.TH SIGNAL 5 "1 March 2010" "WFDB software 10.5.0" "WFDB Applications Guide"
.SH NAME
signal \- WFDB signal file formats
.SH DESCRIPTION
WFDB signal files exist in several formats. Any of these formats
can be used for multiplexed signal files, in which samples from two or more
signals are stored alternately. See \fBheader\fR(5) for information on how to
identify which of the formats below is used for a particular signal file.
.SS Format 8
.PP
Each sample is represented as an 8-bit first difference; i.e.,
to get the value of sample \fIn\fR, sum the first \fIn\fR bytes of the
sample data file together with the initial value from the header file.
When format 8 files are created, first differences which cannot be
represented in 8 bits are represented instead by the largest difference
of the appropriate sign (-128 or +127), and subsequent differences are
adjusted such that the correct amplitude is obtained as quickly as
possible. Thus there may be loss of information if signals in another
of the formats listed below are converted to format 8.
Note that the first differences stored in multiplexed format 8
files are always determined by subtraction of successive samples from
the same signal (otherwise signals with baselines which differ by 128
units or more could not be represented this way).
.SS Format 16
.PP
Each sample is represented by a 16-bit two's complement amplitude stored
least significant byte first.
Any unused high-order bits are sign-extended from the most significant bit.
Historically, the format used for MIT\-BIH and AHA database distribution
9-track tapes was format 16, with the addition of a logical EOF (octal 0100000)
and null-padding after the logical EOF.
.SS Format 24
.PP
Each sample is represented by a 24-bit two's complement amplitude stored
least significant byte first.
.SS Format 32
.PP
Each sample is represented by a 32-bit two's complement amplitude stored
least significant byte first.
.SS Format 61
.PP
Each sample is represented by a 16-bit two's complement amplitude stored
most significant byte first.
.SS Format 80
.PP
Each sample is represented by an 8-bit amplitude in offset binary form
(i.e., 128 must be subtracted from each unsigned byte to obtain a signed
8-bit amplitude).
.SS Format 160
.PP
Each sample is represented by a 16-bit amplitude in offset binary form
(i.e., 32,768 must be subtracted from each unsigned byte pair to
obtain a signed 16-bit amplitude). As for format 16, the least significant
byte of each pair is first.
.SS Format 212
.PP
Each sample is represented by a 12-bit two's complement amplitude. The first
sample is obtained from the 12 least significant bits of the first byte pair
(stored least significant byte first). The second sample is formed from the
4 remaining bits of the first byte pair (which are the 4 high bits of the
12-bit sample) and the next byte (which contains the remaining 8 bits of the
second sample). The process is repeated for each successive pair of samples.
Most of the signal files in PhysioBank are written in format 212.
.SS Format 310
.PP
Each sample is represented by a 10-bit two's-complement amplitude. The first
sample is obtained from the 11 least significant bits of the first byte pair
(stored least significant byte first), with the low bit discarded. The second
sample comes from the 11 least significant bits of the second byte pair, in the
same way as the first. The third sample is formed from the 5 most significant
bits of each of the first two byte pairs (those from the first byte pair are
the least significant bits of the third sample). Note that the unused bit in
each byte pair is set to zero when using the WFDB library to write a format 310
signal file. The entire process is then repeated for each successive set of
three samples.
.SS Format 311
.PP
Each sample is represented by a 10-bit two's-complement amplitude. Three
samples are bit-packed into a 32-bit integer as for format 310, but the layout
is different. Each set of four bytes is stored in little-endian order (least
significant byte first, most significant byte last). The first sample is
obtained from the 10 least significant bits of the 32-bit integer, the second
is obtained from the next 10 bits, the third from the next 10 bits, and the two
most significant bits are unused (note that these bits are set to zero when
using the WFDB library to write a format 311 signal file). This process is
repeated for each successive set of three samples.
.SS Formats 508, 516, and 524
.PP
Signals are compressed using the FLAC (Free Lossless Audio Codec) format,
either using 8 bits per sample (in format 508), 16 bits per sample (in format
516), or 24 bits per sample (in format 524). See
https://xiph.org/flac/format.html for the full specification of these formats.
Note that these formats impose some restrictions on the signal data that can be
stored:
.IP \(bu 4
The number of signals must equal the number of channels in the FLAC stream,
which implies there can be at most eight signals stored in a single file.
.IP \(bu 4
Every signal in the file must have the same sampling frequency, and therefore
the same number of samples per frame.
.PP
Conversely, some limitations apply when generating a FLAC stream to be used as
a WFDB signal file:
.IP \(bu 4
The FLAC `bits per sample' field must be 8, 16, or 24.
.IP \(bu 4
Every sample in the stream must be within the range of values indicated by the
`bits per sample' field.
.PP
Note that the number of samples per frame is unrelated to the size of the
blocks used in FLAC compression; a single FLAC block may contain multiple WFDB
frames, and a single WFDB frame may be split across multiple FLAC blocks.
.PP
Also note that the FLAC `sample rate' field should be set to 96000, regardless
of the actual sampling frequency of the signals. For compatibility with older
FLAC implementations, this field should not be set to 88200, 176400, or 192000.
.SH SEE ALSO
\fBannot\fR(5), \fBheader\fR(5), \fBwfdbcal\fR(5)
.br
\fIWFDB Programmer's Guide\fP
.SH AUTHOR
George B. Moody (george@mit.edu)