Harmonizer Schematic

[rocker-D82]

I would build an harmonizer. Anyone have a schem??

Thanks

[ESPguitar]

When we're at it...  :roll:  :wink:

What is an Harmonizer?

Thanks, and sorry for the stupid question..

RB

[Paul Perry (Frostwave)]

It's very difficult to build an analog harmonizer.. you need a bunch of BBDs & each of these samples the signal for a short while, then reads it out at a different clock rate, and they all take turns doing this, and you try to fade them in and out so that it souds as good as it can.
I'm afraid that harmonising is one thing that a DSP chip is a million times better at.
BUT!!! those cheapo "gender bender" telephone voice distorter chips... anyone made a novelty 'harmoniser' from one of those? I doubt the harmony would be a very musical one:shock:

if anyone has a line on those chips let me know, though.

[SnooP_Wiggles]

digital is best at harmonizing/pitch shifting by far. Don't assume analog is always best - its only best when its the reference point as it is for most other effects. as far as actually making a pedal, well Peter mentioned a DIY dsp thing a while back could bring that up again. otherwise just buy something, theres digitech rack units on ebay (eg ips33-b, dhp-33 etc)

[jmusser]

There's an EMM Harmonic Generator on Mark Hammer's site, I believe on Page 8. I have been looking at it myself. It looks like there's not much blending, it's mostly single octaves or 5ths and 3rds at a time, but it's supposed to do 3 octaves up , and 3 octaves down, using a 4040 counter chip. I have no idea what it would sound like, but it looks like a medium level build.

[StephenGiles]

The snag with all these "harmonisers" is that you really need an intelligent one, which will generate a sensible harmony from a given key and type of scale. In other words it will provide changing harmony intervals and sound musical  - perhaps like the solos on Boston records. A fixed interval can sound very odd.
Stephen

[puretube]

oneday, when I`m grown up, I want an Eventide...  

[Mark Hammer]

Okay, Aron and Peter, snag this one for an FAQ.  If there are other categories, folks, please feel free to add them:

The Kingdom of "Other Notes"[/u]

Many players want to be able to produce notes other than the one they are currently playing for a fuller sound.  A variety of methods have been developed for doing this, each with their pros and cons.  Here is a brief synopsis of all the available methods.

Division  Note-division is used for the purposes of octave dividers and production of tones *lower* than the one being played.  This method uses a flip-flop switch/device that moves back and forth between high and low states, depending upon the state of the input signal.  Since it moves up and down for each 2 or 4 times the input note moves up and down, it can produce  one and two octaves below the input note.  Since it can only keep track of one note at a time, it is limited to single note runs.  More than one note at a time confuses it.  Complex or unstable waveforms confuse it too.  Since it is a threshold-based device, it does not react very predictably to the onset and decay of notes, unless additional circuitry conditions the signal.  It does NOT produce a pitch shift.  Rather, it causes another tone to be produced, which is a square wave, and, happily for the musician, is related to the input signal.  Because it is a product of integer-division (i.e., even numbers) there is nothing it can do other than notes below the original in divisors of the original.  So, it can divide the original by 2, 3, 4, 5, etc., but it can not divide the original by .66 or 1.5 to yield other intervals.  Because it is a square wave, the sound will be raspy, unless it is appropriately filtered to mimic a mellower note.  Because it depends on the input exceeding a fixed threshold, it requires the user to adapt picking style and unless the input is heavily compressed will generally not respond equally well to all notes on the fingerboard.  This is a largely analog, partly digital, method and can be accomplished with modest cost.  Many commercial and DIY schems are posted around for this.

Rectification  This method is used for producing notes *above* the input note.  In contrast to the division method, it is not especially threshold-sensitive, although hotter input signals tends to produce more audible octave up sounds.  While the division method actually produces a second tone, based on the input signal, the rectification method produces what might be called an artifact of the original signal.  In other words, nothing is actually producing a second note above the original.  Rather it produces a complement of harmonics, based on the way it is fuzzed, which creates the illusion that an octave note is also being generated.  This illusion depends on the appropriate balancing of versions of this rectified signal.  While multiple notes do not confuse this type of circuit in the same way that they confuse a division circuit, nevertheless, playing one note will yield a reasonable octave, while playing two or more will yield a sloppy-sounding fuzz with not obvious octave-up sound.  Although it is technically possible to use such a method and produce 2 octaves up by cascading a duplicate circuit, realistically, this would be hard to do, and in any event, few if any commercial or DIY circuits have ever attempted it.  Probably the simplest and cheapest of "other note" circuits, however, and one of the least frustrating because it can still yield an interesting fuzz, even if it doesn't nail the octave you want.

