AutomateDirectUSAMotion LibraryShop the parts catalog

How-to

How to Identify Geared Resolver and Encoder Feedback on Old Servos

Sensors, encoders and machine vision on the machine Published 7 min read

Short answer

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.

On this page
  1. What the three devices are doing
  2. Bench identification, step by step
  3. The ratio check with numbers
  4. Reading each channel's waveform
  5. What to record before you close it up
  6. The retrofit decision
  7. Questions asked

Start with an ohmmeter on the unpowered feedback connector: a device with isolated winding pairs of similar low resistance and no supply pin is a resolver, and devices with supply, ground and signal pins are encoders. Then count ratios by turning the marked motor shaft exactly 20 revolutions: in this case the geared devices came out near 1:4 (5.00 device turns in 20) and 1:80 (a quarter device turn, 90°, in 20). Treat the assembly as one integrated coarse-and-fine position package: the resolver reports absolute angle inside a single shaft revolution, and the two geared devices, turning at roughly 1:4 and 1:80, count revolutions on top of it. The resolver's model number alone tells you neither how the drive commutates the motor nor how the controller recovers position after several turns, so identify every channel on the bench before you select a replacement drive. The procedure below needs a multimeter, a marker, an oscilloscope with isolated inputs, and a few hours. At the end you will know which signals close the commutation and servo loops, which ones hold multi-turn position, and whether a modern drive can take the motor as-is or needs the feedback stack replaced.

What the three devices are doing #

A brushless resolver is an absolute angular sensor. You drive one winding (the reference) with an AC carrier and two stator windings return sine and cosine envelopes of shaft angle; the angle comes from the ratio of the two envelopes and is valid over exactly one revolution. That is the fine channel. The geared devices turn slower than the shaft: one at about 1:4 completes a revolution every 4 motor revolutions, one at about 1:80 every 80. Together they form the classic coarse-fine cascade, built mechanically instead of electrically. A true multispeed resolver gets its extra speeds from the winding pattern alone, with no gears, so this stack is not one of those. Synchro systems of the same era switched between coarse and fine channels near the null with a relay; the gears here do the equivalent split in hardware. Both ratios are bench observations from the case at hand, approximate until you count them yourself.

Slowing each channel also slows its signal rate, which mattered to mid-1990s controllers: a channel turning at one-eightieth of motor speed stays easy to read even with the motor at full speed.

Channel One cycle spans Role in the stack Confirming check
Resolver 1 motor revolution Fine angle; commutation and velocity source Sine and cosine envelopes on the reference carrier
Geared device near 1:4 4 motor revolutions Intermediate turn count Counted 4 shaft turns per device turn, repeatable state
Geared device near 1:80 80 motor revolutions Coarse turn count Counted 80 shaft turns per device turn, repeatable state

Bench identification, step by step #

  1. Lock out the machine and prove zero energy before opening anything: disconnect, lock and verify every energy source, confirm the drive's DC bus reads 0 V, and block any gravity-loaded axis. That verification step is what US practice under OSHA 1910.147 requires before this kind of work, and the machine's risk assessment governs anything beyond the bench.
  2. Document before you disconnect. Photograph each connector in place, log wire colors from every device to its board terminal, copy every marking on every device, and scribe alignment marks across each gear mesh and coupling. Do not unbolt the gears: if their phase encodes absolute position, re-indexing them destroys the machine's position memory.
  3. Map the wiring with an ohmmeter on continuity, device pin by device pin, terminal by terminal. Three devices feeding one board means the board conditions, combines, or converts signals, and the wire map is what lets you probe the right pins later without the original schematic.
  4. Classify each device electrically, still unpowered. A resolver reads like a transformer: isolated winding pairs, similar low resistance on each pair, no supply pin anywhere. A powered device, optical or magnetic, shows supply and ground pins plus signal pins with low or diode-like readings to them. This test alone separates the resolver from anything electronic.
  5. Count the ratios. Mark the motor shaft and both device shafts, turn the motor exactly 20 revolutions, and count what each device did, with the arithmetic below. Record the direction each device rotates as well, because gear stages can reverse it.
  6. Excite and capture. Power the original controller only if the machine can run it safely with guards in place; otherwise bench-excite the resolver reference from a signal generator (carriers commonly sit between 400 Hz and 10 kHz as a rule of thumb; take the exact frequency from the resolver or drive datasheet) and probe each channel at the board with isolated scope inputs.
  7. Capture one screen per channel at a marked shaft position, then repeat after 5 revolutions forward and 5 back. An amplitude-modulated sine and cosine pair riding the carrier is the resolver; a pulse train is incremental; stepped states that hold between moves are absolute or commutation information.
  8. Run the repeatability test. Turn forward through several revolutions to a mark, record all three channels, turn backward to the same mark, and record again. Identical channel states at identical marks mean the stack holds absolute position mechanically. Different states mean slip, a miscount, or a rate signal, so resolve that before trusting any retrofit plan.

The ratio check with numbers #

Inputs for this worked example: motor shaft turned exactly 20.0 revolutions, devices at the approximate ratios observed above.

  • 1:4 device: 20 revolutions ÷ 4 = 5.00 device revolutions, five full turns of the device shaft.
  • 1:80 device: 20 revolutions ÷ 80 = 0.25 device revolution; 0.25 × 360°/revolution = 90°, so the device ends a quarter turn from its mark.

