Chapter 21 — Exercises

How to use these. Difficulty runs ★ (warm-up) to ★★★ (stretch). Time estimates are honest averages — slower is fine, skipping is not. Parts A and D need paper or a notes file; Part B needs headphones and the named tracks (any streaming service); Part C needs your DAW and the session you balanced in Chapter 20 — these exercises are this chapter's Project Checkpoint, expanded. Do the listening work with fresh ears, at your calibrated level from Chapter 8, not at the end of a long session. Selected answers and worked solutions live in the back matter (Answers to Selected Exercises); exercises marked ✎ have an entry there.

Part A — Conceptual (about 50 minutes total)

A1. ★ Gain or volume? ✎ For each move, say whether it changes what a channel's plugins see, what the bus receives, or only what you hear — and name the stage involved: (a) clip gain on a vocal region; (b) the channel fader; (c) an EQ's output trim, with a compressor in the next slot; (d) the monitor volume knob on your interface; (e) the master fader. One sentence of justification each. (6 min)

A2. ★ The Theo equation. ✎ Explain, in plain words a bandmate would follow, how a session can have every channel fader below zero and a 2-bus peaking at +4 dBFS. Your explanation must use the words summing and upstream correctly at least once each. (5 min)

A3. ★★ The -18 deposition. You're on a forum (condolences) and someone writes: "-18 dBFS is a myth — digital doesn't care." Write a four-sentence reply that (a) concedes what's true in their claim, (b) gives the historical reason the number exists, (c) names the modern practical reason it still matters, and (d) frames the whole thing honestly as convention rather than law. Bonus point for working in A4 = 440 Hz. (8 min)

A4. ★★ Crest-factor triage. ✎ Rank these sources from largest to smallest expected crest factor, and for each, state which meter (peak or average) you'd trust to set its level and which you'd use only to check the ceiling: close-miked snare, sustained string pad, lead vocal, heavily distorted rhythm guitar, fingerpicked acoustic guitar. Then answer: two of these could read identical peaks while differing hugely in loudness — name a pair and explain. (7 min)

A5. ★★ Where the red light bites. ✎ For each location, say whether clipping there is permanent, survivable-but-costly, or harmless — and what your immediate move is: (a) the A/D converter during a once-in-a-lifetime vocal take; (b) a synth bus inside the DAW's 32-bit float engine; (c) the D/A converter while you mix; (d) the 24-bit WAV you just bounced for your collaborator; (e) the input of a tape-emulation plugin you drove on purpose. (7 min)

A6. ★★ Headroom as a promise. "Headroom is a workflow, not a number." Defend this sentence using the headroom ladder: name who leaves room at the tracking stage, the track stage, the bus stage, and the 2-bus — and for whom each margin is left. End with one sentence on what goes wrong when the mix delivers a 2-bus at -0.2 dBFS to mastering. (8 min)

A7. ★★★ The tradeoff essay. Driving a character plugin harder changes its sound and spends headroom and biases your judgment if the output runs hot. Using the unity-in/unity-out discipline, explain how a producer can have all three: deliberate drive as a flavor decision, a staged channel, and an honest A/B. Where exactly does each of the three concerns get handled in the insert chain? One solid paragraph. (8 min)

A8. ★★ Pan law, briefly but correctly. Your collaborator opens your session in a different DAW and says the lead vocal (panned center) now feels quieter against the hard-panned doubles. Explain what probably happened, why DAWs do this on purpose, and what you'd both check before touching a single fader. (5 min)

Part B — 🎧 Listening Lab (about 70 minutes total)

Trained ears with cheap tools beat the reverse — and this chapter's listening skill is hearing level relationships and crest instead of just "loud" and "quiet." Headphones on, calibrated listening level, listening journal open. Each exercise ends with a written observation.

B1. ★ Crest factor, extreme spiky — Led Zeppelin, "When the Levee Breaks." The drum intro, three passes. Pass one: feel how hard each hit lands. Pass two: ask how loud the drums actually are between the hits — attend to the space and decay, not the impacts. That gap between the strike and the body is crest factor made audible. Journal: if you had only a peak meter, would it tell you how loud these drums feel? One sentence on what the peak meter would miss. (8 min)

B2. ★ Crest factor, near zero — Brian Eno, "An Ending (Ascent)." Sustained pads, almost no transients: peaks and average nearly holding hands. Notice how the music feels louder than it should at a given volume setting compared to B1 — steady energy reads loud even when nothing ever spikes. Journal: describe the difference between this track's "loud" and the Zeppelin drums' "loud" in your own words; you've just described average versus peak. (7 min)

B3. ★★ Low crest by design — AC/DC, "Back in Black." The rhythm guitars: saturation flattened their peaks long before the mix (amp distortion is a crest-factor decision — Chapter 26 will say why). Compare the guitars' steadiness against the drums living beside them, which still spike. Journal: rank the guitar, snare, and vocal by crest factor as you hear them, highest to lowest, with one phrase of evidence each. (8 min)

