posted 26 February 2006 23:02            

Swarm.kym

Thanks!

H

posted 28 February 2006 20:41            

RescaledRangeQueenSwarm.kym

I multiplied the FrequencyTracker by SignalProcessor halfSampleRate so that !Queen is a value in the range of hertz (rather than 0 to 1). Then I removed the scaling that you had in the Oscillators so they follow the actual frequency value.

posted 02 March 2006 23:43            

MotionGeneratesValue.kym

The next step is to have resynthesized copies of the violin following the live one. I will explain my reasoning here a little, and maybe you can tell me if I'm on the right track.

The frequency of the resynth is controlled by the live input, and the only way I can affect it is through the FreqScale parameter. The pitch of the synthetic violin should be the same as the live one, plus the deviation caused by the swarm behavior when there is motion (I'm thinking of continuous glissandi in this case). So the FreqScale should be a result of the amount of deviation taking place, and it is this parameter that the swarms should follow.

This is similar to a simple sound I made to use with the tablet, so that there is only sound when I move the pen (or fader). I used the direct sound minus a delayed copy of itself to calculate this deviation. This works fine with a control signal, but I'm not sure how I should implement this with the pitch from the violin. I suppose I should be able to subtract the result of the freq tracker from itself delayed, but this doesn't seem to work.

I've uploaded a copy of the sound I made, with the FreScale resulting from the motion of Fader1, and a copy with the freq tracker added (for the latter case I am using a copy of Mikka by Xenakis as source, as it has the type of behavior that I want from the live violin). Scaling the difference of the freq tracker improves the behavior, but it's still not accurate (and quite honestly I don't know what the scale for this should be or why I should need to make it * 20).

Advise on how to improve this would be greatly appreciated. Thank you.

posted 03 March 2006 08:23            

Do you know where we can get a copy of Mikka by Xenakis or part there of to try some experiments with.

I am assuming that the Mikka file is a non polyphonic file other wise the pitch tracker wouldn't work. I wouldn't have used the analysis and resynth, unless the intention was to do something more funky than pitch shifting a non poly sound. Instead I would have used Frequency scaling as is used in the auto cord prototype. This way the sound is less deconstructed and has less artifacts that would need to be dealt with.

If I understand correctly what you are trying to produce, is a second (or more) violin that follows the pitch of the first but with the overshoots that swarm provides.

You already have the formula that converts the pitch detectors output into hz. So to get the pitch shifter ratio required by the frequency scaler you would need to divide the output of the swarm formula (which is acting on the FrequencyTrackers output converted to hz) by the FrequencyTrackers hz converted output ( before it goes into the swarm formula). This formula could be place in the FrequencyScale field in the FrequencyScale module, and the same frequency tracker would also by placed in the FreqTracker input.

You would not need the SoundToGlobalController unless you needed to see a visual moving fader on the VCS.

There would also be no need for a delay unless you wanted to delay the signal going into the Frequency shifter to match the delay inherent in the pitch detector. This would be quite a short delay well below 0.5 secs.

Of cause if you wanted three following violins you would need three copies of the FrequencyScaler but they could all share the same Pitch detector.

Hope this helps

Pete

posted 03 March 2006 08:36            

Instead of !Queen being the frequency of the violin, !Queen should be the *ratio* of the current violin frequency to the previous frequency (through a delay). In the SoundToGlobalController that generates !Queen, you would have something like:

[freqTrak] L / [delayedFreqTrak] L

Then, in the FreqScale parameters of your resyntheses, you could use !Result0 through 2.

For example, say the violin jumps up an octave in the delay time. Then value of !Queen would change to 2 and all the !Results would try to follow it. So the resyntheses would go up. But now the violin is staying on the same pitch for a while so its ratio becomes 1 and the swarms eventually settle on a FreqScale of 1.

posted 03 March 2006 09:40            

(( [FrequencyTrackerViolin] L )swarmFollowFromPosition: 0 velocity: !Vel acceleration:
!Acc friction: !Damping) / [FrequencyTrackerViolin] L

No Delay needed.

posted 05 March 2006 00:21            

Thanks for your reply. Yes, Mikka is non-polyphonic. It consists of a single violin line in brownian motion through continuous glissandi. If you give me your address I can email you an mp3 excerpt of the piece. But as I said, it is just a violin line glissing up and down at changing speeds and with occasional stops.

I use a resynth because that is what I've had most success with when doing pitchshifting for intrsuments. I tend to find the results more real-sounding than with the FrequencyScale.

I am using the SoundToGlobalController for visual feedback at this stage.

I tried using the ratio directly without the delay, and it works, but I seem to be getting some artifacts, and the pitch never seems to settle into a unison when there is no motion. Interestingly, the small shift that's always present give more separation to the two signals. I will have to play with it more and try to understand better what is going on.

Thanks for your help!

posted 05 March 2006 13:34            

You can email me on pete@bantusound.com .

The reason I ask for the source sound is because all source sounds are different even if we hear them as similar. For example we may hear a mono phonic sound but if it has some reverb the computer will see it as a polyphonic sound (due to the previous decaying notes. Also some signals have harmonics that are loader than the fundamental (mobile phone ring tones are a good example of this). Also the rate of pitch change can sometimes confuse the frequency tracker if the settings are wrong although we as a listener can hear nothing odd at all. Also the scratches at the beginning of violin notes can confuse things even though we don't even notice them. Every sound source can pose different problems.