DRV8833 vs MX1508 for N20 Motors on a 2S Pack: Why the Cheap One Stays on the Shelf

I had both a DRV8833 and an MX1508 breakout on the bench and briefly considered designing the motor stage so either could be swapped in as a field spare. After reading both datasheets against the actual operating conditions, the answer was clear: the DRV8833 stays, the MX1508 does not go anywhere near the main drive.

The operating point

• Supply: 2S LiPo, 6.0 V empty to 8.4 V fully charged
• Load: four N20 gear motors, two per driver
• Per-motor stall current: roughly 0.8 A, so ~1.6 A per driver worst case
• PWM at 20 kHz, and the robot brakes and reverses hard in normal use

That last point matters more than the current figure.

Side by side

Why the DRV8833 is hard to kill here

At 1.6 A per driver the DRV8833 is well inside its rating. If a wheel jams and the current climbs, the over-current or thermal protection trips before the silicon is damaged. What you see from the outside is: the robot suddenly loses power on one side, pauses for a few hundred milliseconds, then recovers. That is the protection cycling, not a failure.

This is worth stating plainly because the symptom looks like a dying driver. Across the whole project, with several hard stalls and one wiring mistake that ran a motor at full speed for minutes, no DRV8833 has failed. The one incident that looked like driver failure turned out to be an unpowered GPIO domain on the ESP32-P4. Don’t replace the driver until you have measured the inputs.

Given that, a swappable-driver design is solving a problem that does not occur.

Why the MX1508 does not get the main drive

The number that rules it out is the 9.6 V ceiling against an 8.4 V full-charge supply. That is 1.2 V of margin, and an H-bridge driving a motor does not keep the rail at the battery voltage.

When a spinning motor is braked or reversed, it acts as a generator and pushes current back into the supply. The battery absorbs some of it; the protection board and wiring resistance mean the rail rises during that pulse. On a small robot doing quick direction changes, transient excursions of a volt or two above the battery voltage are normal. On the DRV8833 that lands in the 10.8-11.8 V window with room to spare. On the MX1508 it crosses the absolute maximum, and a discrete-MOSFET bridge with no protection does not recover from over-voltage; it fails shorted.

The lack of thermal shutdown compounds it. Two N20s stalled into an MX1508 will heat the MOSFETs until something gives, with no cut-off in between.

Where the MX1508 is fine

Low voltage, light load, no aggressive reversing. A single small auxiliary motor on 5 V, a fan, a pump. It is a perfectly good part for the price in that role, and that is the only role it has on this robot.

Practical notes for the DRV8833

• Leave nSLEEP high (or pull it up) if you are not using it; a floating nSLEEP gives a driver that randomly stops responding.
• The ISEN pins can implement hardware current limiting with a sense resistor (limit ≈ 0.2 V / R). For these motors the limit needed to protect a weak battery would have cut stall torque to a third, so the pins are grounded and the battery was upgraded instead — see the battery article.
• nFAULT is worth wiring. It is open-drain, active-low, and asserts on over-current, over-temperature or under-voltage. Wiring both drivers’ nFAULT pins to one GPIO with a pull-up costs one input and turns “the robot went soft for a moment” from folklore into a log line. You lose the ability to tell which driver tripped, which is a fine trade at this scale. (Planned, not yet wired on my build.)
• 20 kHz PWM with slow decay is quiet and gives smooth low-speed control on N20s. Fast decay gives sharper braking but audible whine at lower frequencies.

If you do want to size a sense resistor, this does the arithmetic and shows what it costs in stall torque: