Conducting an RFID site survey: from objectives to a measured read-zone map

01 / FIELD NOTE

Keep the decision tied to the operating context.

A site survey exists because a catalogue is the wrong instrument. The read range on a tag’s data sheet is measured in open air against a bench antenna, and real sites supply none of those conditions: metal racks, damp stock, concrete walls, moving carts and other radio devices all act on a UHF signal before the reader gets a say. The survey converts one site’s physics into hardware decisions — which tag families, which reader positions, which power and network provisions — and it does that by producing evidence rather than expectations. Its output is a map, a target set and a pilot plan, and its discipline is that nothing is installed on a promise.

The survey starts with objectives, because objectives decide every later choice. The questions come first: which assets are to be tracked, how they move, what decision each read will support, and whether the answer has to be real-time, near-time or checkpoint-time. A portal at a dock answers what crossed the doorway; a shelf reader answers what sits on the shelf; a handheld answers what is where now. Each is a different system, and the survey’s first job is to name which one the operation actually needs. From the objectives fall the performance targets: the read range that has to hold, the number of tags a single scan has to resolve, and the latency allowed between a read and the record updating.

Sample tags go on the real product before the whole site is walked. A tag that reads well on a carton in the hand can fail on the same carton at the bottom of a pallet, against a metal shelf, or inside a plastic tote, because each of those changes the tag’s environment more than the tag changes it. The sample test is the tag and its attachment together, on the actual item, in the actual position in which it will be read. A bench result measures the tag; a site result measures the installation, and the installation is what is being bought.

The walkthrough is a documentation exercise. With a floor plan as the base, trace how items actually move: receiving, staging, aisles, packing, dispatch, and every doorway, tunnel or chokepoint between departments. Photograph the zones, sketch shelf heights and aisle widths, note the construction of floor, walls, racks and ceilings, and record which areas are busy at which times of the shift. The walkthrough produces the material list the radio analysis will work against, because the signal at any position is decided by the building before it is decided by the reader.

The walkthrough’s material notes are the survey’s first hard data. Solid metal blocks and reflects: steel racks, pallet frames and conveyor structure detune tags near them, so a tag mounted on metal needs a purpose-built on-metal construction, or a standoff that lifts it off the surface. Anything wet absorbs: stock with high moisture content and liquids in containers take energy out of the field, so tags keep their distance from direct contact and the read point is chosen accordingly. Stacked pallets shadow the tags on their inner faces, which is why the tag position on the unit load is specified as deliberately as the tag itself. Concrete, glass, dense board and insulated cladding each bend or block the signal in their own way, and each earns an entry in the register.

The radio background is measured, not assumed. A spectrum analyzer sweep across the UHF allocations around 860–928 MHz shows the noise floor the readers will have to work against, and spikes inside the band mark the equipment that will compete: motors, frequency converters and welders on nearby power, wireless devices in adjacent spectrum, and the everyday machines a facility never thinks of as radios. The key discipline is that interference is time-varying. Machines idle at night and start at shift change; devices switch on and off through the day. Scans at peak production, at startup and shutdown, and during maintenance, repeated over several days and logged rather than summarized, are what separate a persistent blocker from a transient spike — and a transient spike is still a design input.

Read-zone mapping turns the objectives into positions on the plan. Each checkpoint the operation named gets a zone: a portal antenna at the dock, an overhead reader over a conveyor, side antennas at a gate. For each zone the survey fixes the antenna family — directional for a focused chokepoint, circular-polarized where tag orientation cannot be controlled, omni-directional for area coverage — and the height, angle and spacing are then tuned by a grid test that walks a tagged item through the zone and marks where reads hold and where they drop. The null zones the grid finds, where reflection or orientation defeats the signal, are either redesigned or recorded as workflow fallbacks; either way they go onto the map instead of being discovered after installation.

Power and network are surveyed like the signal. Every reader position needs either a power-over-Ethernet drop or a nearby outlet, so the plan records the cable runs from the switch to the read zone, the outlet positions, surge protection, and a decision about backup power. Outdoor positions add weather-rated enclosures and protected connections. The network side records where the readers will connect, the wired or wireless coverage at each position, the firewall rules the readers must pass, and a data-routing decision — whether reads are processed at the site or sent to a platform — because that decision is the latency budget the targets measure against.

The targets are written down before the pilot, and the pilot measures them. For each zone the survey records the minimum acceptable range, the throughput expected at peak — a portal serving a conveyor resolves tags an order of magnitude faster than a handheld held by a person, and the target set says what each must manage — and a tolerance for missed reads, expressed as what the workflow does when one occurs rather than as an ideal. The pilot runs at one door or one aisle with real transactions, and its job is to reconcile the map with reality: antenna positions move, angles tilt, power settings adjust, until the measurements meet the targets that were written down at the start.

The survey ends with the report and the honest limit of the report. The report is the installation plan: the zones on the floor plan, the antenna positions and angles, the cable routes, the interference zones that remain, the tag families that tested well on the real products, and the target set the pilot ran against. Its limit is that a survey is a snapshot. Layouts change, equipment arrives, stock composition shifts with the season, and the radio environment drifts along with the building. When any of that changes materially, the correct response is a re-survey rather than a belief that the first map still holds.