Troubleshooting
Choosing the RS-485 protocol for one HMI and multiple PLCs
Short answer
Count the conductors before you open the HMI driver. On a 2-wire RS-485 DL-06 Port 2, Modbus RTU is the only protocol the port runs; rewire as 4-wire and K-sequence, DirectNET and Modbus RTU all open up, Modbus RTU fastest of the three. Give every PLC a unique station address, match baud and framing at both ends, and bring PLCs onto the bus one at a time.
On this page
Ask what the wiring is before you ask what the protocol should be. On the DL-06, Port 2 is the multi-function port — RS-232C, RS-422 or RS-485 on one 15-pin connector — and the documentation ties its protocol list to the way the port is wired. 2-wire RS-485 runs Modbus RTU and nothing else. 4-wire RS-485 opens Modbus RTU, DirectNET and K-sequence, with Modbus RTU the fastest of the three. That is the whole decision for one HMI polling several PLCs: count conductors, pick the driver the wiring allows, give every PLC its own station address, and set identical baud and framing at both ends. A short run and a slow application are on your side — the worked example below puts a number on how fast a 9600-baud 2-wire network can poll them.
The wire count decides the driver #
The DL-06 manual gives Port 2 three serial protocols: Modbus RTU, on which the port acts as master or slave; DirectNET, the protocol the DL-06 serial communications appendix lists as non-sequence; and K-sequence, which Port 2 runs with the PLC answering as a slave. Its RS-485 mode comes in two electrical layouts: 2-wire, one differential pair shared by both directions, and 4-wire, separate transmit and receive pairs like RS-422. The documentation is explicit about where the lists meet: 2-wire RS-485 on the DL-06 is Modbus RTU only. The three-protocol list belongs to 4-wire wiring, where the HMI does the polling and every PLC answers.
That constraint is electrical before it is a software menu. A 2-wire segment shares one pair for both directions, so the driver must turn the line around and release it between frames; the two extra conductors of 4-wire wiring give each direction its own pair. If the cabinet is already wired 2-wire, the decision is made for you: run Modbus RTU. If you are pulling new cable, 4-wire buys the protocol choice, and laying out a multidrop of DirectNET PLCs is its own design task — the guide to networking DirectNET PLCs and multiple HMIs covers that side.
| Port 2 wiring | Protocols it runs | One HMI, several PLCs |
|---|---|---|
| 2-wire RS-485, one shared pair | Modbus RTU only | Works; the HMI is the single master |
| 4-wire RS-485, separate TX and RX pairs | Modbus RTU, DirectNET, K-sequence | All three work; Modbus RTU is the fastest |
Do not pick a driver because it appears in the HMI software's device list — the port's wiring and mode decide, and a PLC simply ignores a protocol it is not running. Control means data here: every one of these protocols reads and writes the PLC's registers and bits from the panel, and none of them downloads or edits the ladder program — why Modbus, K-sequence and DirectNET cannot write ladder covers that boundary. If the plan needs the program changed from the panel, no protocol on this port does that.
Working example: poll time for three DL-06s at 9600 baud #
Inputs. Three DL-06 PLCs and one HMI on a 2-wire RS-485 segment — the layout in question, which pins the protocol at Modbus RTU. Baud 9600 with 8 data bits, even parity, 1 stop bit — assumed example settings and a common serial default, so read the real values off both ends before trusting the result. Each poll reads 10 holding registers per PLC — assumed, sized like a screen of setpoints and statuses. From the serial format: 1 start bit + 8 data bits + 1 parity bit + 1 stop bit = 11 bits per character. From the Modbus RTU framing rule: at least 3.5 character times of silence before and after every frame. From the Modbus RTU frame layout: a read-holding-registers request is 8 bytes (slave address, function code, start register, quantity, 2-byte CRC) and the response is 5 + 2 × register count bytes (address, function, byte count, data, CRC).
The rule in words. Seconds per poll equal the payload bits — request plus response bytes, times 11 bits per byte — plus the two silent gaps at 3.5 characters × 11 bits each, all divided by the baud rate. Multiply by the PLC count for a full cycle.
The arithmetic.
- Response: 5 + (2 × 10) = 25 bytes, so request plus response = 8 + 25 = 33 bytes.
- 33 bytes × 11 bits = 363 bits on the wire.
- Silence: 2 × 3.5 × 11 = 77 bits.
- 363 + 77 = 440 bits per poll.
- 440 ÷ 9600 = 45.8 ms per PLC.
- Three PLCs: 3 × 45.8 ≈ 138 ms per full cycle, so every PLC is refreshed about seven times per second (1 ÷ 0.138 ≈ 7.2).
What moves it. Twenty registers per poll instead of ten puts the response at 45 bytes and the poll at 53 bytes × 11 = 583 bits, + 77 = 660 bits, or 68.8 ms per PLC — a 206 ms cycle, half again as long. Doubling the baud to 19200 halves the cycle to about 69 ms.
