Rossum Panharmonium Resynthesizer

[starthief]

I was literally dreaming about this module just now (though in my dream a singer visited me to record a few songs for a collab EP, and she was annoyed by my synth corner's terrible acoustics, noisy aquarium and lack of decent mics ). And I woke up thinking about how cool this would be even just as a weird reverb. I might be a bit obsessed now.

[Dragonaut]

This is definitely one of the more interesting and powerful modules to come along in a while. It has a lot of different use cases which is great but also a problem for people who have trouble concentrating on one idea at a time. Like me.

[peripatitis]

This is definitely one of the more interesting and powerful modules to come along in a while. It has a lot of different use cases which is great but also a problem for people who have trouble concentrating on one idea at a time. Like me.

Usually the main issue is selecting the material to send thru. Since the analysis is coupled to the synthesis there are not many things you can experiment with.
In a system like Kyma, staying with the premade solutions, you get a myriad of ways to use the data of the analysis, and of course you can get into the nitty gritty and make your own.

Do they say though this is an fft?

[starthief]

Do they say though this is an fft?

Not directly, but he did say "the analyzer section splits the incoming audio into thousands of frequency bands" in the first video. And in the second, he demonstrated that there's some inherent latency in processing.

To me that screams FFT, rather than the bandpass bank that some people (including me) had speculated.

[Zymos]

I thought on that first video he was asked if it was FFT and said sort of.

Edit, around 3 minutes in. I think his distinction was that the analysis was like FFT, but the resynthesis part makes it different

[coolshirtdotjpg]

I thought on that first video he was asked if it was FFT and said sort of.

Edit, around 3 minutes in. I think his distinction was that the analysis was like FFT, but the resynthesis part makes it different

The reality is that a lot of DSP starts with FFT (it's what makes digital sound possible in some sense), my guess is that the resynthesis portion is the 32 waves they talked about, and the thousands of sine waves are the analyzed waves that were mentioned in the video. That would make sense given the fact that they offer waveshaping and other forms of manipulation. Having hundreds or even thousands of waves in the analysis section would make sense if you wanted to remain accurate while blurring the sounds together, or focusing on a different part of the spectrum.

This is very exciting to me (at least) because this is part of what I was trying to do by integrating modular synths with Kyma.

[starthief]

I think his distinction was that the analysis was like FFT, but the resynthesis part makes it different

Yeah, FFT analysis, but it feeds an oscillator control algorithm instead of monkeying with the data and then doing inverse FFT as many applications do. (I've written a few VST plugins that play with FFT, all pretty crude but sometimes useful.)

I suspect the algorithm in Panharmonium has a bit more sophistication than "pick the strongest bands" -- and I wonder if it also takes the selected oscillator waveform into account. (As in, if you've got a square wave it knows you're getting the third harmonic of any band for free.)

[mojopin]

Been looking for a cool all-in-one vocoder for the modular and this keeps popping into my head. Can't wait for some more demos!

[coolshirtdotjpg]

I suspect the algorithm in Panharmonium has a bit more sophistication than "pick the strongest bands" -- and I wonder if it also takes the selected oscillator waveform into account. (As in, if you've got a square wave it knows you're getting the third harmonic of any band for free.)

That wouldn't surprise me either, considering how accurate the results were.

[ishi]

Was only starting to understand Trident and now this... it just never ends

[cg_funk]

On the Sonic State video, around 10:30 the music processing demo starts.

It's pretty impressive what this thing can do.

Are there any other modules that can do live pitch transposition of real-time audio? The whole re-synthesis thing sounds like it will enable us to do absolutely crazy things with modular.

For example, imagine using this as a guitar input. You could have the modular going and changing key signatures and things, and use the V/oct input on the Panharmonium to key-change the guitar so you can just play in A the whole time.

This effect works great with keyboards, and I already love it. However, being able to do that with guitars would be unreal!!

[PietroC]

[quote="Dragonaut"](In Arnold Swartzeneger voice) IT IS NOT A VOCODAH![/quote]
hahahahaha

[bandenoire]

Preordered directly after watching that sonic state video. I’ve been wanting a Technos Acxel but obviously could never afford one and bought a kawai k5kr years ago hoping to get the sound diver program to output waves to the synth. This is just a brilliant execution with a great price tag.
It brings up an interesting paradox about sampling. Does copyright still hold true if this is a synth and not the original sound? Could one feed a scratchy break into it and get a fairly clean (albeit with possible digital artifacts) resynthesized version out of it that is technically a new sound.
Super stoked on it either way. Between assimal8or and the zplane, I think rossum might be my new favorite in design and execution.

