???Question about GS3???

[cptexas]

Hey, everyone!

I haven't purchased GS3 yet, but hope to within the next year or so. (I REALLY need A LOT of time to think things over before I spend any money :D )
I was reading the downloadable manuel for how to create instruments. This is called, as I learned, the "Instrument Editor" I think. It doesn't say anything about controlls that digitally modify the sound, like adding viberato to a NV instrument. (this is orchestral stuff we're talking about here) Is there a way I can make the mod wheel control viberato by slightly pitch bending the NV sample?
Just a yes or no would suffice. Knowing how to do it wouldn't help me here. I haven't purchased the software yet!!! :D

Also, what exactly is a 'dimension'? It says in the manuel that there is a limit of 127 of them. Does that mean that I can only have 127 different samples per region or that I can only have 127 different controls per region?
:confused:
Thanks for y'all's time!

Chris

[gugliel]

Yes.

Dimensions: controllable alternatives, via mod wheel, midi control message, note velocity, key on, key off, etc.

[JonFairhurst]

To go into more detail...

Let's start off with a single stereo sample. That's two dimensions. Let's go with four velocity layers. Now we have eight dimensions. Let's say you use the mod wheel to crossfade between the first sample set and a second. Now you have sixteen. Now let's say you have four keyswitches, so you can select between slow-attack sustain, fast-attack sustain, marcato or staccato. Now you're up to 64 dimensions (2x4x2x4 = 64). We have one more split available. Okay, let's redo the whole thing with eight velocity layers. We've now hit the ceiling. (2x8x2x4 = 128 - or zero through 127.)

The dimensions are really the number of individual mono samples that you can have related to a single key on the keyboard for a given instrument. For GS2.5 the limit was 32.

-JF

[cptexas]

OK, I gets it now!
But what about digitally modifying sounds?

Chris

[JonFairhurst]

Let's say you automate a parameter of downstream EQ or GigaPulse. That doesn't change the sample-side of things, so there are no additional dimensions required for that kind of process. Dimensions just concern the number of concurrent samples that need to be ready when you play a note.

-JF

[cptexas]

But what parameters can you change with GS3? Can I turn a non-viberato instrument into a viberato instrument with viberato controled by the mod wheel? Can I have one control control the viberato speed and one the depth all derived from one sample?
Like I've said before, I'm not familiar with GS vocabulary, so if this isn't making scence, just let me know and I'll try to rephrase all of this. :o

Thanks for your help! :)

Chris

[JonFairhurst]

Regarding vibrato there are three paths...

1) Apply the LFO (low frequency oscillator) to the frequency control to modulate the pitch. This has a very synthy sound, but it has its place. This takes no extra dimensions.

2) Use a control to blend between non vibrato and vibrato samples. This requires more dimensions. It works okay for ensemble instruments, but fails for solo instruments. Instead of one violin slowly adding more vibrato, you mix between two violins - one with and one without vibrato.

3) Use an advanced algorithm, such as Gary Garritan's Sonic Morphing that will soon be released with his solo violin. GS3 doesn't include any such features, but this stuff is very new to the industry. (Use the LFO approach for solo instruments in Giga, if you need control. Otherwise use a sample that has escalating vibrato built into the recording.)

-JF

[cptexas]

gotcha! :)

[meroveus]

FWIW, I think the limit of 128 dimensions is archaic.
It is the only limit imposed by GS3 that I bump up against when programming/mapping samples. :mad:

I can't understand why Tascam didn't make it something more useful like 65536 or even 2^32, considering the cheapness of large HDs. :confused:

[Mark_Knecht]

FWIW, I think the limit of 128 dimensions is archaic.
It is the only limit imposed by GS3 that I bump up against when programming/mapping samples. :mad:

I can't understand why Tascam didn't make it something more useful like 65536 or even 2^32, considering the cheapness of large HDs. :confused:

It's not only the cost of the drive, it's the cost of the memory ($, power, etc.) to cache the initial samples. 65536 dimensions would be quite a few versions of middle C to have sitting around in memory... ;-)

That said I fundamentally agree with you. Why not make the file format support it and leave it to you developers to figure out how many you can really make work.

- Mark

[JonFairhurst]

I can understand the limit for mixed layers and stereo - those samples need to be streamed simultaneously when the notes are played. But for keyswitches and dynamic layers, you only need to play the samples selected when the note is struck.

It would make sense to have unlimited "static" dimensions, like keyswitches and dynamics, but a limited nubmer of "dynamic" dimensions like multichannel voices and mixed layers.

