Case Study 2 — The Hum-Hunt: Aisha Debugs Her Chain with the Half-Split Method

This picks up where the chapter's opening left off — Aisha Brooks, 11 p.m., an episode due at noon, a vocal recorded at whisper level under three different species of noise. A fictional narrative, but the method is the real one: the systematic fault-finding discipline that telephone-line engineers used to debug circuits that ran for miles, scaled down to a desk in a Chicago apartment. Read it with your own chain map from the Project Checkpoint beside you.


Triage first: the deadline is not the debug

Aisha does the professional thing first, which is to admit that 11 p.m. before a deadline is not when you re-wire a studio.

Episode 41 ships on the recording she has. She boosts her whisper-level track and accepts the hiss that comes up with it, carves what she can, level-matches herself against her guest by ear, and exports at 2 a.m. It is, by her own description in the show's group chat, "listenable, barely." She's not wrong. She's also not done.

Saturday morning, coffee, and the flipped-over piece of junk mail with her chain drawn on it. She rewrites it properly in a notebook, two columns, the way you did in this chapter's Project Checkpoint:

 CAPTURE:  voice → dynamic broadcast mic → XLR (balanced)
           → 2-input USB interface (gain knob ?!)
           → unpowered USB hub ← also: external SSD, keyboard dongle
           → laptop (charger: two-prong wall wart)
           → Logic

 MONITOR:  Logic → interface D/A → headphone out → headphones
           and → interface line outs → RCA cables (unbalanced!)
           → powered desk speakers → plugged into WALL OUTLET
                                      across the room
           (laptop + interface: into the DESK power strip)

Then she writes the symptom list, because she read somewhere — fine, in a forum thread at 1 a.m. — that you name the animal before you hunt it:

  1. Hiss — broadband static under her voice, appeared after she boosted the quiet track. Suspect: gain structure.
  2. Hum — smooth, low, steady. Present in her speakers even when nothing plays. Suspect: grounding.
  3. Buzz — edgy, harmonic, intermittent. Comes and goes by some schedule she hasn't decoded. Suspect: power.
  4. The whisper-level recording itself — peaks at -38 dBFS. Suspect: operator. (She writes this with a sigh.)

Four problems. Her old method — wiggle everything, re-plug everything, buy something — treats them as one big haunted-studio problem. The chapter's method treats them as four addresses to find. She tapes a rule to the desk lamp, borrowed from the practice section: one change at a time, and write down what each test proves.

Hunt 1 — The hum, by half-split

The hum is the most annoying, so it goes first.

Test 1: split the chain at the interface. The classic midpoint. She plugs headphones straight into the interface's headphone jack, speakers up but audio cable route bypassed. Plays a quiet section of the episode. Listens.

Silence. Clean, black silence under the audio.

She writes: Headphones at interface = clean. Everything upstream of the interface's outputs — mic, hub, laptop, Logic, the interface itself — is innocent of the hum. The hum lives in the speaker leg. One test, and half the chain — the half with the most components — is eliminated. This is the whole half-split idea: don't check stages one by one from the start; cut the search space in half with each test. A ten-stage chain falls in three or four cuts.

Test 2: split the speaker leg. The remaining suspects: the RCA cables, the powered speakers, and whatever electrical relationship connects them. She pulls the RCA cables out of the speakers entirely. Speakers powered, volume up, connected to nothing.

Silence from the speakers.

Speakers alone = clean. They don't hum by themselves. The hum needs the connection to exist. This is the spooky result the chapter's Productive Struggle box warned her about: two innocent devices, and the connection between them is guilty. Which means she's not hunting a broken component. She's hunting a shape.

Test 3: collapse the loop. She looks at her own map and sees it drawn right there: laptop and interface grounded through the desk power strip; speakers grounded through the wall outlet across the room; an audio cable bridging the two ground territories. A circle of conductor the size of her office — a ground loop, the textbook case, almost insultingly classic.

She gets the extension she uses for the holiday lights, brings the speakers' power around to the desk strip, re-plugs the RCAs, and turns everything on.

The hum is gone. Not quieter. Gone.

She writes: Hum = ground loop, speakers on far outlet. Fix: one strip, one ground path. Cost: $0. And then, because the chapter told her the loop will try to come back every time she rearranges the room, she adds the belt-and-suspenders item to her shopping list: a pair of balanced TRS cables to replace the RCAs — her interface's line outs and her speakers' inputs both support balanced connection, and have all along. The hum she lived with for two years required two unbalanced cables, two outlets, and nobody ever looking at the map. It got all three.

Hunt 2 — The buzz that keeps office hours

The buzz is intermittent, and intermittent faults are where superstition breeds. So she stops trusting memory and starts logging: every time the buzz appears or disappears for two days, she writes the time and what changed.

The log cracks it embarrassingly fast. Buzz present: Saturday 2:14 p.m. — laptop charging. Buzz absent: 3:40 p.m. — charger unplugged, on battery. Present again: 7:02 p.m. — charger back in. The buzz keeps the charger's hours.

