Buyer's Guide 📅 September 5, 2026 ⏱ 11 min read

How to Read a UHF RFID Reader Datasheet, Line by Line

A UHF RFID reader datasheet page annotated line by line, with transmit power, receive sensitivity and antenna port specifications marked up for comparison

A UHF RFID reader datasheet is a set of claims, and most of those claims are true. The difficulty is that a claim only means something when its conditions travel with it, and the conditions are the first thing cut when a specification is laid out on two pages. Two readers can print the same transmit power and a sensitivity figure within a decibel of each other, then behave very differently on the same dock door — because one vendor measured under a published method and the other quoted the best number the radio produced on a good afternoon.

This article is a decoder: it reads the lines in the order they appear and says what each one commits to. It separates the measured quantities with a defined test behind them from the configuration limits, and shows which lines become specifications the moment their conditions are stated. Where a public datasheet does the job well, we point at it — a well-qualified fixed-reader specification is used throughout as a worked example, because it qualifies its numbers in exactly the way a specifier should demand.

We design, tune and manufacture our own UHF readers and write the software that runs on them, so this is written from the side of the page that has to defend the numbers when a customer puts a spectrum analyser on the port. The last two sections are the practical payoff: a receipt-inspection protocol you can repeat, and a comparison template that makes three suppliers genuinely comparable.

Why two readers with the same headline numbers behave differently

Three lines carry most of the real difference between readers, and all three are routinely printed without the conditions that make them mean anything: how transmit power is specified, how receive sensitivity was measured, and what the thermal and duty-cycle numbers assume about how hard you intend to run the radio.

Everything else on the page is either a configuration limit (port count, connector type, frequency range) or a mechanical fact (dimensions, weight, mounting). Those are easy to compare because they are unambiguous. The three RF lines are the ones where two honest engineers can print different numbers for identical hardware, simply by testing differently.

Read range and read zone are different quantities

Before any of that, one vocabulary correction saves a great deal of argument. atlasRFIDstore defines read range as “the hypothetical range at which an RFID system can read a tag in a given direction in an ideal environment” — a single-direction figure under favourable conditions. The read zone is “a 3-dimensional area around an RFID antenna in which RFID tags can be read”.

You deploy into the zone. You are quoted the range. A better-shaped zone comes from geometry and per-port control; headline range tells you the link budget, and the zone tells you the deployment. On a portal a tightly bounded zone is usually worth more than reach — it is what keeps the reader on the pallet in front of it rather than the one in the next aisle. Range is a useful sanity check on the link budget; the zone is the thing you commission, map and sign off.

Read the datasheet with that split in mind and the marketing separates from the measurement almost on its own.

Transmit power: conducted, radiated, ERP and EIRP

Transmit power is the most compared line and the most misread, because two different quantities share the name.

Conducted is what the reader makes; radiated is what the site emits

Conducted power is measured at the antenna port. Impinj states this explicitly for its R700 series: the user configures “the conducted power available at the RF port on the reader”. Everything between that port and the air — coaxial loss, connector loss, antenna gain — changes what actually leaves your building. The cable run belongs to your site, so the port figure is the honest thing for a manufacturer to publish, and the one you add your own losses to.

atlasRFIDstore puts typical fixed-reader transmit power as configurable from 0 dBm up to a maximum “usually between 27 dBm and 31.5 dBm”. Published datasheets sit in and slightly above that window. The worked-example datasheet lists 10–30 dBm on PoE, 10–31.5 dBm on PoE+ in the ETSI lower band and 10–33 dBm on PoE+ for FCC and the ETSI upper band, with 0.25 dB resolution and 0.5 dB accuracy. The maximum is conditional on how the reader is powered — the higher-specification model in the same family sits at 20 dBm on plain PoE. That single dependency has decided the outcome of more pilots than any sensitivity difference.

