Group subset

The FLAC format specifies a subset of itself to ensure streamability and limits the decoding requirements for hardware implementations. The reference FLAC encoder will enforce this subset unless specifically disabled.

The files in this group are considered a baseline for general decoders: these files should be properly decoded or properly rejected before playback is attempted by any decoder. A decoder can choose to reject certain files, for example multichannel files, files with high or unusual samplerates, files with a high bit depth. Crashing or mangled playback of these files is probably going to be noticed by unsuspecting users of a decoder. Read the README.txt in the directory subset for details on each file.

Files #1 - #27

The first 10 files tests 44.1kHz, 16-bit audio with various blocksizes that are within subset.

Files 11 through 23 tests the ability to decode FLAC files with various features that are within subset but aren't used often.

File 11 uses the maximal allowed rice partition order (8)
File 12 uses the maximal allowed qlp precision (15)
File 13 uses the smallest sane qlp precision (2)
File 14 uses wasted bits
File 15 uses only 'verbatim' frames
File 16 uses rice escape codes and partition order 8
File 17 uses all possible fixed orders (especially order 0 which isn't used often)
File 18 is encoded with precision search, using qlp precisions between 3 and 15
File 19 uses a samplerate of 35467Hz
File 20 uses a samplerate of 39kHz
File 21 uses a samplerate of 22050Hz
File 22 has 12 bits per sample
File 23 has 8 bits per sample

Files 24 through 27 test the ability to decode a FLAC file with a variable blocksize. This is a subset feature which is currently (August 2021) only implemented in the Flake decoder and its forks/decendants and is not enabled by default.

With the release of FLAC 1.2.0 in July 2007, the FLAC specification was augmented to more clearly signal variable blocksize streams by the use of a special bit in the header. File 24, 25 and 26 use this format. File 27 follows the old specification, which is much harder to detect

File 24 uses the current format and is created by flake r264
File 25 uses the current format and is created by a modified flake r264 creating smaller blocks
File 26 uses the current format and is created by CUETools.Flake 2.1.6
File 27 uses the old format and is created by flake 0.11

Files #28 - #37

Files 28 through 37 test the ability to decode various high-resolution FLAC file (96kHz, 24-bit)

File 28 uses default settings
File 29 uses the largest allowed blocksize (16384)
File 30 uses non-standard blocksize 13456
File 31 uses only 32th order predictors
File 32 uses escape codes and partition order 8
File 33 is upsampled to 192kHz
File 34 is upsampled to 192kHz, uses blocksize 16384, 32th order predictors only, maximum LPC precision and maximum partition order
File 35 uses non-standard samplerate 134560Hz
File 36 is upsampled to 384kHz
File 37 has 20 bits per sample

Files #38 - #44

Files 38 through 43 test the ability to decode various multichannel FLAC files. Each file contains a voice description of the channels present, so as to see whether the channels are decoded in the correct lay-out.

File 38 is 3.0-channel (left, right, center)
File 39 is 4.0-channel or quadraphonic
File 40 is 5.0-channel
File 41 is 5.1-channel
File 42 is 6.1-channel
File 43 is 7.1-channel

File 44 tests the ability to decode a file with the highest possible data input per second, staying within subset and using a standard samplerate. It also only uses 32th order predictors at the highest possible predictor precision and the largest blocksize allowed within the FLAC subset making it especially challenging to decode.

Files #45 - #59

Files 45 through 59 test the ability to handle various streams with valid but rather unusual or extreme metadata.

File 45 has 'unknown number of samples' in STREAMINFO
File 46 has maximum and minimum framesize set to 'unknown'
File 47 has only a STREAMINFO block
File 48 has an extremely large SEEKTABLE
File 49 has an extremely large PADDING block
File 50 has an extremely large PICTURE block (JPG of 15.8MB)
File 51 has an extremely large VORBISCOMMENT block
File 52 has an extremely large APPLICATION block
File 53 has a CUESHEET block with absurdly many indexes
File 54 with the same 20 VORBISCOMMENTs repeated 1000 times
File 55 has the metadata of track 47-52 combined
File 56 has a PICTURE with mimetype image/jpeg
File 57 has a PICTURE with mimetype image/png
File 58 has a PICTURE with mimetype image/gif
File 59 has a PICTURE with mimetype image/avif

Files #60 - #64

Miscellaneous, later additions

File 60 is mono audio
File 61, 62 and 63 are signals with rather extreme characteristics that might trigger overflow if a decoder uses 32-bit integers to calculate the predictor where 64-bit integers are appropriate
File 64 contains rice codes with escape code zero

5.0 KiB Raw Blame History