-JF

[cptexas]

Yeah, I just added up all the middle C samples for a violin that I was planning to record. It was almost 200! I think I need to cut down on those dynamic levels. The 200 was with six dynamic levels. I'm afraid if I go to five or four, the user will hear the transition from the p sample to the mp sample. I encountered this in an oboe of Garry's. The note just happened to be one of those in the crossover, and it sounded like two oboes. It was VERY suddle, but for my library I'm shooting for maximum realism and playability. It seems a library is usually either playable and somewhat realistic, or EXTREMELY relaistic but you have to go fishing through samples upon samples of middle C's.
But that's beyond the point.

So what other tricks are there in Giga that I can use without making more dimensions?
I'm just wondering if any of them can substitute for another dimension.

And also, I have a keyswitch for pizz samples, and that KS won't use any controller other than velocity. Can I do this or do I have to map all of the other controller values to the same sample?

Chris

[JonFairhurst]

200 samples? Six layers? That sounds great!

The solution that Kirk Hunter has used in the past with GS2.5 is to have a number of keyswitch instruments. The user chooses the one that's closest to their immediate needs. GS2.5 only supports 32 dimensions, so with GS3, you're far ahead of the game.

You could have one keyswitch instruments with half of the articulations, then another with the other half. You could then have some that combine articulations from the two halves with some practical considerations.

You might also solve the problem with linked ports. You would load the first instrument in channel 1 port one and the second in channel 1 port 2. Link ports 1 and 2. Make it so the keyswitches are independent, but all of the keys for instrument 1's active sounds map to silence in instrument 2. The keys that bring up articulations in instrument 2 map to silence in instrument 1.

Now you get the whole shebang available for live playing.

From a practical perspective though, this eats lots of RAM for samples that one might not use. If the user can choose only the articulations of interest, they can load a lot more instruments at once.

Here's a trick I use. I set up my instruments to use unique banks. For instance Westgate Studios Flute might be on Bank 11. Westgate Oboe might be on Bank 12. Kirk Hunter's Solo Violin might be Bank 20, with the Viola on 21, Cello 22 and Bass 23. I then map all of the sustain vibrato patches to program 0, sustain non vibrato program 1, portato program 2, marcato P3, staccato P4...

Sure, not all articulations overlap for all instruments. There's no pizz flute, for instance.

The advantage of this approach is that I can load up the articulations that I want, and set each track for the needed bank. I then use program changes between articulations. The nice thing is that I can remember that vibrato sustain is always P0 and staccato is always P4. It's easy to remember.

Anyway, some mega keyswitches will satisfy the live players who want it all. Some smaller keyswitches with practical groupings works with more RAM efficiency. And the Bank/Program approach works for people who want the best efficiency, but who select articulations after the fact. By alway using the same program numbers for your articulations, users easily learn the setups. Using consistent keyswitch mappings also helps.

-JF

[cptexas]

200 samples? Six layers? That sounds great!
Great as in realism? Or quality? What? I want to know so I don't change whatever it is you like.


You could have one keyswitch instruments with half of the articulations, then another with the other half. You could then have some that combine articulations from the two halves with some practical considerations.

I see where your coming from, but since my library is (hopefully) going to be based upon layered solo instruments, two channels per instrument will eat up your 128 channels quick if you want a large orchestra. Maybe I could have different style instruments........like an aggressive instrument, a light instrument, an airie instrument, a dramatic instrument, a normal instrument. Any others? I think I like this idea, but it wouldn't be that great if your music incorperates a bunch or styles.


You might also solve the problem with linked ports. You would load the first instrument in channel 1 port one and the second in channel 1 port 2. Link ports 1 and 2. Make it so the keyswitches are independent, but all of the keys for instrument 1's active sounds map to silence in instrument 2. The keys that bring up articulations in instrument 2 map to silence in instrument 1.
Now you get the whole shebang available for live playing.

Again, with only 128 channels to work off of and solo instruments building enseambles, this could be a problem with large orchestras.


From a practical perspective though, this eats lots of RAM for samples that one might not use. If the user can choose only the articulations of interest, they can load a lot more instruments at once.
I thought that problem was solved with direct from disc streaming? Or will it still eat a lot of RAM?


Here's a trick I use. I set up my instruments to use unique banks. For instance Westgate Studios Flute might be on Bank 11. Westgate Oboe might be on Bank 12. Kirk Hunter's Solo Violin might be Bank 20, with the Viola on 21, Cello 22 and Bass 23. I then map all of the sustain vibrato patches to program 0, sustain non vibrato program 1, portato program 2, marcato P3, staccato P4...

BIG words!
please rephraise if possible. :o

Overall, these were great suggestions. What do you think about the different style instrument idea?

Thank you sooooooooooo much for your help! :)
Chris

[JonFairhurst]

Great as in realism? Or quality? What? I want to know so I don't change whatever it is you like.

I like the 200 samples, because that means there are lots of articulations, and I think six velocity layers is perfect. Four is often not enough. Eight can be costly. Six is a nice goal.


