1 <?xml version="1.0" standalone="no"?>
2 <!DOCTYPE section PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN"
3 "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd" [
7 <section id="vorbis-spec-floor1">
13 <title>Floor type 1 setup and decode</title>
16 <title>Overview</title>
19 Vorbis floor type one uses a piecewise straight-line representation to
20 encode a spectral envelope curve. The representation plots this curve
21 mechanically on a linear frequency axis and a logarithmic (dB)
22 amplitude axis. The integer plotting algorithm used is similar to
23 Bresenham's algorithm.</para>
28 <title>Floor 1 format</title>
30 <section><title>model</title>
33 Floor type one represents a spectral curve as a series of
34 line segments. Synthesis constructs a floor curve using iterative
35 prediction in a process roughly equivalent to the following simplified
40 <listitem><simpara> the first line segment (base case) is a logical line spanning
41 from x_0,y_0 to x_1,y_1 where in the base case x_0=0 and x_1=[n], the
42 full range of the spectral floor to be computed.</simpara></listitem>
44 <listitem><simpara>the induction step chooses a point x_new within an existing
45 logical line segment and produces a y_new value at that point computed
46 from the existing line's y value at x_new (as plotted by the line) and
47 a difference value decoded from the bitstream packet.</simpara></listitem>
49 <listitem><simpara>floor computation produces two new line segments, one running from
50 x_0,y_0 to x_new,y_new and from x_new,y_new to x_1,y_1. This step is
51 performed logically even if y_new represents no change to the
52 amplitude value at x_new so that later refinement is additionally
53 bounded at x_new.</simpara></listitem>
55 <listitem><simpara>the induction step repeats, using a list of x values specified in
56 the codec setup header at floor 1 initialization time. Computation
57 is completed at the end of the x value list.</simpara></listitem>
63 Consider the following example, with values chosen for ease of
64 understanding rather than representing typical configuration:</para>
67 For the below example, we assume a floor setup with an [n] of 128.
68 The list of selected X values in increasing order is
69 0,16,32,48,64,80,96,112 and 128. In list order, the values interleave
70 as 0, 128, 64, 32, 96, 16, 48, 80 and 112. The corresponding
71 list-order Y values as decoded from an example packet are 110, 20, -5,
72 -45, 0, -25, -10, 30 and -10. We compute the floor in the following
73 way, beginning with the first line:</para>
77 <imagedata fileref="floor1-1.png" format="PNG"/>
80 <phrase>[graph of example floor]</phrase>
85 We now draw new logical lines to reflect the correction to new_Y, and
86 iterate for X positions 32 and 96:</para>
90 <imagedata fileref="floor1-2.png" format="PNG"/>
93 <phrase>[graph of example floor]</phrase>
98 Although the new Y value at X position 96 is unchanged, it is still
99 used later as an endpoint for further refinement. From here on, the
100 pattern should be clear; we complete the floor computation as follows:</para>
104 <imagedata fileref="floor1-3.png" format="PNG"/>
107 <phrase>[graph of example floor]</phrase>
113 <imagedata fileref="floor1-4.png" format="PNG"/>
116 <phrase>[graph of example floor]</phrase>
122 A more efficient algorithm with carefully defined integer rounding
123 behavior is used for actual decode, as described later. The actual
124 algorithm splits Y value computation and line plotting into two steps
125 with modifications to the above algorithm to eliminate noise
126 accumulation through integer roundoff/truncation. </para>
130 <section><title>header decode</title>
133 A list of floor X values is stored in the packet header in interleaved
134 format (used in list order during packet decode and synthesis). This
135 list is split into partitions, and each partition is assigned to a
136 partition class. X positions 0 and [n] are implicit and do not belong
137 to an explicit partition or partition class.</para>
140 A partition class consists of a representation vector width (the
141 number of Y values which the partition class encodes at once), a
142 'subclass' value representing the number of alternate entropy books
143 the partition class may use in representing Y values, the list of
144 [subclass] books and a master book used to encode which alternate
145 books were chosen for representation in a given packet. The
146 master/subclass mechanism is meant to be used as a flexible
147 representation cascade while still using codebooks only in a scalar
152 1) [floor1_partitions] = read 5 bits as unsigned integer
153 2) [maximum_class] = -1
154 3) iterate [i] over the range 0 ... [floor1_partitions]-1 {
156 4) vector [floor1_partition_class_list] element [i] = read 4 bits as unsigned integer
