EXS 24 Logic Sampler Users Group

Legend has it that Roland had some nice custom sample-processing 
software that they used to create some of their sample libraries.

If you were doing this seriously, and had access to your sampler 
editor's code, I guess you'd put in a feature to overlay two samples, 
set one sample to be the master, click a set of peaks or other 
identifiable points on the master waveform, click the corresponding 
peaks or points on the slave waveform and then have the automatically 
run through the slave waveform cutting or padding data until those 
points aligned, after which you'd hopefully have a guaranteed 
phaselock, and could try breaking the thing into components for 
resynthesis (selecting nice clean unvibrato'ed source samples 
probably helps). 
Roland's R&D helped create a range of resynthesised piano/epiano 
units (MKS20/P330/etc), so maybe their epiano samples were a spin-off 
of that development ... or maybe the samplists simply took one of the 
resynthesised piano patches, turned off the modulation, and got nice 
clean in-phase pairs of samples that way. 

PS: For complex-sounding "real" piano multicrossfades, I guess part 
of the problem is getting the notes played precisely enough in a 
repeatable way, across the keyboard, with precisely-defined force 
intervals -- that might allow you to minimise some of the phase-
variation issues. Yamaha's piano factory has automated machines that 
repeatedly thump the individual keys on completed pianos as 
a "shakedown" exercise -- since seeing that, I've always wondered how 
much easier it'd be to sample a piano properly, played with properly-
calibrated strengths using one of those machines. 
You could sit there with your analysis software, telling your 
autoplayer robot to keep playing "layer seven" strength on a 
particular key until the software beeped told you that the phase was 
a close enough match to the existing "layer six" sample, then click 
ok and repeat the exercise for "layer eight" on that note, and so on. 
Once you'd got the technique right, you could cover the whole 
keyboard in a reasoable amount of time, then do other pianos in the 
same way, then start looking at ways to reduce the amount of data 
when you got back to your lab. 


--- In exs-users@yahoogroups.com, "Eppo Schaap" <eschaap@w...> wrote:
> I still remember the best crossfading instrument I came across in 
years: the
> e-piano of the Roland U-2(2)0. It crossfaded from tiny bells to a 
pleasant
> roar without any audible switching (which can't even be said of e-
piano's in
> todays high tech workstations). Any tricks there, or just very good 
sampling
> :-)?