1. When a Motor Stumbles Before It Spins, Look at the Hall Plate

A small production run of 24 V brushless fans came back with three complaints: the motor jerked once before it spun up, a few units needed a hand-start to get going, and two units ran backwards even though nobody had touched the wiring. The three-phase bridge checked out. The gate driver checked out. The firmware waveforms checked out. The fault was on a circuit board holding three small parts — the Hall plate — where the electrical spacing of the three sensors did not match the commutation table in the firmware.

In a brushless DC machine the Hall plate is the least expensive part of the drive and the one most often treated as a commodity. Three Hall switches, a few pull-up resistors, a five-conductor harness — nothing about it looks critical. Yet this block decides which phase pair the controller energizes and when. Get it wrong and every downstream stage amplifies the error: cogging at start-up, reverse rotation, missed commutation, audible noise.

What makes these faults expensive is how poorly they reproduce. A bench test at room temperature with no load usually passes. Then summer arrives, or the magnet vendor ships a different lot, or the Hall plate comes from a second source, and failures arrive in batches. Engineers naturally look at the power stage first because it has waveforms, current, and heat to inspect, while the Hall plate only offers three signal lines that look fine as long as they toggle. The mismatch only shows up when Hall edges and phase-current overshoot are captured on the same time base.

2. Three Symptoms That All Trace Back to the Hall Layer

Jerk or reverse rotation at start-up. On power-up the controller reads one Hall state and energizes the corresponding phase pair, producing torque in one direction. If the Hall state sequence does not match the commutation table, that phase pair is wrong and the rotor is yanked backwards for a moment before the next state pulls it forward. That is the clunk. Reverse rotation usually means a phase-to-Hall mapping inversion somewhere in the chain.

Missed commutation at speed. As RPM rises, each commutation sector gets shorter. The Hall device itself has a sampling period; the output edge, the RC time constant formed by the pull-up resistor and harness capacitance, the MCU input filter, and the interrupt latency all consume part of that window. Once the sum approaches the sector width, the commutation point starts to drift: the motor tops out below its target speed, the current waveform grows spikes, and noise climbs.

A second, sneakier contributor to missed commutation is a lazy rising edge. Oversize the pull-up resistor and the Hall output climbs slowly; the MCU input threshold and its Schmitt hysteresis then translate that slow ramp into a jittery switching instant. The same design behaves differently with a longer harness or a different MCU pin. Faults of this kind routinely consume a week of debug time.

One Hall channel frozen high or low. When a channel never toggles after assembly, the usual cause is a magnetic threshold set above the field actually present at the air gap. A slightly weaker magnet, a slightly larger gap, a slightly higher temperature, and the device never operates at all. These units pass pilot builds and fail in the field.

3. 120 and 60 Degrees Are Electrical, Not Mechanical

This is where the whole discussion starts. Let p be the number of pole pairs (magnet pole count divided by two). Mechanical angle θm and electrical angle θe are related by:

θe = p × θm

One electrical cycle of 360°e corresponds to a mechanical rotation of 360° / p. Six-step commutation divides that electrical cycle into six sectors of 60°e, each with its own energizing combination. The job of the three Hall sensors is to identify which of the six sectors the rotor is in.

The terms "120-degree Hall" and "60-degree Hall" refer to the electrical spacing between the three output signals:

One relationship that regularly surprises people: flip any one sensor in a 120° arrangement so its sensing face points the other way, and its output inverts. That is equivalent to a 180°e phase shift, and the whole sequence turns into the 60° pattern. A board revision, a second-source Hall plate, or a single sensor mounted backwards will quietly convert a 120° motor into a 60° motor. Leave the firmware alone and the motor starts to cog.

The field check is simple but requires patience. Power the Hall plate on its own, add a pull-up to each of the three outputs, capture HA, HB, and HC simultaneously on a logic analyzer or a three-channel scope, rotate the rotor by hand through exactly one revolution, and write down the levels in time order. Count two things: how many times HA toggles in one revolution, which gives the pole pairs; and whether 000 or 111 ever appears, which gives the arrangement. A multimeter works too if you rotate slowly and record level by level. Run this check every time the Hall plate supplier changes or the plate layout is revised, and put it in the pilot-build checklist.

Worked example 1: count pole pairs first, then lay out the plate.

Take an 8-pole outer-rotor fan (pole pairs p = 4) rated at 12000 rpm.

Step one, confirm the pole count. Disconnect the three phase leads, rotate the rotor one full turn at a reasonably constant rate, and count the transitions on HA. HA toggles twice per electrical cycle — one rising, one falling — and one revolution contains p electrical cycles, so the transition count N = 2p. Eight transitions means p = 4, consistent with the nameplate.

