English is currently available.
Robotics Engineering DISCUSSION
My small robot arm has six hobby servos driven from a microcontroller board, all powered from the board's 5 V pin. One servo on its own works. When several move at once the servos twitch, the board sometimes resets, and even at rest a couple of them buzz and hunt around their position.
Is this a power problem, a signal problem, or both? What is the right way to power and drive a group of servos?
The resets are a supply problem. A hobby servo draws close to its stall current for a moment every time it starts moving, and for standard-size servos that is commonly above 1 A each; the datasheet gives the figure. Six of them can ask for several amps, while a board's 5 V pin fed from USB or a small regulator can supply only a fraction of that. The rail dips, the microcontroller browns out, and the servos lose their position signal and twitch.
Power the servos from a separate 5 to 6 V supply rated for the sum of the stall currents, or at least for the number that move at once.
When you split the supplies, connect the grounds. The control pulse is referenced to ground, so the servo supply ground and the microcontroller ground must be joined, ideally at one point near the servo connector block. Do not route servo current through the board's thin ground traces.
Add bulk capacitance across the servo rail close to the servos, something in the 470 to 1000 µF range, to cover the short current peaks that the wiring resistance and inductance would otherwise turn into dips. Use thick, short power wires; signal wires can be thin.
The buzzing at rest is usually signal timing. Position is set by a pulse of about 1 to 2 ms repeated at about 50 Hz. If 1000 µs of pulse width spans roughly 180°, each degree is only about 5.6 µs, so 10 µs of jitter in the pulse is nearly 2° of commanded movement. Software-timed pulses get stretched by interrupts from serial, I2C or other libraries.
Generate the pulses with a hardware timer output, or use a dedicated 16-channel PWM driver chip on I2C, so that the pulse width does not depend on what the code is doing.
Some hunting is mechanical rather than electrical. The servo is a small position loop with a potentiometer for feedback. If the load holds it slightly off target, or the gear train has backlash, it corrects, overshoots and corrects again, which you hear as buzzing. Analog servos only correct once per pulse; digital ones update much faster and hold harder, but draw more current doing it.
Check that nothing is commanding a position beyond the mechanical end stop. A servo pushed against its stop sits at stall current continuously and will overheat.
Back to Robotics Engineering community