ERP and EIRP use different reference antennas

ERP references a half-wave dipole; EIRP references an isotropic radiator. The dipole has 2.15 dB of gain over isotropic, so EIRP = ERP + 2.15 dB. Quoting an EIRP figure where the regulator specifies ERP overstates the system by 2.15 dB — enough to move a compliant install over the line. Convert once, at the start, and every later number in the budget stays comparable.

The regulatory ceilings are set on radiated power

ETSI EN 302 208 V3.4.1 sets the effective radiated power on each of the four high-power channels in the lower band at no more than 2 W e.r.p. (33 dBm e.r.p.) in a 200 kHz bandwidth, and on the upper-band high-power channels at no more than 4 W e.r.p. (36 dBm e.r.p.) in 400 kHz. The lower-band channels are centred at 865.7, 866.3, 866.9 and 867.5 MHz. FCC-band readers are supplied configured for 902–928 MHz. Our own UHF reader models carry WPC ETA and BIS registration for India’s de-licensed UHF band, and we configure FCC-band or ETSI-band variants per order for export.

The arithmetic, worked

Take one reader at 30.0 dBm at the port, assume 1.5 dB of cable and connector loss, and compare two antennas against the 33 dBm e.r.p. lower-band ceiling.

Line	6 dBic antenna	9 dBic antenna
Conducted power at port	30.0 dBm	30.0 dBm
Cable and connector loss (assumed)	-1.5 dB	-1.5 dB
Antenna gain	+6.0 dBic	+9.0 dBic
EIRP	34.5 dBm	37.5 dBm
ERP (EIRP - 2.15 dB)	32.35 dBm	35.35 dBm
Headroom to 33 dBm e.r.p.	+0.65 dB	-2.35 dB

Same reader, same setting, two outcomes. The 9 dBic build sits inside the ceiling with the port set to roughly 27.6 dBm. Ask any supplier which quantity their number is, and at which point in the chain.

Receive sensitivity and how it is actually measured

Sensitivity is the harder half of the link and the better predictor of field behaviour, because the tag’s backscatter is far weaker than the reader’s carrier. atlasRFIDstore notes that most fixed readers have a maximum receive sensitivity between -84 dBm and -92 dBm, with tag signals arriving anywhere from -30 to -92 dBm.

The method is the specification

The RAIN Alliance is candid about the state of play: among the manufacturers that report sensitivity, “it is doubtful that there is a consistent way of measuring that figure”. Its reader sensitivity whitepaper proposes testing to the international test-method standard for interrogator performance, and defines the result precisely — receiver sensitivity is the smallest signal strength in dBm for which the reader successfully passes 90% of a minimum of 100 consecutive tests, comparing at least the RN16 and ACK, or confirming the reader received a correct CRC-16 of the UII/EPC. The selected criterion has to be recorded, as do Tari, BLF, M, TRext, DR, RTcal and TRcal, the output power and the operating frequency.

That methodology has migrated into regulation. ETSI EN 302 208 measures receiver sensitivity as a conducted measurement, and recommends those same protocol settings for UHF Gen2 air-interface products, with Tari, RTcal, TRcal, BLF, DR and M recorded.

What a properly qualified figure looks like

The worked-example datasheet is a good model. It publishes three numbers rather than one: -94 dBm ideal, -93 dBm typical, -88 dBm challenging, all collected using a CISC conducted sensitivity test at 90% PSR in FCC DRM M8 mode at 30 dBm. The three conditions differ only in the return loss and reflection delay presented to the port — 22 dB or better at 4.5 ns for ideal, down to 12 dB at 38.2 ns for challenging.

Read that again: 6 dB of spread on one reader, from cabling and antenna match alone. That is wider than the gap between most competing headline figures. It is also why a sensitivity number becomes comparable the moment a packet-success-rate criterion and a test method travel with it. Sensitivity is environment-dependent in the field too — the RAIN paper notes it degrades as more readers operate in close proximity, and that reflections from nearby metal generate self-jammer signals into the receiver.

