A small "Waveterm" for Terratec Komplexer (and other products)

[Phil Sonic Academy]

Hi, Loving AudioTerm.

Im having problems with some stuff im trying to do. I have this sample

https://www.dropbox.com/s/ferejnk8mjjww ... 24_001.wav

I resampled it and messed around to get it 1024 cycles x 33 (so it would line up better etc.)

Is there anyway to force it to scan at a lower Octave? when i do it at standard freqs its pitch too high and i lose lots of the top end harmonics.

Thanks!

[blacktomcat666]

The file seems to be damaged somehow, so I can't test it at the moment.

The analyzer can be tuned to +/- 1 octave from the ideal root frequency of 172,265625 Hz using the tuning keyboard on the "Analysis options II" page for the coarse value and the +/- 5 Hz buttons for fine tuning. The range can't be extended, because on lower tunings the analysis window will have a negative effect on the analysis result. Using "overlap-add" on synthesis would cancel out window artifacts, but this isn't possible when creating wavetables because the resynthesized slots don't overlap.

This could only be solved by increasing the analysis window size, but this would also lead to a worse time resolution.

[Phil Sonic Academy]

Thanks for the response.

This could only be solved by increasing the analysis window size, but this would also lead to a worse time resolution.

Is this only in relation to certain types of Wavetables?

If it was for just exporting to wav would increasing the analysis window size be a possibility?

not sure what is up with that previous file. should be 44100 32bit... here is another.

https://www.dropbox.com/s/r7hyyqa3xqnene2/Scan%20p2.wav

[blacktomcat666]

Ok, I managed to open the files. The links included "www" instead of "dl", so they lead to the download page and not to the file itself.

The problem is that these wavetables have a frequency range that cannot be covered with 64 partials only. You'd experience the same limitations when converting these files for Waldorf Blofeld, Microwave I / II or PPG WaveGenerator: These products use slot sizes of 128 samples, which is 63 usable partials in frequency domain.

Further the Komplexer engine AT was intentionally designed for doesn't support arbitrary or unsymmetrical waveforms (it generates no cosine partials). Everything is recomputed with bipolar sine waves. This also leads to a "softer" sound and cannot be changed that easily, because it would require a complete new programming of the KTERM function and would also result in losing compatibilty to half of the supported products.

[blacktomcat666]

Still working on the XFORM - function for VAZ Modular wavetable support and testing transforming algorithms:

https://dl.dropbox.com/s/ipkskb3uxhob8x ... _Ghost.mp3

[blacktomcat666]

Some more wavetable examples (used an alternate algorithm for analysis):

http://dl.dropbox.com/s/mpnurujr3a1gocw ... tables.mp3

Here are the used files (~240 MB) for testing purposes:

http://dl.dropbox.com/s/bjw7swh8se4ysqk ... tables.rar

Don't forget to enable the "crossfade" option when loading in VAZ Modular.

The wavetables come directly from the analysis stage. You may have to put some filters into the signal path to keep control over the top and bottom end of the spectrum.

[blacktomcat666]

Some first impressions of the XFORM function:

[blacktomcat666]

I put v2.18 online, including a first basic implementation of the new XFORM function for VAZ Modular wavetables. Don't expect too much, but you can already resynthesize samples (up to 1.5 minutes mono or stereo) and store the results. Most of the planned transform/editing functions are still under construction, but you can already edit the amplitude profile and the slot order/direction.

[Gone soft]

I put v2.18 online, including a first basic implementation of the new XFORM function for VAZ Modular wavetables. Don't expect too much, but you can already resynthesize samples (up to 1.5 minutes mono or stereo) and store the results. Most of the planned transform/editing functions are still under construction, but you can already edit the amplitude profile and the slot order/direction.

Allright, been looking forward to playing around with this. To my ears the result of my first effort even sounds surprisingly smooth with FM without the artifacts I'd expect from such "abuse". I'm not 100 % sure what the peak boost technically does, I'm just noticing that overdoing it seems to produce a kind of artifacts I've heard in a lot of experimental music that is probably typical of certain FFT based processing intentionally overdone.

[blacktomcat666]

Hm.. "abuse" ... it's possible, so I did it. You could also put a loop on a normal audio file and put it into VAZ for scanning... this is what I would call "abuse" (but you get some interesting phase artifacts like detuning).

