Two Modbus RTU masters on one RS-485 bus: why and the fix

Why one RS-485 segment takes exactly one Modbus RTU master

RS-485 lets many transceivers share one twisted pair, but the electrical standard says nothing about turn-taking: any transceiver may drive the wires at any instant. Modbus RTU fills that gap with the single-master rule. The master issues every request; a slave transmits only in reply to a frame that addresses it.

Two devices that both initiate requests poll on their own timers, so sooner or later their frames overlap on the wire. Overlapping bits corrupt both frames, the CRC check rejects them, and the master that was waiting for a reply times out. A write caught in the overlap is simply lost, which is how an operator command reaches the drive on one press and vanishes on the next. A timetable in which each device transmits only occasionally is not arbitration: a poll timer runs whether or not anyone is touching the screen, so an idle second master stays a hazard.

A few fieldbuses do share a segment between several masters by passing a token around the ring. DH-485 on its own network works that way, and connecting an operator interface to a PLC over DH-485 shows what that stack looks like. Modbus RTU carries no token and no priority scheme, so the one-master rule stands.

Symptoms and what each one points to

• The drive obeys start and speed commands most of the time, then ignores one. A collided write never reached the drive intact. Note whether the misses cluster while both devices are powered and polling.
• CRC errors and no-reply timeouts on the drive link, worse while the screen is in use. The classic two-master signature: operator action adds a second transmitter to a bus that already has one. If timeouts persist on a clean single-master bench setup, look elsewhere; Modbus RTU master no-reply and address errors covers those causes.
• Works while the panel sits on its idle page, fails when an operator presses a command button. The HMI transmits on operator action and lands mid-frame inside the PLC poll cycle.
• Passed a bench test on short leads, fails on the installed run. That points at cable distance, not at the protocol. Check the interface type against the numbers further down.

Checks before you rewire

1. Find today's master. Watch the transmit LEDs on the drive link for a minute. A device that blinks while nobody touches it is polling, and its project will contain a driver aimed at the drive slave address.
2. Read both projects. Write down the port, slave address, and poll interval of every driver on the segment. Two drivers aimed at one drive address are the collision, in writing.
3. Check the spare ports. Read each device's manual for a second serial port and the electrical standard it supports. A port label alone says nothing; the port specification page decides.
4. Measure the cable route. Walk the actual path, add vertical drops and service slack, and keep the total. That number decides the interface on the HMI link.

The fix: one master, two links

Drive segment. The PLC becomes the only Modbus RTU master on the RS-485 pair to the drive. Wire the segment as a daisy chain, fit 120 Ω termination at both physical ends to match the twisted pair's characteristic impedance, and add bias resistors, typically 10 kΩ pull-up and pull-down, so the pair sits at a defined voltage when nobody transmits.

HMI link. Connect the HMI to a second port on the PLC and point the HMI driver at that PLC link, never at the drive. The two buses share no wire, so they can no longer collide.

Registers. Keep a small block of PLC holding registers as the operator interface:

• Command registers, written from the HMI: a command word plus the setpoints an operator may change. PLC logic checks range and interlocks before anything leaves for the drive.
• Status registers, filled by the PLC from values it reads back from the drive: actual speed, fault codes, link health. The screen shows what the drive reports, never an echo of its own command. A value that spans two registers, such as a frequency stored as a 32-bit float, needs its byte order handled once in the PLC; display a PLC float stored in two word registers on an HMI walks through that.

Recovery. When the drive stops answering, the PLC retries a fixed number of times, raises a fault, and holds the last valid status. Set a command-accepted flag only when the drive confirms. The recovery sequence itself, including when to re-open the link, is laid out in make a Modbus master reconnect after a slave outage.

Latency. An operator command now travels one HMI write interval, one PLC scan, and one drive poll cycle before it reaches the drive. Time that chain on the running system; if the process needs it faster, raise the poll rate on the drive segment.

