JP4782614B2 - decoder - Google Patents

Description

(Mutual proof)
This application is filed with US Provisional Patent Application No. 60 / 058,169, filed on September 5, 1997, “5-2.
-5 matrix encoder and decoder system (5-2-5 Matrix
Encoder and Decoder System) ”.

FIELD OF THE INVENTION The present invention relates to the decoding of a pair of stereo sound input signals into the same number of output signals for proper post-amplification reproduction by a plurality of loudspeakers arranged to surround the listener, as well as multi-channel material. The present invention relates to a sound reproduction system including encoding to two channels.

BACKGROUND OF THE INVENTION The present invention provides a constant sound level regardless of the direction of the directionally encoded component of the input sound signal while maintaining a high separation between the various outputs for signals with a defined direction. Improves the balance between the front and rear signals, even when there is a net forward and backward bias to the input signal, or when there is a strong acoustic component in a particular direction, while maintaining a non-directional encoded component at 5 channels that produce smooth sound movement near the 7-channel version of the system and are close to the 7-channel version
Decoding matrix with optimal psychoacoustic performance in playback of encoded multi-channel material, including frequency-dependent circuitry that generates the version of the sound, and between the left and right components of the stereo signal under all conditions The present invention relates to an improved set of design criteria and solutions for generating standard two-channel material, including maintaining a high separation of data.

Furthermore, the present invention provides an improved design basis and solution for generating an encoding circuit for encoding multi-channel sound into two channels for playback in a standard two-channel receiver with a decoder according to the present invention. Regarding the set.

The present invention comprises means for encoding a multi-channel acoustic signal into two individual channels;
It is then part of an ongoing effort to refine the resulting two channels and the means to re-separate them into the resulting multi-channel signal. One of the goals of the encoding / decoding process is to restore the original signal as permanently as possible to the original. Another important goal of the decoder is to extract five or more individual channels from a two-channel source that was not encoded from the five-channel original. The resulting 5-channel representation must be at least as musically entertaining and appreciable as the original 2-channel representation.

The present invention relates to improvements to the acquisition of suitable variable matrix coefficients. To assist in understanding such improvements, the disclosure of this text is based on Griesser's U.S. Pat. No. 4,862,502 (1989), referred to as 1989 U.S. Pat.
136,650 (1992), July 1996 called US patent application July 1996
Reference is made to U.S. patent application Ser. No. 08 / 684,948 of Moon Griesser and 08 / 742,460 of Griesinger, referred to as Nov. 1996 U.S. patent application. The commercial version of the decoder based on said last US patent application is version 1
. 11 (ie V1.11). Some further improvements are in version 1.01 (
That is, US Provisional Patent Application No. 60 / 058,1 filed in September 1997, referred to as V2.01).
69. These versions V1.11 and V2.01, and the present invention,
Collectively referred to as a “logic 7” decoder.

Still other technical references cited are: [1] “Multichannel Matrix Surround Decoder for binaural listeners.
und Decoder for Two-Eared Listeners) "(D.
Griesinger's AES Preprint No. 4402, December 1996, and [
2] "Process in 5-2-5 matrix system" (D. Griesinge
r AES Preprint No. 4625, September 1997).

SUMMARY OF THE INVENTION Means used to recreate the original signal encoded from 5 to 2 channels and embody the two goals of providing a sensuously good reproduction of 2 channel material in a 5 channel format As long as we understand the physical and psychoacoustic events that we do, we have embodied it. The aforementioned US patents and US patent applications presented a philosophy of design that created an effective decoder device.

The present invention relates to the realization of an active matrix having several properties that maximize psychoacoustic performance. In another aspect, the present invention discloses a frequency dependent modification of a portion of the output from the active matrix. In yet another aspect, the present invention uses two decoders according to the present invention, which are both standard two-channel devices, and an industry standard “Dolby Pro Logic” decoder to provide two five-input channels for optimal performance. An active circuit according to the present invention for encoding to a channel output is provided.

The present invention is an active matrix decoder having, in part, matrix elements that vary depending on the directional component of the incoming signal. This matrix element is not included in the intended direction while emphasizing the loudness of the signal in the direction included in the playback in the intended direction, while always maintaining the left and right separation of other signals present at the same time at the input Fluctuates to reduce the loudness of the directionally encoded signal at the output. Furthermore, the matrix element according to the invention restores the left and right separation of a non-correlated two-channel material encoded directionally by increasing or decreasing the mixing between the two inputs, for example by controlling the stereo width. Furthermore, the matrix element according to the invention comprises:
It is designed to preserve as much as possible the energy balance between the various components of the input signal so that the balance between speech and accompaniment is preserved at the decoder output. As a result, the matrix elements according to the invention preserve both the loudness of elements that are not directionally encoded in the input sound and the left and right separation of these elements.

Furthermore, the decoder according to the present invention returns from 2 for the 5 channel decoder to 4 for the 7 channel decoder so that the direction of the sound feels more like the direction of the sound from the 7 channel decoder, and 5 When standard two-channel material that modifies the spectrum of the back channel in the channel decoder is played, it includes a frequency dependent circuit that improves decoder output compatibility.

The encoder according to the invention is such that the direction of the strong input is encoded with the phase / amplitude ratio of the output signal so that when the input level of a particular input is strong, the energy present in this input is stored in the output. 5 input channels (or one that adds one low frequency to all 5 ranges) so that a strong signal can be panned between any two inputs of the encoder and the output is properly directional encoded Is mixed into two output channels. Furthermore, uncorrelated material applied to the two rear inputs of the encoder results in a two-channel output that the in-phase input to the two rear channels of the encoder decodes into the rear channel of the decoder according to the invention and the decoder according to the Dolby standard. And an anti-phase input to the two rear channels of the encoder is applied to the two rear inputs of the encoder so as to produce an output corresponding to a non-directional signal to the decoder according to the invention and the decoder according to the Dolby standard. In such a way that the left and right separation of the input is preserved when the encoder output is decoded by the decoder according to the invention so that the low level reverberation signal is encoded with a 3 dB reduction at the level at the two channel output. Will be encoded into the channel.

Detailed Description of Preferred Embodiments The design presented herein retains many of the design philosophy of the foregoing disclosure, where the actual design has changed in many ways. It is impossible to fully describe the evolution of such a design within the length of a readable document. In order to maintain document consistency, the text presents a mathematical solution to the problem that raises the most important elements of the design philosophy and describes the original solution in the application to claim. While it is useful to consider our previous application on such subject matter, it is not essential.

The background in decoders and encoders as described in the July 1996 and November 1996 U.S. patent applications and the September 1997 U.S. provisional patent application led to further improvements not yet disclosed. It was. This application describes the most important features of the improved encoder and decoder of the present invention, and claims new features that have been added since US patent application Ser. No. 08 / 742,460.

General description of the decoder The decoder in this application is described as consisting of two individual parts.
The first part is a matrix that divides the two input channels into five output channels, usually identified as center, left front, right front, left rear and right rear. The second part consists of a series of delay elements and filters that modify the spectrum and level of the two rear outputs. One of the functions of the second part is to obtain an additional pair of outputs on the left and right when a 7-channel version of the decoder is desired. In US patent application Ser. No. 08 / 742,460, the second part is not clear and two additional channels are derived from a further pair of matrix elements in the original matrix.

In mathematical formulas describing decoders and encoders, vector quantities are in bold capital letters and matrices are in bold lowercase letters, but standard variables are used for most variables where simple variables are shown in italics. Will be used.
Matrix elements that are coefficients from the named output channel resulting from the named input channel are shown in normal capital letters. Simple variables such as lr and cs are described by two-letter names that do not represent the product of two individual simple variables. The other variables l / r and c / s, in a sense, represent the value of the left / right and center / surround ratio, but relate to the control signal voltage obtained from such ratio. These transformations are those used in the aforementioned US patents and US patent applications cited herein. Matlab (
Program segments in the Matlab language are also distinguished by the use of different types of typefaces and point sizes and by retracting these lines. The formulas are numbered to provide a basis for specific features that are distinguished from the Matlab assignment statement and described in the text.

FIG. 1, which is the same as FIG. 4 in US patent application Ser. No. 08 / 742,460, shows a two-channel block diagram for a first portion of the decoder, a five-channel matrix 90. The left half of FIG. 1 delimited by a vertical chain line has two orientations (steeri).
ng) Means for obtaining voltages l / r and c / s. These directional voltages each have a unique or encoded directional component in the left / right or forward / backward direction, respectively. This part of the figure is described in detail in the aforementioned US patent application incorporated herein by reference, and is not specifically discussed herein.

In FIG. 1, the direction detection means of the decoder 90 including elements 92 to 138 is followed by a 5 × 2 matrix to the right of the vertical chain line. The elements of this matrix 140-158 determine the amount of each input channel that is linearly combined with the other input channels to form each output channel. These matrix elements are assumed to be real. (
The case of composite matrix elements is described in US patent application Ser. No. 08 / 742,460 and is not discussed here. ) This matrix element has two directional voltages l / r and c /
is a function of s. US patent application Ser. No. 08 / 742,460 shows mathematical formulas for these functions. Part of the novelty in this application is in improvements to these equations. These mathematical expressions are shown in graphs, and the reason why these expressions take such a form will be described.

2. Brief Description of Directional Voltage As shown in FIG. 1, the directional voltages c / s and l / r are the logarithm of the ratio of the left input amplitude at terminal 92 to the right input amplitude at terminal 94 and the sum amplitude for different amplitudes. And the logarithm of the ratio. In describing the directional voltage matrix elements, it is convenient to express l / r and c / s as angles that vary from +45 degrees to -45 degrees. In the V1.11 and V2.01 decoders, these voltages have units of decibels. In the following, the directional voltage parameter can be converted to an angle.

lr = 90-arctan (10 ^ ((l / r) / 20)) (1a)
cs = 90-arctan (10 ^ ((c / s) / 20)) (1b)
The angles lr and cs determine the angle with respect to the direction component of the input signal. For example,
When the inputs to the decoder are not correlated, lr and cs are both zero. For a signal that arrives only from the center, lr is zero and cs has a value of 45 degrees. For signals coming from the rear, lr is zero and cs is -45 degrees. Similarly, the signal coming from the left is 45
The signal from the right lr value and zero cs right has an lr value of -45 degrees and a cs value of zero. In the present design, it is assumed that the decoder that generates the encode signal has characteristics that lr = 22.5 degrees and cs = 22.5 degrees when encoding the left rear signal. Similarly, the signals applied to the right rear input to the encoder are lr = 22.5 degrees and cs = 22.
Yields a value of 5 degrees.

From the definition of lr and cs and the induction of lr and cs, the sum of the absolute values of lr and cs is 4
It can be seen that it cannot be greater than 5 degrees. The allowable values of lr and cs are absolute values (lr) −
A surface surrounded by a locus of absolute value (cs) = 45 degrees is formed. Any input signal that yields lr and cs values that lie along the boundary of this surface is totally localized-i.e. consists of a single sound encoded to reach from a specific direction.

In the present application, a matrix element graph is widely used as a function in such a two-dimensional plane. In general, the induction of matrix elements is different in the four quadrants of such planes. In other words, matrix elements are described differently depending on whether the orientation is forward or backward and whether the orientation is left or right. Much work has been done to ensure that the surface is continuous across the boundaries between quadrants. This accidental lack of continuity is one of the problems in the V1.11 decoder to which this application is directed.

3. Frequency Dependent Elements The matrix elements shown in FIG. 1 are real and therefore frequency dependent. All signals at the input are directed to the output depending on the induced angles lr and cs. (
In today's technology, low and very high frequencies are attenuated in the induction of lr and cs from the input signal by a filter not shown in FIG. However, the matrix itself is broadband. )

In practice, it has been discovered that there are several advantages of applying a frequency dependent circuit to the post-matrix signal. One of these frequency dependent circuits—the phase shift network 170 of the right output 180 in FIG. 1—is described in US patent application Ser. No. 08 / 742,460 and will not be discussed further herein.

FIG. 2 shows a 5-channel version of yet another frequency dependent circuit. These circuits are
Does not have a fixed parameter. The behavior of frequency and level depends on the directivity values lr and cs.
Depends on. These circuits accomplish several purposes. First, in both 5-channel and 7-channel decoders, the orientation is neutral (lr and cs are 0) or forward (cs
When> 0), an additional element makes it possible to adjust the clear loudness of the rear channel. In US patent application Ser. No. 08 / 742,460, such attenuation was performed as part of the matrix itself and was frequency dependent. Through theoretical research and listening tests, we have found that it is highly desirable that low frequencies be reproduced from the listener's location. Thus, the high frequency is attenuated by the variable low-pass filters 182, 184, 188 and 190 only in the decoder described here.

This is because when the orientation is almost always neutral or forward, the background control signal 186, defined later in this application, is used to cause frequencies higher than 500 Hz in the rear channel by elements 188 and 190 and from 4 kHz. High frequency elements 182, 184
Is achieved by attenuating. If there is an accidental sound that is directed backwards,
Attenuation, which is a feature that automatically distinguishes surround encoded material from normal two-channel material, is reduced.

Elements 192, 194 in the five-channel version are used when the orientation is relative to the back (cs <0) so that the loudspeaker appears to be placed behind the listener even though the actual position of the loudspeaker is on the side. ), The sound spectrum is corrected using the c / s signal 196. The modified left surround signal and right surround signal are connected to terminals 198 and 2, respectively.
Appears at 00. Further details of this circuit are given later in the disclosure.

FIG. 3 shows a 7 channel version of the frequency dependent element. As before, the first set of filters 182, 184, 188, 190 attenuates the high frequencies of the side and rear outputs when the pointing is neutral or forward controlled again by the background control signal 186. . Such attenuation also results in a further forward sound image, which can be adjusted to the listener's preference. As the directivity represented by the c / s signal 196 moves backwards, further circuits 202, 204, 206, 208 serve to discriminate the side outputs from the rear outputs. As the pointing moves backwards, the attenuation previously described in the side speakers is first removed by elements 204, 206 to produce a sound directed sideways. As the pointing moves further back, the attenuation of the elements 204, 206 is restored and increased. The result is that sound travels smoothly from the front loudspeaker to the side loudspeakers, and then to the rear loudspeaker that produces the approximately 10 ms delay caused by the delay elements 202,208. Since the low frequencies are not affected by these circuits, the low frequency loudness in the side speakers (which have the role of spreading recognition) is not affected by the movement of the sound. Further details of the circuit in FIG. 3 are described in subsequent chapters of the text.

4). General Description of Encoder FIG. 4 shows a block diagram of an encoder designed to automatically mix five input channels into two output channels. This architecture is described in US patent application Ser. No. 08/7.
This is completely different from the encoder described in US Pat. No. 42,460. The purpose of this new design is to maintain a 5-channel original musical balance in the two output channels while providing a phase / amplitude cue that allows the decoder to extract the original 5 channels. . Previous encoders had similar goals, but improvements were made in the methods used to achieve these goals. The preservation of musical balance is very important in the encoder. One of the main purposes of the encoder is to use two channels of five-channel recording to perform with the same artistic quality as a five-channel original in a normal two-channel system.
The mix is generated automatically. This new encoder design includes active elements that ensure that musical balance is preserved.

Unlike the encoder in the November 1997 US patent application, this new design allows the input signal to be panned between the five inputs of the encoder. For example, the sound can be panned from the left front input to the right rear input. When the resulting two-channel signal is decoded by the decoder described herein, the result will be very close to the original sound.
Decoding by a conventional surround decoder will also be similar to the original.

  The detailed description of the encoder will be described in the following chapters.

5. Design Goals for Decoder Active Matrix Elements The most basic goal of the present invention is to determine prior decoders, particularly US patent application Ser.
It is the same as the target of the decoder described in No.60. “The present invention reduces the directionally encoded audio components in the output that are not directly involved in playback in the intended direction and outputs in the direction intended to maintain a constant total output for the signal. Emphasize directionally encoded audio components in the output that are directly involved in the playback of
Regardless of whether there is a directionally encoded signal, while maintaining a high separation between the left and right channel components of the non-directional signal regardless of the directional signal, these intended directions are A surround acoustic decoder having a variable matrix value configured to effectively maintain a constant loudness, defined as the total audio output level of the non-directional signal, whatever the case. "

Most of these goals are clearly shared by all matrix decoders. The novelty in this application resides in part in knowing how to implement the above laws more precisely and partly in knowing when to apply the above laws. However, much of the methodology of US patent application Ser. No. 08 / 742,460 is preserved. One of the most important of the aforementioned goals
One is the clear maintenance of high separation between the left and right channels of the decoder under all conditions. The four channels described above provide only one rear channel and therefore cannot maintain separation in the rear. 5 channel decoders by other manufacturers
There are many ways to compromise separation. The decoder described in this application is
While such goals are met in a manner similar to that of V1.11, further goals are also met.

US patent application Ser. No. 08 / 742,460 also includes many relatively small improvements to designs such as circuits that improve the accuracy of directional signals and variable phase shifting networks that switch one phase of the rear channel during strong backward pointing. Is also described. These features of decoder V1.11 are retained in the new design but are not included in the text.

In FIG. 4, forward input signals L, C and R are applied to input terminals 50, 52 and 54, respectively. The C signal is attenuated 372 before being applied to the inputs of both summers 278,282.
Is first attenuated by a factor fcn, but the L and R signals are added by adders 278, 28.
Enter 2 directly. The low frequency effect signal LFE passes with a gain of 2.0 at element 374 and is then applied to both adders 278,282.

Surround input signals LS and RS are applied to two individual paths through two input terminals 62 and 64, respectively, and for LS signals, the path through attenuator 378 is gain fs (l, ls). ) And the RS signal has a corresponding attenuator 380 having a gain fs (r, rs).
Pass through. The outputs of these attenuators are cross-coupled elements 384, 3 having a gain factor -crx.
86 (where crx is nominally 0.383). The cross-coupled signals from these elements are LS and R attenuated from 0.91 attenuators 388 and 392, respectively.
It is sent to adders 392 and 394 that receive the S signal. The outputs of adders 392 and 394 are applied to the inputs of adders 278 and 282. This places these elements 45 degrees left and right behind the center in the decoded space, respectively.

The other signal branches are respectively an attenuator 376 having a gain fc (l, ls) and a gain fc (
r, rs) through an attenuator 382 and then cross-coupled elements 396, 398, 40
2, LS and RS signals are sent through a similar configuration of 404, 406 and 408, and adders 406 and 408 represent the left rear and right rear inputs at 45 degrees left and right of the center rear as before. Has output. However, the left and right signals from adders 278 and 282 pass through phase shift elements 286 and 288, respectively, but the signals are phase shift elements 234 and 24, respectively.
Pass through 6 respectively. Each of these phase shifting elements is an all-pass filter with a phase response of φ (f) for elements 286, 288 and φ (f) for elements 234, 246.
) −90 °. The calculation of the required component values in these filters is well known in the art and will not be discussed further in the text. The result is that the outputs of summers 406, 408 are delayed by 90 degrees from the outputs of summers 278, 282 at all frequencies after passing through an all-pass filter network as shown in FIG. Filters 246, 28
8 outputs are combined by adder 280 to produce a B (ie, right) output signal at terminal 46, while the outputs of all-pass filter networks 234, 286 are now combined by adder 276 to A at terminal 44. (Ie, left) produces an output signal.

The weak surround signal passes through a 90 degree out-of-phase path and maintains a constant output for an uncorrelated “music” signal, but the gain functions fs and fc make the strong surround signal in phase with other sounds. Designed to. The value crx may also change the angle at which the surround signal is heard.

6). Design improvements since US patent application Ser. No. 08 / 742,460 One of the most significant improvements in the present invention associated with US patent application Ser. No. 08 / 742,460
One is the change in the center matrix element and the left and right front matrix elements when the signal is directed in the center direction. It has been found that there are two problems with the central channel as previously encoded and decoded. The most obvious problem is that in a 5-channel matrix system, the use of the center channel is essentially incompatible with the goal of maintaining as much left and right separation as possible. If the two input channels do not have left and right components, the center channel must be driven by the sum of the left and right input channels if the matrix produces an output that can be recognized from conventional two-channel stereo material. In this way, both the left and right decoder inputs are reproduced by the center speaker, and the sound that was initially only left (or right) channel is also reproduced from the center. The result is that the clear location of these sounds is drawn to the middle of the room. The extent to which this occurs is
Depends on the center channel loudness.

U.S. Pat. Nos. 4,862,502 and 5,136,650 used matrix elements with a minimum of 3 dB relative to the left and right channels. When the inputs to the decoder were uncorrelated, the center channel loudness was equal to the left and right channel loudness. As the orientation moved to the central matrix element, it further increased by 3 dB. The effect of such high loudness is to significantly reduce the width of the front sound image. The musical instruments to be played on the left and right sides of the sound image are always drawn toward the center of the sound image.

US patent application Ser. No. 08 / 742,460 used a central matrix element with a minimum 4.5 dB less than the previous value. This minimum was chosen based on listening tests. Such attenuation produced a pleasant spread to the forward sound image when the input material was not correlated as in orchestral music. The sound image in front was not significantly narrowed. In US patent application Ser. No. 08 / 742,460, these matrix elements increased as the orientation moved forward, and finally reached the values used in the Dolby Matrix.

Experience with the V1.11 decoder showed that the reduction in the center channel loudness solved the spatial problem, but the output balance in the input signal was not preserved in the matrix. Mathematical analysis has revealed that not only V1.11 but also the Dolby decoder and other previous decoders are in error. Paradoxically, the central channel was too strong from the perspective of reproducing the width of the front sound image, but too weak to preserve the output balance. This problem is particularly severe for Mandel decoders-standard Dolby decoders. In a standard Dolby decoder, the rear channel is stronger than in our decoder of US Pat. No. 4,862,502. As a result, the central channel must be strong to maintain power balance. The lack of output balance in the central channel remained a problem for the Dolby decoder.
Dolby recommends that acoustic mix engineers always listen to the balance through the matrix, so the lack of output balance in the matrix
Can be compensated in the process. Unfortunately, recent films are mixed for 5 channel releases, and automatic encoding for 2 channels tends to cause problems at the interaction level.

More analysis and listening tests have shown that film and music require different solutions to the balance problem. In the case of film, US patent application Ser.
From 742,460, it has been found most effective to preserve the left and right front matrix elements. These elements remove as much of the center channel information as possible from the left and right front channels. This minimizes interaction leakage to the front left and right channels. In the new “film” design, the orientation moves forward (cs becomes greater than zero)
Accordingly, the output balance is corrected by changing the central matrix elements so that the loudness of the central channel increases more rapidly than the standard decoder. In practice, since such a condition is reached only when the central channel is active, it is not necessary that the final value of the central matrix element be higher than the final value in a standard decoder. It is only necessary that the center level is stronger than the standard decoder when there are approximately equal levels in the center channel and the left and right channels.

The "film" strategy described above increases the center channel loudness to preserve the output balance in the input signal while minimizing the center channel component at all other outputs. Such a strategy seems to be ideal for film, where the central channel is used primarily for interaction, and interaction from locations other than the center is not expected. The main drawback of such a strategy is that the forward sound image is narrowed when there is significant central orientation, as occurs in many types of popular music. However, the benefits to the film—including minimal interaction leakage to the front channel and excellent power balance—can more than compensate for these shortcomings.

For music, use a different strategy. In this case, the loudness of the central channel is increased to the intermediate value of the pointing at the same ratio as in US patent application Ser. No. 08 / 742,460, where cs ≧ 2
It is 2.5 degrees. In order to restore the music balance, the left and right front matrix elements are changed so that the central component of the input signal is not completely removed. The amount of central channel component in the left and right front channels is adjusted without excessive loudness in the center so that the acoustic output from all decoder outputs matches the acoustic output in the input signal.

With such a strategy, all three front speakers reproduce the center channel information present in the original encoded material. The most effective version of such a strategy limits the directing effect when the central component of the input is 6 dB stronger at the central output than either of the other two forward outputs. This is done by simply limiting the positive value of cs.

A novel strategy that limits the directional effect when the center channel component comes from all three speakers and the center is 6 dB higher than the front left and right makes all kinds of music excellent. The encoded 5 channel and normal 2 channel mix decodes with a stable center and sufficient separation between the center channel and the left and right channels. Note that unlike previous decoders, the separation between the center and left and right is not exactly perfect. Signals intended to come from the left are removed from the center channel, but not in the other direction. In the case of music, the high side separation and the stable forward sound image provided by the policy emphasize this lack of complete separation. A listening test with such a setting in the film reveals that the resulting sound image is very stable even if there is interaction from the left and right front speakers. This result is pleasant and not annoying .

