Chapter 19 Exercises

Workflow is the one skill in this book you cannot fake on talent, and the only one where practice is literally free — no plugin, no mic, no treated room required. The DAW Lab is the center of gravity: C1 (the hygiene resurrection) and C2 (the stem-printing drill with verification) are the two exercises Part V silently assumes you've done, and C5 (the real feedback exchange) is the one that changes your trajectory, because it puts a second pair of ears on your work for the first time with a protocol protecting both of you. Work Part A first to load the definitions, do the Listening Lab before you give anyone notes (you'll discover how hard good notes are to write before you're tempted to send lazy ones), and keep your Chapter 4 listening journal open throughout — feedback literacy is listening literacy with a send button. Time estimates are honest averages. Exercises marked ★ have worked answers in the back of the book (Answers to Selected Exercises).

Part A: Conceptual Understanding

A1. Stems or multitracks — four recipients, four verdicts. ★ (15 min) Define both terms from memory, including roughly how many files each produces from a 34-track, 6-bus session. Then, for four recipients — (a) a mix engineer hired to mix the song from scratch, (b) a remixer making a club flip, (c) a vocalist recording a feature, (d) a live playback rig for a festival set — state which package each gets, how many files that probably is, and the one-sentence reason. Finish with the clarifying question you'd ask anyone who emails you the bare word "stems," and what a wrong guess costs.

A2. The version grammar, applied to a crime scene. (15 min) A collaborator's bounces folder contains: mixdown.wav, mixdown_new.wav, mixdown_FINAL.wav, mixdown_FINAL2.wav, mixdownFINALFORREAL.wav, and USE THIS ONE.wav — all with different modification dates, two with identical file sizes. First, diagnose: name the two specific failures this naming system guarantees (think: which file did the label approve? which one is on the singer's phone?). Second, rebuild: rename all six files under the chapter's grammar (version numbers, stage tags), inventing plausible history as needed. Third, write the two-sentence explanation of why "final" is banned that you'd actually send this collaborator without sounding like a textbook.

A3. The 3-2-1 audit, on paper. ★ (10 min) Map your real situation right now: list every copy of your current project that exists, what it's stored on, and where that storage physically lives. Score yourself against 3-2-1 honestly — remembering that copies on the same drive count once, that a backup drive in the laptop bag shares the laptop's fate, and that sync-without-versioning propagates deletions. Then name your single point of failure (everyone has one) and the cheapest move that eliminates it. Write the move down with a date you'll do it by. This exercise has saved more music than any plugin ever made.

A4. Why zero, why same length. (10 min) Explain, in your own words, why every printed file must start at the project's first sample even when the part enters at 2:14 — your answer should include what a bare WAV file does and doesn't know about itself, and what the leading silence actually encodes. Then explain what the same-length rule lets a recipient verify at a glance. Finish with the tradeoff ledger: what do these rules cost (disk space, render time), and why does the chapter rate that cost as trivially cheap against what it buys?

A5. The wet/dry ledger, five elements. ★ (15 min) For each of these five elements in a multitrack print, rule wet (processing baked in), dry (processing omitted), printed separately, or both-versions — and defend each ruling with the chapter's doctrine (identity processing prints; shared space prints separately; the 2-bus prints never): (a) a hard-tuned vocal hook where the tuning is the aesthetic; (b) a riser whose long reverb tail was designed into the sound in your Chapter 14 session; (c) the reverb send that the snare, the vocal, and the pad all share; (d) the pad with the slow filter-open automation that defines the chorus lift; (e) the gentle limiter sitting on your master bus "to make it louder." One of these five has only one acceptable answer — say which, and why it's not a judgment call.

A6. Translate the smoke detector. (15 min) Five real notes, as civilians send them: "it's not slapping," "make the vocals pop more," "something feels off in the middle," "it's too busy," "needs more energy." For each: (1) rewrite it as a protocol-grade note by inventing a plausible timestamp, element, and observation; (2) list two different underlying problems that could produce that exact complaint — the smoke alarm's two candidate fires. (Example of the move, using a sixth note: "too muddy" might be a 200–400 Hz pileup, or might be an arrangement playing too many mid-range parts at once — Chapter 16's disease wearing Chapter 22's symptom.) Notice, when you finish, which step was harder. That difficulty is why the chapter says the observation is the valuable half and the diagnosis is a lead.

A7. The README, from memory, for real. (10 min) Without looking at the chapter, write the complete README.txt for your current track as if it ships tonight to a collaborator you've never met: every technical fact they need, the file list with one-line notes, what you want from them, and how to reach you. Then open the chapter's checklist table and grade yourself: every omission you find is a question your collaborator would have had to email you at 2 a.m. their time. Keep the corrected version — it ships with your Project Checkpoint package.