Sideband method  When a signal is volume modulated by another signal, it tends to produce "sideband" tones that are the sum and the difference of the two tones.  So, a pure tone of 1000hz, modulated by another pure tone of 100hz, will also produce audible tones of 1100hz and 900hz.  The logical extension of this is that when a tone is modulated by itself, the sideband tones are twice the input frequency (e.g., 1khz PLUS 1khz) and silence (1khz MINUS 1khz).  You can see right away that this provides an opportunity to generate an octave up that is, essentially "fuzzless".  Of course, one needs to remember that the sum and difference output is a function of an artificial state, namely "pure" tones, which a guitar does not produce on its own.  This means that unless one wants to get sideband products of all the various harmonics incombination with each other and the fundamental, some serious fixed or adaptive filtering is needed to present the modulation section with the sorts of input and modulation signals that will yield ONLY the octave above, sop it can be mixed back in with the full spectrum original.  This is a tricky method, obviously, and also needs/requires a mono input signal to work and will NOT sound nice with two or more notes at once - unless you WANT a ring modulator.  The advantage for all this extra circuitry is that it comes closest to a nicely tracking distortion-free octave up.  It is generally not considered to be a threshold-sensitive method.  Not many posted circuits use it.

Pitch tracking  This is a method used to nice effect in the Boss DR-2 Super Feedbacker Distortion and the E&MM Harmony Generator, as posted at hammer.ampage.org .  Here, the fundamental of the note played is detected and the pitch identified (rightly or wrongly) is turned into a proportional voltage.  That voltage is then used to drive an oscillator.  Once again, this is not a note that is a pitch-shifted *copy* of the original, but is a second generated tone that possesses the timbral qualities of the original only by accident (and generally not at all).  The proportinal voltage that is produced can actually be used to drive as many voltage-driven sound generating things as you want, and is not limited to either above OR below the original note.  In the DF-2 it is used to drive an oscillator that produce a higher tone that is also divided down with a flip-flop, yielding two tones that are an octave above and the same pitch as the original.  In this case it is used to mimic how a held note sounds in front of a loud amp.  Within limits, it can "survive" bending of the original note, but is not immune to tracking errors.  Perhaps just as importantly, it takes a few miliseconds to reliably detect the pitch of the note played, so tracking is not quite as instantaneous as with rectification or division, both of which are "dumb" techniques, by comparison to this "smart" one.  In the Harmony Generator, this tracker is used to drive a single oscillator at a pitch several octaves above the original, which is then divided down by a bunch of CMOS chips to arrive at a broader variety of intervals than is usually derivable by flip-flop division, both above and below the original.  Not perfect, but capable of much more.  Again, this is a mono (one note at a time) process/circuit, which behaves best with a consistent input signal that doesn't move around too much.  It is also the basis of the earliest guitar synthesizers, that used the proportinal voltage to drive oscillator banks where the waveform could be altered, and where one was not restricted to integer division.  This method is generally more complex than rectification, or division, and is more flexible than the modulation method, though it tends to be fussier than that one about inputs.  If one is willing to make the circuitry more complex, though, it holds possibilities for extending the duration of the "other notes", which makes it quite different than the other methods listed so far.

Resampling  Anyone who has ever fooled around with a recording medium that permits playback at a rate different than the recording rate, knows that one can shift the pitch of the original in a deliberate way by recording at one speed and playing back at another, whether lower or higher.  When the E-H Memory Man first came out, one of the more interesting uses of it was in conjunction with the E-H Hot Foot.  This was a physical pot controller built into a large wah-like foot treadle so that the user could change the delay-time pot on the MM as they played.  If the user timed their playing and pot-rotation just right, notes and indeed whole chords, could be "bent" upwards and downwards, though precision in pitch change was hard to do, and relatively impossible to maintain.  Since one of the things about playing back a sampled sound at a rate faster or slower than the recording rate is that you either move too far ahead of or too far behind the original almost immediately, maintaining a constant pitch change by resampling can't work, in principle.  On the other hand, if several sampling circuits *trade off* resampling duties so that while one is catching up to the input signal, the other is either racing ahead or lagging behind, one can "stitch" together the complementary resampled copies of the original to produce what sounds like a pitch-shifted copy of the original.  This one really IS a copy of a polyphonic original, and does not require single notes.  It does, however, require a ridiculously complex circuit, is generally limited to a single pitch shift at a time unless one is willing to go absolutely NUTS with respect to circuit size, and introduces a certain amount of lag to the shifted signal.  It does, however, retain all the harmonics and dynamics of the original (subject to the limitations of the BBD itself).