[Funch]

Preordered directly after watching that sonic state video. I’ve been wanting a Technos Acxel but obviously could never afford one and bought a kawai k5kr years ago hoping to get the sound diver program to output waves to the synth. This is just a brilliant execution with a great price tag.
It brings up an interesting paradox about sampling. Does copyright still hold true if this is a synth and not the original sound? Could one feed a scratchy break into it and get a fairly clean (albeit with possible digital artifacts) resynthesized version out of it that is technically a new sound.
Super stoked on it either way. Between assimal8or and the zplane, I think rossum might be my new favorite in design and execution.

if you run another persons recording through a Reverb did you violate the copyright if you try to sell the result? Of course. Why would what you propose be any different?

[bandenoire]

Preordered directly after watching that sonic state video. I’ve been wanting a Technos Acxel but obviously could never afford one and bought a kawai k5kr years ago hoping to get the sound diver program to output waves to the synth. This is just a brilliant execution with a great price tag.
It brings up an interesting paradox about sampling. Does copyright still hold true if this is a synth and not the original sound? Could one feed a scratchy break into it and get a fairly clean (albeit with possible digital artifacts) resynthesized version out of it that is technically a new sound.
Super stoked on it either way. Between assimal8or and the zplane, I think rossum might be my new favorite in design and execution.

if you run another persons recording through a Reverb did you violate the copyright if you try to sell the result? Of course. Why would what you propose be any different?

Because this is something entirely different. Running a recording through a reverb or any effect is still the recording through that effect. however, if something run through this process is 100% wet, then technically nothing of the original signal exists. It’s an approximation of the original signal recreated by new waves.

[starthief]

I'm no lawyer, but ask yourself what happens if you perform someone else's copyrighted material without permission, and then think about this again.


Anyway, I just preordered because I'm excited about experimenting with it as both a means of synthesis and an effect (and I just sold a couple of pieces of gear I was letting go of and it suddenly was within budget).

[peripatitis]

Preordered directly after watching that sonic state video. I’ve been wanting a Technos Acxel but obviously could never afford one and bought a kawai k5kr years ago hoping to get the sound diver program to output waves to the synth. This is just a brilliant execution with a great price tag.
It brings up an interesting paradox about sampling. Does copyright still hold true if this is a synth and not the original sound? Could one feed a scratchy break into it and get a fairly clean (albeit with possible digital artifacts) resynthesized version out of it that is technically a new sound.
Super stoked on it either way. Between assimal8or and the zplane, I think rossum might be my new favorite in design and execution.

if you run another persons recording through a Reverb did you violate the copyright if you try to sell the result? Of course. Why would what you propose be any different?

Because this is something entirely different. Running a recording through a reverb or any effect is still the recording through that effect. however, if something run through this process is 100% wet, then technically nothing of the original signal exists. It’s an approximation of the original signal recreated by new waves.

In that case playing it back through another medium would be enough.
Obviously it isn't and in practice obviously it is.
Funny how journalists 20 year latter face the same problem..
Btw I can't imagine a digital processing method that does not involve sampling.

[bandenoire]

I'm no lawyer, but ask yourself what happens if you perform someone else's copyrighted material without permission, and then think about this again.


Anyway, I just preordered because I'm excited about experimenting with it as both a means of synthesis and an effect (and I just sold a couple of pieces of gear I was letting go of and it suddenly was within budget).

For sure, just a thought experiment. Have no plans to nick some famous bonham break or the like and try to get away with it.

[electricfence]

Who cares where a shitty blip came from originally ?

I demand that the shitty blips that I listen to have a distinguished provenance. That's why I only listen to shitty blips made with Cwejman oscillators run through Schippmann filters into a Natural Gate. (Sorry. I couldn't resist.)

In all seriousness, I'm really interested in trying one of these in person. I wonder if they'll be at Moogfest.