Test: the battery test from the chapter. Charger out, sixty seconds of recording. Clean. Charger in, same recording. Buzz — edgy, harmonically rich, the dimmer-switch species, riding in through the laptop's power and out through every audio connection the laptop touches.

The charger is the two-prong wall-wart that came with the laptop — a switching supply with no ground of its own, free to spray its noise into everything downstream. Her fixes, in the order the chapter ranks them: same-strip power (already done — it helps; the buzz is quieter but not gone), then the supply itself. The laptop maker sells a grounded three-prong supply; a colleague has one she can test first. With the grounded supply on the desk strip, the buzz drops below anything her boosted silence print can find.

She writes: Buzz = charger noise. Fix: grounded supply, same strip. Lesson: log intermittent faults against time; they're usually keeping somebody's schedule. Recording on battery power remains her emergency fallback — and a tidy diagnostic she now owns forever.

Hunt 3 — The whine she didn't know she had

While the silence print is boosted to +30 dB anyway, she hears a third resident: a faint pitched whine that changes when she scrolls. She's never noticed it under normal audio — but episode 41's quiet passages, boosted 24 dB, would have carried it to her listeners.

Her map shows the suspect circled in her own handwriting: the unpowered USB hub, bought during the desk move to tidy the cables, now feeding the interface and an external SSD and a wireless keyboard dongle from one overworked port. The interface drinks its power from that shared straw — "bus-powered" was on the box and she'd never once thought about what the bus was.

Test: one change. Interface out of the hub, directly into the laptop's other port. SSD and dongle stay on the hub. Scroll furiously. Record. Boost.

Gone.

She writes: Whine = dirty bus power via shared unpowered hub. Fix: interface gets its own port, always. Hub keeps the peripherals. She remembers the chapter's footnote that hubs also cause the dropout-and-click problems from Chapter 2's buffer discussion, and realizes she'd had one of those in episode 39 and blamed Logic. Apologies are mentally issued.

Hunt 4 — The whisper, or: the operator confesses

Which leaves the original crime: a vocal recorded at -38 dBFS by a person who was wearing headphones the whole time and heard herself just fine.

She reconstructs Saturday-of-the-desk-move like a detective interviewing herself. Everything unplugged, everything re-plugged in a hurry. The interface's input gain knob — bumped, or never reset, down near its minimum. Then, while testing: "hm, quiet," and she reached for the knob she always reaches for — the headphone volume — and turned it until her voice sounded strong. Recorded the episode. The monitoring chain said loud and clear. The recording chain wrote down a whisper. Two chains, one box, and she'd been consulting the wrong one — the meter in Logic had been telling the truth in her peripheral vision the entire time, a flat line with nervous twitches that she never once looked at.

Her fix is the chapter's ritual, performed once, properly: speak at full podcast energy, set the input gain until Logic's meter peaks around -15 dBFS, and then — her own addition, now passed on to you — a thin strip of washi tape on the interface with a dot marking the knob's home position, and a sticky note on the monitor's bezel:

"The headphone knob is not the record knob."

The map on the wall

Sunday afternoon she re-records the silence print — the ten seconds of room at proper gain — boosts it, and listens. Hiss: a faint, even, healthy floor, far below anything that matters. Hum: gone. Buzz: gone. Whine: gone. She prints the final chain map, knob positions inked on it, fixes annotated in red, and tapes it above the desk where the $400 microphone would have hung its halo.

Total cost of the repair: one extension cord she already owned, two balanced cables, a borrowed power supply, and a Saturday. The mic stays. The mic was never the problem.

Episode 42 goes out the following Tuesday. Within a day, a listener emails: "New microphone? Sounds way better."

She reads it twice, laughs once, and types the reply this whole book is about:

"Same mic. New knowledge."


The protocol, distilled

What Aisha actually executed, reusable on any chain, any noise, any deadline:

  1. Ship, then fix. Triage the deadline with the least-bad late fix; schedule the real hunt for fresh ears and daylight.
  2. Draw the chain before touching it. Capture path and monitor path, every cable, every gain stage circled. You cannot half-split a chain you haven't drawn.
  3. Name each animal. Hiss, hum, buzz, whine, crackle — separate symptoms are separate hunts. Four problems treated as one is how studios stay haunted.
  4. Halve, test, conclude, halve again. Each test must eliminate territory. Headphones-at-the-interface and laptop-on-battery are the two highest-value first cuts in any small studio.
  5. One change at a time, in writing. Log intermittent faults against the clock; they keep schedules.
  6. Fix at the source, then re-verify with a silence print. And update the map — the map is the studio's memory, and the next desk move is coming.

The half-split method isn't audio knowledge, strictly speaking. It's debugging — the same discipline that finds faults in code, in telephone lines, in lab equipment. The chain gave it somewhere to stand. That's the quiet gift of this chapter: once your studio is a sequence of stages instead of a fog of gear, every problem you will ever have lives at an address. Some Saturday soon, you'll find one in three tests and feel like a wizard.

You won't be. You'll be someone with a map.