“RFID tag not reading” is how the fault always gets reported, and the tag is usually not the reason. In a live deployment the tag is the least common culprit — in our field experience, regional band configuration and reader settings cause far more misses than dead silicon, and neither shows up on a spec sheet. What helps is a fixed order of elimination, using tools the reader already has: a free-air bench test, a region and power check, RSSI on a handheld, and the reader’s own raw log.
Start here: is it one tag, one reader, or the whole site?
Before changing a setting, establish the shape of the failure.
| What you observe | Most likely layer | Configuration or hardware? |
|---|---|---|
| One tag fails, identical tags read fine | Never encoded, or damaged | Hardware — replace it |
| A whole new batch reads poorly everywhere | Region profile, or wrong tag for the surface | Configuration first |
| Reads at the bench, not at working distance | Antenna gain, angle, environment | Both |
| Reads on the handheld, not on the fixed portal | Session, target, dwell, geometry | Configuration |
| Reader log shows the EPC, your application does not | Filters, dedup window, middleware | Configuration |
| Reads fine, plus tags you did not want | Stray reads: power, aim, shielding | Both |
Why is my RFID tag not reading? First settle which tag you are holding
The question arrives as one question, but there are two unrelated radio systems behind it, and almost every check differs between them. A 13.56 MHz HF or NFC credential couples inductively in the near field and is meant to be presented deliberately at a few centimetres. A passive UHF tag on India’s 865–868 MHz band is read by far-field backscatter at metres. Diagnostic advice written for one is usually wrong for the other, which is why so much of what you will find on this search makes no difference when you try it.
Three examples of how badly it transfers. Raising transmit power is the reflex UHF answer, and on an HF desktop reader there is usually nothing to raise. Checking the reader’s region profile catches more live UHF failures than dead silicon does, and at 13.56 MHz there is no region profile to check. And the fault behind most “the card number appears but nothing happens” calls — sector keys and authentication sitting in the application rather than in the reader — has no UHF equivalent at all.
| What you are holding | Band and standards | Which track |
|---|---|---|
| A card, keyfob, wristband inlay or sticker you tap against a pad or a door | 13.56 MHz HF, with NFC as a consumer profile of it — ISO/IEC 14443 (proximity) or ISO/IEC 15693 (vicinity) | Track 1 |
| A hard tag, label or inlay read from a gate, portal or handheld across a room | Passive UHF, 865–868 MHz in India — EPC Gen2, ISO/IEC 18000-63, marketed as RAIN RFID | Track 2 |
| A thick proximity card or fob on an older door or locker estate | 125 kHz LF — TK4100, EM4200 and similar | Neither; see the note below |
If you cannot tell by looking, the reader will tell you. An HF unit does not see a UHF tag at any distance, and a UHF reader does not see an HF card held flat against the antenna. No setting bridges that gap, and neither does firmware. The same holds for a legacy LF estate: 125 kHz proximity credentials sit at an entirely different frequency, so cards or bands bought for the wrong one are scrap rather than a tuning problem. Confirm what your fitted locks and turnstiles actually read before reordering — many older units are 125 kHz and newer ones 13.56 MHz.
One terminology trap, because it reaches tender documents and then purchase orders: EPC Gen2, ISO/IEC 18000-63 and RAIN RFID all describe passive UHF only. “Gen2 compliant” in an HF reader specification is a drafting error, not a requirement. Our UHF vs HF vs NFC comparison sets out where each band belongs, and RFID frequencies in India carries the UHF channel plan and power limits.
Track 1 — an HF or NFC credential at 13.56 MHz
Steps 1 to 6 below are the UHF track. Almost none of it applies here: there is no country band profile to match, no channel plan to work back to, and usually no power dial to turn. Work down this list instead, which is ordered cheapest check first.
Why does my card work sometimes but not always?
- Presentation, not range. The usable tap depends more on the credential’s antenna area than on the reader. An ISO-size card couples better than a wristband inlay or a small keyfob, so the same reader gives a shorter typical tap on the smaller credential. A hurried pass across the pad can leave a transaction half-completed with no error shown. Hold the credential flat and still for a moment and see whether the intermittence disappears before you change anything else.
- What the credential — or the reader — is sitting against. This is the step most people skip, because HF is sold as metal-tolerant. It is only metal-tolerant relative to UHF. A bare HF antenna detunes against steel: a card in a metal cardholder or clipped to a steel badge reel, a tag stuck straight onto equipment, a reader recessed under a metal counter or resting on a steel desk. For tags on equipment the fix is construction, not configuration — an ABS or anti-metal ABS body reads on metal that would otherwise detune a bare HF coil, and our HF and NFC tag range lists those form factors. For readers, move the unit off the metal or mount it in the housing it was designed for.
- Two read points too close together. Two HF antennas mounted close together at adjacent lanes or adjacent doors can interfere with each other. If the fault follows the lane rather than the card, separate them and retest.
- A cracked coil. Printed cards flex in pockets and the antenna loop breaks. There is no equivalent of the UHF free-air bench test, because free air is already the normal condition, so swap both ends instead: read the failing card on a second reader, and read a known-good card from the same batch on the failing reader. Four outcomes, and between them they tell you whether the fault is the credential, the reader, or the mounting.
Why does the reader show a card number but my software still refuses it?
This is the HF counterpart of an EPC appearing in the reader log but not in the application, and in our field experience it is the HF fault we are asked about most. Reading a card’s UID is not the same as reading its protected sectors. Mifare sector keys and DESFire authentication belong in your application, not in the reader, which is what a bundled SDK and demo app are for. So “supports Mifare” means the reader can talk to the chip, not that it can read your data. Note the symptom this produces, because it is diagnostic: a card refused consistently, not one that taps intermittently. Intermittence belongs in the list above.
The hard version of this problem is key custody. If a previous integrator encoded your estate, the sector keys may simply not be in your possession, and no reader resolves that — what nobody can do is recover keys someone else wrote. We encode sector keys in-house on cards we manufacture, so a fresh batch is straightforward; recovering an old one is not. Settle who holds the keys before planning a migration. Our HF reader page covers what to send us at specification stage.
Why does one card type read and another does not?
ISO/IEC 14443 and ISO/IEC 15693 are different standards sharing the same 13.56 MHz carrier, and a reader has to be specified and enabled for the family you actually hold. Published card lists differ unit by unit across any vendor’s range, ours included: some models name only the Mifare families, while ISO 15693 support, for chips such as the ICODE and NTAG21x families, is listed on others. Library tags and some asset tags are ISO 15693 rather than ISO 14443, which is exactly the case that fails on a reader bought for access cards. Match the chip part number on your credential against the reader’s specification rather than the family name, and send us that part number if you want the match confirmed in writing rather than inferred.
A 15693 vicinity chip on a larger antenna also stretches the usable distance somewhat beyond a normal 14443 tap, so if a tender specified a working distance measured on one card type, do not assume it carries to the other.
Why does the library exit gate alarm on a book that was issued?
Because on HF the gate is deliberately not asking the database. An ISO 15693 library tag carries an AFI or EAS bit, and it is the kiosk that maintains it — cleared when the item is issued, set again when it comes back — so the gate can alarm on the state of the bit alone. That is why an HF gate keeps working through a network outage, and it is also why this particular fault almost always sits in the kiosk software rather than in the gate, the tag or the catalogue. Our Koha and SIP2 integration guide sets out both the HF and UHF designs for that gate.
When HF is not the answer at all
HF does cope well with credentials pressed together — a staff pad reading a stack of books is an HF strength, where a UHF pad has to be shielded to stop it reading the neighbouring books. What HF will not do is reach across a doorway or down an aisle. If the real requirement turned out to be hands-free capture at distance, no amount of tuning gets there and the honest answer is a band change rather than a fault. One point to state plainly: Identium is BIS certified, and it is our UHF readers and antennas that hold WPC approval for India’s licence-exempt UHF RFID band. That approval is specific to UHF hardware and does not carry across to HF units. We hold no CE, FCC, RoHS, REACH or ISO 9001 certification, and as of 2026 we would ask you to confirm the current position for HF equipment with WPC or BIS rather than rely on a supplier’s summary.