Perhaps the next most obvious improvement in this application is the increase in separation between the front and rear channels when the signal is directed from the left front to the left rear direction. The V1.11 decoder is disclosed in US Pat.
862,502 matrix elements were used. These matrix elements do not completely eliminate the backward-directed signal if the backward-directed signal is not directed to the entire rear position—the middle between the left and right rear. When the orientation was left rear or right rear (not completely rear), the left or right front output produced an output that was 9 dB less than the corresponding rear output.
In the present invention, the front matrix elements are modified to remove sound from the front when the orientation is somewhere between left rear and right rear.

7). Improvements to posterior matrix elements Improvements to posterior matrix elements are not immediately apparent to a typical listener. These improvements correct for various errors in the continuity of matrix elements across the boundaries between quadrants. These improvements also improve the output balance between signals directed and not directed under various conditions. The mathematical description of the matrix elements described below includes these improvements.

8). Detailed description of active matrix el
The mathematics used to describe Matlab language matrix elements is not based on continuous functions of variables cs and lr. In general, there are conditions on formulas, absolute values and other non-linear corrections.
For this reason, matrix elements are described using a programming language.
The Matlab language provides a simple way to look up mathematical expressions graphically. Matlab is very similar to Fortran or C language. The main difference is that the variables in Matlab can be vectors, i.e. each variable can represent a series of sequences. For example,
The variable x can be defined as follows:

x = 1: 10;
Such a convention in Matlab produces a string of 10 numbers with values from 1 to 10. Variables x all contain 10 values. This is described as a vector that is a 1 * 10 matrix. Individual numbers within each vector can be accessed. For example, the formula x (4) = 4;
Sets the fourth term of the vector x to the value 4. Variables can also represent a two-dimensional matrix. Individual elements in the matrix can be assigned similarly.

X (2,3) = 10;
Assigns the value 10 to the second row and third column of the matrix X.

The detailed description of the matrix elements below is almost the same as the description published in document [2]. The text was slightly improved. The main differences are: Reference [2] includes the characteristics of the “tv matrix”. This feature reduces the level of rear output when the pointing is forward or neutral. In the present application, this function is obtained by a frequency dependent circuit following the matrix. Therefore, the “tv matrix” correction was excluded. 2. Parts in the central matrix element have been modified to include references to “film”, “music” and “music” setting restrictions. Document [2] describes only the “music” setting without limitation.

9. In the matrix decoder literature [1] in equations and graphics, a matrix that can be described by n × 2 matrix elements
Decoder design was described. n is the number of output channels. Each output can be shown as a linear combination of two inputs, where the coefficients of the linear combination are given by the elements in the matrix. In this paper, elements are identified by simple combination combinations of letters. Reference [1]
Describes 5 and 7 channel decoders. The conversion from 5 channels to 7 channels is done in the frequency dependent part of the decoder, so only the 5 channel decoder is described here.

From the symmetry it is clear that it is necessary to describe the behavior of only six elements: the central element, the two left front elements, and the two left rear elements. The right element can be found from the left simply by switching the left and right identification. These elements are
CL: Matrix element for left input channel to center output CR: Matrix element for right input channel to center output LFL: Left input channel for left front output LFR: Right input channel for left front output LRL: Left input for left rear output Channel LRR: Right input channel for left rear output.

These factors are not constant. Its value changes as a two-dimensional function of the clear direction of the input sound. Most phase / amplitude decoders determine the unambiguous direction of the input by comparing the amplitude ratios of the input signals. For example, the directivity in the left and right directions is determined from the ratio of the amplitude of the left input channel to the amplitude of the right input channel. Similarly, the directivity in the front-rear direction is determined from the ratio of the amplitude of the sum and difference of the input channels. Although the logic 7 decoder is significantly different from the standard decoder in the method of determining the orientation direction, the text does not discuss the method for determining these orientation directions. Assume that the pointing direction is determined. In the text,
Assume that these directions are represented as angles, ie, one angle (lr) for the left and right directions and one angle (cs) for the front and rear (center / surround) directions.
The two directivity directions are variables with signs. When both lr and cs are zero, the input signal is not directed—that is, the two input channels are not correlated.

When the input consists of a single directionally encoded signal, the two pointing directions have their maximum values. However, under such conditions they are not independent. The advantage over expressing the directional values as an angle is that when there is only one signal, the absolute value of the two directional values is 45.
It must be added up to degrees. When the input includes material that is not correlated with a strongly directed signal, the sum of the absolute values of the directional values must be less than 45 degrees.

| Lr | + | cs | ≦ 45 (2)
If the values of matrix elements are plotted on a two-dimensional surface formed by directivity values, the center of this surface has the value (0, 0), and the appropriate value for the sum of directivity values does not exceed 45.
In fact, the sum cannot exceed 45 due to the behavior of the non-linear filter—a circuit that limits the smaller of lr and cs as claimed in US patent application Ser. No. 08 / 742,460.
The sum did not exceed 45 degrees. This claim will not be discussed further in the text. The mathematical relationship to the matrix elements should behave well during overrun. When displaying a matrix element in a graph, the value is arbitrarily set to zero when the appropriate value of the input variable is exceeded. This makes it possible to see directly the behavior of the element along the projection of the boundary,
─ A strong signal follows after projection. The graphics were generated by the Matlab language. In the Matlab language, the Matlab language requires the angle variable to be one greater than the actual angle value, so the undirected position is (46,46). This should not be overly confused.

Previous designs for matrix decoders only consider the behavior of the matrix for strongly directed signals, i.e. the behavior of matrix elements around the boundary of the plane. This is a fundamental mistake in perspective. This is because when the actual signal is examined in either film or music, the boundary of the surface is hardly reached. In most cases
The signal oscillates near the middle of the surface, that is, slightly forward in the middle. The behavior of the matrix under these conditions is very important for sound. When comparing these elements to the previous ones, a significant increase in surface complexity in the intermediate region can be found. It is this complexity that plays a role in improving sound.

Such compositeness has value. Our original 1987 design (see US patent in 1989) was simple in construction with analog components. The new elements are designed to be almost completely described by a one-dimensional index table that is trivial in the digital configuration. Designing analog versions with similar performance is possible but not trivial.

In this application, we contrast several different versions of matrix elements. The oldest is an element from our US patent in 1989. These elements are used in our first surround processor and are the same as the standard (Dolby) surround processor elements in the left, center and right channels (not the same in the surround channel) . In our design, the surround channel is handled symmetrically with respect to the central channel. In a standard (Dolby) decoder, the surround channel is handled differently, and this issue is discussed in detail later in this application.

The elements described here are not always scaled correctly. In general, these elements have a non-oriented value of 1 for non-zero matrix elements for any given channel.
Shown to be. In practice, these elements are usually scaled so that the maximum value of each element is 1 or less. In either case, in the final product, the element scaling is further modified in the calibration procedure. The matrix elements shown herein can be considered scaleable by appropriate constants.

10. Assume that the left front matrix elements cs and lr in the 1989 US patent are pointing directions in degrees on the center / surround and left / right axes, respectively.

In the 1989 US patent, the formula for the forward matrix element is given as:

In the left front quadrant,
LFL = 1-0.5 * G (cs) + 0.41 * G (lr) (3a)
LFR = -0.5 * G (cs) (3b)
In the right front quadrant,
LFL = 1-0.5 * G (cs) (3c)
LFR = -0.5 * G (cs) (3d)
In the left rear quadrant (recall that cs is negative)
LFL = 1-0.5 * G (cs) + 0.41 * G (lr) (3e)
LFR = -0.5 * G (cs) (3f)
In the right rear quadrant,
LFL = 1-0.5 * G (cs) (3g)
LFR = -0.5 * G (cs) (3h)
The function G (x) is determined experimentally in the 1989 US patent and mathematically defined in the 1991 US patent. This function changes from 0 to 1 when x changes from 0 to 45 degrees. When the orientation is in the left front quadrant (lr and cs are both positive), G (x) can be shown equal to 1− | r | / | l |, where | r | and | l |
Is the left and right input amplitude. G (x) can also be described in terms of the pointing angle using various mathematical expressions. One of these formulas is shown in the 1991 US patent and another formula is shown later in the text. LFL plotted in three dimensions against lr axis and cs axis
See FIG. 5 and FIG. 6 for graphs of LFR and matrix elements.

In document [1], these matrix elements have been improved by adding the requirement that the loudness of the unoriented material must be constant regardless of the orientation direction. This means mathematically that the sum of the mean square roots of the LFL and LFR matrix elements must be constant. In the paper it was pointed out that this goal should be changed in the direction of orientation, i.e. the sum of squares of these matrix elements should be increased by 3 dB when the orientation is all left. FIG. 7 shows the sum of squares of these matrix elements, indicating that the matrix elements do not meet certain loudness requirements. In FIG. 7, this value is 0.71 along the axis from the position where it is not directed to the right.
Note that it is constant. Material that is not directed to the left increases by 3 dB to a value of 1 and decreases by 3 dB to a value of 0.5 if it is not directed to the center or back. That part of the graph is hidden by the peak on the left. The backward level is equal to the level in the central direction.

In US patent application Ser. No. 08 / 742,460 and document [1], the amplitude error in FIG. 7 was corrected by replacing the function G (x) in the matrix equation with a sine and cosine. See FIG. 8 for a graph obtained as a result of the sum of squares of correction elements LFL and LFR described below by equations (4a) to (4h).

Note that a constant value of 0.71 in the entire right half of the surface increases to 1 towards the left vertex. For the left front quadrant,
LFL = cos (cs) + 0.41 * G (lr) (4a)
LFR = -sin (cs) (4b)
For the right front quadrant,
LFL = cos (cs) (4c)
LFR = -sin (cs) (4d)
For the left rear quadrant,
LFL = cos (-cs) + 0.41 * G (lr) (4e)
LFR = sin (-cs) (4f)
For the right rear quadrant,
LFL = cos (-cs) (4g)
LFR = sin (-cs) (4h)

11. Improved separation for left front matrix elements In March 1996, several changes were made to these matrix elements. While maintaining the basic function dependency, further boosts were added along the cs axis in the front and cuts were added along the cs axis in the rear. The reason for the boost was to improve performance with stereo music panned forward. The purpose of the cut behind is
It was to increase the separation between the front and rear channels when stereo music was panned backwards.

For the front left quadrant,
LFL = (cos (cs) + 0.41 * G (lr)) * boost1 (cs) (5a)
LFR = (-sin (cs)) * boost1 (cs) (5b)
For the right front quadrant,
LFL = (cos (cs)) * boost1 (cs) (5c)
LFR = (-sin (cs)) * boost1 (cs) (5d)
For the left rear quadrant,
LFL = (cos (-cs)) + 0.41 * G (lr)) / boost1 (cs) (5e)
LFR = (sin (cs)) / boost (cs) (5f)
For the right rear quadrant,
LFL = (cos (cs)) / boost (cs) (5g)
LFR = (sin (cs)) / boost (cs) (5h)
The function G (x) is the same as the function in the 1989 US patent. When expressed as an angle as an input, the function can be expressed equally as:

G (x) = 1-tan (45-x) (6)
The function boost1 (cs) as used in March 1997 is the first 22.5
A total 3 dB linear boost applied over a degree orientation, with the next 22.5 degrees being 0 dB
Reduce to B again. Boost1 (cs) is the following matlab code corr (x
) (Comment lines are preceded by the percent sign%).

Calculate +3 dB boost function at% 22.5 degrees Hold% corr (x) up to 3 dB, raise corrl (x) and then fall again forx = 1: 24x;% x ranges from 0 to 23 Corr (x) = 10∧ (3 * (x−1) / (23 * 20) 9 having a value of 1 to 24 representing degrees; corr (x) = corr (x );
endfor
x = 25: descent for 46% corrl This range exceeds 24 degrees from 25 degrees corr (x) = 1.41:
corrl (x) = cor (48−x);
end
See FIG. 9 for a plot of LFL resulting from equations (5a) through (5h). Note that both boosts are applied along the lr = 0 axis and from the left to the center boundary as the pointing moves toward the center. Note also the drop in level as the pointing moves backwards.

The performance of the March 1997 circuit can be improved. The first problem lies in the behavior of pointing along the boundary between left and center and between right and center. As one strong signal pans from left to center, it can be seen in FIG. 9 that the LFL matrix element vehicle increases to a maximum of half between left and center. Such an increase in value is an unintended result of a systematic increase in levels for the left and right main outputs when the central signal is added to stereo music.

