RS485 to LoRaWAN Converter Not Sending Data? Common Problems and How to Fix Them
Short Answer
“Not sending data” is four different faults, and the network server tells you which one in under a minute. Open the device’s page and look at what has arrived:
| What the server shows | Layer | Where the fault is |
|---|---|---|
| No join, or join requests only | 1 | Radio, keys, band, gateway |
| Uplinks with empty, zero or 0xFF fields | 2 | Interval, payload size, duty cycle, counters |
| Uplinks with empty, zero or 0xFF fields | 3 | RS485 bus or instrument |
| Uplinks with wrong numbers | 4 | Register map or decoder |
Layer 3 is the one people mean when they say the converter isn’t sending: the radio is fine, and the instrument isn’t answering. Nine times in ten the cause is in the first six items of the Modbus checklist below: A/B polarity, ground, parity, baud, slave ID, and the response timeout.
The sections below work through each layer in order, with the one tool that shortens every diagnosis: a laptop with a USB-RS485 adapter.
First, Read the Payload
Before changing anything, decode what the converter actually sent. Network servers show the raw payload in hex or base64; convert it and look at the field bytes.
- Fields full of
FF(base64 strings ending in//////) usually mean the converter timed out waiting for the instrument and filled the field with a marker. - Fields full of
00can mean the same, or an instrument that answered with zeros. - A very short payload (
AQ==is a single byte01) means the converter sent its header with no fields: it got nothing back on the bus. - A payload of normal length with plausible-looking but wrong numbers is Layer 4.
Check the converter’s manual for its timeout marker. [VERIFY: how the LX1 flags a timed-out field in the uplink.] The decoded object in ChirpStack or the live-data view in The Things Stack shows the same thing after the decoder; a decoder that reports warnings for missing fields (as the one in How to Decode LoRaWAN Payloads from Modbus Devices does) tells you directly.
Layer 1: It Hasn’t Joined
Join requests without a join accept, or nothing at all, is the radio side, and it has nothing to do with the instrument. The checklist, in order of frequency:
- Keys. One wrong character in the DevEUI, JoinEUI or AppKey, or EUIs entered in the wrong byte order. Copy from the app or scan the QR code; never retype.
- Band. IN865 on the device, the gateway and the server’s frequency plan. An EU868 unit will never join an Indian network.
- Gateway. Online on this server, and in range of the panel. A panel in a metal room with the converter’s antenna inside it is the usual range problem; mount the antenna outside or use an external one, and check the signal in the app before closing the door.
- LoRaWAN version and class in the device profile.
- Join back-off. After many failed joins the device slows to one attempt every few hours; power-cycle it once the server is fixed.
The full join checklist is in How to Configure a LoRaWAN Sensor and the signal test in Built-In LoRaWAN RF Testing.
Layer 2: Joined, but No Uplinks Arrive
The join accepted, and then silence. Less common, and usually one of these:
- The reporting interval is long and you haven’t waited. Trigger an uplink from the app rather than waiting an hour.
- The payload is too big for the data rate. At SF10–SF12 on IN865 the application payload limit is 51 bytes. A field list that fits at SF7 may not fit when ADR moves the device to SF11 in a poor-signal panel; the converter either splits it, truncates it, or fails to send. Trim the fields or fix the signal.
- Duty cycle. A converter polling every minute with several uplinks per cycle at a slow data rate can hit India’s 2.5% device airtime limit and be held off; see LoRaWAN in India: IN865 and WPC.
- Frame counter rejection. A device reset after a long run, especially on ABP, sends counters the server has already seen and the server discards them. Rejoin (OTAA) or reset the counters on the server.
- Gateway went offline after the join. Check its last-seen time; see What Happens When a LoRaWAN Gateway Goes Offline?
Layer 3: Uplinks Arrive, Fields Are Empty: The Modbus Checklist
The radio works and the instrument isn’t answering the converter. Work this list in order; stop when the field fills.
| Check | How |
|---|---|
| Laptop test | Poll the instrument with a USB-RS485 adapter on the same cable |
| A/B polarity | Swap A and B; labels vary by maker |
| Signal ground | Connect GND to GND; not just the shield |
| Parity | None, even or odd; many meters default to even |
| Baud rate | Match the instrument; 9,600 is common, not universal |
| Slave ID | Read it from the instrument’s menu or DIP switches |
| Stop bits | 1 is usual; some instruments use 2 with no parity |
| Response timeout | Raise to 500–1,000 ms for slow instruments |
| Sensor power | Lengthen warm-up if the converter powers the instrument |
| Termination | 120 Ω at each end of a long run |
| Function code | FC03 vs FC04; check the exception code |
| Register address | 40001 is protocol address 0 |
| Register count | Some instruments allow only a few per request |
| Second master | Disconnect any PLC or SCADA on the same bus |
| Instrument mode | Local/remote, comms enabled, registers unlocked |
The first item is the one that saves the afternoon. If the laptop, on the same wire, gets the right values, the instrument and the wiring are fine and the fault is in the converter’s settings; if it doesn’t, the converter is innocent. The method is in How to Connect a Modbus RS485 Sensor to LoRaWAN.