Three questions to send the manufacturer

What test method produced this figure, and at what packet success rate?
Is the figure conducted or radiated, and is it per-port or a best-port number?
Which RF mode and transmit power were used, and what return loss was presented at the port?

A manufacturer who tunes its own receivers can answer all three in a sentence. Treat the answer, not the number, as the specification.

Antenna ports, multiplexing and what four ports really buys

Port count is a coverage figure. Read it that way and it tells you a great deal.

Ports share the radio

On a conventional fixed reader the ports are multiplexed: one transmit and receive chain, switched between connectors. The worked-example reader has four monostatic RP-TNC ports and supports up to 32 antennas through an optional hub, all served by one radio. Four ports quadruple your coverage geometry; read throughput stays with that single radio. Ask directly whether the ports transmit simultaneously, and the answer tells you which architecture you are buying.

Dwell time is where moving loads are won

This is the arithmetic that matters at a dock door. Suppose a forklift carries a pallet through a portal in three seconds and you are cycling four antennas at a 200 ms dwell. Each port gets roughly 750 ms of the pass, split across four visits, and the tag on the far face may only be favourably oriented for part of that. Halve the dwell and you double the number of looks, at the cost of switching overhead. Ask for the antenna switching time as well as the configurable dwell range — the switching gap is dead air, and on a fast line it is a meaningful fraction of the window. Dwell tuning is the single biggest lever on read success at pallet tracking portals, and it is a configuration setting far more often than a hardware limit.

Monostatic, bistatic and connectors

A monostatic port transmits and receives on one connector, which means the receiver sits alongside its own carrier and depends on the reader’s isolation and cancellation to hear the tag. A bistatic pair splits transmit and receive across two connectors, which relieves the self-jammer problem at the cost of doubling antennas and cable runs. Both are legitimate; they suit different geometries, and the datasheet should say which the port count refers to.

Check the connector type on both the reader and the antennas you intend to buy — RP-TNC, SMA and N all appear in this market. Every adapter in the chain is insertion loss you subtract from the budget in the table above. Also confirm the port impedance and the minimum return loss specified (the worked example states 50 ohms and 10 dB minimum), and whether transmit power and sensitivity thresholds can be set per port. Per-port control is what lets you tune one antenna down so it holds its own aisle in a warehouse deployment while the others stay at full power.

Air interface, host protocol and what the software line commits to

The software lines are short and expensive. They determine how much integration work sits between the reader and your application.

Air interface

Expect EPCglobal UHF Class 1 Gen2, the UHF Gen2 air interface standard; the worked-example datasheet lists Gen2 v2. Gen2 v2 adds features beyond the base protocol, so if you need any of them, ask which are implemented in firmware rather than assuming the version label covers everything. Vendor extensions to the air interface should be named separately from the standard, and where a performance claim depends on one, that dependency belongs on the datasheet.

Host interface

There are three broad options and they cost you differently. LLRP is the standardised control protocol and gives you the best chance of swapping hardware later — but read the qualifier. The worked-example reader supports LLRP 1.0.1 with named vendor extensions, and that phrase is the whole story: portability holds until your application starts using the extensions, at which point you are on that vendor’s protocol wearing a standard’s name. A vendor SDK is faster to build against and locks you in more deliberately; the same reader ships an SDK for .NET and Java, an LLRP library for .NET, Java, C and C++, and an OpenAPI-described REST configuration API. A raw TCP or serial protocol is the cheapest to expose and the most expensive to consume, because you write the state machine yourself.

On-reader application logic

If you intend to filter, deduplicate or make decisions at the edge rather than shipping every read to a server, the processor and memory lines become load-bearing. The worked example publishes a dual-core 1 GHz processor with 1 GB flash and 1 GB RAM on a Linux 5.10 kernel; the higher model in the family raises that to a quad-core 1.4 GHz part. Whatever the figure, remember the firmware lives in that budget too — ask what is actually free for your application, and whether there is a supported container or package format rather than a bare filesystem.

