Three channel panning system

[0002] This application claims priority from provisional application Serial No. 60/102,192, filed Sep. 28, 1998, entitled “Three Channel Panning System,” the disclosure of which is incorporated herein by reference.

[0003] This invention relates to signal processing. Specifically, the present invention relates to encoding and decoding audio signals for driving multiple speakers.

[0004] Many systems have been developed to reproduce three-dimensional sound effects using multiple loudspeakers. One of the most sophisticated systems is the Ambisonic method of recording information about a sound field for reproduction over some form of loudspeaker array so as to produce the impression of hearing a true three-dimensional sound image.

[0005] The Ambisonic system encodes a three-dimensional soundfield in a B-format which is defined as follows:

[0006] X=cos(Az) (front-back velocity signal)

[0007] Y=sin(Az) (left-right velocity signal)

[0008] W=0.707 (pressure signal)

[0009] where the vertical dimension has been ignored and Az is the azimuth (horizontal panning angle) of the sound source.

[0010] Note that the Y component can be positive or negative which requires a signal processing unit which can generate negative values. However, some systems are capable of generating only positive loudspeaker weights and are unable to implement B format encoding.

[0011] According to one aspect of the present invention, a method and system for panning sounds over four speakers employs a three-channel representation of the same. With the present invention, panning sounds over four speakers (denoted L,R,SL,SR) may be achieved while employing a three-channel representation of channels (L,R,W). However, (L,R,W) is more convenient than (W,X,Y) because it already contains the signals for the two front speakers.

[0012] When the three-channels are converted to four speaker feeds, it is ensured that positive loudspeaker weights are produced instead of out of phase feeds in the loudspeakers opposite the direction from where the sound is panned, as is done using the Ambisonic prior art technique.

[0013] According to another aspect of the invention, this is achieved by using a sine-cosine panning technique to pan over a pair of left and right channels, while including the unprocessed input signal (W) in this three-channel form at. A subsequent operation then allows deriving the two rear channel signals for reproduction over four speakers. In this fashion, the three-channel format (L,R,W) could be decoded over any horizontal multi-channel loudspeaker setup.

[0014] According to one aspect of the invention, panning over four loudspeakers is implemented utilizing an internal representation having three channels.

[0015] According to another aspect of the invention, a signal including an audio portion and angular information is encoded in Left, Right, and W channels where all channel weights have positive values.

[0016] According to another aspect of the invention, sine/cosine panning, using the raised cosine function, is utilized to calculate Left and Right speaker weights and the W channel is equal to the input audio signal.

[0017] According to another aspect of the invention, the Left and Right channels are used directly to drive front loudspeakers while the other loudspeakers are driven by combinations of the Left, Right, and W channel.

[0018] The use of only positive weights in the three-channel format is convenient for hardware designers because it is easier to design hardware to implement positive weights. Also, subsequent to decoding, the positive weights allow the panning of signals from front left to right (or vice versa) with no signals sent to the rear speakers, which results in more realistic effects.

[0019] Other features and advantages of the invention will be apparent in view of the following detailed description and appended drawings.

[0020] FIG. 1 is a schematic diagram of the coordinate system utilized to calculate speaker weights for the left and right channels;

[0021] FIG. 2 is a block diagram of an encoder;

[0022] FIG. 3 is a block diagram of a decoder;

[0023] FIGS. 4A-4C are graphs depicting channel weight signals calculated in herical coordinates;

[0024] FIG. 5 is a graph depicting a channel weight signal calculated in Cartesian coordinates;

[0025] FIG. 6 is a schematic diagram of a non-square speaker configuration; and

[0026] FIG. 7 is a graph of a mapping for use with the speaker configuration of FIG. 6.

[0027] The preferred embodiment of the technique for encoding an audio signal into the Left, Right, W (L,R,W) format will now be described with reference to FIG. 1.

[0028] Notations:

[0029] P=input signal.

[0030] Az=azimuth (horizontal panning angle) measured in degrees anti-clockwise in view from top, with 0 degrees in front where this information is included for each signal. The orientation of the azimuthal angle relative to a listener and (x,y) coordinate system is depicted in FIG. 1. Note that the left and right speaker positions are defined for +45° and 45° respectively.

