Appendix: Answers to Selected Exercises
Read this page before you read any individual answer, because the policy matters more than the answers do.
This book's exercises come in five lanes, and only two of them have a "correct" answer to check against.
Part A (Conceptual) and Part M (Mixed/Foundation Practice) ask checkable questions: define a term, run an arithmetic, catch a flawed claim, trace a chain of cause and effect. Those have answers, and a selected, representative set of them is worked below. Not all of them — that would turn the appendix into a crib sheet, and the struggle is where the learning lives. Items marked [A] or ★ in a chapter's exercise file are the ones promised a worked answer; the rest are yours to defend in your journal.
Part B (Listening Lab) and Part C (DAW Lab) are open-ended ear-and-hands work. There is no single right answer to "what happened to the bass when you played 'bad guy' on a phone speaker," because the right answer is what you actually heard on your phone in your room. Grading those against a key would be grading your headphones, not your ears. So for those lanes this appendix gives you assessment criteria — what a strong response demonstrates — and lets you score yourself honestly. A criteria block, not a solution.
Part D (Synthesis) sits in between: the strong answers share a shape even when they reach different conclusions. Where this appendix touches a Part D item, it describes that shape.
One more honesty note before the numbers, because the numbers are about to look authoritative. Every frequency range, every dB figure, every LUFS target on this page is an anchor, not a law — the same warning the Frequency Chart opens with. Your kick, your voice, your room sit a little left or right of every figure printed here. And the platform loudness numbers (Spotify ≈ -14 LUFS, podcasts ≈ -16 LUFS, true peak ≤ -1 dBTP, broadcast EBU R128 -23 LUFS) are true as of this writing and change without telling you.
Worked answers are organized by chapter.
How to grade your own ears (the Listening Lab and DAW Lab rubric)
Use this once, then apply it to every Part B and Part C exercise in the book. A strong response to any ears-or-hands exercise demonstrates four things:
- Band-and-axis vocabulary, not verdicts. "Low-mids piling up in the chorus, around 300 Hz" is an observation. "Sounds bad" is not. By Chapter 4 you have the six bands (sub / lows / low-mids / mids / high-mids / air) and three soundstage axes (width, depth, height-by-frequency). Use them.
- A receipt. A timestamp, a frequency guess, a system named, a number written down. The book hammers this from Chapter 1's listening journal onward because an observation you didn't write is an observation you'll lose.
- Matched loudness on any comparison. If you compared two things and didn't level-match first, you compared loudness, not the thing you meant to compare. Louder always sounds better. Always.
- A decision or a next step. Diagnosis that doesn't end in "so I'll check X" or "so my fix-list gets this item" is just listening. The book's whole arc is observation → diagnosis → action.
Score yourself out of those four. Three or four is a strong response. If you keep losing the same point — usually #2 or #3 — that's not a failing grade, that's your personal curriculum telling you which habit to drill.
Chapter 1 — What Is Sound?
A1. The three numbers. Frequency is how fast the wave cycles, measured in Hz; your ear perceives it as pitch (faster = higher). Amplitude is the size of the pressure swing; your ear perceives it as loudness (bigger = louder). Waveform is the shape of one cycle — the harmonic recipe riding on the fundamental; your ear perceives it as timbre, the quality that lets you tell a trumpet from a flute playing the same note at the same volume. If you wrote "waveform = volume" or "frequency = loudness," you fudged the one distinction the rest of the book stands on: pitch, loudness, and tone are three independent dials.
A2. Octave math. An octave is a doubling of frequency. (a) One octave up from 55 Hz is 110 Hz; two octaves is 220 Hz; three is 440 Hz. (b) 165 Hz is not an octave of 55 Hz — octaves are 55, 110, 220, 440…, all powers-of-two multiples, and 165 is 3× the fundamental, which makes it the third harmonic (a musical fifth above the octave), not an octave. You know this without hearing it because the math is unambiguous: 165 ÷ 55 = 3, and 3 is not a power of 2. (c) Doubling from 20 Hz: 20 → 40 → 80 → 160 → 320 → 640 → 1,280 → 2,560 → 5,120 → 10,240 → 20,480 Hz. That's ten doublings to clear 20 kHz, which is why "roughly ten octaves" is the stock description of human hearing.
