Serial remote I/O is the faster link only when every remote point is discrete and the total is a few cards' worth, roughly 48 points or fewer. Any analog channel, or more discrete cards than that, and Ethernet remote I/O wins: its discrete update window is fixed whether the bases carry 1 point or 1024, and analog arrives in registers in binary with no multiplex logic.
You have a rack PLC whose Ethernet card already serves your HMI stations, and two remote I/O bases to add across the floor. Which link should carry them? Serial remote I/O is the faster choice only when every remote point is discrete and the count is a few cards' worth. Any analog channel, or more discrete cards than that, and Ethernet remote I/O wins on speed and on programming: analog values land in the PLC's registers in binary, with no multiplex logic to write. This article gives the rule, the numbers behind it, the hardware for each case and the decision.
What you are comparing
Both architectures move field signals into the CPU's register map through adapters at the remote bases. They differ in the link and in what your program has to do.
Ethernet remote I/O pairs one master module in the local rack with one adapter at each remote base, tied together over standard Ethernet with IP addressing. Two documented behaviors decide the comparison. The master's own serial-versus-Ethernet comparison states the design plainly: discrete data moves in a fixed window, the same whether the bases carry 1 point or 1024 points, and analog values are written straight into registers in binary, so no ladder logic ever handles them.
Serial remote I/O pairs a serial remote I/O module at the CPU with one adapter per base on a dedicated serial link. Its exchange is quick with a few discrete cards and slows as the count grows. Analog must be multiplexed in program logic; that is the only path the arrangement offers, and it takes programming beyond basic I/O.
The HMI link stays where it is. The Ethernet card that serves your HMI stations is a communications module, not a remote I/O scanner. It keeps that job, and the master takes a second slot beside it. The basics of planning that channel are in connect a PC HMI to a PLC over serial or Ethernet.
Side by side
| Serial remote I/O | Ethernet remote I/O | |
|---|---|---|
| Fastest for | A few discrete-only cards | Many discrete points, or any analog |
| Discrete update | Quick when small, slows as points grow | Fixed window, the same at 1 point and at 1024 |
| Analog | Multiplexed in program logic, no other path | Into registers in binary, no logic |
| Extra programming | Required | None for basic I/O |
| Distance per segment | About 1,200 m (4,000 ft) at 100 kbit/s or less, the usual RS-485-class guideline | 100 m (328 ft) over copper twisted pair, the IEEE 802.3 limit; switches extend it |
| Addressing | Fixed protocol exchange | IP addressing |
| Hardware count | 1 serial module + 1 adapter per base | 1 master + 1 adapter per base, plus a switch if the runs need one |
| Buying | Separate modules | Sold as kits of one master with 1 to 10 adapters |
Where serial remote I/O wins
A remote base holding nothing but discrete points, in cards you can count on one hand, finishes its serial exchange sooner than the Ethernet master finishes its fixed discrete window. That is the whole case, and it is narrow. Typical fit: one or two bases with stack lights, door switches and a parts-present photoeye.
The master's spec-sheet comparison puts serial ahead only while a base carries a few discrete I/O modules. With common 16-point cards, a few cards is roughly 32 to 48 points across all bases. Under that line, with no analog anywhere, serial updates sooner and needs less hardware. Past it, or with one analog channel added, the case is gone.
Where Ethernet remote I/O wins
- Any analog at all. The adapter writes each channel into a register in binary, and your program and HMI read real values. Serial has no equivalent.
- More discrete than a few cards. Past that band, the fixed window beats a serial exchange that grows with the count.
- More than two bases. Adapters are added one per base, and kits cover up to 10 on one master.
- Distance between buildings. Copper runs and ordinary switches carry the network across a plant; longer routes take fiber.
- A dropped base is a known problem. Networked I/O can lose an adapter, and recovery is a defined task; make a Modbus master reconnect after a slave outage shows the pattern.
- A different controller changes the question. If the CPU is a soft PLC on a fieldbus runtime, the choice is protocol-first; choosing Modbus TCP remote I/O for CODESYS covers that case.