Ask which SDK platforms are supplied for the clients you actually build (Android, Windows and Flutter cover most field applications), and whether a working reference application ships with them or you are starting from protocol documentation. Ask too how firmware and configuration are pushed to a hundred readers once they are in the field rather than one on a bench — that is the problem our ReaderSense edge device management exists to solve, and it is worth settling before the order rather than after the rollout.

Environment: IP rating, temperature and duty cycle

These lines decide whether the reader survives the site. They are cheap to check and expensive to get wrong.

Decoding the IP code

The first digit describes protection against solids, the second against liquids. The worked-example reader is rated IP50: dust protected, and specified for indoor mounting, which tells you exactly what to budget — at an outdoor portal or a wash-down line this reader lives inside a sealed housing with its own thermal plan.

Watch the second digit closely. The jet tests behind a 5 and the immersion test behind a 7 are separate tests, so for a wash-down line ask for the jet rating specifically, or for a dual-rated enclosure.

Operating versus storage temperature

These are two different numbers and only one of them describes a running radio. The worked example lists -20 °C to +50 °C operating and -20 °C to +70 °C storage, plus 5–95% non-condensing humidity. The question to ask is where derating begins: at what ambient does the reader start reducing transmit power or duty cycle to protect itself, and where is that behaviour documented? Measure the ambient inside the enclosure at the hottest hour of the hottest month, rather than the shaded outdoor figure — a sealed box on a steel canopy runs considerably warmer than the air around it.

Continuous versus intermittent duty

Ask for the RF duty cycle at which the temperature range and the power figures both hold. At a busy dock door the radio transmits for most of the shift, so a figure stated for continuous duty and one stated for intermittent duty describe two different thermal designs. Either answer is useful and honest — you size the enclosure and the airflow for the one you are given.

Power input and the PoE budget

This is where a tight budget quietly changes the specification. The worked-example series is powered exclusively over PoE (802.3af) or PoE+ (802.3at), with LLDP negotiation deciding the class and PoE as the default. On PoE, maximum transmit power is 30 dBm on the base model and 20 dBm on the higher-specification one, and the datasheet recommends PoE+ when sourcing external USB devices or driving the general-purpose outputs. Specify a PoE+ switch port and verify the negotiated power on site — the reader runs at the power the switch actually grants it, so confirming the class at commissioning locks in the power you specified.

The lines that become specifications once qualified

Some numbers carry their meaning in their conditions. Recognising which ones, and asking for the conditions, is most of the skill.

A bare read range

“Up to 12 m” becomes a specification the moment six things travel with it: the tag, its orientation, the material behind it, the transmit power, the antenna gain and the environment. Change any one of those and the figure moves by metres. Written as a sentence — this tag, at this orientation, on this substrate, at this power and antenna gain, in this environment, reads reliably to this distance — the same claim is immediately usable. Ask for the sentence.

Throughput with a population

A tags-per-second figure needs a stated tag population, air-interface mode and read window, because all three dominate the result. The Impinj R700 datasheet shows the honest form: up to 1100 reads per second, with the explicit note that the maximum “can only be achieved while using reader mode 0 in an FCC-like region”. That qualifier is worth more than the number. A throughput claim is reproducible on your own floor as soon as the mode, the region and the tag population are stated alongside it — so ask for those three, and the number becomes yours to verify.

Ask for the method

When a figure arrives without a test method, the productive move is to ask how it was measured rather than to argue about the figure. A supplier who designs the receiver can describe the bench in a few lines, and the speed and specificity of the answer tell you how close the answering desk sits to the measurement. Three follow-ups cover most cases: what was the pass criterion, what was the sample size, and can the setup be described well enough for us to reproduce it on receipt?

