A thread about a case packer that counted a shadow as a box and then could not see an actual box at all, and how the sensor type was argued back to the physics. Names trimmed.
From: P. Okafor, Production — To: R. Almeida, Design; cc quality, maintenance
Subject: Case packer L7 — counter spike 1,900 pcs on a 240-pcs run, and the reject flap hates the film
Ricardo, the packer logged 1,900 counts on a run that packed 240 cases, so the counter is seeing ghosts, and the ghost has a name, the sensor on the infeed sees the box, then it sees the metal plate behind the gap, and sometimes it sees the shadow of the operator. In the same hour the flap that rejects the shrink film says no box when there is clearly a clear-filmed box in the sensing zone, so we have a sensor that over-counts metal and a sensor that under-counts transparent plastic, both on one line, and the maintenance crew is being blamed for both. I am sending the current models off the drawing: inductive on the case counter, twelve millimetre flush mount, and a diffuse photoelectric on the reject, both bought off a distributor catalogue by a man who left in 2021. The line runs 60 cases a minute, three shifts, and I need a decision that does not start with the word firmware.
From: R. Almeida, Design — To: P. Okafor, A. Berg, Sensor Supplier; cc quality
Subject: RE: L7 — both sensors are the right type for the wrong job
Peter, agreed, and the fix starts with admitting the physics, not changing a parameter. The case counter is counting the metal guide plate because an inductive sensor answers the question is there ferrous metal in range, and on the infeed the box itself is cardboard, which is invisible to it, so what it actually counts is the plate passing, which is another way of saying the counter never saw a box in its life. The reject station is counting on a diffuse photoelectric to see clear film, and a diffuse emitter needs the target to scatter light back, which a shiny web of film does about as well as a mirror, so the answer is no box, every time the film is clean. Neither unit is broken, both are mis-selected, and the substitute either of them deserves is not a firmware release but a different sensing principle, so before I spec anything give me the sensing zone drawings, the gap behind the infeed plate, and the ambient light at the reject, and I will stop defending a catalogue purchase with physics.
From: A. Berg, Sensor Supplier — To: R. Almeida, P. Okafor; cc quality
Subject: RE: L7 — two principles, one counter, and the film is a mirror
Ricardo, I have worked your drawings before the site visit and here is the honest selection, principle first, part number second. The case counter: do not reach for another inductive, because the box is cardboard and no inductive will ever see it, reach for a through-beam photoelectric, emitter and receiver across the infeed, sixty millimetre gap, and the beam is blocked by any case, cardboard, film or empty but wrapped, and a metal guide plate behind the gap becomes irrelevant because the beam only answers is the line of sight clear, which is the actual question the counter was asking. If a through-beam feels like two parts to bolt on, a retro-reflective photoelectric with a polarising filter will do the same job on a coated case with one bracket, but it will not reliably see a clear film web. The reject: your diffuse unit is the wrong tool and no amount of sensitivity will fix it, because clear film returns almost no light to the source, so use a background-suppression photoelectric set to a window of 40 to 60 millimetres, which measures the angle of the returned spot, not the total light, and will reliably fire on a film-wrapped case while ignoring the bright conveyor frame beyond the set point. Any daylight at the reject? If sunlight reaches the sensing zone, say so, because a bare BGS unit at ten thousand lux will pick up the sun as a background and jump, and we will screen or re-angle the mount before we ship.
From: R. Almeida, Design — To: A. Berg, P. Okafor; cc quality
Subject: RE: L7 — measurements and one disagreement about the counter
Site numbers: infeed gap is 55 millimetres, the reject zone sits under the machine deck with the label-light off during film wrap, so daylight is negligible, ambient at the reject was 80 lux on my meter during the midday run. Acceptance: your through-beam for the counter and the BGS for the reject, both 24 VDC PNP with a 200 hertz or better response, and I will not have relay logic in this panel. The disagreement, and it is a small one, is the counter. Production would rather keep a single-point mount and I worry that a through-beam across the infeed catches the operator elbow, so I am asking for the numbers: what is the beam diameter and what is the MTTF of a misaligned emitter in a machine that vibrates at 60 cases a minute, because if the answer is realign it weekly, behind the belt too, then give me the retro-reflective with the polariser and the tolerance on the label web instead, and we will decide with a two-hour trial on Monday before we commit the bracket drillings.
From: P. Okafor, Production — To: R. Almeida, A. Berg; cc quality
Subject: RE: Monday trial — the counter counted 240, the reject saw the film, but two faults first
Ricardo, the trial says the physics wins and the brackets are good, but not without two findings worth a line each. Through-beam on the counter: 240 cases, 240 counts, clean, and the operator elbow did not ghost it because the beam sits inside the guard rail line, so the alignment worry was a paper one. The BGS on the reject fired on every film-wrapped case in the 55-case trial and ignored the conveyor frame, set to the 45-millimetre window, no daylight issue at 80 lux, so that one is done and I will sign the drawing. Fault one: the first BGS unit we mounted ghost-fired twice when the counter beam crossed it during setup, both sensors on the same bracket, and we found the cause is emitter crosstalk, the counter’s beam was rich enough in infrared to leak into the BGS receiver when the two heads sat fifteen centimetres apart, so the bracket now keeps forty centimetres and the ghosting is gone. Fault two: an old spare inductive we kept for a speed pickup is wired PNP into a sink input and reads dark-high, so maintenance burned an hour before we found the polarity mismatch, which is a cabinet labelling failure, not a sensor failure, and I want the panel marked before Friday. Next: the quality run tomorrow on a full shift, then close it out.
