JNXRFID · Know-how Series · Deployment Reality KH‑006

The Count Is Right and the Line Is Still Moving

A correct count that arrives too late to keep the line moving is not a counting problem. It is a throughput problem, and throughput is set by how many tags are in the zone at once, not by how fast one tag is read.

13 min read · Applies to UHF passive RFID · Tag density, anti-collision and read-point throughput · First edition, October 2026

• Tag density
• Anti-collision
• Throughput
• Queueing
• Read point

Correct is not the same as fast

The symptom is worth stating precisely, because it is regularly misread as a broken installation. Nothing is missed. Every pallet is eventually read, and the report is right. What changes is the delay: the pallet has already moved on by the time the result appears, and at the busiest moment that delay is long enough to stop the line.

Three different causes produce this same visible symptom, and they need three different answers:

• The zone holds too many tags at once. A read zone that is generous in coverage is also generous in the number of tags it asks about in a single pass.
• The tags are close enough to interfere with each other. A tight bundle of labels couples them to one another, and each tag detunes slightly when it is next to its neighbours.
• The read point is downstream of something slower. The reader is fast; the pallet lifter, the conveyor gap or the encoder that produced the labels is not.

Anti-collision is a queue, and every queue costs time

The EPC Gen2 air interface is built for a dense world. When several tags answer in the same instant, the reader detects the collision, selects one tag, and asks the rest to hold. It reads the selected tag, then repeats for the next. This is the Q algorithm: the reader works through the population one tag at a time, and the number of round trips grows with the number of tags present.

This is why a quoted tags-per-second figure is a ceiling, not an expectation. It describes the rate at which the reader can process replies when it is not fighting other tags for the same instant. The same reader against a denser population takes proportionally longer, because each additional tag adds a round trip rather than adding a parallel reply.

Two further mechanisms compound it:

• Tag-to-tag coupling. When valuable items are stacked or bundled tightly, their antennas interact and detune. A tag that reads on its own can fail when it is next to its neighbours, which turns a throughput problem into an apparent read-rate problem.
• An oversized zone. One antenna covering a full pallet position also covers whatever is beside it. Dividing a large field into several smaller zones usually raises throughput more than any hardware change, because each zone clears faster.

Measure the lag before you change anything

The measurement that matters is not a read rate quoted in a datasheet. It is the time between a tag entering the zone and its result appearing, measured with the real population present. Four numbers separate the three causes above, and each one points at a different decision.

Run the first and third together. If the sparse-zone rate is close to the manufacturer's figure and the loaded rate is far below it, the cause is density or coupling, and the answer lies in the zone and the population, not in the reader. If the sparse rate is also low, the environment or the tag construction is involved, and that is a different article.

What actually raises throughput

Once density is confirmed as the cause, the controls are ordered by what they change. Reading them in the wrong order is expensive, because a hardware change is the last step, not the first.

1. Reduce the tag count per zone. Read a defined zone rather than a generous one. Fewer tags per zone means fewer round trips, and it is a layout change rather than a purchase.
2. Remove the coupling. Separate the labels physically, or change to a construction that is not affected by a neighbour at close range.
3. Give the read point somewhere to be. A read point positioned where pallets are forced to stop, rather than where they happen to pass, converts a moving target into a stationary one.
4. Only then consider the reader. A reader with a higher quoted ceiling helps only when the zone, the population and the dwell are already settled. It does nothing for coupling, and it does nothing for a bottleneck that sits upstream of the antenna.

A diagnostic sequence that finds the constraint

The steps below are the order that wastes the least. Each one either eliminates a cause or names the next measurement, and none of them requires a purchase to be made first.

The output of step 5 is a number, not a product name. A read point quoted at a peak rate and specified against a measured loaded rate are different purchases, and the second one is the one that can be held to a specification.

What to specify when you buy

When the read point genuinely is the constraint, the specification has to describe a loaded zone rather than an empty one. Four fields decide it.

Ask for the figure to be quoted against a loaded zone, with the population stated. A supplier who quotes only a peak rate is telling you about an empty room. A supplier who quotes a rate at a stated population is telling you about yours.