How did old Harmonizers / Pitch-shifters work?

michael

Bring out the vacuum
So pitchshifters like Eventide's old Harmonizer have been around for decades. Old octave pedals for guitars, likewise.

How did the earliest ones work?

Have they always used sampling / granular synthesis technology?

I got the feeling the earliest ones must be analogue, based on their age, but, if so, I have no idea how they do what they do.

Having had no luck with some googling, thought I'd take a punt that someone here could explain or point me in the right direction.
 

polystyle

Well-known member
Not surehow they worked M
but -
I don't know if they used Granular Synthesis,
would venture that it makes sense if the Eventide is analog , they have been around for awhile.
I can ask my engineer.
I remember them coming along around the same time as Sampling Delays like
the Publison.
There was one outboard device called the Space Station from same period.
Better Studios used to have the pair , an Eventide Harmonizer and less used Space Station.

Some great Harmonizer work came out of Conny Plank's Studio and is heard on the Liasions Dangeruses album cuts, turning and tearing vocals into animal growls , gibberish,
DAF's pitch changed Gabi's vocals into a female chorus.

Also of course the drum sound on Bowie's Low thanks to a Tony Visconti inspiration.
http://www.audiomasterclass.com/arc.cfm?a=the-audio-machine-that-warps-the-fabric-of-time
 
Last edited by a moderator:

doom

Public Housing
It's something to do with varying delay time & the doppler effect. You can build a pitchshifter using a Nord Modular by varying the speed of two chained delays, using no sampling at all. It's how delay/pitchshift pedals like the Boss PS-3 work & the reason that most old pitchshifters get that nice flutter as the pitch rises.

A digital delay is sampling. The Boss PS pedals are very much digital devices.

With guitar, or rather the squarewave you get by running a guitar thru a fuzz box, you can mimic octave up or down by driving a rudimentry oscillator with the squared up signal. This is, roughly speaking, the way analogue octave pedals work.

But thats not a harmonizer / pitch shifter, which as far as I know are all digital at heart. You've gotta be able to grab an input signal, speed it up / slow it down & squeeze it back out again really quickly, if you modulate the playback speed of an analogue delay with a fast LFO you will get a pretty awesome sound, but its not going to useable as a pitch shifter or harmonizer.
 

massrock

Well-known member
The earliest Eventide machines were digital. I think that was kind of their big breakthrough.

http://www.eventide.com/About/History.aspx

But you can do various frequency shifting effects in analogue. Not exactly pitch-shifting though, I think it's usually frequency-addition. And you can also take the octaving process doom describes further to produce other harmonies but I think it only really works well with very simple waveforms.

http://www.modcan.com/modhtml/freqshift.html

http://www.schumannelectronics.com/pll.html
 

doom

Public Housing

PLLs are fun! I would be surprised if he wasn't using a CMOS IC tho, surprised & in awe!! There will always be some lag in these designs no matter how well implemented, but PLLs are often used for freq. multiplication & would make a nice harmonizer, especially if you didn't require exact tracking and embraced the notes between notes that they produce;

<object width="425" height="344"><param name="movie" value=""></param><param name="allowFullScreen" value="true"></param><param name="allowscriptaccess" value="always"></param><embed src="" type="application/x-shockwave-flash" allowscriptaccess="always" allowfullscreen="true" width="425" height="344"></embed></object>

A PLL is doing a very similar thing to what I described with octave fuzz boxes. Take an external signal & use it to drive an oscillator, its just a smarter circuit, it looks at its output, compares that to the input & adjusts itself, so you get the kind over/under-shoot you can hear in the video. Which is not technically ideal but really nice in practice.


Varispeed recording is awesome! Its alot cleaner than using a processor that carves up the signal.
 
Last edited:

michael

Bring out the vacuum
Awesome, am in nerd heaven.

I was never too sure, is "varispeed recording" just changing the tape speed when recording to tape? What Prince did lots of, for example. Used to have a lot of fun doing that on a friend's 4-track. Likewise with a tape delay, although that's more obviously played out.
 

michael

Bring out the vacuum
Also, frequency shifters are nnnuts. My software of choice has one in it, and it obviously makes the tuning completely rooted (sorry, Aus / NZ English for "fucked") very quickly.
 

DJ PIMP

Well-known member
people on gearslutz would be able to tell you all about vintage harmonizers and pitch shifters etc.
 

