From: Jeffrey L. Hayes (tvdog@delphi.com)
Tandy .snd files are created by Sound.pdm, a program that came with the proprietary DeskMate environment. They are used by Music.pdm to create music modules (.sng files). DeskMate Sound and Music require the Tandy sound chip. There is a program to convert RIFF WAVE and other 8-bit PCM formats to .snd, Conv2snd, by Kenneth Udut. Conv2snd v.2.00 comes with Snd2wav, which converts .snd to RIFF WAVE.
There are two types of DeskMate .snd files, sound files and instrument files. Both contain 8-bit unsigned PCM samples.
Sound files are simpler. These are garden-variety sample files with a fixed-length header giving the name of the sound, the recording frequency, and the length of the sound. Sound files may be recorded at 5500Hz, 11kHz or 22kHz.
Instrument files contain samples as well as frequency and looping information used by Music.pdm to represent an instrument. Instrument files provide for attack, sustain, and decay with several samples having different implied frequencies and being used by Music.pdm to represent the instrument in different pitch ranges. Up to 16 different notes (with 16 different samples) can be contained in one instrument file. Instrument files are always recorded at 11kHz. Both sound files and instrument files may be compressed in one of two ways, "music" compression or "speech" compression, or they may be uncompressed. I don't know the compression algorithms, but simple file comparison reveals that "music" and "speech" compression are almost identical.
The DeskMate .snd file header consists of 16 bytes of fixed header information followed by one or more 28-byte note records. The sample information, which may be compressed, follows the header.
| offset | size | what |
| 0 | byte | 1Ah (.snd ID byte) |
| 1 | byte | Compression code: 0 = no compression; 1 = music compression; 2 = sound compression. |
| 2 | byte | Number of notes in the instrument file. 1 if sound file. |
| 3 | byte | Instrument number. 0 if sound file; 0FFh if instrument file with no number set. Valid instrument numbers in an instrument file are 1 to 32. Use this field to distinguish a sound file from an instrument file. |
| 4 | 10 bytes | Sound or instrument name. Filled on the right with nulls if less than 10 characters. |
| 0Eh | word | Sampling rate in samples per second. Note that although a sampling rate other than 5500, 11000 and 22000 can be entered here, Sound.pdm will not actually play at other rates. |
| 10h | variable | Note records begin, 28 bytes each. Number of records given in byte 2 above. |
| 0 | byte | Pitch of the note: 1 = A1 in American Standard Pitch; 2 = A#1; etc. A1 is lowest note allowed; highest note allowed is B6 (3Fh). Sound files have 0FFh here; so do instrument files with no note set. Note that Sound.pdm does not designate notes in the standard manner to the user. Although A1 and B6 in Sound.pdm are the same as A1 and B6 in standard pitch, Sound.pdm starts octaves at A rather than at C (as is standard). Thus, middle C, C4 in standard pitch, is C3 in Sound.pdm. |
| 1 | byte | Sound files, and instrument files with no pitch set, have 0 here. If the pitch is set, this byte is 0FFh. |
| 2 | 2 bytes | Range of the note, first byte is lower limit, second is higher limit. Byte encoding as for offset 0 (i.e., 01h to 3Fh). Sound files have FF FF here; so do instrument files with no range set. |
| 4 | dword | Offset in the file where samples for this note begin (zero-relative), after compression if that was done. |
| 8 | dword | If compressed, the length of the compressed data in the file for this note. Uncompressed files have 0 here. |
| 0Ch | 4 bytes | Unknown. Set to zero. |
| 10h | dword | Number of samples in the note, after decompression if necessary. |
| 14h | dword | Number of sample at start of sustain region for the note, relative to the first (zeroth) sample of the note. For sound files, or if sustain is not set, this field is 0. |
| 18h | dword | Number of sample at end of sustain region for the note, relative to the first (zeroth) sample of the note. For sound files, or if sustain is not set, this field is 0. |
This is the new .snd file format used on the 2500-series. From information provided by John Ball (john.ball@two-t.com).
Like the old format, the new format header consists of a fixed part followed by one or more sample descriptors. The fixed part is 114 bytes; the sample descriptors are 46 bytes each. Samples are still 8-bit unsigned PCM.
Fixed header:
| offset | size | what |
| 0 | 10 bytes | ASCIIZ name of sound. |
| 0Ah | 34 bytes | unknown |
| 2Ch | 2 bytes | New .snd ID: 1Ah 80h. |
| 2Eh | word | Number of samples in file. |
| 30h | word | Sound (instrument) number. |
| 32h | 16 bytes | unknown |
| 42h | word | Compression code (0 = no compression, 1 = music compression, 2 = speech compression). |
| 44h | 20 bytes | unknown |
| 58h | word | Sampling rate in Hz. |
| 5Ah | 24 bytes | unknown |
| 72h | variable | Sample descriptors begin. |
Sample descriptors (number given by word at 2Eh above):
| offset | size | what |
| 0 | dword | Link to next sample descriptor (offset in file of next sample descriptor record). 0 if last. |
| 4 | 2 bytes | unknown |
| 6 | byte | Pitch of note (01h-3Fh), 01 = A1 in American Standard Pitch; 0FFh if not set. |
| 7 | byte | unknown (compare old .Snd format; value is 00 or FF, but seemingly unrelated to pitch setting) |
| 8 | 2 bytes | Range of note. First byte is lower limit,
second is higher limit. Values as for byte at offset 6 above; FF FFh if not set. |
| 0Ah | dword | Offset in file of start of sound data for this sample. |
| 0Eh | dword | Length of sample sound data in bytes. |
| 12h | dword | Uncompressed length of sound data (number of samples). |
| 16h | 24 bytes | unknown |
[ Index | Previous paragraph ]