How EMPATH Grades a Squirrel-Cage Rotor Using the ORNL Method, 57% of Field Tests Marginal or Unacceptable Rotor Grades.
If you've ever pulled a rotor grade off an EMPATH report and wondered what that single number is actually telling you, this is for you. Behind that grade sits decades of work that started in a nuclear reliability lab, a bit of electromagnetic theory, and a very specific frequency component hiding in the motor's current spectrum: the pole pass frequency sideband.
Where EMPATH Came From
Electrical Signature Analysis (ESA) — the technique behind EMPATH — didn't start out as a rotor bar diagnostic. It was developed at Oak Ridge National Laboratory (ORNL) in the 1980s, originally to evaluate bearing and gear wear in motor-operated valves for the nuclear power industry. Researchers there discovered that a motor's own current and voltage waveforms carry a surprising amount of information about the mechanical condition of everything it's connected to — not just the motor itself, but the load, the coupling, and the bearings on both ends.
As the technique matured, it was licensed out of ORNL into general industry, and by the early 1990s the commercialized version of the toolset had become known as EMPATH. Around the same period, a parallel branch of research — led independently in the UK — focused specifically on using current spectra to find broken and cracked rotor bars, a technique usually called Motor Current Signature Analysis (MCSA). EMPATH inherited both lineages: it evaluates rotor bar condition using the same current-spectrum principles as MCSA, layered on top of the broader mechanical diagnostics ESA was originally built for.
Why the Rotor Bars Fail in the First Place
A squirrel-cage rotor is exactly what it sounds like: a cylindrical cage of parallel conductor bars, shorted together at each end by a ring. When the stator's rotating magnetic field sweeps across those bars, it induces current in them, and that induced current is what produces torque.
Two things put stress on the bars over the life of a motor:
- Thermal cycling. Every start pushes a large, high-frequency current through the rotor bars for a brief period, generating substantial I²R heating. Frequent starts, or long acceleration times on high-inertia loads, make the bars increasingly brittle over time.
- Mechanical and inertial stress. In larger motors, the rotor bars and the end rings are effectively separate masses, brazed or welded together rather than cast as one piece. That construction concentrates bending stress right at the joints — which is exactly where bars tend to crack or fracture first.
Once a bar fractures, current can no longer flow through it cleanly. The load that bar used to carry gets redistributed into its neighbors, which heats and stresses them further — a self-reinforcing failure pattern. Left alone, one broken bar tends to become several.
What a Broken Bar Does to the Current Waveform
Here's the part that makes this detectable from the motor leads without ever opening the motor up.
A broken bar creates an asymmetry in the rotor's magnetic field. As the rotor turns, that asymmetry passes each stator pole in turn, producing a small but rhythmic pulsation in torque — and, in turn, a rhythmic pulsation in the stator current. This pulsation repeats once per pole passage, which is where the term pole pass frequency comes from.
The pole pass frequency turns out to be a very simple multiple of a quantity every reliability engineer already knows: slip.
- Synchronous speed is fixed by the line frequency and the number of poles: Ns = 120 × f / P
- Slip is the gap between synchronous speed and actual rotor speed: s = (Ns − Nr) / Ns
- Pole pass frequency (PPF) works out to twice the slip frequency: PPF = 2 × s × f
Because this pulsation modulates the stator current, it doesn't show up as its own isolated peak in the spectrum — it shows up as sidebands straddling the line frequency peak, spaced out at exactly ± PPF on either side. On a 50 Hz motor with modest slip, that might put the sidebands just a fraction of a hertz to either side of 50 Hz, which is why EMPATH and similar tools need fine frequency resolution (long capture windows) to resolve them cleanly.
The key diagnostic quantity is the amplitude of those sidebands relative to the line-frequency peak, expressed in dB. A healthy rotor has a very deep, very small sideband — the current is barely modulated at all. A rotor with fractured or high-resistance bars has a shallower, larger sideband, because the current pulsation is more pronounced.
