[LAU] It's time to release The Salamander Grand Piano

[alexander]

On 05/22/2010 08:20 PM, Philipp Überbacher wrote:
> Excerpts from alexander's message of 2010-05-22 16:54:27 +0200:
>    
>> Gigedit is quite good as long as you're careful when using the "apply to
>> all regions" and "apply to dimension splits" check boxes. It's however
>> not fast when adding samples and regions and dimension as there is no
>> copy/paste. I originally was gonna use giga too but at one point the
>> editor started to behave really odd. There was some unexplainable issue
>> with regions only "seeing" one release velocity layer. Also, I now have
>> two separate release samples, harmonic string resonance and the hammer
>> action. The harmonic releases also comes in 3 different layers, 2
>> veolocity layers that have a high rt_decay value ("The volume decay
>> amount when the region is set to play in *release* trigger mode, in
>> decibels per second since note-on message.") and one that has a low
>> rt_decay value, this is not doable with giga. As far as I know you can't
>> even have 2 separate release samples so you would have to mix the hammer
>> noise and harmonic releases samples down into one sample.
>>
>> For making sfz files there is also a ruby script called "sfzer" I have
>> not tried this tho and I expect you can't make very advanced instruments
>> with it.
>>
>> We should put some pressure on the linuxsampler folks to do a new
>> release ;) with on_locc# and on_hicc# opcodes implemented. (those would
>> give pedal noise)
>>      
> Hi,
> I played with it a bit and seems to have issues, especially with the
> sustain pedal. I thought it was a bug that the sustain was either on or
> off, with nothing in between, my keyboard sending an in between value
> added to the confusion and I even reported a bug:
> https://bugs.linuxsampler.org/cgi-bin/show_bug.cgi?id=146
>
> Anyway, it seems that MIDI only knows on or off for sustain, so it's not
> actually a bug, but releasing a sustain feels very unnatural since it
> releases immediately. I found that changing the release time helps a
> bit: ampeg_release=0.360 to something like ampeg_release=1
>
>    
Yes, midi sustain is either on or off, where anything from 63 and below 
is OFF and 64 and above is ON. So velocity sensitive pedals should exist 
(I haven't looked into it). eg sustain message 120 would be a loud ON 
and 35 would be a moderate OFF.

I know that releasing the sustain sounds quit dry and unnatural and that 
is a common problem with midi. The problem is the note-off message(which 
actually is a note-on message with value 0, at least on my m-audio 
keyrig) is sent even if the sustain is on (>64) and the sample player 
simply ignores it. The optimal solution would be to have the sample 
player both remembering and releasing the note-off message at the same 
time. Then you could have the hammer noise releases play even if sustain 
is on and playing the harmonic string resonance samples when the sustain 
is released(off).

> The other thing is that removing your whole pedal section doesn't really
> hurt. There seem to be some bugs in that section:
>
> <group>  hikey=-1 lokey=-1 on_locc64=64 on_hicc64=127 off_by=2
>
> hikey and lokey take values from 0 to 127 according to:
> http://www.cakewalk.com/devxchange/sfz.asp
>
>    
Yes, this is a bit odd but I found another resource that said you have 
to disable note-on messages by giving hi/lokey a -1 value. Else 
on_hi/locc# wont work. I sort of confirmed it by loading the free sfz 
player in dssi-vst and trying it there, I couldn't properly test it tho 
as it kept crashing when I loaded the whole instrument.

> Someone who works with sfz told me that LS currently has a bug and
> needs offby instead of off_by.
>
> <group>  hikey=0 lokey=0 on_locc64=0 on_hicc64=63 group=2 volume=-30
>
> This line refers to group=2 which doesn't exist.
>    
The group=2 opcode is connected to off_by=2. What it does is that when a 
region that has the group=2 plays, it stops all regions that has 
off_by=2. As for now one could comment out the pedal action section as 
the on_lo(hi)cc# opcodes aren't implemented in LS yet.
> These are just some things that were noticed, I didn't even try to
> understand sfz and didn't check the whole file. I'm very happy that the
> Salamander exists and will certainly use it, so thanks :)
>    
I'm happy you like it!