INTRO
0:00-0:30
ON-SCREEN: Title card "Advanced Diagnostic Scenarios", then the healthy seven-tile baseline dashboard from the D track, which suddenly splits into two overlapping ghost dashboards pulling in opposite directions.
NARRATOR: Everything in the diagnostics track assumed one fault at a time. One fault, one shape, and you learned to name the shape. This module is for the calls where that skill fails on purpose: two faults on one system, each bending the numbers the other way. The combined picture matches nothing on your card, and the tech who pattern matches anyway replaces the wrong part. Here is how a master breaks the tie.
MAIN
[0:30-1:30] Two faults, one set of gauges
ON-SCREEN: M40-compound-fault-web.svg. Animate the two arrows on the head pressure tile: dirty condenser pushing UP, undercharge pulling DOWN, needle settling on 390 with the label "normal, and lying".
NARRATOR: Faults are vectors. Low charge pulls head pressure down. A dirty condenser pushes it up. Put both on one system and head reads 390 at 100 degrees ambient, fifteen over, dead center of the healthy band, and completely false. The masking is not bad luck; real fault pairs push readings in opposite directions all the time. So the master question is never just "which fault matches this picture." It is "does every reading agree." Seven readings telling one story is a single fault, and you already know those. One reading contradicting the story is the second fault waving at you. And the field data says expect it: NIST found over sixty percent of fifty five thousand units already running a wrong charge. Whatever broke today is probably fault number two.
[1:30-2:30] The four cases
ON-SCREEN: M40-case-study-grid.svg. Highlight one row at a time as each case is named.
NARRATOR: The article works four full cases, one for each failure pattern from D22. The head pressure that lied: subcooling 4 says empty, head 390 says full, and the discriminator is cleaning the condenser and re-reading, because both statements cannot be true of charge alone. The capacitor that kept dying: the cap really is ten percent low, but caps do not die of natural causes in three weeks, and the static map finds the crushed return cooking the motor. The charge that was never low: superheat 6 was an airflow verdict, the first tech added refrigerant anyway, and the coil iced. And the compressor that was only hot: OL from common to both terminals with S to R intact is a tripped internal overload, the sum check proves the windings are whole, and the killer was a weak condenser fan capacitor. Four cases, one skeleton: take the full set, name the contradiction out loud, change exactly one variable, restabilize, and read everything again.
[2:30-3:20] Ghosts
ON-SCREEN: M40-intermittent-fault-traps.svg. Step through the five trap rows.
NARRATOR: The compound fault hides in plain sight. The intermittent fault leaves before you arrive. Five usual ghosts: thermal faults that only open at temperature, voltage sags that only appear under compressor inrush, pressure switch flutter, defrost-only faults, and the humidity faults that sleep all winter. You cannot discriminate readings you cannot capture, so the game becomes capture: a recording meter parked on the suspect circuit, a data logger, and on communicating systems the fault history with timestamps, photographed before anyone clears it. And the cheapest logger on the truck is the customer. Bound WHEN before any tool comes out. A ghost bounded in time is half caught.
[3:20-4:15] The autopsy and the time-box
ON-SCREEN: M40-callback-autopsy-flow.svg, then crossfade to M40-escalation-decision-tree.svg at the time-box beat.
NARRATOR: A callback is data, not shame. The system voted against somebody's conclusion, and the autopsy finds the broken assumption on purpose: reconstruct the first visit from the invoice and photos, build the two column list, measured versus assumed, and hunt the assumed column, because that is where broken assumptions live. Came back iced, the airflow assumption broke. Same part dead again, the root cause assumption broke. And when you are the one stuck: time-box it. Forty five minutes without a falsifiable hypothesis, restart the funnel. Ninety minutes, phone a friend, with the packet already built: full readings, what is ruled out and by which number, and the contradiction in one sentence. A master does not get stuck less. A master notices sooner.
OUTRO
4:15-4:30
ON-SCREEN: The four-case grid stacked above the dashboard, end card with module title.
NARRATOR: Read the four cases until you can run each discrimination from memory. Then the demo video hands you a staged compound fault, live. The numbers still make the call. There are just two answers now.