Mitsubishi PLC FX Series: 4 Spec Mistakes That Cost Me $27,000

If you quote or buy Mitsubishi PLC components for a living, here is the conclusion up front: the expensive mistakes are not wrong product families; they are unverified suffix characters at the end of otherwise correct part numbers. In eight years of sourcing Mitsubishi PLC FX series units, VFDs, contactors, relays, and timers, the documented mistakes I am responsible for came to roughly $27,000 in restocking fees, emergency freight, and lost goodwill. Not one of them involved a defective part.

I am the person who turns a request like 'we need the FX PLC again, 32 points' into a purchase order. I do not write machine code and I do not build panels. I do maintain our quote log, and since 2017 I have logged every serious error with its date and dollar value. By 2020, those notes had become a pre-order checklist. The checklist has caught 47 potential errors in the past 18 months. Forty-seven phone calls we did not have to make.

That number matters more than any price discussion. Before I learned the habit, I was proud of fast quotes. Fast is still fine, but I used it as an excuse to skip the questions that actually cost money. A wrong Mitsubishi PLC at a five percent discount is still one hundred percent wrong for the machine. The total cost of rework, not the unit price, is what decides whether a deal was cheap.

Relay outputs click. Transistor outputs switch.

The first and still most common mistake I made was treating output type as a minor option. On the FX family, the middle letter of the part number tells you the output construction: MR is relay output; MT is transistor output. The relay version can switch AC or DC loads at up to about 2 A per point, which makes it great for solenoid valves and contactors. The transistor version is DC-only and usually rated around 0.5 A per point, but it is solid-state and roughly 50 times faster. If your machine needs high-speed pulse output for steppers or servo drives, relay output will not do it. Period.

In my first year, I quoted 30 FX3U-32MR/ES units for a small machine builder. I checked the I/O count: 16 inputs and 16 outputs. I checked the supply voltage: AC-powered with a built-in 24 V DC service supply. I never asked what the outputs were switching. The schematic said '24 V DC output to stepper drives,' and it looked fine on my screen. On the first bench test, the axis just clicked. A relay can switch DC, sure, but a mechanical relay cannot produce a 100 kHz pulse train. That is what the transistor version is for.

The fix cost about $890 in restocking and expedite fees, plus a week of the customer's build schedule. Now the question is on every order form: what are the outputs switching, and does anything on that list need pulse output? If the answer is 'pulse,' the part number needs a T.

To be fair, relay outputs are often the better choice. For slow, high-current, or AC loads, relay is more robust and simpler to replace. The mistake was not choosing relay output. The mistake was choosing it without checking the load.

'24 V DC inputs' is not a complete sentence

By September 2021 I thought I had the output question under control. Then a packaging line order taught me the input-side version of the same lesson.

The customer ordered 40 Mitsubishi PLC FX series units and wrote: '24 V DC inputs, transistor outputs.' I confirmed the model, confirmed the price, and processed the order. Both sides said '24 V DC inputs' and meant different things. I meant there is a 24 V DC sensor supply. They meant the sensors are PNP, which is a wiring mode, not a voltage.

With NPN sensors (sinking), the PLC input common is wired to the positive side of the supply. With PNP sensors (sourcing), the common goes to 0 V. On many Mitsubishi FX-series base units you choose the mode by how you wire the S/S terminal. Get it wrong and the input LED stays dark or flickers. That is exactly what happened on site: 40 units, every proximity input on one line staying dark. The product was fine. The installation was correct for the sensors they had not told us about.

The correction cost roughly $1,100 of our support time, and the line learned an embarrassing lesson: specify the sensor type in the purchase order. NPN or PNP. Not just '24 V DC.'

One more detail on the same line: transistor outputs also come in sink and source versions, and the difference can hide in one character. With FX3U and FX5U transistor main units, you will see /ES and /ESS suffixes. That extra S is easy to miss when you are moving fast. It is not optional. Ask whether the load side of the output wiring is positive-common (sink output) or negative-common (source output). If this is starting to sound repetitive, good.

