rms
HF_rms
-30 HF_EQ_rms

-35

-40

Figure 6 Computed RMS values for five signal chains of the Centre Channel with 2500 Hz crossover.

RMS levels of complete audio files of three screen channels

-12
-14
-16
-18
Level dB re FS

-20 Centre
-22 Right
-24
Left
-26
-28
-30
-32
-34
BB BB eq LF HF HF eq

Figure 7 Computed RMS values for five signal chains of the three screen channels with 1200 Hz crossover.

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 6 of 11
Leembruggen EQUALIZING THE EFFECTS OF PERFORATED MOVIE SCREENS

sample 1
Unequalised Broadband Signal
5 sample 2
0 sample 3

-5 sample 4
sample 5
-10

Level re 0dBFS
sample 6
-15
sample 7
-20
sample 8
-25 sample 9
-30 sample 10

-35 sample 11
Complete
-40
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
Statistical Exceedance Level Lx dB

Figure 8 Exceedances values 0% to 100% for unequalized broadband centre channel signal. (1200 Hz crossover)

Equalised Broadband Signal Centre sample 1

5
sample 2
0 sample 3
-5 sample 4
-10 sample 5
Level re 0dBFS

-15 sample 6
sample 7
-20
sample 8
-25
sample 9
-30 sample 10
-35 sample 11

-40 Complete
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
Statistical Exceedance Level Lx dB

Figure 9 Exceedance values 0% to 100% for equalized broadband center channel signal ((1200 Hz crossover)

Unequalised High Frequency Chain sample 1

5
sample 2
0
sample 3
-5 sample 4
-10 sample 5
Level re 0dBFS

-15 sample 6
sample 7
-20
sample 8
-25
sample 9
-30 sample 10
-35 sample 11
-40 Complete
20% 18% 16% 14% 12% 10% 8% 6% 4% 2% 0%
Statistical Exceedance Level Lx dB

Figure 10 Exceedance values 0% to 20% for unequalized high frequency center channel signal. (1200 Hz crossover)

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 7 of 11
Leembruggen EQUALIZING THE EFFECTS OF PERFORATED MOVIE SCREENS

Equalised High Frequency Chain sample 1

5
sample 2
0
sample 3
-5
sample 4
-10 sample 5

Level re 0dBFS
-15 sample 6
-20 sample 7

-25 sample 8

-30 sample 9
sample 10
-35
sample 11
-40
20% 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% Complete

Statistical Exceedance Level Lx dB

Figure 11 Exceedance values 0% to 20% for equalized high frequency center channel signal. (1200 Hz crossover)

Difference of maximum Lx value between with and without eq

4

3
Differencein dB

0
20% 18% 16% 14% 12% 10% 8% 6% 4% 2% 0%
Statistical Exceedance Level Lx dB

Figure 12 Difference of exceedance values for equalized and unequalized high frequency center channel chains
with 1200 Hz crossover.

Figure 13 Difference of exceedance values for equalized and unequalized high frequency center channel chains
with 2500 Hz crossover.

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 8 of 11
Leembruggen EQUALIZING THE EFFECTS OF PERFORATED MOVIE SCREENS

Histogram of samples in 1 dB wide bins for high frequency chain

1000000

100000

10000

Samples per bin

1000 no eq
100 eq
10

0
0.1
-40 -30 -20 -10 0 10
Bin Position in dB re 0 dBFS

Figure 14 Histograms of signal levels in the high frequency chain with and without equalisation. Note the ordinate
scale is logarithmic axis to allow comparisons at bin-position values close to 0 dBFS. (1200 Hz)

Crossover point
Parameter
1200 Hz 2500 Hz
RMS level of high frequency chain relative to broadband level without -11.1 dB -17.5 dB
equalization
Increase in RMS level of broadband signal with equalization 0.2 dB 0.1 dB
Increase in RMS level of high frequency signal with equalization 1.3 dB 3.5 dB
Exceedance levels of broadband signal affected by equalization Top 1% Top 1%
Increase in exceedance levels of high frequency signal with equalization approx. 1 dB approx. 3.2 dB
Increase in peak level (L0) of high frequency signal with equalization 3.4 dB 4.0 dB
Number of samples in top 1% bin (between L1 and L0) (Figure 14) 3
Number of samples in top 3 exceedance bins (between L3 & L0) (Figure 14) 140 127
Crest factor of unequalized/equalized high frequency signal 25/27 dB 29/30 dB
Increase in crest factor 2 dB 1 dB

Table 1 Calculated parameters with crossover points of 1200 Hz and 2500 Hz.