Frequency shifting  I'm not going to pretend to comprehend this one, since others here comprehend it much more profoundly than I ever could, but in general it could be described as a "really, REALLY smart ring modulator".  What it DOES, though, is shift absolutely *everything* in your input signal over by a fixed amount, whether upwards or downwards, in a continuously variable fashion.  Is this the one you want?  Probably not.  The quirk of frequency shifting is that it simply adds or subtracts a constant from the input signal content.  So, not only is the fundamental increased by a certain number of hertz, but so are the harmonics.  Of course, that almost immediately puts the harmonics as non-multiples of the fundamental, which means that the more shift introduced, the more dissonant the output sounds.  I suppose that if one used it monophonically, and seriously filtered the input note to eliminate most of the harmonics, it could be used as a continuously variable pitch shifter, but this is an extremely complex circuit for accomplishing such a task.

There, I have to go pick up my kid from Sunday school.  This will get you started.  I'll check here later and will happily edit in your respective contributions to what is intended to be a reference document for newbies.

[Mark Hammer]

Duplicate deleted to reduce scrolling headaches.

[jmusser]

I think I'm going to cry :cry: ! This is what I've been hoping someone of learning would do for a long time! I have gotten fragments of this subject through my time here, but this is absolutely the first time I can remember it ever being condensed into one article and then articulated for the common bozo like myself. If no one else appreciates this, I sure do. One thing didn't see you hit on, is if different octaves tend to cancel each other out? The reason I'm asking this, is that a circuit like the Harmonic Generator doesn't allow for simultaneous blending of the various octaves. You switch to one or the other, but you can't blend any of them together at the same time, or in the amount you want to. If I was to build the HG, it sure would be nice to have that option. I would imagine that this all has to be done in parallel circuits, like Tim has done with his Octup Blender. This concept for some reason, has been a quest of mine to be accomplished. I don't have you article fully digested yet, but it's wonderful and a great teaching tool as usual. BTW, did you notice that the forum is starting to act like it's old self again? If you'll excuse, I have to go finish my "Mark Hammer Rocks" banner.....now, where did I put that glitter?

[rocker-D82]

I found 2 thing:

http://ampage.org/hammer/files/AnalogHarmonizer.ZIP
http://ampage.org/hammer/files/EMMHarmoGen.PDF

The first seems to be a very complex effect! It's good for me!!!!!!
The second it's simplest...

[RickL]

I've built the E&MM Harmony Generator and can confirm that it works. Build it if you want to prove to yourself that you can do it yourself (like me) otherwise buy one.  With careful adjustment of the controls and attention to your picking style it is possible to get something approaching parallel harmonies out of it but it ain't easy or predictable.

Even the pitch shifter in the Zoom 505 (which I've picked up for as little as $40) works better. Something like a Boss Harmonist pedal or a Digitech Vocalist (to name only a couple) will do that much better job and are capable of 'intellegent' pictch shifting for not a lot more money.

[jmusser]

RickL, I was wondering exactly how this thing works, not circuit wise, since it's fairly well explained, but to find out what it's capabilities are. I see 3 octaves up, 3 octaves down, a 3rd, 5th, Interval, and Unison. I thought i'd go ahead and ask this on the forum, instead of a PM, so that everyone one have an idea. I'm just going to try to guess what it does, and see how close I am. Do the octaves play all over the fingerboard or do only the down octaves do that, and the up octaves play from the 12th on up per usual? When you go to an octave, can you play the octave and say the 3rd at the same time? If we're in the key of G, would a B note be a 3rd, and a D note be a 5th? "Unison", I guess means all octaves at once, "Interval' I don't know. Are all these chipss available through Mouser, and also the transistor BC212L, or did you sub it for something else? That ought to be enough questions to keep you busy  I have probably 2/3 of the parts already. I don't expect it to sound like Boston harmonies (I think they used two guitarist any way), but it would be cool to have all the different octaves to play with. there is one last question, would I be able to mod it, so I could play 2 or 3 octaves of my choice at a time, or would it just end up making a tonal mess to tie them together?

[Transmogrifox]

I think I'll add my 2 cents to the last topic "Frequency Shifting" of Mark Hammer:

This sounds to me like something I have been wanting to try for a while as a kind of a unique chorus effect.

Here's the background (from my angle):

In my advanced telecommunications class from university we were introduced to the topics of modulation, yadda yadda...

When you amplitude modulate, for example, an audio signal onto a high frequency carrier (RF, say 900 kHz), and look at the spectral content, you see "side bands", which are all of the frequencies of the original signal as sums and differences with the RF carrier.  