A3. The dB ladder. (a) +20 dB SPL is 10× the sound pressure, so a snare 20 dB above the vocal is pushing roughly ten times the pressure. (b) Using the rough "+10 dB ≈ twice as loud" rule, 20 dB is two doublings, so the snare sounds about four times as loud. (c) The answers differ wildly — 10× vs 4× — because physical pressure and perceived loudness are different scales. Pressure is what the air and the meter do; loudness is what your brain does, and your brain compresses the range hard (that's why we hear across a trillion-to-one pressure span without going deaf). When you're riding faders by ear, the perceived number is the one that matters — you're mixing for a listener's brain, not for a barometer.
A4. Recipe reading. Fundamental 110 Hz with partials at 220, 330, 440, 550. (a) Those are harmonics 2, 3, 4, and 5 (each an integer multiple of 110). (b) The two that are octaves of the fundamental are 220 Hz (one octave) and 440 Hz (two octaves) — the power-of-two multiples again. (c) Strengthening the harmonics above 1 kHz makes the sound brighter, because brightness is just where the harmonic energy concentrates. The pitch does not change — pitch is set by the fundamental (110 Hz) and the spacing of the harmonic series, not by which harmonics are loudest. This is the whole principle behind EQ and saturation: you can change a sound's character without changing its note.
A6. The transient question. (a) What got deleted is the transient — the attack, the first few milliseconds where the note is struck before it settles into steady sustain. (b) Beyond instrument identity, transients carry rhythm (they're how your brain locates exactly where the beat is — the edge, not the body) and perceived loudness/punch (a sharp transient reads as "hits hard" even when the average level is modest). They also carry articulation — the difference between a soft and a hard piano keystroke lives in the attack. (c) Reversing the note turns a "hit" into a "bloom" because you've reversed the envelope: a piano note normally has a fast attack and a long decaying tail, so played backward it becomes a long swelling rise into a sudden stop. The hit moves to the end. That slow crescendo-into-silence is exactly why reversed piano and cymbal swells are the universal transition tool — they build tension and then hand off.
A8. Spot the flawed claim. (a) Wrong: higher frequencies do not travel faster — all audible frequencies move at essentially the same speed in air (~343 m/s), which is why music arrives coherent and not smeared by pitch. Hi-hats feel immediate because of their transients, not their speed. (b) Wrong: "85 dB" alone is incomplete — dB is a ratio and needs a reference. 85 dB SPL (sound in a room) and -85 dBFS (a level inside your DAW) are wildly different statements. Always name the reference. (c) Wrong: timbre is determined by the harmonic recipe and envelope, not the fundamental — that's the entire point of A4. Two instruments share a fundamental and still sound nothing alike. (d) Wrong: even if you can't hear above 16 kHz, plenty of listeners (and the harmonics that shape lower sounds) live up there, and content above your personal ceiling isn't "wasted on everyone" — it's wasted on you, today, on this system. Mix for the audience, not your fatigue line.
D1. The translation post-mortem (answer shape). A strong response is diagnosis only — the exercise forbids fixes, and that restraint is half the grade. It names a likely band for each symptom: weak bass on bookshelf speakers = the small woofers can't reproduce the sub/low fundamentals the headphones delivered; buried vocal in the car = the vocal's mid/high-mid presence is being masked or the low-mid mud is blanketing it; ear-splitting cymbals in the car = high-mid/air harshness the car's tuning exaggerates. The key insight the answer must reach: the failures differ by system precisely because each system has a different frequency response — so the same file meets a different filter each time. That's not random; it's deterministic. Then it proposes gathering more receipts: the exact frequencies, a spectrum reading, the same track on a fourth system to triangulate.
Chapter 4 — Listening
A2. The map from memory (answer key for the six bands). Sub: roughly 20–60 Hz, felt more than heard — 808 fundamentals. Lows: 60–200 Hz, warmth and punch — kick body, bass meat. Low-mids: 200–500 Hz, body and the mud zone — nearly everything lives here. Mids: 500 Hz–2 kHz, intelligibility — vocal vowels, the telephone band. High-mids: 2–6 kHz, presence and attack — consonants, pick attack, the harshness tightrope. Air: 6 kHz and up, sheen and breath — cymbal shimmer, sibilance. The chapter predicts most people misplace the mids / high-mids border first; if you slid it, note which way you bent it, because that's the band your ear under- or over-weights.
