Real versus Sampled drums: the neglected flamewar.

[Dasheesh]

where do you stand on ursine appendages?

LoL that was as one I had to look up. Doesn’t Happen often. From what I’m reading.. I don’t have a say one way or the other. It’s in God’s hands.

[whyterabbyt]

where do you stand on ursine appendages?

mostly on the ursine feet.

[aciddose]

It gets a bit more iffy with things like hihats and cymbals, which ring out for a lot longer. Snare too, to a certain extent, depending on the snare setting. If the drum/cymbal is still ringing, and you hit it again, it behaves differently than if it is not vibrating. So you lose a bit of that, and you lose the ability to play all the little hi-hat games with the pedal. You can get close though with a good sample library and judicious programming. Close enough, that most of the time it doesn't matter.

things are better now with BFD3 cymbal swell modeling.
works for the hat as well

How does that work exactly (not being a BFD3 owner)?

It just does. I have no idea what they did.

one assumes it models the dynamics of hitting something that's already in motion instead of every cymbal hit a hit of a cymbal in repose.

I used to do a lot of work around this issue, no more.

One possibility on this BTW is that you sample not just single hits, but variations on multiple hits in series. You can then do things like two hits where the last hit is allowed to ring out and you manually (massive amount of effort) cut the sample in the right place and catalog the timbre/intensity of that variation.

The difficult part is then interpreting the input sequence to find where a similar pattern is encountered, and filling in with the correct samples taken from that same type of sequence of hits rather than isolated hits.

Once accomplished and understood, simplification can take place and I believe one would find a sort of revelation is experienced where this effect is really not at all as complex as might first be assumed. That would become the "model", and reproducing the effect from a certain collection of samples would be the next step.

It would be very interesting to understand how they accomplished the task without offline processing (or knowing drum patterns long in advance.) One solution would be to have the plug-in itself handle the sequence entirely inside the plug-in rather than being driven from an external source via MIDI or otherwise. This would allow "predicting the future" and near real-time accurate modelling of the associated phenomena.

I say "near real-time" because you would likely be stuck with building short sequence segments and have the ability to switch between them via external control in real-time, but not on a dynamic hit-by-hit basis.

Any model that did work in a fully dynamic real-time manner would not be accurate, and we would then be forced to admit that accuracy is clearly either less relevant or completely irrelevant vs. a rough model of the phenomena; not just to only the casual listener but to anyone, even an experienced drummer.

[fluffy_little_something]

For Pop, HipHop, R&B etc. sampled drums are usually enough, or synthetic drums even. But for Jazz, Latin and such more demanding genres, programming drums so that they sound alive and authentic is much more difficult.
In many cases only a drummer will notice the difference.

[aciddose]

For Pop, HipHop, R&B etc. sampled drums are usually enough, or synthetic drums even. But for Jazz, Latin and such more demanding genres, programming drums so that they sound alive and authentic is much more difficult.

To sequence the drums accurately enough requires an understanding of drums, so this fact does not prove anything about the core issue of realism but rather is a basic observation about the reality of an experienced professional vs. those who are unqualified.

[fluffy_little_something]

People can also get that understanding from decades of listening experience.

I think drums are much easier to emulate via samples than guitar or bass.

[aciddose]

Any model that did work in a fully dynamic real-time manner would not be accurate, and we would then be forced to admit that accuracy is clearly either less relevant or completely irrelevant vs. a rough model of the phenomena; not just to only the casual listener but to anyone, even an experienced drummer.

Just to be clear what I'm referring to here: I'm talking about the variations emergent from multiple strikes like velocity that are not included in the sequence.

If the modelling process was all about reproducing the exact input pattern (as if you had measured the impact velocity and direction exactly from a real drummer), the model wouldn't then be part of the plug-in but rather provided by the person doing the sequencing.

In order for "model" to make sense we'd need to be talking about sending an input of the raw sequence (position, timing, intended velocity) without that kind of extremely precise "micro management"esque information of actual individual strike parameters (exact actual velocity: speed, acceleration during the hit/contact, direction and 3d position of both the instrument and stick.)

The model may be simply a model of a hihat or cymbal that will perform the same as a real one, which is a lot more trivial. This though means you would need to understand the influence of the drummer on the performance: you'd be the drummer, there would be no "drummer + hat/cymbal" model at play to fill that information in for you.

