Haas

Widener — Position 6

The Haas module widens the stereo image using two complementary techniques: M/S (mid/side) processing with independent gain control, and Haas effect comb filtering. It is named after Helmut Haas, whose 1951 doctoral research demonstrated that a short delay (1–40 ms) between the ears causes the brain to perceive a single fused sound, and that the direction and width of that image can be sculpted by controlling the delay and level relationship.

Not Transaural XTC

The Haas module is a creative stereo widener, not a transaural crosstalk cancellation (XTC) processor. It works conventionally on stereo buses and is mono-compatible when using Side Comb mode.

Signal flow:

In L/R → M/S encode → Mid gain / Side gain
       → [Side Comb or Wide Comb delay]
       → M/S decode → Output trim (RMS-safe) → Dry/Wet blend → Out L/R

Default position in chain: Before Punch — so the clipper catches any widener-induced peaks before they exceed the ceiling.

---

Controls

Control	Range	Description
Bypass	On/Off	Bypasses all Haas processing.
Mid Gain	−12 to +6 dB	Gain applied to the M/S mid (sum) channel before the comb filter. Reduce to push the mix wider; boost to increase mono focus.
Side Gain	±6 dB	Gain applied to the M/S side (difference) channel. Increasing side gain makes the stereo image wider; reducing narrows it.
Comb Depth	0.0–1.0	Amount of comb filter contribution. 0 = pure M/S width control. 1 = maximum Haas comb effect.
Comb Time	1–20 ms (skewed)	Delay time for the comb filter. Shorter (1–5 ms) = tight, subtle widening. Longer (10–20 ms) = pronounced spatial depth.
Mode	Side Comb / Wide Comb	Comb filter routing (see below).
Mix	0.0–1.0	Dry/wet blend. 1.0 = fully processed; 0.5 = equal blend for parallel widening.

---

Comb Modes

=== “Side Comb” The comb filter is applied only to the side channel (L−R). A polarity-flip comb delay is injected into the side signal — because side energy sums to zero on mono collapse, the comb contribution is fully mono-compatible.

**Characteristic:** WOW-Thing–style widening. Tight, coherent center with enhanced stereo width in the diffuse field. The mono image is not affected.

**Recommended for:** Master bus, vocal bus, any context where mono compatibility is critical.

=== “Wide Comb” A delayed (L−R) signal is injected with opposing signs into L and R. This creates a more diffuse, spacious effect than Side Comb.

**Characteristic:** Broader, more ambient widening. Comb depth is internally clamped to 0.5 to cap the L-channel peak at ≈+3.5 dB worst case and stay within the RMS budget.

**Recommended for:** Synth buses, ambient tracks, reverb returns. Check mono compatibility before committing on a master bus.

---

Techniques

Subtle Bus Widening

Control	Value
Mid Gain	0 dB
Side Gain	+2 dB
Comb Depth	0.3
Comb Time	7 ms
Mode	Side Comb
Mix	1.0

A small side boost + light comb gives noticeable width without disturbing the center image. Works well on guitar and synth buses.

Mono Focus (Narrowing)

Control	Value
Mid Gain	+2 dB
Side Gain	−4 dB
Comb Depth	0.0
Mix	1.0

Reduce side gain to pull an overly wide mix toward center. Useful for mono-checking or on bass buses where too much side energy causes low-end smear.

Ambient Depth (Wide Comb)

Control	Value
Side Gain	+3 dB
Comb Depth	0.5
Comb Time	14 ms
Mode	Wide Comb
Mix	0.7

The wider comb time at 0.7 mix creates a natural room-like widening without the pitched quality of a short comb. Effective on synth pads, strings, and ambient buses.

Mono Compatibility Check

After applying any widening, briefly set Mix to 0.0 (full dry) and compare the stereo image. Then listen to your DAW output in mono. Side Comb mode is safe for mono — Wide Comb mode may introduce comb coloration at the mono sum, particularly noticeable with high Comb Depth values.

---

Implementation Notes

• Delay interpolation: Hermite 4-point cubic interpolation for smooth, artifact-free delay time automation. Delay time is additionally smoothed by a one-pole LPF (τ = 20 ms) to eliminate zipper noise during sweeps.
• Ring buffer: Power-of-two length (4096 samples), wrap via bitwise AND mask — avoids modulo on the audio thread.
• Anti-denormal: 1e-20 alternating dither written into the delay line every sample (Airwindows convention).
• Output trim: RMS-safe automatic compensation based on mid+side peak budget: 1 / sqrt(1 + |side_gain| * comb_depth). Prevents the processed signal from exceeding the input RMS.
• MAX_DELAY_SAMPLES: Hard limit at DELAY_BUF_LEN − 4 (4 samples reserved for Hermite headroom). At 192 kHz, this covers ~20 ms.