The decision. For a slow application, 2-wire Modbus RTU at 9600 baud is already enough: a full refresh of all three PLCs takes about 138 ms. The cycle crosses half a second only at roughly eleven PLCs at 10 registers each, or seven PLCs at 20 registers each — move to 19200 baud before touching the wiring. Rewire to 4-wire for protocol choice, not for speed: at equal baud, Modbus RTU is already the fastest protocol the port offers.
Give every PLC its own station address #
Two slaves set to one station number answer the same poll at the same moment and both replies are lost, which shows up later as random dropouts rather than as an error. Assign 1, 2, 3 in a fixed order and keep the map — HMI device entry, station number, machine — written down at the panel. In the HMI project each PLC is its own device entry carrying its own station number; a copy-pasted device entry with an unedited station number is the classic commissioning fault.
Verify mapping separately from the link. Write a known value to a scratch register from the HMI and read it back in the ladder instead of trusting tag names. For K-sequence and DirectNET the PLC-side data lives in V-memory, and the K-sequence and DirectNET slave V-memory configuration guide shows how that mapping is laid out.
Set identical serial settings at both ends #
Protocol name, baud, data bits, parity and stop bits — five lines that must read the same in the HMI device entry and in the PLC's port configuration. A mismatch never establishes a link, and it looks exactly like broken wiring: the receiver hears bytes but never one valid frame. So read the values off both ends before changing anything, and change them at both ends together; dropping the baud on one side alone, or stepping down to 4800 because 9600 feels wrong, burns an hour and proves nothing. The example above is the sanity check — at matched settings the math predicts the reply rate, and replies arriving near the predicted pace mean the link is done and the faults left are addresses and mapping.
Commission one PLC at a time #
Put the machine in a safe state before you change port protocols, station numbers or wiring — a mid-run change can drop every HMI write at once. Then work the list:
- With power off, count the conductors you actually landed against the DL-06 serial wiring appendix, pair polarity included — a swapped differential pair shows up as no communication, not as a message.
- Set PLC 1's Port 2 to the wiring you have — 2-wire RS-485 and Modbus RTU here — plus its station number, baud and framing.
- Create the matching device entry in the HMI project: same protocol, same serial settings, station 1.
- Connect PLC 1 alone and confirm reads plus one test write to a scratch register.
- Add PLC 2 on station 2 and retest both, then PLC 3 on station 3 and retest all.
- Finish with representative writes to each controller, read back in each program, before calling the network done.
Follow the termination and biasing instructions in the DL-06 serial documentation rather than adding resistors on instinct; on a 10 m run a gratuitous 120 ohm resistor is more often a new fault than a fix.
What the symptoms tell you #
- No PLC responds. The fault sits in the shared items: driver protocol against port protocol (a K-sequence driver on a 2-wire port never talks), serial settings, port mode, pair polarity. The fix for HMI Modbus device no response over RS-485 walks these causes end to end.
- One PLC answers, the rest are silent. Test each silent PLC on its own with the same cable; its station number, port mode or drop wiring is the suspect.
- Every PLC works alone, the bus fails when all connect. Duplicate station addresses; check the written map against each PLC.
- Replies are stable but values look wrong. The link is fine and the driver mapping or data type is not; write a known value from the HMI and read it in the ladder.
- All PLCs answer but the screens lag. Compare the observed refresh with the poll-time math above; if the wire is not the bottleneck, the software-side causes sit in why an HMI freezes or updates slowly over serial.
The decision, one last time: count conductors first. 2-wire RS-485 pins Port 2 to Modbus RTU — unique station numbers, matched settings, PLCs added to the bus one at a time. 4-wire RS-485 opens K-sequence and DirectNET alongside it, and Modbus RTU remains the fastest pick of the three. A 10 m, slow-polling multidrop does not argue with any of it.
Questions asked
Which Modbus RTU station addresses can I assign to the PLCs?
Modbus RTU slave addresses run from 1 through 247. Address 0 is broadcast: a write to 0 goes to every slave at once and collects no reply, so never assign it to a PLC. For three DL-06s, stations 1, 2 and 3 are fine. Write the map down next to the machine, because the duplicate-address fault never announces itself.
Can a second HMI join the same RS-485 segment?
Keep one master on the segment. Modbus RTU slaves only answer polls; two masters polling the same wire will talk over each other and corrupt replies. If a second station needs the same data, put a serial-to-Ethernet converter or a gateway at the panel and let the second display poll through the plant network, leaving the RS-485 segment with a single master.
Does a 10 m run need termination resistors?
Usually not. At 10 m and these baud rates the signal round trip settles well inside one bit time, so reflections do not corrupt the sampling, and every resistor you add loads the transceivers. Terminate and bias the way the DL-06 serial wiring diagram shows; if you measure errors on the bus, start with pair polarity, shielding and station settings before reaching for resistors.
Why do updates drop out instead of showing an error?
Modbus RTU protects each frame with a 16-bit CRC. A receiver whose CRC check fails simply discards the frame, so noise, a marginal connection or a framing mismatch shows up as missing or stale values, not as a message on the screen. Count the discarded frames if your HMI diagnostics offer it; a count that climbs when a motor or contactor runs nearby points at shielding and routing first.