B4. ★★ Sparse and spiky — Billie Eilish, "bad guy." A nearly empty arrangement: kick and bass thumps with real transients over a very quiet floor. Each hit jumps far above the track's average — yet the song never feels quiet. Listen for how few loud moments it takes to feel full when the spaces are clean. Journal: two sentences — where does this track's loudness live, in its peaks or its body, and how is that different from B2? (8 min)

B5. ★★ The self-deception demo (do this one — it changes people). ✎ Take any finished track you love. Play 20 seconds at your normal volume. Now play the same 20 seconds about 3 dB louder (one or two volume-key taps; small, not dramatic) and ask yourself honestly which playback sounded "better — fuller, more exciting, more expensive." Repeat the pair twice. Nothing about the music changed; only level did. Journal: write down what the louder pass seemed to improve (most people say bass, air, and "energy" — equal-loudness curves at work), and then write the moral in one sentence beginning "Any A/B where the levels don't match is really a test of…" Then run the reverse: same passage about 3 dB quieter, and notice how readily "thinner, duller, less exciting" arrives — about music that also didn't change. The bias cuts both ways, which is exactly why every plugin demo, preset, and "improved" bounce must be matched before your verdict counts. (10 min)

B6. ★★ Average versus peak in the wild — the ad-break study. Stream anything with commercials, or compare a dialogue-heavy show against its trailer. The program and the ads are peaking in similar territory; the ads spend nearly every moment near the top while the program saves its top for accents. That's a crest-factor difference, not a peak difference. Journal: describe the sensation of the transition in band-and-dynamics vocabulary (Chapter 4 + this chapter), not in feelings. One extra sentence: which meter would catch this difference, and which would swear nothing changed? (8 min)

B7. ★★★ Drive sweet spots — hearing input level as tone. Find any saturation, tape, or console-style plugin you own (free ones abound) and a loop of full-mix material. Set the plugin's output so bypassed and active match loudness. Now audition three input levels into it — about 12 dB too quiet, the healthy -18-ish zone, and 8–10 dB hot — re-matching the output loudness each time. Listen for what changes when loudness can't vote: thin and inert when starved; gently thickened at nominal; increasingly fuzzy, dense, and low-mid-heavy when slammed. Journal: describe each of the three in one phrase, and mark which one you'd choose on purpose for this material — there's no wrong answer, only an undeclared one. (12 min)

Part C — 🎛️ DAW Lab (about 110 minutes total; uses your Chapter 20 session)

This is the chapter's core homework. C1–C4 executed in order are the red-light audit on your own mix; the rest build the reflexes around it. Work at your calibrated monitoring level, take a five-minute break between C2 and C3 (fresh ears make honest unity matches), and resist every urge to "improve" anything while you stage — the mixing resumes in Chapter 22, on foundations that deserve it.

C1. ★ The before snapshot. ✎ Open your session. Before touching anything: screenshot or write down the master meter's peak reading on the loudest section, every bus's peak, the master fader's position, and the three hottest channel meters. Save as audits/ch21-before.txt (the audits/ folder you started in Chapter 3). You cannot measure improvement without a baseline — and in three chapters you'll enjoy reading this. (8 min)

C2. ★★ The clip-gain pass. ✎ Loop the loudest section. Track by track, watching an average (RMS/VU-style) meter, clip-gain each track's body toward the -18 dBFS neighborhood — spiky tracks will peak way above that and should. Give the lead vocal a finer pass: section by section, bringing quiet verses and hot choruses within a couple of dB of each other. Log the two biggest trims you made and why they were there (hot sample pack? timid render? visual-waveform mixing?). That log is a portrait of your habits. One caution while you work: trim regions and sections, not individual drum hits — hit-to-hit dynamics are performance (Chapter 13's velocity work) and they belong to Chapter 23's compressor, shaped on purpose, not flattened here by meter. (25 min)

C3. ★★ The unity pass. ✎ For every insert already in your session (you have few, this early — good): set its output trim so toggling bypass produces no loudness change, judged by ear first, confirmed by meter. Where a plugin has no output control, add a gain plugin after it. Note the single worst offender — the plugin that was adding the most free loudness — and write one sentence about whether you liked it for its processing or its level. Be honest; nobody's reading. (15 min)

C4. ★★ The full audit, steps 3–7. Restore your static balance with the faders (fast — you know the destinations from Chapter 20), check FX returns aren't screaming, trim buses toward peaks around -8, and confirm the 2-bus lands near -6 dB peaks with the master fader at unity and zero red lights from source to 2-bus. Save the after-snapshot beside C1's as audits/ch21-after.txt. Compare: where did your hidden gain live? (20 min)

C5. ★★ The matched-bypass lie detector. ✎ Put any EQ on a track and boost something flattering (a high shelf, say, +3 dB) with no output compensation. A/B it. Note your verdict. Now trim the EQ's output until bypass and active match loudness, and A/B again. Write down both verdicts and what changed between them. Keep this exercise in your back pocket forever: it's the sixty-second proof that an unmatched A/B is a volume test wearing an EQ costume. (10 min)