A3. The two traps. (a) Mixing for six hours at whisper volume corrupts bass decisions specifically: the Fletcher–Munson (equal-loudness) curves say your ears are least sensitive to lows at quiet levels, so you'll hear the bass as too weak and overcompensate — bass-boost the mix to sound right quietly, and it'll be boomy and bloated everywhere else. (b) The limiter comparison is rigged because one version is "noticeably hotter" — louder always wins by reflex, so they're not comparing limiter quality, they're comparing level. The one procedural rule that fixes it: match loudness before judging. Same perceived volume, then decide. (This is the law Chapter 22 builds and the whole mastering section re-ratifies.)
A6. The threshold, defended (answer shape). A strong paragraph defends "you hear with your brain, not your ears, therefore hearing is trainable" using at least two named phenomena — the cocktail-party effect (your brain isolates one voice in a noisy room, proving hearing is active attention, not passive recording), phonemic restoration or the McGurk effect (the brain fills in or overrides what the ear sends), or the missing fundamental (you "hear" a bass note the speaker never produced). The chain of logic: if perception is constructed by the brain rather than delivered by the ear, then perception can be trained like any brain skill. The practical payoff the answer must land: a producer on a budget should invest in ears (free, trainable) before gear (expensive, capped) — the book's first theme stated as a spending priority.
Chapter 10 — Room Acoustics
A1. The three lies, from memory (answer key). The three families of small-room problems: (1) Modal problems / standing waves — caused physically by low frequencies whose wavelengths fit the room's dimensions, reinforcing at some spots and cancelling at others; sound like bass notes that boom or vanish depending where you sit; the $0 test is the walk-the-room bass hunt (crouch and stand while a bass line loops); first fix is bass trapping the corners and moving your chair off the null. **(2) Comb filtering / early reflections** — caused by a reflection arriving a few milliseconds after the direct sound and summing with it; sounds hollow, phasey, smeared imaging; the $0 test is the mirror trick; first fix is absorption at the first-reflection points. (3) Flutter echo / reverberation — caused by sound bouncing between parallel hard surfaces; sounds like a metallic zing or a long ring; the $0 test is the clap test; first fix is breaking up one of the two parallel surfaces with absorption.
A4. The null arithmetic. Your chair sits in a 60 Hz null; you boost the monitors +6 dB at 60 Hz. At the chair, almost nothing improves — a null is a physical cancellation at that point in space; you're feeding more energy into a spot where the room is subtracting it, so you spend headroom and still hear a hole. At the corner behind you, where 60 Hz already piles up (modes peak at boundaries), the boost makes the boom worse. Four feet to your left, off the null, 60 Hz is suddenly far too loud, because there the boost adds to energy that was never cancelled. The one-sentence rule: the only two things that genuinely address a null are moving your chair (or speakers) out of it and bass trapping to reduce the mode's strength — and the first one is free. You cannot EQ a hole in space; the hole isn't in the signal.
A6. Thickness is the purchase. A 2-inch mineral-wool panel works by absorbing air motion, and air velocity is highest a quarter-wavelength off the wall. At 2 kHz the wavelength is short (~17 cm), so a 2-inch panel sits right in the high-velocity zone — honest, full absorption. At 300 Hz the wavelength is longer (~1.1 m), so 2 inches only catches a fraction of the motion — partial work. At 80 Hz the wavelength is enormous (~4.3 m) and a 2-inch panel sits in near-dead air against the wall — nearly nothing. Two ways to extend the same panel downward without buying thicker stock: (1) mount it on an air gap off the wall (the gap moves it into the higher-velocity region for lower frequencies), and (2) straddle it across a corner (a "superchunk" position) where bass energy concentrates. The blow test checks porosity — you should be able to blow air through the material; if you can't, it's too dense to absorb and will reflect instead, costing the panel its broadband reach.
B1. The clap-test tour (assessment criteria). A strong five-room survey delivers the journal table the exercise asks for — room / flutter? / pitch? / tail rank / one-word character — and shows three things: that you can hear flutter (the metallic zing between parallel hard surfaces) and name which two walls cause it; that you can hear modal pitch (a singable note in the tail vs. a pitchless decay); and that you can rank the five rooms by tail length, which is a hands-on RT60 survey. The payoff isn't a "right" set of rooms — bathrooms ring, closets are dead, that's physics — it's that your studio room's row stops being a vibe and becomes a measurement you made with your hands. If your studio room rings or booms, you've just written your first treatment to-do.
