Averaging FX2N-2AD Analog Input on FX1N PLC Integer Methods

Problem Overview

The Mitsubishi FX2N-2AD analog input module delivers two 12-bit channels into an FX1N CPU as 16-bit integer values in data registers. Field installations commonly see noise from thermocouples, RTD transmitters, 4-20 mA loops, and VFD-induced EMI on the analog cable. Engineers want to smooth the signal, but the FX1N family does not natively handle 32-bit REAL (floating-point) arithmetic the same way an FX3U or FX5U does. The FX1N programming manual (JY992D88101) section 4.15 documents FLOAT, not REAL, and ladder-instruction libraries such as DINC, DADD, and DMOV require careful data-type declaration to avoid the "type mismatched" compile error reported in the original PLCTalk thread.

Prerequisites

• FX1N-14MR/MT or FX1N-24MR/MT CPU with firmware that supports the FROM/TO instructions
• FX2N-2AD analog input module (catalog FX2N-2AD User's Manual, JY992D74701)
• GX IEC Developer 7.04 or later, or GX Works 2 with FX1N compatibility mode
• FX1N, FX1S, FX2N, FX2NC Programming Manual - JY992D88101

FX2N-2AD Hardware and Read Procedure

The FX2N-2AD occupies eight I/O points and identifies itself as a special-function block. Default number range is K0-K7, configurable by rotary switch. Average-mode and high-speed conversion are enabled by writing control word H0000 (normal) or H1000 (averaging) to buffer memory address 0 with the TO instruction, then read the digitized 12-bit value from buffer memory address 1 (CH1) and 2 (CH2) with the FROM instruction.

[ TO  K0  K0  H1100  K1 ]   ; enable averaging + 4-20 mA mode, BFM 0
[ FROM K0  K1  D100  K1 ]   ; read CH1 digital value (0-4000) into D100
[ FROM K0  K2  D102  K1 ]   ; read CH2 digital value into D102

Integer-Only Averaging Methods

Three approaches are practical on FX1N without breaking the integer data type. All three rely on the FX1N's integer math, and the choice depends on noise profile and available D-register budget.

Method 1: Hardware Averaging in the FX2N-2AD

The module's H1000 averaging bit samples 8 internally and returns the mean. No D register, no ladder scan overhead, no REAL required. The result is already integer in D100/D102. Use this first; it is the cleanest solution.

Method 2: Running Average (Single Register, Memory-Light)

The running-average formula A₁ = (I + (n-1) × A₀) / n requires only one previous-average register, not n samples. For n=16, the multiplier is 15. With D110 declared as DOUBLE for the previous average, the program becomes:

DADD D100 D114 D116       ; running_sum = I + A_prev*(n-1)  // D116 declared DOUBLE
DDIV D116 K16 D118        ; A_new = running_sum / 16
DMOV D118 D114            ; shift current average into "previous" slot

This technique is documented in the Industrial Monitor Direct knowledge base and scales well to FX1N's limited register budget.

Method 3: Sliding Window (N Samples)

For non-periodic spikes, a 16-deep circular buffer and integer mean is more accurate. Use a pointer in D200 and store raw samples in D300-D315. After 16 reads, copy the buffer to D320-D335 and apply a simple moving average. For larger N, indirect addressing with Z0/Z1 (index registers) is required; FX1N supports up to 16 index points in some firmware revisions, but Z0 is always present.

Fixing the "Type Mismatched" Compile Error

The error reported in the PLCTalk discussion occurs when 16-bit single-word variables are wired into 32-bit instructions. The fix sequence is:

1. Open the Global Variable editor in GX IEC Developer.
2. Locate D114, D116, D118, D120 and any other register feeding DADD, DMUL, DDIV, or DMOV.
3. Change the data type from WORD to DWORD (DINT).
4. For D100/D102 raw inputs being combined into a 32-bit sum, declare D100 explicitly as WORD but pre-merge with DMOV D100 D114 so the destination is DWORD.
5. Rebuild the project. The type mismatch should clear.

If GX IEC Developer still rejects the project, switch the project's CPU target to FX1N (not FX1S) - FX1S has no 32-bit instructions, which is the most common root cause when developers attempt to use DADD on an FX1S.

Engineering Formulas for Filter Tuning

Filter performance in a control loop is measured against the analog scan period (T₀) of the FX2N-2AD, which is 15 ms per channel in high-speed mode or 6 ms with averaging disabled. The equivalent time constant τ of a running average of n samples is τ = (n-1) × T₀ / 2. For n=16 and T₀=15 ms, τ = 112 ms - fast enough for temperature, slow enough to reject 50/60 Hz pickup when combined with the module's own H1000 averaging.

Comparison: Smoothing Strategies

Verification Procedure

1. Connect a precision 4-20 mA source (Yokogawa CA150 or similar) to CH1.
2. Force the FX2N-2AD averaging bit OFF first; force a step change from 4 mA to 20 mA.
3. Monitor D100 in GX Works 2 monitor mode. Confirm raw counts jump 0-4000 with visible noise of typically ±5 LSB.
4. Enable Method 2 with n=8. Confirm the noise envelope drops to ±1 LSB while step response is acceptable for the controlled process.
5. Repeat for CH2.
6. Run the project for 24 hours with the process live; confirm no watchdog trip and D-register values remain within expected engineering range.

FAQ