The 'new FX3U' trap

The second expensive belief is that FX5U is simply a newer FX3U and old projects will transfer over without pain, because it is 'still Mitsubishi PLC FX series.' The idea has some truth: the two families share a name and many instructions. But family name does not mean identical platform. The FX5U programming environment is GX Works3, not GX Works2. A GX Works2 project will not just open; it has to be converted, and conversion requires manual review of instruction support, device assignments, and module configuration.

I hit this in October 2022. A customer with 24 existing FX3U-controlled machines wanted to standardize on the newer hardware. They asked for FX5U-32MT/ES. We validated I/O count, output type, and supply. We did not ask who was responsible for the program conversion. Their contract programmer found that the GX Works2 file could not be loaded into GX Works3 without going through the converter and then fixing multiple instruction and device mismatches by hand. The whole panel schedule slipped.

We took back 17 of the 24 units at a restock penalty and helped source FX3U units to keep the line moving. That cost about $1,900 in restocking, freight, and a damaged relationship. Today, when someone says 'we are moving from FX3U to FX5U,' the first thing I ask about is program migration and who has reviewed it. This is not a product criticism. The FX5U is a solid upgrade with built-in Ethernet. It is just not an automatic one.

Drives: voltage, amps, duty, braking

The same pattern shows up in VFDs, just with higher voltage and, sometimes, a bigger bill. If you are a VFD distributor, the returns are familiar. The order says '7.5 kW drive.' The motor nameplate says something else. The plant supply is 400 V, but the quote was written for the 200 V class because the old drive was 200 V.

Do not hold me to an exact split, but I would guess most of our VFD returns are not failed units. They are mismatched ratings. Here is the drive specification guide I keep in the quote template:

1. Supply voltage and phase. 200 V class or 400 V class? Single-phase input or three-phase input? Confirm the actual panel feed.
2. Motor current, not just motor power. The kW on the motor nameplate is a starting point. Match the drive's continuous output current to the motor's full-load current under the drive's rated ambient and switching frequency.
3. Duty. A centrifugal fan or pump is a different load from a conveyor or a plunger pump. Check whether the drive rating you are looking at assumes normal duty or heavy duty.
4. Braking. If the load is high inertia or overhauling, a standard drive without a braking resistor will eventually trip on overvoltage. Ask about the stopping profile.
5. Options and feedback. Does the application need closed-loop feedback, a communication option, or a specific EMC filter? That changes the model number too.

Voltage. Current. Duty. Braking. In that order. Every drive return I have processed skipped one of those four lines.

The rest of the catalogue: relays, timers, contactors

It is tempting to think this only applies to 'complex' products. Then I added an AC vs DC coil question to our checklist in Q1 2024 after a timer order came back with the note 'coil voltage wrong.' A timer with a 24 V AC coil looks identical to one with a 24 V DC coil, and the model number is not always printed on the outside of the box where a rushed stock clerk can see it.

Contactors have the same issue, plus an extra layer. According to IEC 60947-4-1, a contactor's rated current is tied to its utilization category: AC-1 for resistive loads, AC-3 for starting and stopping squirrel-cage motors during running, and AC-4 for inching and plugging. Two contactors with the same frame can have very different application ratings. If you buy a contactor the way you would buy a doorbell, you discover the difference when it welds shut.

The point is not to memorize the whole Mitsubishi catalogue. The point is that the easy products punish sloppy habit just as fast as the programmable ones do. One of our regular timer distributor customers tells the same story: the returns stopped when he started putting voltage type and time range on every line item.

Where I would still ignore this article

If you are building a one-off machine for your own use and you have time to rewire a panel, you can get away with a lot. A relay-output PLC can switch a solid-state relay, and a missing PNP sensor can be fixed with a little interface wiring. The strict pre-order checklist is for people who sell or install the same thing over and over, where a two percent error rate means weekly firefighting.

If the application is genuinely simple, the rules relax. A few panel relays in the output circuit, all NPN sensors, and a conveyor running at constant speed will not need the full interrogation. But the habit costs no vendor money and very little time, and it turned out to be worth $27,000 to me. The list now sits on a laminated card next to our quote terminal. Read the full part number out loud before sending. Price is a number. The part number is a promise. The current score is 47 errors avoided in the last 18 months, and I would like to keep it at zero for a full year.