[bandenoire]

If they make an expander for this module I hope they consider a 33 channeled pitch and envelope follower that’s fed from the analyzers/fft output. Ridiculous, yes, but in theory could be amazing

[coolshirtdotjpg]

Preordered directly after watching that sonic state video. I’ve been wanting a Technos Acxel but obviously could never afford one and bought a kawai k5kr years ago hoping to get the sound diver program to output waves to the synth. This is just a brilliant execution with a great price tag.
It brings up an interesting paradox about sampling. Does copyright still hold true if this is a synth and not the original sound? Could one feed a scratchy break into it and get a fairly clean (albeit with possible digital artifacts) resynthesized version out of it that is technically a new sound.
Super stoked on it either way. Between assimal8or and the zplane, I think rossum might be my new favorite in design and execution.

if you run another persons recording through a Reverb did you violate the copyright if you try to sell the result? Of course. Why would what you propose be any different?

Because this is something entirely different. Running a recording through a reverb or any effect is still the recording through that effect. however, if something run through this process is 100% wet, then technically nothing of the original signal exists. It’s an approximation of the original signal recreated by new waves.

In that case playing it back through another medium would be enough.
Obviously it isn't and in practice obviously it is.
Funny how journalists 20 year latter face the same problem..
Btw I can't imagine a digital processing method that does not involve sampling.

It definitely sampled in order to do FFT (the digital to analog conversion is sampling in some sense), but those samples are not audible synth the audio is resynthesized, that is, of course, unless you crossfade it in. I have no idea what the legal ramifications of this are, but I suspect you'd have a lot of trouble explaining how something that sounds exactly the same isn't the same to a judge. I'm not saying it's impossible, but from a practical perspective, you're not going to win against a large corporation.

There's some weird confusion here (not saying you, but in this thread generally) about how digital audio works, what different types of synthesis mean, etc.) Let me lay it out, for those who don't entirely understand:

There are samplers, aka, devices that include an audio file that is at one point held in a buffer and recalled. Some can recall multiple files from storage (such as an SD card or harddrive), some can record directly into the buffer, some can do both.

There is granular synthesis, which takes a very small sample and uses this as the basis for synthesis. Some of these allow for repitching and this works essentially as it would in a sample, but because these small grains can loop, the create the perceived effect changing the pitch of a sound without affecting the length. It also possible to sample multiple grains. If you think that a sample and a grain are different aside from their application, you are misinformed.

There is no reason sampling has to be digital, the same is true for granular synthesis (see many of Xenakis' tape compositions for evidence of this), but it certainly helps facilitate this process.

Then there is vocoding where a sound is sent into X number (usually 16 or more) of band pass filters spaced such that there is very little overlap between each filter. Each filter output has it's own slew limiter (aka envelope follower), which then controls the amplitude envelope of X bandpass filters for another sound. Most commonly this is a synthesized sound. You could also use X number oscillators tuned to the center frequency of each bandpass filter, put send them into a VCA and control their amplitude to produce essentially the same effect.

Then there is this, which is a black box to us, but we can assume it samples the audio into a small buffer (since there are some time-based analysis, making it similar in some sense to convolution aka impulse response reverb aka phase vocoder) at some point in the chain Fast Fourier Transform is performed which allows them to determine the amplitude of the most important frequencies at a given moment, from this a variable number of these frequencies are chosen for resynthesis, thus preserving the character of the sound and allowing for modification. If this sounds like a digital version of a vocoder, that's because it is a similar concept. It's not directly comparable like granular synthesis and sampling (ever notice how the phonogene becomes a granular synth when you make the samples really small?), but it's obviously much closer to a vocoder than a traditional sampler since the "sample" is never heard when the effect is fully wet.

[bandenoire]

Preordered directly after watching that sonic state video. I’ve been wanting a Technos Acxel but obviously could never afford one and bought a kawai k5kr years ago hoping to get the sound diver program to output waves to the synth. This is just a brilliant execution with a great price tag.
It brings up an interesting paradox about sampling. Does copyright still hold true if this is a synth and not the original sound? Could one feed a scratchy break into it and get a fairly clean (albeit with possible digital artifacts) resynthesized version out of it that is technically a new sound.
Super stoked on it either way. Between assimal8or and the zplane, I think rossum might be my new favorite in design and execution.

if you run another persons recording through a Reverb did you violate the copyright if you try to sell the result? Of course. Why would what you propose be any different?

Because this is something entirely different. Running a recording through a reverb or any effect is still the recording through that effect. however, if something run through this process is 100% wet, then technically nothing of the original signal exists. It’s an approximation of the original signal recreated by new waves.

