# 📡 S7-1500 Advanced Signal Filtering Library
### SCL Function Blocks for Siemens TIA Portal

---

## Overview

This library provides a collection of advanced signal filtering Function Blocks (FBs) written in **Structured Control Language (SCL)** for the **Siemens S7-1500 PLC** platform. All blocks share a consistent interface and are designed for real-time analog signal processing in industrial automation environments.

Each filter uses a **circular buffer** architecture — no memory shifting per scan — keeping CPU load minimal and deterministic.

---

## Contents

| FB Name | Filter Type | Best Use Case |
|---|---|---|
| `FB_MovingAverage` | Simple Moving Average (SMA) | General noise reduction |
| `FB_EMA` | Exponential Moving Average | Fast response + smoothing |
| `FB_WMA` | Weighted Moving Average | Emphasize recent samples |
| `FB_MedianFilter` | Median / Order Statistics | Spike & glitch rejection |
| `FB_RateLimiter` | Slew Rate Limiter | Actuator protection |
| `FB_SignalFilter` | Unified wrapper (all above) | Single-block deployment |

---

## Requirements

- **PLC:** Siemens S7-1500 (any CPU)
- **TIA Portal:** V16 or later
- **Language:** SCL (Structured Control Language)
- **OB:** Can be called from any OB (OB1, cyclic OB, etc.)
- **No special libraries required**

---

## Interface — Common Pattern

All FBs follow the same interface convention:

```pascal
VAR_INPUT
    xEnable     : BOOL;     // Enable filtering; FALSE = passthrough + reset
    rInput      : REAL;     // Raw analog input value
    iWindowSize : INT;      // Buffer/window size (filter-specific)
END_VAR

VAR_OUTPUT
    rOutput     : REAL;     // Filtered output value
    xReady      : BOOL;     // TRUE when buffer is full (SMA/WMA/Median)
END_VAR
```

> **Note:** When `xEnable = FALSE`, each FB resets its internal state and passes `rInput` directly to `rOutput`. This allows safe runtime switching between filter modes without output jumps.

---

## Filter Reference

---

### 1. `FB_MovingAverage` — Simple Moving Average (SMA)

Computes the arithmetic mean of the last *N* samples using a circular buffer and a running sum. No per-scan array iteration — the sum is updated incrementally.

**Parameters:**

| Parameter | Type | Default | Description |
|---|---|---|---|
| `xEnable` | BOOL | — | Enable filter |
| `rInput` | REAL | — | Analog input |
| `iWindowSize` | INT | 10 | Window size (1–100) |
| `rOutput` | REAL | — | Filtered output |
| `xReady` | BOOL | — | TRUE when buffer full |

**Formula:**

```
Output(n) = [ Sum of last N samples ] / N
```

**Characteristics:**
- Equal weight to all samples in window
- Output lags behind input by N/2 samples
- Larger window = more smoothing = more lag
- Very low CPU load (no division per element)

**Typical window sizes:**

| Signal Type | Recommended Window |
|---|---|
| Thermocouple / temperature | 20–50 |
| Pressure sensor | 5–15 |
| Flow meter | 10–20 |
| Current / power | 5–10 |

---

### 2. `FB_EMA` — Exponential Moving Average

Applies exponentially decreasing weights to past samples. No buffer required — only the previous output value is stored. Ideal when fast response with smoothing is needed.

**Parameters:**

| Parameter | Type | Default | Description |
|---|---|---|---|
| `xEnable` | BOOL | — | Enable filter |
| `rInput` | REAL | — | Analog input |
| `rAlpha` | REAL | 0.1 | Smoothing factor (0.0–1.0) |
| `rOutput` | REAL | — | Filtered output |

**Formula:**

```
Output(n) = α × Input(n) + (1 − α) × Output(n−1)
```

**Alpha selection guide:**

| Alpha | Effect | Equivalent SMA Window |
|---|---|---|
| 0.05 | Heavy smoothing, slow response | ~39 samples |
| 0.1 | Balanced (default) | ~19 samples |
| 0.2 | Light smoothing, fast response | ~9 samples |
| 0.5 | Minimal smoothing | ~3 samples |

**Characteristics:**
- Minimal memory footprint (no array)
- Infinite impulse response — past samples never fully forgotten
- Best choice when memory is constrained or window must be very large

---

### 3. `FB_WMA` — Weighted Moving Average

Similar to SMA but assigns linearly increasing weights to more recent samples. The newest sample gets weight *N*, the oldest gets weight *1*.