Step two, convert to physical placement. For a 120°e target, adjacent sensors sit 120° / 4 = 30° apart mechanically, at 0°, 30°, and 60° around the bore. For a 60°e target the spacing becomes 60° / 4 = 15°, placing them at 0°, 15°, and 30°.

Step three, check the electrical frequency. fe = p × n / 60 = 4 × 12000 / 60 = 800 Hz, giving an electrical period of 1250 µs and a commutation sector of 208 µs. This number reappears in worked example 3.

Thirty degrees versus fifteen degrees is a different set of pads on the same small board. No amount of firmware tuning recovers from the wrong pads.

4. Why Commutation Requires a Latch, Not a Plain Switch

Hall switches come in three behavioral flavors: unipolar, omnipolar, and bipolar latch. Choosing the wrong one produces a waveform that looks like a square wave while the commutation point drifts continuously.

Unipolar switches respond to one magnetic pole only. Assume the air-gap field varies roughly sinusoidally with angle: B(θ) = Bm × sin θ. The device turns on when B rises past the operate point and turns off when B falls back through the release point, so the conduction angle depends on the ratio of Bm to the thresholds. For a unipolar switch with a 60 Gs operate point and a 40 Gs release point:

• Bm = 800 Gs (strong magnet, small gap): the operate point is crossed at θ = arcsin(60/800) = 4.3°, the release point at 180° − arcsin(40/800) = 177.1°, giving a conduction width of 172.8° and a duty cycle of 48.0%
• Bm = 100 Gs (weak magnet, large gap): conduction runs from arcsin(60/100) = 36.9° to 180° − arcsin(40/100) = 156.4°, a width of 119.6° and a duty cycle of 33.2%
• Bm = 50 Gs: the peak never reaches the operate point, so the output never switches

In other words, a unipolar switch delivers a duty cycle that floats anywhere between 0% and 50% depending on magnet strength, air gap, assembly tolerance, and temperature drift. The six equal-width sectors assumed by the commutation table are not equal here, so the motor cogs, runs noisily, and cannot reach its target speed.

Omnipolar switches respond to both poles, which does not help either. Such a device conducts near the north-pole peak, releases in the zero-crossing region, and conducts again near the south-pole peak — two conduction intervals per electrical cycle, which is not a commutation signal at all.

Bipolar latches are the type this job requires. The output switches when one pole exceeds the positive threshold and holds; it releases only when the opposite pole reaches the negative threshold. The switching instant is therefore set by the field zero crossing and is independent of field amplitude — as long as the peak exceeds the threshold, the duty cycle stays at 50%. That is the fundamental reason a latch works for commutation, and the reason magnet aging, temperature drift, and assembly tolerance do not move the commutation point.

5. H6011MI: Four Parameters Mapped to Four Commutation Requirements

The H6011MI from HXY Electronics is a bipolar latching Hall switch in a SOT-23-3L package whose parameter set lines up with the four requirements above.

Latching and bipolar, aligned to the zero crossing. The operate and release points are +45 Gs and −45 Gs (the Bop/Brp symbol convention follows the datasheet), a 90 Gs window between them. That window means the output, once switched, will not drop out until an opposing field of equal magnitude arrives, which is exactly the chatter immunity commutation needs. At the same time, a 45 Gs threshold does not demand much from the magnetic circuit: a conventional NdFeB magnet delivers several hundred gauss at a gap of a few millimeters without effort.

Supply range of 3.8 to 60 V, straight from the control rail. A 24 V or 48 V system needs no dedicated low-voltage rail for the sensors, and 5 V and 12 V systems sit inside the range as well. The saving shows up in both BOM cost and board area.

Operating temperature of −40 to 150 °C. The Hall plate is normally mounted at the motor end, close to the windings, where summer full-load operation can push local temperatures well past 100 °C. Devices rated to 85 °C have no margin at that location; a 150 °C rating provides one.

Output rise of 1 µs and fall of 1.5 µs, open-drain with a 0.4 V output low and 30 mA output current. The edges are fast enough for clean input-capture timing, and the open-drain stage lets the pull-up rail follow the MCU. The 0.4 V output low together with the 30 mA sink capability sets the lower bound on the pull-up resistor.

Three related parts are worth evaluating alongside it: the H6011NS is the TO-92S through-hole version with identical parameters, useful when the Hall plate is built with insertion-mount processes or hand rework; the H6013MI is another SOT-23-3L latch with operate and release points of +25 Gs and −25 Gs, a 50 Gs window, and a better fit when the air-gap field is marginal; and the H6042MI is a latch with a 500 kHz sampling rate, kept for high-speed commutation.

The H6711MI is included for contrast only. As a unipolar non-latching device, it is not recommended for three-phase commutation.

6. The Application Circuit: From Magnet to Three-Phase Bridge

The complete commutation signal chain has five stages.