When the stereo signal is panned forward, it is desirable that the left and right front outputs should increase in level to compensate for the matrix removal of correlation components from these outputs. However, the method used to increase the level under these conditions should only occur when the input lr component is minimal, i.e. there is no net orientation left or right. 199
The method chosen to achieve such an increase in March 2007 did not depend on the value lr and resulted in an increase in level when a strong signal was panned across the boundary.

Boost is only needed along the lr = 0 axis. When lr is non-zero, the matrix element should not be boosted. This problem can be solved by using addition terms instead of multiplication for matrix elements. A new directivity index which is a boundary limit cs value is defined by the following Matlab code.

Both lr and cs are> 0, ie in the left front quadrant (cs
And the matlab convention where lr varies from 1 to 46)

ｃｓ＜２２．５およびｌｒ＝０（マトラブ規約では、ｃｓ＜２４およびｌｒ＝１）であ
るならば、ｂｃｓはｃｓに等しい。しかし、ｌｒが増加すると、ｂｃｓはゼロに低減する
。ｃｓ＞２２．５ならば、ｌｒが増加するときｂｃｓもまた低減する。

If cs <22.5 and lr = 0 (in the Matlab convention, cs <24 and lr = 1), bcs is equal to cs. However, as lr increases, bcs decreases to zero. If cs> 22.5, bcs also decreases as lr increases.

Next, to find the necessary correction function, find the difference between the boosted and non-boosted matrix elements along the lr = 0 axis. This difference is cos
Called _tbl_plus and sin_tbl_plus. Using matlab code

ベクトルｓｉｎ＿ｔｂｌ＿ｐｌｕｓおよびｃｏｓ＿ｔｂｌ＿ｐｌｕｓは、平坦な正弦お
よび余弦と、ブーストされた正弦および余弦との間の差異である。次に、下記のように定
義する。

The vectors sin_tbl_plus and cos_tbl_plus are the differences between the flat sine and cosine and the boosted sine and cosine. Next, the definition is as follows.

LFL = cos (cs) + 0.41 * G (lr) + cos_tbl_plus (bcs) (7a)
LFR = -sin (cs) -sin_tbl_plus (bcs) (7b)
LFL and LFR in the front right quadrant are similar but there is no + 0.41 * G term. These new definitions lead to the matrix elements shown in the graph of FIG.

Note in FIG. 10 that the new element has the proper amplitude along the left-to-center boundary, as well as along the center-to-right boundary.

The orientation in the backward quadrant is also not optimal. When the orientation is backwards, the matrix element is given by:

LFL = cos_tbl_minus (-cs) + 0.41 * G (-cs) (8a)
LFR = sin_tbl_minus (-cs) (8b)
These matrix elements are almost identical to the elements in the 1989 US patent. Consider the case where a strong signal pans from left to back. The matrix element of the 1989 US patent is that this signal is completely backward (cs = −45, l
Only when r = 0), it was designed to completely cancel the output from the front left output.
However, in the logic 7 decoder, it is desirable that the output from the left front output is zero when the encoded signal reaches the left rear direction (cs = -22.5 and lr = 22.5). The left front output must remain zero as the signal pans further all the way back. The matrix element used in March 1997—described above—results in an output on the front left channel of approximately −9 dB when the signal is panned to the left rear position. This level difference is not as good as it can be, even if it is sufficient for the excellent performance of the matrix.

The performance can be improved by changing the LFL and LFR matrix elements in the left rear quadrant. Note that it is of interest here how the matrix elements change along the boundary between left and rear. In order to find the behavior of the matrix elements along the boundary, the mathematical method shown in document [1] can be used. Suppose that when t changes from 0 (left) to -22.5 degrees (left rear), the amplitude of the left front output is reduced by the function F (t). This method gives the following matrix elements:

LFL = cos (t) * F (t)-/ + sin (t) * (sqrt (1-F (t) ^ 2)) (9a)
LFR = (sin (t) * F (t) +/- cos (t) * (sqrt (1-F (t) ^ 2))) (9b)
If F (t) = cos (4 * t) is selected and the correct signal is selected, these simplify to the following equation.

LFL = cos (t) * cos (4 * t) + sin (t) * sin (4 * t) (9c)
LFR = (sin (t) * cos (4 * t) -cos (t) * sin (4 * t) (9d)
Refer to FIG. 11 for plots of these coefficients LFL (solid curve) and LFR (dotted curve) against t. (Since all angles in the Matlab code are integers, a slight defect in the middle is due to the lack of a point at 22.5 degrees) These matrix elements work well-t is between 0 and 22.5 degrees When changing to degrees, the front left output is reduced smoothly to zero. When the pointing continues from 22.5 degrees to 45 degrees (all rear), the output is required to stop at zero. Along this part of the boundary,
LFL = -sin (t) (10a)
LFR = cos (t) (10b)
Note that these matrix elements are far from the matrix elements along the lr = 0 boundary. Here, in the literature [1], this value is LFL = cos (cs) (10c)
LFR = sin (cs) (10d)
Met.

Note that these matrix elements are designed to behave properly with strong directional signals-where both cs and lr have maximum values. When lr is almost zero
That is, when there is a stereo signal panned backwards, the previous matrix elements were good for the signal. When lr and cs approach the boundary, a method of smoothly transforming the old matrix elements into newer matrix elements is required. Linear interpolation can be used. In the processor used in Lexicon products that become expensive as the number increases,
A good strategy is to define a new variable—minimum values of lr and cs, as defined by the Matlab segment below. That is,

そして、ｂｐに依存する新たな補正関数

And a new correction function depending on bp

次いで、当該象限におけるＬＦＬおよびＬＦＲを下記のように定義する。即ち、
LFL＝cos(cs)/(cos(cs)+sin(cs))-front＿boundary＿tbl(bp)+0.41*G(lr)（１１ａ）
LFR＝sin(cs)/(cos(cs)+sin(cs))+front＿boundary＿tbl(bp) （１１ｂ）
ｃｏｓ（ｃｓ）＋ｓｉｎ（ｃｓ）の補正に注目されたい。ｃｏｓ（ｃｓ）をこの係数で
除すと、関数１−０．５＊Ｇ（ｃｓ）を得るが、これは当該象限におけるドルビー・マト
リックスと同じものである。ｓｉｎ（ｃｓ）をこの係数で除すと、往時の関数＋０．５＊
Ｇ（ｃｓ）を得る。

Next, LFL and LFR in the quadrant are defined as follows. That is,
LFL = cos (cs) / (cos (cs) + sin (cs))-front_boundary_tbl (bp) + 0.41 * G (lr) (11a)
LFR = sin (cs) / (cos (cs) + sin (cs)) + front_boundary_tbl (bp) (11b)
Note the correction of cos (cs) + sin (cs). Dividing cos (cs) by this factor gives the function 1-0.5 * G (cs), which is the same as the Dolby matrix in that quadrant. When sin (cs) is divided by this coefficient, the forward function + 0.5 *
G (cs) is obtained.

Similarly, in the right rear quadrant,
LFL = cos (cs) / (cos (cs) + sin (cs)) = 1-0.5 * G (cs) (12a)
LFR = sin (cs) / (cos (cs) + sin (cs)) = 0.5 * G (cs) (12b)
See FIGS. 12 and 13 for a graphical representation of these values.

Note the large correction along the left rear boundary in FIG. 12, which shows the coefficient graph from the left rear. This causes the front left output to be zero when the orientation goes from left to left back. As the pointing moves all the way back, this output remains at zero. The function is the same as the Dolby matrix along the lr = 0 axis and in the right rear quadrant.

Note the large peak at the boundary from left to back in FIG. This works for LFL matrix elements to keep the front output at zero along the boundary when the orientation goes from left rear to full rear. Again, the matrix elements are the same as the Dolby matrix in the backward quadrant along the lr = 0 axis.

One of the main design goals of the Logic 7 matrix design is that the loudness at any given output of undirected material present at the decoder output must be constant regardless of the direction of the directional signal at the same time. It is not to be. As explained earlier, this means that the sum of squares of the matrix elements for each output must be 1 regardless of the orientation direction. As explained earlier, this requirement must be changed when there is a strong orientation in the direction of the output in question. That is, if it is towards the left front output, the sum of squares of the matrix elements must increase by 3 dB when the orientation is all left. The above matrix elements also slightly change the requirement when the orientation moves in the front-back direction along the lr = 0 axis.

However, the success of our design can still be verified by plotting the square root of the sum of squares of the matrix elements. See FIGS. 14 and 15 for these plots for the modified design.

In FIG. 14, note the 3 dB peak in the left direction and the slightly weak peak when the signal is not directed in the center direction at 22.5 degrees. (For this plot, we can remove the 1 / (sin (cs)) correction in the rear quadrant and know how exactly the resulting sum is 1). This is the result of careful boosting of the left and right outputs in half forward orientation. Note that in other quadrants, the sum of rms is very close to 1, as was the design intent. The value in the rear left quadrant is not very close to 1 when the method used to produce the matrix elements approximates but the match is slightly better.

In FIG. 15, the axis that is not oriented to the right (middle) has a value of 1, and the central vertex has a value of 0.
. 71, the rear vertex has the value 0.5, and the left vertex has the value 1.41. Note the peak along the middle with respect to the central axis.

12 Back Matrix Element in Forward Orientation The back matrix element in the 1989 US patent (except that a scaling of only 0.71 was introduced here to show the effect of a standard calibration calibration procedure) is given by It is done. That is,
For the front left quadrant,
LRL = 0.71 * (1-G (lr)) (13a)
LRR = 0.71 * (-1) (13b)
For the rear left quadrant,
LRL = 0.71 * (1-G (lx) + 0.41 * G (-cs)) (13c)
LRR = -0.71 * (1 + 0.41 * G (-cs)) (13d)
(The right half of the face is the same, but switches between LRL and LRR)
Back matrix elements in Dolby Pro Logic (after a similar calibration calibration)
For the front left quadrant,
LRL = 1-G (lx) (14a)
LRR = -1 (14b)
For the rear left quadrant,
LRL = 1-G (lr) (14c)
LRR = -1 (14d)
(The right half of the face is the same, but switches between LRL and LRR)
Note that when cs = −45 degrees, the Dolby element and the 1989 US patent element are calibrated to be equal in the rear left quadrant.

13. Small deviation in surround level in Dolby Pro Logic This Dolby element is similar to our 1989 US patent, but there is no cs-dependent boost in the back. Such differences are actually very important when the elements have very different values for signals that are not directed after a standard calibration procedure. In general, this matrix element description does not consider the calibration procedure for three decoders. Get all matrix elements with relatively arbitrary scaling. In most cases, the element is provided as if it had a maximum value of 1.41. In fact, for technical reasons, the matrix elements are finally scaled to have a maximum value less than one. Furthermore, when the decoder is finally put into use, the gain of each output to the loudspeaker is adjusted.
To adjust the matrix elements, four main directions with equal acoustic power-left, center,
The right and surround encoded signals are reproduced and the gain of each output is adjusted until the sound output is equal at the listening position. This actually means that the actual level of the matrix elements is scaled so that the four outputs of the decoder are equal under omnidirectional conditions. Clearly, such a calibration was included in the formula for the posterior element described above.

The 3 dB difference in elements under forward-oriented or non-oriented conditions is not meaningless. In undirected conditions, an element from a 1989 US patent has a value of 0.71,
The sum of squares of the elements has the value 1. This is not valid for the Dolby element when calibrated. LRL has a non-oriented value of 1, and the sum of squares is 2 to 3 dB higher than the output of the 1989 US patent. Note that when the matrix is not oriented, the calibration procedure results in a matrix that does not correspond to a “Dolby Surround” passive matrix. Dolby Surround Passive Matrix has a rear output of 0.71 *
It must have a value of (A _in −B _in ) and the Dolby Pro Logic matrix specifies that such a specification is not met. The result is that the rear outputs are 3 dB stronger than the other outputs when inputs A and B are uncorrelated. If there are two speakers sharing the rear output, each 3 dB more than a single rear speaker when the decoder inputs are uncorrelated
It is adjusted to be soft, which gives all five speakers approximately the same sound output. When matrix elements from the 1989 US patent are used, the same calibration procedure results in 3 dB less acoustic output from the back when the decoder inputs are not correlated.

