Odd and even harmonics

MikeRaphone · Post by **MikeRaphone** » Wed Jul 07, 2004 7:31 am

Hello,
does anybody have an idea how to filter out either odd or even harmonics from music material. I have Pulsar II stock plugins and MixMaster and SynthSampler packs.
Thank you

borg · Post by **borg** » Wed Jul 07, 2004 8:51 am

geeh... wouldn't know really, but i think your best bet would be asking in modular section or adern forum.

others will probably be more helpful. only thing i know is, that odd harmonics are caracteristic for square waves, even harmonics are caracteristic for a sawtooth (with a lack of the fundamental). maybe this will make other peoples lamp burning.
i'm not sure, and can't investigate myself at the moment, but i guess flexor's 'shapers' are built for this purpose... or am i wrong?

alfonso · Post by **alfonso** » Wed Jul 07, 2004 9:07 am

Flexor shapers are really made to treat the saw oscillators mainly...

Filtering harmonics of a complex sound seems not the easiest task to me...maybe those processing tools made to increase sound quality like spectral enhancers etc. could help someway, but nothing exactly comes to my mind...

[ This Message was edited by: alfonso on 2004-07-07 10:07 ]

BingoTheClowno · Post by **BingoTheClowno** » Wed Jul 07, 2004 10:46 am

I think the odd frequencies are 1/3, 1/5, 1/7 etc of the fundamental frequency while the even frequencies are 1/2, 1/4, 1/6 of the fundamental frequency. I guess you could use multiple notch filters to filter those frequencies (and some adiacent frequencies too) out.

alfonso · Post by **alfonso** » Wed Jul 07, 2004 11:13 am

On 2004-07-07 11:46, BingoTheClowno wrote:
I think the odd frequencies are 1/3, 1/5, 1/7 etc of the fundamental frequency while the even frequencies are 1/2, 1/4, 1/6 of the fundamental frequency. I guess you could use multiple notch filters to filter those frequencies (and some adiacent frequencies too) out.

This should be theorically possible on a single constant fundamental sound...but as I said a complex sound is another thing...nothing that can be achieved by static freq. processing...there should be a constant analysis of the sound and a huge processing changing in time as well...

What should be this harmonics filtering for?
changing distortion? smoothen a sound?
Any of these specifical tasks could have a workaround...

borg · Post by **borg** » Wed Jul 07, 2004 11:17 am

i'm just remembering things i read in books here... although the nodge filter technique sounds sensible, it might be very hard to do in real live situation. there's a lot of frequencies! a square for instance has an infinite number of odd harmonics (theoretically).

it might help to tell us what you're trying to achieve. looking forward to what the knowing folks have to say on this.

borg · Post by **borg** » Wed Jul 07, 2004 11:19 am

http://www.auditory.org/mhonarc/2002/msg00258.html

haven't checked it yet...

nprime · Post by **nprime** » Wed Jul 07, 2004 11:45 am

Slight correction:

Harmonics are are not fractions of the fundamental frequency, they are whole number multiples of the root frequnecy.

i.e. odd being 3,5,7,9 etc. times the fundamental

even being 2,4,6,8 etc. times the root.

I can't imagine how someone could write a program that could do what you want in real time on constantly shifting material.

but we do have some genius' here...

[ This Message was edited by: nprime on 2004-07-07 12:47 ]

ChrisWerner · Post by **ChrisWerner** » Wed Jul 07, 2004 1:33 pm

Hey Borg,
I read it and his idea to solve this problem with a comb filter, sounds interesting.
Do you know the Spekral Delay? It can work as a comb too, has 1024 bands that can be muted moved or whatever. Maybe this would be an entry for the solution?
Just a thought.

