Bell Derivations

~ Sound Design Illustration: All from 1 Source ~

prepared by A Endrich

sounds developed for Crossing the Dark Rift

Goal: a variety of tonal effects derived from a single source

Source: a single stroke on a Tibetan singing bowl: bellae.wav (4.9 sec).


Sequence 1: contrasting tones

These tones need to be short because they are intended for rapid, multi-event textures. Se we CUT the source from 0 to 1.1 seconds and lightly DOVETAIL the start and finish. I can't find the exact values for this dovetail; a close approximation is 0.25 second fade-in, linear, and 0.85: exponential decay. The outcome of these two operations is baegcdt.wav

Here are four sounds made with this version of the original source:

  1. HI-PASS FILTER –: baegcdthipc.mp3 is produced with a Pass frequency of 1568 Hz and a Stop frequency of 10: everything above the Pass frequency is retained, and everything below it, down to 10 Hz, the bottom of the sound, is gradually filtered away.
  2. EQUAL INTERVALS FILTER – baegcdthipcei.mp3 results when the filter bands are set at equal intervals of 100 Hz, using the Hi-Pass Filter result as input. This is set to occur between 100 and 3200 Hz, with a Gain of 6 to compensate for the filtering, and a Q (filter focus) of 200, which is fairly high.
  3. SHRED – baegcdtshred2.mp3 shows how the shredding operation breaks up the sound. The parameters here were set to 3 repeats, average_duration (of the shreds) at 0.2 seconds, and random_scatter of 3 (where in the sound the shreds are taken from).
  4. SUBHARMONIC FILTER – baegcdtsubh4dt.mp3 is a soft plunk remaining after subharmonic filtering between a high_frequency of 1000 and a low_frequency of 50, with a great deal of Gain: 53, and a high Q (filter focus): 350. A DOVETAIL smooths out the beginning and the end, values not specified in my notes.

Sequence 2: a very soft, lightly bubbling sound

This time the 4.9 second baeg.wav source is CUT between 1 and 4.9 seconds, i.e., a whole second of the attack transient is removed. The edges are greatly smoothed with exponential DOVETAILS of 1 second both at the beginning and the end. The use of exponential here (effect starts slowly and gets faster) is designed to maximise the length of the sustain part of the sound. We therefore end up with a gentle and resonant sound: bresdte.wav .

FOCUS ACCUMULATE introduces a slow glissando inside the partials of the sound, by accumulating data across several time zones of the sound (analysis windows). The glissando is set at 2.8 octaves per second and the sound is tidied up with GAIN x 2 and a DOVETAIL which doesn't alter the beginning (0.02 sec.) and brings the end down to silence over 0.5 sec. with an exponential curve. Our 2nd Sequence results in bresdteaccu2&8gdt.mp3. I can't work out why there's a rapid fluttering towards the end – perhaps it's caused by the exponential curve.

Sequence 3: richly modulating sounds

Now we set off to make some big changes. The original baeg.wav is DOVETAILED with 1 second slopes at the beginning and the end, this time with the beginning exponential (rises increasingly faster) and the end linear (steady reduction in amplitude). This gives us bdtG5.wav .

Time-varying RING MODULATION makes a big change to the sound. Ring modulation both adds and subtracts a given frequency, thus splitting the sound by pushing its frequencies both upwards and downwards. In this case, it does this by different amounts at different times – all rather small amounts, as specified in bdtrm.brk:

0.0	15
1.2	25
3.0	17
4.5	20
The modulation_frequency changes gradually here, e.g., moving gradually from 15 to 25 Hz over the first 1.2 seconds. Thus it is changing all the time, undulating between the modulation frequencies provided. Now our sound is: bdtrmtv1.mp3. The letters 'tv' in the output soundfile name refer to this time-varying modulation frequency.

The amplitude is a little low after this operation, so we raise it by a factor of 1.7 with GAIN: bdtrmtv1g.mp3. The next transformation might work better if the sound were a little lower, so we move it down 2 semitones without changing its length by using REPITCH TRANSPOSE: bdtrmtv1gd2.mp3.

Now we're ready for the final step in this sequence which is to SHUFFLE the internal order of the sound. This is a wavecycle DISTORTION process which rearranges little time-zones (analysis windows) of the sound. It increases the length because we create an extended pattern of rearragements which repeats in cycles. The pattern has a domain part which names three letters: abc and an image part which duplicates them 3 times in a different order: cab-acb-bca. [The hyphens aren't used; they are only put here for clarity.] This new image repeats in 5 cycle_groups and gives us: bdtrmtv1gd2shuf5.mp3, which is quite a long way from our original sound.

Sequence 4: a warm, vibrant tone

This sound begins with yet another CUT & DOVETAIL operation applied to baeg.wav :

The next step is to enrich this rather gentle sound by TRANSPOSING it up and down by a mere 0.1 semitone (10 cents) – more than this gives beat interference patterns which we don't want here. This gives three versions which are then MIXED together, all starting at time 0 but with slightly different amplitude levels. This is the MIX file baeg3mx.mix:

baeg3cdt.mp3			0	1	0.85	0
baeg3cdtu&1.mp3		0	1	0.75	0
baeg3cdtd&1.mp3		0	1	0.9	0
and this is the result: baeg3mix.mp3. Another CUT (0 to 2 sec.) and DOVETAIL 0.01 linear at the start (i.e., virtually no change) and 0.6 linear at the end (smooth decay) gives us our tidied up results: baeg3mixcdt.mp3.

This subtly enriched, gentle sound shows its special qualities in an arpeggio texture: soft, with a strange resonant, reverberant character. Describing how these arpeggios are made goes beyond the scope of this series of examples, so we can just listen to the result: baeg3mixcdtflob1mtf.mp3 (random factors make it deliberately irregular in its timing.)

Last updated: 30 September 2003
mp3 format changes: 21 February 2010 (sources or near sources are kept as wav)

©2003-2010 Archer Endrich Chippenham, Wiltshire  England, both texts and music. All rights reserved.