A8. Roles in three configurations. (10 min) Assign the producer-roles vocabulary (producer, tracking engineer, mix engineer, mastering engineer, featured artist, songwriter/top-line) to three scenarios: (a) you, alone, making and releasing your own track; (b) a four-piece band self-recording in the drummer's basement with one member behind the laptop; (c) you sell a beat to a rapper who records at a commercial studio and hires an outside mixer. For each, note which roles share one skull, which role holds final say at each stage, and one place where ambiguity about a role could cost money or friendship later — and which chapter resolves that ambiguity (preview only; the percentages scene is Chapter 36).

Part B: 🎧 Listening Lab

This lab trains the two listening skills the chapter added to your stack: hearing what a rough mix communicates, and converting reactions into notes another human can use. Matched listening levels whenever you compare anything to anything — Chapter 4's law didn't retire when the faders got social.

B1. Notes on a stranger's mix. ★ (45 min) Go to the community mix forum attached to Mike Senior's free multitrack library (cambridge-mt.com — the discussion zone where people post their mixes of the library's songs for critique; as of this writing it's active and free). Pick one song you don't know, then pick two different members' mixes of it. For ONE of them, write a full protocol-grade consolidated note doc: open with two sentences on what the mix is doing well (orientation, not flattery — what should they protect?), then five to eight timestamped notes in [time] — element — observation form, each measured against what the mix seems to be going for rather than your taste. Then — this is the half people skip — compare the two mixes and notice how different two committed interpretations of identical multitracks can be. Write one sentence on what that does to the idea of a "correct" mix. File the note doc; you'll want the format again in C5.

B2. What a reference mix communicates. (30 min) Pick a professionally released track you know well and run this thought experiment: it just arrived in your inbox as the reference mix in a send package — the artist's way of saying "this is the intent." Listen twice (once full-attention, once on your worst playback) and write down everything the file tells you that a text brief never could: how loud the vocal rides over the track, how much space everything sits in, where the low end's weight lives, how wide the choruses bloom versus the verses, what the loudest and quietest moments are. Aim for ten concrete observations. That list is what your rough mix silently tells every collaborator who receives it — whether or not you meant to say it.

B3. Demo versus release — reconstruct the feedback round. (35 min) Find an officially released early version of a song alongside its final release — real options include Taylor Swift's voice-memo demos on the 1989 deluxe edition, the Beatles' Anthology takes against the canonical versions, or any artist's officially released demo/early upload you can pair with the album cut (Billie Eilish's early "Ocean Eyes" upload against later masters works too). Listen to both at matched loudness and write the consolidated note doc that — hypothetically — got the song from version one to version two: timestamped, specific, goal-anchored. You're reverse-engineering decisions, not reading minds; the exercise is noticing how many discrete, nameable changes separate a sketch from a record, and how few of them are "better gear."

B4. The stem census. (30 min) Take any multitrack you have rights to explore — your own session, or a download from the Cambridge library — and build the six-bus families from Chapter 6 if they don't exist (drums, bass, harmony/synths, guitars, vocals, FX). Solo each family for one full chorus and log: what frequency territory does this family own? What does the mix lose when it's gone — weight, rhythm, words, glue? Then play all six and write one sentence per family on its job. This is stem thinking: when you print six files for a stranger next week, each file should be a coherent answer to "what does this one do?"

B5. You, a month later, as the stranger. (20 min) Open a bounce of something you finished (or abandoned) at least a month ago — long enough that you-now is genuinely not you-then. Write ten timestamped protocol-grade notes to the producer (you-then), honoring every rule: open with what to protect, locate every observation, separate problem from prescription, measure against what the track was trying to be. Notice how much easier honest notes are when the artifact isn't wearing today's face. That ease is the target state for receiving notes on current work: the file is always just a file; the only thing that changes is how recently you bled on it.

Part C: 🎛️ DAW Lab

Hands on. Appendix E translates every command name (consolidate/glue, render in place/freeze/commit, collect all and save/copy media) into your DAW's dialect. Save a new session version before each drill — which is, conveniently, the first drill.

C1. The resurrection: full hygiene pass on your messiest session. ★ (60–90 min) Find the worst offender in your projects folder — every producer knows exactly which one it is. With a timer running, execute the chapter in order: (1) Save-as under the version grammar before touching anything (songname_v0.9_pre-hygiene is honest). (2) Name audit: every track renamed to pass the stranger test; count how many were called some flavor of Audio 12 and report the number in your journal without shame. (3) Color by family, matching your bus skeleton. (4) BENCH audit: every muted maybe-someday region moved to a BENCH folder/track group — exiled, not deleted. (5) Collect-all-and-save so every referenced sample actually lives in the project folder; note how many lived elsewhere (downloads folder findings are traditional). (6) Start _session-log.txt with one retroactively honest line. (7) Save-as the cleaned version with a stage tag. Deliverables: before/after screenshots, the orphaned-sample count, and the elapsed time — which is the number that converts you, because it's always smaller than the dread predicted.