You can then use a synchronous homodyne demodulator, and translate all of those frequencies back to exactly their place in the audio range by multiplying the modulated signal with a sine wave of exactly the same frequency of the carrier.  If the frequency of this "demodulating" frequency drifts from the carrier, then all the audio range frequencies together drift by this amount.  For small changes, this looks like a mild "detune", but for large changes it sounds noisey since the harmonic components retain the same frequency intervals so that they are no longer integer multiples of eachother.  The maximum frequency shift for voice such that it is possible to understand the speech is 50 Hz.   The ideal homodyne receiver has very little error between the carrier frequency and the local oscillator frequency.

I think my explanation was a bit convoluted, but the answer to that involves modulating the audio signal onto a high frequency carrier.  then demodulating it by frequency translation (multiply with another frequency to get a sum and difference of all frequencies) back to a pitch shifted state.

As one can see, this would be a viable method for an effect such as a detune or chorus type effect requiring very minor deviations.  As a harmonizer, this is probably not a good solution.

[Mark Hammer]

Thanks j, much appreciated.  The part I always loved best about teaching was looking up and seeing all the wheels turning in 200 minds simultaneously.  Always a pleasure to get a little taste of that now and then.

Something like the Harmony Generator CAN be rejigged to produce multiple notes, although clearly that isn't going to happen with the PCB layout accompanying the article.  If you've ever taking a peak at any of the old mid-1970's poly keyboards, you'll see that they pretty much all started with a single high frequency clock which was fed to a top-octave generator chip like the old (now long since out of production and hard as heck to find) Mostek 50240, which would produce a full octave from the divided-down HF clock.  Those 13 notes would be further divided by 2, 4, 8 to derive lower octaves.  Once all the notes were divided down, you could lay both arms out flat acorss the keyboard and produce all 37, 49. 61, 76, or even 88 notes at once.  The trick was having a high enough HF signal to start with so that it could be divided down several times.  Seems to me that one of those 4046-based trackers COULD be rejigged to produce a high-er frequency signal that tracked the fundamental, only several octaves up, and divide it down so that multiple notes could be generated at once and mixed together in a fashion not much different than an old keyboard.  It'd be complicated, though.

As for frequency shifting, Modcan (http://www.modcan.com/) sells a frequency shifter that Mike Irwin designed for them.  There are some samples posted at the site.  Very impressive, although I think you'll realize after hearing them that this is a device really more for the experimental electronic music types, and is not the sort of thing for doing Tom Scholz solos in a bar band.

[loki]

Thanks Mark,
your posts are always much detailed and explanatory  :wink:

just one thing i didn't understand... what's a BBD?

[puretube]

Bucket Brigade Device = analogue delay chip

http://www.geocities.com/scaprile/fxa.html#bbd
http://www.soundonsound.com/sos/mar04/articles/synthsecrets.htm
http://www.soundonsound.com/sos/jun04/articles/synthsecrets.htm
http://www.uaudio.com/webzine/2005/April/content/content2.html
http://www.du.edu/~etuttle/electron/elect39.htm

[jmusser]

Thanks Mark, that's good to know about the Harmony Generator. I would be making it on perf board anyway, so the PC board doesn't matter to me. The main thing I was fishing for there, was to know if a selected combination of octaves would get "confused" trying to handle a single note at the same time or not. I figured since there was a "Unison" feature, that it was probably doable with what ever octave configuration you'd come up with.

[Mark Hammer]

Think of it this way.  An octave down circuit like the Rocktave that lets you simultaneously mix in one and two octaves below doesn't really "know" or "care" that these are divide by 2 and divide by 4.  For all it cares, these could be divide by 37, divide by 119, and divide by 523.  As long as you can derive the suitable divided-down outputs in some isolatable manner, they can be mixed down.

The Super Octavizer that is posted at my site, uses two different methods to produce above-the-note and below-the-note tones, neither of which are really pitch tracking.  If one used a pitch-tracking/oscillator combo as in the E&MM HG, though, both notes above the actual fundamental and those below can be simultaneously produced divisions of a generated note that is, say, 3 or even 4 octaves above the original.  For instance (assuming the pitch tracking is reasonably accurate), feed in C2 and the oscillator produces C6.  Divide that by 4, 8 and 32, and you get C4, C3, and C1, or an octave below, and one and two octaves above.

The general principal to observe is that the more choice you want in available notes, the higher the original generated tone has to be, so that integer division of it can get you somewhere useful with respect to just intonation.

[jmusser]

One thing you touched on briefly was that it was very difficult to do another up octave with the rectification process. Would there be a way to use the 1st octave up "artifact" as a reference for something like a flip flop to achieve a perceived second octave up. Since it's quick and dirty as you say to accomplish this, there are several effects that use transformer rectification to get the artifact, but none I've seen that use division to achieve a second octave up. I'm thinking that something like the SOU would be a good candidate, since it has such an intense fuzz.