When the inputs are uncorrelated, the question of how loud the rear channel should be is ultimately a matter of preference. When surround-encoded recordings are played, you want to reproduce the balance that the producer heard when the recordings were mixed.
Achieving such balance is a design goal for decoders and encoders in combination combinations. However, in the case of standard stereo material, the goal is to reproduce the output balance of the original recording while creating an interesting and unobtrusive surround. The problem with Dolby Matrix elements is that the output balance in conventional two-channel recording is not stored in the matrix. The surround channel is too strong and the center channel is too weak.

To know the importance of such a problem, consider what happens when there is an input to a decoder consisting of three components: a left and right component and a separate central component.

A _in = L _in -0.71 * C _in (15a)
B _in = R _in + 0.71 * C _in (15b)
When A _in and B _in are played by a conventional stereo system, the acoustic output in the room is proportional to L _in ² + R _in ² + C _in ² . If all three components have approximately the same amplitude, the output ratio for the component with the right added to the left of the central component is 1: 2.

L _in and R _in regardless of the output ratio of C _in respect, I want to play the sound output indoors decoder in stereo and substantially the same power ratio. This can be expressed mathematically. If all other matrix elements are used as given, the substantially equal power ratio requirement defines the functional form of the central matrix along the cs axis. Given a Dolby matrix element calibrated so that the rear acoustic output is 3 dB less than the other three outputs when the matrix is fully oriented, ie 3 dB less than the standard calibration, the central matrix element is Should have the shape shown in FIG. The same can be done for a standard calibration, resulting in FIG.

In FIG. 16, assuming that the output ratio at the decoder output should be the same as the output ratio in stereo, if the rear Dolby matrix element is calibrated 3 dB lower at the level typically used, the actual value is While this produces reasonable results for undirected and fully directed signals, this actual value is about 1.5 dB lower in the middle.

In FIG. 17, assuming that an equal output ratio to stereo output results in calibration with matrix elements actually used in Dolby Pro Logic, the actual value is less than 3 dB, which is too low for all values of cs. Note the high price.

These two figures show that mix engineers often notice that a mix prepared for reproduction in a Dolby Pro Logic system requires more central loudness than a mix prepared for stereo reproduction. Show you get. Conversely, a mix prepared for stereo loses the clarity of the sound when played on a professional logic decoder. Ironically, this is not true for a passive Dolby Surround decoder. When discussing the central matrix element, we will again address this issue.

14 Generation of Two Independent Back Outputs The main problem with both the 1989 US patent element and the Dolby element is that there is only one back output. A 1991 US patent discloses a method for generating two independent lateral outputs, the mathematical processing of which is described in 1996 [1] and US patent application Ser. No. 08 / 742,460. It was included in the front left quadrant. The goal of the element in that quadrant was to remove the output of the signal directed from left to center while maintaining some output from the left rear channel for the simultaneously undirected ZAR. In order to achieve such a goal, the LRL matrix element has the following form.

For the left front quadrant,
LRL = 1-GS (lr) -0.5 * G (cs) (16a)
LRR = -0.5 * G (cs) -G (lr) (16b)
As can be seen, these matrix elements are very similar to the elements of the 1989 US patent, but with the G (lr) term added to the LRR and the GS term added to the LRL. G (lr) was included to add the signal from the decoder's B input channel to the left rear output and provide some undirected signal output when the directed signal is being removed. Next, a solution for the function GS (lr) was obtained using the criterion that there should be no signal output when a fully directed signal moves from left to center .

In document [1], these factors are corrected by giving a boost of (sin (cs) + cos (cs)) to approximate a certain loudness for materials that are not directed to themselves. Although quite good in the right front quadrant, such correction is not so good in the left front quadrant. See FIG. (For the right front quadrant, the matrix elements are the same as the LRL and LRR elements in the 1989 US patent.)

In FIG. 18, in the front left quadrant, 3 dB along the line from the middle to the left vertex.
Note that there is a boost at a level of approximately 3 dB along the boundary between the left and center. The mountain range in the rear quadrant will be discussed later. In the same figure, FIG.
In order to allow a good contrast to the present invention at 0, the “tv matrix” correction at V1.11 has been removed.

FIG. 18 illustrates some problems with the sound output. First consider the slack in the sum of squares along the cs = 0 axis. This slack exists because the functional form of G (lr) in LRR is not optimal. The selection of G (lr) is arbitrary, and this function already exists in the decoder in the design of the past, and its realization in the analog circuit is easy.

Ideally, the function includes the function GR (lr), and the sum of squares of LRL and LRR is cs =
We want to choose GS (lr) and GR (lr) to keep constant along the zero axis and to keep the output at zero along the boundary between the left and center. This is possible. We also want to confirm that the matrix element is the same as the matrix element in the right front quadrant along the lr = 0 axis. For this reason, the following is assumed. That is,
LRL = cos (cs) -GS (lr) (17a)
LRR = -sin (cs) -GR (lr) (17b)
We want the sum of squares to be 1 along the cs = 0 axis.

(1-GS (lr)) ² + (GR (lr)) ² = 1 (18)
And we want the output to be zero for the directed signal, i.e. t changes from zero to 45 degrees.

LRL * cos (t) + LRR * sin (t) = 0 (19)
Equations 18 and 19 result in cumbersome quadrants for GR and GS,
This is solved mathematically and graphed in FIG. The use of GS and GR as shown results in a significant improvement in the output sum along the cs = 0 axis. However, a peak in the sum of squares along the boundary between the left and center remains.

In actual design, compensating for such errors is probably not as important, but we decided to do this heuristically by the following strategy. Divide both matrix elements by a factor that depends on the new variable combination based on lr and cs. This new variable is called xymin. (In actuality, multiplication is performed by the reciprocal of the coefficient described below without using division.
) In Matlab notation, find the minimum of% x or y xymin = x;
if (xymin> y)
xymin = y;
end
if (xymin> 23)
xymin = 23;
end
% Indicates that xymin varies from zero to 22.5 degrees.

Next, xymin is used to know the correction for the matrix elements along the boundary.
In the front left quadrant,
LRL = (cos (cs) -GS (lr)) / (1 + 0.29 * sin (4 * xymin)) (20a)
LRR = (-sin (cs) -GR (lr)) / (1 + 0.29 * sin (4 * xymin)) (20b)
In the front right quadrant,
LRL = cos (cs) (20c)
LRR = -sin (cs) (20d)
In document [2], these elements are further multiplied by a “tv matrix” correction. FIG. 20 herein shows matrix elements that do not perform “tv matrix” correction. In the present application, such correction is processed by a frequency dependent circuit following the matrix and will be described later.

Note in FIG. 20 that the sum of squares is close to 1 and continuous except for an intentional increase in the level at the back.

15. Back Matrix Element in Back Orientation The rear matrix element shown in the 1991 US patent is not suitable for a 5-channel decoder and has been modified practically in our CP-3 product. Document [1] and US patent application Ser. No. 08 / 742,460 showed a mathematical method for obtaining these elements along the boundary of the left rear quadrant. The method works along the boundary, but along the lr = 0 axis and cs =
This resulted in discontinuities along the 0 axis. In the March 1997 US patent, these discontinuities were (almost) repaired by further corrections to the matrix elements, which retained the behavior of the matrix elements along the orientation boundary.

For the elements described in this application, these errors were corrected by interpolation. The first interpolation fixes the discontinuity along the cs = 0 boundary for the LRL. This interpolation method matches the value to the value of GS (lr) when cs is zero, and smooths this value to the value given by the previous match when cs increases in the negative direction backwards. Increase to. The second interpolation method is L
RR is interpolated along the cs = 0 axis to the value of GR (lr).

16. Left / Right Output in Right-to-Right Back Orientation First consider the left back left and left back right matrix elements when the orientation is neutral or somewhere between all right and right back. That is, lr can vary from 0 degrees to -45 degrees, and cs can vary from 0 degrees to -22.5 degrees.

Under these conditions, the directional component of the input must be removed from the left output-there should be no output from the rear left channel when the directivity is to the right or right rear.

The matrix elements described in the 1991 US patent achieve this goal. These elements are substantially the same as the back matrix elements in a 4-channel decoder with the addition of sin (cs) + cos (cs) correction for undirected loudness. When this is done, the matrix elements are simple sine and cosine. That is,
LRL = cos (-cs) = sri (-cs) (21a)
LRR = sin (-cs) = sric (-cs) (21b)
New function sri (x) equal to sin (x) over the range from 0 degrees to 22.5 degrees
And define sri (x) equal to cos (x). We will again use these functions in the definition of the left rear matrix element in the left orientation.

17. Left / Right Output in Backward Orientation from Right Back to Back Next, consider the same matrix element when cs is greater than -22.5 degrees. As stated in document [1] and the two aforementioned US patent applications, LRL should be increased to 1 or more over the range and LRR should be reduced to zero. A simple function meets this (recall that cs is negative and varies from -22.5 degrees to -45 degrees in the following equation).

LRL = (cos (45 + cs) + rboost (-cs)) = (sri (-cs) + rboost (-cs)) (22a)
LRR = sin (45 + cs) = sric (-cs) (22b)
Rboost (cs) is defined in document [1] and US patent application Ser. No. 08 / 742,460. This is very close to the function 0.41 * G (cs) in the old matrix element, except that rboost (cs) is zero for 0>cs> −22.5, and cs is − It changes from zero to 0.41 when changing from 22.5 degrees to -45 degrees. Its exact function form is determined by the requirement to keep the loudness of the rear output constant when the sound is panned from left rear to full rear.

  The left rear matrix element during right orientation is now complete.

18. The behavior of the left rear element left rear left and left rear right element in left-to-left rear orientation is much more complex. lr is 45 degrees to 22 degrees
. When decreasing to 5 degrees or to zero degrees, the left rear left element increases rapidly from zero to approximately the maximum value. The matrix element shown in document [1] does this, but as indicated earlier, there is a problem due to continuity at the cs = 0 boundary.

In the March 1997 release, a solution was found that uses a function of one variable and several conditional terms. In the literature [1], on the front side of the boundary (cs ≧ 0), the LRL matrix element is given by GS (lr), thus causing a problem at the boundary of cs = 0. On the rear side (cs <0), the function given by document [1] has the same termination point, but is different when lr is not zero or 45 degrees.

The mathematical method in document [1] provides the following equations for the left rear matrix elements over the range 22.5 <lr <45 (when rewriting these equations from document [1], t = 45−lr I want to recall that).
LRL = cos (45-lr) * sin (4 * (45-lr))-sin (45-lr) * cos (4 * (45-lr))
= Sra (lr) (23a)
LRR =-(sin (45-lr) * sin (4 * (45-lr)) + cos (45-lr) * cos (4 * (45-lr))) =-srac (lr) (23
b)
Over this range, two new functions, sra (lr) and srac (lr)
Note that

If cs ≧ 22.5, lr can still vary from 0 to 45. Reference [1]
(When lr has the range 0 <lr <22.5) LRL and LRR are defined as follows (see FIG. 6 in document [1]).
LRL = cos (lr) = sra (lr) (23c)
LRR = -sin (lr) =-srac (lr) (23d)
Two functions sra (x) and srac (x) are now defined for 0 <lr <45.

19. March 1997 version The March 1997 version uses an interpolation technique to modify the LRR along the boundary. Here there are two discontinuities. Along the cs = 0 boundary, the rear LRR needs to be aligned with the forward LRR indicating LRR = −G (lr) along the cs = 0 boundary.

The option used in March 1997 is to use interpolation based on the value of cs over a range of 0 to 15 degrees, albeit somewhat emphasized computationally. In other words, c
When s is zero, use G (lr) to find the LRR. When cs increases to 15 degrees, s
Interpolate to the value of rac (lr).

There is also the possibility of discontinuities along the axis lr = 0. In March 1997, this discontinuity was corrected (although somewhat) by adding a term to the LRR determined using the new variable cs_bounded. The correction term is simply sric (cs_bounde
d). This term guarantees continuity across the axis of lr = 0.

  First, cs_bounded is determined by the following Matlab notation.

20. LRL implemented in Logic 7 in August 1997
In the present invention, the LRL is calculated using an interpolation method like the LRR. In Matlab notation:

21. When the rear output directing between left rear to full rear direction goes from left rear to full rear, the element adds a correction to the rear volume according to what is shown in reference [1] Accompanied by. In Matlab notation:

これにより、左指向の間に、ＬＲＬおよびＬＲＲのマトリクス・エレメントが完成され
る。右指向に対する値は、定義における左と右とを交換することによって求めることがで
きる。

This completes the LRL and LRR matrix elements while facing left. The value for right orientation can be determined by exchanging left and right in the definition.

