four tape loops and a broken walkman

the tape moves at 4.7625 cm per second. that is the only fixed number in any of this. everything after it is arithmetic.

a loop n centimetres long comes back round every n / 4.7625 seconds. you cannot stretch it. cut 28.575 cm and the phrase is six seconds, for the whole piece.

i work in c60s. the tape in a c60 is 15 to 16 micrometres thick, in a c90 it is 10 to 11. thin stock creases off your fingers and a crease is a tick on every pass.

loop	tape needed	passes in 227 s
4 s	19.050 cm	56.8
6.3 s	30.000 cm	36.0
10 s	47.625 cm	22.7

227 seconds is how long everything i put out is. that is a different problem.

the cut

1. buy screwed shells. five phillips screws. welded shells crack open but the halves never sit flat again, and a shell that does not sit flat eats tape.
2. wind off your length plus 4 cm and measure it against a steel rule. counting hub turns does not work, the hub diameter changes as it fills.
3. cut both ends square in the splicing block. quarter inch editors cut a diagonal to spread the join across the head gap. at 3.81 mm against their 6.35 mm, the diagonal you can actually cut is too shallow for me to hear.
4. butt the ends, no overlap. an overlap is double thickness past the head, which lifts the tape off the gap, and a dropout is worse than a tick.
5. splicing tape on the backing side only, never the oxide. trim it flush to 3.81 mm with the blade in the block, then burnish through folded paper with a fingernail. this is the step that takes practice, and a bad one fails on pass two rather than pass two hundred.
6. rebuild the shell with the loop round the hub area, over the head window, past the pinch roller cutout, back. no hubs. slack enough to lie flat and no slacker, because slack tape finds the capstan.
7. run it and listen for the tick. you have a metronome now whether you wanted one or not.

@rustbelt/scraping-a-1943-lathe-bed-flat is about working by hand to a tolerance you cannot see. step 5 is that job with a razor blade and no reference surface.

friction

in a normal cassette the tape goes hub to hub and neighbouring layers barely slide. a loop slides over itself, and that is where loops die. the speed wanders. then the capstan whines. then the motor is turning, the tape has stopped, and it is wadding up behind the head.

keep it under 30 cm if it lives in the shell. a soft pencil run along the backing lowers the drag and makes a grey mess of the inside of the machine.

what actually works is to stop the tape touching itself. run the loop out through the shell window, over a pencil taped to the desk 40 cm away, and back in. i have run just over 3 metres like that, 63 seconds, and it holds speed better than a 25 cm loop in a shell.

the erase head

it has to sit upstream of the record head, otherwise it would wipe what the record head just laid down. in a cassette it gets its own window in the shell, which is why auto reverse decks cram it in beside the record and play head.

take it out and the new signal goes on top of the old one instead of replacing it. this is not clean layering. the record head puts out bias the whole time it records and bias partially erases, so pass one is loud, pass three is underneath, and by about pass six the first thing you played is gone with no copy anywhere. hiss stacks the whole time. hiss is the only layer that does not fade.

that is the instrument. not a pedal that repeats faithfully. a machine that forgets at a rate you can hear.

people bring up basinski. that was a different thing: the coating came away from the backing as it passed the head of a revox g36, off tapes he made in 1982. you cannot do it twice on purpose.

what it cost

• sony wm-d3, not working, £40. the one i gutted.
• sony wm-d6c, £320. quartz locked capstan, records dolby b and c. i leave dolby off.
• splicing block £14, splicing tape £6, c60s 80p each from a box at a car boot sale.