C4. Compute your mode map (assessment criteria + worked arithmetic). The arithmetic is checkable even though your room isn't: for each dimension, the first axial mode is f ≈ 565 ÷ L (L in feet), then 2f, 3f, 4f. A 10-foot wall gives 56.5, 113, 170, 226 Hz; an 11-foot wall gives 51, 103, 154, 205 Hz; an 8-foot ceiling gives 70.6, 141, 212 Hz. A strong response: builds the combined table sorted by frequency, flags any two modes within a few Hz of each other (that's stacking — those frequencies bloom hardest), flags wide bass gaps (the "broken xylophone" zones), maps the chair as a fraction of each dimension (mode 1 nulls at the halfway line, mode 2 at the quarter points), and then verifies by ear — plays sines at the predicted trouble frequencies and walks them. The grade is in the comparison: predicted and confirmed means you own the physics; predicted and absent or surprise means the simple rectangular model missed your bed, doorway, and bookshelf — which also vote.
D1. Choose the room (answer shape). A strong memo computes both rooms' first-mode series, assesses stacking (a square room is worst — both horizontal dimensions land on the same modes and double the trouble) and bass gaps, then weighs the non-modal factors the chapter taught: window real estate stolen from panel real estate, floor reflection, where left-right symmetry is achievable (your speakers want matched side walls), and tracking-vs-mixing duty. The verdict names a pick, the three strongest acoustic reasons, the honest cost of what the loser gave up, and a first $100 of treatment specific to that room. The shape that earns it: every claim is tied to a number you computed or a reflection you can point at — not "Room A feels better."
Chapter 22 — EQ
A1. The vocabulary (answer key). High-pass removes everything below its corner (lets highs pass); low-pass removes everything above (lets lows pass); shelf lifts or cuts everything beyond a corner by a fixed amount (a plateau); bell boosts or cuts a band centered on one frequency (a hump or notch). Q is the width of a bell — and here's the direction trap: higher Q = narrower, not wider. The rule of thumb: wide (low Q) for musical, broad tone-shaping; narrow (high Q) for surgical problem-solving — finding and killing one resonance.
A2. Three reasons, three sentences. The subtractive-first doctrine: cut before you boost. Reason one (masking): cutting a crowded band un-masks the parts you want to hear, which a boost can't do. Reason two (headroom): cuts lower level and preserve your gain structure, while boosts eat headroom and stack toward clipping. Reason three (boost-bias trap): a boost adds level, and added level always sounds "better," so you'll keep boosts you shouldn't unless you match loudness. The boundary clause that keeps "cut first" from curdling into "never boost": when a featured element's character lives in a band nothing else occupies, or you're fixing a dull capture you can't re-record, a boost is the right first move — doctrine with reasons survives exceptions.
A4. Anatomy of a committee. Mud at 200–400 Hz is the most common mix problem on Earth because of where energy piles up. Fundamentals of many instruments (low vocals, guitars, snare, piano, pads, toms) sit in or just below this band, and the low-order harmonics of even-lower sources land right in it — so nearly every track makes a deposit here. Recorded room sound adds its own low-mid thickness on top. And masking spreads upward, so this congestion smears the clarity of everything above it. The reason solo interrogation acquits every suspect: each track sounds warm and fine alone — the mud only exists in the sum, where all the small deposits add up. That's why the fix is distributed (a small cut on each of the two or three biggest contributors) rather than concentrated (one big cut on one track that wasn't individually guilty).
A5. The Matched-Loudness A/B Law. From memory: no EQ move is judged until the processed and bypassed signals are at the same perceived loudness. The four-step ritual: (1) make the move; (2) trim the output until toggling bypass produces no jump in size or level; (3) toggle blind, eyes away, so you're not voting with your eyes; (4) render one of three verdicts. The three verdicts: better (keep), worse (undo), different-but-not-better (bypass wins by default — because if a move only changes the sound without improving it, you've added complexity and risk for nothing, which is theme T2, less is more). It matters more for boosts than cuts because boosts add level, and added level is the exact thing that fakes "better" — the bias rides on top of the boost.
A7. The two lies of the solo button. Lie one: solo makes a track sound like it matters more than it does — isolated, every element seems to need more presence, more low end, more everything, because nothing is masking it. Lie two: solo hides committee crimes — a problem that only exists in the sum (like 200–400 Hz mud) disappears when you solo any single contributor, so you'll never find it one track at a time. The working rule that survives: solo to find a problem frequency, mix to make the decision. The lie that cost Jaylen his three days in the chapter hook is lie two — his vocal sample was buried by a committee of mud he couldn't find because soloing each track acquitted it; "the crime only exists in the sum" is the precise description of a committee, because no single member did it.
