Notes for Nerds

I want strong preferences with an escape route. Keep the source, signal path, and listening conditions in view, then ask what would change my mind. A favorite explanation should survive a difficult question. If I protect it from the evidence, I’ve built a shrine. The room is already short on space.

An experiment needs a musical question. I check the operating conditions before borrowing a technique. A clean-gain measurement won’t tell me how the same equipment behaves under deliberate overload. Peak matching after saturation can leave a loudness bias as crest factor changes. I keep audibility, cause, and preference separate.

The lyric gets first refusal on my production ideas. When a songwriter brings me a melody, I respect it as the starting point, then read what it’s trying to say. A neat rhyme can still spotlight the wrong word. I look for the specific image and the point where the thought turns. The melody needs to carry that thought.

I check lexical stress, sentence stress, and musical accent against one another. Duration, register, and an entrance after silence can outweigh the bar line. I want a displaced stress to earn its emphasis in the lyric. Prosodic friction can be useful. An accidental change in meaning is a less interesting experiment.

Eminem’s elastic pronunciation is a useful reference for the freedom I enjoy. I listen for what vowel length, consonant placement, and articulation do to the line before treating pitch accuracy as the verdict. Guitar taught me the same suspicion of correct notes. A phrase can be in tune and still miss its point.

A repeating progression needs an anchor, not a new identity every four bars. I can change voicing, inversion, bass motion, voice-leading, reharmonization, or harmonic rhythm. I can also leave the harmony alone and move a part’s register, entrance, or exit. First I identify what makes the repetition satisfying. Then I change the surroundings without operating on the thing that makes the return feel good.

I use the Nashville Number System’s Arabic scale-degree numbers to keep harmonic relationships legible when singers ask for a new key. The chart saves rewriting, but transposition still changes the vocal tessitura and the available instrumental voicings.

Classical training still helps me hear how the individual lines move, including the parallel fifths I learned to avoid in independent part-writing. Those habits aren’t a ban on the intervals I enjoy as a rock guitarist. I like knowing the convention well enough to hear what happens when I ignore it. I’m happy to find something by accident too. I just want to understand why it worked before calling the accident a method.

I want the articulation to carry as much character as the notes. I’m a six-string rock and metal guitarist by training, with a large Paul Gilbert-shaped dent in my playing. Alternate picking, string-skipping phrases, rhythmic attack, bends, and vibrato are the territory I keep returning to. Speed is available. Whether the phrase needs it is a separate investigation.

Standard tuning is my starting point. Depending on the song, I may retune the whole guitar a semitone down or up. I don’t use a capo. That keeps the interval layout familiar while changing where the open strings sit against the voice. I still have to judge the register and the part in the new key. A comfortable fingering isn’t automatically a good arrangement.

My beloved Fender Jim Root Jazzmaster has EMG pickups. The Ibanez RG Prestige in the studio has passive DiMarzios and technically belongs to a friend. It appears to have acquired permanent residency. That old Ibanez image of Gilbert with an RG still makes sense to me.

The first choices happen before the signal reaches a processor. I know what my SCHOEPS V4 U and Sony ECM-100U pair are likely to give me, which lets me respond to a performance without starting a microphone audition every time. The V4 U also looks excellent. I’m capable of appreciating an object and still asking it to do a job.

My Line Audio CM4 and Omni1 let me place a microphone where a larger body would get in the way, in the studio, live, or on production sound. Working at Vienna Synchron Stage sharpened that attention to placement. I want the direct-to-reverberant balance and off-axis contribution to make sense at capture. Equalizing later is a different intervention.

With sushi, I’m interested in what happens after the ingredients are chosen. The direction and thickness of the cut affect texture. The packing of the shari changes how it holds together and comes apart. Those decisions respond to the material in front of the chef. In a recording, I want to hear what’s worth preserving before I decide what needs changing.

Before softening a vocal consonant, I want to hear what it contributes to the phrase. The edge I’m removing may be carrying the intention. Good source material gives me something worth preserving. It doesn’t make that judgment for me. I’ll stop the food analogy there. I’m not dry-aging the snare.

The Black Box Analog Design MM-1 is my favorite preamplifier. I’ve tried and owned plenty of equipment, but this one stayed, and its rebuild made it more personal. Clean capture is my default, especially when a client’s production may change direction. On my own records, when the sound is clear in my head, I’ll drive it heavily and print that character. The mix will inherit the decision. I’d better mean it.

The MM-1 rebuild matters to me because of where it changes the gain structure. The stock preamplifier already has interacting Pentode and Triode stages followed by passive output attenuation. Saturation was available before the operation. I want to know which stage I’m asking to misbehave and what is driving it there.

Robert Wainscott, Black Box’s film mixer and engineer with a satellite and microwave communications background, rebuilt mine. His report places the added gain stage after the input transformer and before the first tube stage, alongside changes to both tube-amplifier stages. That gives the rebuild a specific location in the gain structure. An output-level match won’t make those operating conditions equivalent.

Once I’ve printed heavy saturation, output trim can’t restore the clean waveform or its original crest factor. Further processing remains possible. Recovering the unrecorded alternative is another matter. That distinction sets my commitment threshold. The undo department has no jurisdiction upstream of the converter.

I’ve kept some equipment as toys and some as tools. For a tool, recall and CPU cost change what I can afford to explore inside a real session. A small sonic preference isn’t much help if it makes every revision awkward. That’s why stability belongs in the sound decision.

I wanted two sections to feel different without replacing the band. “If I Could Drink Enough” moves from post-punk toward pop-punk with the same guitar and amp, bass, and drums. I changed the mix and sense of space. The puzzle was how far the production could turn while the song still held together.

If I Could Drink EnoughFirst Chorus · Post-Punk01:05–01:25

PlayStop

If I Could Drink EnoughSecond Chorus · Pop-Punk02:12–02:32

PlayStop

At that transition, I slipped in a skateboard sound. The next verse opens on a half-pipe. Choi heard a tape recorder, while I heard a skateboard immediately. It was a useful reminder that the sound in my head doesn’t arrive with its label attached.

The lyric says “background noise,” so I put noise behind it. The drinking reference in “shotgun this” gets the sound of a can opening. Choi liked those small bits. I draw on film and commercial sound work, but I don’t want the song to turn into a demonstration reel. The effect has to sit inside the mix and earn its moment.

If I Could Drink EnoughBackground Noise01:45–01:56

PlayStop

If I Could Drink EnoughShotgun This02:03–02:14

PlayStop

“Wait for You” begins with another little scene. The lyric has a bus stop, and I built an atmosphere around that idea. To my ears, it ended up closer to a bus terminal. I also synthesized the clicky effects because I liked their sound. A colleague loved them too. Not every sound needs to illustrate a noun. Some can suggest a place, and some can simply make the record more fun to hear.

On “Breaking Dawn” from Vagabond, I first sent over a guitar riff, then recorded the part all the way through instead of looping that recording. It became a habit I like. A phrase can return without the audio being an exact copy. The written idea repeats, but the next performance gets its own timing and touch. I’m not trying to make every bar announce a new invention. I want repetition to keep moving.

“Devil Eyes” makes a different use of time. I added a fermata before each chorus and held the second one slightly longer, around “I’m.” The arrangement is sparse, so a small delay in arriving can carry a lot of expression. That isn’t the same decision as moving notes a few ticks off the grid. I’m changing the phrase’s breathing room on purpose. Sometimes the part that needs producing is the wait.

Devil EyesThe Breath before the Chorus00:32–01:07

PlayStop

On “Lullaby,” I wanted an airy voice inside heavy distorted guitar, present without sitting on top of the record. Making it louder was available, but it would have weakened the relationship I wanted. Level, envelope, and space decide how those parts coexist. I want to preserve the contrast that made the record interesting, even when a favorite sound gets less room. My solo button has useful information. It doesn’t get the final vote.

LullabyInside the Sound02:00–02:30

PlayStop

I used a UAFX Golden Reverberator reverb pedal on “Devil Eyes” because it was fun. I’m a guitarist. Pedals are familiar territory, and producing a record gives me an excuse to play with hardware, synthesizers, and things I can put my hands on. I mix entirely in the computer. That doesn’t make production a hardware-free zone. Sometimes the detour is part of why I want to make the record.

Before the pedal, I sent the dry feed through a Louder Than Liftoff Chroma+ and its Implode Colour module, adding saturation and leaning hard into the compression. Implode uses 1176 Revision F-derived circuitry. The order matters. I was changing the envelope and harmonic content that excited the reverb, rather than compressing an already finished tail. I loved what came back.

The enjoyable part still had to fit the studio’s electrical limits. Universal Audio specifies a +12.2 dBu maximum input for Golden and supports line-level connections, with the source turned down when necessary. Its jacks are unbalanced TS. Those are usable conditions for a studio send, not permission to throw any line-output level at it. Driving the upstream analog circuit and overloading the pedal input are two different experiments.

Once the sound is in the session, I still decide where the reverb phrase begins and where it stops deserving space. Send automation can stop new excitation while preserving the existing decay. Return automation can interrupt the tail itself. One chooses the syllable that opens the space. The other can evict what remains. The fun part still has to fit the phrase.

The CLA-10A is a working tool. I want a reference I understand, not a new surprise every morning. A flatter magnitude response doesn’t by itself fix decay, directivity, or the next seat. When I consider changing the setup, I want a named problem and a way to tell whether it improved. Otherwise I may be changing my ruler halfway through the measurement.

I keep two listening positions because they do different jobs. My main Avantone CLA-10A pair connects straight to MOTU, with physical level and tone controls set using REW measurements. At the sofa, my Dante setup feeds the Dutch & Dutch 8c pair through its AES3 (AES/EBU) inputs. Room adjustments are guided by the manufacturer’s REW procedure. That’s my client-preview position. I want two useful perspectives, with the conditions of each kept clear.

I’ve also mixed a feature longer than two hours on a 7.1.4 setup assembled from different speaker pairs. Knowing that room was important, but familiarity didn’t cancel placement, level, timing, or tonal consistency. I needed to judge the space being built across the speakers. That’s a different task from deciding whether one pair is pleasant to listen to.

I want a vocal ride to change the thing I meant to change. On a self-keyed compressor, pre-detector automation, post-compressor trim, and reverb-feed automation do different jobs. Reducing detector input can trade some of the level reduction for less gain reduction. A downstream ride leaves that interaction alone. Before drawing the curve, I trace which part of the phrase I’m actually operating on.

The word Output doesn’t guarantee a clean level change. On a modeled compressor, that control may drive a nonlinear stage. If so, I need a separate trim for a level-only comparison, then I trace what follows it. Another compressor can still react to the changed level, and an external key needs its own routing check. I want the complete dependency chain before deciding that a ride changed only loudness. “Post” needs an address.

The reverb-feed point makes that choice audible in another dimension. A vocal ride and a reverb ride can move together or separate, depending on the routing. I’m listening for level, apparent distance, and the consonants that still need to reach the listener. Before writing automation, I want those dependencies settled. A phrase getting quieter shouldn’t accidentally walk out of the scene.

Film sound gives me a different kind of puzzle. Dialogue, Foley, ambience, and music can be assembled separately, but the audience meets one scene. I want the technical choices to agree about where we are and what deserves attention. The sound doesn’t need to announce how many pieces I had to fit together.

A shared room doesn’t require a shared source perspective. On THE DETACHMENT, I used ARVerb Room on ADR and other elements, including the lighter’s Foley in the warehouse. The enclosure needed to agree while the sources occupied different positions within it. My correspondence with developer Alexandre Ratchov helped me separate those controls. The trailer lets you hear some of that work in the scene, rather than as isolated effects.

