Sample playback loops when not expected, and there's a slight click in playback

[ericclack]

Hi,
Apologies, this is perhaps not an issue, more of a question.

Is this expected behaviour: when I playback a freesound sample with release option the sample loops? I was hoping to use this option to fix a slight click at end of sample playback.

(def stick (freesound 82280))

(stick) ; this plays with a click at the end

(stick :release 0.1) ; this loops a little 

(stick release 2) ; this loops a lot

The player in the file:https://github.com/overtone/overtone/blob/master/src/overtone/sc/sample.clj

[ericclack]

From conversation in Slack: https://clojurians.slack.com/archives/C053Q1QSG/p1546177050008000

A simple player has no click sound -- and obviously none of the other features of the regular sample player:

(def stick-buff (load-sample (freesound-path 82280)))
(definst stick [] (play-buf 1 stick-buff))
(stick)

I'd like to fix the looping and click and submit a pull request.

Is anyone willing to review and comment on my work, when ready?

Cheers,
-Eric.

[hlolli]

I will review it. But how are you thinking about solving non looping of sample envelope if the total envelope time is greater than one? How do we know if this isn't used in many music tracks already and we'd be making breaking changes? Note that release time is time added to an envelope, if you want to shorten the sample, you may want to decrease the sustain time. Are you sure that this click in your stick sample doesn't also appear when you play it in a DAW (some players like VLC may have anti-click faders on pause/end of media).

[hlolli]

Behind every sample, freesound or non-freesound, there's a buffer id, that you can use with a phasor to iterate over the samples in the sample. You have total freedom to manipulate it the way you want. I'd say, we should stick with the default behaviour of Supercollider, which is gated based envelopes. If a gate triggers after the whole sample is played, then it loops, except when looping is turned off in PlayBuf.

[ericclack]

Thanks for your comments and advice. I don't have any answers to your questions right now, but it's good to know the broader requirements. I'll have a think and post my progress to this thread...

[ericclack]

I've a question: what's the advantage of using the phasor and buf-rd over the simpler play-buf? Is this to support the variable start and end points? Could this be done with a envelope instead?

Here's where I've got to in my own repo:
https://github.com/ericclack/overtone-loops/blob/68116ddceb9bdb8917529882bbad95b6a90ae51b/src/overtone_loops/loops.clj#L128

I'm way off a Pull Request and really just experimenting at the moment. Comments welcome.

I think my next step is to explore how widespread the click problem is with the standard implementation (e.g. trying some other computers), and looking at documenting the expected behaviour with the release and loop? options.

[hlolli]

Clicks are inherently constant problem in every audio signal. For example a punchy kick with low attack time has always click, but it just dies in the loudness other things happening. An poorly enveloped audio recording will click, just because the spatial sounds attack otherwise silent speaker on full force. So declicking is something I always bear in mind, a battle of tradeoffs that never ends. The advantage you ask about, doesn't matter if you play the sample linearly, for granular effects, hopping back and forth on a sample, you may need more low level tool, like a phasor and buffer read.

[theJohnnyBrown]

I've had the same problem, also solved by straightforward use of load-sample.