Four detectors are not four science channels

The source defines a channel precisely: each BPW34 shall have an independent bias, return path, amplifier, comparator and microcontroller input, with a shaping stage and an injection point, and under 1 per cent response in adjacent channels. I had been reading four devices on a sheet as four channels because four is the number I intended.

Going back and tracing it gave a different answer. The four BPW34 devices sit on a shared net rather than four separate bias and return paths. There are three op-amp sections and three comparators for four intended channels, and one of those comparators drives a status LED rather than an event input. The per-channel feedback resistor, feedback capacitor, shaping network and injection point are absent — the sheet's entire discrete complement is four resistors and three capacitors.

I wrote the conclusion into the case study as it stands rather than softening it: Rev A shows four detector devices but does not yet demonstrate four complete independent science paths. Completing and documenting those paths became the first Rev B item.

The acceptance criteria are what make independence testable: injecting a calibrated pulse into one channel at a time and measuring all four outputs, then disconnecting or shorting one channel and confirming the rest keep working. Neither test can be run on Rev A, because the injection points are not there.

Why it matters is not tidiness. If the detectors share a bias or a return, a disturbance on that node appears on every channel at once — which is indistinguishable from a real multi-channel event. So coincidence, the technique that is supposed to identify penetrating particles, is the technique most damaged by the sharing.

Independence is the precondition for both things four channels are for: comparing differently shielded responses, and trusting a coincidence.