Coin Comparator Token Jams: Optical Rejection Loop Calibration
A coin comparator that swallows a token and spits back a jam fault is almost never a mechanical problem—it is an optical rejection loop that has drifted out of calibration and is now misreading valid tokens as slugs. The three-eye infrared sensor array inside a standard comparator compares alloy signature, diameter shadow, and edge reflection against a stored baseline, and any deviation in the phototransistor gain or the emitter duty cycle triggers a hard reject. When that rejection fires mid-token, the solenoid gate snaps shut on a coin that already passed the entry fork, and you get the classic jam. 🎯
The optical rejection loop is where every technician should start, because the mechanical tolerance stack on these units is tight enough that dirty optics mimic worn gears almost perfectly. A comparator running at 12V with a failing IR emitter will show a 30 to 40 percent drop in photocurrent across the sensing gap, and the firmware interprets that drop as a token that is too dense or too reflective—an instant reject command to the divert gate. That is your jam, and it is electrical, not physical.
Why Baseline Drift Turns Valid Tokens Into Slugs
Every comparator factory-calibrates its optical thresholds against a reference token set, but those thresholds are stored in EEPROM and they drift with temperature and emitter aging. A unit that calibrated cleanly at 22 degrees Celsius will run 8 to 12 percent outside its acceptance window at 38 degrees in a busy machine room. The token jam rate climbs correspondingly, and operators start swapping comparators that are actually fine. The real fix is re-teaching the acceptance window against a live token under load, not bench-testing with a static sample.
Here is the hard truth most field techs refuse to accept: a comparator that jams one token in fifty is not marginal, it is dangerous. That single jam event holds the divert gate closed for 400 to 900 milliseconds, and in a high-throughput hopper that window is enough to stack three more tokens behind the obstruction and shear the gate return spring. Once the spring deforms, the rejection loop never fully resets and you get cascading jams that look like a firmware bug but are pure mechanical fatigue. 🛠️
Calibrating the Rejection Loop Without Guesswork
Pull the comparator, connect a scope to the phototransistor output, and watch the pulse train as a known-good token passes the three sensing eyes. You want clean square transitions with rise times under 200 microseconds and no ringing on the trailing edge. If you see rounded shoulders or a slow decay, the emitter is degrading and no amount of gate adjustment will fix the rejection loop. Replace the emitter pair first, then re-teach the acceptance window with the new optical baseline.
The calibration sequence that actually holds is: set emitter current to factory spec, capture the photocurrent signature for three passing tokens, average the peaks, and set the reject threshold at 85 percent of that average. Anything below 85 percent gets rejected as a slug, anything above passes. This single threshold eliminates most coin comparator token jams because it accounts for real-world optical variance instead of the idealized factory reference. 🎲
The Duty Cycle Trap Nobody Talks About
Comparators pulse their IR emitters at 38 to 40 kHz to reject ambient light interference, and that duty cycle is the most overlooked variable in the whole rejection loop. If the driver transistor heats up and the duty cycle shifts by even 2 percent, the demodulator loses phase lock and the comparator reads garbage. You will see intermittent jams that correlate with machine uptime—cold machine runs fine, hot machine jams every twenty minutes. That is not a token problem, it is a thermal drift problem in the emitter driver stage.
Scope the emitter drive line under load and confirm the duty cycle stays within a half-percent of spec across the full operating temperature range. If it walks, the driver transistor is the culprit, and swapping it is a ten-minute job that saves you from replacing an entire comparator assembly. The rejection loop only stays in calibration when the optical pulse train is rock stable, and stability starts at the driver, not the sensor.
Gate timing matters just as much. The divert solenoid must release within 15 milliseconds of the accept decision, or the token clips the gate edge and jams. Measure the solenoid response with a current probe and confirm the release profile is crisp. A lazy gate and a drifted optical threshold together produce exactly the jam pattern that sends operators into replacement cycles they never needed.