A Tablet Miscount Costs More Than the Bottle

Six ways tablets miscount, ranked by how often I find them

Counting errors are rarely random. They cluster by lane, by time of shift, or by speed, and the cluster is the diagnosis.

Feed system design: where the count is decided

Multi-stage channels with independent amplitude

A single long vibratory channel has to do two contradictory jobs: bulk transport and singulation. Splitting them works better. I specify a bulk hopper stage feeding two or three cascade stages, each with its own amplitude control, usually 35–75% of maximum, tuned so the stream thins visibly from one stage to the next. At the final stage you want single-file tablets with a clear gap, not a moving carpet. If it looks like a carpet, no detector will save you.

The gate is where overcounts are born

Cut-off response time is the most under-specified number on a tablet counting machine. A pneumatic gate at 40–60 ms lets a trailing tablet through on every cycle; at 45 cycles per minute across twelve lanes that is a slow overcount the in-process check will only catch intermittently. A servo gate closing in under 18 ms removes it. I also ask that the gate closes on a predicted count and that the last stage decelerates as the target approaches, so the final tablets arrive slowly enough to be cut cleanly.

Geometry, drop height and extraction

Singulation geometry follows the tablet: lane width roughly 1.3 to 1.6 times tablet diameter for round tablets, and a shallower channel angle for capsule-shaped tablets, which stack rather than roll. Keep the gate-to-shoulder drop at 90–120 mm with a spiral or soft-catch funnel. Extraction sits at the counting head, not at the ceiling: 300–500 m³/h at about 0.5 m/s capture velocity clears the sensing zone without pulling tablets off the channel, which is a real risk if you oversize it.

Detection: infrared array versus camera, and false rejects

An infrared or laser array is fast, robust and cheaper, and it handles most compressed tablets without complaint. It struggles when the product is small, very thin or translucent, because the beam edge goes soft. A high-speed camera with image processing reads shape as well as interruption, so it separates two touching tablets that an array reads as one, and it can be trained on banded capsules, odd shapes and minitablets down to about 2 mm. It costs two to three times more and needs a clean window and stable lighting.

The tuning mistake I see most is a team widening the detection window to suppress false rejects until the machine stops reporting real faults. Do it the other way round. Quantify the true defect rate, set the window to catch it with margin, then drive false rejects down at the feed end. Above roughly 1% false rejects, the fault is upstream, not in the algorithm.

Product damage control

Every handling stage removes a little of the tablet. On a film-coated round tablet we measure 0.1–0.3% friability loss across a full bottling line; on a fragile chewable or an orally disintegrating tablet it can reach 1.5%, and most of it happens in the hopper and first cascade stage rather than at the counting head. The dust from that damage then feeds straight back into the sensing problem, so the two are one loop. A deduster and polisher upstream, running at low elevation with gentle air, removes surface dust and loose fragments before the counter sees them. Control hopper fill at roughly 30–70% as well: a deep bed compacts fragile tablets, and a bed that runs empty and gets topped up gives you a surge of damage followed by a surge of dust.

From unscrambler to reconciliation

Data integrity and qualification across regions

Recipe management with three access levels is the baseline: an operator selects a recipe, a supervisor edits setpoints, only an administrator creates one. Every edit writes to a secure audit trail with user, timestamp, old value and new value. The batch report should carry count distribution per lane, rejects by reason code, in-process check results and operator ID.

Qualification expectations differ mainly in the wrapper. In the EU, OQ challenge tests with known counts sit inside the IQ/OQ/PQ structure of Annex 15, and Annex 11 covers the computerised system. In the US the same electronic record controls fall under 21 CFR Part 11 and the drug CGMP obligations under 21 CFR 211. Plants audited against PIC/S guidance, the practical reference across much of Southeast Asia, and those supplying Gulf authorities or India's CDSCO, generally accept the same package. I have repeatedly seen audits stall because the protocol and reports existed only in the builder's language, so insist on English-language protocols, raw data and signed reports in the contract rather than as a post-order extra.

Field case: EU contract manufacturer at 120 bottles per minute

The plant ran two formats over three shifts, five days: an 8 mm film-coated round tablet at 100-count and a 14 mm capsule-shaped tablet at 30-count, both into 100 ml HDPE bottles with a desiccant sachet and a child-resistant cap. Line rate was 120 bottles per minute on the round tablet and about 100 on the capsule-shaped one.

The plant's own view was that the biggest gain was not the count but the reconciliation: with rejects logged by reason code, the quarterly review stopped being an argument about whether a bottle had gone missing.

Buying mistakes that only show up after installation

Speed is quoted on the easiest tablet in the range, so ask for the rate re-run on your largest and your most fragile product, and ask whether the figure is mechanical cycles or sustained output over an hour. Second is utilities: 400 V at 50 Hz is straightforward in the EU, but a Gulf plant on 220–240 V at 50 Hz with 45 °C ambient needs derated drives and a cooled cabinet. Third is line balance, because a counter rated at 140 bottles per minute feeding a capper that holds 115 gives you a capper that starves and a counter that stops and starts, and every restart is a chance for a gate timing error. Finally, put a spare gate actuator, a spare sensing window and a full set of the most-used change parts into the first order.

Questions buyers ask me most

Can the counter be retrofit into an existing line?
Usually yes, if conveyor centreline height and bottle handling are standard. Measure infeed pitch and available floor length first; a twelve-lane head with its funnel assembly needs more depth than most people allow for.

How do I reduce false rejects without hiding real faults?
Fix the inputs. Most high false-reject rates come from dust, bounce or double feeding, not from the threshold. Quantify your true defect rate, set the window around it, then work on the feed end.

Do I need a checkweigher if the tablet counting machine is accurate?
Yes, for any regulated product. The counter is a process control; the checkweigher is the only 100% verification of what went into the bottle. Set its band tight enough to see a single tablet.

What changeover time should I expect?
With tool-free change parts and stored recipes, 15–25 minutes for a diameter change on the same bottle. Closer to 45–60 minutes if you also change bottle size, capper chuck and label format.

Is the qualification package different for the US and the EU?
The structure is largely the same, IQ, OQ and PQ with challenge tests using known counts. The difference is the wrapper: Part 11 and 21 CFR 211 for the US, Annex 11 and Annex 15 for the EU. PIC/S-audited plants in Southeast Asia accept the same package.

How much tablet damage should I accept?
Below 0.5% friability loss across the line for a conventional film-coated tablet, below 1% for a fragile chewable or ODT. Above that, look at the hopper and first cascade stage before the counting head.

Written by Helen Xu | Chief Industrial Application Engineer

Helen Xu is a Chief Industrial Application Engineer with 9 years of specialized experience in packaging machinery and liquid filling machine design, equipment model selection, and full production line process optimization. He focuses on delivering customized packaging & filling solutions for pharmaceutical, food, and chemical manufacturing industries, with mature practical expertise in GMP compliance, ISO 9001 quality management standards, and turnkey large-scale filling & packaging production line integration.