Test the handover with the drive in a safe state before production: a command path that used to drop frames now delivers every one, so verify what each screen button actually does.

How far each serial interface reaches

RS-232, the port most panels and small PLCs carry first, is short-range. The recommended planning figure is about 50 ft with standard cable, a number tied to cable capacitance at data rates of 19.2 kbps and below. Under ideal conditions the same guidance allows toward 100 ft, and you treat that as a ceiling, never a target. Special low-capacitance cable can extend a link toward 1,000 ft, but count on a distance only when the cable's own datasheet states it.

RS-422 and RS-485 are built for distance: around 4,000 ft at low data rates is the usual figure from the electrical standard, and repeaters extend it further.

The thresholds fall out of those numbers. Inside the recommended span, RS-232 is fine. Between the recommendation and the best case, RS-232 survives only with verified cable. Beyond the best case, plan RS-485. For a worked example of moving a panel-to-PLC link from RS-232 to RS-422/485, this RS-232 to RS-422/485 panel link note walks the interface swap.

Worked example: an HMI 100 ft from the PLC

Inputs, and where each comes from:

• HMI-to-PLC run: 100 ft. The case that triggers this question: the PLC sits more than 100 ft from the panel, so 100 ft is the best case.
• RS-232 recommended span: 50 ft with standard cable, at data rates of 19.2 kbps and below. RS-232 distance guidance.
• RS-232 best case: 100 ft under ideal conditions. Same guidance, treated as a ceiling.
• RS-422/RS-485 reach: 4,000 ft at low data rates, more with repeaters. The usual figure from the electrical standard.

Rule in words: an interface suits a fixed run when its recommended span clears the measured distance with margin to spare. A run that sits at the interface's best-case figure has nothing left when the cable is ordinary and the plant is electrically noisy.

Arithmetic:

• RS-232 against its recommendation: 100 ft - 50 ft = 50 ft over the line. Fails.
• RS-232 against its best case: 100 ft - 100 ft = 0 ft of margin. No production link runs at zero margin.
• RS-485 against its reach: 4,000 ft - 100 ft = 3,900 ft of margin, and 100 / 4,000 = 0.025, so the run uses 2.5% of the interface's budget.

Result: RS-232 fails both tests. RS-485 passes with 39 times the run length in spare reach (3,900 / 100 = 39).

Decision: put the HMI-to-PLC link on RS-485, or RS-422 where the ports offer it, terminate both ends, and leave the drive segment on its existing RS-485 pair. The threshold to carry forward: above 50 ft, specify RS-485; between 50 and 100 ft, RS-232 is best-case only; beyond 100 ft, RS-232 is out.

The decision for each case

• The HMI is master today and the PLC must also command the drive. Move the master role to the PLC, add a second PLC port for the HMI link, and route commands through the register block above.
• The PLC is already master and an HMI driver also polls the drive. Delete the drive driver from the HMI project and point the HMI at PLC registers. The collision ends with no new cable.
• The HMI ships a Modbus RTU slave driver. Keep one cable: the PLC master addresses drive and panel as slaves on one segment, and operator entries sit in the panel slave registers for the PLC to read. Many panels ship master drivers only, so check the driver list before planning this variant.
• The HMI-to-PLC run exceeds 100 ft on RS-232. Replace that link with RS-485 per the numbers above. The drive segment is unaffected.

None of this changes the machine HMI, which stays the operator's interface to the controls. If the floor also needs a status board for people walking past, that is a separate screen outside the control loop: our shop-floor screens run 12 to 36 V DC, are rated from -10 to 60 °C, carry an IP65 front face, and play offline content or your own Android app.

Recommendation: make the PLC the only Modbus RTU master on the drive segment, connect the HMI to a second PLC port over RS-485, and pass every drive command through PLC registers the PLC validates and forwards at its next scan. Specify RS-485 for any serial run beyond 50 ft and treat 100 ft as RS-232's ceiling. Use the single-cable variant only when the HMI's driver list includes a Modbus RTU slave.

Questions