How many layers are really necessary?

[Daniel]

what do you think?

i just imported a snare form superiour drummer with dsoundtool and choosen 99 layers.

how many are do you guys use?

[Baby Samus]

I think the best rule here is to just use as many layers as you see fit - I have created .dsnd files with only 16 layers and they still sound very good to my ear - perhaps to others it may not be enough, so trust your ear and follow your ear and instinct.

[nonoduweb]

60 layers for the snares, it's sufficient for me. Don't forget to put enough time between each hit (3 secs or more if you use some reverb).

[Slap the drummer]

There's also the question of the range between softest and loudest hit.  It seems to me that if there are too many samples
in a narrow dynamic range, then the random selection won't give the proper feel.

I've noticed that the 2box dsnd files tend to have a pretty high level softest hit -- ie minus 30 db, whereas say
minus 50 is very quiet but potentially audible (if you're a bat).

Mostly the 2box files sound great (??) so I guess they know what they're doing.  Their files do vary quite a lot tho.
On the whole an average (I'm guessing a bit) seems to be 25/30 hits with range from -30db to -3db.  But for instance
the Swiss 22 ride is 47 hits in range 0db to minus 20db.  Then again the EU 20 ride is 25 hits, range -6db to -32db.
The EU ride is my favourite but I have noticed it can sound a bit mechanical.  Is there a connection???

Perhaps somebody whose logic is better than mine can offer some opinions!

[Daniel]

i had this thought that maybe fewer layers will cause LESS "machine gunning".

if i remember correctly the 2box never plays the same sample one after the other, (as superior drummer does also) to prevent machine gunning.

but if you have a snare with lets say of 99 layers, and play something soft, the software will automatically choose the second hit from the soft velocity area too.
but because you have so much soft velocity layers, and the difference in sound is so small from each sample (of the soft velocity area) that it sounds actually like machine
gunning. (btw: if you have 99 layers, i would refer to the first 25 to be in the soft velocity area)

anybody can confirm this?

i hope you understand what i mean, its pretty hard to explain in english, sorry.  

[nonoduweb]

Yes I think you're right, that's why in the case of sounds made with a VST drum software and the midi method of DSoundtool, we don't need to have a lot of layers.

The method of linear velocities works well, but it's true that many 2box sounds are not built like this, and their realism is sometimes better than the sounds which I was able to make. Perhaps that the ideal would be to listen to every sample and to determine its velocity arbitrarily. If the designer of the 2box sounds could give us some advice, it would be cool.

[Slap the drummer]

i had this thought that maybe fewer layers will cause LESS "machine gunning".

Yes, that's what i was trying to say.  Too many layers and you must reach a point where
machine gunning starts....

then again, not enough and.....  your kit will sound like a roland

[Deve Loper]

Hi.

The layers should each be recorded at the original level it sounded when hit. NO NORMALIZE HERE.
Then they are put into a dsnd with the top layer normalized and the rest at the original relative level with respect to the top layer.
This is how our Sound Editor uses each wav file to build a dsnd file. You can see how it sorts the layers by level.
So, yes, the MORE original layers, the LESS machine gunning because the playback engine in the unit alternates between those layers nearest the trigger level.
You can change this behaviour with the LAYER parameter:

- Fixed, a certain trigger level will always start the same layer which
best matches the trigger level amplitude.
- Variable (default), a certain trigger level will start a semi-random layer as
long as it is close to the trigger level amplitude. This allows for a
very natural response.
- Top, any trigger level will start the top most layer.
In all 3 cases, the resulting layer amplitude will be as the trigger
amplitude.

Deve.

[Slap the drummer]

Hi Deve -

But can you tell us what you think is (roughly) the optimum number of layers??

I suppose some research went into it when the kit was being developed..... 

thanks

[pro424]

I'd like to bump this one. I sampled a cymbal of mine, and for some reason the edge hit sounds good, but the bow and bell have low volume, compared to other crashes, even though i haven't done any kind of normalizing or stuff.

Also it would be interesting to here how many samples is optimal from more persons.

[fishmonkey]

i don't think there is any easy answer, as it depends on how the recording/sampling was done.

in general more layers is better than less layers however there will be some point where you can't hear the difference any more, depending on how sensitive you are, and how you play. and for more layers to be useful, the sampling needs to have been done very sensitively and methodically for all those layers to be meaningful...

[Jovato]

I'm starting to think it depends on the number of layers that the sample library you are recording used in the original sample.  I have sampled a couple of kit pieces from Addictive drums and found that 15 layers gives a notably more natural response curve than 24 layers.  With 24 layers I got weird random spiking at the top and little dynamic variance in the middle.  Addictive only has 12-15 layers usually, and the rest are horizontal samples.  I am about to try to sample a couple of pieces from BFD Eco this morning.  Those kits all have 24 layers and no horizontal samples.  I'm guessing 24 layer dsnds should be a perfect match.  I'll let you know.

Edit.  Some further exploration of this is provided in this thread.
http://www.2box-forum.com/index.php?topic=1398.0