A factory cable does not fail politely. It is pulled through a tray with motor leads, landed on a terminal by someone in a hurry, exposed to ground difference between cabinets, and asked to carry a clean bit while contactors, drives and welders make the local ground move. A rugged interface begins by accepting that the wire is part of the product, not an accessory.

The same board may need RS485 for a long multidrop run, CAN for deterministic machine traffic, a small SPI CAN controller because the host has no CAN block, digital isolation because two cabinets should not share fault current, and an isolated driver because a power switch cannot be trusted to share the logic ground. These parts belong together. They decide where noise is allowed to flow, where a fault stops, and how much abuse the controller is allowed to see.

RS485 stays useful because it respects the cable

Why RS485 holds up on the factory floor is a better starting question than which transceiver is fastest. The bus uses a differential pair, so the receiver reads the voltage between two wires rather than treating one wire as a fragile truth against local ground. That does not make it immune to bad wiring. It gives the designer a fighting chance when the two ends of the cable sit in different cabinets, with different ground noise and long parallel runs near power wiring.

The ruggedness comes from the whole link. Twisted pair limits magnetic pickup because induced noise tends to appear on both wires. A receiver with a useful common-mode range can keep reading while the pair floats away from local logic ground within its allowed window. Bias resistors can define an idle state when no driver is active. Termination can stop reflections when the run is long enough for the edge rate to matter. Surge and ESD protection can send a strike away from the transceiver rather than through it. None of those choices is exotic. Leaving one out is how a bus that looked fine on a bench becomes a field-return story.

RS485 also forces discipline in firmware. A multidrop half-duplex bus has to decide who speaks, when a driver is enabled, how long the transmitter waits after the last byte, what happens when a node disappears, and whether a cable short should lock the product in a transmit state. The electrical layer gives distance and noise margin, but it does not solve collisions, retries or address discipline. A durable interface is the electrical layer and the protocol behavior married tightly enough that the product recovers without a service visit.

That is the reason RS485 still appears in new equipment. It is not nostalgic. It is a plain way to move data through a place that treats cables badly.

The isolated interface is a boundary with rules

MAX22505GTG+T in an isolated industrial interface sits in the class of parts chosen when the communication pins need more than a transceiver. The isolation boundary has to pass signal information while blocking ground current. That sounds simple until the board layout has to preserve creepage, clearance, transient rating, power domain separation and return paths without letting a copper pour, test pad, connector shield or mounting screw quietly bridge the boundary.

Isolation is often added after a prototype misbehaves, then treated like a bandage. That is the wrong order. If the product connects to a field cable leaving the enclosure, the isolation plan should be drawn before the PCB placement starts. Which side owns the connector? Which side receives field power or isolated power? Where do TVS parts return? What happens to cable shield current? Can a technician probe the field side without touching the logic side? Does firmware know whether a communication fault came from the isolated side or the host side? The schematic symbol cannot answer those questions.

The isolation boundary also changes testing. A board can pass a bench serial test with the two grounds clipped together and then fail in a cabinet because the field side was never allowed to float during validation. A better test fixture powers the sides the way the product will be powered, exercises hot plug and cable fault cases, and checks that a transient does not corrupt the host, latch the isolator or leave the transceiver driving a damaged line. The part choice matters, but the boundary behavior matters more.

When isolation is designed as an architecture, not a patch, the interface has a clear promise: a field fault may stop communication, but it should not become a logic-board fault.

Classic RS485 and CAN solve different factory conversations

MAX485 for classic RS485 transceiving points to the older, familiar half-duplex bus style. It is small, direct and easy to understand. Driver enable, receiver enable, A/B pair, termination, bias, protection and a UART behind it are enough to build many field links. That simplicity is the strength and the trap. The chip will not decide the protocol, guard timing, node addressing or line-fault recovery. The designer has to handle those choices in the controller and in the bus rules.

SN65HVD230 for CAN bus transceiving belongs to a different conversation. CAN was built around message arbitration, error handling and a bus that can let several nodes share time without a central poller in the same way. A CAN transceiver still faces the physical cable, but the protocol above it carries stronger ideas about priority, frame validity and fault confinement. That is why CAN is common in machines and vehicles where control messages have to share a bus without turning every exchange into a custom serial schedule.

The choice between RS485 and CAN is not a contest over which line driver is tougher. It is a question of traffic shape. RS485 is attractive when the product wants a controlled master-slave conversation, simple framing, long cable reach and freedom to define the higher protocol. CAN is attractive when many nodes need to talk with defined priority and the controller should participate in arbitration and error handling. Both can be laid out badly. Both can be made durable. The interface layer should follow the traffic model rather than forcing the traffic to fit the cheapest transceiver on the shelf.

There is also a service story. RS485 faults often look like silence, collisions, wrong bias or stuck transmitters. CAN faults often expose bus-off behavior, error counters, termination errors or one node disturbing the whole network. The firmware and the diagnostics should match the bus. If service software reports every fieldbus failure as communication error, the technician is left to guess whether the cable, node, terminator, transceiver, protocol or ground reference is guilty.

