The motor current waveform looks like a pure sine wave at 60 Hz. It is not. Every structural and electromagnetic asymmetry in the motor modifies the current at specific predictable frequencies. MCSA finds those modifications.
Step 1: Current Measurement at the Motor Terminals
A current clamp or Rogowski coil wraps around one phase conductor at the motor control panel. The sensor captures the instantaneous current as a time-varying signal. No contact with the motor itself is required.
The measurement can be taken while the motor runs at normal operating conditions. The technician does not need to interrupt production, open the motor enclosure, or attach anything to the shaft.
For three-phase motors, measuring all three phases allows imbalance and negative-sequence analysis in addition to standard spectrum analysis. A single-phase measurement is sufficient for most fault frequency identification tasks, but three-phase capture provides a more complete electrical picture of motor health.
| Sensor Type | Typical Use | Notes |
|---|---|---|
| Split-core current clamp | Periodic portable MCSA | Clips onto conductor without disconnection; common for field measurements |
| Rogowski coil | Online continuous MCSA | Flexible, suitable for large conductors and tight panel spaces |
| Solid-core CT (current transformer) | Permanently installed monitoring | Higher accuracy; conductor must pass through core during installation |
Step 2: Signal Conditioning and Sampling
The raw current signal is conditioned before digitization. Conditioning typically involves anti-aliasing filtering to remove frequencies above the analysis range, and amplification to bring the signal into the optimal input range of the analog-to-digital converter (ADC).
The ADC samples the conditioned signal at a rate determined by the Nyquist requirement: the sample rate must exceed twice the highest frequency of interest. For bearing fault detection at frequencies up to 1,000 Hz, a minimum sampling rate of 2,000 Hz is required. High-frequency analysis uses sampling rates between 10,000 and 50,000 Hz.
Sample length determines frequency resolution. A 1-second sample provides 1 Hz resolution. A 10-second sample provides 0.1 Hz resolution. The required resolution depends on how close the expected fault sidebands are to the supply frequency, which is a function of slip.
Resolution and measurement duration
Frequency resolution = 1 / measurement duration (seconds). A motor running at 2% slip on a 60 Hz supply produces rotor bar sidebands at approximately 57.6 Hz and 62.4 Hz. Separating the lower sideband from the supply requires better than 2.4 Hz resolution, which means at least 0.42 seconds of data. In practice, longer windows (2 to 10 seconds) are used to improve signal-to-noise ratio.
Step 3: Fast Fourier Transform
The FFT converts the time-domain current signal into the frequency domain. The output is a spectrum: amplitude on the Y axis (typically in decibels relative to a reference level) and frequency on the X axis in Hz.
The dominant peak in a healthy, lightly loaded induction motor appears at the supply frequency (50 or 60 Hz). Every other peak in the spectrum has a physical origin inside or connected to the motor. The task of MCSA is to identify which peaks indicate developing faults and which represent normal operating characteristics.
Spectral resolution is a direct function of the number of samples and the sampling rate. A 10-second capture at 10,000 Hz produces 100,000 samples and a frequency resolution of 0.1 Hz. Higher FFT resolution allows the analyst to separate closely spaced sidebands that would merge into noise at lower resolution.
Step 4: Fault Frequency Identification
Each fault type in an induction motor produces current modulation at mathematically predictable frequencies. Those frequencies are calculated from the motor's nameplate data and the measured operating slip. The analyst or automated monitoring system then examines the spectrum for amplitude peaks at those calculated locations.
Slip is the key variable. It is not constant and must be measured at the time of the MCSA capture, not assumed from nameplate values.
Where f1 = supply frequency (Hz), P = number of poles, Nr = measured shaft speed (RPM).
Example: a 4-pole motor on a 60 Hz supply running at 1,746 RPM has a synchronous speed of 1,800 RPM and a per-unit slip of 0.0333. The rotor bar fault sidebands appear at 60 x (1 +/- 0.0667) = 56.0 Hz and 64.0 Hz.
| Fault Type | Frequency Location | Physical Cause |
|---|---|---|
| Broken rotor bar | f1 x (1 +/- 2s) | Asymmetric rotor resistance modulates torque at twice slip frequency |
| Static eccentricity | f1 +/- (Ns / 60) | Fixed air gap irregularity produces spatial harmonics in flux |
| Dynamic eccentricity | f1 +/- k x (Ns / 60) x (1 - s) | Rotating air gap variation; pattern changes with load |
| Stator winding fault | Supply harmonics (3rd, 5th, 7th) | Inter-turn shorts alter winding impedance symmetry |
| Bearing fault (outer race) | f1 +/- BPFO | Recurrent impact at ball pass frequency modulates current via air gap |
Step 5: Amplitude Assessment and Trending
A peak at the correct fault frequency location is not, by itself, evidence of a fault. Every induction motor has some residual asymmetry and produces small peaks at fault frequency locations under normal operating conditions.
