Motor Driver PCB Design: Layout and Thermal Rules
A motor driver is a power design wearing a logic design’s clothes. It receives low-level control signals, and it delivers currents that are large enough to destroy traces, cook devices and corrupt the ground reference of everything else on the board. Designing a motor driver PCB therefore means deciding the function and the performance targets first, then treating the power path, the gate drive and the heat as the primary layout problems.
Deciding What the Circuit Has to Do
The functional questions come first because they determine the topology.
Does the motor turn in one direction or two? A unidirectional drive needs only a single high-power transistor, field effect transistor or relay. Bidirectional rotation needs either an H-bridge of four power devices or a double-pole double-throw relay.
Does the speed need to be controlled? If not, a relay is sufficient. If it does, a switching device and pulse width modulation are required, and the design shifts from a switch to a converter.
Performance Targets for a PWM Drive
Five parameters define the design.
Output current and voltage range set the largest motor the circuit can drive.
Efficiency matters twice: it saves supply energy, and it reduces heat in the driver itself. The two levers are keeping the power devices fully in their switching states rather than in between, and preventing shoot-through — the condition in which both devices of a bridge arm conduct at the same time and short the supply.
Isolation from the control input. The power stage must not let high voltage or large current reach the controlling circuit. A high input impedance or an optocoupler provides that separation.
Effect on the supply. Shoot-through produces a momentary collapse of the supply voltage, which is felt as high-frequency pollution by every other circuit sharing the rail. Large currents also cause the ground potential to shift, which is the failure mode that produces mysterious behaviour in unrelated parts of the system.
Reliability. The driver should be safe under any control signal and any passive load, which is a design goal rather than a test result.
The Input Stage
The input side is where the control and the power domain are joined, and where the grounding scheme is decided.
On a typical driver board the input signal arrives on one pin, with pin one as the ground connection, and a resistor of around two kilo-ohms is connected between that ground pin and board ground. That resistor does two different jobs depending on how the system is powered. When the driver board and the controller have separate supplies, it provides a return path for the signal current. When they share one supply, it prevents a large current from flowing along the interconnecting wire into the controller’s ground and disturbing it — in effect separating the two grounds and achieving single-point grounding.
A high-speed operational amplifier is often used as the comparator that converts the incoming logic level into a square wave swinging close to the power supply, against a reference of about 2.7 volts derived from an indicator and a diode. The input range of such an amplifier cannot approach the negative supply, so clamping diodes are added at the input to prevent the range being exceeded. Of the two input resistors, one limits current and the other pulls the input low when it is left floating.
One substitution is worth avoiding: an open-collector comparator cannot replace the amplifier in this position, because its high output state has an impedance above one kilo-ohm. The voltage drop that produces is enough to stop the following transistor from turning off.
The Gate Drive
The stage between the comparator and the power devices has two jobs: raise the signal to the level the gate needs, and guarantee that the two devices in a bridge arm are never on together.
In a typical discrete implementation, transistors, resistors and a zener diode drive the gate, and the gate capacitance itself — around 1000 picofarads — provides the timing that prevents shoot-through. The behaviour is easiest to see by looking at the two transitions separately.
When the drive transistor turns on, its on-resistance is far lower than the pull-up resistor, so the charge on the gate capacitance is released quickly and the device turns off fast. When it turns off, the gate has to charge through the resistor, which takes time. The device therefore turns off faster than it turns on. If both drive transistors switch simultaneously, the arrangement makes the upper and lower devices turn off before either turns on, which is exactly the break-before-make behaviour required.
The real circuit is safer still, because the amplifier output takes time to slew between its extremes, and during that interval both drive transistors are conducting and both power devices are off.
The zener diode at the gate has a specific purpose. Gate oxide withstands about 18 or 20 volts, so applying the full 24 volt supply directly would destroy the device; a 12 volt zener limits it. A resistor of about two kilo-ohms can serve the same purpose, since it produces the same division, but a normal diode cannot.
The Output Stage
Large power MOSFETs contain an intrinsic diode in parallel with the source and drain. Used in an H-bridge, that means the four diodes normally added to clamp voltage spikes are already present inside the devices, and no external diodes are needed.