An MCU can borrow CAN, but it still has to own recovery

MCP2515 adding a CAN controller to an MCU is useful when the host has no built-in CAN controller and the design still needs CAN traffic. The part gives the microcontroller a CAN controller over SPI, with message buffers, filtering and interrupt behavior outside the host. That can save a redesign or let a small MCU join a CAN network, but it also inserts a second timing path between application code and the bus. SPI latency, interrupt service, buffer strategy and error handling now become part of the network behavior.

A borrowed CAN controller should not be treated as a transparent cable. Firmware has to decide how received frames are drained, what gets filtered in hardware, what happens when transmit buffers fill, how bus-off recovery is handled, and how a reset of the MCU interacts with the external controller. If the controller keeps state while the host restarts, the bus may see behavior the application did not intend. If the host clears the controller too aggressively, a recoverable noise event may become a repeated network drop. The chip adds capability; it also adds state that must be owned.

Layout is still a fieldbus layout. The CAN transceiver sits between the controller and the cable, with termination, protection and common-mode behavior placed around the connector. A clean SPI route inside the board does not make the outside cable safe. The common mistake is to spend attention on the MCU-to-controller link and then give the bus edge the same generic protection used on a short internal header. On a factory floor, the outside edge is where the product earns its life.

Automotive CAN parts are chosen for the abuse around the bus

TJA1051 for automotive grade CAN transceiving shows why a CAN transceiver is not only a logic translator. Vehicle and industrial environments both punish ground, wiring and power states. A transceiver in this class may be selected for fault behavior, standby behavior, electromagnetic performance, bus pin protection and how it behaves when one side of the system is powered and the other side is not. Those details decide whether a module is a polite participant on the bus or a load that drags the bus down during a fault.

Automotive grade does not automatically mean the part fits every factory product. The supply voltage, logic level, common-mode range, package, protection scheme, standby current, wake behavior and compliance needs still have to match the design. A part used in a vehicle may have excellent bus manners in its intended circuit and still be the wrong fit if the industrial controller needs a different isolation scheme, cable length, connector grounding method or operating temperature assumption. The right question is not whether the part is stronger in the abstract. The right question is which failure it was designed to survive.

Isolation also appears where power devices switch

SI8244BB-D-IS1R as an isolated gate driver belongs beside the communication isolators because both are about controlling where energy is allowed to cross. A gate driver may sit between a logic controller and a MOSFET or IGBT that switches a motor, solenoid, heater or power stage. The logic side wants clean command timing. The power side may move at high voltage, high dv/dt and ugly current edges. The driver has to pass the command while keeping the switching node from shoving noise back into the controller.

The isolated driver is not only a voltage rating. It sets timing, propagation delay matching, output drive strength, undervoltage behavior and fault response. Too little drive can make the switch heat. Too aggressive an edge can raise emissions and ringing. A poor return layout can make the gate command bounce against the source or emitter reference it is trying to control. A good isolated driver choice has to be paired with the gate resistor, local decoupling, bootstrap or isolated supply plan, clearance, thermal path and the measurement method used during validation.

This is where the interface article and the power article meet. A fieldbus can tell the controller what to do, but the isolated gate driver may be the device that turns that command into current through a load. If the isolation strategy is split across communication and power without one ground plan, the product can pass messages cleanly and still fail when the output switches. The noise has to be designed out of the whole path, not each schematic page in isolation.

Small digital isolators keep ordinary signals from becoming ground paths

ADUM1201 for two channel digital isolation is the kind of part that appears when only a few logic lines need to cross a boundary. It may isolate UART, enable, fault, chip-select, reset or status lines. The electrical idea is modest: pass digital information without a direct conductive path. The system effect can be large because one forgotten logic wire can undo the isolation provided by the main communication part.

The hard part is not drawing the isolator. It is auditing every crossing. A debug header, shield connection, programming adapter, LED reference, pull-up rail, test pad or measurement ground can sneak across the gap. In production, the board may never see those paths. In service, a technician may connect a laptop or scope and create a temporary bridge. If the isolation boundary matters for field safety or system resilience, those service paths have to be planned as carefully as the main signal.

ADM2483 isolating an RS485 field bus brings the idea back to the cable. An isolated RS485 transceiver combines the fieldbus function and the isolation boundary in the same design area. It can reduce the number of decisions a layout has to coordinate, but it does not remove them. The field side still needs isolated power, surge strategy, termination, biasing, connector grounding and a place for cable energy to go. Integrated isolation helps only when the rest of the board lets it stay isolated.

The interface is rugged when its failure is contained

A factory interface should be judged by what happens after the first bad cable, not by the first clean packet. RS485 earns its place by respecting long noisy pairs. CAN earns its place when shared machine traffic needs arbitration and error behavior. External CAN controllers help small hosts join the bus but add state that firmware must own. Automotive-grade transceivers bring bus manners shaped by abuse, yet still need the right circuit around them. Digital isolators, isolated RS485 devices and isolated gate drivers draw the line where fault current, common-mode noise and switching edges should stop. The goal is not a board that never sees noise. The goal is a board that lets the outside world be noisy while the controller stays in charge.