Sampling and Resynthesis of real glass

[IngoWeidner]

NOTE: I'll first discuss the idea and the procedure and examples + downloads will follow in following posts.

I was always a fan of "Glass" and "Bell" like sounds and Wavetable, FM based and additive synths seemed to be a great choice for those (somehow you could combine all 3 of those in Icarus but that's another story...).
In the past i had also doe attempts to create such waveforms from scratch and got some nice results alraedy.

Anyway yesterday while working with Icarus it striked me that it is finally time to get at least such Glass like sounds from the "real thing".
From my parents (who passed away in 2005 and 2015) i inherited a bunch of different vessels in differnt sizes and shapes that are made from leaded crystal. Most of those semed to be hand made in the former German Democratic Republic, before Germany was re-united in 1990.

I selected a few differnt ones with differnt shapes, sizes and thickness of the glass and what i needed was a microphone. I am no singer but i have a microphone from the vocoder in my Novation UltraNova that fits into the microphone connector at a mono channel of my small anlog mixer.
For recording and editing of the resulting sample (like many times in the past for doing samples from hardware synths) i used Steinberg WaveLab Elements 7. (much cheaper than the full WaveLab version).

Based on the vessels i selected i used different ways of striking them to produce sound, filed 2 of them wit h water to change the pitch and used differnt shapes and thickness of the glass.


The resulting recordings were cut/edited to be used for Resynthesis in Icarus then. What i got from the recordings were sounds with decreasing loudness and changing tone and all seemed to end at a pure Sine wave at some point.


It was possible to do Resynthesis of those recordings in Icrarus using the "Create wavetable: do Resynthesi from Noisy sounds" feature (of cousre you coudl also try other Resynthesis modes). The imported wavetables were cut where necessary, especially as there were a lot of resuting Sine waveforms taht wre not really necessary to keep. The full Resynthesis sound was posible to be played back by modualtion of teh wavetable position (with a LFO or envelope). Like in the original recording this resulted in decrtaesing volume over time (for proper rplication you would have to set the Sustain level in the Amp envelope to 0) and adjust the Decay time to a value that seems to fit, same for the mod speed/time of the wave position




Everyone could post his own result of sampling and Resynthesi of real glass here. I will maybe do more exmples/recordings in the future. FWIW i had started all this around 24 hours ago (or even less) so theer is room for improvements and/or more examples for sure.


More concerning downloads etc. in the next posts here...

[IngoWeidner]

At the moment i want to keep the original results of the Resynthesis for myself but i had created edited versions with a minimum of waveforms and all waveforms set to a maximum volume so they all got the same loudness opposing to the original recording. This offers an opportunity to get waveforms/sounds you would have totally ignored in the original recording or would not even be able to recognize. This is one case where Resynthesis could lead to quite differnt results than "simple" sampling as with Resynthesi a recording/sample is divided into many small waveforms that could be used each on their own or as a part of a wavetable.


Here is a download for a wavetable where i combined waveforms of all recordings into a single Icarus wavetable and added 3 crossfaded waveforms between waveforms from each of the recordings (overall 166 waveforms in the table):
https://dl.dropboxusercontent.com/u/53230726/Crystal%20Glass_all%20IW.zip


While this is not identical to the original recordings and the fisrt Resynthesis of those it includes the "essence" of what was received from those and the single waveforms could offer options not possible from the original samples and Resynthesis results.
In the waveforms inckuded in teh table not only the volumes were normalized/maximized but also a bucnh of veyr small partials in the spectrum editor were removed that mostly have no influence on the sound of the waveform or only offer very subtle differnces..

You could either use this for using single waveforms or parts of the wavetable. PLaying the whole table is possible too but the resuklt would maybe not be that great then.
To get smooth result from a smaller amount of waveforms you could increase the "Fade" knob in the Os setio nof Icarus (higher values offer better/smoother interpolations).

While you could also use others i recommend testing the wavetable at an Octave (setting in the Oscs) of +1.

[IngoWeidner]

UPDATE 1 (3.9.2016):

Today i did some more sampling of different glass vessels (again hit/striked to produce sound), now made of "normal" glass, not from crystal glass like the first wavetable.
This included two Wine glasses, 2 shot glasses and a fish bowl (actually that never had fish inside...).

Again after Resynthesis i edited/cleaned the partials in the spectrum editor to remove the very small partials that have no or a very small effect on the sound and could be just noise.
I also maximized the volume for all waveforms and selcted only the ones that are really different which in some exmples are just a few.

Here is a 2nd ZIP file with 5 glass wavetables:
https://dl.dropboxusercontent.com/u/53230726/Glass%20Resynthesis%202%20IW.zip

To play the full tables i recommend using the Fade knob i the OSc at a higher amount to get smooth results.

UPDATE:
For those who are interested here are two wavetables of the Wine glasses (included in the previouas 2nd table) directly after the Resynthesis of the sample/recording and just cutting waves at the beginning and the end of the table:
https://dl.dropboxusercontent.com/u/53230726/Glass%20Resynthesis_Wine%20Glasses_no%20edit%20%20IW.zip

To play this properly you need to scan trough the whole table. I had used a Saw Up LFO with slow speed, the LFO set to ENV andthe Wave mod LFO1 knob in the Osc at maximum amount.

No waveforms were edited in this (none of the very small partials removed) and while the wavetable is normalized the relative amplitudes of the waves are the same as in the original recordning.

