When every PLC serial port and rack slot is taken, the third operator panel still has two clean ways in: free one port with a module swap or remote I/O move, or put a gateway on the existing line so the gateway alone holds the master role. Never wire a second polling master onto an RS-422 pair, and keep any shared RS-485 line inside 32 unit loads and about 1,200 m at 100 kbps.
Every serial port is taken, the rack has no free slot, and a third operator panel still needs its data. One master per serial line is the rule that decides the job. RS-422 gives each pair a single driver, and polling masters on an RS-485 pair have no way to take turns, so a second master panel on an existing line fails no matter how you cable it. The clean paths are to free one serial port, or to put a gateway on an existing line so the gateway alone holds the master role and the new panel hangs off it. The RS-485 numbers to keep in view are 32 unit loads per segment and about 1,200 m (4,000 ft) at 100 kbps. Here is the decision, with the arithmetic.
One master per serial line, and why
RS-422 and RS-485 look alike in the cabinet: same screws, same twisted pair, same little DIP switches. The roles they allow are different.
RS-422 is full duplex. Each direction has its own pair, and EIA/TIA-422 gives each pair one driver with up to 10 unit-load receivers. Two master panels would both have to drive the pair that reaches the controller's receiver. Two drivers on one pair is outside the standard, and no daisy-chain order fixes it. One safety note: when a port is committed to a controller-to-controller link, leave it alone until the change is approved, because that exchange feeds processes you cannot see from the panel.
RS-485 is half duplex on one pair, and its electrical standard, TIA/EIA-485, allows several tri-state transmitters on that pair. Who transmits is decided by the protocol, not the wire. Modbus RTU defines one master per segment. Ordinary operator panels are polling masters: each transmits on its own clock, neither listens for the other's turn, and overlapping frames are lost. Two such panels on one pair produce intermittent, timing-dependent faults that get worse as poll rates rise.
A few protocols were designed for several nodes on one network from the start, and DH-485 is one of them, which is why connecting an operator interface over DH-485 works as a shared network. If your panels and controller speak only a single-master protocol, sharing needs a gateway, not a clever cable.
What the RS-485 numbers let you count on
TIA/EIA-485 defines the unit load and requires every transmitter on a segment to drive 32 of them. A reference receiver is 1 unit load; the counting comes from the standard's own unit-load definition. Modern transceiver chips draw 1/4 or 1/8 of a unit load, so a segment built with them carries 128 or 256 receivers inside the same drive budget. When a segment is genuinely full, a repeater joins two loaded segments into one network.
The length and speed pairing most vendors quote is 1,200 m (about 4,000 ft) at 100 kbps. Panel lines usually run 9.6 to 38.4 kbps, so the same cable has margin at that length. The point that decides this job is in the counting itself: 32 counts electrical load, not permission to transmit. A line well under 32 unit loads still fails if two polling masters transmit together, because the electrical standard has no referee. Sum the loads to size the segment; settle the master question in the protocol.
The four ways to add the panel
Compare the routes against three constraints: free rack slots, free serial ports, and the role the panels can play on the wire.
| Route | What it needs | Where it works | The catch |
|---|---|---|---|
| Free a serial port | a dual-port communications module in place of a single-port one, or I/O moved to a remote block | a slot can be emptied without losing required I/O | a module and engineering time; no new roles anywhere |
| Share one RS-485 line | the controller as the only master, panels that accept a slave role | 32 unit loads per segment; 1,200 m at 100 kbps leaves room | master-only panels cannot join, and Modbus RTU allows one master |
| Serial multiplexer or RTU-to-TCP gateway | panel-side ports on the gateway, one gateway-owned master port on the controller | ports full, slots full, panels are polling masters | the gateway must speak the panel protocol on one side and the controller protocol on the other |
| Ethernet upgrade | an Ethernet port on the controller or a communications module, panels that speak Ethernet | the system is being reworked anyway | usually takes a slot and a project |
Still choosing the panels themselves? Give real weight to slave-capable multi-drop modes now, because that is what makes the shared-line row available later; planning a touchscreen operator interface for factory controls is where that choice belongs.
The gateway row is the serial cousin of a decision that comes up over Ethernet too, when several PCs poll one controller: direct, or through a relay.
Example: a third panel on a two-port controller with a full rack
Scenario. A controller has two serial ports. Port 1 feeds panel A, a master polling at 19.2 kbps, 8 data bits, no parity, 1 stop bit. Port 2 feeds panel B, set up the same way. The processor's remaining port runs a controller-to-controller link. All rack slots are occupied. Panel C is wanted on the line, and all three panels are master-only.
Step 1, port. A dual-port communications module needs a slot; the rack has none, and moving the discrete I/O to a remote block is a shutdown-length job. Set it aside unless the next planned outage covers it.
Step 2, share. Panels A, B and C all poll as masters. RS-422 forbids two drivers on one pair outright, and RS-485 gives two polling masters no way to take turns. Rejected before any cable is cut.