[ This Message was edited by: ChrisWerner on 2004-07-07 14:36 ]

BingoTheClowno · Post by **BingoTheClowno** » Wed Jul 07, 2004 1:41 pm

Yes, but, like nprime and alfonso said, how would one cue in onto the shifting fundamental?

astroman · Post by **astroman** » Wed Jul 07, 2004 1:51 pm

imho a formant filter has that moving/fixed characteristic. Some docs of the Yamaha FS1 might be revealing.

cheers, Tom

BingoTheClowno · Post by **BingoTheClowno** » Wed Jul 07, 2004 2:43 pm

Formants are a little different, they are dependent on the characteristics of the instrument:

http://www2.sfu.ca/sonic-studio/handbook/Formant.html

Post by **garyb** » Wed Jul 07, 2004 3:07 pm

blazesboylan · Post by **blazesboylan** » Wed Jul 07, 2004 3:22 pm

Fast Fourier Transforms are generally pretty slow. But judging by the specs people list for their machines these days... One might be able to find fundamental by taking the largest amplitude band. Of course the harmonics could be stronger than the fundamental, but maybe it's a starting point -- ?!?

Just a thought from a non-plugin-designing cheap seats commentator.

Johann

BingoTheClowno · Post by **BingoTheClowno** » Wed Jul 07, 2004 4:19 pm

On 2004-07-07 16:22, blazesboylan wrote:
Fast Fourier Transforms are generally pretty slow. But judging by the specs people list for their machines these days... One might be able to find fundamental by taking the largest amplitude band. Of course the harmonics could be stronger than the fundamental, but maybe it's a starting point -- ?!?

Just a thought from a non-plugin-designing cheap seats commentator.

Johann

The Spectral Delay from NI is processing (amazingly!) the data in the frequency domain and it was designed a couple of years ago.
http://www.native-instruments.com/index ... aldelay_us

nprime · Post by **nprime** » Wed Jul 07, 2004 11:53 pm

this is such a weird coincidence.

I have been thinking about a new kind of eq, one which bases itself upon actual notes, not just even number frequencies, something much more purposeful than a 1/3 octave eq.

it's a twelve note per octave eq, the centre frequencies are actually the frequencies of notes, as in A being 440. Drastic octave dependant band pass filters.

I thought it would be useful if each band would also have a Synthesizer-like filter and be able to self-oscillate. Now if you let each band enter slight oscillation and invert the phase of the oscillation you could cut a specific frequency and it's harmonic.

I still don't know how you would do this with complex material...

MikeRaphone · Post by **MikeRaphone** » Sun Jul 11, 2004 2:45 am

Well here's what i had in mind- a single monophonic source. All else should be far too complicated. But i was interested if someone who knows a lot about acoustics would come up with something. I imagine if you succeeded in this, one would be able to change timbre of the source in a very peculiar way. Thanks all for your input...

astroman · Post by **astroman** » Sun Jul 11, 2004 6:54 am

On 2004-07-07 08:31, MikeRaphone wrote:
...does anybody have an idea how to filter out either odd or even harmonics from music material...

you got us on the wrong trail with your original question

On a single mono source it's actually pretty simple: like the filtering Alfono describes and you control filter frequency by keytracking. Imho...

cheers, Tom

kensuguro · Post by **kensuguro** » Tue Jul 13, 2004 10:24 am

interesting idea. I wonder what the effect will sound like tho. Perhaps we can do a test with a simple static sound, just to check out if the effect is worth implementing. Automatically tracking the pitch would require flexor tho.

The filtering part may be more complex than a bunch of peak/knotch filters tuned in harmonic relationship. These peack/knotch filters are only so steep, that as the harmonics get closer together in the higher range, the filters would start to overlap. I'd probably prefer spectral processing or an stft analysis/resynth combination for such task. Still you'd only have so much accuracy in the high range.

I wonder if there's any way you can make a counter in mod (with or without flexor). Then we can overdrive a monophonic sound, change it into a near perfect square wave, and actually count the frequency, or predict it using a short time window. Sounds a bit brute, but quite funky if possible. Atleast we'll have a pitch tracking mechanism.

astroman · Post by **astroman** » Tue Jul 13, 2004 1:36 pm

another suggestion:
given the base note is known (I assume that from the 'single monophonic sound', a sample or whatever) one could generate the desired spectrum of harmonics (odd or even) and phase reverse it so it extinguishes the original (like the SPL de-esser).
For the cases in which the original doesn't contain 'enough level' of that part some rectifying would be needed to not add an artificial extra.
Reminds a bit on how the Kawai K5000 deals with sound, but I never got really into it.
Looks like some Modular tweaking applies...

cheers, Tom