The Manufacturing Side: Aluminum Die-Cast Rotor Defects
Not every fault a rotor grade flags develops in the field. A large share of rotor bar problems on today's motors are baked in on day one, because most rotors below 1000 kW are built with aluminum die-cast bars and end rings rather than fabricated copper. Die casting is fast and cheap compared to brazing individual copper bars — but it's also a process where the metal has to fill a cavity full of thin, deep slots almost instantly, and that leaves plenty of room for defects:
- Porosity. Gas bubbles trapped during the high-pressure injection, or gas released from mold coatings and the molten aluminum itself, get frozen into the casting as it solidifies. Porosity reduces the effective cross-section of the bar, which raises its electrical resistance in that spot.
- Shrinkage voids. As aluminum cools and solidifies, it contracts. If the mold design doesn't feed enough molten metal into that shrinking volume, it leaves a void behind — mechanically and electrically similar to porosity, but usually larger and more localized.
- Cold shuts. These occur when two streams of molten aluminum meet inside the mold but have already cooled enough that they don't fully fuse. The result looks like a bar, but it can carry a weak, high-resistance seam right where the two flows met.
- Incomplete end-ring bonding. The end rings are supposed to be one continuous casting with the bars. If flow or temperature isn't controlled well, individual bars can end up poorly bonded — or in the worst cases, completely disconnected — from the end ring even though the rotor looks intact from the outside.
None of these are things a rotor "develops" through wear. They're already there when the motor ships, sitting dormant, and they matter for two connected reasons.
The heat developed in the rotor is transmitted across the air gap and degrades the Stator Winding Insulation.
First, they're a leading cause of early-life failures: a rotor with distributed porosity or a cold shut has weak points built in, and normal thermal cycling from repeated starts is often enough to turn a manufacturing flaw into a full crack far sooner than you'd expect from a defect-free rotor. This is why a brand-new motor can occasionally show a poor rotor grade on its very first baseline test — the fault didn't develop in service, it was cast in.
Second, they complicate the current-spectrum interpretation. Distributed porosity spread across many bars doesn't always mimic a single broken bar's sharp signature; it can instead raise the general noise floor around the fundamental or produce a milder version of the classic sideband pattern. This is part of why the literature explicitly calls out die-cast porosity as a source of false rotor fault indications — a rotor can register as marginal on ESA even though no single bar is fully broken, simply because resistance is unevenly elevated across the cage. It's also why a first-time reading on a new rotor should be treated as its true baseline rather than assumed to be a "perfect" grade 1: distributed manufacturing variation is normal enough that even healthy die-cast rotors don't all read identically.
From dB to Grade: The ORNL Severity Scale
This is where the single "rotor grade" number comes from. The sideband-to-fundamental amplitude, measured in dB, is mapped onto a severity scale that traces back to the original ORNL/EMPATH work and has since become a de facto industry convention:
| Sideband level (dB below fundamental) | Condition | Typical action |
| >60 dB Grade 1 | Excellent | None |
| 54–60 dB Grade 2 | Good | None |
| 48–54 dB Grade3 | Moderate | Trend at next scheduled test |
| 42–48 dB Grade 4 | Cracked bar or high-resistance joint suspected | Increase test frequency, trend closely |
| 36–42 dB Grade 5 | Broken rotor bar(s) likely | Confirm with vibration; plan repair |
| 30–36 dB Grade 6 | Multiple cracked/broken bars, possible ring damage | Repair or replace as soon as possible |
| <30 dB Grade 7 | Severe rotor fault | Repair or replace immediately |
Note the direction: a Larger dB number is worse, because it means the fault sideband is closer in amplitude to the fundamental — less separation means more current modulation, which means a more damaged rotor. When this scale gets compressed into a simple ordinal grade (1 through 7, best to worst) for reporting purposes, grade 1 corresponds to that ">60 dB, excellent" band, and grade 7 corresponds to the "<30 dB, severe" band at the bottom.
Can You See a Rotor Defect in the Common-Mode Current?