A small capacitor across the outputs reduces the voltage spikes the motor generates, but under pulse width modulation it also produces spike currents, so its value must be kept small. For small motors it can be omitted entirely, and where it is fitted it must have a high voltage rating: an ordinary ceramic capacitor can fail short and take the bridge with it.
A typical board built this way handles a supply of 15 to 30 volts, 5 amps continuously per motor with 10 amps for ten seconds, and PWM up to 30 kilohertz, though one to ten kilohertz is the practical range. With four logically independent power stages whose outputs are paired into bridges, it can be driven directly from a microcontroller to give bidirectional rotation and speed control.
The Power Path
Heavy-current conductors should be as short and as wide as the layout allows, and vias should be avoided in them.
Where a via cannot be avoided, make it large — more than a millimetre — and surround the pad with a ring of small vias that fill with solder during assembly. A single undersized via in a motor current path is a fuse waiting to open.
If zener diodes are used in the gate circuit, the conductors from the source of the power device to the supply and to ground must be as short and wide as possible. Under high current, the voltage drop along a thin conductor can forward-bias the zener and turn on the transistor, and the pair then destroys itself. In one design, a 0.15 ohm resistor placed between the source of the low-side device and ground to sense current became the component that repeatedly destroyed the board — a clear case of a conductor whose voltage drop was treated as a measurement and behaved as a fault.
The geometry of a bridge arm is where this is decided, and the reasoning is the same as that described in this note on half-bridge converter layout.
Thermal Design
Motor drivers need deliberate cooling, because the board material is a poor conductor of heat while copper is an excellent one. Increasing the copper area is therefore the most effective measure available.
Thicker foil removes heat better than thin foil — two ounce copper is roughly 68 micrometres — but it costs more and is harder to pattern finely, which is why one ounce, around 34 micrometres, has become the common choice, with outer layers typically using half an ounce to one ounce.
Solid copper planes on inner layers conduct heat well, but because they sit in the middle of the stack, heat tends to accumulate inside the board rather than escaping. Adding copper on the outer layers and stitching it to the inner layers with many thermal via connections moves heat outwards. On a two-layer board the problem is harder, because traces and components break up the copper; providing as much solid copper as possible, with a good thermal connection to the driver device, is the practical objective. Pouring copper on both outer layers and connecting the pours with many vias also spreads heat between the regions that traces and components have divided.
The principles are the same ones that apply to any thermally loaded board, and they are set out in this discussion of thermal management on dense boards.
Trace Width for Motor Currents
Wider is better, and on motor drivers it is not a matter of preference.
Current into and out of the driver can exceed 10 amps, so trace sizing has to ensure the resistive loss neither wastes significant power nor heats the conductor. A trace that is too small behaves as a fuse and burns open.
The usual reference is IPC-2221, whose charts convert current and permitted temperature rise into a copper cross-sectional area, and from there into a width for a given copper thickness. As an illustration, a one ounce copper trace carrying 10 amps needs to be slightly wider than 7 millimetres to limit the rise to 10 degrees Celsius, while a 1 amp trace needs only 0.3 millimetres.
Those figures assume a long, uniform trace, which is an important qualification. A short trace has lower resistance and any heat it generates is absorbed by the wider copper at each end, which acts as a heat sink. Short, narrow sections that terminate in large copper areas therefore carry far more current than the chart suggests.
The other qualification concerns layer. Inner-layer traces cannot dissipate heat as well as outer-layer traces because the substrate beneath them conducts poorly, so inner-layer conductors should be about twice as wide as their outer-layer equivalents. The calculations behind these figures are set out in this note on trace width calculation.
FAQ
What is shoot-through and why does it matter? It is the condition where both devices in a bridge arm conduct simultaneously and short the supply. It wastes power, collapses the supply voltage and injects high-frequency noise into every circuit on the rail, and it is prevented by ensuring the devices turn off before the opposite one turns on.
Why is a wide trace not always needed on a motor driver? Because the IPC charts assume long, uniform conductors. A short trace with wide copper at both ends has a low resistance and a heat sink built into its ends, so it can carry considerably more current than its width alone suggests.
Why should inner-layer power traces be wider than outer-layer ones? Because the laminate around an inner trace conducts heat poorly compared with the air and mask around an outer trace. The same current produces a higher temperature rise inside the board, so the conductor has to be roughly twice as wide to compensate.
