Boards of Canada: Tape Hiss Won't Save a Weak Melody

Play it clean first

A tape-style plug-in is good at making every note sound as if it has a history. It can't decide which notes deserve one. Put an undecided phrase through a damaged signal path and you get an undecided phrase that has been somewhere. So write the motif on a plain sine, triangle or saw with every effect bypassed, and leave the texture alone until you'd be happy to hear the phrase on its own.

What the duo said, and what they didn't

The duo's own account is more specific than "they used tape". In Pitchfork's interview of 25 September 2005 ↗ (external site, opens in a new tab), Mike Sandison describes learning to compose by recording between two cassette recorders and re-recording until the sound had degraded. In a Clash interview about The Campfire Headphase ↗ (external site, opens in a new tab), he describes spending a lot of time "destroying the sounds, throwing parts onto overloaded tape", and re-recording them through low-quality mics. Marcus Eoin says they set things up one way and then try to make you hear the melody in "a completely different way", changing what the original melody does to give it "a kind of uneasy undercurrent".^[1][2]

Read that carefully. The damage serves a plan for the melody, and the melody comes first: it is something to be heard differently. The first is about how they learned; the second is about one 2005 album, and its page doesn't date the conversation. Neither describes every record the duo have made, and nothing in it ties a machine, a setting or a signal path to any particular track. This article doesn't either.

Two bars to test

What follows is invented for this article and is not taken from any recording. Set 72 bpm in 4/4. Hold a D2 (about 73 Hz) on a plain sine under both bars. Play this on a plain triangle or saw, with no effects:

• Bar 1, beat 1: A3, two beats.
• Bar 1, beat 3: D4, one beat.
• Bar 1, beat 4: F4, one beat.
• Bar 2, beat 1: E4, two beats.
• Bar 2, beat 3: D4, two beats.

It rises a fourth from A to D, steps up a minor third to F, falls a semitone to E, then a tone home to D. Against the D bass, ignoring octaves, the five notes are the fifth, the root, the minor third, the second and the root again. Play it a few times until you can hum it.

Change one note

Raise the F to F sharp and leave everything else alone. Over the D bass, F is a minor third above the root and F sharp is a major third: three semitones against four, which is the difference between a D minor and a D major colour. In equal temperament the ratios are 2^(3/12), about 1.189, and 2^(4/12), about 1.260. F4 is 349.2 Hz and F sharp 370.0 Hz, 20.8 Hz apart. The line into the next note changes too: F to E is a semitone, F sharp to E a whole tone.

Don't take our word for the mood of either. Play both with the effects bypassed, because that is where one note is easiest to judge. It is a compositional decision you can name, and it is yours to make before any texture arrives.

Now damage it, one variable at a time

Keep notes and levels fixed and change one thing per pass. Start with slow pitch drift, applied to the lead only. As an illustrative range, not a recipe, try ±5 to 10 cents (a cent is a hundredth of a semitone) at 0.1 to 0.3 Hz. At ±8 cents the F4 moves about 1.6 Hz either side, under a tenth of its 20.8 Hz gap to F sharp, so the third stays a minor third. Push towards ±50 cents and the F strays half a semitone at its extremes, a noticeable deviation, though still the same note being modulated rather than a new interval. This is a slower, subtler modulation than typical vibrato, which runs at several hertz and reads as expression, and it is not transposition, which moves every interval together. Watch the tempo too: at 72 bpm a bar lasts 3.33 seconds, so a 0.3 Hz LFO realigns with the bar once a bar (0.1 Hz, once every three) and can turn into rhythm. A free-running rate such as 0.17 Hz takes about 1.8 bars to cycle, though free-running alone doesn't guarantee it never coincides with the bar. A tape speed change moves everything recorded on that tape together. Drift on the lead alone also moves it against the bass, so decide whether the bass shares the LFO.

Next, bandwidth. A gentle low-pass, 6 to 12 dB per octave, with its cutoff at 2 to 3 kHz, is not a brick wall: it turns the harmonics above the cutoff down progressively rather than removing them. To see where they fall on a saw, F4's eighth harmonic is at 2.79 kHz (8 × 349 Hz), so eight sit below 3 kHz and four below 1.5 kHz. The fundamentals and the interval stay audible, so the melody survives. What softens is edge and attack, so the onsets of the shortest notes, the one-beat D4 and F4, become less distinct. Close the filter until you notice, then open it a little.

Noise comes last, and only if the phrase wants it. Noise fills the gaps between harmonics and covers quiet detail first: soft attacks and the tails of short notes. Start where you can't hear it, raise it until the one-beat notes begin to blur, then back off. None of this is bitcrushing, which cuts bit depth or sample rate and adds quantisation error and aliasing. That is a different kind of damage from tape's, and neither one is a universal sound of age.

The same damage on a weaker tune

Now the control. Keep the D bass, the tempo, the rhythm of two, one, one, two and two beats, and the same processing at the same settings. Swap in a plainer phrase: D4, E4, D4, E4, D4. It only ever shuttles between the root and the note above it, and never plays a third, so against the bass the harmony is undecided.

In our opinion it is the weaker phrase: there is no leap to remember and no note whose change would alter the harmony. Switch the damage on and off for both and ask what you would change in each with the texture bypassed. The first motif gives you a note to raise or lower. If the only interesting thing in the second loop is the ageing, then the loop is a texture, not a tune.

Bypass first, then choose

Write the tune with the texture bypassed. Then choose which imperfection should carry the atmosphere. As a starting hypothesis, and our reading, not the duo's: drift for unsteadiness, a darker top end for distance, noise for grain. Leave out the rest. Don't degrade the master to rescue a loop that needs a second part; that is an arrangement problem, and an eight-bar loop isn't a track covers it.

A stock synth, a filter, an LFO and a noise source are all this needs, so there is nothing to buy. If the phrase only works with the hiss on, keep writing.