Free tool
Read filter
A pallet goes through a dock door and the reader reports a few thousand times. Move the filter, dedup and quiet-timeout settings and watch what reaches the ERP.
Move the filters
- Tag reports from the reader
- After the filter
- After deduplication
- Tags seen
- Events
Sightings
| Tag | First | Last | Reports | Peak |
|---|
What leaves the edge
How it works
What you’re looking at
A 40-carton pallet goes out through a dock door over four seconds. The reader has two antennas. Antenna 1 faces the door. Antenna 2 faces a staging shelf, where a tagged carton sits all day. Behind the door there’s a parked pallet that throws the odd faint read.
The reader reports every tag it sees, every time it sees it. The IoTReady reader can report up to 700 tags a second. A four-second pass in this simulation comes to about 3,000 rows: an EPC, which antenna, how loud, and when. Nobody wants those in an ERP.
So the edge does four things. It drops reads below a signal floor and from antennas that don’t count. It treats the same tag inside a short window as one sighting. It keeps one line per tag: first seen, last seen, how many reports, how loud at best. And when the gate goes quiet for a few seconds, it decides that whatever it saw has gone, and that’s the event.
The settings are where deployments go right or wrong. Switch antenna 2 on and the shelf carton keeps the gate open forever. Drop the floor to -70 dBm and the parked pallet’s reflections join the shipment, three tags that were never on it. Shorten the quiet timeout and the event closes sooner: about 3.7 s at 0.3 s of quiet, 6.4 s at 3 s. Stretch it to 5 s and it doesn’t close inside the eight-second window at all. In this stream the pallet never splits in two. On a real dock, with a gap while the forklift turns, it can. Try each one.
What leaves
An EPCIS 2.0 ObjectEvent, the GS1 standard for saying what was seen, when, where and why. The JSON on this page is the shape the standard describes: a list of EPCs, an event time, a read point, a business step (shipping) and a disposition (in_transit) from the Core Business Vocabulary. That one record is what an ERP connector turns into a delivery note. IoTReady’s platform does this step for you; see how the platform receives, processes and dispatches and the reader that feeds it.
What it assumes
The stream is simulated and deterministic: the same settings give everyone the same rows. Signal strengths follow a bell through the door, from about -72 dBm at the edges to about -45 dBm in the middle, which is an assumption, not a measurement. Readers that follow GS1’s LLRP standard (TCP port 5084 by default) report each tag with a microsecond timestamp; the IoTReady reader sends JSON with a millisecond timestamp. Either way, the rows here are what the edge sees after decoding, not a wire format. A real door also has a sensor or a light curtain to close the event; the quiet timeout stands in for it.
Where the numbers come from
The 700 tags a second is the peak rate on the IoTReady RFID Reader datasheet; Impinj rates the E710 chip from 50+ tags a second in its most sensitive mode to 1000+ in its fastest, 700+ in the FCC hybrid mode (chip datasheet, v1.4, large tag population, quiet RF). The EPCs use GS1’s documentation company prefix. The pipeline is the one in src/scripts/rfid-stream.ts, and the part 12 film runs a copy of it, so the counts match.