A loop makes a promise: play me again and nothing changes. Most hold that promise for a bar or two. Leave one running under a vocal for a minute, though, and the seams start talking. There’s a tick on every downbeat, a hi-hat that vanishes, a groove that stiffens the longer it repeats. Nearly every “bad” loop traces back to a small, fixable error: an edit in the wrong place, a length that doesn’t quite match the bar, a tail sliced mid-decay, or a timing quirk that only becomes obvious once it happens thirty times in a row.
Here’s what to check, roughly in the order things tend to go wrong.
The click at the seam
Nothing exposes a loop faster than a click on every repeat. It usually comes from one of two sources: the waveform doesn’t line up across the join, or there’s DC offset pushing the whole file off the centre line, so even a careful cut lands on a non-zero value.
Zoom in hard on the edit and snap it to a zero crossing, the point where the waveform crosses the centre. That alone kills the worst of it. If a faint tick survives, add a very short crossfade across the seam, somewhere in the region of two to five milliseconds. That’s brief enough to be inaudible on drums and guitars. Be more cautious with sustained bass and sub-heavy material, where longer fades can cause a dip or a phase wobble, so keep them short and check the low end on headphones and on speakers.
Before any of that, run a DC offset removal on the source file. Most editors have one built in, it takes a second, and it prevents a whole category of clicks you’d otherwise chase around the timeline. It’s also worth checking whether a plugin on the export chain is adding a few samples of latency — some limiters with lookahead do — which shifts the whole loop sideways against the grid.
Loop length maths you can’t skip
A loop that is four milliseconds short is a loop at war with the click. At 120 BPM a bar is exactly two seconds, so four bars is eight seconds. Cut that by eye and you’ll be a few milliseconds out, and that error repeats every cycle. Thirty repetitions later you’re a tenth of a second adrift, which isn’t feel anymore, it’s a flam.
Set the length mathematically rather than by dragging. At 44.1 kHz, two seconds is 88,200 samples, so four bars at 120 BPM is 352,800. You don’t need to do the arithmetic by hand — let the DAW snap the selection from bar line to bar line, then trim to that selection rather than nudging the edges until it looks about right.
Two more length traps worth a mention. First, cutting three bars when the phrase needs four, or vice versa, which produces a loop that either stumbles early or leaves a gap of silence. Second, cutting to the last hit instead of the end of the bar, so the final beat is clipped. Select the whole bar, including the empty space at the end of it.
Cutting through tails, notes and one-shots
Reverb and delay tails are where most otherwise decent loops fall apart. If the last snare in bar four decays into nothing and you cut precisely on the bar line, that decay never arrives — the loop ends dry and restarts with a bang. The fix is to render the part with its tail, then fold the tail back over the start of the loop so the decay from the final hit continues into bar one. Paste it sample-accurately; a millisecond of misalignment here will smear the attack of the first beat.
The opposite problem is the one-shot sitting on beat one: a crash, a riser, a vocal ad-lib, a tom fill. Repeated every four bars, that crash sounds like a machine. Three options: remove it and add it in the arrangement instead, rebuild the loop so the one-shot lands at the end rather than the start, or accept the loop as a single-use section rather than something to repeat.
Timing that stiffens or drifts
Over-quantising is the most common culprit. Nudge every note to the grid and you remove the small inconsistencies that made the part feel played; the loop then sounds like a diagram of the performance. Quantise lightly, or keep the original feel and fix only the notes that are genuinely wrong.
Watch for groove applied twice. If the source part was already swung and you then drop a swing template over the loop, you get a lurch. Equally, a part recorded with monitoring latency may sit a few milliseconds late against the grid. If it’s consistently late, shift the whole thing by the same amount. Don’t correct each note individually — that’s how a tight performance becomes a stiff one.
Live takes are their own case. If bar four runs fifteen milliseconds longer than bar one, cutting and repeating it produces an audibly late join. Either warp the section so it holds steady, or pick a stretch that already sits well, or use it once and move on.
Monitoring mistakes
Judging a loop from one pass at high volume hides almost everything. Work through these instead:
- Play it at least sixteen times, ideally under a vocal or a melody, and listen for anything that starts to irritate on the eighth repeat.
- Check it in mono. Phase problems that are inaudible in stereo become obvious the moment the channels collapse.
- Listen quietly. Ticks and clicks hide at volume; they don’t at low level.
- Check the first cycle separately from the rest. Some errors only exist at the join, not inside the phrase.
- Test it against a click track at the intended tempo. If it drifts, you’ll hear it within a few bars.
Render and file slips
Small export errors undo good editing, and they’re easy to miss because the loop sounds fine in the project and wrong everywhere else. Fade-in and fade-out boxes left ticked will shave the attack off your first hit and the decay off your last one. Normalising every loop to the ceiling makes them inconsistent in level and leaves no headroom. Bouncing 48 kHz material into a 44.1 kHz project, or exporting a loop as a compressed file, adds smearing you can’t remove later. Export at the project’s sample rate, in a lossless format, with the same bit depth as the session.
A last pass before you commit
Once the loop sounds right, do one final sweep, and do it in the same order every time so nothing gets skipped.
- Check the loop’s start and end points in samples, not by eye, and confirm they match the bar length at your tempo.
- Play it for a solid minute against the click and something melodic.
- Listen in mono at low volume for ticks and phase issues.
- Confirm the tail from the final hit carries into the first bar rather than stopping dead.
- Check the exported file for fades, silence at the head, and the correct sample rate.
- Name it with the tempo and key so the next session doesn’t guess.
Do that and the seam disappears, which is the whole point. A good loop isn’t one that sounds impressive on the first pass; it’s one nobody notices is repeating at all.
Photo: theglassdesk / Pixabay