Some of these deserve a paragraph.
Polarity and ground. A and B are swapped on more installations than anyone admits, because manufacturers disagree on which terminal is which. If a swap doesn’t fix a timeout, add the ground: on long runs, a missing signal ground lets the common-mode voltage drift outside the transceiver’s range and the link fails intermittently or completely, while a laptop two metres away works fine.
Parity. Meters and flow meters from European makers often default to even parity; a converter left at none will never see a valid frame. One real case: a CS Instruments flow meter at 19,200 bps, even parity, that returned timeouts until the converter’s parity matched.
Timeout and warm-up. The converter waits a fixed time for the reply. Multifunction meters answer in tens of milliseconds; some sensors and controllers take 300 ms or more, and an instrument powered from the converter’s switched output needs its warm-up (the LX1 allows 0 to 2,000 ms) before the first request. Too short a timeout or warm-up looks exactly like a dead bus.
Locked and conditional registers. Instruments can be answering correctly and still return nothing useful. The ABB meter case above involved a feature that blocks the energy registers until the meter has recorded 1 kWh; reading live voltage instead of total energy showed the bus was fine. VFDs ignore reads of some parameters until set to remote mode; some meters need a password register written first. Read the manual’s notes column, and test a register that can’t be locked (a voltage, a serial number) to prove the link.
Two masters. The most under-diagnosed fault on brownfield sites. The converter is wired onto a bus that a PLC or SCADA already polls. Two masters collide; both see intermittent timeouts; the SCADA engineer blames the new box. There can be one master per RS485 bus. Either the converter becomes the only master and feeds SCADA through a register image (the pattern in LoRaWAN to SCADA, PLC or BMS via Modbus TCP), or the instrument’s second port, if it has one, is used.
Unsupported protocol. A converter that speaks Modbus RTU can’t read Modbus ASCII, Modbus TCP over a serial gateway, or a proprietary protocol; the LX1 is RTU-only. Non-Modbus serial devices need transparent mode or a different product; see Transparent Mode vs Polling Mode.
Reading Modbus Exception Codes
If the converter reports the instrument’s reply rather than a timeout, an exception code tells you exactly what the instrument objected to. The function code comes back with its high bit set (0x83 for a failed FC03) followed by the exception:
| Code | Meaning | Fix |
|---|---|---|
| 1 | Illegal function | Use FC03 or FC04 as the manual says |
| 2 | Illegal data address | Fix the register offset (40001 → 0) or count |
| 3 | Illegal data value | Count too large, or a write out of range |
| 4 | Slave device failure | Instrument fault or locked state |
| 6 | Slave device busy | Poll less often; lengthen timeout |
An exception is good news in one sense: the wiring, baud, parity and slave ID are all correct, because the instrument understood the request well enough to reject it.
Layer 4: Uplinks Arrive with Wrong Numbers
The bus works; the interpretation doesn’t. Values 10× or 100× out are scaling; 32-bit values that are huge or random are word order; a power factor of 64,551 is a signed value read as unsigned; a voltage appearing in the current field is field order or header length; everything shifted by one register is the 40001-versus-0 offset. Each fix is in the decoder post and in the step-by-step guide; the rule is to make the converter and decoder match the laptop test, never to adjust the decoder until the numbers look right.
One further Layer 4 case: frozen readings. The dashboard updates but the values never change. Either the instrument stopped answering and the converter is re-sending its last good values (check the timestamp and any timeout flag), or the register being read is a configuration value rather than a measurement.
Intermittent Faults
Data that comes and goes is almost always one of four things.
- Polling too fast. A converter asked to read many registers every few seconds builds a queue it can’t drain; responses overlap and the uplinks carry corrupt or variable-length data. In one documented case a flow meter’s 29-byte response arrived as 14 or 58 bytes until the polling interval was lengthened. Poll at 30 seconds or more per instrument, and let the LoRaWAN interval be longer still.
- Noise. An unshielded pair run alongside VFD or contactor cables, or an unterminated long bus, corrupts frames at random. Shielded twisted pair, earthed at one end, terminated at both ends, routed away from power.
- Marginal radio. The converter joins and reports from a panel until the door is closed or a truck parks beside it. Check RSSI and SNR on the uplinks; a margin under 10 dB will fail in the monsoon. See the placement guide.