160 5) [maximum_class] = largest integer scalar value in vector [floor1_partition_class_list]
161 6) iterate [i] over the range 0 ... [maximum_class] {
163 7) vector [floor1_class_dimensions] element [i] = read 3 bits as unsigned integer and add 1
164 8) vector [floor1_class_subclasses] element [i] = read 2 bits as unsigned integer
165 9) if ( vector [floor1_class_subclasses] element [i] is nonzero ) {
167 10) vector [floor1_class_masterbooks] element [i] = read 8 bits as unsigned integer
171 11) iterate [j] over the range 0 ... (2 exponent [floor1_class_subclasses] element [i]) - 1 {
173 12) array [floor1_subclass_books] element [i],[j] =
174 read 8 bits as unsigned integer and subtract one
178 13) [floor1_multiplier] = read 2 bits as unsigned integer and add one
179 14) [rangebits] = read 4 bits as unsigned integer
180 15) vector [floor1_X_list] element [0] = 0
181 16) vector [floor1_X_list] element [1] = 2 exponent [rangebits];
182 17) [floor1_values] = 2
183 18) iterate [i] over the range 0 ... [floor1_partitions]-1 {
185 19) [current_class_number] = vector [floor1_partition_class_list] element [i]
186 20) iterate [j] over the range 0 ... ([floor1_class_dimensions] element [current_class_number])-1 {
187 21) vector [floor1_X_list] element ([floor1_values]) =
188 read [rangebits] bits as unsigned integer
189 22) increment [floor1_values] by one
197 An end-of-packet condition while reading any aspect of a floor 1
198 configuration during setup renders a stream undecodable. In
199 addition, a <varname>[floor1_class_masterbooks]</varname> or
200 <varname>[floor1_subclass_books]</varname> scalar element greater than the
201 highest numbered codebook configured in this stream is an error
202 condition that renders the stream undecodable.</para>
204 <section id="vorbis-spec-floor1-decode">
205 <title>packet decode</title>
208 Packet decode begins by checking the <varname>[nonzero]</varname> flag:</para>
211 1) [nonzero] = read 1 bit as boolean
215 If <varname>[nonzero]</varname> is unset, that indicates this channel contained
216 no audio energy in this frame. Decode immediately returns a status
217 indicating this floor curve (and thus this channel) is unused this
218 frame. (A return status of 'unused' is different from decoding a
219 floor that has all points set to minimum representation amplitude,
220 which happens to be approximately -140dB).
224 Assuming <varname>[nonzero]</varname> is set, decode proceeds as follows:</para>
227 1) [range] = vector { 256, 128, 86, 64 } element ([floor1_multiplier]-1)
228 2) vector [floor1_Y] element [0] = read <link linkend="vorbis-spec-ilog">ilog</link>([range]-1) bits as unsigned integer
229 3) vector [floor1_Y] element [1] = read <link linkend="vorbis-spec-ilog">ilog</link>([range]-1) bits as unsigned integer
231 5) iterate [i] over the range 0 ... [floor1_partitions]-1 {
233 6) [class] = vector [floor1_partition_class] element [i]
234 7) [cdim] = vector [floor1_class_dimensions] element [class]
235 8) [cbits] = vector [floor1_class_subclasses] element [class]
236 9) [csub] = (2 exponent [cbits])-1
238 11) if ( [cbits] is greater than zero ) {
240 12) [cval] = read from packet using codebook number
241 (vector [floor1_class_masterbooks] element [class]) in scalar context
244 13) iterate [j] over the range 0 ... [cdim]-1 {
246 14) [book] = array [floor1_subclass_books] element [class],([cval] bitwise AND [csub])
247 15) [cval] = [cval] right shifted [cbits] bits
248 16) if ( [book] is not less than zero ) {
250 17) vector [floor1_Y] element ([j]+[offset]) = read from packet using codebook
251 [book] in scalar context
253 } else [book] is less than zero {
255 18) vector [floor1_Y] element ([j]+[offset]) = 0
260 19) [offset] = [offset] + [cdim]
268 An end-of-packet condition during curve decode should be considered a
269 nominal occurrence; if end-of-packet is reached during any read
270 operation above, floor decode is to return 'unused' status as if the
271 <varname>[nonzero]</varname> flag had been unset at the beginning of decode.
275 Vector <varname>[floor1_Y]</varname> contains the values from packet decode
276 needed for floor 1 synthesis.</para>
280 <section id="vorbis-spec-floor1-synth">
281 <title>curve computation</title>
284 Curve computation is split into two logical steps; the first step
285 derives final Y amplitude values from the encoded, wrapped difference
286 values taken from the bitstream. The second step plots the curve
287 lines. Also, although zero-difference values are used in the
288 iterative prediction to find final Y values, these points are
289 conditionally skipped during final line computation in step two.
290 Skipping zero-difference values allows a smoother line fit. </para>
293 Although some aspects of the below algorithm look like inconsequential
294 optimizations, implementors are warned to follow the details closely.