[OneOfManyPauls]

Interesting work - thanks for sharing.

That wavetable seems a good basis for electric-piano style patches.

[IngoWeidner]

Interesting work - thanks for sharing.

That wavetable seems a good basis for electric-piano style patches.

Hi,

many thanks for the nice comment.  

A good starting pouint for patches basd on it would be:
- single osc with "HyperStereo 4" unison mode (+ some detuning for this)
- Octave at +1
- Lowpass filter
- "FAde" knob to 0 if you want to "browse" trough single waveforms, set Fade to higher values if you want to scan trough parts of the table
- some Reverb (there is actually a mode called "Reverb Glass"...) and/or delay added
- Amp envelope with Suatain at 0 and Attack slightly bigger than 0 (to avoid clicks) and Decay at a medium to higher value.

For pads and other synth stuff you could of course also use higher Sustain levels and/or longer Attack and Release.

[OneOfManyPauls]

It's funny how easy it is to get stuff that sounds like it fell out of the omnisphere factory presets library with it just by having a volume env with decay/release, routing a little bit of positive aux env 1 with its default curve to wave 1 position, adding pretty much any delay/reverb fx to it, and then manually moving the wave 1 start point to find workable patches.

[IngoWeidner]

It's funny how easy it is to get stuff that sounds like it fell out of the omnisphere factory presets library with it just by having a volume env with decay/release, routing a little bit of positive aux env 1 with its default curve to wave 1 position, adding pretty much any delay/reverb fx to it, and then manually moving the wave 1 start point to find workable patches.

I tried finding a term for the sound source of this. Maybe we could call it "analog glass"...  

Concerning "analog warmth" it could be interesting to check sound diffferences with warmed up or cooled down glass... (while warming cooling too fast could destroy glass if not being careful).

The leaded crystal glass used for the first recordings is a quite special material. Would be also interesting to compare it to "normal" and/or "cheap" glass.

[OneOfManyPauls]

Just realized I've lost over an hour playing with this wavetable!

How's about this for one-key-press instant meditation cd?

https://www.dropbox.com/s/2qp1jz3mapxwp4m/crystalglass_iw_wt_preview.wav?dl=0

and patch in case anybody wants it:

https://www.dropbox.com/s/2dnt081kun0fwgs/CrystalGlass_IW_WT.fxp?dl=0

(ok, there's an extra bit 24s in where I was pressing 2 notes an octave apart)

that's just doing what I said before but with the factory preset nightdriverBC arp + ping pong delay + glass reverb.

thanks again for posting this wavetable.

[IngoWeidner]

Just realized I've lost over an hour playing with this wavetable!

How's about this for one-key-press instant meditation cd?

https://www.dropbox.com/s/2qp1jz3mapxwp4m/crystalglass_iw_wt_preview.wav?dl=0

and patch in case anybody wants it:

https://www.dropbox.com/s/2dnt081kun0fwgs/CrystalGlass_IW_WT.fxp?dl=0

(ok, there's an extra bit 24s in where I was pressing 2 notes an octave apart)

that's just doing what I said before but with the factory preset nightdriverBC arp + ping pong delay + glass reverb.

thanks again for posting this wavetable.

Hi,

interesting patch indeed also due to using the Arp in it.

Maybe in the future Markus must add a new category like "meditation" or "meditative" to thz Icarus_sounds folder... 

[IngoWeidner]

Today i did some more sampling of different glass vessels (again hit/striked to produce sound), now made of "normal" glass, not from crystal glass like the first wavetable.
This included two Wine glasses, 2 shot glasses and a fish bowl (actually that never had fish inside...).

Again after Resynthesis i edited/cleaned the partials in the spectrum editor to remove the very small partials that have no or a very small effect on the sound and could be just noise.
I also maximized the volume for all waveforms and selcted only the ones that are really different which in some exmples are just a few.

Here is a 2nd ZIP file with 5 glass wavetables:
https://dl.dropboxusercontent.com/u/53230726/Glass%20Resynthesis%202%20IW.zip

To play the full tables i recommend using the Fade knob i the OSc at a higher amount to get smooth results.


I also added this to the 3rd post here (that was reseerved for updates).

[IngoWeidner]

Today i did some more sampling of different glass vessels (again hit/striked to produce sound), now made of "normal" glass, not from crystal glass like the first wavetable.
This included two Wine glasses, 2 shot glasses and a fish bowl (actually that never had fish inside...).

Again after Resynthesis i edited/cleaned the partials in the spectrum editor to remove the very small partials that have no or a very small effect on the sound and could be just noise.
I also maximized the volume for all waveforms and selcted only the ones that are really different which in some exmples are just a few.

Here is a 2nd ZIP file with 5 glass wavetables:
https://dl.dropboxusercontent.com/u/53230726/Glass%20Resynthesis%202%20IW.zip

For those who are interested here are two wavetables of the Wine glasses (included in the previouas 2nd table) directly after the Resynthesis of the sample/recording and just cutting waves at the beginning and the end of the table:
https://dl.dropboxusercontent.com/u/53230726/Glass%20Resynthesis_Wine%20Glasses_no%20edit%20%20IW.zip

To play this properly you need to scan trough the whole table. I had used a Saw Up LFO with slow speed, the LFO set to ENV andthe Wave mod LFO1 knob in the Osc at maximum amount.

No waveforms were edited in this (none of the very small partials removed) and while the wavetable is normalized the relative amplitudes of the waves are the same as in the original recordning.