How we state our own figures

We hold ourselves to the same test. Every performance number we publish carries the conditions it was taken under, and where a figure is site-dependent we say so and confirm it by survey. Regulatory claims stay attached to the specific models they cover: our UHF reader models carry WPC ETA and BIS registration for India’s de-licensed UHF band, and we build FCC-band 902–928 MHz and ETSI-band 865–868 MHz variants configured per order. Because we design the hardware and write the firmware and application software in-house, the person who can explain a measurement works in the same building as the person who took it — which is the practical argument for buying from the manufacturer that designs and exports the hardware when the specification matters.

Measure it yourself: a protocol you can repeat on receipt

Every claim above can be checked on your own floor in an afternoon. Do it once, record it properly, and it becomes the acceptance evidence for the purchase order.

Equipment and setup

You need the reader under test, the antenna and cable you actually intend to deploy (rather than a shorter bench cable), a tape measure, a non-metallic stand, at least 50 tags of the exact model you will buy, and a laptop logging reads with timestamps and RSSI. Fix the antenna at deployment height and tilt, and mark floor positions at regular intervals along the boresight. Keep people and forklifts out of the aisle for the duration; a body in the beam changes the measurement.

Run a power sweep, and log it

Define the pass criterion before you start. A useful benchmark comes from the requirements-specification guidance published by RFID News, which recommends quantifying success as a first-pass read rate — 99.5% or higher for supply chain applications — and tying acceptance criteria directly to those metrics with measurable pass and fail thresholds. Then sweep: at each transmit power step, at each distance, in each of three tag orientations (parallel, perpendicular, 45°), present the tag 20 times and record how many presentations produced a read. Log every attempt. A read rate that gets eyeballed becomes a memory within a week; a read rate in a spreadsheet is evidence.

Repeat on the materials you actually ship

The same tag behaves like three different tags depending on what is behind it. On cardboard it performs close to its datasheet. Against liquid it loses substantial range, because water absorbs at UHF frequencies and detunes the antenna. On metal, an on-metal tag with a spacer or ground plane restores the read — which is why the material test decides the tag choice. Run the sweep three times — free air, on a loaded pallet of your densest product, and on the metal surface most common in your operation. The gap between those three curves determines your real design, and it is the number your own floor gives you: for design purposes it outranks every published figure.

Turn it into an annex

Write the results as a one-page sheet: equipment list, geometry, tag model, criterion, and a table of read rate by power, distance, orientation and material. Attach it to the purchase order as the acceptance-test annex, with the condition that the delivered units reproduce it within a stated tolerance. It costs an afternoon and it converts a negotiation about adjectives into a test both sides have agreed to. For a multi-site asset tracking rollout, run it once per site type rather than once per site — the material and geometry drive the result far more than the postcode.

A side-by-side comparison template

Once you have decoded three datasheets, put them into one sheet in the order a specifier actually reads. Blank cells are informative: they mark the questions to send back.

Line	Record exactly this	Why it moves the decision
Frequency range	Band in MHz, and which variant is being quoted	Determines the export market and the certification path
Transmit power	dBm range, conducted or radiated, and at which power source	PoE versus PoE+ can set the maximum
Power resolution and accuracy	dB steps and tolerance	Controls how finely you can shape the read zone
Receive sensitivity	dBm, test method, PSR criterion, RF mode	Comparable once the method and PSR travel with it
Antenna ports	Count, monostatic or bistatic, connector, impedance, return loss	Sets cabling cost and adapter losses
Port switching and dwell	Configurable dwell range and switching time	Decides read success on moving loads
Read rate	Reads per second plus mode, region and tag population	Those three make the figure reproducible on your floor
Air interface	Gen2 version, air interface standard, extensions named separately	Governs tag compatibility and future features
Host interface	LLRP version, SDK platforms, REST or MQTT support	Directly sets integration effort
Edge compute	Processor, RAM, flash, free space for your application	Determines whether filtering runs on the reader
Environment	IP rating, operating and storage range, duty cycle	Decides enclosure and thermal budget
Power input	PoE class and injector option, consumption at full power	Drives switch and UPS sizing

Compare installed cost, not unit cost

The reader is one line in the build. A fixed reader needs antennas, coaxial cable, brackets, an enclosure where the IP rating calls for one, a PoE+ switch port, and the labour to run and terminate every cable. An integrated reader with the antenna built in removes most of that at the cost of geometry flexibility. Price the whole portal both ways — the cheaper reader frequently loses on four antenna runs, and the more expensive one frequently wins by removing them.

