Magnetic rotary encoders split into two dominant sensing families, Hall-effect and magnetoresistive (AMR, GMR, TMR), and the choice dictates resolution, temperature drift, package size, and BOM cost for any shaft-feedback design [S2][S5].
Hall-effect encoders integrate both the sensing element and the digital processing on the same silicon die, which makes them physically smaller than typical magneto-resistive encoder electronics [S2].
How each principle actually converts flux into counts
A Hall element outputs a voltage proportional to magnetic flux density perpendicular to the chip surface, and is built mainly from compound semiconductors: InSb for raw sensitivity, GaAs for temperature-stable parts, InAs as the balanced middle option, and Si where full monolithic integration is required [S4]. The Hall signal is typically a few tens of millivolts per Tesla, which is why a Hall IC almost always ships with an on-chip amplifier and ADC [S4].
A magnetoresistive sensor changes resistance when an in-plane field is applied, so the wiring is a Wheatstone bridge rather than a single voltage tap [S3]. TMR, the most recent MR variant, uses a tunnel-junction stack and has been characterised as offering high resolution, low power draw, and a compact footprint suitable for battery-powered angle sensing [S5]. TMR cells are also produced as monolithic ICs, which keeps the package small even though the bridge itself is more complex than a Hall plate [S5].
Sensitivity, field direction, and linearity
Magnetoresistive devices have much higher sensitivity than Hall devices, and that sensitivity is tunable by selecting film thickness and line width during manufacture, a degree of design freedom that Hall processes do not offer [S3]. The trade-off is operating field: an MR sensor responds to parallel fields, while a Hall sensor responds to fields perpendicular to the die, so the magnet shape and air-gap geometry cannot be reused between the two families without re-design [S3].
Hall is the preferred choice when a highly linear output is required with no saturation out to very high field strength, which is why Hall is the default for proximity detection of ferrous targets such as gear teeth in CNC spindles and transmission-speed sensing [S3]. MR sensors, by contrast, are omnipolar and excel at unipolar precision displacement measurement, including magnetic encoder wheels, which is the application area that overlaps directly with rotary encoder feedback [S3].
Resolution, accuracy, and the optical comparison
Optical encoders remain the resolution benchmark because they read a coded disc with a photodiode array and routinely reach sub-arc-minute accuracy in clean environments [S5]. Magnetic encoders using TMR have closed much of that gap and are specified for high-resolution angle feedback, while Hall-effect encoders typically land in the 8 to 14 bit absolute range and are preferred where robustness outweighs raw precision [S5].
Compared with optical designs, magnetic encoders of either type tolerate dust, oil mist, and condensation that would fog or scratch an optical disc, which is the main reason they are specified for elevator traction sheaves, off-highway drives, steel-mill rollers, and servo motors on dirty factory floors [S4][S5]. The 22 mm outer-diameter kit-encoder format now common in the market is feasible largely because the Hall IC carries the entire analog front-end on-die [S5].
Temperature behaviour, drift, and failure modes
Field reports from automotive parking-sensor production show that magnetoresistive parts, despite their sensitivity advantage, suffer from real-world temperature stability problems, and compensating the offset and gain drift in firmware took significant engineering effort to bring within spec [S3]. Hall-effect parts, particularly those built on GaAs, were characterised as having stable temperature behaviour suitable for industrial environments from the outset [S4].
For high-temperature service above 125 degC, the reliability analysis literature contrasts Hall's simpler construction against MR's higher sensitivity, with Hall generally favoured for sustained thermal cycling and MR reserved for applications where the magnet can be kept cool or the package is actively managed [S8]. Cylindrical SmCo magnets, with temperature-stable remanence, are the usual partner for either sensor type when the encoder sits near a hot motor or gearbox [S4].
Size, integration, and power budget
Hall-effect magnetic encoders integrate both the sensor element and the processor on the same chip, and that single-die architecture is what allows through-hole and hollow-shaft encoder housings down to roughly 22 mm OD without sacrificing the on-board diagnostics that used to require a separate ASIC [S2][S5]. The integrated Hall IC is also the lowest-power option of the three magnetic families, which matters in battery or energy-harvested feedback loops [S5].
TMR parts consume notably less current than earlier AMR designs while keeping the resolution benefit, which has pushed TMR into wearable robotics, surgical handpiece feedback, and small drone gimbal motors, where the Hall IC's lower sensitivity would cap the usable resolution [S5]. AMR and GMR still appear in older industrial designs and remain in the supply chain for legacy replacement, but new platform designs are migrating either down to Hall for cost or up to TMR for accuracy [S5][S8].
Selection matrix: Hall vs AMR vs TMR
On sensitivity per unit field, TMR and AMR sit roughly an order of magnitude above Hall, and that gap is the single biggest reason MR is picked over Hall in precision angle modules [S3]. On temperature stability, Hall ranks highest and MR ranks lowest, with MR offset drift often needing firmware compensation to deliver a usable signal in automotive-grade temperature swings [S3][S4].
On package size, Hall wins because of full monolithic integration; on power consumption, TMR and Hall are close, both well below older AMR bridges; on BOM cost, Hall is the lowest and TMR the highest [S2][S5]. For absolute single-turn resolution above 14 bit, or for hollow-shaft through-bore designs above 30 mm, the spec sheet is currently dominated by MR-based products, while sub-14-bit kit encoders below 30 mm OD are almost universally Hall [S5].
Who each technology is for, and where it fails
Hall-effect encoders are for design engineers who need a small, cheap, robust absolute or incremental feedback device on a servo or stepper motor, where dust, oil, and vibration would cripple an optical disc and sub-degree accuracy is not required [S4][S5]. They are the wrong choice when the application demands sub-arc-minute repeatability, when the air gap to the magnet is large and variable, or when the magnetic field on the chip is below the Hall threshold of a few millitesla [S3].
Magnetoresistive (AMR, GMR, TMR) encoders are for precision angle modules, battery-powered IoT actuators, surgical and humanoid robotics, and any position loop where the higher sensitivity and the in-plane field response open up magnet geometries that Hall simply cannot resolve [S3][S5]. They are the wrong choice when the operating temperature is high and uncompensated, when the budget cannot absorb the bridge trim and compensation circuitry, or when the application does not need the extra resolution and would only pay the cost of temperature-drift mitigation [S3][S8].
Engineering recommendations
For new designs on a BLDC or stepper below 50 W, start with a Hall-based absolute or incremental kit encoder in the 22 to 36 mm OD range, which keeps the BOM and firmware simple and lets you leverage the existing on-chip diagnostics [S2][S5]. For new designs above 14 bit single-turn or with hollow shafts above 30 mm, evaluate TMR first, then AMR for cost-down, and budget a temperature-compensation routine in firmware from day one [S3][S5].
Track two signals in the next sourcing cycle: the migration of TMR kit encoders into sub-30 mm OD packages, which would compress Hall's size advantage, and the release of GaAs or Si Hall ICs with on-chip temperature compensation, which would compress MR's accuracy advantage at the low end of the market [S4][S5][S8]. For a deeper look at how the Hall element is built into a complete feedback module, see the magnetic sensor primer, and for the magnet materials that sit on the other end of the air gap, the magnetic material reference covers the SmCo versus NdFeB decision directly [S4].
For related coverage, see Positive vs Centrifugal Discharge Bucket Elevator for Fragile Material.