Wondering about the 1ms Attack minimum on the Sampler

[bitvoid]

Hello,

I've been enjoying working in Bitwig and especially the beta with the new sampler features!

Though I've always notice the 1ms attack minimum on the attack and decided to dig in to it a little bit and check for any audible differences. I did some null tests with sharp transients and tried various different configurations and wasn't really able to get consistent results - though I didn't actually perceive any issues with the attack in my testing.

I just tested some things briefly so I won't say one way or another about any issues I may or may not have found because I wasn't super thorough, so I don't won't to make any inaccurrate assumptions/statements.

That being said, what are others' view/perceptions of the 1ms attack minimum on the sampler? To my ear, I don't hear any differences in the transients for eg. chopping breakbeats. And from my tests, I couldn't say for sure I found any noise or artifacts on the attack in my null tests. Also, have the devs ever had any input on this matter. I've checked other samplers and they go to 0ms, though some say that under the hood they never actually do.

As well, I've seen online some mention to put it in 1-shot mode and use a ADSR device on the volume, though I was most definitely getting clicks on the transients when doing null tests with that method.

Anyways, your thoughts/input would be greatly appreciated. I'm inclined to just continue on with my workflow using the sampler to chop breakbeats as I haven't actually heard anything amiss, though if anyone or the devs were inclined to chime in on this it would be nice for peace of mind anyway

thanks!

[Benedict]

Once you go below 1ms in an Env or LFO movement, you start to get clicks/pops. Known as DC Clips/Pops and not uncommonly used in early drum machines for initial impact/cut. BUT they can be really problematic and seeing many won't understand that hard transitions are an issue synths often default to 1-5ms and even take 0ms off the table to avoid people thinking something is wrong with the device.

We also have to consider how doable a full 0ms actually is as everything takes time so while a knob can read 0ms, how close-far from a 'zero kelvin' is that action really? Perhaps a really experienced person in DSP coding etc could help answer that one.

[bitvoid]

Thanks Benedict,

Yes, I've sort of come across similar conclusions and reasoning. As well, there seems to be the case made that a true "0" time is still being affected by the block size. Anyway, it's ( I don't think) a huge issue, and I couldn't really get anything audible to cause me concern, though I had always took notice of it and thought it worth following up.

I'd still be interested to hear/learn more if there's anything to know, though really, I just want to make sure the transients on my sample slices are getting retained. I did notice that if I moved the sample start position on a kick drum sample 1ms, it was actually a significant portion of the transient - though bouncing it to audio it only had the slightest bit of smoothing on the waveform in a range of less then a 100 samples or something (don't quote me on that).

Anyway, thanks for the info - cheers

[pdxindy]

You can use the Sampler Module in the Grid and test with different env's...

[bitvoid]

You can use the Sampler Module in the Grid and test with different env's...

just tested in Grid - on a quick look - very snappy. Using a breakbeat in sliced mode, if I use the ADSR and set the DSR all to 0, then move the Attack between 0 & 1ms, there quite a significant amount of sound in that range. Is that all DC offset that is better off being smoothed...I don't know but it is interesting - thanks for the tip!

[Benedict]

Thanks Benedict,

Yes, I've sort of come across similar conclusions and reasoning. As well, there seems to be the case made that a true "0" time is still being affected by the block size. Anyway, it's ( I don't think) a huge issue, and I couldn't really get anything audible to cause me concern, though I had always took notice of it and thought it worth following up.

I'd still be interested to hear/learn more if there's anything to know, though really, I just want to make sure the transients on my sample slices are getting retained. I did notice that if I moved the sample start position on a kick drum sample 1ms, it was actually a significant portion of the transient - though bouncing it to audio it only had the slightest bit of smoothing on the waveform in a range of less then a 100 samples or something (don't quote me on that).

Anyway, thanks for the info - cheers

Just be aware that moving a sample start by 1ms is not the same as an env running at 1ms. The move is literally chopping off the first x, the env is opening over that time. Even if linear, that means that the first ms of that sample is already 50% level by 0.5ms. If the env is shaped, that would be more like 75% by 0.5ms. A bit pedantic perhaps but really changes what we perceive. Try it yourself with an env with shapable ADSR (and a real osc preferably) and you may be surprised how the feel changes. This is what matters and why people want old devices for those decisions made, e.g., the same ppl who abscess over 24/96 oversampled 16x will pay out the nose for a 12 Bit Sampler with a "snappy" 5ms env.

Ultimately tho, worrying about such tiny details gets in the way of what matters.

[Tj Shredder]

If you cut a dc with 0ms attack or with 1ms attack, you will get a click. They sound different but I wouldn’t want any of them. If you want that 0ms click, simply add 1ms silence to the sample and you get it.
Personally my minimum attack would be 10ms…