I see where your coming from, but since my library is (hopefully) going to be based upon layered solo instruments, two channels per instrument will eat up your 128 channels quick if you want a large orchestra. Maybe I could have different style instruments........like an aggressive instrument, a light instrument, an airie instrument, a dramatic instrument, a normal instrument. Any others? I think I like this idea, but it wouldn't be that great if your music incorperates a bunch or styles.
Again, with only 128 channels to work off of and solo instruments building enseambles, this could be a problem with large orchestras.

Oooh. Layered solo instruments. Cool! This could be done well with stacking. Just stack as many instruments as you want for a big ensemble. Stack fewer for divisi. Or use them separately, and program them independently.

By stacking, they share channels, so the 127 limit isn't a problem.



I thought that problem was solved with direct from disc streaming? Or will it still eat a lot of RAM?

Every sample that gets loaded eats some RAM. The reason is the need for low latency. When you play a key, the audio is instantly streamed from RAM. The disk starts seeking the data immediately, and finally catches up before the RAM part of the sample is used up.


BIG words!
please rephraise if possible. :o

It's pretty simple really. Give each of your instruments its own MIDI bank number. Give each articulation (sustain, pizz, etc) its own MIDI program number. Think of it like a big grid, with instruments on one axis and the articulation on the other axis.


Overall, these were great suggestions. What do you think about the different style instrument idea?

A variety of styles can be really useful. For instance, Kirk Hunter's Solo Strings has both Normal and Gypsy violins. The Normal has moderate vibrato and short slides. The Gypsy has aggressive vibrato and huge slides. Also, they physically different instruments.


Thank you sooooooooooo much for your help! :)
Chris

Hey, this is fun! Who knows what cool lib might result!

-JF

[cptexas]

All right.
I've gots an idea.
My problem really is that each keyswitch takes up so many dimensions
So.................
Drum roll please.....................
I can stack instruments with as many keyswitches as the 128 dimensions will allow on one instrument, with another instrument with more articulations, with another instrument with MORE articulations. Since every MIDI signal goes to every stacked instrument, I can create some keyswitches to trigger silence so that I don't have three violins playing at once when I only want one.
I don't know if I said that correctly.
For example,
I could have C1-F1 be all styles of long bows, and F#1-C2 to trigger silence. On the other matching 'pair' F#1-C2 will trigger different styles of trills, tremlos, pizz, short bows, and everything else, and C1-F1 will trigger silence.
I feel so smart right now! :D

Chris

EDIT: Actually, it would be more efficient if each instrument had only one keyswitch that didn't trigger silence. That way, the user doesn't have to waste any RAM at all with unused samples!
I'm so smart! :D

[cptexas]

So what exactly can you do with strings?
Here's what I've come up with so far:

Long bows intense, normal, and light in viberato and non-viberato, muted and unmuted, in upbows and downbows, slurred and not slurred, all with attacks that are either normal, cruch, or click and with six dynamic layers.

Pizz with snap pizz from velocity 120-127 in muted and unmuted.

Trills in whole step and half step, in starting from top note and starting from bottom note, in muted and unmuted, slurred or not, upbows and downbows, attacks in normal crunch and click, and in the three styles of intense, normal, and light.

Tremlos in the three styles, the three attacks, and muted and unmuted.

Anything else I'm forgetting?
Short bows will be done by creating a very short long bow note, and everything above will be done in the six dynamic levels.
Everything added up above would be 2623 samples per note!!! I surprised myself! :eek:
Any suggestions for cutting that number down?

And also, with crossfading between different dynamic samples, how would I do that with viberato samples? Without making the viberato jolt up or down in pitch when it crosses over to another sample?

Thanx,
Chris

[JonFairhurst]

...And also, with crossfading between different dynamic samples, how would I do that with viberato samples? Without making the viberato jolt up or down in pitch when it crosses over to another sample?...

You could have no, light, medium and hard vibrato. It will jump, but not as much. Yeah, I know. More samples...

Since you're looking at ensemble building (very sample intensive, BTW), you can have each instrument's mod-wheel crossover points come in at slightly different places. That will help mask the effect. Also, the crossfades can be gradual, rather than on/off, so you won't hear any harsh click.

-JF

[cptexas]

Good idea, JF!
With the light, normal, and intense styles, the viberato will be different for each. That'll cover it a bit.

But what about cutting down that slightly large number of articulations? I just need to cut it down so that I don't drive musicians crazy when recording.
Just imagine: "Hello, Sarah. I'm going to need to you play 2600 chromatic scales for me today." :D
That'd be funny!

Chris

[cptexas]

I REALLY don't mean to be a pest, but I really need to figure out how to lower that number. :o

Thanx, :)
Chris

[cptexas]

Never mind.
I've figured it out.
I can't explain it in words, so that's that! :)

Chris