In that case playing it back through another medium would be enough.
Obviously it isn't and in practice obviously it is.
Funny how journalists 20 year latter face the same problem..
Btw I can't imagine a digital processing method that does not involve sampling.

It definitely sampled in order to do FFT (the digital to analog conversion is sampling in some sense), but those samples are not audible synth the audio is resynthesized, that is, of course, unless you crossfade it in. I have no idea what the legal ramifications of this are, but I suspect you'd have a lot of trouble explaining how something that sounds exactly the same isn't the same to a judge. I'm not saying it's impossible, but from a practical perspective, you're not going to win against a large corporation.

There's some weird confusion here (not saying you, but in this thread generally) about how digital audio works, what different types of synthesis mean, etc.) Let me lay it out, for those who don't entirely understand:

There are samplers, aka, devices that include an audio file that is at one point held in a buffer and recalled. Some can recall multiple files from storage (such as an SD card or harddrive), some can record directly into the buffer, some can do both.

There is granular synthesis, which takes a very small sample and uses this as the basis for synthesis. Some of these allow for repitching and this works essentially as it would in a sample, but because these small grains can loop, the create the perceived effect changing the pitch of a sound without affecting the length. It also possible to sample multiple grains. If you think that a sample and a grain are different aside from their application, you are misinformed.

There is no reason sampling has to be digital, the same is true for granular synthesis (see many of Xenakis' tape compositions for evidence of this), but it certainly helps facilitate this process.

Then there is vocoding where a sound is sent into X number (usually 16 or more) of band pass filters spaced such that there is very little overlap between each filter. Each filter output has it's own slew limited (aka envelope follower), which then controls the amplitude envelope of X bandpass filters for another sound. Most commonly this is a synthesized sound. You could also use X number oscillators tuned to the center frequency of each bandpass filter, put send them into a VCA and control their amplitude to produce essentially the same effect.

Then there is this, which is a black box to us, but we can assume it samples the audio into a small buffer (since there are some time-based analysis, making it similar in some sense to convolution aka impulse response reverb aka phase vocoder) at some point in the chain Fast Fourier Transform is performed which allows them to determine the amplitude of the most important frequencies at a given moment, from this a variable number of these frequencies are chosen for resynthesis, thus preserve the character of the sound and allowing for modification. If this sounds like a digital version of a vocoder, that's because it is a similar concept. It's not directly comparable like granular synthesis and sample (ever notice how the phonogene becomes a granular synth when you make the samples really small?), but it's obviously much closer to a vocoder than a traditional sampler since the "sample" is never heard when the effect is fully wet.

if it's anything like the technos axcel, then what you are hearing is 100% multiple waveforms (in this case 33) with envelopes applied approximating the sound coming in through fft processing. if i understand it correctly, we don't hear any of the fft processing. it only divides it to what is essentially pitch and envelope trackers which then spits the info out to vcos. what's beautiful about this module is that there are multiple oscillators vs. strictly sine

[coolshirtdotjpg]

Preordered directly after watching that sonic state video. I’ve been wanting a Technos Acxel but obviously could never afford one and bought a kawai k5kr years ago hoping to get the sound diver program to output waves to the synth. This is just a brilliant execution with a great price tag.
It brings up an interesting paradox about sampling. Does copyright still hold true if this is a synth and not the original sound? Could one feed a scratchy break into it and get a fairly clean (albeit with possible digital artifacts) resynthesized version out of it that is technically a new sound.
Super stoked on it either way. Between assimal8or and the zplane, I think rossum might be my new favorite in design and execution.

if you run another persons recording through a Reverb did you violate the copyright if you try to sell the result? Of course. Why would what you propose be any different?

Because this is something entirely different. Running a recording through a reverb or any effect is still the recording through that effect. however, if something run through this process is 100% wet, then technically nothing of the original signal exists. It’s an approximation of the original signal recreated by new waves.

In that case playing it back through another medium would be enough.
Obviously it isn't and in practice obviously it is.
Funny how journalists 20 year latter face the same problem..
Btw I can't imagine a digital processing method that does not involve sampling.

It definitely sampled in order to do FFT (the digital to analog conversion is sampling in some sense), but those samples are not audible synth the audio is resynthesized, that is, of course, unless you crossfade it in. I have no idea what the legal ramifications of this are, but I suspect you'd have a lot of trouble explaining how something that sounds exactly the same isn't the same to a judge. I'm not saying it's impossible, but from a practical perspective, you're not going to win against a large corporation.

