Time-aligned sound samples

Question

if you were to provide an application and a set of samples that a user can use to compose rhythms to be played back by a sequencer, whithout having to care about time-aligning, how would you save those samples?

Percusive instruments seem to align to the beat at the start of the sound (at the onset) but, is this the case for all instruments?

Should I align my samples at an offset, e.g. 100ms, leaving leading space for sounds that align with the beat after some sound has already been made?

Should I dynamically align the samples? E.g. is it correct to say that the sample aligns with the beat at the point where the sound has its peak volume?

Thank you.

Answer 1

If a human is playing the samples, I would prefer it if this latency was consistent, and as close to zero as possible.

Whenever I make my own samples, I zoom right in and trim any 'silence' from the start of the sample. All my samples are saved like this.

I'm not 100% sure if I would notice 100ms of silence at the start of my samples, but I would expect them to play as soon as I'd triggered them. Some smartphone apps of this nature are unplayable because they're not responsive enough.

You can always factor in a quantisation feature if your sounds work best when absolutely synchronised. (delay the playback if triggered early, play sample from just-after-the-start if triggered late, like aligning all the sounds to a 'grid')

---

When sequencing, it's just as important to have samples aligned correctly too. The easiest way to do this is also to trim any silence from the start of your samples. That is usually all that is necessary.

If you find some of your samples are 'out of time' even with the samples trimmed, then you're either:

• Playing the sample at the wrong time. Many sequencers allow you to offset the trigger for sample playback, e.g. not starting immediately, but after a few 'ticks'. This is one way of achieving a 'triplet' when sequencing in 4/4 time.

• Playing the sample at the wrong rate. Can be true of longer, rhythmic samples.

• Playing a sample that contains leading noise e.g. reverse cymbal, in which case, it's being triggered at the wrong time, you can't add a delay to all the other samples to compensate, you simply have to sequence it to play earlier.

Should you dynamically align the samples? That's very ambitious, you'd need to calculate a known-offset for each sample, either pre-calculated or on-the-fly. It might also take a bit of getting used to for the user, e.g. if I placed my reverse cymbal sound where I wanted it to peak rather than where I wanted it to begin. I guess it's no different than left/center/right justification for word processors.

It might actually be a cool feature, but I wouldn't want it to be dynamically calculated (but an operation to calculate a suggested point might help), I would want to be able to set my own 'alignment markers' to control where the sample starts, since playing a sample like this at different pitches/rates would all require them to be started at different times.

Answer 2

Maybe I'm not understanding something about the question, but why do you say non-percussive instruments don't align to the beat?

If I'm playing a MIDI keyboard, for example, I want the sample to begin playing as soon as possible when I press the key. If there's any kind of appreciable delay, it throws off my performance. Especially at higher tempos. The onset of sound should occur as soon as possible.

If you're thinking in terms of a typical ADSR envelope, the peak volume will occur at some brief time after the initial onset of sound, and then decay from there (this initial period of sound, between the onset and the peak is the Attack Phase). So maybe you're asking if the onset of the sample should precede the beat by the length of the attack phase? (so the peak occurs on the beat?). I don't think this is the case, since it would imply that musicians are actually constantly playing slightly ahead of the beat.

Furthermore, it will probably depend on what instrument you're using -- lots of instruments don't actually follow the ADSR envelope. Basically, any bowed string instruments, woodwinds, or brass are capable of starting at a very soft pitch, and gradually increasing in volume over the entire length of the note, so that the peak comes right at the end. Or they could even swell and fade multiple times.