The diagnostic indicator is amplitude relative to a baseline taken when the motor was in known good condition. A sideband rising from -60 dB to -45 dB relative to the supply frequency peak over six months indicates a developing rotor bar fault. A static reading of -45 dB without a baseline has little diagnostic value because there is no reference point for what is normal for that specific motor.
General severity thresholds used in practice:
- Greater than -54 dB below supply frequency: monitor; possible early-stage rotor asymmetry
- -45 to -54 dB: elevated; schedule inspection at next planned outage
- -36 to -45 dB: significant; plan repair
- Above -36 dB: critical; risk of rapid deterioration
These thresholds are starting points. Motor-specific baselines and trend rate are more useful than absolute values alone. A motor that has always read at -40 dB with no upward trend is less concerning than one that moved from -58 dB to -46 dB in 90 days.
What Affects MCSA Accuracy
Load level
Fault sidebands weaken at low load because slip decreases, moving the sidebands closer to the supply frequency where they are harder to separate from the main peak and its noise floor. Below 40% of rated load, reliable rotor bar detection requires longer measurement windows and more careful spectral analysis.
Some analysts set a minimum load threshold of 50% for routine MCSA. For motors that regularly operate at low load, continuous monitoring with longer averaging windows compensates for the reduced sideband amplitude.
Measurement window length
Frequency resolution equals 1 divided by measurement duration in seconds. To separate a sideband at 56.4 Hz from the supply at 60 Hz when slip is 6%, you need resolution better than 3.6 Hz, achieved with a measurement of at least 0.3 seconds. For small slip values in the 1 to 2% range, the sidebands sit much closer to 60 Hz and require 2 to 10 second measurements for reliable separation.
Supply quality
Harmonic distortion in the supply voltage produces corresponding harmonic content in the motor current. The 5th and 7th supply harmonics (300 Hz and 420 Hz on a 60 Hz system) are particularly common in facilities with variable frequency drives and switching power supplies. These supply-side harmonics appear in the MCSA spectrum and can mask or be mistaken for fault signatures. A power quality analyzer run in parallel with MCSA helps identify which spectral components originate in the supply versus the motor.
Variable frequency drives
VFD-driven motors present a different analysis challenge. The drive's carrier frequency and its harmonics overlay the current spectrum. The fundamental supply frequency presented to the motor changes with speed command. Specialized algorithms separate drive artifacts from motor fault signatures. Standard MCSA analysis methods designed for line-fed motors cannot be applied directly to VFD-driven motors without modification.
Online vs Periodic MCSA
Periodic MCSA
A technician visits the motor, clamps a portable analyzer to the motor panel, takes a measurement of defined duration, and compares the result to the previous measurement from the same motor. The comparison reveals whether any fault frequency amplitudes have increased since the last reading.
Periodic MCSA is appropriate for non-critical assets. The sampling interval defines the maximum detectable fault progression rate: a rotor bar fault that develops and causes failure in 60 days will not be detected by a quarterly measurement schedule.
Online continuous MCSA
Permanently installed current transducers stream data to a monitoring system continuously. The system calculates fault frequencies in real time, trends sideband amplitude over time, and generates alerts when amplitudes cross defined thresholds or when trend rate accelerates beyond expected bounds.
Continuous MCSA provides the earliest possible detection because every operating hour generates data. It is standard practice for critical motors in process industries where an unplanned motor failure causes production loss, safety risk, or equipment damage.
| Attribute | Periodic MCSA | Online Continuous MCSA |
|---|---|---|
| Fault detection lead time | Limited by measurement interval | Maximum possible; every operating hour analyzed |
| Data density | Snapshot at each visit | Continuous time series; trend rate visible |
| Labor requirement | Technician visit per measurement | Automated; review only on alert |
| Installation cost | Low; portable instrument only | Higher; permanent sensors and monitoring infrastructure |
| Best fit | Non-critical motors; established baseline programs | Critical motors; process lines where failure has high consequence |
Continuous MCSA, Automatically
Tractian runs continuous MCSA on connected motors, computing fault frequencies from nameplate data and trending sideband amplitude automatically.
See Electrical Monitoring