295 Deviation from implementing a strictly equivalent algorithm can result
296 in serious decoding errors.</para>
299 <title>step 1: amplitude value synthesis</title>
302 Unwrap the always-positive-or-zero values read from the packet into
303 +/- difference values, then apply to line prediction.</para>
306 1) [range] = vector { 256, 128, 86, 64 } element ([floor1_multiplier]-1)
307 2) vector [floor1_step2_flag] element [0] = set
308 3) vector [floor1_step2_flag] element [1] = set
309 4) vector [floor1_final_Y] element [0] = vector [floor1_Y] element [0]
310 5) vector [floor1_final_Y] element [1] = vector [floor1_Y] element [1]
311 6) iterate [i] over the range 2 ... [floor1_values]-1 {
313 7) [low_neighbor_offset] = <link linkend="vorbis-spec-low_neighbor">low_neighbor</link>([floor1_X_list],[i])
314 8) [high_neighbor_offset] = <link linkend="vorbis-spec-high_neighbor">high_neighbor</link>([floor1_X_list],[i])
316 9) [predicted] = <link linkend="vorbis-spec-render_point">render_point</link>( vector [floor1_X_list] element [low_neighbor_offset],
317 vector [floor1_final_Y] element [low_neighbor_offset],
318 vector [floor1_X_list] element [high_neighbor_offset],
319 vector [floor1_final_Y] element [high_neighbor_offset],
320 vector [floor1_X_list] element [i] )
322 10) [val] = vector [floor1_Y] element [i]
323 11) [highroom] = [range] - [predicted]
324 12) [lowroom] = [predicted]
325 13) if ( [highroom] is less than [lowroom] ) {
327 14) [room] = [highroom] * 2
329 } else [highroom] is not less than [lowroom] {
331 15) [room] = [lowroom] * 2
335 16) if ( [val] is nonzero ) {
337 17) vector [floor1_step2_flag] element [low_neighbor_offset] = set
338 18) vector [floor1_step2_flag] element [high_neighbor_offset] = set
339 19) vector [floor1_step2_flag] element [i] = set
340 20) if ( [val] is greater than or equal to [room] ) {
342 21) if ( [highroom] is greater than [lowroom] ) {
344 22) vector [floor1_final_Y] element [i] = [val] - [lowroom] + [predicted]
346 } else [highroom] is not greater than [lowroom] {
348 23) vector [floor1_final_Y] element [i] = [predicted] - [val] + [highroom] - 1
352 } else [val] is less than [room] {
354 24) if ([val] is odd) {
356 25) vector [floor1_final_Y] element [i] =
357 [predicted] - (([val] + 1) divided by 2 using integer division)
359 } else [val] is even {
361 26) vector [floor1_final_Y] element [i] =
362 [predicted] + ([val] / 2 using integer division)
368 } else [val] is zero {
370 27) vector [floor1_step2_flag] element [i] = unset
371 28) vector [floor1_final_Y] element [i] = [predicted]
384 <title>step 2: curve synthesis</title>
387 Curve synthesis generates a return vector <varname>[floor]</varname> of length
388 <varname>[n]</varname> (where <varname>[n]</varname> is provided by the decode process
389 calling to floor decode). Floor 1 curve synthesis makes use of the
390 <varname>[floor1_X_list]</varname>, <varname>[floor1_final_Y]</varname> and
391 <varname>[floor1_step2_flag]</varname> vectors, as well as [floor1_multiplier]
392 and [floor1_values] values.</para>
395 Decode begins by sorting the scalars from vectors
396 <varname>[floor1_X_list]</varname>, <varname>[floor1_final_Y]</varname> and
397 <varname>[floor1_step2_flag]</varname> together into new vectors
398 <varname>[floor1_X_list]'</varname>, <varname>[floor1_final_Y]'</varname> and
399 <varname>[floor1_step2_flag]'</varname> according to ascending sort order of the
400 values in <varname>[floor1_X_list]</varname>. That is, sort the values of
401 <varname>[floor1_X_list]</varname> and then apply the same permutation to
402 elements of the other two vectors so that the X, Y and step2_flag
403 values still match.</para>
406 Then compute the final curve in one pass:</para>
411 3) [ly] = vector [floor1_final_Y]' element [0] * [floor1_multiplier]
412 4) iterate [i] over the range 1 ... [floor1_values]-1 {
414 5) if ( [floor1_step2_flag]' element [i] is set ) {
416 6) [hy] = [floor1_final_Y]' element [i] * [floor1_multiplier]
417 7) [hx] = [floor1_X_list]' element [i]
418 8) <link linkend="vorbis-spec-render_line">render_line</link>( [lx], [ly], [hx], [hy], [floor] )
424 11) if ( [hx] is less than [n] ) {
426 12) <link linkend="vorbis-spec-render_line">render_line</link>( [hx], [hy], [n], [hy], [floor] )
430 13) if ( [hx] is greater than [n] ) {
432 14) truncate vector [floor] to [n] elements
436 15) for each scalar in vector [floor], perform a lookup substitution using
437 the scalar value from [floor] as an offset into the vector <link linkend="vorbis-spec-floor1_inverse_dB_table">[floor1_inverse_dB_static_table]</link>