Signal flow
Defines the ordered path and the location of every block, meter, converter, and boundary.
Routing, Phase, Crossovers, Dynamics, And Verification uses the Wondom ADAU1701 reference project to expose matrix arithmetic, mute/slew, timing, acoustic filters, sidechains, limiters, FIR/IIR resources, export, reload, and regression.
Defines the ordered path and the location of every block, meter, converter, and boundary.
Defines which input feeds which output and at what coefficient.
Defines how multiple contributions add, cancel, or consume headroom.
Defines how a detector changes gain over level and time.
The first practical version uses one enclosed two-input, four-output Wondom ADAU1701 unit and a full bench oscilloscope. The same logic applies to other DSPs, but block names, number formats, meters, time constants, and analog limits vary.
ADC input L/R
→ input trim and input meters
→ routing / matrix stage
→ four parallel output paths
→ output meters
→ DAC outputs 1–4
| Output | Teaching path | Purpose |
|---|---|---|
| 1 | Left unity reference | Known clean comparison |
| 2 | Right unity reference | Channel and physical-output verification |
| 3 | Adjustable L+R sum | Correlation, coefficients and headroom |
| 4 | Compressor / limiter path | Detector, timing, gain reduction and ceiling tests |
Complete the DSP-only lab before connecting an amplifier. Use the bench scope according to its grounding and input limits, and verify the Wondom's actual unit and revision rather than hard-coding generic voltage claims.
Include physical input, analog input buffer, ADC, digital input node, input trim, matrix, processing blocks, dynamics detector, controlled gain stage, output trim, DAC, line driver, physical connector, amplifier, and speaker.
Mark each meter as pre- or post-processing and record its detector type and reference. A meter is evidence only for its own node.
Do not assume a display value means percent. Translate the actual unit:
linear gain = 10^(dB / 20)
dB = 20 log10(linear gain)
Useful anchors:
The worst coherent condition sets the deterministic headroom requirement. Music correlation varies with time and frequency.
Use 0.5 L + 0.5 R as a conservative starting point for fully correlated center material, then verify anti-phase loss, source-specific EQ or delay, the post-sum meter, physical output voltage, and downstream gain requirements.
A -3 dB-per-input convention can be useful for preserving the average power of uncorrelated inputs, but it produces +3 dB for identical coherent inputs. The target must be stated.
Create a deliberate duplicate route to one output, measure the level increase, then remove it. This makes a common configuration error visible: a signal can be routed directly and through another bus at the same time.
A user fade and a protection mute are different functions. One is chosen for feel; the other is chosen for fault response.
audio path ─────────────→ controlled gain ─→ output
└→ sidechain → detector → curve → timing ─┘
Place meters on the audio input, detector envelope, gain reduction, and output. The detector can use the same audio, a filtered copy, another channel, or a linked channel combination.
Use stepped input levels to record the static curve. For a simple 4:1 hard-knee compressor, 8 dB above threshold produces approximately 2 dB of output rise above threshold. Then compare hard and soft knee.
Do not call the threshold a ceiling. Timing, detector and makeup gain are not shown by the static curve.
Use bursts and observe overshoot, gain-reduction rise, hold, recovery and waveform modulation. The displayed time value may be defined differently by each algorithm, so measure the result rather than transferring numbers blindly between platforms.
Drive one side harder than the other. Independent compression can move the stereo image. Linked maximum, average, or other detector rules preserve different aspects of the relationship. Record the link rule instead of only writing “stereo linked.”
Record the block diagram, matrix, native coefficient units, meters, summing assumptions, mute sequencing, detector tap, detector type, threshold or curve, ratio, knee, attack, hold, release, channel link, makeup gain, limiter ceiling, analog output, amplifier state, test signals and measured failures.
Revalidate after any preset, routing, EQ, bass-restoration, source, firmware, controller or amplifier-gain change.
| Symptom | First checks |
|---|---|
| Mono path is 3 or 6 dB too high | matrix coefficient unit, coherent L/R signal, duplicate path, post gain |
| Some stereo content disappears in mono | opposite-polarity or side content, source processing, differential factory signal |
| Preset switch pops | DC offset, no slew, state-load order, channel mismatch, source phase or latency |
| Compressor pumps with bass | sidechain tap, detector type, sidechain filtering, release, link rule |
| Limiter still clips the amplifier | gain after limiter, DAC or line clipping, amplifier sensitivity, supply, load, overshoot |
| Stereo image moves during compression | independent detectors, link method, unequal path gain, asymmetric source |
The amplitude doubles, producing a 6.02 dB increase if the next stage has enough range.
No. Linear 0.5 is approximately -6.02 dB; -3.01 dB is approximately 0.707.
No. It changes what the detector reacts to unless the audio path is also filtered.
No. Detector timing, overshoot, later gain, DAC and analog limits, amplifier gain and operating conditions still matter.
The complete 12-question data set is stored with this lesson as quiz.json for future interactive use.