Track 2 — a passive UHF tag on India’s 865–868 MHz band
Steps 1 to 6 below are this track, already in cheapest-first order: bench-test the tag, check the region and power, then the environment, then antenna geometry, then reader settings, then software. Two things trip people up repeatedly, and neither is a step so much as a piece of arithmetic the steps leave to you.
Why did the tag read on the bench but not on site?
Because the bench removed four variables at once, and the fix is to put them back one at a time rather than all together. In order, cheapest first:
- Re-mount it and read it in the same spot. This isolates the surface on its own — see Step 3. If the read dies the moment the tag goes back on the asset, it is detuning, and no setting recovers it.
- Walk it out to working distance with the same reader. This isolates geometry, gain and nulls — Step 4.
- Move to the real reader, antenna and cable run. This is where the region profile, the power sum and a damaged coax or mis-mapped port enter — Step 2.
- Only then settings, and only then software. Steps 5 and 6.
Two honest caveats. A bench test at low power proves the silicon answers; it does not predict range, and read ranges quoted anywhere are typical, environment-dependent figures rather than a specification. And the site had things the bench did not: a full rack where the pilot had an empty one, condensation after cold storage, and an operator’s body between antenna and tag, which is mostly water and a genuine attenuator.
The power sum the datasheets leave to you
Step 2 makes the point that what is regulated is radiated power rather than the number on the reader dial. The arithmetic behind that is yours to do, and it is worth doing before you conclude you have a tag problem: start from the reader’s conducted output in dBm, add the antenna’s gain in dBi, subtract cable and connector loss over the run you actually installed — a long or lossy run can cost a few dB and behave like a lower-gain antenna — and compare the result against the e.r.p. ceiling in the governing notification. Check which scale each figure is quoted in before you subtract: e.r.p. and e.i.r.p. differ by a fixed standard offset of about 2.15 dB, and mixing them quietly moves your answer.
The gap this exposes is real and specific. Our fixed readers on the R2000 generation are specified up to 33 dBm output, and our two-port UHF gate kit is listed with 9 dBi antennas; India’s 2021 short-range-device rules permit up to 2 W e.r.p., and only in the four channels of up to 200 kHz centred at 865.7, 866.3, 866.9 and 867.5 MHz. Run a reader at full output into a high-gain antenna on a short cable and the radiated figure can sit above that ceiling. We are deliberately not publishing a permissible conducted-dBm number here, because it depends on your antenna, your cable and on how the limit is measured, and we have not seen a WPC text that settles the measurement method or any feed-loss allowance. Do the subtraction with your own figures, confirm the method with WPC, and ask your reader supplier which figure their type-approval was granted against. Our BIS and WPC compliance guide lists what to ask for in writing. In the meantime, increase dwell before you touch power — it is free and it does not create stray reads.
Why does a reader bought abroad read our tags badly rather than not at all?
Because the failure is gradual, which is what makes it hard to spot. A unit left on a US 902–928 MHz or a wide global profile will often read Indian-band inlays, but typically not at its best: reduced range, inconsistent captures, a portal that works on Tuesday and not on Thursday. Separately from how the reads look, it is transmitting outside the slice India permits, and that is a compliance problem whether or not the performance is acceptable to you.
The regional trap is narrower than it is usually described. India and ETSI both use 865–868 MHz, so European equipment is generally in the same window; it is North American equipment at roughly 902–928 MHz that genuinely mismatches. Our frequency-by-country guide and the note on 900 MHz band fragmentation in sourcing cover what to check on an import.
