+.B sample_format \fIBITS\fB/\fIRATE\fB/\fICHANNELS
+Describes the sample format expected by the \fBspeaker_command\fR (below). The
+components of the format specification are as follows:
+.RS
+.TP 10
+.I BITS
+The number of bits per sample. Optionally, may be suffixed by \fBb\fR or
+\fBl\fR for big-endian and little-endian words. If neither is used the native
+byte order is assumed.
+.TP
+.I RATE
+The number of samples per second.
+.TP
+.I CHANNELS
+The number of channels.
+.PP
+The default is
+.BR 16/44100/2 .
+.PP
+With the
+.B network
+backend the sample format is forced to
+.B 16b/44100/2
+and with the
+.B coreaudio
+backend it is forced to
+.BR 16/44100/2 ,
+in both cases regardless of what is specified in the configuration file.
+.RE
+.TP
+.B signal \fINAME\fR
+Defines the signal to be sent to track player process groups when tracks are
+scratched. The default is \fBSIGKILL\fR.
+.IP
+Signals are specified by their full C name, i.e. \fBSIGINT\fR and not \fBINT\fR
+or \fBInterrupted\fR or whatever.
+.TP
+.B speaker_backend \fINAME\fR
+Selects the backend use by the speaker process. The following options are
+available:
+.RS
+.TP
+.B alsa
+Use the ALSA API. This is only available on Linux systems, on which it is the
+default.
+.TP
+.B coreaudio
+Use Apple Core Audio. This only available on OS X systems, on which it is the
+default.
+.TP
+.B oss
+Use the OSS (/dev/dsp) API. Not available on all platforms.
+.TP
+.B command
+Execute a command. This is the default if
+.B speaker_command
+is specified, or if no native is available.
+.TP
+.B network
+Transmit audio over the network. This is the default if
+\fBbroadcast\fR is specified. You can use
+.BR disorder-playrtp (1)
+to receive and play the resulting stream on Linux and OS X.
+.RE
+.TP
+.B sox_generation \fB0\fR|\fB1
+Determines whether calls to \fBsox\fR(1) should use \fB-b\fR, \fB-x\fR, etc (if
+the generation is 0) or \fB-\fIbits\fR, \fB-L\fR etc (if it is 1). See the
+documentation for your installed copy of \fBsox\fR to determine which you need.
+The default is 0.
+.TP
+.B speaker_command \fICOMMAND
+Causes the speaker subprocess to pipe audio data into shell command
+\fICOMMAND\fR, rather than writing to a local sound card. The sample format is
+determine by
+.B sample_format
+above.
+.IP
+Note that if the sample format is wrong then
+.BR sox (1)
+is invoked to translate it. If
+.B sox
+is not installed then this will not work.
+.TP