Velocity Layer Test Experiment

dore_m

"X many velocity layers" is such a sampled instrument selling point, that I thought we could crowdsource an experiment here on PC.

I was thinking about making prototype sample libraries to share with people here to get opinions on. I'd make 5 playable instruments, and volunteers would download/play the five and rank them in terms of number of velocity layers that they can detect by just using them. It's both an experiment and a game/challenge. I'm looking for volunteers.

Here's what I'd like if you do agree to try:

Download and play libraries
Rank by least to most velocity layers
State how you characterize your playing ability and controller you use

I'm considering instruments with 4 ,8, 12, 18, and 24 layers. Oh, and you'd have to be able to use an SFZ file (Sforzando or other SFZ player).

If you're interested, post here that you are and if I get enough people, I'll go ahead and create something for test. I also will distribute the libraries via DM as I don't want to publicly post them.

*** Argggh. "test experiment" how redundant can I be?

QuasiUnaFantasia

I'm in! I realize this means "no peeking at the files". 🙂

dore_m *** Argggh. "test experiment" how redundant can I be?

Well, it could be an experimental test experiment. 😄

Joannchr

dore_m count me in . It is an instructive test . My guess is that 4 layers will show serious obvious limitations specially in passages like rach 2 nd concerto with progressions of chords from pp to ff . 8 and 12 will layers will show limitations when dynamics equality is required ( for instance (Schubert op 90-3) as there may unwanted adjustment to closest layer that can generate audible gaps and 18 and 24 will be very hard to differentiate. But it is just a guess , confirming it by experiment would be fantastic.

amc252

Sure, I'd love to.

dore_m

@amc252 Cool! Let's recruit more....

David B

I have such a bad ear and very little actual experience and understanding when it comes to layers/samples. I don't even know how to discern loops. I hear people talk about it, but I can never discern it with the onboard sounds in different pianos.

It's an exercise in futility to try and get any meaningful data from me.

God Bless,
David

dore_m

@David B , I don't know - you'd probably know the difference and you could see where, on the spectrum, you lie.

But it does require getting SFZ files to work, if that's too much trouble.

Del Vento

I would participate too, if I had a suitable controller, but unfortunately at the moment I don't.

FWIW, I think a test like that would have worked even better if it were made "more blind" in that you did not tell the player what the difference was among the variants, just that you changed something and wanted to know if it were a meaningful change for the player.

PS: no I'm not dead, but I've been out of town to assist my father, who underwent major surgery, and my mother who had to juggle many other things. I'm still playing catch up and will continue to for the foreseeable future

dore_m

@Del Vento if you're still interested and are able when you get home, let me know. We could possibly wait to announce any results or opinions....

Del Vento

dore_m if you're still interested and are able when you get home

Thanks! I am in fact back home (otherwise I would not be "playing catch up", but would be still losing ground).
Even home, Mybrid is not ready (not even in a single octave) and at the moment I don't have any other working controller (I still have the grand and the clavichord, but those can't be used for this purpose)

HZPiano

Hello,

I shouldn't promise to participate, however curious I am, because I simply wouldn't find the time and energy to actually follow up on it.

What I can offer is a tip:

All five .sfz sets should be the same file size. Otherwise the sheer act of downloading and installing would give away what is what.

And/or maybe create one folder with all sound files in it for the five sets together, and provide five .sfz script files of same length that access the sound files each needs.

Hope this is some form of contribution,

Cheers and a happy experiment,

dore_m

@Joannchr that's great! The more the merrier. Advanced players will give me the most info, but also newer players can learn something as well.

HZPiano All five .sfz sets should be the same file size. Otherwise the sheer act of downloading and installing would give away what is what.

Actually the sample set will be for this experiment only, so I've decreased the recording lengths and taken all of the bells and whistles out so as to focus mostly on attack loudness. All versions use the same sample set, just less of the files are mapped to the kb. More to follow...

dore_m

To set some expectations, it's not like these libraries for comparison will be the most playable libraries in your collection. They are custom made for the experiment and the sustain recordings only last 3-4 seconds. Believe it or not, after stripping away all of the release samples, only including pedal up recordings, limiting sample length, and using flac compression, the library is only about 120mb.

EB5AGV

I am in if you think I can help 👍🏻

SouthPark

If pianoteq/roland acoustic modeling people plus the sample playback crew can get around to it in the future, and as was mentioned in other posts before, I would really like to see a collaboration effort, where the pianoteq/roland gang (in the future) come up with their best 'model' - and then generate for example -- 1023 or 2047 individual 'samples' - for the various different velocities. And pass those samples onto the sample playback gang to do whatever they need to do or include. And put out a sample playback digital piano based on that effort. This is regardless of whether 1023 or 2047 samples is overkill or not. And they can use (if needed) some sympathetic string resonance and other effects - if needed that is. The approach is to generate 'samples' from pianoteq/roland acoustic models for the sample playback gang to work some wonders. At the moment - if the pteq sounds aren't quite up to scratch yet, then that's ok. We just keep waiting - time for more evolving - and then we can do all this when the sounds make the grade.

Or, if they want to start smaller, then go for 254 or 511 'samples'.

SouthPark

SouthPark Or, if they want to start smaller, then go for 254 or 511 'samples'.

Actually … even 127 individual samples generated from 'best' pteq/roland acoustic models would be a nice start.

Also, I'm currently not using external 'vst' sound sets yet. But definitely interested to check out results from dore's user tests.

amc252

Okay, I went ahead and tried the samples, here are my first impressions.

I tested the library with liquidsfz on linux, my old clavinova clp115, and a pair of sony headphones, which is my usual setup, nothing fancy.

I first played my test MIDI track a few times, which I know very well 'cause I recently tested several VSTs. Then I proceeded to play each sfz with my piano, switching back and forth.

Now, I'm not a big fan of pianoteq, but the first surprise was that dore's samples actually sounded better. Of course, there were limitations in sustain and playability, but the timbre itself seemed fuller. It might have to do with liquidsfz vs pianoteq player or just my personal taste.

All in all, very interesting, I'll post my notes when everyone is done.

dore_m

amc252 Now, I'm not a big fan of pianoteq, but the first surprise was that dore's samples actually sounded better.

Sometimes I wonder if Pianos are like wine - brand reputation has a large influence on perception (?)

dore_m

If anyone is interested, we're holding off on making guesses until others try. Feel free to let me know here and I'll add you to the private group that can download the instruments.

SouthPark

dore_m Sometimes I wonder if Pianos are like wine - brand reputation has a large influence on perception (?)

For some people it probably is like that. For myself, it's only about enjoying the tinkering on any piano that makes interesting and enjoyable music. And learning and accumulating just enough experience in piano playing and music to have unlimited fun on any piano. And the rest is just continue to learn more and develop more, and have more unlimited fun in playing pianos. This is also probably similar to lots of other people - as well.

Joannchr

I encourage PC members to do the test . It’s a good way to evaluate factually the different number of velocity layers and feel what impact it makes on both playability and recording . Have started the test last night. It confirms some on my initial views , but the differences are more subtle that what numbers could suggest . The more testers we have , the more conclusions we can draw for the test . Thank to Mark for facilitating the testing.