22. Both the Center Matrix Element 89 patent and Dolby Pro-Logic have the following matrix elements:

  For forward pointing:

後方指向に対しては以下のようである。

For backward orientation:

マトリクス・エレメントは、左右軸について対称であるので、右指向に対するＣＬおよ
びＣＲの値は、ＣＬとＣＲとを交換することにより求めることができる。このエレメント
の図式表現に関しては、図２１を参照せよ。

Since the matrix element is symmetric about the left-right axis, the values of CL and CR for the right orientation can be obtained by exchanging CL and CR. See FIG. 21 for a graphical representation of this element.

In FIG. 21, the value of the middle and right and rear vertices of the graph is 1. The value of the central vertex is 1.41. In practice, this element is scaled so the maximum value is 1.

In application number 08/742460 and reference [1], these elements are replaced by sine and cosine.

  For forward pointing:

これらの式は実施されなかった。１９９７年３月の製品は８９年の特許におけるエレメ
ントを使用したが、異なるスケーリングおよびＧ（ｃｓ）と異なるブースト関数を用いて
いる。我々は中央出力の指向されないレベルを低減させることが重要であったことを発見
し、プロ−ロジックのレベルよりも４．５ｄＢ小さい値を選択した。ブースト関数（０．
４１＊Ｇ（ｃｓ））は、ｃｓが中央方向に増加するにつれてプロ−ロジックの値に戻すよ
うにマトリクス・エレメントの値を増加するように、変更された。１９９７年３月バージ
ョンにおけるブースト関数は、傾聴テストを通じて発見的に選択された。

These formulas were not implemented. The March 1997 product used the elements in the 89 patent, but with different scaling and different boost functions with G (cs). We discovered that it was important to reduce the undirected level of the central output and chose a value 4.5 dB less than the pro-logic level. Boost function (0.
41 * G (cs)) has been modified to increase the value of the matrix element to return to the pro-logic value as cs increases in the middle direction. The boost function in the March 1997 version was selected heuristically through listening tests.

In the March 1997 version, the cs boost function starts as 0, as before, and in a manner that CL and CR increase by 4.5 dB as cs goes from 0 to 22.5 degrees. Ascend with. This increase is a constant dB for each dB increase in cs. The boost function then changes the slope in the next 20 degrees in such a way that the matrix element rises an additional 3 dB and then remains constant. That is, when the orientation is “half front” (8 dB or 23 degrees), a new matrix
The element is equal to the neutral value of the old matrix element. As the orientation continues to move forward, the new and old matrix elements will be equal. Thus, the output of the center channel is 4.5 dB less than the old output when directed to neutral, but rises to the old value when the directivity is fully centered.
See FIG. 22 for a three-dimensional view of this element.

Note that in FIG. 22, the median and right and rear vertices are reduced by 4.5 dB. As cs increases, the center increases to a value of 1.41 at two slopes.

It has been discovered that the central element used in March 1997 is not optimal. The central part of conversations in popular music recordings and movies tends to be lost when switching between stereo (two-channel) playback and playback through a matrix. Shown by experience. Furthermore, if the central channel changes level, a listener who is not equidistant from the front speaker can notice the movement of the apparent position of the central audio. This problem was further analyzed in the development of the new central matrix element presented here. As will be seen later, there is a problem when the signal pans from left to center or right to center along the boundary. The matrix element in application number 08/742460 causes the output from the central speaker to be too low when panning is in progress.

23. Whenever the central channel orientation in the new design is heading forward but not biased to the left or right, it is possible to remove the strongly directed signal from the central channel output using matrix technology. The central channel must reproduce the sum of the A and B inputs with some gain factor. In other words, it is not possible to remove uncorrelated left and right material from the central channel. The only thing you can do is adjust the volume of the center speaker. So what should be the volume?

This question depends on the nature of the left and right main outputs. The above matrix values for LFL and LFR are designed to remove the central component of the input signal as the pointing moves forward. We use the above-mentioned (elements of the 1996 AES paper, elements of March 1997, and this) when the input signal is encoded using a mixer, such as stereo width control. It can be shown that all of the matrix elements (presented earlier in the specification) are completely restored to their original separation.

However, if the input to the decoder consists of uncorrelated left and right channels with an unrelated center channel added, i.e.

の場合には、Ｃ_inのレベルがＬ_inおよびＲ_inに相対的に増加すると、デコーダのＬおよび
Ｒの前方出力のＣコンポーネントは、Ｃ_inがＬ_inおよびＲ_inと比較して大きくない限り、
完全に除去されない。一般に、Ｃ_in左方のビットがＬおよびＲの前方出力内にある。傾聴
者は何を聞き取るであろうか。

In the case of, the level of C _in increases relatively to L _in and R _in, C components of the front output of the decoder of L and R, unless C _in is not large compared to L _in and R _in ,
It is not completely removed. Generally, the _Cin left bit is in the L and R forward outputs. What will the listener hear?

There are two ways to calculate what the listener hears. When the listener is equidistant from the left, right and center speakers, the listener hears the sum of the sound pressures from each speaker. This is equal to the sum of the three forward outputs. Under these circumstances, it is easy to show that if the central components of the left and right speakers are reduced, the resulting sound pressure from the central component will result in residual loss regardless of the amplitude of the central speaker. It is. This is because the center speaker is always derived from the sum of the A and B inputs, and when its amplitude is increased, the amplitude of the L _in and R _in signals must increase with the amplitude of the C _in signal. is there.

However, if the listener is not equidistant from each speaker, the listener will tend to hear the sum of the sound output from each speaker, which is equal to the sum of the squares of the three front outputs. In fact, listening in detail shows that in fact, the sum of all speaker outputs is actually important, so we need to consider the sum of the squares of all the outputs of the decoder, including the rear outputs. is there.

When switching the matrix between stereo playback and matrix playback, L _in , R _in
, And Cin from the central output if it is desired to keep the ratio _in amplitude of Cin
The sound output of the _in component must rise in proportion to the reduction in its sound output from the left and right outputs and its reduction in the rear output. To make things more complicated,
The left and right front outputs have level boosts up to 3 dB as described above. This requires the middle one to keep the ratio constant by increasing the volume somewhat. This requirement can be written as a set of expressions for sound output. These equations can be solved for the gain function required for the center speaker.

Previously, a graph showing the energy relationship of a Dolby Pro-Logic decoder under various circumstances has been shown. Pro-logic decoders are not optimal. The same can be done for our new decoder.

FIG. 23 shows the central gain (solid curve) required when the energy of the central component of the input signal is maintained in the three forward channels as the pointing increases in the forward direction. As can be seen, the required rise at the level of the central channel is fairly steep, which is a large amplitude dB relative to the dB of the directed value. In addition, the gain (dotted curve) of the standard decoder is shown.

As mentioned above, there are two solutions to this problem. First, the “film (movie)” solution is described. This solution is not completely mathematical. In fact, the rise in the function shown in FIG. 23 is too steep. The change in the level of the central channel is too obvious. We decided to relax the required power a little and center it about 1 dB less than ideal. When recalculating the center value, the result shown by the solid line in FIG. 24 is obtained. In fact, the linear rise can be replaced with the initial part of the curve. In fact, using these median values was the best for the film.

Referring to FIG. 24, the rise of the solid curve is actually too steep. The linear gradient shown by the dashed line works better.

Music needs another solution. Given the matrix elements previously given for LFL and LFR, the central attenuation shown in FIGS. 23 and 24 is obtained. What if you used different elements? Specifically, do we need to be aggressive in removing the central component from the left front and right front outputs?

The listening test shows that the previous left and right front matrix elements are unnecessarily aggressive in removing the central component during music playback. Acoustically, they do not have to do such things. The energy removed from the left and right front needs to be applied to the central loudspeaker. If this energy is not removed, it comes from the left and right front speakers and the central speaker need not be strengthened. The sound output in the room is the same. This devise gives just enough energy to the central speaker to produce a forward image that is convincing for off-axis listeners, while stereo for listeners equidistant from the front left and front right speakers. It is to minimize the width reduction.

As described in application No. 08/742460, the optimum center volume can be found by trial and error. Next, answers can be obtained for the matrix elements required on the front left and front right to maintain the output of the Cin component _in the room. As before, assume that the level of the center channel is reduced 4.5 dB below the level of our 89 patent decoder, or a total attenuation of -7.5 dB. In addition,
-7.5 dB is equal to 0.42. The matrix element for the center can be multiplied by this factor, and a new center boost function (GC) can be defined as:

  For forward pointing:

  For backward orientation:

幾つかの関数をＧＣ（ｃｓ）に対して試みた。以下に示すものは最適ではなかも知れな
いが、十分であるように思える。これは、度で表した角度ｃｓに関して指定されたもので
あり、何度かの試行錯誤により得られたものである。

Several functions were tried for GC (cs). The following may not be optimal, but seems to be sufficient. This is specified with respect to the angle cs expressed in degrees, and is obtained by several trials and errors.

  In the MATLAB standard notation:

関数（０．４２＋ＧＣ（ｃｓ））を図２５に示す。値０．４２（ドルビー・サラウンド
よりも４．５ｄＢ低い）からの速い上昇に、緩やかな上昇が続き、最終的に、値１への急
な上昇が続くことに留意されたい。

The function (0.42 + GC (cs)) is shown in FIG. Note that the fast rise from the value 0.42 (4.5 dB below Dolby Surround) is followed by a gradual rise and eventually a sudden rise to the value 1.

Given the functions for LFL, LRL and LRR, the required function for LFR can be solved. It determined the rate at which C _in component in the left output and right output decreases, then it is desirable to design the matrix element to reduce its rate. These matrix elements also have _Lin components and R
_It should provide some boost of the _in component and should have the current shape at the left-to-center boundary and right-to-center boundary.

  Here, it is assumed as follows.

前方左および前方右からの出力は以下のように計算できる。

The output from the front left and front right can be calculated as follows.

中央からの出力は以下のようである。

The output from the center is as follows.

後方からの出力は、使用するマトリクス・エレメントに依存する。後方チャネルは前方
指向の間に３ｄＢ減衰され、ＬＲＬはｃｏｓ（ｃｓ）であり、ＬＲＲはｓｉｎ（ｃｓ）で
あると仮定する。１つのスピーカからは以下のようである。

The output from the back depends on the matrix element used. Assume that the back channel is attenuated by 3 dB during forward pointing, LRL is cos (cs), and LRR is sin (cs). From one speaker, it is as follows.

Ｌ_in ²≒Ｒ_in ²であると仮定すると、２つのスピーカに対して以下のようである。

Assuming that L _in ² ≈R _in ² , for two speakers:

３つの全てのスピーカからの合計出力は、ＰＬＲ＋ＰＣ＋ＰＲＥＡＲである。

The total output from all three speakers is PLR + PC + PREAR.

Ｃ_in出力の、Ｌ_inおよびＲ_in出力に対する比率は以下のようである（Ｌ_in ²＝Ｒ_in ²であ
ると仮定）。

The ratio of C _in output to L _in and R _in output is as follows (assuming L _in ² = R _in ² ):

通常のステレオに対しては、ＧＣ＝０、ＧＰ＝１、ＧＦ＝０である。従って、中央対Ｌ
Ｒ出力比は以下のようである。

For normal stereo, GC = 0, GP = 1, GF = 0. Therefore, the center pair L
The R output ratio is as follows.

アクティブ・マトリックスに対するＣ_in ²／Ｌ_in ²の値に拘わらずこの比が一定である場
合、

If this ratio is constant regardless of the value of C _in ² / L _in ² for the active matrix,

上記の方程式は数値的に解くことができる。我々がＧＣを上記のようにそしてＧＰ＝Ｌ
ＦＬを前のように仮定すると、我々はその結果を図２６に見ることができる。

The above equation can be solved numerically. We have GC as above and GP = L
Assuming FL as before, we can see the result in FIG.

In FIG. 26, the solid curve is a graph of the GF required for a constant energy ratio with a new “music” center attenuation GC. The dashed curve is the March 1997 LFR element, sin (cs) ^* corr1. The dotted curve is sin (cs), that is, the LFR element without the correction term corr1. Note that GF is close to zero until cs reaches 30 degrees and then increases rapidly. We have found that it is best to actually limit the cs value to about 33 degrees. In fact, the LFR derived from these curves has a negative sign.

