What's new

Realidrums: no choke?

If you have a cymbal in it's own instance of Kontakt, and then set the maximum number of voices in that instance to 1, you'll be able to choke the cymbal by hitting it a second time with a low velocity. The only downside is that it won't sound as natural as a player actually grabbing the cymbal.
 
I thought I had added chokes in the 2.1 update, but checking just now ... we didn't. We're working on a 3.0 update that includes chokes, although that obviously doesn't do you any good now.

If you have the full version of Kontakt, here's a workaround for now. Click the wrench icon to open up the editor, then open "Group editor." Then find and select which group has the crash you're playing. How???, you might ask? You can tell which group is used by seeing which speaker icon is lit when you play a crash note. (See image below. That speaker icon is handy!)

Once the correct group is selected, scroll to the bottom to where the ADHSR envelope setting should be. The obvious thing to do is shorten the release, but that won't work because since these are all drums, I set them all to "Play entire sample," which makes the Release setting irrelevant. D-oh!

So instead, we need to use the Hold, Decay and Sustain settings. Set Sustain to 0 (actually negative infinity), then Hold for some time between 100 and 250 ms (depending on how quick you want to "grab" the cymbal), and set Decay to taste, most likely under 200 ms.

The downside is that group is now going to always choke that particular cymbal. Also, of course, the choke is artificial, so it won't sound as good as sampled chokes. Sadly, I didn't think to sample real chokes, and I'm pretty sure that even in the 3.0 update, those are artificial, too, although there is some new content and we might have done some chokes.


How to choke in RealiDrums.png
 
Just thinking outloud here: it might be possible to fake pretty good chokes by taking a few snippets of audio along different times in the cymbal's decay, quickly fading each one out and fading in a later one, and applying an envelope to fade out the whole combined blob of sound.

Could also do it with smaller grains, but I'd just try 5-6 chunks of around 0.2 seconds first. It might be that simple. If the first chunk is from earlier in the file than the moment when the cymbal's being choked, it should do a decent job of making the sound get momentarily a bit brighter, which a real choke seems to, from looking at their spectrograms.

I got a few cymbals with chokes sampled, I could try this sometime and see how close it comes to the real thing.
 
Just thinking outloud here: it might be possible to fake pretty good chokes by taking a few snippets of audio along different times in the cymbal's decay, quickly fading each one out and fading in a later one, and applying an envelope to fade out the whole combined blob of sound.

Could also do it with smaller grains, but I'd just try 5-6 chunks of around 0.2 seconds first. It might be that simple. If the first chunk is from earlier in the file than the moment when the cymbal's being choked, it should do a decent job of making the sound get momentarily a bit brighter, which a real choke seems to, from looking at their spectrograms.

I got a few cymbals with chokes sampled, I could try this sometime and see how close it comes to the real thing.
That's an interesting idea. If you do try that, please let me know how it sounds.
 
That's an interesting idea. If you do try that, please let me know how it sounds.
Tried it, seems pretty decent. Attaching the results.

First is the crash with sampled chokes - two quick chokes and one slower one on a lighter hit. Then the same thing with reconstituted chokes. Decent enough, I think. Then four sampled chokes and four fake ones by themselves - there you can tell they're not quite the same, but for a half hour of trial and error this isn't bad, and it'd be possible to do something much better. Basically, there are three chunks taken from a hard crash hit. More doesn't seem necessary.

First chunk skips the attack and starts playback 3000 samples into the file, fades in in 0.15 seconds, fades out in 0.45, 30% volume.

Second chunk starts 8000 samples into the file, fades in in 0.15 seconds, out in 0.75, 30% volume.

Third starts playback 0.15 seconds later, starts 320000 samples into the file, fades in in 0.15, out in 2 seconds, 50% volume.

All three get a lowpass filter with an envelope with a hold time of 0.3 seconds and 2-second decay.

This would chew up a lot of polyphony voices once you got into multiple room mics etc, but I think it's a decent approach. Make it more sophisticated with awareness of how hard the original hit was and how much time has passed since the hit, and this could actually be much better and more flexible than sampled chokes. Could even make the envelopes longer, for gently muting a cymbal rather than grabbing it quickly.

Here's the SFZ I used for the quick test:

Code:
<master>
key=51
amplitude_cc101=30
locc101=1
pan_cc102=100
pan_curvecc102=16
tune_cc120=1200
ampeg_sustain=0
fileg_hold=0.3
fileg_decay=2
fileg_sustain=0
fileg_depth=14000
cutoff=20

<group>
offset=3000
group=61
ampeg_attack=0.15
ampeg_decay=0.45
<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr1.wav

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr2.wav
seq_position=2

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr3.wav
seq_position=3

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr4.wav
seq_position=4

<group>
offset=8000
amplitude_cc101=30
ampeg_attack=0.15
ampeg_decay=0.75
<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr1.wav

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr2.wav
seq_position=2

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr3.wav
seq_position=3

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr4.wav
seq_position=4

<group>
offset=320000
amplitude_cc101=50
ampeg_delay=0.15
ampeg_attack=0.15
ampeg_decay=2
<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr1.wav

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr2.wav
seq_position=2

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr3.wav
seq_position=3

<region>
sample=../Samples/crash_17/cr/oh/cr_vl5_rr4.wav
seq_position=4
 

Attachments

  • choke_test.mp3
    578.6 KB · Views: 21
Hmmmm.... sorry but after all your hard work, I have to confess those examples don't sound like chokes to me. I think one would be better off drawing in an instant volume drop curve and simply letting a little room sound help smooth out the abruptness. It probably wouldn't fly for an exposed cymbal solo, but I think it would sate most ears in a standard mix.
 
If the sampled chokes don't sound realistic to you either, that would be good news - it would mean that we're not missing out on anything by not actually sampling chokes. If a simple fadeout is as good as it's gonna get, that would be really good news, because that's really really easy to do.

Or maybe a fadeout plus a tiny little sound from earlier in the sample to add brightness. That would also be simple. I am convinced a choke makes the sound momentarily brighter, though, especially when it's a quick choke very soon after the hit. Gentle, gradual chokes are probably mcuh closer to a fadeout. Too bad I don't have the raw samples I cut those chokes out of, or any cymbals laying around at the moment.
 
That's a good test, and thanks for posting it. I don't think it sounds better than the basic volume envelope I'm using now, though, which I was surprised that it actually isn't too bad at all.
 
The only chokes I've ever played are full ones, where you grab the cymbal and the note is instantly done except for some lingering vibration that probably wouldn't be detectable in a full mix. A gradual choke would be harder to pull off, and would have to be either sampled (ideally) or recreated with envelopes and filters.

In the near future, I will be taking on the task of recording cymbal chokes and implementing them as release samples, because in some styles of Polish polka music, continual cymbal choking is a technique sometimes used on the backbeat.
 
Yeah, simple fadeouts are surprisingly decent. Doing somewhat better is possible if you really need to do, say, big exposed crash chokes for hard rock, where the choke sound will be really really obvious.

I think sampling chokes will work out pretty well for backbeats since the duration between the hit and choke is decently consistent. Making a sampled choke that was recorded a second after a hard whack work 10 seconds after a softer hit, that's where it starts to sound too obvious, so you make it quieter, and that just gets you back to the point of sounding like a fadeout. But for polka backbeats, I think a few samples should work fine and more realistic than a fadeout.
 
Top Bottom