There's one more angle worth covering, because it's a different measurement entirely from the phase-current spectrum EMPATH normally works with, and it answers a question that comes up whenever someone's already got a common-mode current probe on a motor for VFD bearing-current work.
In a healthy motor, the three phase currents sum to zero, so the common-mode current should stay flat around line frequency with no real structure of its own. A rotor defect changes that. The rotor's back-EMF, generated by the rotating magnetic flux, doesn't stay perfectly balanced when there's an impedance imbalance from a cracked or broken bar — and that imbalance distorts the otherwise-zero sum. The signature isn't a clean new spectral peak the way a pole pass sideband is; it shows up as a small, repeating inflection in the common-mode sine wave, appearing just before or after each line-frequency peak, cycle after cycle. That recurring ripple near line frequency is the classical rotor-defect fingerprint in this particular measurement.
It's worth being precise about where to look, because common-mode current is far better known for a completely different purpose: VFD-driven bearing-current diagnostics. There, the signal of interest is the fast switching noise from the drive's PWM output — sharp transients synchronized to the switching frequency, often resonating in the low-MHz range. That switching noise is a property of the drive electronics, cabling, and grounding, not the rotor's condition, and single large spikes there usually point to a cable-shielding or grounding issue rather than anything mechanical. It sits in a completely different part of the frequency spectrum from the slow, line-frequency ripple a rotor defect produces.
So the two phenomena can coexist on the exact same current waveform without interfering with each other, provided the measurement bandwidth covers both ends of the spectrum: a Rogowski coil rated for 10 megahertz of bandwidth can resolve the fast switching content and the slower line-frequency distortion in one capture. In everyday practice, though, standard MCSA/ESA on the phase current — or negative-sequence current analysis for inverter-fed motors — remains the primary tool for rotor bar diagnosis. Common-mode current is a secondary, much less commonly used window into the same underlying rotor asymmetry, useful mainly when the instrumentation is already in place for another reason.
Do Rotor Defects Cause Shaft Currents Too?
It's worth pulling one more thread, because it connects rotor condition to a completely different failure mode: bearing damage from circulating shaft current.
A healthy rotor produces a magnetically balanced field as it spins. A broken or cracked bar — like an eccentric air gap or residual shaft magnetization — breaks that symmetry. The result is axial shaft flux: flux that leaks along the length of the shaft instead of staying confined to the normal air-gap path. As the shaft turns through that leakage flux, it induces a small voltage between the two shaft ends. If that voltage is high enough to break down the thin oil film inside the bearings, it discharges as current through the bearing balls and races on its way back to the frame — the circulating current itself.
Stated in IEEE1415 Any Break in Magnetic Symmetry creates Shaft Current.
Since that Current flows Rotor Ends through the Motor Frame, grounding the Shaft with brushes is ineffective.
That discharge is genuinely damaging. It causes pitting, cratering, and metal transfer on the bearing surfaces, which increases friction and heat, degrades the lubricant, and accelerates the bearing toward failure. So a rotor bar problem doesn't necessarily stay contained to the rotor — it can manifest downstream as a bearing that's wearing out faster than it should.
That said, rotor defects are far from the leading cause of shaft voltage in the field today. On VFD-fed motors — the majority of installed motors at this point — the dominant driver is the fast-switching common-mode voltage from the drive itself, which capacitively couples onto the rotor and shaft regardless of the rotor's actual condition. A broken bar is one well-documented mechanism among several for line-fed motors, sitting alongside air-gap eccentricity and magnetic asymmetry in the rotor or stator iron, rather than the primary suspect whenever shaft current shows up.
This is exactly why it's worth looking at DE and NDE bearing findings side by side with rotor grade rather than treating them as unrelated line items on a report. A motor with a poor rotor grade and a bearing defect isn't automatically two independent problems — a degrading rotor can be a contributing cause of the bearing wear. If that link were strong across a fleet, you'd expect bearing defect rates to climb noticeably as rotor grade worsens. Worth keeping an eye on that pattern the next time a larger dataset is available to check it.
And on Generators?
