[78-L] Pitch shifting

[Michael Biel]

From: DAVID BURNHAM <burnhamd at rogers.com>
> Michael, I assume you work with Digital Audio Workstations, (DAWs),

Rarely.  Actually, I work more with REAL equipment, which has a heritage
and a set of terminology which existed before the computer and before
the geeks who design the DAW programs were even born.  These terms,
including "Pitch Shifting" were originated before computers were
invented.  I use pitch shifters that are not attached to a computer.  I
use turntables with speed controls that are not attached to a computer. 
I use tape recorders with variable speed controls that are not attached
to a computer.  I do these operations BEFORE feeding the signal into the
computer because some of these operations, as you state: "if it done to
any extent [in the computer], usually introduces undesireable audio
artifacts."

Michael Biel wrote:
>> This is the problem with the terminology again.  You can't just pitch
>> shift.  You have to speed adjust.  The term "pitch shifting" was coined
>> to mean changing the pitch without changing the tempo. 

> any professional DAW has a plug-in called a "Pitch shifter", and any pitch shifter
> has a setting which allows it to adjust tempo along with the pitch, (like speeding
> up an analog tape), or change the pitch without changing the tempo, (which if it's
> done to any extent, usually introduces undesireable audio artifacts). 

That was exactly my point.  The change in the terminology allows for a
misunderstanding of what is needed, and that happened in this case.   If
you adjust ONLY the pitch, that will not correct the problem of the
original question.  The pitch and tempo must be LOCKED together, and
that is not an inherent understanding if the only the word "pitch" is
used.  Because the word "speed" had a 100 year history of meaning just
that -- the simultaneous adjusting of the pitch and tempo -- it would be
helpful if "speed" would be used to describe this function, leaving
pitch to mean adjusting only the pitch, and tempo meaning adjusting only
the tempo.  The terms were changed by geeks who did not know much beyond
computers and did not adequately think things through. 

Let me explain how the problem arose.  Most of the geeks who designed
computer programs and computer-based recorders had never used a
turntable or a tape recorder.  Those were old fashioned and obsolete. 
So their concept of "speed" only meant how fast their computer
processors could pass bits (or possibly the drugs they used to allow
them to stay up endless hours to meet production deadlines). 

 Variable speed turntables go back to the Berliner days of the 1890s,
and for 100 years turntables and variable speed tape recorders had a
control called "Speed".  Not "pitch".  Not "tempo".  SPEED.  It was well
known that adjusting the speed of a turntable or tape recorder would
adjust the pitch of the sound.  It was well known that adjusting the
speed of a turntable would also adjust the tempo of the sound.  But for
some reason every turntable or tape recorder ever made only had a
control marked "SPEED" -- at least till the geeks took over.  

The term Pitch Shifter originated before the first DAW, and was adapted
by the geeks who didn't quite know what the earlier devices actually
did.  As I mentioned in an earlier posting, there were tape oriented
pitch shifters going back to 1935, and the first all-electronic pitch
shifter was the Eventide Harmonizer which was introduced around 1972. 
These operated independent of any recording device, and shifted the
pitch up or down of a live performance or of the playback of a recorder.
 Obviously there was no time expansion or compression possible with live
performances -- the tempo remained the same -- but if the source was a
recording, there was a way of adjusting the tempo of the recording by
changing the "speed" of the playback and then using the pitch shifter to
RESTORE the original pitch.  Thus, that was TEMPO SHIFTING.  

So probably without realizing that there had been a 100 year history of
using a "speed" adjustment which would change the pitch and tempo of a
recording, and a fifty year history of using "pitch shifting" to mean
the changing of the pitch without changing the tempo, and "tempo
shifting" to mean the changing of the tempo without changing the pitch,
they dropped the name of the "speed" control and started to instead call
it "pitch shifting".  

Now you occasionally see variable speed controls on turntables labeled
"Pitch".  Some of them now have a "pitch lock" toggle which turns the
speed control into a tempo control because the pitch does not vary
because it is being pitch shifted.  But if they would add a tempo lock
option, then the control WOULD be a pitch control because the tempo
would remain the same as the rotational speed changes.  But the speed
control without either lock is still a speed control.  Calling an
unlocked speed control anything else is stupid and confusing because the
locks change the function.

David's posting contained several other examples where terminology
changed, but I find the history he cites to be faulty.

> The term "equalizing" was coined to mean a circuit which was employed to
> compensate for the action of an earlier stage which distorted the frequency
> response for some purpose - like recording characteristics.  Today an
> equalizer is any unit which manipulates frequency response and really
> isn't "equalizing" anything.

Actually the term was first used in the motion picture industry for
devices which adjusted the sound quality of one sound pickup with
another so that the voices sounded "equal" throughout the film.
Parametric equalizers were used, and the Lang-Pultec was one of the
later types in the post-war years.  Your definition of using the word
for pre-emphasis and de-emphasis recording curves is a later one.  I
don't have the original documents in front of me right now, but I think
the paperwork that were published in the late 30s by the NAB and RCA for
their transcription recording and reproducing curves refer to them as
"compensation" not "equalization".  "Compensator" is the term RCA used
for its six-position frequency curve modifier in its series 70
professional broadcast turntables.  They also called it a "filter".  But
never an equalizer.  Looking thru some of the manuals from the 30s and
40s I never see the word "equalization".  The terms Pre-emphasis and
De-emphasis were also used along with compensation. I don't think that
the term equalization was used for recording curves was used until the
late 40s or early 50s. In fact, the first use of it I see of the word in
RCA literature is in 1950 for the 56-E Telephone Line Equalizer which
allowed for the adjustment of the sound quality of different non-linear
broadcast telephone lines by two multi-position resister switches.  


> A Synthesizer used to refer to an electronic instrument which could generate an electronic signal with appropriate harmonics to simulate the sound of real acoustic instruments, (like electronic organs and pianos).  Now a synthesizer just refers to any instrument which generates electronic musical sounds which usually don't sound like any real musical instrument.

Actually the word Synthesizer was coined to define an electronic
instrument which was NOT designed to simulate any one specific
instrument.  Electric organs existed decades before the use of the term
synthesizer but were never called synthesizers at the time.  It has only
been called a synthesizer retrospectively in recent years by people who
don't know that the term was not used when the electric organ was
introduced in the 30s and 40s.  The Columbia-Princeton Synthesizer in
the 50s did not sound like a real instrument. Listen to the recordings.
The Moog Synthesizer in the 60s did not sound like real instruments
without a lot of manipulation, and even then rarely sounded real.  Did
the "Well Tempered Synthesizer" sound like a real acoustical
instrument?????  Truly real sounding synthesizers actually came later --
and many of them were really "samplers" which manipulated or controlled
the algorithms derived from recordings of real instruments.  Some of the
cheaper ones are really just samplers.  They take a sample of one note
of an instrument or a voice and pitch-shift it upwards and downwards
with the keyboard. The sounds of first truly real sounding electric
piano, the 1983 Kurzweil, were developed from analyzing samples of real
sounds and building these algorithms into the device rather than
requiring the user to create the algorithms.     

> I think the action you're describing above is called "transposing", like what you can do with an electric piano or an electric organ..

Actually you can transpose on any musical instrument.  It does not have
to be electronic. (And Irving Berlin had his famous transposing piano
with a special movable keyboard.)  But, since the first electronic
(non-tape) pitch shifter was the Harmonizer in 1972, one of its uses was
to change keys, and the adjustments were marked in musical half-tones
rather than in percentages.  But my point was precisely that this is NOT
what you want to do with an analogue recording that must be speed
adjusted like the World records, but the first proposed solution to the
problem offered the transposing type of pitch shifting.  



 > Obviously, in the case of a constant velocity disc recording, (and I
realize constant velocity can have an entirely different meaning when
applied to equalization), you will want the pitch and tempo to change
together to compensate for the situation on the record when played at
constant speed - where the pitch and tempo will lower as the record is
played, (if from the outside in).  db

This is the real conundrum -- does the recording need to be speeded up
or slowed down!  It depends on what speed you played it at as well as
whether it is a center or rim start.  Let's assume that the World is a
rim-start disc.  The original questioner later in the discussion
mentioned that he had played the record at 33.  Thus as the record
continued to play it will sound like it was slowing down because it had
been speeded up when it was recorded.  Thus the computer must gradually
speed UP the recording.  BUT, what if he had played it at 78?  It would
start too fast and seem to slow down while playing it back because the
record would continue to approach the playback speed of the end of the
recording.  Thus, the computer will have to initially slow down the
recording but gradually reduce the amount of the speed reduction.

If it is center start like CDs or CAV Laser Discs, the disc was recorded
at the start with a faster rotation in the center and slows down towards
the rim.  If played at constant rotational speed things will work
opposite the above.  If you start playing it at 33 in the center it will
start slow and speed up to normal.  If you start at 78 in the center it
will start correctly and will speed up as it goes towards the rim.  

Offhand, I think it would be better to play the record at the slowest
speed in either case.  The frequency response of the slow rotational
part of the recording will possibly exceed the response of the
re-recording system if it is played at more than double speed.  That
will negate the purpose of the constant linear velocity recording -- but
it will make the dubbing time longer!   

Mike Biel  mbiel at mbiel.com