Raw signal commonly refers to unprocessed data or measurements captured directly from sensors, instruments, or equipment, before any filtering, scaling, aggregation, or interpretation has been applied.
What a raw signal includes
In industrial and manufacturing environments, raw signals typically include:
- Analog readings from field devices, such as temperature, pressure, vibration, or flow sensors
- Digital states from equipment, such as on/off, open/closed, or fault bits
- Electrical waveforms or time-series data captured by PLCs, data historians, or condition monitoring systems
- Unprocessed network or protocol frames captured from OT or industrial communication buses
The raw signal is usually stored as-is, using engineering units or device counts (for example, voltage levels, ADC counts, or raw integer register values), depending on the source device and interface.
What a raw signal is not
A raw signal is not:
- A calibrated, scaled, or validated measurement (for example, a temperature value adjusted for sensor drift)
- A derived metric or KPI, such as OEE, cycle time, or scrap rate
- A summarized or aggregated value, such as hourly averages or batch totals
- A contextualized event record, such as a quality deviation or maintenance work order
Operational use in manufacturing systems
Raw signals are the starting point for many OT and IT workflows, including:
- Real-time control in PLCs and DCS, which read raw sensor values to make control decisions
- Condition monitoring and predictive maintenance, where vibration or current signals are analyzed for anomalies
- Data collection in historians and IIoT platforms, storing raw time-series data for later analysis
- Transformation layers feeding MES, quality, and ERP systems, where raw signals are converted into counts, states, or events aligned with ISA-95 style models
In regulated or high-compliance environments, retaining or being able to reconstruct the raw signal can be important for traceability, investigations, and independent verification of derived values or decisions.
Processing and transformation
Raw signals often go through one or more processing steps before they are used in operations, reporting, or compliance records:
- Filtering and cleaning: Removing noise, invalid readings, or communication artifacts
- Scaling and calibration: Converting counts or voltages into engineering units with calibration factors
- Feature extraction: Turning high-frequency signals into features like peaks, RMS, or frequencies
- Contextualization: Linking processed values to equipment, products, batches, operators, and time windows
After these steps, the signal is no longer considered raw; it becomes processed data or derived information used by MES, quality systems, or analytics tools.
Common confusion
- Raw signal vs raw data: Raw signal is often used when the data directly reflects a physical measurement or real-time equipment state. Raw data can be broader and may include unstructured logs, files, or text not tied to a specific physical sensor.
- Raw signal vs event: A raw signal is continuous or sampled data over time. An event is typically a discrete occurrence derived from the signal, such as a limit exceeded, a machine start, or a fault code logged.
- Raw signal vs KPI: KPIs (such as OEE or yield) are calculated metrics built from multiple processed data points. They never refer to raw signal values.