Everything above carries over to generators, and it's actually one of the oldest documented problems in the industry — the mechanism was first described by Alger and Samson back in 1924. Any magnetic asymmetry, stator or rotor, still creates a small net axial flux linkage through the shaft, still induces a voltage from one shaft end to the other, and still discharges through the bearings if that voltage is high enough to break down the oil film. The physics doesn't change.
What changes is the machine architecture, and that makes the problem more consequential:
- Multi-bearing shaft trains amplify it. A large turbine-generator isn't a single rotor on two bearings — it's often a turbine, a generator, and an exciter all coupled together on a common shaft with several bearings in between. That shaft-and-bearings arrangement effectively behaves like a transformer secondary loop, so an asymmetry anywhere along the train can drive circulating current through multiple bearing sets, not just the two closest to the fault.
- The rotor-defect mechanism looks different depending on generator type. The squirrel-cage rotor bar story from earlier in this article applies directly to induction generators, common in wind turbines — a broken or cracked bar creates the same asymmetric flux there as it does in a motor. Large synchronous generators, though, don't have that same squirrel-cage structure in their main field circuit. There, the closer analogue to a "rotor defect" is a field-winding insulation asymmetry, a turn-to-turn short in the rotor winding, or a fault in the damper winding — which is itself a squirrel-cage-like structure and behaves much like a broken rotor bar when it's damaged.
- Generators pick up a few extra sources of their own. Static excitation equipment can inject voltage ripple directly onto the rotor winding, and brush-type excitation systems add another possible asymmetry path. Large turbines also accumulate electrostatic charge independent of any magnetic effect at all — steam impinging on turbine blades, or water droplets striking wind turbine blades — which raises shaft potential on its own.
- Mitigation is standard design practice, not an afterthought. Because this failure mode is so well established on generators, insulated bearings (typically at the non-drive end), Insulated Housings and shaft grounding brushes are essentially mandatory rather than optional add-ons. Monitoring usually means watching the voltage across the oil film through a brush contact, since a lifted ground brush or degraded bearing insulation lets shaft voltage build up unchecked until it arcs.
The devices are not fixing the Root Cause being magnetic symmetry and effectively are a bandaid. MCA Testing can Identify the Source.
What About Permanent Magnet Motors?
PM motors have their own version of this problem, and in some designs it's arguably worse than in an induction motor, because the asymmetry can be built into the rotor topology itself rather than only arising from a fault.
The baseline issue is intrinsic design asymmetry. Unlike a squirrel-cage rotor, which starts out symmetric and only becomes asymmetric if a bar cracks, some PM rotor designs are inherently asymmetric from the outset. Consequent-pole PM motors — a design that uses permanent magnets for only half the poles and plain iron for the other half, to cut magnet cost — have been shown to produce substantially higher intrinsic shaft voltage than conventional interior PM motors, purely as a consequence of their structural asymmetry. That means the shaft-voltage risk exists from day one, before any degradation ever occurs.
The fault-driven mechanism, the closer analogue to a broken rotor bar, is uneven demagnetization. Magnets can lose strength from overheating, overcurrent events, or long-term aging, and that loss is rarely perfectly uniform across every magnet. When one or a few magnets weaken more than the rest, it creates the same kind of asymmetric flux imbalance a cracked bar creates in an induction rotor, which again drives a circular flux around the machine and induces an axial shaft voltage. Partial demagnetization also leaves its own characteristic sideband signature in the stator current spectrum — conceptually similar to the pole pass sidebands described earlier, though tied to the demagnetization pattern rather than to slip.
This tends to matter more in practice for PM motors because of how they're typically deployed. EV traction motors and similar PMSM (Permanent Magnet Synchronous Motors) applications almost always run through inverters, so they already carry VFD-driven common-mode shaft voltage. That gets layered directly on top of any intrinsic or demagnetization-driven asymmetry voltage, and the two sources add together at the bearing. Bench testing has found that the deep-groove ball bearings most PMSMs use often can't maintain a stable enough oil film at low speed to fully insulate against the combined voltage, which puts electrical bearing corrosion risk across nearly the whole operating range rather than just at speed extremes. IEC 60034-17 sets a guideline limit of roughly 500 mV for intrinsic shaft voltage, while actual oil-film breakdown in bearings is reported to begin somewhere in the 5–30 V range — not a large margin once multiple voltage sources are stacking on the same shaft.
