Acid Dub: What Happens When a 303 Meets a Dub Delay

Every note to the delay

Program a squelchy sixteenth-note line, turn the delay send up on the whole track and it sounds like an idea for about four bars. Then every note has an echo, every echo has an echo, and the line has become a wash of resonance that has swallowed the kick and the bass. The delay isn't the fault. A 303 pattern already fills the grid, and dub works by leaving space. Until you put space back in, the line can't behave like a dub instrument, and the place to do that is the send. (If you've read the dub techno piece, this is the opposite problem: a chord needs echoes to go anywhere, while a 303 line needs most of its echoes taken away.)

One bar, two throws and a gap

Here's an invented pattern to argue with: one bar of sixteenth steps in A minor at around 120 BPM, with the notes and their treatment written out. It's an illustration, not a transcription of any record.

Steps 1 and 3: A1, plain.
Step 4: A2, accented.
Step 6: G1, plain.
Step 7: A1, slid into C2 on step 8.
Step 10: A1, plain.
Step 13: E2, accented, the last hit of the phrase. Steps 14 to 16 are empty.

Set the delay to a dotted eighth, three steps, and think of each note as a source and each repeat as a copy that only speaks after the source has finished. Send step 13 only. Its first repeat lands on step 16, in the empty tail, so the line answers itself. The second lands on step 3 of the next bar and the third on step 6, quieter by then, tucked under notes you wrote. The phrase ends with a comma instead of a full stop.

Send the accent on step 4 instead and the repeats land on steps 7, 10 and 13: the slide, a plain note, then your other accent. The first three copies each have a note to hide behind, and only a fourth, quieter one, on step 16, falls in a gap. You have doubled the line at a lower level. It is the swamp again, just slower to arrive.

Now the gap. Steps 14 to 16 are silence, and that is worth featuring. Push the send level or feedback up on step 13 and let the bar turn, and the space is filled by the echo of a note you played, not by more notes. Mute the dry line for the following bar and the echoes carry on alone. In both cases the rhythm of the track changes, not just the length of the notes.

What the echoes are allowed to keep

A 303's filter is part of the note. Whatever the cutoff and resonance are doing at step 13 goes into the send, so the echoes carry that moment's brightness, and opening the filter later doesn't reach back to them. Filter on the return and it's the opposite: every echo changes with the knob, including ones already in flight. Closing a return's low-pass mid-bar is a mixing move that re-voices echoes the synth already played.

Neither is filtering inside the feedback path, where each repeat is filtered again and darkens one after another. A fixed EQ on a return thins every echo by the same amount. Sound On Sound's 2007 dub-mixing tutorial (external site, opens in a new tab) builds its loop on a desk by feeding the return back into its own aux send, which puts an EQ on that return inside the loop, and suggests cutting highs there to tame the phasing digital delay can produce. In a DAW, an EQ after a delay plug-in usually sits outside the plug-in's own feedback, unless the delay has a filter in its loop.^[1]

Then there's the low end. A 303 doesn't have to play the bass, and low notes that get echoed pile up under the kick and the real bass as sub-frequency nobody mixed. Keep the A1 and G1 out of the send, and high-pass the return around 200 Hz, with the low-pass somewhere near 5 kHz, so the echo is the squelch and the dry note supplies the weight. Both numbers are starting points. Keep feedback modest, around 30 to 40 per cent, and put a limiter on the return; the same tutorial advises one because echoes can build until they distort. Whether the 303 itself is a stock mono synth or a hardware clone matters much less than any of this. If you do want to compare 303 emulations, that comparison is separate and about something else.^[1]

Play the delay, don't sync it

Dotted-eighth arithmetic is tidy, and it isn't the only way. The Acid Dub Studies albums by Om Unit are a well-known meeting of 303 and dub delay, though the idea doesn't start there: in a 2024 Ban Ban Ton Ton interview (external site, opens in a new tab), Om Unit names Andreas Tilliander's acidub album as TM 404 as an influence and says they're still digging for others who did something similar. Treat acid dub as a meeting point, not a fixed genre.^[2]

What Om Unit says is useful. Asked about processing, the artist calls the digital delay the project's "sonic identity" (external site, opens in a new tab) and says repetition adds an extra melodic layer "if timed right" (Sonic State, 2025). For the live show, Om Unit describes a delay that isn't tempo-synced and that can be manipulated by hand (SSL, 2025 (external site, opens in a new tab)); that is a stage rig, and the article doesn't claim the albums were made that way. The point for a DAW is that turning the delay time by hand slides repeats off the grid, which is a different effect from any quantised answer, and sometimes the right one.^[3][4]

Roland's article on taking the TB-303 beyond acid (external site, opens in a new tab) advises applying accents and slides sparingly when nudging a filter sweep. That fits here: an accent worth echoing is one that stands out, and if every other step carries one, none of them reads as a call that gets an answer.^[5]

Try it tonight

Load a mono synth with a resonant low-pass, write the pattern above, and send everything to a fully wet delay on a return. Listen to the swamp for a minute. Then pull the send down to nothing, automate it to open on step 13 alone, and hear the tail fill. Move the same throw to step 4 and hear the line double instead. Last, high-pass the return and mute the dry line for a bar. You don't need a 303, a Space Echo or a new plug-in. You need to decide which note gets an answer.