This contains exactly 36 quantized subband samples for a channel and a subband.

After requantization and channel dematrixing this gives the input for the subband synthesizer.

Member of Audio Frame

Coding of the subband samples depends and only depends on the Allocation.

Samples are huffman coded by fixed huffman tables.

The following allocations are currently defined (allocations below 0 are not possible for subband 0):

alloc	sample value are . . .	typical entropy per sample	No. of Codes
-32	Reserved		-
. . .	. . .		-
-18	Reserved		-
-17	Intensity stereo using 8 vector bit, all further bands also uses this encoding	0.00 bit	-
-16	Intensity stereo using 7 vector bit, all further bands also uses this encoding	0.00 bit	-
-15	Intensity stereo using 6 vector bit, all further bands also uses this encoding	0.00 bit	-
-14	Intensity stereo using 5 vector bit, all further bands also uses this encoding	0.00 bit	-
-13	Intensity stereo using 4 vector bit, all further bands also uses this encoding	0.00 bit	-
-12	Intensity stereo using 3 vector bit, all further bands also uses this encoding	0.00 bit	-
-11	Intensity stereo using 2 vector bit, all further bands also uses this encoding	0.00 bit	-
-10	Intensity stereo using 1 vector bit, all further bands also uses this encoding	0.00 bit	-
-9	Intensity stereo using 8 vector bit	0.00 bit	-
-8	Intensity stereo using 7 vector bit	0.00 bit	-
-7	Intensity stereo using 6 vector bit	0.00 bit	-
-6	Intensity stereo using 5 vector bit	0.00 bit	-
-5	Intensity stereo using 4 vector bit	0.00 bit	-
-4	Intensity stereo using 3 vector bit	0.00 bit	-
-3	Intensity stereo using 2 vector bit	0.00 bit	-
-2	Intensity stereo using 1 vector bit	0.00 bit	-
-1	-0.5 or +0.5 generated by a random number generator	0.00 bit	-
0	0. all further bands also uses this encoding	0.00 bit	-
1	0.	0.00 bit	-
2	[-1,+1], 6 samples are grouped At most 2 samples per group can be nonzero.	0.58 bit	73 (table_03)
3	[-1,+1], 4 samples are grouped	1.11 bit	81 (table_04)
4	[-1,+1], 3 samples are grouped	1.37 bit	27 (table_05)
5	[-2,+2], 3 samples are grouped	1.56 bit	125 (table_06)
6	[-2,+2], 2 samples are grouped	1.73 bit	25 (table_07)
7	[-3,+3], 2 samples are grouped	2.25 bit	49 (table_08)
8	[-4,+4], 2 samples are grouped	2.72 bit	81 (table_09)
9	[-5,+5], 2 samples are grouped	3.22 bit	121 (table_10)
10	coded using 0 linear bits and heavy shaped table encoded range [-16,+15]	3.74 bit	32 (table_01)
11	coded using 0 linear bits and light shaped table encoded range [-16,+15]	4.26 bit	32 (table_02)
12	coded using 1 linear bits and heavy shaped table encoded range [-32,+31]	4.74 bit	32 (table_01)
13	coded using 1 linear bits and light shaped table encoded range [-32,+31]	5.26 bit	32 (table_02)
14	coded using 2 linear bits and heavy shaped table encoded range [-64,+63]	5.74 bit	32 (table_01)
15	coded using 2 linear bits and light shaped table encoded range [-64,+63]	6.26 bit	32 (table_02)
16	coded using 3 linear bits and heavy shaped table encoded range [-128,+127]	6.74 bit	32 (table_01)
17	coded using 3 linear bits and light shaped table encoded range [-128,+127]	7.26 bit	32 (table_02)
18	coded using 4 linear bits and heavy shaped table encoded range [-256,+255]	7.74 bit	32 (table_01)
19	coded using 4 linear bits and light shaped table encoded range [-256,+255]	8.26 bit	32 (table_02)
20	coded using 5 linear bits and heavy shaped table encoded range [-512,+511]	8.74 bit	32 (table_01)
21	coded using 5 linear bits and light shaped table encoded range [-512,+511]	9.26 bit	32 (table_02)
22	coded using 6 linear bits and heavy shaped table encoded range [-1024,+1023]	9.74 bit	32 (table_01)
23	coded using 6 linear bits and light shaped table encoded range [-1024,+1023]	10.26 bit	32 (table_02)
24	coded using 7 linear bits and heavy shaped table encoded range [-2048,+2047]	10.74 bit	32 (table_01)
25	coded using 8 linear bits and heavy shaped table encoded range [-4096,+4095]	11.74 bit	32 (table_01)
26	coded using 9 linear bits and heavy shaped table encoded range [-8192,+8191]	12.74 bit	32 (table_01)
27	coded using 10 linear bits and heavy shaped table encoded range [-16384,+16383]	13.74 bit	32 (table_01)
28	coded using 11 linear bits and heavy shaped table encoded range [-32768,+32767]	14.74 bit	32 (table_01)
29	Reserved		-
30	Reserved		-
31	Use a subframe based allocation theme		-

All subband samples are 0.

    int  i;

    for ( i = 0; i < 36; i++ ) {
        sample [i] = 0;
    }

Subband samples has the range [-1,+1], 6 samples are grouped.

