Drum Multi-Samples: Round Robin vs. Velocities

[tommyzai]

Which is more important for virtual drum realism, round robin or velocity layering?

Obviously, having both is better, but there is software that only allows users to choose one of these features at a time.

[Teksonik]

I voted Round Robin because the samples can still represent different velocity strikes as well as positional differences in each strike.

There is more variation than locking a specific sample to a specific velocity range.

It depends on if the RR is truly random instead of just playing sample 1 then sample 2 then sample 3 and so on.

[tommyzai]

Teksonik makes an excellent point about Round-Robins.

Personally, I'm at a crossroads with drum multi-samples. I'm not sure whether to go all-in and do my best to simulate "realistic" drums or simply use one sample per instrument (kick, snare, toms, etc.) like an old drum machine, focus on performance, then ask a drummer to replace my "scratch track." I guess I need to decide if I'm making a demo, going for a pro final product on my own . . . or with help from a professional drummer.

[NAD]

If I want real sounding drums I'll go with a dedicated plugin like Addictive Drums. If I'm using individual samples then I'm probably after an oldschool sound so I wouldn't care about variations.
That being said, if I had to choose it would be a toss-up. It would depend on the actual beat I'm trying to program. If I have rapid repeated notes then round-robin would be more important, but if I want ghost notes for example on a snare or a kick then velocity layers would be the priority. So that's my non-answer

[tommyzai]

One afterthought about Teksonik's choice and post . . . Round-Robins need to be more similar in timbre than velocities or they would sound like a different drum with each strike, whereas velocities timbres are expected to be different.

[donkey tugger]

I voted Round Robin because the samples can still represent different velocity strikes as well as positional differences in each strike.

There is more variation than locking a specific sample to a specific velocity range.

It depends on if the RR is truly random instead of just playing sample 1 then sample 2 then sample 3 and so on.

I'd agree with this too. Nice to have both, but it's a lot easier to mimic softer hits (filter cutoff mapped to velocity, dynamic eq etc..) than it is to mimic variations by way of pitch/sample start etc.

[tommyzai]

There is another factor that supports RR over V . . . RR typically would include a collection of the louder strikes, which would generally have less noise. The softer V strikes tend to have a bunch of noise . . . not very noticeable when used as intended . . . soft strike = lower MIDI velocity level, but still . . . they are usually noisier.

[donkey tugger]

There is another factor that supports RR over V . . . RR typically would include a collection of the louder strikes, which would generally have less noise. The softer V strikes tend to have a bunch of noise . . . not very noticeable when used as intended . . . soft strike = lower MIDI velocity level, but still . . . they are usually noisier.

Judicious automation of transient shaping is also a good way to add to the feel of softer/harder hits.

[pljones]

I'd expect two RRs and six or seven VLs per strike, L&R per position (rim, edge, centre), per kit piece. Preferably stereo with a fairly dry mix (i.e. not much room mic). Hihats should have seven to twelve degrees of openness. All kit pieces should have muting (not a strike - cutting off the current sound naturally). Oh, and it shouldn't cost more than 3.99.

[enroe]

... Oh, and it shouldn't cost more than 3.99.

I think it should cost more than 4.49 and have cowbell!

[pljones]

Damn! Forgot the cowbell...

[DSmolken]

Some developers, like TK Drums, just do no round robins and a ton (like 127 for the core pieces/articulations) of velocity layers. With Virtuosity Drums we figured it's meant for expressive, non-repetitive contemporary jazz, so the kick and cymbals have round robins (because they're used to keep a steady pulse), but snare and tom stuff is just velocity layers and a big variety of techniques. Not 127, but "a lot for free stuff", anyway. So you can try that kit and see how you get along with the no-RR approach for the snare and toms.

Another hybrid approach is Soniccouture Moonkits - IIRC there are no true round robins, just a bunch of velocities, but repeated hits at the same velocity will use hits from neighboring velocities. But I might be misremebering that, if somebody feels like checking the walkthrough feel free to correct me.

I think it's very style-dependent, though. For death metal you'd probably want 8-10 RRs for the hard kick and snare hits because you'll want a lot of very-similar-but-not-identical hits as the core of the drum pattern there, though soft hits wouldn't need that much as they would probably not be used repeatedly. Moonkits and Virtuosity Drums are the opposite of death metal.

[elxsound]

You’re not going to get the best answer here and I’ll explain why.

What will affect making the right choice is your style of music, the number of hits of the same type of within a time frame that could overlap and how well it was recorded.

A good example of a bad result: There are some “respected” sample library developers that have promoted an insane number of samples in their kits (multiple round robins per multiple velocity layers), but the end result is heavy noticeable phasing issues on playback.

If your drums hits are sparse/with decent time between each hit (same drum part) then this is unlikely a concern.

*If this is software, can’t you record your passages and compare it on playback with RR enabled and then switch to velocity layers?

How it sounds is everything. If you can’t hear a difference on playback then that’s not a major problem or else you’ll hear what you prefer.

[kinwie]

Depends on how to use it.

For paino-roll style note input drum programming, which usually the velocity is the same, then Round robin is better choice.

For a playable kit i.e. on an edrum kit, more velocities is better choice

[jancivil]

Well, if something does not have both I'm not going to bother with it.

Sometimes more is not more, tho