The 90° figure is the quick bench tell: scribe the mark, turn the shaft 20 counted revolutions, and the coarse device should stop a quarter turn away. A count that lands far off, 4.5 revolutions where 5.00 is expected for instance, is a miscounted shaft turn or a slipping coupling far more often than an odd ratio, because gear ratios come from whole tooth counts. Recount before concluding anything unusual.

Position window from the same inputs: if each geared device produces one electrical cycle per mechanical revolution, the resolver repeats every 1 motor revolution, the 1:4 channel every 4, the 1:80 channel every 80. The stack is therefore unambiguous across an 80-motor-revolution span, a calculated result for these ratios, not a measured one. Beyond that span the coarse channel repeats, and the turn count has to live in the controller or be re-established by homing.

Reading each channel's waveform #

The resolver's two signal windings carry the reference carrier modulated by the sine and cosine of shaft angle, and the drive's resolver-to-digital converter derives angle, velocity and commutation from them. If your scope shows that pair, the fine channel is confirmed.

A pulse train from a geared device means incremental: the original controller counted pulses for its turn count, so the position memory was electronic, battery or capacitor backed, not in the gears. Stepped states that hold between moves mean absolute: the device shaft position itself encodes the count. A three-signal pattern of square waves stepping through fixed states is commutation if it repeats once per electrical revolution (the motor's pole pairs set that period) and a turns code if it repeats once per 4 or 80 motor revolutions. The period is the distinguishing number.

What to record before you close it up #

  • The counted ratios with rotation direction, replacing the approximate 1:4 and 1:80.
  • The wire map from device pin to board terminal, with wire colors.
  • One waveform capture per channel at the marked position.
  • Every device marking, and the motor's pole count from its nameplate.
  • Photos of the gear-phase marks, so a future repair can restore them.

The retrofit decision #

Case A: the geared channels are pulse trains. The controller held the turns electronically, so nothing in the gears must be preserved. Decision: replace the stack with a multi-turn absolute encoder, or keep the resolver for the new drive and re-home the axis at every power-up; recovering position that way after a feedback swap is exactly what multi-turn motor homing with encoders and geared potentiometers covers.

Case B: the geared channels are absolute and repeatable. The stack is the machine's position memory. Decision: keep it and interface its outputs into the new controller, where level shifting, scaling and a fault line are usually the whole task, or fit a multi-turn absolute encoder and re-master the axis once. A resolver-only drive will not recover position across revolutions, so do not plan on it.

Case C: one geared channel carries the three-wave commutation pattern. That channel existed to commutate the original drive. Decision: the new drive must accept resolver feedback and derive commutation itself, or you replace feedback with an encoder whose commutation tracks match the motor's pole count, a number taken from the motor's nameplate or datasheet, never guessed.

If none of the captured waveforms matches these three patterns, treat the motor and its board as one undocumented unit: obtain the feedback specification from the machine or motor builder before buying a drive, because commutation and position recovery cannot be verified by inference.

Questions asked

How do I tell a resolver from an encoder with just a multimeter?

Put an ohmmeter across pin pairs with every connector removed. A resolver reads like a transformer: isolated winding pairs of similar low resistance, typically tens to a few hundred ohms as a rule of thumb, and no pin that behaves like a power supply. An optical or magnetic encoder instead shows a supply-to-ground pair plus signal pins that read low or diode-like against it. Anything with supply pins is not a resolver.

Why would a 1990s servo use three feedback speeds instead of one encoder?

Two reasons justify the stack on their own: absolute multi-turn position without battery backup, and coarse channels slow enough for an early-1990s controller to read. A third motive often suggested, that the controller's counting rate could not follow one fast encoder at full motor speed so it switched to a reduced-speed channel, fits the era, but the ratios alone do not prove it. The captured waveforms tell you what the controller actually consumed.

Can I run this motor on a modern drive using only the resolver?

Yes when two conditions hold: the new drive accepts resolver feedback for commutation and loop closure, and the axis re-homes at every power-up, because a resolver repeats every revolution and carries no turn count. No when the machine must retain absolute position across power cycles; then keep the geared channels and interface them to the new controller, or replace the stack with a multi-turn absolute encoder.

How far does the 1:80 channel extend absolute position?

Eighty motor revolutions for the ratios in this case. The resolver repeats every 1 revolution, the 1:4 channel every 4, the 1:80 channel every 80, so the stack is unambiguous across an 80-revolution span, a calculated result that assumes one electrical cycle per device revolution. Past that span the coarse channel repeats, and the controller must track turns or re-home.

Is it safe to turn the shaft by hand for the ratio count?

Yes for a horizontal axis once the drive is disconnected and the DC bus reads 0 V. Block or support any gravity-loaded axis first, because back-driving the motor lets a suspended load fall, and expect the motor to generate while it turns, so treat its power leads as live during the count.

Keep reading

  • How-to

    Match a Stepper Feed Rate to Conveyor Speed With an Encoder

    Continuously scale the conveyor encoder's measured count rate into the pulse-output velocity register: divide the count rate by the counts per inch of belt travel, then multiply by the label-feed ratio and the drive's pulses per inch of label travel. A fixed feed rate matches one belt speed only; the per-scan register update is what makes the labeler follow the belt.

    Servo and stepper motion