- Power. A converter on a panel’s control-circuit supply drops out when that circuit is isolated, or browns out when a contactor pulls in. Give it a fused auxiliary supply.
The Fifteen-Minute Protocol
When a converter “isn’t sending data”, do these in order and stop when it works:
- Read the server. Join? Uplinks? Empty fields or wrong numbers? That picks the layer.
- Decode the last payload and look for timeout markers or missing fields.
- Poll the instrument from a laptop on the converter’s own cable. Record the settings that work.
- Compare those settings with the converter’s: baud, parity, stop bits, slave ID, function code, address, count.
- Swap A/B, then add ground, if the laptop also fails.
- Raise the timeout and warm-up, and poll one register that can’t be locked.
- Disconnect any other master from the bus.
- Check the signal from the panel with the door closed.
- Reload the decoder matching the firmware version, and test with known bytes.
- Escalate with the raw hex of three uplinks, the converter’s configuration export, the instrument’s register map and the laptop-test result. Support can solve almost anything from those four things, and almost nothing without them.
Mistakes We See
Changing five settings at once. Then nothing can be learned from the result. One change, one test.
Blaming the radio for a bus fault. The server shows the uplinks arriving; look at their contents.
Skipping the laptop test. It separates instrument from converter in two minutes.
A second master on the bus. There can be only one.
Reading a locked register to prove the link. Prove it with a voltage first.
Polling every five seconds. The instrument and the converter both need breathing room; the radio needs far less data than that anyway.
Help When the Checklist Runs Out
Macnman designs and manufactures LoRaWAN devices in India, and supports them from Pune. The MacSync LX1 RS485 Modbus to LoRaWAN converter is configured and tested in the field through the Maya app, with a per-request timeout, a switched sensor-power output with warm-up delay, a signal check at the panel, and decoders with field-level warnings in the MacnMan LoRaWAN decoders repository. It pairs with the Setu LX indoor and MacLink Gen-2 outdoor gateways, whose on-board ChirpStack shows every join and uplink. How it compares with other converters on these points is on the versus page, and the selection criteria are in How to Choose an RS485 to LoRaWAN Converter.
Stuck at step 10? Send the raw hex of three uplinks, the converter’s configuration, the instrument’s register map and what the laptop test returned, and a product engineer will diagnose it within one working day: contact Macnman. For a new site, the pilot kit ships a converter and gateway pre-registered to each other so the first panel comes up without the radio layer in doubt.
Frequently Asked Questions
Why is my RS485 to LoRaWAN converter not sending data?
Check the network server first. No join means a radio, key or band problem; joined with no uplinks means interval, payload size or counters; uplinks with empty fields mean the instrument isn’t answering on the RS485 bus; uplinks with wrong numbers mean the register map or decoder. Most cases are the third: A/B polarity, ground, parity, baud, slave ID or timeout.
My converter joined fine but every field is 0xFF or empty. Why?
The converter is timing out waiting for the instrument. Poll the instrument from a laptop on the same cable to confirm the settings, then check A/B polarity, signal ground, parity, baud, slave ID, the response timeout and the sensor-power warm-up, and make sure no other master is on the bus.
Does swapping A and B really matter?
Yes. Manufacturers label the pair inconsistently, and a swapped pair produces no valid frames at all. It’s the first physical check after the laptop test.
What does Modbus exception code 02 mean?
Illegal data address: the register address or count isn’t valid on the instrument. Usually the 40001-versus-0 offset, or reading past the end of a block. It also proves the wiring and serial settings are correct.
Why does my meter answer the laptop but not the converter?
A settings mismatch (parity, stop bits, timeout), a warm-up too short on switched power, a function code or count the instrument rejects, or a second master on the bus that interferes only when the converter polls on schedule.
Can two devices poll the same RS485 bus?
No. RS485 Modbus allows one master. A PLC or SCADA already polling the instruments must be disconnected or fed from the converter’s data instead.
Why are my readings corrupt or variable in length?
Polling too fast. The converter’s queue overflows and responses overlap. Lengthen the polling interval to 30 seconds or more per instrument and keep the LoRaWAN uplink interval longer still.
Why do uplinks stop when I close the panel door?
The steel enclosure blocks the 865 MHz signal. Mount the converter or its antenna outside the panel and check RSSI and SNR with the door closed.
Why are my values wrong by a factor of 10 or 100?
Scaling. Apply the instrument’s divisor per field in the converter or decoder. Random or huge 32-bit values are word order; 65,000-ish values are signed read as unsigned.
What should I send to support?
The raw hex of three uplinks, the converter’s configuration export, the instrument’s register map, and the result of the laptop poll. With those four, most faults are diagnosed on the first reply.