Chapter 30 — Mix Troubleshooting
A1. The symptom table (answer key). The eight rows, each as symptom → what you hear → address → 60-second check → first fix: - Muddy → blanket over the mix → 200–400 Hz → mute your top low-mid depositors during the thick section → distributed small cuts. - Harsh → fatiguing edge, you turn it down → 2–6 kHz → low-level listen + mute bright tracks one at a time → narrow cut at the worst offender; check for distortion stacking. - Thin → no body, stands on stilts → missing 60–250 Hz → solo the low end, check what's high-passed too aggressively → restore lows / back off an over-zealous high-pass. - Boxy → cardboard, small-room honk → 300–600 Hz → sweep a -4 dB bell across the vocal → cut at the found frequency. - Boomy → one note hangs and dominates → 60–200 Hz, often a room mode → walk the room / check the note against others → tame the resonant note; suspect the room. - Sibilant → ess/tee spit → 5–10 kHz → solo and scrub the esses → de-esser before broad EQ. - Lifeless → flat, no movement → no frequency address — it's the dynamics/automation axis → does the second chorus add anything? → automation and arrangement, not EQ. - Small → fine here, tiny everywhere else → no frequency address — translation/balance axis → the quiet test and the phone test → balance and arrangement.
The three rows with no frequency address are lifeless, small, and cluttered — lifeless lives on the dynamics/movement axis, small on the translation/balance axis, cluttered on the arrangement/density axis. A spectrum analyzer is blind to all three, which is why ears outrank meters here.
A3. The attribution test. A symptom that shows up on one system and nowhere else: the protocol is to play a mastered reference track through the same system. If the reference also exhibits the symptom, it's the system's accent, not your mix — cross it off your fix list. If the reference plays clean and only yours is broken, it's your mix — keep it. A mastered reference is the right measuring instrument because it's a known-good signal: a record that's already proven it translates everywhere, so any failure it shows is the system's, not the music's. Outcome A (reference fails too): do nothing to the mix; learn the accent. Outcome B (reference is fine): convict your mix and fix it.
A6. Off the map. Lifeless lives on the dynamics/movement axis — a spectrum analyzer is blind to it because the frequency balance can be perfect while nothing moves; 60-second check: does the energy curve build, does the second chorus add layers? First fix: automation/arrangement. Small lives on the translation axis — the analyzer can't see it because it's about how the balance survives a level drop or a worse speaker; check: the quiet test and the phone test; first fix: balance/arrangement. Cluttered lives on the arrangement/density axis — the analyzer shows a full spectrum and calls it healthy, but too many parts occupy one band; check: the mute test; first fix: mute parts (arrangement). The one described as "not a mix problem at all" is cluttered — and that's the table's most important row, because it's the threshold concept from Chapter 20 in disguise: a mix problem is usually an arrangement problem, and no EQ move fixes too many parts.
B1. The archive autopsy (assessment criteria). A strong autopsy of your two oldest mixes names the top three symptoms per mix, each with an address (band or axis), a timestamp ("chorus only," "from the bridge on"), and a suspect list — and fixes nothing, because these are teaching patients. The grade is in the discipline (diagnosis without surgery) and in the vocabulary (addresses, not verdicts). The payoff the exercise is fishing for: across a stack of honestly-diagnosed mixes, you'll see your mixes fail in characteristic ways — the same two or three diseases recur — and that pattern is the single most valuable page in your journal, because it tells you what to prevent in your next production.
B3. The three-differences drill (assessment criteria). Strong response: a reference level-matched down to your mix (so neither feels louder), then exactly three differences written in band-or-axis vocabulary, each converted into a fix-list item with a guessed stage (arrangement / source / edit / mix). Why exactly three, not one or ten: one difference is too few to reveal a pattern and lets you fixate; ten overflows working memory and turns into a vague "it's just worse" — three is the number a human can hold, prioritize, and act on in one sitting. The drill trains the muscle that converts a comparison into a to-do list.