[0031] El=elevation (vertical panning angle) measured in degrees (from −90° to +90°), with 0° in the horizontal plane and +90° directly on top.

[0032] Encoding: 3-channel formulas producing (L,R,W) format in spherical coordinates uses the raised cosine function to produce only positive loudspeaker signals:

[0033] W=P;

[0034] L=P*(1+cos(Az−45°))/2;  (Eq. 1)

[0035] R=P*(1+cos(Az+45°))/2.  (Eq. 2)

[0036] Encoding: an alternative 3-channel panning formulas producing (L,R,W) signals using Cartesian coordinates:

[0037] W=P;

[0038] L=P*(min{max[(x−y), −1], 1}+1)/2  (Eq. 3)

[0039] R=P*(min{max[(x+y), −1], 1}+1)/2  (Eq. 4)

[0040] where the max[ ] function clamps (x−y) or (x+y) to be greater than −1 and min( ) clamps the output of max[ ] to be less than 1, and where X and Y are defined by X=cos(Az)*cos(EL) and Y=sin(Az)*cos(El) where X points towards the front and Y towards the left. In the above, the coefficients of P are the (L,R,W) weights.

[0041] Generally, the (L,R,W) panning function has to have L and R components having a magnitude between 0 and 1 which are also symmetric about 0.5. Also, in each case L and R are related by:

[0042] R(Az)=L(Az+90°)

[0043] FIG. 2 is a block diagram of part of an encoder 30 for implementing horizontal encoding. As described above, each wave-table synthesizer 32 outputs a signal including an audio portion and azimuthal angle information. A controller 33, under control of program code stored in store 33a, calculates the L and R weights according to equations (1) and (2) based on the azimuthal information output by the wave-table synthesizers 32. The output of each wave-table synthesizer is split into three channels and the amplitude of audio signal is modified by adjustable gain amplifiers 34 under control of the controller 33 according to the above formulae to generate the L, R, and W signals. Note that all the weights applied by the amplifiers are positive.

[0044] Decoding to 4 speakers (L,R,W) to (L,R,SL,SR)

[0045] L=L;

[0046] R=R;

[0047] SL(surround left)=W−R;

[0048] SR(surround right)=W−L.

[0049] FIG. 3 is a block diagram of a system for implementing the above decoding. A decoder matrix 40 combines the (L,R,W) channels to form the (L,R,SL,SR) output signals.

[0050] Thus, as described above, in (L,R,W) encoding the L and R channels are fed directly to the front loudspeakers. In the preferred embodiment W=1 and, as described above, L and R are symmetric about 0.5, SR is the reflection of L about 0.5 and SL is the reflection of R about 0.5.

[0051] The L, R, SL, and SR signals with L and R calculated by Eqs. 1 and 2 are depicted in FIGS. 4A-4C. In this case:

[0052] SR(Az)=1−L(Az)=L(Az+180°); and

[0053] SL(Az)=1−R(Az)=R(Az+180°).

[0054] The L signal calculated by Eq. 3 is depicted in FIG. 5.

[0055] The above (L,R,W) encoding scheme places the speakers in square with L and R at ±45°. However, the system can be utilized for other configurations utilizing simple linear mapping. For example, if the speakers were in a rectangle with L and R at ±30° and SL and SR at ±120° (depicted in FIG. 6) then the azimuthal angle would be designated Az₁. Equations (1) and (2) require mapping Az₁to Az. For this case the mapping is shown in FIG. 7. Note that for the Az₁to Az mapping, +30° is mapped to +45°; −30° is mapped to −45°; +120° is mapped to +135°; and +240°(−120°) is mapped to +225°(−135°). Thus, the L,R,SL, and SR signals are mapped directly to the loudspeakers at the appropriate angles. The mapping is linear between these points.

[0056] The invention has now been described with reference to the preferred embodiments. Alternatives and substitutions will now be apparent to persons of skill in the art. In particular, the (L,R,W) channels can be combined to drive more than four loudspeakers. A linear combination of L, R, and W, e.g., a*L+b*R+c*W would be used as a feed to each loudspeaker with an appropriate triplet (a,b,c) for each speaker. Further, both the encoder and decoder could be implemented digitally or in analog parts. Accordingly, it is not intended to limit the invention except as provided by the appended claims.