As cs increases from zero to center, GF gives the shape of the LFR matrix element along the lr = 0 axis. We need a way to mix this behavior with the behavior of the previous LFR element, which must be preserved between the left and center and along the boundary from right to center. The method of doing this when cs ≦ 22.5 degrees is GF and sin (c
s) to define the difference function. We then limit this function in various ways. In the notation of Matlab, it is as follows.

ＬＦＲエレメントは、ここにマトラブの表記法で次のとおり書くことができる。

The LFR element can be written here in Matlab notation as follows:

Note that the sign of gf_diff is positive in the above equation. Gf
_Diff cancels the value of sin (cs) and reduces the value of the element to zero along the first part of the lr = 0 axis. See FIG.

In FIG. 27, note that as cs increases along the lr = 0 axis to ˜30 degrees, the value is zero in the middle of the plane (without steering) and remains zero. . The value then drops because it matches the previous value along the boundary from left to center and right to center.

24. Panning error at the center output
After all, if we write a new central function like this:
[Equation 26]
CL = 0.42 ^* (1-G (lr)) + GC (cs) (34a)
CR = 0.2 + GC (cs) (34b)
The new center function works well along the lr = 0 axis, but causes panning errors along the boundary between left and center and right and center. The value in reference [1] (which was never performed) gives a smooth function of cos (2 ^* cs) along the left boundary. These values produce a smooth panning between the left and center. we,
We would like our new central function to have a similar behavior along this boundary.

We can correct for the matrix elements that will do the job by adding an additional function of xymin (in matlab notation).

[Equation 27]
center_fix_tbl = 0.8 ^* (corr1-1)
Therefore,
[Equation 28]
CL = 0.42−0.42 ^* G (lr) + GC (cs)
+ Center_fix_table (xymin) (35a)
CR = 0.42 + GC (cs)
+ Center_fix_table (xymin) (35b)
See FIG. 28 for a three-dimensional display of CL matrix elements. Although not perfect, this correction actually works well.

In FIG. 28, note the correction for panning along the boundary between the left and center, which is quite smooth.

Note that in FIG. 29, which is a graph of the left front (dotted curve) and center (solid curve) output, the center operation is the left of the plot and the full left is the right. "musics"
In the strategy, we are currently limiting cs values to about 33 degrees (about 13 on the axis), where the center is about 6 dB stronger than the left.

25. Technical details of the encoder There are two main purposes of the logic 7 encoder. First, it should be able to encode 5.1 channel tapes in a way that allows the encoded version to be decoded by a logic 7 decoder with minimal subjective change. Second
In addition, the encoded output should be compatible with stereo, i.e. the encoded output should sound as close as possible to a manual two-channel blend of the same material. One factor in this stereo compatibility is that when the encoder output is played back in a standard stereo system, the encoder output is the same perceived loudness for each sound source in the original five-channel mix. Should be given. The apparent position of the sound source in stereo should also be as close as possible to the apparent position in the 5-channel original.

In the discussion by the Institute of Broadcasting Technology (IRT) in Munich, it was revealed that the purpose of stereo compatibility of stereo signals cannot be fulfilled by passive encoders, as mentioned above. Five channel recordings where all channels have equal foreground importance must be encoded as described above. This encoding requires that the surround channel be mixed to conserve energy in the encoder output. That is, the total energy of the encoder output should be the same regardless of which input is being driven. This constant energy setting may be necessary for most film (cinema) sources and for a five channel music source with instruments equally assigned to all five speakers. It is the author's opinion that even though such music sources are not currently common, they will become common in the future. Music recordings in which the foreground instrument is placed in the front three channels and has the main reverberation in the rear channel require different recordings.

After a series of tests of encoding (in IRT and elsewhere), this type of music recording was successfully encoded in a stereo-compatible format when the surround channel was mixed with 3 dB lower power than the other channels It was confirmed. This -3dB level was adopted as the standard for surround encoding in Europe, but the standard is
Note that other surround levels can be used for special purposes. The new encoder includes active circuitry that detects strong signals in the surround channel. When such a signal is occasionally present, the encoder uses a full surround level. If the surround input is consistently -6 dB or lower compared to the front channel, the surround gain is gradually reduced by 3 dB to accommodate European standards.

These active circuits are also described in patent application no. In the encoder in 08 / 742,460. However, an early encoder test at the Institute of Broadcasting Technology (IRT) in Munich found that the direction of one encoded sound source was not encoded correctly. A new architecture has been developed to solve these problems.
The performance of the new encoder is clearly superior for a wide variety of difficult materials. The original encoder was first developed as a passive encoder. The new encoder also operates in a passive mode, but is primarily intended to operate as an active encoder. Active circuitry corrects for some small errors inherent in the design. However, its performance is better than previous encoders even without active correction.

Extensive listening has found several other minor problems with the first encoder. Many (but not all) of these issues have been addressed in the new encoder. For example, when a stereo signal is applied simultaneously to both the front and rear terminals of an encoder, the resulting encoder output is too biased too far relative to the front. New encoders compensate for this effect by slightly increasing the back bias. Similarly, we have found that conversations can sometimes be lost when the film is encoded with substantially surround content. This problem has been greatly improved by the aforementioned changes to power balance, but the encoder is also intended to be used with a standard (Dolby) decoder. New encoders compensate for this effect by slightly increasing the center channel input to the encoder under these conditions.

26. Design Description The new encoder is left when the central damping function fcn is equal to 0.71 or -3 dB,
The center and right signals are processed the same as the previous design and the same as the Dolby encoder.

Surround channels look more complicated than they are currently. The functions fc () and fs () direct the surround channel either to a path with a 90 degree phase shift with respect to the forward channel or to a path without a phase shift. In the basic operation of the encoder, fc is 1 and fs is zero, ie only the path with 90 degree phase shift is active.

The value crx is typically 0.38. It controls the negative cross feed amount for each surround channel. As in the previous encoder, one of the surround channels
When there is only one input in one, the A and B outputs have an amplitude ratio of 0.38 / 0.91 resulting in an operating angle of 22.5 degrees towards the rear. As usual, the total power of the two output channels is 1 (unit), ie the sum of squares of 0.91 and 0.38 is 1.

The output of this encoder is relatively simple when only one channel is being driven, but that output is problematic when both surround inputs are being driven simultaneously. We are L
When the S and RS inputs are driven by the same signal (usually occurring in film), all signals at the summing node are in phase, so the overall level of each output channel is 0.3.
8 + 0.91, or 1.29. This output is too strong by a factor of 1.29, or 2.2 dB. Active circuitry is included in the encoder to reduce the value of the function fc to 2.2 dB when the two surround channels are similar in level and phase.

Another error occurs when the levels of the two surround channels are similar and out of phase. In this case, the two attenuation factors are reduced so that the A and B outputs have equal amplitude and phase and a level of 0.91-0.38, ie 0.53. This signal will be decoded as a center direction signal. This error is severe. The previous encoder design produced an unsteered signal under these conditions, but that signal is reasonable. It is not reasonable for the signal applied to the rear input terminal to result in a signal that is centered. Thus, an active circuit is provided that increases the value of fs when the two rear channels are similar in level and in opposite phase. The result of mixing both the back path and the phase shifted path for the back channel is a 90 degree phase difference between output channels A and B. This result results in an unmanipulated signal, which is what we want.

As mentioned earlier, I discovered that a European standard surround encoder exists during the IRT discussion in Munich. The encoder is simply two surround
Attenuate the channels by 3 dB and then add them to the front channel. Thus, the left rear channel is attenuated and added to the left front channel. This encoder has a number of drawbacks when encoding multi-channel film sound, or recordings that have a particular instrument in the surround channel. Both the loudness and direction of these instruments will be encoded incorrectly. However, this encoder works rather well for classical music, where the two surround channels are the main reverberations. 3dB
Attenuation was carefully selected through listening tests to produce a stereo compatible encoding. I found that our encoder was 3d when classical music was being encoded
We concluded that it should include B attenuation, and that one can detect this condition through the relative levels of the forward and surround channels in the encoder.

The main function of the function fc in the surround channel is to reduce the level of the surround channel in the output mix when the surround channel is much softer than the forward channel. A circuit is provided to compare the front and back levels, and when the back is less than 3 dB, the value of fc is reduced to a maximum of 3 dB. Maximum attenuation is achieved when the rear channel is 8 dB less intense than the front channel. This active circuit seems to work well. It makes the new encoder compatible with European standard encoders for classical music. The action of the active circuit causes the instrument intended to be strong in the rear channel to be encoded at full level.

There is another function fs of the path that mixes the real coefficients for the surround channel. As the sound moves from the left front input to the left rear input, the active circuit detects that these two inputs are similar in level and phase. Under these conditions, fc is reduced to zero and fs is 1.
Will be increased. This change to the real coefficient in the encoding is not limited to this type of pan (p
an) a more accurate decoding. In practice, this functionality is probably not essential, but it looks elegant and refined.

There are additional active circuits not yet released in the product. The level detection circuit looks at the phase relationship between the center channel and the front left and right. Some popular music recordings using 5 channels mix vocals into all three forward channels. When a strong signal is in all three inputs, the encoder output will have excessive vocal power because the three front channels join together in phase. When this occurs, the active circuit increases the center channel attenuation by 3 dB to restore the power balance at the encoder output.

In summary, active circuitry is provided for:
1. Reduce the surround channel level by 2.2 dB when the two channels are in phase.
2. The real coefficient mixing path for the rear channel is sufficiently increased to generate an unoperated condition when the two rear channels are out of phase.
3. When the surround level is much smaller than the front level, the surround channel level is reduced by 3 dB.
4). When the rear channel level resembles the front channel, the rear channel level and the opposite phase are increased.
5. When the sound source is panning from the front input to the corresponding rear input, surround channel mixing uses real coefficients.
6). When the center level and the front and surround input levels are approximately equal, the level of the center channel in the encoder is increased.
7). When the common signal is in all three forward inputs, it reduces the level of the center channel in the encoder.

Future improvements to the encoder are likely to include features similar to feature 2 above for the front channel. In current encoders, when the two front channels are out of phase, the encoding causes the encoder to place the sound behind. We intend to detect this condition and prevent the resulting output from being manipulated.

27. Frequency Dependent Circuit in the Encoder FIG. 2 shows a block diagram of a frequency dependent circuit according to a matrix in a 5-channel version of the decoder. Three sections: variable low-pass filter, variable shelf
Filter and HRTF (Head Related Transf
er Function)) filter. The HRTF filter changes its characteristics according to the value of the backward directing voltage c / s. The first two filters change their characteristics in response to signals intended to represent the average direction of the input signal to the decoder during the pause between the strongly directed signals. This signal is called a background control signal.

28. Background control signal One of the main goals of current decoders is that a 5-channel surround signal can be optimally generated from two normal channel stereo signals. It is also highly desirable for a decoder to regenerate a 5-channel surround recording encoded in two channels with the encoder described as part of this application. These two applications differ in how the surround channel is perceived. With normal stereo input, the majority of the sound needs to be in front of the listener. Surround speakers are enclosed (e
nvelopment) and the pleasant feel of the atmosphere, but should not draw attention to them. Encoded surround recordings require surround speakers to be stronger and more aggressive.

In order to optimally play back both types of input without any adjustment from the user, it is necessary to distinguish between 2-channel recordings and encoded 5-channel recordings. The background control signal is designed to make this discrimination. Background control signal (BC
S) is similar to and derived from the backward pointing signal cs. BCS represents the negative peak value of cs. That is, when cs is more negative than BCS, BCS is made equal to cs. When cs is more positive than BCS, the value of BCS decays slowly. However, the attenuation of the BCS involves further calculations.

Many types of music consist of a series of strong foreground sounds, or in the case of songs, a series of sung words. There is a background in the middle of the foreground notes. The background can consist of other notes played by other instruments, or it can consist of reverberation. B
The circuit that derives the CS signal tracks the peak level of the foreground note. When the current level is ˜7 dB below the foreground peak level, the level of cs is measured. The value of cs in these gaps between foreground peaks is used to control BCS attenuation. If the material in the gap between notes is reverberation, it tends to have a net backward bias in the recording made by encoding the five channel original. This is because the reverberation on the original rear channel will be encoded with a backward bias. The reverberation in a normal two-channel recording will have no net backward bias. The cs for this reverberation will be zero or slightly forward.

