I would appreciate any donations. Wishlist or send e-mail type donations to maekawa AT daemon-systems.org.

Thank you.

SPEAKER(4)                   Device Drivers Manual                  SPEAKER(4)

     speaker - console speaker audio device driver

     spkr*     at pcppi?
     spkr*     at audio?

     #include <machine/spkr.h> /dev/speaker

     The speaker device driver allows applications to control the console
     speaker on machines with a PC-like 8253 timer implementation or a
     synthesized speaker from an audio device/soundcard.

     Only one process may have this device open at any given time; open(2) and
     close(2) are used to lock and relinquish it.  An attempt to open(2) when
     another process has the device locked will return -1 with an EBUSY error
     indication.  Writes to the device are interpreted as 'play strings' in a
     simple ASCII melody notation.  An ioctl() for tone generation at
     arbitrary frequencies is also supported.

     For the pcppi device sound-generation does not monopolize the processor;
     in fact, the driver spends most of its time sleeping while the PC
     hardware is emitting tones.  Other processes may emit beeps while the
     driver is running.

     For the audio device speaker, the speaker uses one of the virtual audio
     channels.  Enabling this device will also provide a wsbell(4) keyboard

     Applications may call ioctl() on a speaker file descriptor to control the
     speaker driver directly; definitions for the ioctl() interface are in

     The tone_t structure is as follows:

     typedef struct {
             int     frequency;      /* in hertz */
             int     duration;       /* in 1/100ths of a second */
     } tone_t;
     A frequency of zero is interpreted as a rest.

     At present there are four ioctls:

     SPKRGETVOL    Returns an integer, which is the current bell volume as a
                   percentage (0-100).

     SPKRSETVOL    Accepts an integer, which is the desired volume as a

     SPKRTONE      Accepts a pointer to a single tone structure as third
                   argument and plays it.

     SPKRTUNE      Accepts a pointer to the first of an array of tone
                   structures and plays them in continuous sequence; this
                   array must be terminated by a final member with a zero

     The play-string language is modelled on the PLAY statement conventions of
     IBM BASIC 2.0.  The MB, MF and X primitives of PLAY are not useful in a
     UNIX environment and are omitted.  The `octave-tracking' feature is also

     There are 84 accessible notes numbered 1-83 in 7 octaves, each running
     from C to B, numbered 0-6; the scale is equal-tempered A440 and octave 3
     starts with middle C.  By default, the play function emits half-second
     notes with the last 1/16th second being `rest time'.

     Play strings are interpreted left to right as a series of play command
     groups; letter case is ignored.  Play command groups are as follows:

     CDEFGAB -- letters A through G cause the corresponding note to be played
     in the current octave.  A note letter may optionally be followed by an
     accidental sign, one of # + or -; the first two of these cause it to be
     sharped one half-tone, the last causes it to be flatted one half-tone.
     It may also be followed by a time value number and by sustain dots (see
     below).  Time values are interpreted as for the L command below;.

     O <n> -- if <n> is numeric, this sets the current octave.  <n> may also
     be one of 'L' or 'N' to enable or disable octave-tracking (it is disabled
     by default).  When octave-tracking is on, interpretation of a pair of
     letter notes will change octaves if necessary in order to make the
     smallest possible jump between notes.  Thus "olbc" will be played as
     "olb>c", and "olcb" as "olc<b".  Octave locking is disabled for one
     letter note following by >, < and O[0123456].

     > -- bump the current octave up one.

     < -- drop the current octave down one.

     N <n> -- play note n, n being 1 to 84 or 0 for a rest of current time
     value.  May be followed by sustain dots.

     L <n> -- sets the current time value for notes.  The default is L4,
     quarter notes.  The lowest possible value is 1; values up to 64 are
     accepted.  L1 sets whole notes, L2 sets half notes, L4 sets quarter
     notes, etc..

     P <n> -- pause (rest), with <n> interpreted as for L.  May be followed by
     sustain dots.  May also be written '~'.

     T <n> -- Sets the number of quarter notes per minute; default is 120.
     Musical names for common tempi are:

                          Tempo          Beats per Minute
           very slow      Larghissimo
                          Largo          40-60
                          Larghetto      60-66
                          Adagio         66-76
           slow           Adagietto
                          Andante        76-108
           medium         Andantino
                          Moderato       108-120
           fast           Allegretto
                          Allegro        120-168
                          Presto         168-208
           very fast      Prestissimo

     M[LNS] -- set articulation.  MN (N for normal) is the default; the last
     1/8th of the note's value is rest time.  You can set ML for legato (no
     rest space) or MS (staccato) 1/4 rest space.

     Notes (that is, CDEFGAB or N command character groups) may be followed by
     sustain dots.  Each dot causes the note's value to be lengthened by one-
     half for each one.  Thus, a note dotted once is held for 3/2 of its
     undotted value; dotted twice, it is held 9/4, and three times would give

     Whitespace in play strings is simply skipped and may be used to separate
     melody sections.


     audio(4), pcppi(4), wsbell(4), sysctl(8)

     This speaker device was originally for the pcppi PC timer interface.
     Support was added for a synthesized device by Nathanial Sloss, first
     appearing in NetBSD 8.0.

     Eric S. Raymond <esr@snark.thyrsus.com>

     Due to roundoff in the pitch tables and slop in the tone-generation and
     timer hardware (neither of which was designed for precision), neither
     pitch accuracy nor timings will be mathematically exact.

     There is no volume control.

     In play strings which are very long (longer than your system's physical
     I/O blocks) note suffixes or numbers may occasionally be parsed
     incorrectly due to crossing a block boundary.

NetBSD 8.99.34                   June 13, 2017                  NetBSD 8.99.34