Audio DSP: Delay

---

---

Overview

A simple single-tap digital delay line implemented in plain C++, intended as a learning resource for understanding the circular buffer; the foundational data structure behind almost every time-based audio effect.

A delay line stores incoming audio samples in a fixed-size buffer and reads them back after a configurable time, producing an echo. The wet/dry mix controls how much of the delayed signal is blended into the output. This same structure underpins chorus, flanger, reverb, and the Haas effect; all of which are variations on the same principle.

---

Files

File	Description
Delay.h / .cpp	Single-tap delay line with wet/dry mix
main.cpp	Example: applies the effect to a 440Hz sine wave at three different delay and mix settings

---

How It Works

The delay is implemented as a circular buffer - a fixed-size array with a write index that wraps around when it reaches the end. On every sample:

1. The sample at writeIndex is read — this is the sample that was written delayInSamples frames ago
2. The new input sample is written into writeIndex
3. writeIndex advances by one, wrapping at the buffer boundary
4. The dry and delayed samples are blended by wetMix and returned

output = (1.0 - wetMix) * input + wetMix * delayedSample

writeIndex:  0   1   2   3   4   5  ...  N-1   0   1  ...
             ↑
             read oldest sample here, write new sample here, then advance

Because the buffer size is fixed at construction, no memory is ever allocated or freed during processing, which is critical for real-time audio.

---

Parameters

Parameter	Type	Range	Description
delayInMilliseconds	int	> 0	Time between dry signal and echo.
wetMix	float	0.0 – 1.0	Blend between dry and delayed signal. 0.0 = fully dry, 1.0 = fully wet.
sampleRate	float	> 0.0	Audio sample rate in Hz. Used to convert ms to samples.

Delay time and perceived effect:

Delay Time	Perceptual Effect
1 – 10ms	Phase/comb filtering — not perceived as echo
10 – 40ms	Haas effect range — stereo widening (see Audio Haas Effect)
40 – 100ms	Pre-delay / ambience — adds space without a distinct echo
100 – 500ms	Slap-back echo — common in rockabilly and vocal production
500ms+	Distinct repeating echo

---

Usage

Basic

// 500ms delay, equal wet/dry mix, at 44100Hz
Sherbert::Delay delay(500, 0.5f, 44100.0f);
 
float output = delay.ProcessSample(input);

Fully Wet

// Only the delayed signal is heard — silence for the first delayInSamples frames
delay.setWetMix(1.0f);

Changing Parameters at Runtime

// setDelayInMilliseconds calls reset() internally to avoid buffer artefacts
delay.setDelayInMilliseconds(250);
 
// setWetMix does not require a reset
delay.setWetMix(0.3f);

Resetting State

delay.reset();

---

API Reference

Method	Description
Delay(delayInMs, wetMix, sampleRate)	Construct with initial parameters.
ProcessSample(input)	Process one sample. Returns the blended dry/wet output.
reset()	Clears the delay buffer and resets the write index.
setDelayInMilliseconds(value)	Update delay time. Calls reset() internally.
setWetMix(value)	Update wet/dry mix. Range: 0.0 – 1.0.
getDelayInMilliseconds()	Returns current delay time in ms.
getWetMix()	Returns current wet mix value.
getSampleRate()	Returns the sample rate set at construction.

---

Limitations & Next Steps

This is an intentionally minimal single-tap delay for learning purposes. Real-world delay effects build on this foundation with additional features:

Feedback: a feedback delay feeds a portion of the output back into the input, producing multiple decaying repeats rather than a single echo. This is the most common next step and only requires one additional line: input += feedbackSample * feedbackAmount before writing to the buffer.

Modulation — modulating the delay time with an LFO produces chorus (very short delay, slow LFO) or flanger (very short delay, fast LFO) effects. This is the same LFO structure used in the phaser implementation.

No filtering in the feedback path — most hardware delay units apply a low-pass filter to the feedback signal, so each repeat gets progressively darker. This simulates the high-frequency absorption of real acoustic spaces.

If you want to explore further, the natural next steps from here are:

• Adding a feedback parameter (input += output * feedback before processing)
• Adding an LFO to modulate delayInSamples for chorus/flanger
• A low-pass filter in the feedback path for tape-style decay
• Multiple tap points in the buffer for multi-tap delay patterns

---