Papercut

cut to the bone
CHAOTROPIC is right, put simply, its a really small delayed signal (milliseconds) that modulates the original sound so that the pitch raises up and down slightly. The doppler effect part comes in because the frequency (pitch) is percieved to change because there are two signals that are not in phase with each other but not noticeably enough for them to sound like different things, so they harmonize to our ears. thats how i understand it anyway.

delay is the basis of lots of other early effects (like chorus.)




Also, I've been reading this forum for a while but never posted before, so hello to everyone also.
 

poetix

we murder to dissect
There are two pitchshifters I use in Buzz, the TunaMan and the Neptune. The first chops the incoming signal into very short (and slightly overlapping) slices, and replays each slice looped at a faster or slower rate. This is at least in principle reproducible via some analogue mechanism. The second runs everything through an FFT to get a frequency-domain snapshot of what's going on in a short window, shifts all the frequencies up or down, then does a reverse FFT - the result is much cleaner and less fluttery, but you have a lag of (at least) the number of samples in the FFT window, plus processing time for the FFT algorithm. There is no analogue way of doing this that I know of.

Early pitch-shifting pedals (not octavers, but things like the whammy pedal Tom Morello was so fond of) used the first technique AFAIK. They were all digital, using sampling technology. I remember being astonished by the sounds Steve Vai was getting out of the Eventide Harmonizer, which seemed impossible at the time. The prize for Harmonizer abuse goes to Reeves Gabrels, who used it to make all manner of unpleasant noises with Tin Machine.

The better modern FFT pitch-shifters for voice or guitar know about transients, and shift only the parts of the sound that actually change as the pitch of the instrument goes up and down. I don't know of any that can do this quickly enough to be usable for live signal processing, though.
 

massrock

Well-known member
The better modern FFT pitch-shifters for voice or guitar know about transients, and shift only the parts of the sound that actually change as the pitch of the instrument goes up and down. I don't know of any that can do this quickly enough to be usable for live signal processing, though.
Do you mean real time pitch shifting with formant correction?

I think the latest Eventides (H8000FW) with their UltraShifter algorithm will do this. As does RBC Voicetweaker (like Autotune) and maybe some others.

Eventide said:
ULTRASHIFTER™ THE APEX OF PITCH MANIPULATION

The H8000FW incorporates Eventide’s most advanced formant-maintaining pitch shifter optimized for the human voice. The UltraShifter’s real-time adaptive resynthesis allows for unmatched modification of vocal timbre. So, if you are changing gender or going for something completely bizarre, we’ve got you covered.
So as poetix says it's resynthesis, not based on delays in this case.
 
Last edited:

poetix

we murder to dissect
I wonder whether anyone's done any psychoacoustics-based optimisation of resynthesis-based pitch shifting. Not so much optimising for the human voice, as optimising for the human ear, as with mp3.
 

poetix

we murder to dissect
On a related note, does anyone have any insight into how the guitar modelling technology in Line6's Variax guitars works? It seems to be a form of resynthesis which takes the relatively pure output from a piezo pickup on each string and maps it to the frequency response patterns of various real guitars. I played one a while back and it wasn't bad, although personally I'd be more interested in getting it to create wacky unheard-of sounds rather than not-quite-convincingly emulating a vintage Les Paul...
 

michael

Bring out the vacuum
CHAOTROPIC is right, put simply, its a really small delayed signal (milliseconds) that modulates the original sound so that the pitch raises up and down slightly. The doppler effect part comes in because the frequency (pitch) is percieved to change because there are two signals that are not in phase with each other but not noticeably enough for them to sound like different things, so they harmonize to our ears. thats how i understand it anyway.

delay is the basis of lots of other early effects (like chorus.)

Also, I've been reading this forum for a while but never posted before, so hello to everyone also.
Hello hello. :)

I have to say what you describe actually sounds more like a chorus than what I was thinking of. I was meaning something where you can transpose a signal up or down something like an octave. I can't imagine how you could do that by modulating a delay line because the modulation would be too extreme.

Sorry if I wasn't clear.
 

massrock

Well-known member
Actually it's hard to see how there wouldn't still be some considerable latency in those resynthesising pitch shifters.

.

I always thought more should have been done with resynthesis but other than the old Fairlights not many commercial systems have used it afaik. Now there's Camel Audio's Alchemy synth which looks very interesting for that kind of thing.

There's also this which has been knocking around forever and hasn't been updated since 1998.

http://scrime.labri.fr/InSpect/

It's actually not hard to do. I did some experiments of my own a while ago but my computer at the time wasn't really fast enough.
 
Last edited:
Top