Editing Multi sampled acoustic piano

[Oggfish]

Hi All, i just finished sampling a Yamaha acoustic piano (8 velocity layers > 608 25sec samples) with Samplerobot. Now I try to enhance the playability/realism in my Kronos 2. I have a few questions regarding this. Maybe someone has the experience to point out a few things?

1) When it comes to realism should I avoid using Xfade range in OSC?
Playing around with this does not really make it clear to me.

2) Do I need to make a combi if I want Kronos damper noise added to the multi sampled piano? Samplerobot 6 does not support release samples YET. I can’t seem to find a way to load the multi samples directly into a SXG2 program. It must be a HD1 program right?

3) It`s odd for me that I have to change volume levels per velocity layer in OSC in order to make the transition between those layers more smooth. It should automatically get louder the harder I press it? I don't think the samples got normalized when exporting.

If you have some other tips to try out they would be very welcome. Editing synth sounds and such are much more easy but an acoustic instrument like a piano needs some more practice.

[AntonySharmman]

1) Xfade is actually a mixer amplifier so the average point has a result of adding 2 different sounds and can create either phase cancellation of near frequencies
nor phase modulation effects so if velocity layers are properly created it should be avoided IMO.
2) Definitely HD1 is your only option and you also need pedal noise , keys mechanical noises , releases and hammers FX and others like sympathetic resonance
and another 8 velocity samples of sustain pedal.
Program mode is very poor to host all those articulations as you may have discovered , you can add the most pedal or hammer articulation , so you have to use
Combination mode to combine Programs with sustain velocity samples and Programs with FX and articulation.
3) I have calculated average velocity curve of Kronos and for the best smooth transition of layers I advice every lower velocity layer to be max -2db lower than
the higher one , so just gain low volume velocity layers accordingly but never normalize.

If you believe that you need sound developers assistance to create it for you , then you can use WavesArt Service https://wavesart.eu/Hire

Hope this helps

[Oggfish]

Thank you very much Antony for this clear reply. At least i know what i could tryout and look out for. Otherwise i would surely look for assistance!

[KK]

Hi there,

Theory and practice can differ a lot and be very surprising when it comes to sampling, just like creating conventional (or not) instrument sounds. Also, a very important aspect is that the Kronos, just like any other digital machine, has a polyphony limit. In HD-1, it is 148 voices (it will actually start to cut notes before that depending what is played).

Crossfading : it will mainly depend on your samples and the only way to know how it will sound is to experiment. But remember that any key in a crossfade will double the voices played, per OSC. If you use stereo samples, each key in a crossfade will play 4 voices and this, for each OSC. So a simple RH 4-note chord with just one program of two OSCs and crossfading on both sides uses 32 voices. Add the LH and of course pushing on the sustain pedal and you rapidly end up out of voices.

Release/noise samples : personally I always found them annoying on real grand pianos, so to me there are useless in digital realm. Oh and they of course use voices too, so an even better way to reach polyphony limit.

Amplitude of layers : as mentioned, there is theory and practice. Here another enemy is your samples' SNR. As they decay, long piano notes show the quality of your samples. Make sure to record with minimum noise. If long piano chords transform into a white noise generator after a few seconds, play with filter envelope parameters to try to improve results.

[Oggfish]

Amplitude of layers : as mentioned, there is theory and practice. Here another enemy is your samples' SNR. As they decay, long piano notes show the quality of your samples. Make sure to record with minimum noise. If long piano chords transform into a white noise generator after a few seconds, play with filter envelope parameters to try to improve results.

Yeah.. the white noise is a big problem. Strangely this happens only with the three lowest velocity samples. The higher vel samples are fine, even if i keep holding the sustain pedal (noise will stack up). The recording quality should be fine actually. I will try to fool around with those filter parameters. Why is this noise coming from the lower vel samples only?

[KK]

Strangely this happens only with the three lowest velocity samples. Why is this noise coming from the lower vel samples only?

When sampling always record at optimal level, regardless if it is a soft or loud sound. If one records piano samples of keys played fortissimo and then uses the same recording level when sampling keys played very softly, that is for sure noise will be much more obvious on the soft sounds. One trick is to find the loudest key of a given velocity set, then set it just a notch under max level with no distortion (or lower depending if you plan to edit/treat samples with EQ, effects and so on before transferring them in your Kronos). Then do the same for each velocity set. Keep track of the dB adjustments between each set, then reproduce this later in the sample playing parameters.

[bpoodoo]

Keep track of the dB adjustments between each set, then reproduce this later in the sample playing parameters.

That sounds like a good strategy. A typical way a digital piano can sound artificial or synthetic - that it doesn't sound right - is the sound volumes and timbres for different velocity layers are mismatched or discontinuous compared to a real piano.

[Oggfish]

Hmm i will try that. Before it seemed like the result will be the same thing. Like cranking up a soft sound before recording, or just record it soft. In my previous assumption the noise amount would be about same. I need to look at samplerobot how to do this properly. Thanks for this advise!

Strangely this happens only with the three lowest velocity samples. Why is this noise coming from the lower vel samples only?

When sampling always record at optimal level, regardless if it is a soft or loud sound. If one records piano samples of keys played fortissimo and then uses the same recording level when sampling keys played very softly, that is for sure noise will be much more obvious on the soft sounds. One trick is to find the loudest key of a given velocity set, then set it just a notch under max level with no distortion (or lower depending if you plan to edit/treat samples with EQ, effects and so on before transferring them in your Kronos). Then do the same for each velocity set. Keep track of the dB adjustments between each set, then reproduce this later in the sample playing parameters.

[KK]

As with anything else, perfection is elusive and great results come with lots of work and trial and error. Sampling is an art and involves many technical details and tricks, especially for instruments like acoustic piano. There are entire books on the subject and they only cover the theory.