C1. The diagnostic pass (assessment criteria). This is the chapter's centerpiece, run across two days. Day one: a multi-system tour with symptoms written in band-and-axis vocabulary and nothing fixed. Day two, fresh ears: 60-second checks on each symptom, a ranked fix list (five items max) sorted by impact on vocal and groove, executed one change at a time at matched loudness, each re-verified on the system that exposed it, with every verdict logged — including the items that turn out to be deliberate choices, not diseases. A strong pass demonstrates all three disciplines: diagnosis and treatment in separate sittings, matched loudness on every verdict, one change at a time. The deliverable is the fix log, and the chapter's warning is the grade: the mastering stage "will not accept it sounds good in here."
C4. Re-balance versus rescue (assessment criteria + the expected finding). Two versions of your worst old mix: a rescue path (60 minutes, symptom table and plugins, repair it as it stands) and a re-balance path (every fader to silence, rebuild the static balance most-important-element-first per Chapter 20, 90 minutes faders-and-mutes only before any plugin). Bounce both gain-matched, listen blind the next morning on a phone and one other system, and write which translates better. The expected finding — and the chapter's claim — is that the re-balance usually wins, because most "mix problems" are really balance problems that no amount of plugin rescue can out-process. A strong response either confirms that with a clear verdict and a foundations lesson, or honestly reports the opposite and explains why (the patient's foundation was healthier than its symptoms looked) — both are real findings; the only failing answer is one with no level-matched verdict.
Chapter 33 — Loudness, Streaming, and the Modern Platform
A1. The four-decibel family. dB SPL — sound pressure in the air, what a meter reads in a room (entered Chapter 1). dBFS — level inside the digital domain, where 0 is the ceiling and everything useful is negative (Chapter 2). dBTP (true peak) — the real peak including the inter-sample overshoots that appear when the signal is reconstructed, read by an oversampling meter (Chapter 32). LUFS — loudness units full scale, a perceptual average weighted to how you actually hear, the unit platforms normalize to (Chapter 33). The forum sentence "My master peaks at -14 dBFS so Spotify won't touch it" has at least two confusions: it mixes up peak (dBFS) with loudness (LUFS) — Spotify normalizes by integrated LUFS, not peak — and it misstates the reference, since ≈ -14 is a LUFS target, not a dBFS peak. A master can peak near 0 dBFS and still measure -14 LUFS, or peak at -14 dBFS and measure something else entirely. Peak and loudness are different axes.
A2. Three windows, three questions. Momentary (~400 ms window) answers "how loud is it right now?" — watch it while riding a limiter or catching a stab. Short-term (~3 s window) answers "how loud is this section?" — watch it to check that your chorus actually lifts above your verse. Integrated (whole file) answers "how loud is the entire song, on average?" — the single number that defines the master. Platform normalization uses integrated, because it has to turn the whole track up or down by one consistent amount; momentary and short-term jump around constantly, so normalizing to them would make the playback level lurch second to second — unusable.
A3. The gate, explained to a drummer. Reassurance in plain words: LUFS measurement has two gates that protect exactly his worry. The absolute gate throws out everything below roughly -70 LUFS — true silence — so it doesn't count. The relative gate throws out everything more than 10 LU below the running average — which means the 40 seconds of quiet brushed cymbals get excluded from the integrated number entirely. The song will report a loudness based on the full-kit sections that actually carry the energy, not dragged down by the intro. So nothing gets "turned up into distortion": the quiet intro plays quiet, the loud part plays at its real level, and the platform's one offset is computed from the parts that count. The relative gate exists to close exactly this loophole — without it, you could game the system (or accidentally sabotage yourself) by padding a song with silence to lower its average.
A4. PLR arithmetic. PLR (peak-to-loudness ratio) = true peak minus integrated loudness. (a) -1.0 dBTP and -9.0 LUFS → PLR = 8.0 dB — a loud, modern/aggressive neighborhood. (b) PLR 14.5 dB with true peak -1.5 dBTP → integrated = -1.5 − 14.5 = -16.0 LUFS — a dynamic, open master. (c) Two masters both at -11.0 integrated, one at -1.0 dBTP and one at -4.0 dBTP: same loudness, but the -4.0 one has 3 dB more peak headroom, i.e. a higher PLR and more preserved transient — it almost certainly met a gentler limiter, leaving its peaks more intact. Same average, different punch.