One related thing that looks like a contradiction and is not. A supplier quoting 865–867 MHz on an older certificate may be entirely accurate: India’s licence-exempt UHF RFID band is 865–868 MHz under G.S.R. 853(E) of 10 December 2021, made in supersession of the 2005 rules that set 865–867 MHz, but Rule 1(3) of the 2021 rules grandfathers equipment type-approved under the older rules for its service life. That is a question about the date on the certificate, not evidence that the current allocation is still 865–867 MHz. Notifications get amended, so confirm the position with WPC before you write a figure into a specification.
What if the tag is both?
A true single-IC dual-frequency NFC and UHF tag carries an HF coil and a UHF dipole wired to one die, so the EPC bank and the NFC record are separate views of the same memory. On that part, a tag that taps on a phone but never answers the portal — or the reverse — is one interface failing rather than a dead tag, so run the two tracks independently instead of condemning the batch. Establish which part you are actually holding first, though: a laminated “combo” label is a separate NFC inlay and a separate UHF inlay with two identities, and there one silent interface really can mean one dead chip. Either way, an encoding station has to write and verify both interfaces, and not every printer or service bureau does.
Step 1 — prove the tag is alive with a free-air bench test
Take the failing tag off the asset. Hold it in free air on a wooden or plastic bench, away from metal, liquid and your own hand, and read it at low power. You are not measuring range here, only whether the silicon answers.
- Valid EPC returned. The tag is alive; the problem is downstream in environment, geometry or settings. This is the usual outcome.
- It answers, but the EPC is blank, all zeros or a factory pattern. It was never encoded, or encoding failed at the printer — a process issue, and fixable. Reading the TID bank usually confirms the chip is healthy even when EPC memory is locked or empty.
- No response at all, even held against the reader antenna. Treat the tag as dead: cracked chip, broken strap, delamination, moisture ingress or crush damage. No setting recovers this.
Sample five or ten from the same reel. One failure is attrition; ten in a row points at the encoding step or the application method — the kind of avoidable error covered in ten common RFID implementation mistakes.
Step 2 — check the region: band, power limit and reader profile
More live failures hide here than in the tags. Passive UHF sits between roughly 860 and 960 MHz worldwide, but each regulator opens only a slice of it. In India, as of 2026, the WPC Wing of the Department of Telecommunications has allocated 865–868 MHz for delicensed UHF RFID under the 2021 short-range-device rules (G.S.R. 853(E), which superseded the 2005 865–867 MHz rules), with a radiated-power ceiling set out in that notification. Notifications get amended, so confirm the current rule with WPC before you change any setting.
Then check four things: that the reader’s region profile matches the country it sits in, since a unit left on a US 902–928 MHz or wide global profile will typically read Indian-band inlays weakly and erratically — and is also transmitting outside the slice India permits; that the region survived the last firmware update, as some readers revert to a factory default after flashing; that the inlays are tuned for your band; and that power sits inside the legal ceiling once antenna gain and cable loss are counted, because what is regulated is radiated power, not the number on the reader dial.
Never raise transmit power above the permitted limit to chase a read. It is not lawful, and it tends to manufacture stray reads rather than solve the original miss. Our BIS and WPC compliance guide sets out what to ask a vendor for.
Step 3 — the environment: metal, liquid, moisture and bodies
Passive tags are tuned to radiate in free air. Metal reflects and detunes the tag’s antenna; water absorbs the energy. Human bodies are mostly water, so an operator standing between antenna and tag is a real attenuator, not an excuse.
The diagnostic is simple: if the tag reads in free air at your working distance but fails when mounted, the environment is the answer — not the batch, not the reader. Watch for labels applied directly to steel or aluminium, tags on liquid-filled containers, condensation after cold storage, wash-down moisture in soaked cartons, and racking that was empty during the pilot and full in production. Where the surface is the problem, the fix is a tag built for it — an on-metal construction, which typically uses a spacer and a ground plane, as explained in our guide to on-metal tags for tools and assets.
Step 4 — antenna geometry: distance, angle, polarisation and nulls
Read ranges quoted anywhere, datasheets included, are typical figures that depend on the tag, the surface, the antenna, the cable and the room. Bench-test your own combination and design to the distance you actually measure.