Magnetic circuit. The rotor magnets, or a dedicated sensor magnet ring, produce a field in the air gap that varies approximately sinusoidally with angle. What the sensor sees is the component perpendicular to its package sensing face, so the mounting attitude of the Hall plate must be aligned with the magnetizing direction.

Sensing. Three H6011MI devices are placed at the mechanical spacing derived from the 120°e or 60°e target. VCC comes from the control rail anywhere in the 3.8 to 60 V range; each output gets a pull-up resistor to the MCU I/O rail; a decoupling capacitor sits close to the device between VCC and ground.

Transmission. A five-conductor harness (VCC, GND, HA, HB, HC) normally links the Hall plate to the controller. This harness is the main risk in the chain: it is distributed capacitance, and it is also an antenna that couples phase-line interference into the sensor signals.

Control. The MCU connects the three Hall outputs to GPIOs with input capture or external interrupt capability. Every state change triggers a commutation: the firmware looks up the next energizing combination and emits six PWM channels with dead time. The table must correspond exactly to the physical arrangement, 120° or 60°.

Power. Six PWM signals are amplified by the gate driver and applied to six N-channel power MOSFETs in a three-phase bridge, switching the bus voltage onto the motor windings per the commutation table. Dead time prevents shoot-through in a leg; 1 to 3 µs is a typical starting point, with the final value taken from the gate driver and MOSFET datasheets and bench measurements.

String the delays of those five stages together and you have the full time budget for one commutation: the magnet passes the threshold position, the sensor sees it at the next sampling instant (up to one sampling period of uncertainty), the output edge flips (1 to 1.5 µs from the device), the edge propagates through the pull-up and harness RC network (0.5 to 3 µs), the MCU input filter adds its share (1 to 5 µs if fitted), interrupt latency and table lookup take a few microseconds more, and finally the gate driver propagation delay and MOSFET turn-on complete the action. In that chain the Hall sampling period is usually the dominant term, which is why high-speed commutation problems surface at the sampling rate long before they surface at the power stage.

Where the H6011MI sits in this chain determines three practical outcomes: its 60 V supply ceiling lets the Hall plate hang directly off a 24 V or 48 V control rail and eliminates a dedicated low-voltage supply; its open-drain output lets the pull-up follow the MCU, so the same design serves 3.3 V and 5 V logic; and its SOT-23-3L package suits a space-constrained plate, with the TO-92S H6011NS available when an insertion-mount process is required.

7. Two Worked Examples You Can Apply Directly

Worked example 2: sizing the open-drain pull-up.

Given: the H6011MI has an open-drain output with a 0.4 V output low and a 30 mA output current limit; the pull-up returns to the MCU 5 V I/O rail; the harness from Hall plate to controller is 30 cm, and the load capacitance is estimated at 100 pF (MCU input capacitance plus device output capacitance — confirm by measurement).

The lower bound comes from sink current: R ≥ (5 − 0.4) / 0.03 = 153 Ω. Start at 1 kΩ in practice, which leaves more than six times margin.

The upper bound comes from rise time. With τ = R × C, the 10% to 90% rise time is approximately 2.2τ:

• R = 2.2 kΩ: τ = 0.22 µs, tr ≈ 0.48 µs; current while low is (5 − 0.4) / 2200 = 2.09 mA, resistor dissipation 9.6 mW
• R = 4.7 kΩ: τ = 0.47 µs, tr ≈ 1.03 µs; current 0.98 mA, dissipation 4.5 mW
• R = 10 kΩ: τ = 1.00 µs, tr ≈ 2.20 µs; current 0.46 mA, dissipation 2.1 mW

For short on-board traces, 4.7 kΩ is a reasonable compromise: the rise time is the same order as the device‘s own 1 µs edge while dissipation stays low. Stretch the harness to 1.5 m (capacitance around 300 pF) and the 4.7 kΩ rise time grows to 3.1 µs; drop to 2.2 kΩ (tr around 1.45 µs) and add a capacitor to ground at the MCU along with a Schmitt-trigger input to keep ringing under control.

One trap worth calling out: do not return the pull-up to the motor bus.

Worked example 3: delay budget and whether the sampling rate is enough.

Take the motor from example 1 (p = 4) and raise the target speed to 24000 rpm:

• Electrical frequency fe = 4 × 24000 / 60 = 1600 Hz, electrical period 625 µs
• Each commutation sector spans 60°e, so 625 / 6 = 104.2 µs
• The H6011MI samples at 20 kHz, a 50 µs sampling period, which is 48% of the sector width
• Add a 1.5 µs falling edge, 1.0 µs of pull-up RC rise, and a few microseconds of MCU input filtering and interrupt latency, and the phase uncertainty of the commutation point approaches half a sector

The conclusion is straightforward: at this speed a 20 kHz sampling rate is running out of room. Back the speed down to 12000 rpm (fe = 800 Hz, sector 208 µs) and the sampling period drops to 24% of the sector, which is acceptable in practice. If the application genuinely needs four pole pairs at 24000 rpm or above, move to a latch sampling at 500 kHz such as the H6042MI: a 2 µs sampling period, 1.9% of the sector, at the cost of a 30 V supply ceiling that requires a dedicated 5 V or 12 V rail in a 48 V system.

