Lesson 18 mapped who owns which decision. This lesson teaches where the information for those decisions comes from, and it lands one idea above all others: a decision is only as good as the read it's made from. Don't let this turn into a parts tour. The center of the hour is a photoeye you draw live, square versus angled, until the room sees the registration point move on its own. If they leave calling identification "just a scanner," or assuming every barcode reads, the lesson didn't land.
| Segment | Min | What happens |
|---|---|---|
| A bad read is confident | 6 | Open on the hook, cold. A decision is only as good as the read it's made from, and the system doesn't hesitate on a bad read, it commits. Ask the room: if the scan is wrong, what does the sorter do? Steer them to the answer that the wrong lane happens fast and correctly. That's the whole reason this layer matters. |
| Photoeyes, drawn live | 14 | Key teaching moment, below. Split the photoeye's two jobs on the board: presence versus registration. Then draw square versus angled and make the room predict what happens to the trigger point as carton width changes. Do not tell them. Draw the beam breaking at two widths and let them watch it move. This is the center of the hour. Do not cut it. |
| Encoders and tracking | 10 | How a carton scanned upstream still diverts at the right instant. The encoder measures belt travel; the PLC holds the carton as a position and fires when the position lines up. Land the CNC/robot parallel: same coordinate-tracking idea. Then tie it back, the encoder can only trust a read that repeats. |
| The identification family and no-reads | 14 | Scanner, scan tunnel, camera, inline scale, dimensioner, each chosen by what it reads. Then the load-bearing point: no system reads everything, so the no-read rate is a design input. A failed read needs somewhere to go or it stops the line. Name the hospital lane as that destination; hold the sizing for Lesson 23. |
| Riverside sensing plan | 10 | Run the single-scanner-versus-scan-tunnel debate below on Riverside's real product mix and two-pick-zone origin. Make them defend the call from label orientation, not cost. Collect the three-device plan and the named no-read path. |
| Forest and close | 6 | Read the forest aloud. Every layer left in Part V acts on what this one reports. Close on the one line the whole lesson hangs on: a decision is only as good as the read it's made from. |
| Total | 60 | Baseline session. Expand with the stretch options below if you have 90 minutes. |
Put a photoeye mounting decision on the board: square to travel versus angled across the belt. Before you draw anything, ask the room to predict what each does to the registration point as carton width changes. Take every answer. Most students won't see the drift until it's drawn, and that's the point.
Now draw it. Square first: the beam is a single line across the belt, and a narrow box and a wide box both break it at that same line. Then angled: the beam crosses on a diagonal, so draw the leading edge of a narrow box breaking it at one spot, then a wide box breaking it at a different spot. The moment the trigger point visibly moves is the moment the rule stops being a slogan. Say it plainly, an angled registration eye trades a repeatable trigger for a variable one.
Then tie it to the encoder while it's fresh: once the read repeats, the encoder can trust it and track the carton to the divert. Once it drifts, no amount of tracking fixes a read that started in the wrong place. The encoder commits to the bad read precisely. That link, repeatable read then trustworthy tracking, is what makes the two halves of the lesson one idea.
Three failure patterns, each driven back to the same line: a decision is only as good as the read it's made from.
Run the debate on Riverside's real conditions, not a generic one. The carrier is tied to the barcode, so the sort point has to read it, and the cartons come off two pick zones onto throw-on lines. That origin is the whole question: how confident are you which way the label faces when it arrives? Make students defend the single-versus-tunnel call from label orientation, not from cost alone. A student who reaches for the cheaper single head without asking the orientation question has skipped the reasoning that matters.
Instructor-only note, don't preview: the scan-point location is deliberately left open in the student deliverable because Ray's WMS response-time number isn't confirmed yet. That resolves in Lesson 22, when the confirmed number arrives and the scan-to-divert distance gets locked. If a student tries to pin the scan point now, hold the line, they have the devices, not the distance. Don't tell them what the confirmed number turns out to be.
Square, the beam is a fixed line across the belt, and a leading edge trips it at the same instant no matter how wide the carton is or where it rides. Angle the eye and the beam crosses on a diagonal, so where a leading edge first breaks it depends on the carton's width and its lateral position. The wider the carton, the further its leading edge reaches across the belt, so its break point shifts furthest from a narrow carton's, and a wide box riding to one side shifts furthest of all. The registration instant moves box to box, the tracking clock starts from the wrong place, and the divert fires early or late by enough to miss the lane. The widest cartons get hit hardest because the trigger-point error grows with width. Source rule, verbatim: "Never use an angled photoeye for product registration. The angle of the eye combined with the position of the package across the face of the conveyor will produce inconsistent trigger points, and the system will behave as though product is in a different position than it actually is." Full credit names the fix out loud: mount square to travel.
No identification system reads everything. Damage, print quality, orientation, glare, and a label that lands face-down all produce no-reads regardless of how good the customer's labels are on an average day. It's a rate, not a possibility, so it's a design input on every real system. The exception destination isn't optional insurance, it's the release valve for the reads that fail. Leave it out and the first no-read is a carton stopped at the decision point with no routing answer and nowhere to go, so it can't divert and the line backs up behind it. Skipping it doesn't save money, it converts a routine event into a line-down event during a peak wave. The destination is the hospital lane; how big it has to be and how the routing around it works is Lesson 23, so don't let the debate drift into sizing. Full credit is the argument that a no-read rate is a design input and that one unhandled no-read stops the line.
The sort-point sensing plan the students write is a real piece of the Riverside controls architecture package, and it goes into the same running project note they've kept since Lesson 1. Have them date the entry and file the device list, the single-versus-tunnel call with its justification, and the named no-read path where they can find it again.
Don't explain where it leads. The payoff is theirs to find later, and it lands hardest for the students who kept their notes without being told why. Your job is to make filing this feel like ordinary professional discipline, one more layer added to a document that's quietly becoming the controls package they'll assemble at the end.