The same registers can also feed a wall status board away from the machine; we build factory and shop-floor displays for that job.
Example: two remote bases on a packaging bench
Scenario: the bench line's CPU sits in the main panel with one Ethernet card serving three HMI stations. Two bases go in: the infeed base is 40 ft from the panel, the discharge base 80 ft. Base 1 carries one 16-point input card, one 8-point output card and 2 analog sensors: 24 discrete points and 2 analog channels. Base 2 carries one 16-point input card and one 8-point output card: 24 discrete points, no analog.
Count it up:
- Discrete: 24 + 24 = 48 points across five cards of 8 and 16 points.
- Analog: 2 channels, both at the infeed base.
- Farthest run: 80 ft (24 m), well inside the 100 m (328 ft) copper segment, so no fiber and no media converters.
- Hardware over Ethernet: 1 master in the local rack + 2 adapters, with the HMI card keeping its slot and its job.
- Hardware over serial: 1 serial remote I/O module + 2 adapters, plus multiplex logic written and maintained for the 2 analog channels.
Decision: 48 points sits at the top of serial's winning band, and the 2 analog channels fall outside it entirely, so this line takes Ethernet remote I/O. Configure the master, fit one adapter per base, and the analog values appear in registers in binary while the discrete window runs as if the bases held 1 point instead of 48.
Flip the scenario to see the other side: two discrete-only bases at 8 points each, 16 points across two cards and no analog, would update sooner over serial at lower hardware cost. That is the job serial wins.
The decision
- Discrete points only, one or two bases, roughly 48 points or fewer across all of them: serial remote I/O, and expect to write the program logic it needs.
- Any analog channel anywhere, or more discrete points than that band, or more than two bases: Ethernet remote I/O, one master in the local rack and one adapter per base.
- Either way, the HMI link keeps its own card and never moves onto the remote I/O network.
Recommendation: default to Ethernet remote I/O for any new remote base. Choose serial remote I/O only when every point is discrete, all bases together hold three or fewer 16-point cards, and no analog channel is present. In every other case the Ethernet master's fixed discrete window and register-mapped analog pay for themselves in scan time and in logic you never write.
Before you order
- Compatibility is per CPU model. The master's documentation lists the CPUs it works with; confirm your CPU appears on that list before powering anything up.
- Count slots and power. The local rack now holds the CPU, the HMI card and the master: two module slots and the base's power budget.
- Order the kit shape. Kits pair one master with 1 to 10 adapters in one order; a two-base job is one master and two adapters.
- Check every run. Copper segments have a fixed maximum length; any run past it takes a switch or a fiber link.
- Keep the I/O network separate. Put remote I/O traffic on its own subnet, and leave internet reachability off the controls network entirely.
Power the rack down before inserting any module, and test each remote point from the PLC's I/O tables with the machine in a safe state before the process uses it.
Questions
Does the Ethernet card that serves my HMI stations also scan remote bases?
No. That card is a communications module for the HMI stations, and remote bases need a remote I/O master in the local rack. The two modules have separate jobs and sit side by side, so the HMI stations do not ride the remote I/O network. Before ordering, count one free slot for the master and check the base power budget.
How many remote bases can one Ethernet master handle?
In the rack-PLC family this setup usually comes from, kits pair one master with 1 to 10 adapters, and a two-base job uses one master and two adapters. Each master has a stated adapter and point limit in its manual; two or three bases sit well inside it. If you outgrow one master, add a second master in a free slot and split the bases between them.
If the remote bases are only 80 ft away, can I use local expansion instead?
Only if your CPU is a model that lists local expansion. In the rack family this setup comes from, the expansion modules are listed for the 250-1 and 260 CPUs, not the 240, and local expansion carries distance limits of its own. A short cable run changes none of that; use remote I/O for the bases and keep expansion for the CPUs that list it.
Planning information, not legal or engineering advice. Rules differ by place: check codes and permits with your local authority and your licensed trades.