[Dasheesh]

Yeah... programming drums... just hang me with a neck tie. I envy the guys/gals that can get off on it. At this point it’s mostly just a “let’s throw shit together until I find something that works” event.

[fluffy_little_something]

Any model that did work in a fully dynamic real-time manner would not be accurate, and we would then be forced to admit that accuracy is clearly either less relevant or completely irrelevant vs. a rough model of the phenomena; not just to only the casual listener but to anyone, even an experienced drummer.

Just to be clear what I'm referring to here: I'm talking about the variations emergent from multiple strikes like velocity that are not included in the sequence.

If the modelling process was all about reproducing the exact input pattern (as if you had measured the impact velocity and direction exactly from a real drummer), the model wouldn't then be part of the plug-in but rather provided by the person doing the sequencing.

In order for "model" to make sense we'd need to be talking about sending an input of the raw sequence (position, timing, intended velocity) without that kind of extremely precise "micro management"esque information of actual individual strike parameters (exact actual velocity: speed, acceleration during the hit/contact, direction and 3d position of both the instrument and stick.)

The model may be simply a model of a hihat or cymbal that will perform the same as a real one, which is a lot more trivial. This though means you would need to understand the influence of the drummer on the performance: you'd be the drummer, there would be no "drummer + hat/cymbal" model at play to fill that information in for you.

But do those multiple strikes even matter with simpler music such as Pop, HipHop, R&B etc.? I don't think so.
I was just listening to Sade's Sweetest Taboo, it sounds authentic and alive, but I think it might have been programmed, not sure...

[aciddose]

They do, which is why 16ths or 8ths hats in electronic music sound so electronic. If you listen to any live kit/synth you should notice 8ths and even quarters almost always occur during the decay of the cymbal (hihat or otherwise) and the difference even between a "real live tr-808" vs. a single sample is well known and acknowledged in the computer music scene.

Just like recreating an authentic tr-808 requires some modelling and many samples or a tr-808 cymbal pulses + filters synthesis process, real cymbals require exactly the same thing.

It is very easy to tell when a single sample has been used and yes, it is very common but it clearly sounds worse than many samples or a real "live" drum machine.

Why would the same not apply to real cymbals? It clearly does unless you're using a single sample like a tr-909 or similar where the effect is intentional. I would argue that if you were to take any real tr-909 recording and replace the hihat track with multiple samples, it would instantly sound way better. I'm not talking subjective tastes or anything like that, I'm talking clear and obvious differences.

The majority of the most popular electronic music uses layered cymbals, often with many different samples and sounds and these are used differently at different points in the track to add interest. Only rarely do you hear popular music (chart toppers) where you have a loop with consistent un-modulated cymbals played over and over throughout the track, even in rap and r&b.

If you listen carefully, even in instances where the hihat sounds very similar you should notice small variations in intensity, attack and decay. The authors of this music take great pride in producing the ideal "loop" with these sorts of variations tweaked to perfection. I would argue that this attention to detail is what often makes a significant difference between the unskilled and unknown vs. those who are most successful.

That said, we're not talking about that in this thread, not even close. Would you use a tr-909 in a country track?

[Dasheesh]

Well, first acid... the 606 has the best hats. Even though the 808 is very good. The reason these machines have been such a standard is because they sound good. They do. In thier own right, as a type of sound. In fact they are a sound you can’t get from live drums. The problem isn’t the sound so much as the programming aspect. Sequencing drums is flipping BORING!

[jancivil]

I have to have a good high energy level to get in there and program drums at this point. But I get deep into it.
I started with drums (leaving the thwarted primary school trumpet episode aside).
I have ideas for drums. When I had the zendrum and more computer I would play much of it in live. I'm doing pencil tool in the key editor with this pitiable setup. I had a fantastic drum track last August, computer stolen right after I finished the thing and that's gone forever.

drums in last section of this monstrosity
viewtopic.php?f=14&t=498913&start=15

[jancivil]

you wanna make idiots dance?

Not so much.

[herodotus]

every argument i get involved in lately, someone brings out the robot playing the glockenspiel!

It's a robot dog whistle. Like in the movie Serenity, where Summer Glau sees a commercial with a subliminal message and suddenly starts beating the crap out of everyone.

In fact, we should probably stop mentioning the robot glockenspiel meme altogether.

[jancivil]

Summer Glau