XFORM uses a set of 65536 x 33 x 2 (the latter because of stereo) magnitude partials. The original phases of the source file are only used to improve the stereo reconstruction. For resynthesis, a fixed set of random phases is used. These phases are used for both channels and for all slots. This is what makes the sound scanable.

I had to imlement a fixed low bandstop around 50Hz. Otherwise you would blow your speakers ( the lowest possible frequency would be around 0.3 Hz ). There will be filters and much more possibilities for adjustments in future.

The PEAK BOOST parameter: The peaks of the FFT are detected in 6 stages. The first stage holds all data (PEAK BOST 0%). Every stage up the peaks of the previous stage are interpolated and a new detection run is done. This leads to that only the strongest peak survive in stage 6.

PEAK BOOST is useful when the sound is too noisy or has to much "wobbling" - for example when anlyzing single notes or very short samples.

SMOOTHNESS blends to a smoothed version and reduces interference between the slots. The artifacts you hear are the result of the wavetable approach: XFORM uses fixed frequencies, which is required for storing the data in a wavetable file. It would be better to have a realtime synth with 65536 partials and pitch envelopes for each of them, but this wouldn't run with <10% CPU like a small VAZ patch. And it would take age to compute.

When analyzig a stereo sound, the PEAK STEREO LINK parameter appears: This can be used to widen the spectrum without phase problems. The parameter blends between separate peak detection for each channel and a version with holds peaks of the sum. This parameter requires that PEAK BOOST is > 0.

[yemski]

Kind of a noob set of questions here,since I've never done this sort of thing before.

At the moment I'm looking to produce a set of wavetables for Dune 2.
The included ones are cool and all but there are a number of single cycles I have created and wish to use as wavetables.
I've got a number of synths and utilities that I have used to create single cycle waves over the years. From what I understood of the videos referring to Dune I should be able to put an arbitrary set of single cycle waves into the directory Audio-term refers to and then assign each wave to a slot. Having done that I can then use Audio-term to resynthesize them and have it produce the necessary interpolation or crossfading between each wave to produce a wave sequence,which I can import into Dune. Am I correct or am I missing something?

Also,what length does each single cycle have to be?

[Gone soft]

Kind of a noob set of questions here,since I've never done this sort of thing before.

At the moment I'm looking to produce a set of wavetables for Dune 2.
The included ones are cool and all but there are a number of single cycles I have created and wish to use as wavetables.
I've got a number of synths and utilities that I have used to create single cycle waves over the years. From what I understood of the videos referring to Dune I should be able to put an arbitrary set of single cycle waves into the directory Audio-term refers to and then assign each wave to a slot. Having done that I can then use Audio-term to resynthesize them and have it produce the necessary interpolation or crossfading between each wave to produce a wave sequence,which I can import into Dune. Am I correct or am I missing something?

Also,what length does each single cycle have to be?

You may loose some high frequency content, and possibly some other details as well, as they will be resynthesized as 64-partial waves and mirrored ones at that. Should that be technical mumbo jumbo to you can just use the chance to auditon them and you'll hear if you think you're loosing something important or not.

I believe there's a video covering assembling single cycles into a wavetable in the "ASSEM" mode in K-TERM, and if I remember correctly it even covers saving the resulting wavetable in Dune 2 format. I'll have a look for it, and if I find it I'll edit the YouTube video in here.

You will want to venture into the TEF page after importing some single cycles into the "ASSEM" page and try both algorithms in the "IPOL" page to see which one fits best. I think that should all be covered in that video, which I'll hopefully find for you.

EDIT: found it!

The length for single cycles recognized by Audio-Term is covered in the pdf, "Audio-Term.pdf", that should be in the same folder as "Audio-Term.exe". Scroll down to where it says "Input file formats for samples and single cycle waves in *.WAV format".

And blacktomcat666, thanks for the last couple of updates and your continued dedication to this wonderful creation of yours.

[yemski]

Removed

[blacktomcat666]

v2.25 is online. You can export your KTERM wavetables directly to Xfer Serum WT format (without using the resynthesis). Here are some example conversions (mostly resynthesized stuff):

https://dl.dropbox.com/s/q80lagb2bpauti ... m.rar?dl=0

[aMUSEd]

Still Windows only?