AutomateDirectUSAMotion LibraryShop the parts catalog

Troubleshooting

Why a Servo Motor Oscillates Between Two Positions

Servo and stepper motion Published 7 min read

Short answer

A servo that hunts in a stable 30-50 degree band and restarts on a light shaft nudge has a working position loop running on wrong motor or load data. Fix the configuration — engineering units, load description, pole count, resolver excitation — before swapping phases or retuning: in the documented case, converting metric values into the drive's lbf-ft fields and correcting the load entry stopped it.

On this page
  1. What the oscillation signature means
  2. The checks, cheapest first
  3. One wrong unit, by the numbers
  4. Resolver wiring and pole count
  5. When to swap two motor leads
  6. Proving the fix
  7. Questions asked

A servo that swings in a stable 30-50 degree band, settles, and starts swinging again the moment you turn the shaft by hand has a closed, working position loop running on the wrong motor or load data. This is a configuration fault, not a phasing fault, and almost never a tuning fault. In one documented repair, metric numbers from a motor datasheet had been typed into drive fields that expected lbf-ft-based values, and the load entry described a massive inertia on what was really a free-running shaft; converting the values and correcting the load stopped the oscillation with no gain changes. Work the checks below in order — every one of them costs only time — and keep phase swapping and retuning for the symptoms that actually call for them.

What the oscillation signature means #

A bounded, repeatable swing tells you the loop is closed and the feedback is usable. Genuinely reversed feedback does not hunt — the axis runs away or faults the instant it is enabled, because every correction pushes it further from the target. What is broken here is the model the loop computes with: torque constant, pole count, load. Corrections leave the drive scaled by data that does not describe the motor, so the shaft parks alternately at two positions instead of converging, and a light nudge injects fresh error that restarts the cycle.

Read the whole symptom before changing anything:

Observation What it tells you
Bounded 30-50 degree swing at a steady rhythm Loop closed, feedback usable, motor model wrong
A light shaft turn restarts it The drive cannot hold position against small errors with this data
Swapping two motor leads changes nothing Phase order is not the cause — stop pursuing it
The same motor runs correctly on another drive Motor and feedback are serviceable; suspect this drive's setup
Buzz that grows, or swings that widen until a fault Gain-driven instability — a different fault, covered in check 6

The checks, cheapest first #

  1. Audit the units of every motor parameter. In the commissioning software, write down each value together with the unit shown next to its field, then compare against the datasheet of the exact motor. Two drives of the same family can even carry different unit systems for different markets, so read the field and never assume it. Confirming check: any field where a metric number was entered unchanged into a non-SI field, or the reverse, is the prime suspect.
  2. Correct the load description. A shaft running free on the bench carries essentially no inertia and no process friction; an entry claiming a heavy load makes the drive schedule its gains for a machine that does not exist. Confirming check: compare the load entry with what the shaft actually carries today, not what the machine will eventually do.
  3. Fix the pole count. A wrong entry mis-times commutation across the whole rotation; the arithmetic is in the pole count section below. Confirming check: the nameplate or datasheet of that exact motor, and where the documentation is ambiguous, the number in writing from the motor's manufacturer.
  4. Verify the resolver connections. Method and reference requirements are in the resolver section below. Confirming check: every connection matches both the drive manual and the motor's feedback documentation, terminal names included.
  5. Check the feedback coupling. Slip between shaft and feedback device makes the reported angle lag the real one. Confirming check: with the axis locked out, scribe a match mark across coupling and shaft, rotate by hand, and confirm the reported angle tracks exactly. Hand work on the shaft belongs inside your lockout and tagout practice — 29 CFR 1910.147 in the US requires stored electrical and mechanical energy to be isolated and locked before hands go in — and inside the machine's risk assessment.
  6. Only now, gains. Halve the integral gain and watch. Amplitude that scales down with the gain is tuning; a 30-50 degree band that ignores the change is the model, and you go back to checks 1 through 4.
  7. Phase order last. Swap two motor leads only when the axis runs away or faults on enable; that reversed-direction signature is what a phase swap actually fixes.

One wrong unit, by the numbers #

Calculated example for the inputs shown, not a measured result. The motor datasheet states a torque constant of 0.75 N-m/A. The drive field expects lbf-ft/A. The number 0.75 is entered unchanged.

The conversion basis: 1 lbf = 4.448 N and 1 ft = 0.3048 m, so 1 lbf-ft = 1.356 N-m.

  • The drive now believes the torque constant is 0.75 lbf-ft/A x 1.356 = 1.02 N-m/A, overstated by 1.02 / 0.75 = 1.36, or 36 percent.
  • When the position loop demands 1.5 N-m, the drive commands 1.5 / 1.02 = 1.47 A.
  • The motor actually delivers 1.47 A x 0.75 N-m/A = 1.10 N-m.
  • The correction arrives 0.40 N-m light — 26 percent short — so the loop's effective stiffness sits near 0.74 of what was tuned.

One wrong field alone usually reads as sluggish response, not oscillation. Stack it on a load entry that claims a massive inertia on a free shaft and the drive's computed gains belong to a different machine: the loop still corrects, with the wrong magnitude, and settles into a bounded limit cycle between two positions instead of at the target. The decision follows the arithmetic: read the unit on the field, convert the number, then describe the load the shaft actually carries.