The issue with using Shaft Voltage vs Common Mode Current is the ability to determine Circulatory Current to Pulse Wave Modulation peaks.
Mitigation looks similar to what's already been covered for induction motors and generators — insulated or ceramic/hybrid bearings, or EMF Cores — plus one technique.
There's a real patent — US 6,218,757, "Minimizing Shaft Currents in Dynamoelectric Machines" — that documents this exact shorted-secondary principle, and it cites earlier foundational work by Costello ("Shaft Voltages & Rotating Machinery," IEEE Paper PCIC-91-13, 1991) and Walker ("Preventing Motor Shaft-Current Bearing Failures," Plant Engineering, 1990).
It's a conductor built into or alongside the rotor shaft itself.
How it's really built: The shaft is bored out (hollow), and an insulated conductor is run through that bore, then electrically connected — solidly, not through a switch — to the shaft at both ends. In an alternative version covered by the same patent, instead of drilling all the way through, the conductor runs through a partial bore covering just the section of shaft exposed to the problematic flux, or it's simply mounted alongside the shaft on the outside and bolted to it at both ends.
Why it works, per the patent's own explanation:
- The rotating shaft, plus the return path through the bearings and frame, forms a loop. Magnetic asymmetry in the stator or rotor produces a stray flux that links this loop, inducing an EMF along the shaft — exactly the mechanism we discussed.
- The inserted conductor is wired so it forms its own closed loop, tightly coupled to that same stray flux, running parallel to the shaft.
- Because that conductor is solidly connected (shorted) at both ends, it behaves as a shorted transformer secondary. Any current the stray flux tries to induce in it gets opposed by Lenz's law — the loop's own induced current creates a field that fights the change in flux.
- The patent states plainly that with this conductor in place, the current through the bearings drops to an "insignificant magnitude" — the low-resistance path through the insulated conductor essentially steals the current away from the bearings, rather than reducing the underlying flux to zero.
It provides a very low-resistance alternate path for any induced current (so current preferentially flows through the conductor instead of through the bearings), and the current flowing in that conductor generates an opposing field that suppresses the net flux somewhat. Both effects point the same direction: less current ends up going through the bearings.
It senses the circular flux caused by the asymmetry and, when shorted, drives a canceling current that suppresses the shaft voltage at its source rather than simply isolating the bearing from it.
A Few Practical Caveats
A rotor grade is a maintenance-planning signal, not an exact bar count. Several factors can shift the sideband amplitude independent of actual bar condition:
- Load matters. Broken-bar sidebands are far easier to see under load; a lightly loaded or unloaded motor can mask a real fault. Loads below 10% are difficult to detect. High Sampling rates of EmPower help at low loads.
- Measurement point matters. Readings taken from panel current transformers, rather than directly at the motor leads, tend to dampen the sideband peaks — worth remembering when comparing grades across a fleet with mixed instrumentation.
- Manufacturing artifacts exist. As covered above, die-cast rotors can show a small baseline sideband from uneven bar resistance — porosity, shrinkage voids, cold shuts — even when no bar is actually broken, so a single reading is less informative than a trend over time.
That last point is really the philosophy behind the whole grading system: EMPATH's real value isn't the grade on any one report, it's watching that grade — or the underlying dB number behind it — move over successive tests, and catching a rotor that's sliding from "good" toward "plan a repair" long before it becomes an unplanned outage.
Measured Rotor Grades (Field Testing)
Blog Free Advice on Electric Motor Reliability and Energy Savings
Feel free to Copy Link and use on your Website (Blue Chain Button)- MENU ABOVE
Gain Electric Motor Reliability & Return Energy Savings