On 1/28/07, BJaY <bruceslists(a)ntlworld.com> wrote:
I was just reviewing some projects (again) and noticed that during the
loading of the projects, the stating of rosegarden set the volumes of some
(but not all) of the linuxsampler channels to 48. I've done nothing that I'm
aware of to make it do this, all channels should be at full volume. When I
change the volume control for the midi channel in rosegarden, the volume of
the channel changes in qsampler but not to its full extent. I'm using planet
CCRMA on FC6 (32 bit)- a recent version of linuxsampler and qsampler that
Nando did to fix the 'immediate crash' bug.
As someone who participated in a *LOT* of the LS testing in the early
days as it moved from an intersting idea and into a piece of software
that became more usable let me suggest that in general you do *not*
want all channels at full volume, most especially if you are loading a
lot of gig files. After the first phase where a sample gets played LS
has to mix these samples into one or more audio output streams. LS, as
opposed to GSt, has very little signal overhead in it's channel level
mixing buses as well as it's output mixing buses. This will create
severe distortion if you mix even a few notes at the same time. GSt
has very much more overhead and doesn't run into this problem to such
an extent although it has the same sort of limitations. If you start
driving LS very hard with channel volumes up then you will likely run
into this pretty quickly and have to pull them down anyway.
Whether you are an LS user, or a GSt user such as I am, I suggest that
you look very carefully at MIDI velocity events in Rosegarden, most
especially if you are using any of the larger orchestral gig files.
Lowering MIDI velocity will generally give you more realistic sampling
of what a real player does most of the time and allows you to save the
high velocity samples for times when players are really laying into
their instruments. Set output volumes in such a way as to get the
highest S/N you think you can without clipping LS outputs. With LS I
would always output each instrument individually and then record and
mix them in Ardour where the mix buses worked better.
On an aside - the most immediately usfull tools for general musicians are
the software synths - qsynth and qsampler - and I've had loads of trouble
with them. With ZynAddSubFX having problems with real time (and soft synths
are only any use with real time), users of linux audio are left with very
few options. What can we do ?
No on-list comments from me on this one other than to say I didn't
have much trouble at all with qsampler a year or two ago when I was
still involved with the LS project.