A5. Normalization offsets. Reference ≈ -14 LUFS, turn-down-only platform. A master at -7.8 gets turned down 6.2 dB to play at -14. -11.2 gets turned down 2.8 dB to -14. -14.0 is already there — no change, plays at -14. -17.5 is quieter than reference, so on a turn-down-only platform it gets no change and plays at -17.5 (quieter than its neighbors); the master whose experience differs most by platform policy is this one — if the platform also turns quiet tracks up, it gets boosted to -14, but if it's turn-down-only, it just plays soft. The judgment question: none of these four numbers alone proves anything is wrong — loudness is a creative choice, and -17.5 might be a perfectly intentional dynamic folk master. What you'd check before deciding: the true peak (is it clipping?) and the PLR / how it sounds at matched loudness against references — the meter is a measuring tape, not a verdict.
A7. Does / doesn't (answer key). What normalization CAN change: the playback level of your track (one offset, up or down). That's it — one entry, which is rather the point. What it CANNOT change: your dynamics/PLR, your frequency balance, your distortion (a slammed master stays slammed, just quieter), your transients, your stereo width, the relationships between elements in your mix, or whether your low end translates. The one hedged exception: some platforms apply a protective limiter when they turn a quiet track up and its peaks would exceed the ceiling — that's actual processing, and it triggers only on the turn-up path, only on platforms that turn up at all. The headline: normalization changes your volume knob, not your master.
B1. Death Magnetic, both verdicts (assessment criteria). Strong response logs Death Magnetic in symptom vocabulary (crushed transients, fatiguing sustain, distortion riding the loud passages, the timestamp fatigue first registers) and then the earlier catalog at matched loudness by ear, same three observations. The honesty the exercise demands: name one thing the 2008 master genuinely bought — it was loud and aggressive on un-normalized playback, which in 2008 was most playback (radio, CD, shuffle), and aggression suited the music — the war wasn't run by fools. Then the verdict: that purchase largely evaporates under normalization, because the platform turns the loud master down to match everything else, leaving only the distortion and crushed dynamics with no loudness advantage to show for them. A strong answer reaches that "the weapon was confiscated" conclusion from its own ears.
B3. The normalization toggle (assessment criteria). Strong response: a cross-era playlist played with normalization OFF, hands off the volume, journaling every lurch (the 1970s master quiet, the 2008 slam jumping out), then the same playlist ON, journaling that the lurches flatten — and crucially, the observation that nothing changed within each track, only the level between them. In this chapter's vocabulary, the feature is integrated-LUFS normalization doing the loudness-matching the loudness war tried to win by brute force — the comparison got matched at playback, by the platform, which is the whole point of the chapter.
B6. The stats receipt (assessment criteria). On the video platform that shows its loudness math (the content loudness readout in the stats panel, as of this writing), a strong response logs five tracks' readings and decodes each: the platform measured the track's integrated loudness, compared it to the platform reference, and turned it down by the difference (a negative content-loudness value = "we attenuated this much"; near zero = "this track was already at reference"). Connecting hearing to number: the track showing a large negative value should sound like a slammed, dense master that's been turned down (and may feel oddly flat for its density); the near-zero one should sound like it was made for this world. The grade is reading the number and hearing what produced it.
C1. Your master's passport (assessment criteria). Strong response: a whole-file analysis (not a mid-song glance) of your Chapter 32 master logging integrated LUFS, true peak, PLR (subtraction done by hand once), and LRA, then running gate one — true peak ≤ -1.0 dBTP. If it fails, back to the limiter with true-peak detection on, re-bounce, re-measure. The lesson the exercise wants you to write down: the fix usually costs nothing audible — a master that respects -1 dBTP sounds the same as one that doesn't, minus the inter-sample clipping risk on lossy playback. The discipline is "integrated means whole-file."
C2 / C4. The three-masters experiment & reference cartography (assessment criteria). C2 is the chapter's highest-value evening: three prints of your track varying only the limiter push (your keeper, a deliberate honest slam, a breather backed off), all measured whole-file, then trimmed to a common ≈ -14 playback to simulate normalization, then blind-judged by at least two jurors besides you on punch / size / fatigue / playlist-pick. A strong response collects scores, decodes them, and writes the ledger — what each master bought, what each paid — landing on a settled target with one sentence of justification. The near-universal finding: once loudness is matched, the slam stops winning and often loses on punch and fatigue, because PLR is what survives the match. C4 grounds it: whole-file numbers on your three references give you your genre's measured loudness and PLR neighborhood, and where your master sits in that family.