C2. The stem-printing drill, with verification. ★ (60 min) On your current track: print the six bus stems AND the full multitracks, by the five rules — every file from zero, every file the same length (tails included), session sample rate and bit depth, numbered names in mixer order, wet/dry decided per element with the decisions written into the README (and the 2-bus chain bypassed; if you've been sneaking a limiter onto your master "to make it louder," this is where that habit gets caught). Then the half that makes it professional: create a brand-new empty session at the same sample rate, import the stems, drag everything to zero, faders flat — and A/B against your rough mix bounce at matched loudness. It should be nearly indistinguishable. Hunt every discrepancy to its cause: a still-muted track from the Chapter 16 mute test, a forgotten return, automation that lived on the master. Repeat the import test with the multitracks (expect a balance-less but complete version). Log every error you caught — each one is a disaster that didn't ship.

C3. The send package, assembled and tested. (40 min) Build the complete package from the chapter's checklist table for a named, real-or-realistic recipient: reference mix (version-tagged), the audio for the job, README.txt (your corrected A7 version), the three-sentence brief with one to three reference tracks, filenames carrying BPM/key/version, zipped under a proper name. Then run the stranger test for real: copy the zip to a different computer — or a different user account, or hand it to a friend — unzip it cold, and start a stopwatch. How long from "open folder" to "could begin working"? Under five minutes is the professional bar. Every minute past that, find the cause and fix the package, not the stopwatch.

C4. 3-2-1, implemented, with a restore drill. (45 min plus shopping) Stop auditing and build it: working copy, second medium (external drive — the cheapest insurance in audio), one offsite (cloud with version history, or a drive that lives at someone else's house). Automate or calendar the routine. Then the drill that makes it real: pick one old project and restore it using only the backup — original drive unplugged, no peeking. Open the session, press play, render ten seconds. Log everything that was missing or broken (orphaned samples are the classic; collect-all-and-save is the cure) and fix the routine so next month's restore is boring. A backup you've never restored from is a rumor; today you make yours a fact.

C5. The real feedback exchange. ★ (90 min across a week) The chapter's centerpiece, and it requires a second human: a producer friend, a bandmate, or a community like the one in this chapter's story (the Cambridge forum from B1, or any collab community you already lurk). The exchange runs both directions, by protocol: (1) Send your track as a proper package — C3's zip, including the brief that says what kind of notes you want. (2) Receive their consolidated doc; reply with thanks and nothing else for 24 hours. (3) Answer every note explicitly — accept, defer (with a reason: "that's a mix-stage fix"), or veto (with a reason measured against the goal). Implement the accepts; bounce the next version; send it back with a change list. (4) In return, write them a full protocol-grade doc on their track: orientation paragraph, five to eight timestamped notes, problems before prescriptions. Deliverables: both note docs and a three-sentence debrief — which note stung, which note you vetoed and why, and which of their problems you almost prescribed a fix for before catching yourself.

C6. Version-control week. (15 min/day for 7 days) For one week, run the full grammar on everything you touch: minor version at the end of every session, stage tags, one session-log line per sitting, exports carrying the version of the session that made them. On day seven, the proof: pick one decision from day three or four that you've since changed your mind about, and roll back — open the right historical version, recover the material, and bring it forward into the current session. Time the recovery. When it takes ninety seconds instead of a despairing evening, you'll understand why the chapter calls hygiene a creativity tool: cheap reversibility is what boldness costs.

C7. The time capsule. (45 min) Take one finished (or honorably abandoned) project and build its complete archival package: collected session with media inside, consolidated multitracks from zero, stems, the final mix or rough print, and an archival README — tempo, key, structure map, and the plugin manifest with settings noted for any irreplaceable sound. Zip it, size it, store it per 3-2-1. Then write the one-sentence answer to this prompt in your journal: "In 2036, with none of today's plugins, what could I rebuild from this capsule?" If the honest answer is "everything that matters," you've understood the sidebar. If not, the gap is your to-do list.

Part D: Synthesis & Critical Thinking

D1. Your collaboration one-pager. ★ (40 min) Write the document you'll send at the start of every future collaboration — one page, your voice, no legalese. It must cover: deliverables in both directions (formats, from-zero rule, naming), the version grammar you use, the feedback protocol (consolidated rounds, timestamps, who holds the baton, turnaround windows), approval rights (who has a veto over what — steal Demi's comp-approval clause if you sing, or grant it if you produce singers), and one line scheduling the percentages conversation before release (the page itself is Chapter 36's business; the scheduling of it is this chapter's). Pressure-test it against the chapter's disaster: would this document have prevented the 4 GB folder? The three days of silence? The drip-feed? Revise until yes, yes, and yes.

