Cracking the unisons

[Ric Brekne]

Robert !!

Thank you so much for your straight forward and on topic response. I 
would enjoy greatly it if you could go into a little more detailed 
description of the algorithm behind the phase display  and  how it 
measures pitch.  It would also be cool to hear a bit about how measuring 
of several partials  at once (ie inharmonicity readings) is done on 
different machines. 

Cheers as always
RicB

.................................................................................................
Ric Brekne asked for it, so here's the deal on FFTs and TuneLab.

The comments are correct that the FFT is used only for the Spectrum 
Display, which is simply a graph of the FFT power spectrum in the 
vicinity of the desired pitch.  The fine tuning is usually done with the 
Phase Display, which is functionally like the Accu-Tuner lights because 
a phase measurement is being made over and over again.  It is the phase 
between the selected partial and an internally-generated ideal signal.  
In the case of TuneLab, this is all simulated in software.  In the case 
of Accu-Tuner, there is an actual quadrature electrical signal 
synthesized by the computer chip which is used to strobe transistor 
switches that drive the LEDs.  (I recommend the original Sanderson 
patent which describes this.  It is very readable.  I don't have the 
patent number handy, but you can find it on-line at the Patent Office 
website.)  By watching how the phase drifts forward or backward, the 
user determines if the piano pitch is high or low.  The advantage of a 
phase-type displa!
 y is  
that there is a perfect trade-off between how long you watch the display 
and how precise the pitch comparison is.  The limiting factor is not the 
technology, but rather the length of the sustain of the note being 
measured.  Every time you strike the note, the phase starts out in a 
random place, so every strike is a new independent measurement, not a 
continuation of the measurement.  This analysis applies to the Yamaha 
PT-100, the Conn Strobo-tuner, the Peterson strobes, the Accu-Tuner, and 
TuneLab.  I don't know if Dave or Dean are using a phase-type 
calculation.  Even though both programs use a spinner, which looks like 
a phase-type display, it is entirely possible to synthesize the movement 
of the spinner from calculations that are based on FFTs.  In principle, 
it should be possible to extract the same amount of information from a 
FFT, provided the sample period was optimally-chosen.  The description 
of the CyberTuner in Dean's patents seems to imply that he is using a  
time-domain bandpass filter to select a particular partial, and then 
measuring the time between a large number of zero-crossings to determine 
pitch.  Again, in principle, given the optimal bandpass filter and 
number of zero-crossings, it should be possible to extract the same 
amount of information as any of these other methods.  In my opinion, 
whether you use quadrature demodulation (like TuneLab and Accu-Tuner) or 
FFTs, or bandpass filters with zero-crossings, there is no intrinsic 
difference in the potential of these methods.  The differences, if there 
are any, would lie in the details of how each method was implemented.

Robert Scott
Real-Time Specialties