For those per-source inserts, his starting point was one Room Size with similar Decay and Brightness. I could then vary Distance, Diffusion, Pan, and Mix. Crucially, Distance changes the reverberation pattern without attenuating the direct component. I still had to build the direct-to-reverberant relationship. A larger Distance value alone wasn’t going to put the lighter farther away.

I also keep ARVerb’s internal Pan distinct from post-plugin panning. The first participates in the room model, while the second moves its completed output. That distinction mattered for these inserts. On a fully wet shared return I’d be solving a different routing problem, so I wouldn’t transplant the same Mix decision by habit.

Lip sync doesn’t settle an ADR edit. A replacement can hit the mouth movements and still disturb the handoff between speakers. I want to protect the overlaps, interrupted endings, and delayed replies that give the exchange its rhythm. That’s the performance I want to preserve while repairing the words. Perfectly disinfected syllables can still leave a very dead scene.

A signal path has to keep earning its maintenance. I built the studio around Dante, then stopped needing that much infrastructure for tracking. I wasn’t recording orchestras here, and my immersive speakers weren’t on Dante. MOTU and Cranborne Audio C.A.S.T. covered the tracking job with less to maintain. The sofa’s separate Dante listening route still had a job, so it stayed.

N22 and N22H boxes put the C.A.S.T. analog runs where I need them. The Cat 8 cable installed during the build was already more than that application required. I’m comfortable with that excess staying in the wall. I don’t need to preserve an unnecessary workflow just to justify its cable.

My recording environment is very quiet. I judge noise at the actual source impedance, gain, and bandwidth I’m using. A weighted equivalent-input-noise figure and an unweighted output measurement don’t answer the same question. Unwanted hum and hiss are separate from the distortion I deliberately choose. Once noise is inconsequential to the job, I’d rather spend the attention on the performance than optimize a comparison whose conditions have quietly changed.

My Linux desktop gets the same treatment as my Cockos REAPER setup. I start with defaults I like, then change the parts that interrupt me repeatedly.

Beside the clock, I put a countdown to local midnight. I can see how much of today is left without doing the subtraction. The widget updates on the system clock’s minute events and recalculates after sleep or a clock change. It doesn’t launch a shell command every second to tell me I’m running out of time. That would be an unnecessarily elaborate way to waste some of it.

English interface text uses IBM Plex Sans, with IBM Plex Sans KR for Korean. I set the sans-serif fallbacks without replacing every explicit font or the terminal’s monospace face. Language switching, shortcuts, and the status bar get the same attention. These are things I use repeatedly. A small irritation is worth fixing when the interface keeps handing it back to me.

I want less travel between deciding and doing. In Cockos REAPER, my useful shortcuts live around my left hand while my right hand works the mouse. Watching esports players operate at speed gave me the idea. The target is the small interruption between hearing a problem and acting on it. I’m not trying to defeat a vocal take through superior clicks per minute.

My MIDI editing uses ESDF rather than WASD. The one-key shift suits my hand. E and D handle pitch, S and F handle time, so the same directional shape survives the move from a game to a piano roll. I want the physical layout to stay obvious while the operation changes.

Fast mistakes are still mistakes. Modifiers change the operation, though Shift timing nudges use pixels and therefore depend on zoom. Related edits get related gestures, with a few exceptions worth knowing. I check whether I’m editing items or MIDI notes, which window is active, and what’s selected before trusting muscle memory with the specimen.

My archive action sets normal track FX offline on selected tracks. Repeating it is harmless. Parent routing also has explicit off and on actions, so a mixed selection converges on the state I asked for. I leave bypass as a toggle for comparisons. Deterministic state changes belong in housekeeping. Auditioning has a different job.

Scope is part of the shortcut. My main split follows the mouse without changing the selection and ignores grouping. Splits at the edit cursor or across a selected range are separate actions. For the adjacent item boundaries I make with trim-behind replacement, I use a crossfade action that moves the edges to create an overlap. Merely asking an existing overlap to crossfade would leave the real problem untouched.

