posted 23 May 2013 20:21            

MemoryMachine.kym

I encapsulated a sound! Hopefully it is attached. Nothing too spectacular, but it is based on the MemoryWriter prototype, but with a lowpass and high pass in the loop, and it is stereo! so I created some stereo offset controls and some gentle modulation. Is great for all your modulated delay ringmod type stereo effects all in one segment of your sound architecture!

Careful with the feedback, sometimes gets out of hand if is over 0.75 and you twang the filter, but perhaps you like that sort of thing.....

Sorry about the icon. Need to investigate photoshop.. and acquire some artistic merit.

Let me know what you think...
.
Happy Days

posted 24 May 2013 15:20            

MemoryMachine2.kym

Thanks for the sound. I hope you don't mind but I made an adjustment (attached) and given the two feedback files different names else they were interfering with each other.

I notice that you tend to get more effect (less original) when input is zero and fade is 0.1318. I don't know why this is, maybe you have some idea.

Thanks

Pete

P.S tidied up the fields a bit and through in some more presets labeled defaultxx.
P.P.S I notice it seems to be right heavy and I haven't worked this one out either.

posted 25 May 2013 13:22            

I notice in the feedback path you are using the Single Side Band Ring Modulator [SSBRM]. This version of the SSBRM is a one frequency band only type as any signals outside the band do not get shifted by 90 degs and become amplitude modulate instead of frequency shifting. i.e the bulk of the signal acts more like an original dual side band ring modulator.

There is another version of the SSBRM in Kyma, which uses the Hilbert transform. The Hilbert shifts all frequencies by 90 degs (at unity gain) and hence makes a true SSBRM over the whole audio spectrum.

The problem is that the Hilbert introduces about a thousand samples of latency (about 22ms at 44.1 Khz), which is not too bad for live work, but is no good in a feedback path as you cannot get short delays, and it stops being a flanger.

There is however an alternative, where by you start with the Hilbert transform on the input, and and feed the real and the imaginary into two similar feedback circuits (2 per channel). Then with a quadrature oscillator complex multiply in the feedback path, you get a true SSBRM happening inside the feedback path with zero latency, allowing short delays. The latency is still there but it's on the input, not in the feedback path.

If your interested, I'll try to make one.

posted 25 May 2013 18:56            

Interesting, I figured out the Input, it is a left over - 0.3 on the input.

Not sure what the LR oddness is about. I will go though the small details.

I am interested in ways to improve the patch, would like to learn more about the 'quadrature oscillator complex'.

I was wondering if anything interesting can be achieved by modulating the frequency field on the feedback sample playback. I shall try this.

Thanks

posted 22 June 2013 12:11            

MemoryMachine3.kym

As promised see attached the true single side band RM pitched shifter version of your sound.

I made a few tweaks as I went through though. As the SSBRM shifts in terms of frequency not pitch, the interval is not related to an interval just an arbitrary pitch so I've removed the logTwo to allow frequ up, frequ down and not frequ change. You may notice that if you shift the frequency down enough it will start to rise again. This is because the frequencies become negative, which sound just the same as positive frequencies.

I've also altered the gain structure of the feedback so that 1 is unity if the filters are opened. You can still put the feedback to greater than one to compensate for the filters attenuation.

Also I noticed that you had tapped the effected sound off the feedback loop just after you had added the original signal, which meant that the original signal was always mixed in with the effected, even if the the fade was set to (zero) all effected only. I have set the memory writers to silent, and tapped the sound off just after the filters (before the original gets mixed in). Now the fade introduces the original only when you start to turn it up.

The Hilberts in the module does mean that there is an overall latency of about 1000 samples with it not too bad for live use (not great), but the latency is not inside the feedback loop so should not effect the process.

I also got rid of the stereo imbalance. I didn't re encapsulate it as I thought you may want to see what was going on inside.

If you need to re-encapsulate it will take a bit of work but you don't have to redo all the work you have already done before, as there are techniques for copying the setting over. let me know if you need help with this. You can also copy and paste all the VCS setting as well but I've already done that bit.

Enjoy, and I hope it helps

posted 28 June 2013 21:44            

You made it so much better! I can't get over the differences. I have so much more to learn.

The stereo width and how it reacts to short delay times is really great.

the imaginary vs real feedback concept is fascinating, I am also struggling to conceptualise how the quad oscillators being enveloped from the incoming signal, but as I type this, It strikes me that this must be related to the cancelling process.

Ok enough rambling, I am going to study this further.

Thanks so much for taking the time. The layers are peeling away .

Many thanks

posted 28 June 2013 21:53            

Fascinating.

Thanks so much.

posted 01 July 2013 17:57            

When I first came up with the idea it did my head in as well as it seemed to be an oxymoron.

looking at it one way shows it defiantly will work but looking at it another way shows it defiantly couldn't work.

Looking at it one way.
Starting with the real and imaginary by putting it into two delay lines and making a complex multiply in the feedback path (crossfading between them evenly with one inverted) said that each time it went round the feedback path the phase would be rotated and form a pitch shifter going higher and higher with each iteration. It should work.

Looking at it another way.
The two delay lines are simply crossfading the signals between each other in the feedback paths so it should not be possible to produce frequencies that didn't exist in the original signals as it is all linear. Neither of the two input signals had the higher frequencies so why should simple delaying and cross fading them create new pitches. It simply shouldn't work

I had to try it out to find out that it did work after all, Though I still haven't got my head fully round it.

One thing you may notice is that the frequency shifter lets you up the feedback quite a lot more than unity without howl round, ( that is the nature of frequency shifters) but as soon as you slightly clip the sound, the real and imaginary signals stop being the real and imaginary signals and it stops pitch shifting and the howl round kicks in. You have to turn the feedback way down and let it flush out quite a but before you can get back to normality

If it doesn't clip the first rule above applies but as soon as it clips even slightly the second rule takes over completely and destroys the first rule. It becomes a completely different animal.

I hope this makes sense

Pete

posted 03 July 2013 09:51            

What an adventure.

I am currently processing speech samples with it!

Thanks