Delay Effects

From DSPWiki

Jump to: navigation, search

A digital delay unit or digital delay line (DDL) takes in a stream of input samples and stores them in a circular queue for a brief period before sending them out again. A single-tap delay has one write position in the queue, and one read position. A multi-tap delay has two or more read positions in the queue. Both types of delay increment the read position(s) along with the write position on every read.

lecture17_image4.gif

Contents

Generic Effects Processor

Many delay effects can be created with a generic effects processor.

lecture17_image1.gif

Fixed Delay Effects

Several audio effects can be produced with a fixed delay unit, which is a delay unit where the delay time does not change while the sound is passed through it. Various fixed delay effects can be grouped into categories based on their time spans, and the perceptual effects they create:

Short (less than 10 ms):

Short delays are perceptible in terms of the frequency-domain anomalies they introduce. Single sample delays create lowpass filters, and 0.1 to 10 ms delays create comb filter effects.

Medium (10 to 50 ms):

Medium delays can be used to enhance a thin signal by giving the illusion of loudness without a corresponding increase in measurable amplitude. Delays in this category fuse with the original sound to create a doubling effect which may be enhanced with subtle time-varying pitch shifts and delays.

Long (greater than 50ms):

Long delays create discrete echoes. Since sound travels at 1100 feet per second in air, a delay of 1ms corresponds to a total sound path of about 1 foot. This implies that a distance of about 50 feet from the source to the reflective surface to the listener is the same as a time delay of 50 ms.

Variable Delay Effects

In a variable delay unit, the delay time is constantly changing, which is implemented by varying the tap points at each sample period.

lecture17_image5.gif

delay ( i - j ) is dynamic.

Flanging

Electronic flanging derives from an effect made popular by a reel to reel tape technique of playing two identical copies of a recording, but pressing on the flange of the tape reel to slow down one of the tapes, varying the phase offset compared with the other tape. Flanging was discovered by Christian Huygens in 1693, but Les Paul was the first to use flanging as a sound effect in 1945, but he did it with disk recorders, not tape. In place of a thumb on the flange of a tape reel, electronic flanging varied by an LFO.

Flanging could be called a swept comb filter effect, where several nulls sweep up and down the spectrum. Filter peaks are located at frequencies that are integer multiples of 1/D, where D is the delay time. The depth of flanging is maximum if the amplitudes of the original signal and the delayed version are equal.

lecture17_image2.gif

Phasing

Phasing is similar to flanging, but is usually not as pronounced. Phasing is produced by running a signal through a series of allpass filters that shift the phase of the original signal. An LFO is used to sweep the amount of phase shift introduced by each allpass filter, and the outputs of the filters are mixed with the original signal.

Chorusing

Chorus effects are a general classification of techniques that aim at creating a full chorus from just a single instrument sound. This is generally done with a mixture of varying multitap delay lines, detuning, and splitting the signal into various octave-wide frequency bands that will be slightly detuned.

lecture17_image3.gif

Fractional Delay:

A fractional delay interpolates fractional samples, which makes for a smoother modulation. Without a fractional delay, we hear a jump in frequency with a modulated delay.

One problem with interpolating is that an average of two samples is the equivalent to a 1st-order lowpass filter. The amount of filtering depends on where you interpolate between the two samples, so the change in amplitude that this causes creates amplitude modulation.

One possible solution is Allpass Interpolation, as proposed by Dattoro, which is good for relatively small pitch changes.

lecture17_image6.gif

Personal tools
Ads: