Cascading tuned delay

This is an effect that sometimes showed up on "digital delay" devices in the early 1980s. Anything that goes into the effect comes out as an arpeggio (a quick, overlapping sequence of pitches) that can either be used to decorate the original sound or by itself as an abstract-sounding musical cloud.

The network consists of a delay line (just a delwrite~) that is read out by 19 copies of an abstraction. The abstraction contains a delread~ and a bandpass filter (bp~ will work fine). Each copy gets a different delay and a different filter setting so that if you put a short sound in (for example) you hear the 19 filters in order of the delay times. In the example here, the delay times are just 17, 34, 51, ... (the multiples of 17) - of course, the number 17 is just a setting that you can change later. The filters are tuned to a whole-time scale, specifically, the MIDI pitches 72, 74, 76, ..., 108. (these are chosen roughly to span the formant frequencies of most vowels so that this setting works particularly well with speech.) The steps are as follows:

• Make a sampler that can play a voice sample - you can use this one or supply a short one of your own. The sampler example 3j.sampler.pd from class might provide you with a good starting point (but for this purpose you don't really need the pitch control so you could simplify it considerably). Connect the output into a delwrite~ with plenty of room, say 10 seconds. Also connect the sample output to the patch's output so we hear the "dry" sample along with what will come out of the delays.
• Make an abstraction to read and filter the signal from the delay line; the abstraction should have only one delread~ and bp~ (or vcf~ would be OK too). Use the "clone" object to make 19 copies. In each copy, "$1" will be a number from 0 to 18.
• Fix the abstraction to use its argument as a multiplier for the delay time (that is, multiply a given base delay received through a "receive" object by the numbers 1, ..., 19), and also to make a pitch offset, counting by twos: 0, 2, .., 36, added to a base pitch, so that if the base pitch is is 72 you should hear the pitches 72, 74, ...108).
• Inside the abstraction, feed the calculated delay time to the delread~ object. Send the calculated pitch (converted to a frequency) to the filter's resonant frequency input, and also to the input for Q so that the bandwidth, equal to frequency/Q, is one Hz. This will be a selective, resonant filter.
• Outside the abstraction, in the main patch, make number boxes feeding into sends to control the base delay time (17 for me) and the base pitch (72). If you like find your own values that go well with whatever sample you're using. As usual, fix a loadbang and message box to get the good values into the number boxes. Finally, collect the output of all the filters, and multiply it by something to raise its level to something that mixes well with the "dry" unprocessed sample (I used 50; the filter outputs are quite soft).

  Your successful patch should sound something like this.

  back to music 171 main page