The synth patch makes use of PWM for animation within the timbre. For thickness and the ensemble tone, I layered detuned oscs together. This worked extremely well in achieving the sound I wanted, but presented a problem when sampling in that sometimes the osc phases would align and create a hot peak that would clip the A/D converter.
I was not willing to compromise the overall sampling level of the sound to prevent A/D clipping, but I also did not want to use any compression or limiting on the source sound. My solution was simply to take 30-second-long samples such that I would get a long-enough sample region that was "clean," even if I did get a clipped peak somewhere during the sampling. This worked quite well - if I got a clip, it was never more than 1 over within the 30-second sample.
I made a total of 21 samples from C1-C6, sampling every minor third, i.e., C - D# - F# - A, at 44.1KHz, 16-bits depth, in mono. Each raw sample was ~2.5MB. I looped the raw samples such that the loops were ~17 seconds in length, and 1.5MB in size. This resulted in a sample set that was ~32MB in size - perhaps a little larger. If so, not using the C6 sample would be the easiest solution for a 32MB sampler.
After considering the results I began to think I could still achieve my original goal (ensemble string sound without obvious short-loop) with shorter loops, resulting in a more memory-efficient sample set. I went back to my original raw samples and re-looped them such that the loops were ~9-10 seconds in length, and ~800KB in size. This resulted in a sample set that was ~16MB in size - perhaps a little larger. If so, not using the C6 sample would be the easiest solution for a 16MB sampler.
I had one issue to address before I was done with the samples, and that was the issue of levels compression. These samples are quite dynamic, with all the life and energy of the original oscillators. This sounds quite pleasing to me, but I also understand that it may not be the best choice when preparing a sampleset to be used in production. My thinking was to use a gentle compression such that their energy was maintained, but with the overall "lumpiness" somewhat tamed, resulting in a more consistent polyphonic sound.
I did just that, and 21 samples later, I had what I considered a nice sampleset going that preserved my original timbral intent. I am pleased with the way the samples came out in the end. These samples are absolutely clickless and smooth-sustaining - if you sustain a note for more than 9-10 seconds, that is.
Overall, these samples are fairly bright, at least in the high-end. This is purposeful, so there are options as it concerns later filtering in a sampler architecture, and/or EQ during mix and production. This may also be of benefit if they are used with a vocoder.
Aside from the gentle compression, no effects processing was used in the creation of these samples. They are absolutely dry and raw.
The files are named as such:
nn_keynameoct.wav
where nn is the MIDI note number for key assignment (36-96), keyname is C, D#, F#, or A, and oct is the octave of the sample (1-6).
And yea, I am sharing them with you.
Download them here:http://rapidshare.com/files/198978380/aeon_strings.zip
You will need to assign these samples to a keymap and assign filter and amplitude envelopes so they properly play as "strings" would. Yea indeed, there is some programming work involved, minimal as it might be. That said, I hope you enjoy these samples and can put them to good use in your music.
cheers,
Ian
Let your ears be the judge - if it sounds good to you (or not) it is good (or not). I didn't want there to be bias as it concerns the use of this thing vs. that thing.
