Follow

Any of my followers a pro at audio formats used in late 90s grooveboxes? Dumped the sound ROM from my Korg EM-1, synth waveforms were simple 8 and 12 bit audio but the drums seem… compressed? Repeating 256 bit blocks with an internal 7 bit pattern. Screenshot shows raw bits in rows of 256 of what it probably a single sample. Bits get more sparse as it gets quiet.

tech bla 

Each audio frame consists of a 40 bit header and 31 7 bit samples. The header contains multiple scaling and offset parameters and the 7 bit samples are differential PCM.
Top channel is my decoding attempt, next are the 7 bit DPCM samples, a 16 bit offset, a 6 bit and a 10 bit scaling value for the DPCM data which look almost the same and what I assume to be some kind of exponent.
I'm getting close to the original sound, but it's still kinda messed up…

re: tech bla 

@Slyka

Makes me think of DANCE audio from the Japanese MUSE analog bandwidth-compressed HDTV format of the 1990s. One of these days I'm going to get onto my project of building an encoder… which will presumably involve implementing a decoder as well.

re: tech bla 

@tsukkitsune ohh interesting, do you have links to any technical docs on that format?

re: tech bla 

@Slyka

Most of what I have is either (a) on paper, (b) in Japanese, or (c) both. I wrote to NHK Labs back around '05, & they sent me a big envelope full of A3 format, what I think must have been, microfilm offprints. I have, however, made some attempts to summarize the information.

MUSE stands for "Multiple Sub-nyquist Sampling Encoding", and DANCE for "Dpcm Audio Near-instantaneously Compressed and Expanded". (They seem to have gotten hold of NASA's backronym generator.)

DANCE has two modes. A-Mode provides four channels of audio, 15 bit sampling at 32 kilohertz, compressed to 8 bits × 8 ranges. B-mode is 2-channel, 16 bit sampling × 48 kHz, compressed to 12 bits × 5 ranges. DPCM frame length (the period for which ranges are calculated) is 1 ms for both modes.

After shuffling & the addition of error-correction, the resultant bitstream is ternary-encoded into the 48 lines per frame of vertical interval. (MUSE has 1080 total lines, 1032 active lines, as against 1035 active in the original Japanese "Hi-Vision" production standard.)

Anyway, not all the information I have is totally in agreement. That's one reason I think I may need to implement a decoder ; although, since I have a number of consumer-market decoders, I could just try it & see what works!

Sign in to participate in the conversation
The Vulpine Club

The Vulpine Club is a friendly and welcoming community of foxes and their associates, friends, and fans! =^^=