**Parameters:**

| Parameter | Type | Default | Description |
|---|---|---|---|
| `xEnable` | BOOL | — | Enable filter |
| `rInput` | REAL | — | Analog input |
| `iWindowSize` | INT | 8 | Window size (1–16) |
| `rOutput` | REAL | — | Filtered output |

**Formula:**

```
Output(n) = Σ [ weight(i) × sample(i) ] / Σ weight(i)
weight(i) = N for newest, 1 for oldest
```

**Characteristics:**
- Responds faster to changes than SMA at same window size
- More computationally intensive than SMA or EMA
- Good compromise between lag and smoothing

---

### 4. `FB_MedianFilter` — Median / Order Statistics Filter

Collects the last *N* samples, sorts them using insertion sort, and outputs the middle value. Completely immune to impulse spikes regardless of magnitude.

**Parameters:**

| Parameter | Type | Default | Description |
|---|---|---|---|
| `xEnable` | BOOL | — | Enable filter |
| `rInput` | REAL | — | Analog input |
| `iWindowSize` | INT | 7 | Window size, odd recommended (1–15) |
| `rOutput` | REAL | — | Filtered output |

**Characteristics:**
- **Best** filter for rejecting sensor glitches, wire bounce, EMI spikes
- Non-linear filter — does not distort signal shape between spikes
- Use **odd** window sizes (3, 5, 7, 9...) for a clean median value
- Higher CPU load than SMA/EMA due to in-place sort each scan

**Recommended window sizes:**

| Application | Window |
|---|---|
| Remove single-sample spikes | 3 |
| General spike rejection | 5–7 |
| Heavy transient environment | 9–11 |

> **Tip:** Combine Median + EMA for best results: Median removes spikes, EMA smooths residual noise.

---

### 5. `FB_RateLimiter` — Slew Rate Limiter

Limits the maximum rate of change (rise and fall) of a signal per scan cycle. The output tracks the input but cannot change faster than the configured rate. Protects actuators from sudden setpoint jumps.

**Parameters:**

| Parameter | Type | Default | Description |
|---|---|---|---|
| `xEnable` | BOOL | — | Enable filter |
| `rInput` | REAL | — | Analog input |
| `rMaxRiseRate` | REAL | 10.0 | Max units/second rising |
| `rMaxFallRate` | REAL | 10.0 | Max units/second falling |
| `rCycleTimeS` | REAL | 0.1 | OB cycle time in seconds |
| `rOutput` | REAL | — | Rate-limited output |

**Formula:**

```
MaxStep = MaxRate × CycleTime
Output(n) = Output(n-1) + clamp(Input(n) − Output(n-1), −MaxStep, +MaxStep)
```

**Characteristics:**
- Not a noise filter — purely a change-rate constraint
- Asymmetric rise/fall rates supported (e.g. slow open, fast close)
- Must be configured with correct `rCycleTimeS` to function correctly
- Does not affect steady-state accuracy

> **Important:** Set `rCycleTimeS` to match your OB cycle time. For OB1 at 100ms, use `0.1`. You can read this from `OB_CYCL.CYCLE_TIME` if needed.

---

### 6. `FB_SignalFilter` — Unified Wrapper

A single FB that encapsulates all five filters. Select the active filter at runtime via `iMode`. Internally instantiates all FBs; only the selected one processes each scan.

**Parameters:**

| Parameter | Type | Default | Description |
|---|---|---|---|
| `xEnable` | BOOL | — | Enable filter |
| `rInput` | REAL | — | Analog input |
| `iMode` | INT | 0 | Filter selection (see below) |
| `iWindowSize` | INT | 10 | Window for SMA / WMA / Median |
| `rAlpha` | REAL | 0.1 | Alpha for EMA |
| `rMaxRate` | REAL | 10.0 | Rate for RateLimiter |
| `rCycleTimeS` | REAL | 0.1 | Cycle time for RateLimiter |
| `rOutput` | REAL | — | Filtered output |
| `xReady` | BOOL | — | Buffer ready flag |

**Mode selection:**

| `iMode` | Filter |
|---|---|
| 0 | Passthrough (no filtering) |
| 1 | Simple Moving Average |
| 2 | Exponential Moving Average |
| 3 | Weighted Moving Average |
| 4 | Median Filter |
| 5 | Rate Limiter |