My humanizer offsets selected note starts by −5 to +10 PPQ ticks and velocity by −10 to +10, preserving pitch and duration. I audition that late-biased window in the phrase rather than treating it as a groove preset. Its millisecond displacement depends on tempo and PPQ resolution, and successive runs accumulate.

My stereo and immersive sessions need different engine settings, so I gave the switching an explicit launcher. Cockos REAPER’s anticipative-processing and offline-render notification settings are global. Opening a project doesn’t restore the profile by itself. The launcher checks the active project and stops if playback or another open session makes the change ambiguous. A faster switch is only useful if it lands in a known state.

I use artificial intelligence to help build this infrastructure. Preparing sessions, checking file references, running exports, and organizing the resulting files are labor I’m happy to hand to a supervised system. Writing, playing, producing, and mixing are the reasons I came into the room. The distinction is what I want to spend my time doing. My laboratory assistant can count the test tubes. I’m keeping the experiments.

Purpose can change the music itself. All eight cues in THE DETACHMENT have different film and album versions. The film timings serve the picture. For the soundtrack album, I adjusted them to follow the music’s own flow. A cue can arrive at the right moment in a scene and still need a different shape when someone plays the record.

THE DETACHMENTPicture, Then Record00:55–01:25

PlayStop

I want mastering to bring another pair of ears to a mix whose balances already work. My stereo mix bus is unprocessed. I build the balances and movement inside the mix, then send it to a mastering engineer I trust. That’s how I divide the work.

John Hanes makes a useful distinction in his delivery advice. An unprocessed lead, a lead with compression and equalization but no reverb or delay, and separate effect returns preserve different choices. What interests me is the separation of tone from space. The vocal’s character can stay intact while its apparent distance changes. “Dry” alone doesn’t tell the next person which decisions are printed. I want the intended combination identifiable before experimenting with the alternatives.

In recent Apple Vision Pro work with LATENCY, I used a handoff from Cockos REAPER through DaVinci Resolve Studio built around separate export copies. The original mix sessions stay intact. Each handoff has an explicit source, range, sample rate, channel layout, and expected output set. The repeated steps can be automated because those decisions have already been made. A missing file or an unconfirmed setting should stop the process, not invite a plausible substitute.

I check the rendered file, not just the export dialog. Sample counts and frame timing can disagree at a boundary. Loudness and true peak need a fresh readback, and spatial metadata has to survive the handoff. A fixed binaural reference lets collaborators discuss the music on ordinary headphones. It doesn’t verify head tracking or sync with the final picture on the intended system. Hashes protect file identity. Listening decides whether the scene works. Neither can do the other’s job.

Theatrical and near-field home playback also ask different things of a mix. Dialogue, effects, and dynamics may need another judgment. Changing the file format can’t make that judgment for me. I want the delivery plan to leave time for listening under the intended conditions, so the scene still works after it leaves this room.

I want an efficient system so I can be inefficient with the music. There. The entire lab has been caught funding creative detours. A left hand that knows the keys and a session that opens predictably buy me room to try the part that might go nowhere. I’m removing the delay in operating the equipment. Wondering what to do with it can take as long as the idea deserves.

Doing many jobs myself taught me how much another person can change a record. I play guitar, but another guitarist can hear a part I’d never invent. Singers, songwriters, instrumentalists, and mastering engineers bring different instincts into the room. I like having a direction and then discovering a better one together.

Music also keeps escaping the studio into images, clothes, language, and the places people encounter it. That wider world helps me decide what a sound is for. The engineering can be meticulous without making the listener study it. I’d rather the record make sense than require an explanation of every clever edit.

For pleasure, I can enjoy music on wired earbuds or a phone speaker. I love listening with other people too, for much the same reason I like a full movie theater. The working reference is important when I’m making decisions. It doesn’t have to accompany me to every good song.

Making music is fun. That’s the whole motive, under all the measuring and rewiring. I like the absurd amount of care a tiny sound can invite. I like the part I can’t settle without picking up the guitar or trying one more take. The experiment has produced a person who wants to do another experiment. Funding continues.