The BCS derived in this way tends to represent the type of recording. The BCS will always be strongly negative whenever there is significant backward-oriented material. However, B
CS can be negative even if the reverberation in the recording has a net backward bias, even in the absence of a strong backward orientation. We can use BCS to tune the filter that optimizes the decoder for stereo input versus surround input.

29. Frequency dependent circuit: 5 channel version The first of the filters in FIG. 2 is a simple 6 with adjustable cut-off frequency
It is a low pass filter of dB / octave. When BCS is positive or zero, this filter is set to a value that is user adjustable, but is typically about 4 kHz. As BCS goes negative, the cut-off frequency is increased, and when BCS is finally behind 22 degrees, the filter is not active. This low frequency filter reduces the rear output protrusion when normal stereo material is played. The filter is at least V1.11
Since then it was part of the decoder, but in early decoders it was controlled by cs and not by BCS.

The second filter is a variable shelf filter. The low frequency section (pole) of this filter is fixed at 500 Hz. The high frequency section (zero) varies depending on user adjustment and BCS. This filter performs a “sound stage” control in the current decoder. Patent application no. In 08 / 742,460, a “sound stage” is performed through a matrix element with a “tv matrix” correction. A forward decoder based on this work reduced the overall level of the rear channel when the orientation was neutral or forward. In the new decoder shown here, the matrix element is "
Does not include "tv matrix" correction.

In the new decoder, when the sound stage control is set to “back”, the high frequency section of the shelf filter is set equal to the low frequency section.
In other words, the shelf has no attenuation and the filter has a flat response.

When the sound stage control is set to “neutral”, the setting for high frequency zero changes. When the BCS is positive or zero, the zero moves to 710 Hz, resulting in a higher frequency 3 dB attenuation. For high frequencies, the result is the same as the initial decoder.
There is 3 dB attenuation when the pointing is neutral or forward. However, low frequencies are not attenuated. They come full level from the side of the room. The result is a greater low frequency richness and siege without diverting the high frequencies behind. B
As CS goes negative, the high frequency zero moves towards the pole, so when the BCS is about 22 degrees to the rear, the shelf filter has no attenuation.

When the sound stage control is set to “forward”, the effect is similar, but zero is B
Move to 1 kHz when CS is zero or positive. This gives 6 dB of attenuation at high frequencies. Again, as BCS goes negative, the attenuation is removed.

The third filter is controlled by c / s and not by BCS. This filter is designed to emulate the frequency response of the human head and pinna when the sound source is approximately 150 degrees azimuth from the front of the listener. This type of frequency response curve is
It is called “head related transfer function” or HRTF. These frequency response functions have been measured for many angles for many different people. Generally, when the sound source is about 150 degrees from the front, there is a strong notch in the frequency response at about 5 kHz. A similar notch exists when the sound source is in front of the listener, and only in this case the notch is about 8 kHz. A sound source to the side of the listener does not generate these notches. The human brain uses the presence of a notch at 5 kHz as one way it detects that the sound source is behind the listener.

Current standards for 5-channel sound reproduction recommend that the two rear speakers be placed slightly behind the listener and ± 110 or 120 degrees from the front. This speaker position provides a good surround at low frequencies. However, sound from the side of the listener does not produce the same level of excitement as the sound completely behind the listener. Film directors very often want sound effects that come from behind the listener and not from the side.

Also, the listening room often does not have the appropriate size and shape to place the speaker completely behind the listener, and the lateral position is often the best that can be achieved.

The HRTF filter in the decoder adds a frequency notch for the back source, so that the listener listens to the sound further behind the actual location of the speaker. The filter is designed to change with cs. When cs is positive or zero, the filter is maximum. This makes the ambient sound and reverberation appear to be behind the listener. As cs becomes negative, the filter is reduced. When cs is approximately -15 degrees, the filter is completely removed and the sound source appears to come completely from the side. As cs goes more negative, the filter is applied again, so the sound source seems to go behind the listener. When cs is sufficient for the rear, the filter is slightly modified to accommodate the HRTF function for sound that is sufficient for the rear.

30. Frequency Dependent Circuit: 7 Channel Version FIG. 3 shows the frequency dependent circuit in the 7 channel version of the decoder. These are shown in practice as being composed of three sections, even though the second two sections can be combined into one circuit.

The first two sections are identical to the two sections in the 5-channel decoder and perform the same function. The third section is specific to the 7 channel decoder. V1.1
1 and patent application no. At 08 / 742,460, the side and back channels had separate matrix elements. The effect of the element was such that when cs is positive or neutral, the side and rear outputs are identical except for the delay. The two outputs remained the same until cs was more negative than 22 degrees. As the pointing moved further backward, the side output was attenuated by 6 dB and the rear output was boosted by 2 dB. This makes the sound appear to move from the side of the listener to the rear of the listener.

In current decoders, the discrimination between the side output and the rear output is achieved by a variable shelf filter at the side output. The third shelf filter in FIG. 3 has no attenuation when cs is forward or zero. When cs becomes more negative than 22 degrees, the zero in the shelf filter moves quickly towards 1100 Hz, resulting in a high frequency attenuation of about 7 dB. Although the shelf filter has been described as a separate filter from the shelf filter that provides the “sound stage” function, the action of the two shelf filters can be combined into a single shelf filter via appropriate control circuitry. Can be combined.

While preferred embodiments of the present invention have been described and illustrated herein, there are many other possible embodiments and those and other modifications and changes will occur to those skilled in the art without departing from the spirit of the invention. It will be obvious.

The novel features believed characteristic of the invention are set forth in the appended claims. The invention itself, as well as other features and advantages of the invention, will be best understood by referring to the following detailed description of the embodiments with reference to the accompanying drawings.
FIG. 1 is a block diagram of a 2 to 5 channel matrix section and a direction detection section of the decoder of the present invention. FIG. 2 is a block diagram of a 5-channel frequency dependent active signal processor circuit connected between the output of the matrix section of FIG. 1 and the decoder output. FIG. 3 is a block diagram of a 5 to 7 channel frequency dependent active signal processor connected to any of the outputs of the matrix section of FIG. 1 and a decoder output. FIG. 4 is a block diagram of the active 5-channel to 2-channel encoder of the present invention. FIG. 5 is a three-dimensional graph of a prior art left front left (LFL) matrix element from the 1998 US patent and Dolby Pro-Logic scaled to a matrix value of 1. Show. FIG. 6 shows prior art from a 1998 US patent and Dolby Pro-Logic scaled to 0.71 with a minimum value of -0.5 and a maximum value of +0.5. Figure 3 shows a three-dimensional graph of a left front right (LFL) matrix element. FIG. 7 shows a three-dimensional graph of the square root of the square sum of the LFL and LFR prior art from the 1989 US patent, scaled to have a maximum value of one. FIG. 6 shows a three-dimensional graph of the square root of the sum of LFL and LFR from application number 08 / 42,460, scaled to have a maximum value of one. FIG. 9 shows a three-dimensional graph of the left front left (LFL) matrix of V1.11. FIG. 10 shows a particularly complete three-dimensional graph of the left front left matrix element of the present invention. FIG. 11 is a graph showing the action of LFL and LFR of the present invention along the rear boundary between the left and the front-rear. FIG. 12 shows a three-dimensional graph of the complete left front left (LFL) matrix element of the present invention as viewed from the left rear. FIG. 13 shows a three-dimensional graph of the complete left front right (LFR) matrix element of the present invention. FIG. 14 shows a three-dimensional graph of root means for squaring the sum of LFL and LFR of the present invention. FIG. 15 shows a three-dimensional graph representing the square root of the sum of the squares of LFL and LFR of the present invention being corrected for the rear level viewed from the left rear. FIG. 16 is a graph showing a solid line curve of the central matrix element to be used in the conventional Dolby Pro Logic as a DB unit CS function and a dotted curve of the actual value of the central matrix element of the Dolby Pro Logic decoder. FIG. 17 is a graph showing the solid curve of the ideal value of the central matrix element and the dotted curve of the actual value of the central matrix element of the conventional Dolby Pro Logic. FIG. 18 shows a three-dimensional graph representing the square root of the sum of the square of LRL and LRR used in the conventional element of V1.11. FIG. 19 is a graph representing a numerical solution of GS (lr) and GR (lr) for a constant power level along the cs = 0 axis, and zero output along the boundary between left and center. FIG. 20 shows a three-dimensional graph representing the square root of the sum of the squares of LRL and LRR using values for GR and GS of the present invention. FIG. 21 represents a conventional center left (CL) matrix element of a 1989 US patent 4-channel decoder (and Dolby Pro Logic decoder) representing a center right (CR) matrix element with the left and right interchanged. A three-dimensional graph is shown. FIG. 22 shows a three-dimensional graph representing the center left (CL) matrix element of the Logic 7V1.11 decoder. FIG. 23 is a graph showing the center output channel attenuation solid curve required for the new LFL and LFR, and the center attenuation dotted curve for a standard Dolby Pro Logic decoder. FIG. 24 is a graph showing a solid curve for ideal center attenuation for the “film” strategy of the present invention, a long curve for values that perform significantly better, and a dotted curve for attenuation for a standard Dolby decoder for comparison. FIG. 25 shows the central attenuation used in the “music” strategy of the present invention. FIG. 26 is a graph showing a solid line curve of a GF value that requires a constant energy ratio with a “music” center attenuation GC of the present invention and a dotted curve of a sin (cs) value. FIG. 27 shows a three dimensional graph representing the left front right (LFR) matrix element of the new invention during mid-level correction along the lr = 0 axis. FIG. 28 shows a three-dimensional graph representing the center left (CL) matrix element with the new center boost function. FIG. 29 shows a graph in which the output level is plotted from the left front output (dotted line), and the central output (solid line) is plotted as a strong signal from the center to the left.

Claims (3)

A surround sound decoder for decoding a plurality of sound input signals including a right input signal and a left input signal into a plurality of sound output channel signals, wherein the plurality of sound output channel signals includes a left front output signal and a right sound signal. The decoder determines the directional content of the left input signal and the right input signal, and determines at least a left-right directional signal and a center-surround directional signal from the directional content. Including means for generating,
The decoder
A left input terminal and a right input terminal for receiving the left input signal and the right input signal;
Left delay means and right delay means for generating a delayed left signal and a delayed right signal from the left input signal and the right input signal;
A plurality of multiplier means arranged in pairs, wherein the first multiplier means of each pair of the plurality of multiplier means receives the delayed left signal and the plurality of multiplier means Each pair of second multiplier means receives the delayed right signal, and each of the plurality of multiplier means converts the received input signal with one of a plurality of variable matrix coefficients. Multiplying to provide a multiplier output signal, wherein the plurality of variable matrix coefficients are controlled by at least one of the left-right directional signal and the center-surround directional signal. Instrument means,
A plurality of adding means, each of the plurality of adding means receiving the multiplier output signal from one multiplier means of the pair of multiplier means and out of the plurality of output channel signals. A plurality of adding means for generating one of
Each matrix coefficient is a function of the directional signal that is plotted when the function is plotted using the left-right directional signal and the center-surround directional signal that define the principal axis. Defining a surface in a plurality of quadrants as a graphical representation of a plurality of matrix coefficient values, the entire surface being substantially continuous across the plurality of quadrants ;
At least two different operating modes are provided, and the matrix coefficients are controlled differently by the directional signal in the different operating modes,
The film operating mode is suitable for playing surround-encoded audio signals derived from film soundtracks and other video sources, and the music operating mode is suitable for playing music recordings or music broadcasts And
For decoding in the film mode, the matrix values for the left front output and right front output are configured to remove or attenuate the center component of the input signal, and the matrix value for the center output is the center value As the output attenuation provides a center output of 4.5 dB less than the previous standard Dolby Pro-Logic decoder, the center / surround directional signal becomes more positive Configured to reduce, the center / surround signal varies between -45 ° and 45 °, with a value of -45 ° corresponding to the signal from the back, and a value of 45 ° from the center Corresponding to the signal, an intermediate matrix value keeps the power ratio of the central component to the uncorrelated component of the input signal the same at the output of the decoder It is determined by the determined, surround sound decoder.   The decoder according to claim 1, wherein the adding means includes a plurality of adders.   The decoder of claim 1, wherein a total power in the plurality of acoustic output channel signals is substantially equal to a total power in the plurality of acoustic input signals.

Images