5 Specs That Mismatch What the Datasheet Claims — Mitsubishi Electric vs Siemens PLC Under Real Load

Number: Mitsubishi Electric MELSEC iQ-F FX5U executes a basic instruction in ~34 ns; the Siemens S7-1200 CPU 1214C standard (non-G2) takes ~85 ns. On paper, Mitsubishi PLC is 2.5× faster. Mechanism: That 34 ns is measured on an empty program — zero I/O, no comms overhead, no background tasks. The IEC 61131-3 execution model forces the CPU to interleave I/O refresh, communication stack (PROFINET on Siemens PLC, CC-Link/Ethernet on Mitsubishi), and operating system housekeeping every cycle. The real instruction throughput collapses when the program uses mixed data types or indirect addressing, which both CPUs handle differently. Worked consequence: In a 10,000-step ladder program with 64 digital I/O and one PID loop, the FX5U’s effective cycle time is about 1.2 ms (illustrative) vs the S7-1200’s 1.8 ms (illustrative) — not a 2.5× gap, but still ~33 % faster. That 0.6 ms difference matters if your machine’s critical path is a high-speed counter that must sample every 1 ms. If your application is a batch process with 100 ms cycle tolerance, the delta is irrelevant. When reversed: If the program uses heavy indirect addressing or floating-point math on arrays, the S7-1200’s architecture with local memory caching (100 KB work memory) can narrow the gap — or even surpass the FX5U if the FX5U spills to its slower external memory bus. The provenance: the 34 ns is a best-case, one-instruction measurement; never multiply it by your step count.

Number: The MELSEC iQ-F FX5U includes 2-channel 12-bit analog input and 1-channel 12-bit analog output built into the CPU. The Siemens S7-1200 CPU 1214C has no on-board analog — you must add a signal board or module. Mechanism: 12-bit resolution means 4,096 discrete steps over a 0–10 V range, ~2.4 mV per step. For a pressure transducer with 0–10 bar output, that’s ~2.4 mbar resolution — fine for many hydraulic systems. But the critical detail is that the FX5U’s built-in analog shares the same internal supply as the CPU logic; on a 24 V DC bus that dips below 20 V (common on aged machines or long cable runs), the analog reading drifts up to ±3 % (roughly, based on typical unregulated architecture). The S7-1200, with no on-board analog, forces the designer to use an external isolated signal module — which costs more and takes a slot but guarantees galvanic isolation and a regulated reference. Worked consequence: If you spec the FX5U for a simple temperature monitor (0–10 V from a PT100 transmitter), the 12-bit resolution is adequate and you save the cost of an expansion module. But if that analog input controls a servo valve in a closed loop, the unregulated drift can cause oscillation at low setpoints. The decision: use an isolated external module anyway, defeating the on-board advantage. When reversed: For a machine that runs on a clean, regulated 24 V supply (e.g., a new panel with a dedicated PSU), the FX5U’s built-in analog works without issue. The Siemens approach forces an extra purchase even when you don’t need isolation — a hidden cost.

Number: Both the FX5U and the S7-1200 claim “built-in positioning” — the FX5U with high-speed counters and pulse outputs on the CPU; the S7-1200 with integrated motion (PTO) and PID. Mechanism: The FX5U can drive up to 4 independent axes using transistor outputs (assuming 4 PTOs). The S7-1200 CPU 1214C can output up to 4 pulse trains, but the actual axis count depends on how many high-speed outputs you configure and whether you use the on-board signal board for additional channels. The critical constraint is not the published max axes, but the simultaneous, coordinated motion — both CPUs can run independent PTOs, but neither does true interpolation across all axes without an external motion module. Worked consequence: For a simple pick-and-place with two independent linear axes, both work. For a three-axis CNC-style contouring move, the FX5U’s built-in positioning can’t execute G02 circular interpolation — you’d need an iQ-R or a dedicated motion controller. The S7-1200 likewise lacks S-curve profiling on all axes simultaneously. The datasheet “motion” label applies to open-loop pulse/direction or stepper drives, not closed-loop servo with electronic gearing. When reversed: If your machine uses simple point-to-point moves with two servo drives controlled via PTO + direction, the integrated motion is sufficient and saves the cost of a separate motion module. The provenance trick: “motion” in the micro-PLC space means “pulse train output” — it does not mean multi-axis coordinated servo control, regardless of brand.

Number: The S7-1200 ships with a built-in PROFINET interface for programming, HMI, and PLC-to-PLC networking. The FX5U includes built-in Ethernet and RS-485. Mechanism: PROFINET uses a dedicated hardware stack with real-time (RT) and isochronous (IRT) classes — the stack runs independent of the CPU scan, meaning a PROFINET cycle of 1 ms does not add jitter to the user program. The FX5U’s Ethernet is standard TCP/UDP over a software stack; every packet interrupt steals CPU cycles. Under heavy data exchange (e.g., 32 bytes to a VFD every 10 ms), the FX5U’s scan time can increase by 15–25 % (illustrative). The RS-485 port runs Modbus RTU master/slave, also handled by the CPU. Worked consequence: In a machine that communicates with 8 VFDs over Modbus RTU (RS-485) and one HMI over Ethernet, the FX5U’s cycle time inflates enough to miss a 2 ms high-speed counter interrupt (practical failure mode). The S7-1200, with dedicated PROFINET silicon, handles the same HMI + drive traffic with negligible scan impact. The decision: if your application has even moderate real-time network overhead (3+ devices, When reversed: For a standalone machine with one HMI and no networked drives, both Ethernet interfaces perform identically — the RS-485 is rarely used. The FX5U wins on connector cost (RJ45 vs PROFINET connector). The provenance: the number “built-in Ethernet” tells you nothing about how much CPU load that network adds.

Number: The FX5U lists program capacity up to 64k steps; the S7-1200 CPU 1214C offers 100 KB integrated work memory. Mechanism: Steps and kilobytes are not convertible without knowing the instruction encoding. One Mitsubishi step is roughly 4 bytes of object code (typical for a basic LD instruction); 64k steps ≈ 256 KB of compiled code. The Siemens 100 KB work memory is for compiled code plus data blocks — but the actual user program size limit depends on the code/data split. The critical difference: the S7-1200’s retentive memory is part of the work memory (you allocate) while the FX5U uses a retentive latch area (4 kB typical). If your application needs to retain 100+ analog setpoints across power cycles, the FX5U may run out of latch memory even though the program size is below 64k steps. Worked consequence: A packaging machine with 80 recipes of 60 bytes each = 4,800 bytes of retentive data — the FX5U’s latch area is tight; the S7-1200 can allocate 10–20 KB from work memory as retentive. The decision: if your machine has a large recipe database, the Siemens memory architecture is more flexible. When reversed: For a small program (