Why Coin Comparators Jam Continually: Optical Token Rejection Loop Calibration
The jam is almost never mechanical β it is a calibration failure inside the optical token rejection loop, and anyone who keeps swapping solenoids instead of fixing the feedback threshold is burning money. Coin comparators do not "wear out" in the way operators assume. They drift. The infrared emitter and phototransistor pair that defines the accept window slowly falls out of alignment with the reference signature stored in the comparator's logic, and once that delta crosses the tolerance band, every legitimate token gets rejected, re-routed, and cycled back into the entry throat until the mechanical gate physically stalls. That is your continual jam. π
The Optical Rejection Loop Is a Closed System, Not a Sensor
Most floor techs treat the comparator as a passive eye. Wrong. The optical head, the discrimination circuit, and the divert gate form a closed feedback loop, and the loop has a settling time. When the emitter current sags β and it always sags as the LED junction ages β the received signal amplitude drops, the discrimination circuit tightens its acceptance window to compensate, and the reject rate climbs. Tokens that should pass at a 96 percent acceptance rate fall to the low 80s, the gate cycles faster than the spring return can reset, and you get the classic stack-up jam where three tokens wedge against the gate arm. The root cause is optical token rejection loop calibration drift, not a bent coin path.
Reference Signature Corruption and the Calibration Trap
Here is the part that infuriates me. Field calibration routines let a tech "learn" a new reference token, and techs do this the moment a machine starts rejecting β without cleaning the optical window first. A film of nickel dust and finger oil scatters the IR beam, the learned signature gets written with that scatter baked in, and now the comparator is calibrated to a contaminated baseline. Six weeks later the window gets cleaned, the scatter disappears, and the machine rejects everything because the stored signature no longer matches the clean optical path. The fix is brutal and simple: clean the optics, reset to factory signature, then recalibrate. Never learn a token through a dirty window. π―
Emitter Degradation and the Acceptance Window Collapse
Infrared emitters lose roughly 15 to 20 percent of their radiant output over 18 months of continuous duty. That degradation is silent. There is no fault code, no diagnostic flag β the comparator simply narrows its acceptance envelope until marginal tokens start bouncing. On a multi-token comparator handling three or four denominations, the narrowest signature dominates the loop, and the rejection cascade begins. If your coin comparator jamming incidents cluster around one denomination, the emitter is the prime suspect, not the gate spring.
Gate Timing, Spring Fatigue, and the Rejection Cascade
The divert gate fires on a millisecond pulse timed to the discrimination output. When the optical loop misfires, that pulse arrives early or late relative to the token's physical position, and the gate clips the token edge instead of clearing it. Do this forty times an hour and the return spring takes a set, the gate rests two degrees off home, and now even correctly discriminated tokens catch the gate lip. You cannot calibrate your way out of a fatigued spring β replace it, then re-run the loop calibration from a clean optical baseline. Anything less is a band-aid on a hemorrhage. βοΈ
Environmental Scatter and the Dirty-Window Feedback Trap
Casino floors are hostile optical environments. Cigarette smoke, cleaning solvent residue, and ambient IR from nearby displays all bleed into the comparator head. The discrimination circuit interprets ambient IR as a phantom signal, the rejection loop oscillates between accept and reject states at high frequency, and the gate chatters until it seizes. Shield the head, seal the window gasket, and audit the calibration interval against actual duty cycle rather than the manufacturer's optimistic schedule. Your optical token rejection loop calibration is only as good as the cleanest optical path in the bank.
The Diagnostic Sequence That Actually Works
Stop guessing. Pull the comparator, scope the emitter drive current against the datasheet spec, measure phototransistor output amplitude at the reference token position, and compare the live signature to the stored reference byte-for-byte. Any deviation beyond 8 percent means recalibrate; beyond 15 percent means replace the optical head. Log the ambient temperature and humidity at the head β thermal drift shifts IR wavelength and quietly re-tunes your entire acceptance window. Run the loop calibration twice and confirm the settling time converges under 40 milliseconds. If it does not converge, the discrimination circuit's gain stage is failing, and no amount of token learning will save it.