There's some weird confusion here (not saying you, but in this thread generally) about how digital audio works, what different types of synthesis mean, etc.) Let me lay it out, for those who don't entirely understand:

There are samplers, aka, devices that include an audio file that is at one point held in a buffer and recalled. Some can recall multiple files from storage (such as an SD card or harddrive), some can record directly into the buffer, some can do both.

There is granular synthesis, which takes a very small sample and uses this as the basis for synthesis. Some of these allow for repitching and this works essentially as it would in a sample, but because these small grains can loop, the create the perceived effect changing the pitch of a sound without affecting the length. It also possible to sample multiple grains. If you think that a sample and a grain are different aside from their application, you are misinformed.

There is no reason sampling has to be digital, the same is true for granular synthesis (see many of Xenakis' tape compositions for evidence of this), but it certainly helps facilitate this process.

Then there is vocoding where a sound is sent into X number (usually 16 or more) of band pass filters spaced such that there is very little overlap between each filter. Each filter output has it's own slew limited (aka envelope follower), which then controls the amplitude envelope of X bandpass filters for another sound. Most commonly this is a synthesized sound. You could also use X number oscillators tuned to the center frequency of each bandpass filter, put send them into a VCA and control their amplitude to produce essentially the same effect.

Then there is this, which is a black box to us, but we can assume it samples the audio into a small buffer (since there are some time-based analysis, making it similar in some sense to convolution aka impulse response reverb aka phase vocoder) at some point in the chain Fast Fourier Transform is performed which allows them to determine the amplitude of the most important frequencies at a given moment, from this a variable number of these frequencies are chosen for resynthesis, thus preserve the character of the sound and allowing for modification. If this sounds like a digital version of a vocoder, that's because it is a similar concept. It's not directly comparable like granular synthesis and sample (ever notice how the phonogene becomes a granular synth when you make the samples really small?), but it's obviously much closer to a vocoder than a traditional sampler since the "sample" is never heard when the effect is fully wet.

if it's anything like the technos axcel, then what you are hearing is 100% multiple waveforms (in this case 33) with envelopes applied approximating the sound coming in through fft processing. if i understand it correctly, we don't hear any of the fft processing. it only divides it to what is essentially pitch and envelope trackers which then spits the info out to vcos. what's beautiful about this module is that there are multiple oscillators vs. strictly sine

That is correct, insofar as that's what the manufacturer has implied (and how it sounds). Having wave-shaping rather than sine waves make it very interesting, also the time-domain processing is very unique in euro.

[cg_funk]

Well, then this module isn't strictly a vocoder either... Otherwise they would have just gone with a bank of many sine waves. Instead, the intent of this module is to apparently sculpt and create new sounds derived from inputs, so it uses saws and squares to build up something with the same spectrum as the input... Although I am not exactly sure what that means. To my ears the demos on human speech sounded a lot like low-quality MP3 compression.

Also, after listening to the demo like 50x times, I am hearing a whole lot of high-pitched overtones and clangs that I would honestly need to filter out. Is that caused by the aliasing? Originally I thought it was just noise from the auditorium. I really need to hear some more demos of this thing, when is divkid gonna get a test-unit??

One of the things that Noise Engineering has done with their digital stuff is to change the sampling rate dynamically with the fundamental pitch, which means that the harmonics and overtones are always in-key. Now I can really start to understand why this is really important! I don't know how this module handles aliasing, so it's really up to the demos for me.

[coolshirtdotjpg]

Well, then this module isn't strictly a vocoder either... Otherwise they would have just gone with a bank of many sine waves. Instead, the intent of this module is to apparently sculpt and create new sounds derived from inputs, so it uses saws and squares to build up something with the same spectrum as the input... Although I am not exactly sure what that means. To my ears the demos on human speech sounded a lot like low-quality MP3 compression.

They don't choose your waveshapes for you; as far as I can tell from the front panel, you can shape them yourself. All that means is, essentially that you can add/remove harmonics/partials to the waveforms they resynthesizes, I'm guessing they start with sine waves, but it doesn't really matter.


Yes, it does sound a bit like a low quality MP3, because MP3 compression removes the waveforms that aren't loud at a particular moment in time. In theory this works perfectly because of so called "masking" phenomena. Jonathan Sterne's book, MP3: The Meaning of a Format is a useful history of compression/perceptual coding, if you want to know a bit more about that.