C6. ★★ Build the headroom budget. In your project notes, write your session's ladder as actual numbers: track averages, typical track peaks (note your spikiest track and its crest), bus peaks, 2-bus peaks, master fader position. Then write one sentence per rung naming who the margin is for (plugins, summing, mastering). Tape it next to the Chapter 3 chain map. This is your studio's gain-staging contract — revise it when experience earns a revision. (12 min)

C7. ★★★ The crest-factor safari. Drop a peak meter and an average meter side by side (or one plugin showing both) on four different tracks in your session: your spikiest drum, your steadiest pad or sustained part, the lead vocal, and the bass. Play the chorus and log peak, average, and the crest gap for each. Compare your numbers against the chapter's table — your session, your sources, your proof. One journal sentence: which track surprised you, and what does its crest tell you about how to stage it? (12 min)

C8. ★★★ The deliberate disaster (sandbox only — Save As first). In a copy of the session, commit every sin: normalize three clips to 0, push every insert's output a few dB hot, drag faders down to compensate, and "fix" the resulting hot 2-bus with the master fader. Now listen at matched overall loudness against your real, staged session. Often the disaster doesn't sound dramatically worse yet — that's the honest, important finding: bad staging mostly poisons what comes next (every threshold, every drive, every A/B). Write two sentences: what you can hear now, and what you know is waiting. Delete the copy. (10 min)

Part D — Synthesis (about 40 minutes total)

D1. ★★★ The rescue protocol. ✎ Write a one-page, numbered "session rescue" checklist for the Theo scenario — a years-old session with hidden gain at every stage — ordered in signal order, with a meter and target at each step, and a final verification line. Include one warning about what not to destroy (the balance) and one about the master fader's job description. Then prove the document works: run it, step by step and without improvising, on the deliberate-disaster copy you built in C8 — anywhere you had to deviate from your own checklist, revise the checklist. You'll use this document on a real session within a year; write it for tired-you. (20 min)

D2. ★★ Teach it in five sentences. A friend says: "I keep my faders low so nothing clips — is that gain staging?" Write a five-sentence reply that names what they got right (protecting a ceiling), what they missed (the inserts never found out; the level the plugins see), and the one-sentence version of the fix. No jargon they don't have; you may assume they've plugged in a mic before. (8 min)

D3. ★★★ The convention with reasons. In your own words — one paragraph, journal-quality — explain why this book teaches -18 dBFS averages and -6 dB 2-bus peaks as habits with history rather than as laws, and what you'd actually lose (name two concrete things) by ignoring them versus what you wouldn't lose (name one). The goal is to own the convention rather than obey it. (8 min)

D4. ★★★ The relationships test — and the yardstick test. Design a thirty-second demonstration — for a skeptical friend, using any session — that proves "levels are relationships, not absolutes." Write the steps and the one question you'd ask them at the end. (The chapter contains one such demo; yours may improve on it.) Then, after reading Case Study 1, run the pink-noise balance once on a copy of your session — noise at a fixed reference level, each track raised alone until just audible through it, noise off, everything unmuted — and listen for one full chorus. Write a three-part verdict: (a) what the procedure got usefully close to, (b) the two most important things it got wrong (think hierarchy and sections), and (c) whether you'd use it again, and for which specific situation — reset ritual, cold-start scaffold, ear-training demo, or never. There's no required answer; there is a required reason. (20 min)

Part M — Mixed Practice (interleaving Chapters 1–20; about 30 minutes)

M1. ★ (Ch 2 + Ch 21) Your session is 24-bit, 48 kHz. A friend worries that running tracks 18 dB below full scale "throws away resolution." Using Chapter 2's bit-depth-equals-noise-floor model, explain in three sentences why the worry is misplaced at 24-bit — and name the one era-specific reason old-timers tracked hotter to tape and early converters. (6 min)

M2. ★★ (Ch 4 + Ch 21) Equal-loudness contours explain both why quiet mixing tells the truth about balance and why unmatched A/Bs lie. Connect the two in one short paragraph: what single fact about your hearing underlies both rules? (6 min)

M3. ★★ (Ch 11/12 + Ch 21) Chapter 11 told you to track vocals with peaks around -18 to -12 dBFS; this chapter told you to keep track averages near -18. Are those instructions in conflict? Resolve them using crest factor and the difference between protecting a converter and feeding a plugin. (6 min)

M4. ★★ (Ch 16 + Ch 20 + Ch 21) A mix's 2-bus keeps creeping hot even after a clean audit, and the loudest section feels cluttered no matter how carefully it's staged. Using Chapter 20's threshold concept and Chapter 16's density budget, name the diagnosis the meters can't show you — and the pass you'd run before re-touching a single level. (6 min)

M5. ★★★ (Ch 13 + Ch 21) Your programmed drums from Chapter 13 use velocity layers and ghost notes. Explain why heavy-handed clip-gain "smoothing" of individual drum hits would undo that chapter's work — and which tool, in which later chapter, is designed to shape hit-to-hit dynamics musically instead. (6 min)