Motion feedback — encoders, resolvers and load cells

An incremental encoder outputs pulses as it turns. One channel tells you the shaft moved; two channels 90° apart tell you which way. That phase relationship is quadrature, and it is why encoders have A and B rather than just A.

Because the counter can see all four edge combinations, a controller can count 1×, 2× or 4× the nominal pulses per revolution. A 1024 PPR encoder decoded 4× gives 4096 counts per revolution. Vendors quote both numbers, and mixing them up is the most common cause of a machine that moves exactly four times too far.

Most encoders add a Z (zero, index, marker) channel: one pulse per revolution. That is what a homing routine uses to establish a repeatable reference. Without Z, homing depends entirely on a home switch, and repeatability is limited by that switch.

An incremental encoder counts changes, so its position is only meaningful relative to wherever the counter was zeroed. Cut the power and the count is gone; the machine must home before it can be trusted. An absolute encoder reports its actual angular position on demand, so at power-up the controller already knows where the axis is.

Single-turn absolute gives a unique position within one revolution. Multi-turn adds a gear or battery-backed counter, so it also knows how many revolutions have passed — which is what you need on a ballscrew axis or a hoist where one revolution is a fraction of the travel.

The economics are simple. Absolute costs more per axis and saves you a homing sequence, the mechanics of a home switch, and the production time lost homing after every power cut. On a machine that stops and starts often, or where homing is dangerous or slow, absolute wins quickly.

A resolver is a small rotating transformer: an excited rotor winding, two stator windings 90° apart, and outputs whose amplitude ratio gives the shaft angle. There is no glass disc, no LED, no electronics at the sensor — which makes it exceptionally tolerant of heat, shock, vibration and radiation.

That is why resolvers persist in traction motors, presses, foundry equipment and military kit long after optical encoders took over the general market. The price is the electronics at the other end: a resolver needs excitation and a resolver-to-digital converter, so the drive must support it.

Not every axis needs position. If all you need is speed, an inductive proximity sensor watching a toothed wheel, or a Hall sensor watching a magnet ring, gives a pulse train the PLC can measure — either by counting pulses over a window or by timing between them.

Choose the method by speed. At high speed, count pulses per fixed window: the count is large and the resolution is good. At low speed, time the gap between pulses instead, because a fixed window may contain one pulse or none and the reading collapses to a stutter. Many high-speed counter modules do both and switch over automatically; if yours does not, write it that way in code.

A load cell is a spring element with strain gauges bonded to it in a Wheatstone bridge. Under load the element deflects microscopically, the gauge resistances change, and the bridge produces a tiny differential voltage — typically 2 mV per volt of excitation at full scale, so about 20 mV at 10 V excitation for the whole range.

Millivolts is why load-cell wiring is unforgiving. Use a proper weighing transmitter or a dedicated strain-gauge card, keep the cable screened and continuous (a junction in a load-cell cable is a calibration change), and use 6-wire sense connections on long runs so excitation drop at the cell is compensated.

Most weighing problems are mechanical, not electrical. Rigid pipework into a weighed vessel carries load away from the cells; a stay rod tightened during maintenance changes the calibration; thermal expansion of a frame shows up as slow drift. Flexible connections, correct mounting kits and a repeatable check weight solve more weighing tickets than any amount of filter tuning.

Feedback cable shares a machine with the noisiest device in it — the drive that is following it. The mitigations are the same as chapter 1, applied more strictly: differential (RS-422 or the vendor's serial protocol) rather than single-ended output, screened cable with the screen clamped at the drive end, and separation from motor cable along the entire route, not just in the cabinet.

Use the pre-made feedback cable from the drive vendor where one exists. It is not brand loyalty: those cables have the correct pair grouping, screen construction and connector shell bonding, and hand-built substitutes are behind a large share of intermittent position-loss faults.