Questions to send, copy-pasteable

Is the quoted transmit power conducted at the port or radiated, and at which power source?
What test method and packet success rate produced the receive sensitivity figure, and at what RF mode and power?
Are the sensitivity and power figures per-port or best-port?
Do the antenna ports transmit simultaneously, or are they multiplexed? What is the switching time?
Which LLRP version is supported, and which capabilities use vendor extensions?
How much flash and RAM remain free for a customer application after firmware?
At what RF duty cycle do the published temperature and power figures both hold?
What is the maximum transmit power when the reader is powered by plain PoE?

Record the answers verbatim in the same sheet, one column per supplier, with the date and the name of the person who answered. Three suppliers become comparable the moment they are answering the same eight questions in the same units — and the supplier who answers fastest and most specifically is usually the one who designed the thing.

Frequently asked questions

What is a good receive sensitivity for a UHF RFID reader?

atlasRFIDstore puts most fixed readers between -84 dBm and -92 dBm maximum receive sensitivity, and the best published figures reach -93 to -94 dBm under a stated method. Because 3 dB is a doubling of power, a few decibels are meaningful — and they mean most when both figures were measured the same way. Two numbers become directly comparable once both carry a stated packet success rate.

What is the difference between conducted and radiated power on an RFID reader?

Conducted power is measured at the reader’s antenna port, before any cable or antenna. Radiated power is what actually leaves the site, so it adds antenna gain and subtracts cable and connector loss. Datasheets quote conducted power because the cable run belongs to your site, so the port figure is the honest thing for a manufacturer to publish — and the one you add your own losses to. Regulatory ceilings are set on radiated power, so do the arithmetic for your own installation.

How should RFID reader receive sensitivity be measured?

Under the international test-method standard for interrogator performance, which is the methodology the RAIN Alliance references for reader sensitivity. On that basis, sensitivity is the smallest signal strength at which the reader passes 90% of at least 100 consecutive tests, comparing the RN16 and ACK or confirming a correct CRC-16 of the UII/EPC, with the RF parameters and output power recorded alongside. It matters because it makes two vendors’ numbers comparable, and ETSI EN 302 208 recommends the same protocol settings.

How far can a UHF RFID reader read?

An honest range figure is a sentence, not a number: this tag, at this orientation, on this substrate, at this power and antenna gain, in this environment. The same tag can lose several metres moving from cardboard to a liquid-filled pallet. Ask any supplier to state range as that full sentence with all six conditions attached, then verify it on your own materials.

What is the difference between read range and read zone?

atlasRFIDstore defines read range as the hypothetical range at which a system can read a tag in a given direction in an ideal environment — a single-direction figure. The read zone is the three-dimensional area around the antenna in which tags can actually be read. You are quoted the range, but you deploy into and commission the zone, and its shape usually matters more than its reach.

Does a four-port RFID reader read four antennas at once?

On a conventional fixed reader the four ports share one radio, switched between connectors — so four ports multiply your coverage geometry. One widely quoted fixed-reader datasheet has four monostatic ports and supports up to 32 antennas through a hub, all served by that single radio. Ask for the antenna dwell range and the switching time between ports, because those two settings decide read success on moving loads.

How do I compare two RFID readers fairly?

Normalise everything to the same conditions before comparing anything. Put both readers in one sheet with the same twelve lines, note whether each power figure is conducted or radiated, and record the test method and packet success rate behind each sensitivity figure. Where a datasheet is silent, send the question and record the answer verbatim, then verify the shortlist with a repeatable read-rate test on your own tags and materials.