D2. The session plan for a nervous human. (30 min) A singer is coming to you in two weeks for their first real session — talented, green, and frightened of the red light. Write the producer's plan: the prep list completed before they arrive (Chapter 11's mechanics: chain tested, headphone mix drafted, lyrics printed and marked), the first fifteen minutes (what you say, what you deliberately don't), your talkback rules, your push/stop criteria written as observable signs (what does "energy rising" look like in this person? what does "diminishing returns" sound like?), and the gravy-take ask, scripted in your own words. End with one paragraph on ego management: what you will do, concretely, when their best take contains a flaw you love and they want to fix it.

D3. The rot premortem. (25 min) Open your oldest session today — however old that is — and audit it as evidence: what already complains on load (missing plugins? offline files? mystery names?), and what would break given ten more years (which plugins are bets on small companies, which sounds exist nowhere as printed audio, which samples live outside the folder)? Write the dated prediction in your journal: "By [year], this session will have lost: ___." Then spend twenty minutes making the time capsule (C7's recipe) that falsifies the prediction. The premortem habit generalizes: every session you close from now on, one question — what dies if the plugins do? — and one print before you walk away.

D4. The note you can't veto cheaply. (20 min) A collaborator with a real stake in the song — a featured artist, a co-writer — sends a note that collides with your core intent: they want the drumless bridge filled in; you built the song around that emptiness. Write two responses. First: the reply the feedback protocol prescribes (acknowledge, locate the underlying problem behind their note — what are they actually reacting to? — propose what you'd try, state what you're keeping and why, anchored to the shared goal). Second: the harder paragraph — what happens when the disagreement survives good protocol, and who decides? Connect your answer to the roles agreement from the first call (someone was named the decider before feelings were in the room) and to the honest limit: protocol doesn't eliminate creative conflict; it keeps creative conflict from becoming personal archaeology. If your roles agreement didn't name a decider, notice that the time to fix that was before this note arrived.

Part M: Mixed Practice

Interleaved retrieval — earlier chapters, this chapter's stakes.

M1. (Ch 2 + Ch 19) A collaborator sends you takes recorded at 44.1 kHz for your 48 kHz session, with nothing documented. Walk the failure: what happens if the files get dropped in and the mismatch goes unnoticed (think Chapter 2 — what does playing samples at the wrong rate do to pitch and timing?), what modern DAWs usually do automatically and why "usually" isn't a workflow, and which two lines in the README would have made the whole question vanish. Finish by stating the chapter's rule for what sample rate and bit depth prints should use.

M2. (Ch 6 + Ch 19) Inserts versus sends, now with money on it: explain why a shared reverb on a send/return complicates multitrack printing in a way an inserted effect doesn't — where does the reverb live in the routing, which file would it vanish from if you print tracks naively, and what's the chapter's prescription (print the returns as their own files)? Then connect upward: why does the same routing logic make stems-by-bus natural in a session that kept Chapter 6's six-bus discipline, and painful in one that didn't?

M3. (Ch 4 + Ch 19) Your collaborator sends notes on v1.2 after listening at a much louder volume than you mix at, and half the notes say the low end is huge and the vocal is buried. Using Chapter 4's equal-loudness curves, explain why listening level alone could produce exactly those two complaints, why "louder sounds better" corrupts comparative judgments, and what the feedback protocol should specify about listening conditions before anyone's notes are trusted. (One honest wrinkle: their phone-speaker note might still be the truest one in the doc. Why?)

M4. (Ch 11 + Ch 15 + Ch 19) Demi's deal had two clauses with deep roots: every full take delivered with a take log, and comp approval before her voice ships. Explain what the take log (Chapter 11) does for a remote comp that it doesn't even have to do for a local one, and why comp approval is the collaboration-era form of Chapter 15's editing-ethics conversation — whose performance truth is at stake, and what does the approval pass protect that crossfade craft can't?

M5. (Ch 16 + Ch 18 + Ch 19) Less-is-more, three chapters, one law: Chapter 16 benched parts that didn't defend their section; Chapter 18 had you keep violations only on purpose; this chapter bans sending files nobody asked for. State the single principle underneath all three in one sentence of your own. Then apply it to the send package: your track has nine muted alternate ideas you secretly love. What travels to the mixer, what stays in the BENCH, and under what circumstance does an alternate earn a seat in the zip?