Skip to content
Ohmic Audio

DSP Signal Flow

Canonical Definition

Signal flow is the ordered path and branching structure through which audio, control, and detector signals move from source to output, including every gain, routing, processing, metering, conversion, and protection stage.

Draw the chain before tuning it.

A control is meaningful only when you know where it sits, what feeds it, and what comes after it.

Four Paths That Must Not Be Confused

PathWhat it carriesCommon examples
Audio pathThe samples or analog signal being heardADC, mixer, EQ, crossover, delay, compressor gain stage, DAC
Control pathValues that change processingpotentiometer, mute state, preset selector, threshold, matrix coefficient
Detector or sidechainA signal used to calculate gain or trigger an actionpeak envelope, RMS level, linked-channel maximum, external sidechain
Meter pathA tapped view of one nodeinput meter, post-EQ meter, gain-reduction meter, output meter

A Complete Car-Audio DSP Chain

source and source processing
→ analog or digital input interface
→ input buffer / ADC / digital receiver
→ input trim and input meter
→ routing and summing matrix
→ crossover, EQ, delay and other processing
→ dynamics detector and gain stage
→ output trim and output meter
→ DAC / analog line driver
→ amplifier input and output
→ loudspeaker, enclosure and vehicle

Every arrow is a possible level, noise, latency, routing, or clipping boundary. A clean meter after the ADC does not prove the analog input buffer was clean. A clean pre-DAC meter does not prove the line driver or amplifier input remained clean.

Processing Order Changes The Result

Parallel Paths And Hidden Gain

A signal can reach the same output through more than one route. Two identical unity-gain paths add coherently and can raise voltage by 6.02 dB. Parallel processing is useful only when the blend, latency, polarity, phase, and total gain are deliberate.

Trace every output backwards. If two branches rejoin, identify the coefficients and expected worst-case correlation.

Fixed Point, Floating Point, And Physical Boundaries

Word length and accumulator design affect internal range and precision, but the implementation must be documented rather than guessed from a marketing label. For example, the ADAU1701 accepts normal audio data into a 28-bit core representation with additional processing headroom and uses a 56-bit accumulator for multiplication and accumulation. That does not remove the ordinary ADC, DAC, serial-output, and analog-stage ceilings.

Floating-point processing also does not make the complete system unclippable. Output conversion, deliberately saturated blocks, coefficient ranges, analog stages, and downstream hardware still have limits.

Verification Workflow

  1. Label every physical input and output.
  2. Build a unity-reference path with unnecessary processing bypassed.
  3. Verify one route and one physical output at low level.
  4. Add summing and monitor the actual summing node.
  5. Add filters, delay, and EQ while checking level before and after each level-changing stage.
  6. Add dynamics and identify both detector and controlled audio paths.
  7. Verify DAC or line output externally.
  8. Save a diagram, preset, meter map, and physical connector map.

Related Resources