8. Design Pitfalls

Count pole pairs before placing sensors. Pole pairs determine the physical placement, placement determines the state sequence, and the sequence determines the commutation table. In practice: rotate one turn, count HA transitions N, take p = N / 2, compute physical spacing as target electrical angle divided by p, and record all three numbers in the BOM notes and in the firmware comments.

Bind the commutation table to the Hall arrangement; change the plate, change the table. The 120° sequence never contains 000 or 111; the 60° sequence does. Turn that around and it becomes a quick way to identify the arrangement: whether 000 and 111 are handled as fault codes or as valid sectors tells you which layout you have. Re-verify the sequence whenever the Hall plate source, mounting orientation, or layout changes.

Treat the Hall supply and decoupling as an analog rail. Three H6011MI devices draw about 3 × 4.8 mA = 14.4 mA, and with 4.7 kΩ pull-ups to 5 V the resistors add roughly 3 × 0.98 mA ≈ 2.9 mA while the outputs are low, putting the Hall plate in the 17 mA range overall. Harness voltage drop, supply ripple, and inadequate decoupling all show up as threshold uncertainty. Place a decoupling capacitor at the plate between VCC and ground, run dedicated VCC and GND conductors in the harness rather than bundling them with the phase leads, and add bulk capacitance on the controller side.

Select magnetic thresholds for the worst case, not the typical case. The air-gap field falls with assembly tolerance, magnet lot variation, high-temperature demagnetization, and long-term aging. Carry a margin of three to five times the threshold at the field peak. Where the gap field is only in the 100 to 200 Gs range, choose the ±25 Gs H6013MI rather than the ±45 Gs H6011MI.

Do not return the pull-up to the bus. At 24 V a 4.7 kΩ resistor dissipates about 118 mW, and at 48 V about 482 mW, which standard chip resistors are not sized to handle. Return the open-drain pull-up to the MCU I/O rail while the device itself runs from the control supply; the two can sit at different levels, and the output voltage ceiling should be confirmed against the datasheet.

Treat the Hall harness as a noise source. Route the Hall leads away from the phase leads or shield them, add an RC filter and Schmitt-trigger input at the MCU, and where the mechanics allow it leave the Hall plate rotatable so the commutation advance angle can be trimmed to compensate for commutation delay at speed.

9. Vendor and Summary

Hall commutation in a brushless machine comes down to four things lining up: the electrical arrangement matching the commutation table, a latching device rather than a plain switch, enough margin between the magnetic thresholds and the air-gap field, and a pull-up and delay budget that have actually been calculated. Miss any one of them and the symptom in the field is cogging, reverse rotation, or lost commutation.

HXY Electronics (HuaXuanYang Electronics) is a power device brand with ten years of focus on this category, offering more than 5000 part numbers. The Hall line covers 96 Hall switches and 46 Hall ICs, 142 devices in total, spanning unipolar, bipolar latch, omnipolar, two-wire, and programmable types. Output stages include 86 open-drain, 36 push-pull, and 20 current-source parts. Supply voltage spans 1.8 to 60 V, the operating temperature ceiling reaches 150 °C, the sampling rate ceiling reaches 500 kHz, and ESD capability reaches 12 kV. Packages cover SOT-23-3L, TO-92S, SOT-23, PG-SSO-3-52, SOP-8, and TO-94. For brushless commutation the choice between the H6011MI (SOT-23-3L, ±45 Gs) and the H6013MI (SOT-23-3L, ±25 Gs) follows the strength of the air-gap field; the H6011NS (TO-92S) covers insertion-mount processes, and the H6042MI (500 kHz sampling) covers high-speed work. HXY supports selection with FAE engineers and offers sample requests, online ordering, volume quotation, lead-time inquiry, and scheduled delivery.

Disclaimer: this article is compiled from publicly available technical material and HXY Electronics product information for reference and discussion only, and does not constitute any form of design guarantee or commitment. Final device selection, application, and verification must be based on the latest official datasheet and bench measurements. HXY Electronics accepts no liability for consequences arising from use of the information in this article. For reproduction, please credit the source and contact sales@hxymos.com.

声明 / Disclaimer
本文部分参数引自华轩阳产品手册或第三方规格书，数据有更新可能，仅供选型参考，最终设计请以最新版规格书与实测为准。
联系 Contact：sales@hxymos.com