---

## Performance Comparison

| Filter | Noise Reduction | Spike Rejection | Response Speed | CPU Load | Memory |
|---|---|---|---|---|---|
| SMA | ★★★★☆ | ★★☆☆☆ | ★★☆☆☆ | Low | Medium |
| EMA | ★★★☆☆ | ★★☆☆☆ | ★★★★☆ | Very Low | Minimal |
| WMA | ★★★★☆ | ★★☆☆☆ | ★★★☆☆ | Medium | Medium |
| Median | ★★★☆☆ | ★★★★★ | ★★★☆☆ | Medium-High | Medium |
| Rate Limiter | ★☆☆☆☆ | ★★★☆☆ | ★★★☆☆ | Very Low | Minimal |

---

## Recommended Filter Combinations

For best results in demanding environments, chain two filters:

```
Raw Signal → Median (spike removal) → EMA (noise smoothing) → Control loop
```

```pascal
// Example: Median pre-filter + EMA smoothing
fbMedian(xEnable := TRUE, rInput := rRawAnalog, iWindowSize := 5);
fbEMA(xEnable := TRUE, rInput := fbMedian.rOutput, rAlpha := 0.15);
rCleanSignal := fbEMA.rOutput;
```

---

## Usage Example

```pascal
// Declarations
VAR
    fbFilter     : FB_SignalFilter;
    rRawAnalog   : REAL;    // From analog input card (e.g. scaled 0.0–100.0)
    rFiltered    : REAL;
END_VAR

// In OB1 or cyclic FC
fbFilter(
    xEnable      := TRUE,
    rInput       := rRawAnalog,
    iMode        := 1,          // 1 = SMA
    iWindowSize  := 15,
    rCycleTimeS  := 0.1         // 100ms OB cycle
);

rFiltered := fbFilter.rOutput;
```

---

## Circular Buffer Architecture

All array-based filters use a **circular (ring) buffer** pattern:

```
Buffer: [ 12.4 | 13.1 | 12.8 | 13.0 | 12.6 ]
                  ↑ Index pointer (oldest slot, next to overwrite)

New sample arrives → overwrite oldest → advance index
No shifting, O(1) write per scan
```

This ensures constant, predictable scan-time contribution regardless of window size.

---

## Signal Flow Diagram

```
Analog Input Card
      │
      ▼
 ┌─────────────┐
 │  Raw REAL   │  e.g. 0.0 – 100.0 (engineering units, already scaled)
 └──────┬──────┘
        │
        ▼
 ┌─────────────────────────────────────┐
 │         FB_SignalFilter             │
 │                                     │
 │  iMode=1 → FB_MovingAverage         │
 │  iMode=2 → FB_EMA                   │
 │  iMode=3 → FB_WMA                   │
 │  iMode=4 → FB_MedianFilter          │
 │  iMode=5 → FB_RateLimiter           │
 └──────────────────┬──────────────────┘
                    │
                    ▼
             rOutput (REAL)
                    │
          ┌─────────┴──────────┐
          ▼                    ▼
     PID Controller        HMI / SCADA
```

---

## Integration Notes

- **Analog scaling:** All FBs operate on pre-scaled REAL values in engineering units (°C, bar, %, m³/h, etc.). Scale your analog input before passing to any filter FB.
- **First scan:** On the first scan after `xEnable` goes TRUE, output equals input. No startup spike.
- **Mode switching:** Changing `iMode` at runtime resets the newly selected filter on its next call. A brief transient (1–2 scan cycles) may occur.
- **Cycle time:** Only `FB_RateLimiter` is cycle-time sensitive. All other filters are cycle-time agnostic — window size implicitly defines the time constant based on your OB period.

---

## Related Libraries

This library pairs naturally with:

- **`FB_StateAlarmManager`** — State-based alarm management with bitmask suppression, hysteresis, on-delay, and priority. Feed filtered signals into alarm manager for dramatically fewer nuisance alarms.
- **`FB_CUSUM`** *(planned)* — Cumulative sum control chart for slow drift detection
- **`FB_KalmanFilter`** *(planned)* — State estimation and sensor validation

---

## License

Internal use. Adapt freely for your project.  
Tested on: **S7-1500 CPU 1515-2 PN**, TIA Portal V18.

---

*Generated for S7-1500 SCL Signal Filtering Library — Advanced Process Control Series*