At most 2 samples per group can be nonzero.

    static unsigned char  tab [] = { 0x01,0x02,0x03,0x04,0x05,
                                          0x12,0x13,0x14,0x15,
                                               0x23,0x24,0x25,
                                                    0x34,0x35,
                                                         0x45
    };
    int                   code;
    int                   tmp;
    int                   i;

    for ( i = 0; i < 36; i += 6 ) {
        code = Decode_Huffman_Table (table_03) - 1;

        sample [i + 0] = sample [i + 1] = sample [i + 2] =
        sample [i + 3] = sample [i + 4] = sample [i + 5] = 0;

        if ( code >= 0 )
            if ( code < 12 ) {
                sample [i + (code >> 1)] = code & 1 ? +1 : -1;
            }
            else {
                tmp                    = tab [(code - 12) >> 2];
                sample [i + (tmp >> 4)] = code & 1 ? +1 : -1;
                sample [i + (tmp & 15)] = code & 2 ? +1 : -1;
            }
    }

Subband samples has the range [-1,+1], 4 samples are grouped.

    int  code;
    int  i;

    for ( i = 0; i < 36; i += 4 ) {
        code = Decode_Huffman_Table (table_04);

        sample [i + 0] = code % 3 - 1;
        sample [i + 1] = code / 3 % 3 - 1;
        sample [i + 2] = code / 9 % 3 - 1;
        sample [i + 3] = code / 27 - 1;
    }

Subband samples has the range [-1,+1], 3 samples are grouped.

    int  code;
    int  i;

    for ( i = 0; i < 36; i += 3 ) {
        code = Decode_Huffman_Table (table_05);

        sample [i + 0] = code % 3 - 1;
        sample [i + 1] = code / 3 % 3 - 1;
        sample [i + 2] = code / 9 - 1;
    }

Subband samples has the range [-2,+2], 3 samples are grouped.

    int  code;
    int  i;

    for ( i = 0; i < 36; i += 3 ) {
        code = Decode_Huffman_Table (table_06);

        sample [i + 0] = code % 5 - 2;
        sample [i + 1] = code / 5 % 5 - 2;
        sample [i + 2] = code / 25 - 2;
    }

Subband samples has the range [-2,+2], 2 samples are grouped.

    int  code;
    int  i;

    for ( i = 0; i < 36; i += 2 ) {
        code = Decode_Huffman_Table (table_07);

        sample [i + 0] = code % 5 - 2;
        sample [i + 1] = code / 5 - 2;
    }

Subband samples has the range [-3,+3], 2 samples are grouped.

    int  code;
    int  i;

    for ( i = 0; i < 36; i += 2 ) {
        code = Decode_Huffman_Table (table_08);

        sample [i + 0] = code % 7 - 3;
        sample [i + 1] = code / 7 - 3;
    }

Subband samples has the range [-4,+4], 2 samples are grouped.

    int  code;
    int  i;

    for ( i = 0; i < 36; i += 2 ) {
        code = Decode_Huffman_Table (table_09);

        sample [i + 0] = code % 9 - 4;
        sample [i + 1] = code / 9 - 4;
    }

Subband samples has the range [-5,+5], 2 samples are grouped.

    int  code;
    int  i;

    for ( i = 0; i < 36; i += 2 ) {
        code = Decode_Huffman_Table (table_10);

        sample [i + 0] = code % 11 - 5;
        sample [i + 1] = code / 11 - 5;
    }

Subband samples are heavily shaped.

Allocation	linear bits
10	none
12	1
14	2
16	3
18	4
20	5
22	6
24	7
25	8
26	9
27	10
28	11

    int  linear = ...;
    int  code;
    int  i;

    for ( i = 0; i < 36; i++ ) {
        code = Decode_Huffman_Table (table_01);

        sample [i] = (code << linear) + Read_Bits (linear);
    }

Subband samples are lightly shaped.

Allocation	linear bits
11	none
13	1
15	2
17	3
19	4
21	5

    int  linear = ...;
    int  code;
    int  i;

    for ( i = 0; i < 36; i++ ) {
        code = Decode_Huffman_Table (table_02);

        sample [i] = (code << linear) + Read_Bits (linear);
    }

All subband samples are generated by a random generator.

    int  i;

    for ( i = 0; i < 36; i++ ) {
        sample [i] = gaussian_distribution (0.0, 0.5);  // 1st momentum 0, 2nd momentum 0.5
    }