Resolver wiring and pole count #

Finding the excitation winding #

A resolver is a rotating transformer. The drive drives one excitation winding with an AC reference, and two signal windings mounted 90 degrees apart return amplitudes proportional to the sine and cosine of the shaft angle. On an ohmmeter, the excitation winding is the pair with the distinctly lower resistance — that is how it was identified in the documented machine — but the resistance reading only finds the winding; the pinout still has to be confirmed against the drive manual and the motor's feedback documentation before the loop is powered, because signal-winding naming is not standardized between makers.

The excitation reference itself has to match what the resolver is wound for in voltage, frequency, loading and waveform; a mismatched reference weakens or distorts the angle signal before tuning is even relevant. If the feedback type itself is uncertain on an older axis, the wiring and measurement signs that separate a resolver from an encoder are covered in identifying geared resolver and encoder feedback on old servos.

Pole count #

Pole pairs equal poles divided by two, and electrical angle equals pole pairs times mechanical angle. A drive configured for 6 poles (3 pairs) on a 4-pole motor (2 pairs) tracks an electrical angle 1.5 times the true one, so commutation timing drifts as the shaft rotates — fix it from the nameplate before any tuning. In the documented repair the oscillation stopped after the unit and load corrections alone, so the pole count was corrected as a configuration item without being proven the trigger: evidence assigns the cause, not the symptom.

When to swap two motor leads #

A phase swap fixes reversed rotation, and its signature is unmistakable: enable the axis and it runs away, or the drive faults within moments, because each correction drives the motor further from the target. The documented machine did the opposite — it held position between swings and responded to a nudge — and swapping two leads changed nothing. Decision: phase order is the suspect on runaway at enable; on a bounded hunt, spend ten minutes on the configuration instead.

Proving the fix #

Command the same free-running moves that used to oscillate. The motor settles at the target and stays there. Then the disturbance test, hands clear of the pinch points: give the shaft a light twist and confirm the drive pulls straight back to target and holds. Repeat three or four times — every nudge should end at the target, never back in the old cycle. And make the units audit a standing step whenever a drive is replaced or a parameter set is re-entered: reading the unit next to each field costs about a minute and removes this entire fault class.

Questions asked

Can a wrong pole count cause this oscillation by itself?

It can contribute. With the drive set for 6 poles (3 pole pairs) on a 4-pole motor (2 pairs), the computed electrical angle runs 1.5 times the true angle, so commutation timing drifts continuously as the shaft turns. Fix it from the nameplate before tuning, but in the documented repair the oscillation stopped after the unit and load corrections alone, so treat pole count as a configuration item to verify, not the proven trigger.

How do I find a resolver's excitation winding with a multimeter?

Set the ohmmeter to a low range and measure resistance between every pin pair, writing each value down. The excitation winding is the pair with the distinctly lower resistance; the two signal windings sit 90 degrees apart and read similar to each other. Then confirm the full pinout against the drive manual and the motor's resolver documentation, because terminal naming is not standardized between makers, and check that the drive's excitation voltage and frequency match the resolver's rating.

Why does a small shaft nudge restart the oscillation?

At rest the loop sits at whatever position its mis-scaled corrections arrived at. Your nudge injects a position error that the drive chases with torque commands it believes are larger than they are, under gains scheduled for a load that is not there, so the loop re-enters the same limit cycle between the two positions. A correctly configured drive absorbs the same nudge and returns to target.

When is swapping two motor leads the correct fix?

When the axis runs away or faults the instant it is enabled — a correction that pushes the motor further from the target is the signature of reversed direction. A bounded 30-50 degree hunt that holds and repeats is not that signature; in the documented case a phase swap changed nothing and the configuration corrections cured it. Verify motor data, load and pole count before touching phase order.

Keep reading

  • Selection guide

    Choosing Servo or Stepper for a 1% Speed Regulation Pinch Wheel

    Pick the servo: behind a 45:1 planetary reducer, a velocity-mode servo meets a 1% wheel-speed tolerance with margin, because 200-300 lb-in at 80 rpm needs only 4.9-7.4 lb-in at the motor and the loop corrects load-induced error. A stepper qualifies only when its torque-speed curve clears the load with margin at the required motor speed and a measured ripple test holds the wheel inside 0.8 rpm.

    Servo and stepper motion

  • How-to

    Following a Master Axis With a Stepper Traverse

    Gear the traverse stepper to measured spool rotation, not to a fixed speed ratio: commanded traverse speed is v_t = p × n_s, so pitch stays constant as the spool fills. The follower ratio comes from pitch, screw lead and drive resolution — 25 step pulses per 64 master counts in the worked example — and master acceleration must be limited so the amplified slave ramp cannot stall the open-loop stepper.

    Servo and stepper motion

  • How-to

    How to Identify Geared Resolver and Encoder Feedback on Old Servos

    Treat the resolver plus its geared 1:4 and 1:80 channels as one integrated coarse-and-fine, multi-turn position package: the resolver reads absolute angle within one shaft revolution, and the geared channels count revolutions above it. Identify each channel by resistance checks, ratio counting and waveform capture before choosing a replacement drive, because the resolver model number alone establishes neither commutation nor multi-turn position recovery.

    Sensors, encoders and machine vision on the machine