ETDs, PCs, PDAs & cellphones vs tuning fork : how accurate are they ?

[Geoff Sykes]

NIST tones are about as accurate as science can get. They are locked by
means of an atomic particle reference so it's unlikely that if a change in
pitch were to happen at the source that it would even be measurable.

Changing pitch in digital does not happen accidentally. One's are one's and
zero's are zero's. They are on/off switches that are in one state or the
other and don't just randomly change. Both pitch and time changes in digital
require deliberate manipulation of the data, (the one's and zero's), using
very complex algorithms. In order to change the pitch you need to modify the
waveform making it longer or shorter in duration. In analog that's just a
matter of speeding up or slowing down the reference. In digital it means
resampling the entire waveform and then recreating it again in the desired
shape. That means, (for a CD), 44,100 pieces of information need to be
sampled, recalculated and recreated every second in order to change the
waveform in such a way that your output is now different than your input.
Any discernable change in a digital signals pitch or time involves a lot of
math and technology and, I repeat, does not happen accidentally.

If something like a NIST reference tone of 440Hz is recorded digitally, and
it stays in a digital format until you receive it and convert it to analog
for reproduction, it will still be 440Hz. 

A regular phone is analog and subject to all kinds of distortion. VoIP is
digital and subject to no distortion. Using a cell phone as an example, when
you are talking to someone on your cell phone and the other persons voice
gets all garbled up, that is not distortion so much as it is a loss of
signal creating dropouts and lost information. What you DO hear between
those dropouts is actually clean, although it may not seem so at the time.
My point is that comparing a tone from a digital source to a tone from an
analog source is not fair. Now if you were able to compare that same tone
from two different VoIP providers at the same time, that would be fair since
it would then involve two independent signal paths. Someone do that test and
get back to us.

BTW, synchronization is not a problem with digital. All transmissions over
the internet, either data or voice, are via "packets". These packets are
sent out in the sequence in which they are created, but because of the very
nature of the internet they may arrive at their destination in a totally
random order. Fortunately these packets are numbered and one of the jobs of
the cache, or buffer, in your receiver, (computer), is to sort through all
these packets and put them back in the right order before sending them on to
your handset. (One of the biggest obstacles in VoIP technology was
overcoming the processing time required to both receive and process those
packets strings in corrected order fast enough to not cause objectionable
delays.) If a packet or two is delivered too late it is simply dropped. You
are never going to hear the loss of one or two lost packets. But if a whole
bunch of packets get lost then what happens is you may get drop outs. 


-- Geoff Sykes
-- Assoc. Los Angeles



-----Original Message-----
From: pianotech-bounces at ptg.org [mailto:pianotech-bounces at ptg.org] On Behalf
Of Robert Scott
Sent: Saturday, May 27, 2006 3:25 PM
To: pianotech at ptg.org
Subject: Re: ETDs, PCs,PDAs & cellphones vs tuning fork : how accurate are
they ?


If I may dare to jump into the discussion with Philippe and Geoff, let me
say that for five years now I have been recommending that TuneLab customers
use the NIST tones through their telephone to calibrate their Pocket PCs and
laptops.  Throughout all that time, no one has ever reported to me that when
they check the calibration using their cellphone that the pitch is anything
other than rock-solid and repeatable.  If there were any day-to-day pitch
changes due to something that happens through the cellphone network, someone
would have noticed it by now.  I am reasonably sure that cellphones produce
faithful representations of the pitch that was sent.  Precise time-base
synchronization is the bread-and-butter of the cellphone companies.  They
use it extensively to synchronize all the data transmission their network.
If there is a phase-locked-loop in the cellphone, you can bet that it has no
audible frequency modulation.

However, because the question came up a few days ago, I am still
investigating VoIP (Voice Over Internet Protocol) - the service that uses
the Internet to place telephone calls.  Because of the lack of precise
synchronization over the Internet, it is hard to see how pitch precision can
be transferred over the Internet.  VoIP was one of the topics on last
Friday's "Science Friday" on NPR.  After the program I e-mailed the the
guest (the author of a book on VoIP) asking about possible pitch distortion
through VoIP.  Unfortunately, this "expert" was actually unaware that audio
CDs can have pitch distortion, depending on the precision of the playback
speed.  So I was unable to get any useful information from him.  However, if
any of you have VoIP, you can perform the following easy experiment and
report back to the rest of us:

Call the NIST at (303)499-7111 in Colorado on your VoIP connection and
simultaneously on your regular phone.  Listen for any beat between the two
sounds.  There should be none.  But if there is, then we have evidence of
pitch distortion through VoIP.

Robert Scott
Ypsilanti, Michigan