From: K. Yamada, Quality — To: all; cc plant manager
Subject: RE: L7 quality run — 1,020 cases, zero miscounts, and one label gap we see now
Full shift run with the new sensing, and the log is clean: 1,020 cases produced, 1,020 on the counter, zero reject misses and only two legitimately caught damaged wraps, and the packer no longer argues with the night shift about phantom counts. One finding that is not a sensor fault but a sensor dividend: with a real count the line controller now flags when the reject has not fired for a hundred cases, which on the old ghosting counter was a silent signal, so we have gained a diagnostic we did not order. The cabinet is marked PNP left, sink right, and the crosstalk bracket spacing is on the drawing. I call the counter and the reject both closed against the trial data, and I would like the lesson recorded in the design rule: select the sensing principle against the target’s material and the ambient, then pick the part number, never the reverse.
From: R. Almeida, Design — To: all — Subject: RE: L7 close-out, the table, and the rule that survives it
Closing L7 with the comparison that would have saved the 2021 purchase, on one table for the panel folder:
| Target problem | Wrong 2021 choice | Right principle | Why |
|---|---|---|---|
| Count cardboard cases past a metal plate | Inductive, 12 mm flush | Through-beam photoelectric | Beam answers line-of-sight, ignores plate and shadow |
| Detect clear film-wrapped case at a reject | Diffuse photoelectric | Background-suppression photoelectric | Measures spot angle, not returned light; film is a mirror to diffuse |
| Two heads on one machine frame | Any spacing | Bracket keep-out, 400 mm min | Infrared leakage between heads causes crosstalk ghosts |
| Panel wiring clarity | Unmarked | PNP left, sink right, labelled | Polarity mismatch burned an hour |
Reference data — infeed gap 55 mm, through-beam across it; reject window 45 mm BGS, ambient 80 lux; 24 VDC PNP, 200 Hz response; 60 cases/min, 3 shifts; trial 240/240 counts; full shift 1,020 cases, 0 miscounts; crosstalk cleared at 400 mm spacing.
Glossary — inductive sensor: detects ferrous metal in a sensing field, blind to cardboard and plastic; diffuse photoelectric: emits and receives light at one head, relying on target scatter, poor on mirror and clear surfaces; through-beam: separate emitter and receiver, beam blocked by any opaque target, most reliable and longest range; retro-reflective: emitter and receiver at one head with a reflector, blocked when the beam to the reflector is broken; background suppression: measures the angle of the returned light spot to ignore surfaces beyond a set window, ideal for low-contrast targets; crosstalk: stray light from one sensor head leaking into a neighbouring receiver; PNP: a positive-switching output that sources load current, standard on European controls.
Field note — the sensor that saw a box that was not there, and the one that refused the label
Both faults on this packer were the same mistake made twice, choosing the sensing principle by habit instead of by the target. An inductive sensor on a cardboard counter never had a chance, it was counting the metal plate that happened to be in its field, and a diffuse head on a clear-film reject could only have seen the film if the film were sandpaper. The salvage is cheap and the cost of the wrong units was one mis-catalogued purchase and a month of phantom counts, and the rule that lasts is short: identify what the sensor must truly detect, measure the gap and the light and the surface, and only then open the catalogue. The firmware was never going to help, because the part number was the only bug, and when the physics is the specification, the machine stops arguing with the night shift.
From: A. Berg, Sensor Supplier — To: R. Almeida; cc P. Okafor
Subject: RE: one bench note with the spare units
Bench note with the spares we left you: the BGS window is set at 45 millimetres and the teach-in is locked, so a future operator should adjust the bracket before the sensor, and the through-beam is factory aligned, so if the count ghosts again, measure the gap before you touch the screws, because a vibration-loosened mount drifts the beam before it fails the count. Also worth knowing, the BGS ignores a glossy floor reflection when the bracket is tilted past six degrees, which our mount already has, so when somebody rebuilds the frame, keep the tilt. The invoice covers both units plus the keep-out bracket; there is no firmware or subscription, and the only magic on this order is that the right principle was picked on the first week instead of the third month.
From: R. Almeida, Design — To: A. Berg, P. Okafor
Subject: RE: received, and the drawing note for the next machine
Received and filed. For the next machine I have added to the standard: the sensing-principle table above goes on every sensor detail drawing, and the supplier’s keep-out and tilt notes go in with the brackets so the person who maintains the line inherits the alignment intent, not just the part numbers. L7 is closed.
