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

[Geoff Sykes]

Philippe --

OK, I won't say you're wrong. But I do disagree. 

When the 440Hz tone is recorded, the bits and the bytes are all packaged
together in words that are created at a sampling rate of, say, 44.1KHz. Each
word contains a lot of other information, the most important being clock
speed synchronization data and error checking checksums. The clock speed in
the playback device can deviate a little bit without too much concern
because part of the D/A decoding process involves synchronizing circuits
that attempt to keep the playback clock the same as the sampling rate on the
source when it was created. 

When the playback clock can't correctly sync with the source clock, or it
locks at, dare I use the term partial of the source clock, then you start to
hear artifacts like ticks. If these clock rates deviate too much then it
just won't play at all. However, when it comes to decoding the data, if
error checking is not able to automatically correct small playback
conversion errors, then you start to hear things like dropouts. But
regardless of playback decoding and synchronization errors, the one's and
zero's are still in the same sequence and I contend, therefore, that the
tone will still be 440.

I spent many years as a sound engineer in the music industry. My last
corporate gig was as a Mastering Engineer at A&M Records, in Hollywood. I'm
no stranger to digital either. But I don't know it all, and I'm the first to
admit there's always more to learn and understand. Just like piano
technology. 

-- Geoff Sykes
-- Assoc. Los Angeles



-----Original Message-----
From: pianotech-bounces at ptg.org [mailto:pianotech-bounces at ptg.org] On Behalf
Of Philippe Errembault
Sent: Friday, May 26, 2006 5:31 AM
To: Pianotech List
Subject: Re: ETDs, PCs,PDAs & cellphones vs tuning fork : how accurate are
they ?


Hi Geoff,

The problem is not in trasmission. what you listen to are not bytes in a
buffer, but the need to be re converted to analog sound. The problem is in
the digital to analog conversion. Let's assume one use a sampling rate of 44
100 Hz, like on CDs. you can always calculate the samples instead of really
recording them, so let's assume the recording was perfect. So a one second
440Hz tone, will contain 44100 samples which will contain 440 period of a
sine wave.

Now, imagine you play this sample on a sound device whose quartz is 1% too
slow. the real play back will only play 44100-441 = 43659 samples during the
first second. and so you will only play 440 * 43659/44100 = 440*99% = 435.6
periods of the sine wave during this second and so the frequency of your
tone will be slowed down to 435.6 Hz.

Please think well about it before saying I'm wrong, because I may be a
newbee in tuning but I'm definitely _not_ a newbee in digital sound
processing. ;-)

Philippe Errembault

----- Original Message ----- 
From: "Geoff Sykes" <thetuner at ivories52.com>
To: "'Pianotech List'" <pianotech at ptg.org>
Sent: Friday, May 26, 2006 6:54 AM
Subject: RE: ETDs, PCs,PDAs & cellphones vs tuning fork : how accurate are
they ?


> If the transmitting or storage medium of your source was analog then 
> clock speeds, transfer, repeater and amplifier delays, and any number 
> of other mishaps could indeed alter the frequency. Think speeding up 
> or slowing
down
> a record player. (You do remember records, don't you?)
>
> Digital, on the other hand, is simply a collection of one's and zero's 
> in
a
> very specific order. Regardless of clock speed, storage, transmission,
noise
> or any number of inconsistencies in the signal path, even if you 
> deliberately change the speed the transfer process or processor 
> clocks, those one's and zero's will ALWAYS arrive in exactly the same 
> sequence in which they were sent. When they don't, the result is noise 
> and dropouts,
not
> speed or frequency shift. In other words, a calibrated 440Hz tone sent
over
> a digital signal path will still be 440Hz at the receiving end. 
> Changing
the
> pitch of a digital signal is possible, but it involves some serious 
> and deliberate number crunching.
>
> Broadcast tones, ala WWV, I would not trust as absolutely accurate. 
> Tones over the phone, unless you are absolutely sure there is no 
> analog stage in the chain, I would also hold suspect. Also, once a 
> digital signal has been converted back to analog it passes through 
> amplifiers and a speaker so we can hear it. Unless those components 
> are extremely high quality the end result might not be exactly spot 
> on. But unless you have a precision frequency counter or a calibrated 
> spectrum analyzer handy you probably
would
> not be able to tell. Just how accurate do you want to be? We're 
> talkin' pianos here, not rocket science.
>
> -- 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: Thursday, May 25, 2006 6:59 AM
> To: pianotech at ptg.org
> Subject: Re: ETDs, PCs,PDAs & cellphones vs tuning fork : how accurate 
> are they ?
>
>
> Philippe Errembault writes:
>
> > Do you usually have an idea of the precision of the A440 reference 
> >you use ? I mean... I didn't get any the precision information with  
> >my tuning fork, I found it on the net. I wonder what are the 
> >precisions of professional ETDs, and what  precision we can expect 
> >from a pocket PC, about which I wonder  if it even can be as precise 
> >as a normal PC...
>
> If you have back-issues of the PTG Journal you might want to take a 
> look
at
> "Calibration of Pitch References" in the August 2001 issue.  It shows 
> that all major ETDs are calibrated to better than .01 cents. The 
> article also discusses tuning forks, their temperature dependency and 
> accuracy, and how to calibrate them.
>
>
> > ...it might be that PDAs only contain ONE reference clock, for time, 
> > for CPU and for sound processing.
>
> I have found that not to be true.  In fact, even on the Pocket PCs, 
> they sometimes have separate audio sample rate oscillators for the 
> listening
mode
> and the sound-generating mode (recording and playback).  As for the 
> CPU clock, that is definitely not tied to the audio clock because the 
> Pocket
PCs
> take advantage of switching to slower CPU clock speeds during idle 
> times
to
> save on power.  You can't have your audio processing clock tied to 
> such a variable clock source.
>
> It really does not matter that the audio sample rate is not precise, 
> as
long
> as it is stable.  Once you do a software calibration on your ETD, the 
> results are the same as if the oscillator were perfect to begin with.
>
>
> > There was also someone who claimed that listening a tone reference 
> > through a cellphone was as good as with a normal phone. I know the 
> > buffering of cell phone is small, but this doesn't change the fact 
> > that any shift in it's clock should be retrieved in the output!..
>
> You can depend on the frequency of sounds being delivered accurately 
> over the cell phone network.  There is no pitch distortion.  What is 
> unclear is whether the same can be said for the Voice-over-Internet 
> Protocol Internet services that deliver telephone service through your 
> computer.  This is potentially subject to indeterminate TCP/IP 
> buffering and perhaps pitch distortion.
>
> Robert Scott
> Ypsilanti, Michigan
>
>
>
>