Cyber ears

[Jim Coleman, Sr.]

Hi Richard:

Not only is this doable, but it has been done, way back in the late '70s.
Francis Mehaffey used such a machine to listen to specific coincidental
partials and then provided a digital readout. I believe his son is the
one who designed the machine following the ideas of his father. Incidentally
Francis is now in a rest home somewhere in or near Claremont, CA.

Francis used this instrument for several years.

The problem in using such an instrument is that it is as difficult to figure
out how the piano piano wants to be tuned as it is difficult for the aural
tuner to figure this out. For example, the F3-A3 M3rd is not always 6.93
bps at its lowest coincident partials. Due to scaling variations, one piano
may require 6.5 bps whereas another would require 7.1 bps (beats per 
second). The present state of the art electronic tuners have the ability
to smooth curve fit the tuning to the needs of the individual piano. I 
especially like the way the SAT III can do this because of the control of
the double octave beating. Another advantage of the presence popular 
machines is that the smooth curve is mathematically smoothed based upon
certain end points which are established by measurements of certain 
partials. The in-between-points are very accurately established, whereas
in the case of aural tuning or tuning with a beat-readout machine, there
are 12 possible small mistakes which can be additive in their effect and
cause the 12 note temperament to be skewed one way or another. This becomes
quite evident once you extend this tuning beyond that first temperament
octave.

Now, to get off of the negatives for such a machine. It would indeed be
possible to develop such a machine that would compute a smooth curved
tuning, but there might be some problems with the patents which Dr.
Sanderson holds where he claims the right to any type of tuning which is
derived from measured partials. Perhaps one who wishes to pursue this could
get a license to use some of his patent ideas and apply them to a machine
which listens to beats and gives digital readout of the same to be compared
to a pre-calculated set of beat rates. I still think it is much simpler
to do as the present state of the art machines do it.

Jim Coleman, Sr.

On Fri, 16 Apr 1999, Richard Brekne wrote:

> 
> 
> Robert Scott wrote:
> 
> > Several suggestions have been made for an electronic tuning aid that
> > displays beat information only, without any regard to an assumed
> > tuning curve.  Speaking as one who has written ETD software, I would
> > like to address this issue.
> >
> > All major ETDs display beat information between a single note and
> > an internally generated reference signal.  Furthermore, all major
> > ETDs display information on only one partial at a time.  Steve
> > Fairchild has suggested tuning while considering two or more
> > partials at a time and has even constructed some experiements
> > using several SATs at once (set to different partials).  But
> > the comparisons are always between a single note and an
> > electronically synthesized reference.
> >
> > What is different about the present proposals is that the ETD
> > not use any internal reference.  Instead the technician would
> > strike two keys at once, just as in aural interval tuning,
> > and let the ETD "listen" to the beats between the two
> > different notes and display the results in some manner.
> >
> > IMHO, this ETD would be significantly harder to make than the
> > current ETDs.  What you find fairly easy - picking out beats
> > between two simultaneously sounding notes - actually relies
> > on the incredibly complex pattern recognition capabilities on
> > the ear-brain combination.  There are amplitude fluctuations
> > across the frequency spectrum.  Yet to hear beats, you have
> > to be able to pay attention to amplitude fluctuations at
> > only one frequency, ignoring all the rest.
> >
> > There is a similar problem in radio communication with
> > Morse code.  Like beats, Morse code is made up of
> > amplitude variations.  Within the crowded amateur radio
> > bands you may find a number of stations so close together
> > in frequency that the radio is incapable of separating
> > them.  But a highly skilled amateur radio operator can
> > pick out and understand a single transmission admidst
> > all the interference.  When computers came along, it was
> > natural to try to make a computer read Morse code.  This
> > effort has been met with limited success.  Computers can
> > read Morse code, but only when it is fairly clear.
> > Under the most difficult situations, the human listener
> > still copies Morse code better than the computer.
> >
> > If an electronic device can detect beats, it would only
> > be in situations where the beats were easy to hear
> > anyway.  As beats get harder to hear, the electronic
> > device will fail sooner than the human ear.
> >
> > -Robert Scott
> >  Real-Time Specialties
> 
> All this is true as you state, But if the machine was to set to the job
> of listning to the partials of each of the two notes, and then
> calculateing the difference between all coincedentals, would this not be
> an approach which would work. I am not suggesting that a computer should
> "hear" the beats as the human ear does. Rather exploit the number
> crunching capability of a computer.
> 
> We hear beating between two notes, and try to estimate the beat rate as
> you say based on the amplitude or shall I say "hearability" of the
> beating, and its real speed. Excellent tuners are amazing in this
> regard. Trying to emulate this capability humans have directly in a
> computer does sound like a tough one.
> 
> But the computer can sample to notes, and all relavant partials. Compare
> the frequences of coincidents and calculate the difference, and finally
> present the resulting data in terms of bbs.
> 
> Does this sound do-able ?
> 
> Richard Brekne
> 
>