Tablet string control fun

[Hannes_F]

Hi,

strings are still not easy to mock up by midi orchestration. The main problem is that sampled instruments appear to be stiff and more static in tone than real recordings. Although there are libraries with many different layers, articulations and control parameters for vibrato, attack and timbre, it is not easy to combine them or, if they are merged into virtual instruments, to control these parameters in realtime.

One approach to control the main parameters would be to use a pressure sensitive graphic tablet with a software that transforms coordinates and pressure to midi data. For a test with tablet2midi by www.livelab.dk (http://www.livelab.dk) see http://www.northernsounds.com/forum/showthread.php?t=50318

The basic idea is to set one axis (e. g. x-axis) to the volume parameter. It would be preferable to trigger multiple layers with this, with other words the x-axis could replace the modwheel with libraries like GPO or the Garritan Strad.

The y-axis would emulate the timbre control. A string player uses the distance between the bridge and the contact point of bow and strings to make the sound brighter (towards bridge) or warmer (towards fingerboard). The extremes are ponticello (a screeching sound, well known from the shower scene in Psycho) and flautando/sul tasto. Note that almost every string player changes this position within almost every note by default - whereas many samples seem to be recorded as sustains without this changing. With the tablet control this is is emulated by moving the pen up and down.

Combining volume and timbre changings result in characteristic 2-D curves, similar to writing letters. One can draw perpendicular lines, diagonals, bows, circles, and each of them will form a sound in a particular way. After a while the user learns intuitively which move is linked to which result.

The z-axis, triggered by pen pressure or the aftertouch channel of the keyboard, could control the vibrato amount. So the drawing would not be similar to a pencil line but the drawing with a brush or a feather.
http://www.frischat.com/compose/HannesFrischat_SoundCube_001.jpg
When strings begin to play there will always be a little knack - this is the moment when the string looses contact to the moving bow for the first time, slides a little while and gets grip again. This sound is normally referred to as 'attack' and while it will be always there its amound is dependent on two parameters: the pressure of the bow (directly linked to the volume) and the contact point of bow and string. So it should already be defined by the xy-position of the pen at the beginning of the note. However it may be convenient to add a certain influence of the keyboard velocity for intuitive playing.

The idea is to have a relative simple and intuitive realtime control with an easily available and inexpensive controller for moving in a three-dimensional sound array. Each reachable point in this array should produce sounds that are in the range of the real instruments, even if parts of it may be formed by eqing.

What do you think?


Hannes

[joaz]

Hannes, interesting idea.

Potential Problems.
Although string players in a section use the bow exactly as you described. They never move the bow identically to the guy sitting next to them.
This is part of the richness of the string section sound.
In any given phrase they are more likely to achieve unanimity on the climactic note, and then to start to differ as they move away from this peak.
If a sampled section all exhibited the identical EQ curve, this might begin to sound synthetic.

I think your idea could work with an ensemble building approach, and if the primary axis data could be interpreted slightly differently by each solo violin. This is perhaps possible with K2 scripting.

What you are doing for tone, JBacal has done for rhythm with his excellent ensemble maker script, which adds random amounts of sloppy timing to notes.
This works best of all with violas in my experience. :)

regards Joe

[Hannes_F]

Hi Joe,

the idea would be to do the one and not desist the other. :-)



Although string players in a section use the bow exactly as you described. They never move the bow identically to the guy sitting next to them.
This is part of the richness of the string section sound.
In any given phrase they are more likely to achieve unanimity on the climactic note, and then to start to differ as they move away from this peak.
If a sampled section all exhibited the identical EQ curve, this might begin to sound synthetic.

First thing I do to avoid this is that I play the same melody line with three layers in the first violin (one of them is a solo violin). Each of them is played individually with has its own timing and EQ curve which makes the transitions overlap.

Plus, I have developped a technique of making the transition itself more tight even in one line. I did not use it yet with the Barber rendition but here

http://www.northernsounds.com/forum/showthread.php?t=49561

The idea is to add the right amount of a feedbacked delay exactly at the legato transition and fade this in and out. This is a short phrase of one violin section only, first without this smoothing, then with it. Note that the effect is not there at the beginning and the end of the phrases. That could also be done easily but I wanted to avoid overreverbing the whole section in the first step.

http://www.frischat.com/compose/HannesFrischat_Legatotest_02.mp3

The technique is not yet fully developped. First I want to trigger it with the legato pedal because it is tedious to draw cc curves for every legato. I have yet to find something that makes a ramp from a cc step. Or I maybe drive it with a foot pedal which might be the better idea.

Then it would be ideal not to use a feedback delay with a static delay time but to fold (convolute) it with an appropriate impulse that can make a sort of bunch of micro-transitions opposed to an even sequence. I tried that but did not succeed yet.

The reason I am posting this without being ready is that it may show that the legato topic and the dynamics/timbre topic can be solved independent from each other.


Hannes