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 9 of 11
 Leembruggen EQUALIZING THE EFFECTS OF PERFORATED MOVIE SCREENS

Long Term Spectrum of_XF4_LC Long Term Spectrum of_XF4_CC

10
LP chain LP chain
5
5 HP chain HP chain
HP chain eq 0 HP chain eq
0 BB BB
BB with eq -5 BB with eq
-5
-10
-10
Level (dB)

Level (dB)
-15 -15

-20
-20
-25
-25
-30
-30
-35
-35
-40
1 2 3 4 1 2 3 4
10 10 10 10 10 10 10 10
Frequency (Hz) Frequency (Hz)

Long Term Spectrum of_XF4_LC

10
LP chain
5 HP chain
HP chain eq
0 BB
BB with eq
-5

-10
Level (dB)

Upper: Left channel

-15
Centre: Centre channel
-20 Lower: Right channel

-25

-30

-35

1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 15 Power Spectral Density of the five signal chains of the three screen channels. 1200 Hz crossover system shown.

6.1. Spectra 6.2. Power Requirements

The long term power spectra of the three screen The results produced for this sound track segment
channels shown in Figure 15 show surprisingly small with a crossover frequency of 1200 Hz show that
energy at high frequencies compared to low there will be a 1.3 dB increase in the long-term power
frequencies. Of the three channels, the center channel dissipation requirements for the high frequency driver.
contains the greatest amount of high frequency This equates to a 35% increase. With the crossover
energy. frequency of 2500 Hz, the increase is 3.5 dB, equating
to a 124% increase.

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 10 of 11
Leembruggen EQUALIZING THE EFFECTS OF PERFORATED MOVIE SCREENS

However, these increases must be considered in which a calibrated system should have at the reference
relation to the current powers being dissipated by the position.
driver. These powers are estimated as follows in the
The investigation computed the temporal statistical
1200 Hz crossover case:
and RMS parameters for an action-based segment of
 A typical 35 mm throat compression driver is the movie Transformers 4, which was in cinema
likely to have a power rating of 75 watts release format. These parameters were computed for
AES. both the broadband signal and the signal fed to the
 Based on the 6 dB crest factor of the AES high frequency driver of a typical two way high
rating, the driver can accommodate 300 watts frequency cinema loudspeaker with crossover points
instantaneous peak power. of 1200 Hz and 2500 Hz.
 The average power rating of an amplifier The increases in thermal power dissipation and crest
delivering this peak power would be 150 factor resulting from the equalization were computed
watts. (3 dB above the average power rating) from these parameters.
 A power amplifier with 200 watts average The results show that equalization increases the RMS
power is therefore likely to be used with the level applied to the high frequency driver by 1.5 dB
driver. The peak power capacity of this and 3 dB for crossover points of 1200 Hz and 2500
amplifier is therefore 400 watts. Hz respectively. The crest factors increase by 2 dB
 Taking a conservative approach, let’s assume and 1 dB respectively.
that the driver sensitivity is too low for the
Although the RMS increase with equalization appears
situation and the amplifier is being driven
to impose a considerable percentage increase in
into 3 dB of clipping.
thermal power dissipation, the increases in power
 With the crest factor of the no-equalization must be considered in relation to the crest factors of
case at 1200 Hz, this degree of clipping the program content and the actual power being
equates to an RMS level of 22 dB (-25+ 3) dissipated. Estimates of a typical driver and amplifier
below the peak amplifier power, or 2.5 watts combination indicate that the power increases are
average power. negligible.
 With equalization, the thermal power
dissipation would increase by 35% to 3.5 8. REFERENCES
watts. 1. SMPTE. STANDARD for Motion-Pictures —
For the 2500 Hz crossover, similar reasoning predicts Dubbing Stages (Mixing Rooms), Screening Rooms
an average power of 0.9 watts without equalization and Indoor Theaters — B-Chain Electroacoustic
and 2.0 watts with equalization. Response. ST 202:2010.
In relation to the driver rating of 75 watts, the 2. B. Long, R Schwenke, P. Soper, G
increases in thermal power are negligible. Leembruggen. “Further Investigations Into the
Interactions Between Cinema Loudspeakers and
Accommodating the increased crest factor would Screens” . SMPTE Mot. Imag J. November-December
necessitate a greater peak power capacity from the 2012 , Vol. vol. 121 , no. 8 46-62.
amplifier, but with the plethora of power outputs
available in contemporary amplifiers, this capacity is 3. SMPTE B-Chain, TC-25 CSS. Frequency and
readily accommodated. Alternatively, the top 0.5% of Temporal Response Analysis of Theatres and Dubbing
levels in the high frequency chain could be reduced by Stages. 2014.
judicious hard limiting. 4. Welch, P.D. The Use of Fast Fourier Transform for
the Estimation of Power Spectra: A Method Based on
7. CONCLUSION Time Averaging Over Short, Modified Periodograms.
An investigation has been undertaken to determine the IEEE Transactions on Audio Electroacoustics,. AU-
impact on power capacity and dynamic range of a 15, 1967.
cinema sound system when equalization is applied to 5. SMPTE RP200:2012 . Relative and Absolute
the system to flatten the system’s frequency response Sound Pressure Levels for Motion-Picture
at the reference listening position. This equalization Multichannel Sound. 2012.
effectively removes the X curve from the response

AES 57th International Conference, Hollywood, CA, USA, 2015 March 6–8
Page 11 of 11