- Angle and polarisation. A linear antenna typically reads further but demands consistent tag orientation; rotate the tag ninety degrees and the read can collapse. Circular tolerates any angle at some cost in effective range.
- Gain and beam width. A high-gain antenna concentrates energy into a narrower beam, so it reads the centre of a wide doorway and misses the edges.
- Nulls. Reflections in metal-rich rooms create dead spots. The tell-tale symptom is a tag that reads while moving but not while stationary, or in one position and not ten centimetres away. Reposition the antenna or add a second at a different angle — do not answer a null with more power.
- Cabling. Damaged coax, a loose connector or a mis-mapped port looks exactly like a tag failure. Swap the suspect antenna onto a known-good port.
Our antenna guide on gain, polarisation and placement covers the trade-offs.
Step 5 — reader settings that quietly hide good reads
These raise no error. They simply make real tags absent — and most of what genuinely improves an RFID read rate happens here, not at the tag.
- Session. In S0 tags typically reply repeatedly while energised; in S2 and S3 an inventoried tag goes quiet for a persistence period. If a tag reads on the first pass and vanishes on the second, session is the suspect — not a bad tag.
- Target A/B. A reader locked to target A will not see tags already flipped to B until their state persists back.
- Q value. Set too high for a small tag population it wastes time on empty reply slots; too low for a large one it causes collisions and dropped tags.
- Dwell time and RF profile. With several antennas each gets a slice of the cycle, and too short a slice means the far antenna never completes an inventory round. Faster link profiles trade sensitivity for speed. Increase dwell before you touch power.
Hand-tuning this is slow and hard to repeat. Identium’s ReaderSense reader software approaches it by measurement rather than guesswork: its AISense routine applies candidate configurations to the live reader and compares the results — unique tags seen, read rate, RSSI and per-antenna performance — so you pick a setting on evidence. On a handheld, use RSSI as a hunting tool: walk the tag in and out and watch the value rise and fall to map the real coverage edge.
Step 6 — the software layer: filters, dedup windows and middleware
Some reads reach the reader and never reach your screen. Watch the reader’s raw log rather than your application: if the EPC is in the raw stream, the RF chain is fine and the problem is software. Usual suspects are deduplication windows that suppress repeats for a set number of seconds, so an asset going out and coming straight back looks unread; EPC masks left from an older numbering scheme that silently drop a new batch; reader-side select filters; RSSI thresholds that reject weak reads and take genuine distant ones with them; and business rules that decline to display a “retired” asset.
The opposite problem: stray reads and cross reads
Too many reads is the same physics running the other way, and the cure is physical first. Reduce power within the legal ceiling rather than adding more; re-aim antennas pointing past the zone or into a reflective wall; add downtilt, change height, introduce baffles; move staged stock away from the portal. Only then reach for software — an RSSI floor, direction logic from the antenna sequence, or a minimum read-count rule so a single stray glimpse does not register. Filtering can paper over a badly aimed antenna, but only by discarding genuine reads along with the strays.
When to stop tuning and change the tag
Tuning eventually stops paying. Change the tag when it reads reliably in free air at your working distance but never on the asset, even after antenna, session and dwell changes — the surface has detuned it, and only a different construction fixes that, which is what our on-metal tag range exists for. Change it when reads are position-critical, one orientation working and nothing else. Change it when performance degrades in service through moisture ingress, flexing, heat or adhesive failure — a housing problem, not a settings one. And change it when you had to push power to the regulatory ceiling for marginal reads: with no headroom, an ordinary bad day becomes an outage.
We will not promise a fix or a percentage. What a better-matched tag buys is margin. If you are re-specifying rather than repairing, our guide to scoping read range against the environment is where to restart.
Identium manufactures UHF tags and readers in New Delhi, does its own ultrasonic sealing, epoxy potting and encoding, and typically turns custom hard tags around in 5–10 days. Tell us how the tag is mounted, which reader you are using and how it is configured, and we will tell you which layer you are most likely stuck in — get in touch.