C6. The -16 podcast conform (assessment criteria + the spec). The target as of this writing: ≈ -16 LUFS integrated (stereo), true peak ≤ -1.0 dBTP, segments held within about 1 LU short-term of each other. The order is the graded part: process first (subtractive EQ and leveling compression — the Chapter 22–23 voice chain), then measure, then adjust overall gain, then re-measure — never ride a fader toward a LUFS number while listening, because the meter is a measuring tape, not a compass. The one sentence the exercise fishes for: the leveling compression does what a gain trim can't — a trim moves the whole file up or down and leaves the loud-soft swings intact, so a guest who whispers and shouts still whiplashes; the compressor evens the internal dynamics so the integrated target actually reflects a consistent listening experience.
D1. The two-masters memo (answer shape). Strong memo to the club-genre friend: explains the un-normalized world (live DJ systems, club PAs, some DJ software, downloads played raw — places where loudness still competes) versus the normalized world (streaming), states that a second, hotter/club-optimized master buys real loudness exactly in the un-normalized contexts and nothing in the normalized ones, then prices the second master honestly — extra files, extra QA, version-discipline risk (Chapter 19 has the scars). The verdict picks one scenario and commits: weekend DJ at real venues → yes, cut the club master; streaming-only bedroom releases → no, one streaming master, don't pay the version-control tax for a world you don't play in. The shape that earns it: the recommendation follows from where the music will actually be heard, not from "louder is safer."
D3. Three clients, three targets (answer key shape). (a) Techno for club DJ sets: loudness matters in the un-normalized club, so a louder/denser master is defensible (lower PLR), but it must NOT sacrifice the low-end clarity and sub weight the club PA exists to deliver — don't crush the kick to chase numbers. (b) Folk singer-songwriter to streaming, home listeners: prioritize dynamics and a higher PLR — the audience is intimate and normalized, so loudness buys nothing; the master must NOT be limited into flatness that kills the vocal's breath and the arrangement's lift. (c) News podcast, two co-hosts: target ≈ -16 LUFS integrated, ≤ -1 dBTP, and the cardinal rule is NOT to let the two hosts' levels drift apart — consistency between voices and segments matters more than any absolute number. The test of the answer: the three targets differ, and every difference has a stated reason rooted in where and how the audience listens.
M1. (Ch 2 + 32 + 33) Why dBTP. Digital audio "isn't the stairsteps" — the samples are points, and the converter reconstructs a smooth band-limited waveform between them (Chapter 2). Because the real waveform rides between the samples, its true peak can overshoot the highest sample value — an inter-sample over (Chapter 32). So a master that reads -0.1 dBFS on a sample-peak meter can actually peak above 0 when reconstructed or after lossy encoding, clipping on playback. That's why the delivery gate is stated in dBTP (true peak, measured with oversampling) rather than dBFS — and why the recommended ceiling is ≤ -1 dBTP, a decibel of margin to absorb the overshoot the stairstep picture hides.
M2. (Ch 4 + 33) Equal-loudness, twice. One psychoacoustic fact, two consequences. The Fletcher–Munson curves say louder is perceived as better and fuller, especially in the bass and the highs. Consequence one: that fact powered the loudness war — whoever was louder won the A/B on the radio and the CD changer, so everyone pushed limiters until the music bled. Consequence two: the same fact is why your Chapter 8 calibrated listening level matters when judging masters — if you audition at the wrong volume, the curves bend your perception of the bass and air, and you'll make EQ and limiting decisions that only sound right at that volume. One mechanism (loudness changes perceived balance), two places it bites (the war, and your monitoring).
M4. (Ch 21 + 33) Crest factor and PLR. Crest factor (Chapter 21) is peak minus average (RMS) level on a single signal or track — how "spiky" it is. PLR (this chapter) is true peak minus integrated loudness on a finished master — the same idea, scaled up to the whole song and using a perceptual average. The relationship: both measure how much dynamic headroom sits above the typical level — crest factor at the tracking scale (one drum hit's transient above its body), PLR at the mastering scale (the whole master's peaks above its loudness). An unmastered print held at ~-6 dB peaks with a much lower average carries a high PLR — roughly 12–18 dB depending on the music. During Chapter 32, mastering spends that headroom: the limiter pulls the peaks down and the average up, lowering the PLR — the dynamic range "goes" into loudness, which is exactly the trade you're deciding how far to make.
The Listening Lab and DAW Lab exercises have no answer key by design — the answers are in your ears and your session by now, or will be soon. Where this appendix gave you criteria instead of a solution, grade yourself against the four-point rubric at the top, then go re-run the exercise on a different track. The struggle was the lesson; this page was only ever for checking your work after you did it.