Step 3, gateway. Pick a gateway with two independent panel-side serial ports and one controller-side port, speaking the panel protocol toward B and C and the controller protocol on the uplink. Move panel B's cable from port 2 to the gateway's uplink; wire B and C to the gateway's panel-side ports. The gateway alone holds the master role on port 2, and the controller port keeps the slave role it already had. Panel A is untouched.
Check the electrical budget on port 2: gateway transmitter 1 unit load, controller receiver 1 unit load, 2 of 32, on the original few-meter cable. Now the wire time.
At 19.2 kbps with 8 data bits, no parity and 1 stop bit, a byte is 10 bits. A 10-register Modbus RTU poll is an 8-byte request and a 25-byte response: 33 bytes = 330 bits, plus two 3.5-character frame gaps = 70 bits, about 400 bits per cycle. 400 bits / 19,200 bps = 20.8 ms, call it 21 ms.
Panel B polls 10 registers every 250 ms: 4 polls/s x 21 ms = 84 ms of wire time per second. Panel C is specified the same way, another 84 ms/s. Together 168 ms/s, about 17% of port 2. The port has room.
If C were instead specified for 20 registers every 100 ms, its cycle would be 8 + (5 + 2 x 20) = 53 bytes = 530 bits, plus 70, about 600 bits = 31 ms, and 10 polls/s would use 313 ms/s. With B's 84 ms/s the port sits just under 40%, and a fourth panel of that appetite saturates it. That summed wire time is the number to check before you order.
Latency: C's answer waits behind at most one of B's cycles already in flight, about 21 ms, plus the gateway's own translation time, which its documentation prints.
Decision: the gateway keeps every master-slave pair legal, uses no slot, and leaves panel A alone. If the next outage frees a slot anyway, the dual-port module is the simpler long-term layout: one master per port, nothing aggregated.
Checks before you order
Ask the gateway maker three things and write the answers down:
- Panel side: which protocols the panel ports accept, how many independent ports, and whether concurrent polls from the panels are queued or refused.
- Controller side: which protocol and role the uplink presents; it must appear as the single master the controller port expects.
- Timing: the added latency per request, in milliseconds, from their documentation.
Then do the wire-time sum from the example and keep it under half the port's capacity, so retries and a faster poll rate have somewhere to go. Terminate each RS-485 segment at both ends in the cable's characteristic impedance, 120 Ω for the twisted pair normally used on these lines.
Commission one link at a time. After the change, values that arrive late or drop intermittently point at an overloaded aggregate link, the same family of fault as dropped PLC-triggered messages on a serial HMI panel. A comms light that stays dead after rewiring points at the cable or port settings you just touched; the serial timeout check walks that down in order.
The decision
Recommendation: put a gateway between the panels and the controller when the slots are full and the panels poll as masters: the gateway alone holds the master role on the controller's existing port, the new panel plugs into a gateway port, and the summed poll time stays under half the port's wire time. Free a slot for a dual-port communications module when a module swap or a remote I/O move is genuinely available, because one master per port needs nothing aggregated. Share one RS-485 line only when the controller is the single master and the panels accept a slave role, inside 32 unit loads and 1,200 m at 100 kbps. Never land a second polling master on an existing serial line.
This gateway work serves the wired operator interface. A signage board is a separate screen with its own player: if the cell also wants shift targets and posted notices, our shop-floor signage for andon-style boards plays saved content offline and keeps that traffic off the controller's serial ports entirely.
Questions
Can two operator panels share one RS-485 pair if they take turns?
Only if something enforces the turns. RS-485's electrical side allows several transmitters on one pair, and a few protocols build arbitration in from the start. A pair of ordinary polling masters has no such rule: each transmits on its own clock, so overlaps are a question of time, and a layout that works at one poll rate can fail when a screen is added or a refresh is sped up. Use a gateway, or move to a protocol that defines who talks.
How many devices fit on one RS-485 segment?
Count unit loads, not boxes. The standard requires each transmitter to drive 32 unit loads on one segment; a reference receiver counts as 1, and modern transceiver chips are specified at 1/4 or 1/8 of a unit load, so a segment built with them can carry 128 or 256 receivers. Each panel and controller datasheet prints its own load; sum them for the segment. A repeater joins two fully loaded segments into one network when the budget runs out.
How much delay does a serial gateway add to panel updates?
Plan one extra poll cycle. The gateway queues each panel's request behind whatever is already in flight on the controller line, so the worst case is one poll's wire time, about 21 ms for a 10-register request at 19.2 kbps in the worked example, plus the translation time stated in the gateway's own documentation. Requests that arrive while one is already on the wire wait in that queue, which is why the summed poll budget matters.
Planning information, not legal or engineering advice. Rules differ by place: check codes and permits with your local authority and your licensed trades.