DSDIFF: DFF Container
Philips DSDIFF stores DSD audio or its losslessly compressed DST representation with properties and optional production information.
A DSDIFF file, usually named .dff, carries either raw DSD or a losslessly compressed representation called DST. The format reader must identify which path applies before it can supply the one-bit audio sequence.
The file's Form DSD structure contains a format-version description and sound properties. Those properties identify sample rate, channels, and a compression type. Raw DSD data and DST data occupy different sound-data structures.
| Declared sound type | Route to the DSD sequence |
|---|---|
| DSD | Read and interpret the stored one-bit data arrangement |
| DST | Read compressed frames and decode them back into DSD |
DST is Direct Stream Transfer, not a PCM conversion. Its lossless decoding reconstructs the DSD it encoded. Converting that result to PCM is another operation after the storage representation has been interpreted.
A DST frame has time even when its bytes vary
The Philips specification relates a DST frame to 1/75 second. At a DSD64 rate of 2,822,400 one-bit samples per second, that interval contains 37,632 bits, or 4,704 bytes, per channel before compression. Stereo therefore corresponds to 9,408 raw bytes for that interval.
DST can encode that interval in a different number of bytes according to the signal. A reader uses frame structure and, where present, indexing to navigate compressed data; it cannot derive every frame's address from one constant compressed size.
This is the same distinction between represented duration and stored length that appears in many variable-size encodings, applied here to a one-bit signal.
Production descriptions surround the sound
DSDIFF can carry channel descriptions, markers, comments, and edited-master information. These describe the source or intended production use; they do not constitute a second soundtrack.
DSF uses different channel blocking and its specified ID3v2 tagging arrangement. Translating raw DSD between these containers can preserve the signal while changing its byte arrangement and descriptive structures. Where the source uses DST, producing raw DSF also needs lossless DST decoding first.
A conversion that preserves the sound but loses the album descriptions has preserved one layer and changed another. Likewise, a listening copy in PCM has changed the audio representation even when it retains all the titles and markers.
For an Emby source, note DSD versus DST, rate, channel layout, valid duration, and the delivered output path. These properties establish which reading and decoding stages are required. They explain the work far more precisely than the shared description “a DSD file.”