Milwaukee Tool: capacitor disconnect for a cordless tool

Milwaukee Tool, electrical engineering intern

A MOSFET switch that charges part of a tool's bulk capacitance after the battery is seated, to cut the inrush at insertion.

Jun to Aug 2024I designed and tested the circuit

The problem

A cordless tool’s motor drive needs a large bank of bulk capacitors across its battery input. When the battery is pushed in, those capacitors are empty, and they all charge at once through the battery terminals.

On the board I worked on, that inrush measured up to about 280 A, and it arced audibly at the terminals. Each arc wears the contacts.

How it works

The fix puts part of the bulk capacitance behind a power MOSFET, a transistor used here as a switch. At insertion only the capacitors in front of it charge. Once the microcontroller has started, it turns the switch on, and the rest of the bank charges in a second, short pulse.

The capacitor disconnect, block by block. Blocks: Battery pack; Battery terminals; Bulk capacitors (always on the bus); Switched capacitors (behind the MOSFET); Power MOSFET (the switch); Gate drive (level shift); Microcontroller (drives the enable line). Battery pack to Battery terminals; Battery terminals to Bulk capacitors, DC bus; Bulk capacitors to Switched capacitors; Switched capacitors to Power MOSFET; Microcontroller to Gate drive, enable; Gate drive to Power MOSFET, gate.

Click a labeled block to see what it does and why.

The circuit, block by block

Part of the bank sits behind a power MOSFET, which the microcontroller turns on through a gate-⁠drive stage. Click a block to see what it does and why.

In this layer

  • Battery terminals Where the battery pack meets the tool, and where the inrush passes at insertion.
  • Bulk capacitors The part of the bank that stays across the battery input, as before.
  • Switched capacitors The rest of the bulk capacitance, behind a power MOSFET, a transistor used here as a switch.
  • Power MOSFET The switch between the bus and the switched capacitors. I chose it and the other parts, and added the soft capacitor-enable circuit to the schematic.
  • Gate drive A gate-⁠drive stage between the microcontroller's enable line and the MOSFET's gate.
  • Microcontroller The board's STM32 microcontroller drives the enable line.

Battery terminals: where it arced

Where the battery pack meets the tool, and where the inrush passes at insertion.

On the board I worked on, the inrush measured up to about 280 A, and it arced audibly at the terminals. Each arc wears the contacts.

The figure

It draws the current at insertion without and with the fix. The shapes are illustrative; the measured numbers are under Results.

Inrush current, without and with the fix

Two current traces on one time axis, and the capacitor voltages under them. Without the fix, one large inrush spike at battery insertion, labeled up to about 280 A. With the fix, a smaller first spike at insertion, then a second, short pulse later, when the microcontroller turns the MOSFET on. The bulk capacitors charge at insertion; the switched capacitors charge at the second pulse. Illustrative, not to scale.

Without the disconnect, the whole bank charges in one spike at insertion. With it, the first spike is smaller, and the switched capacitors charge in a second, short pulse once the MOSFET is on.

Bulk capacitors: always on the bus

The part of the bank that stays across the battery input, as before.

At insertion

Only these capacitors charge when the battery is pushed in, so the first spike is smaller.

Switched capacitors: the rest of the bank

The rest of the bulk capacitance, behind a power MOSFET, a transistor used here as a switch.

With my advisor I worked out the circuit: switch only part of the capacitor bank, not the whole power path.

What it saved

That needs fewer MOSFETs and less board area than a full disconnect.

What it cost

A new load on the MOSFET, which now carries the second pulse.

The sequence

Blocks: Battery seated; First inrush (front capacitors only); Microcontroller starts; Switch enabled (by my firmware); Second pulse (short, through the MOSFET); Running (whole bank on the bus); Sleep (switch released).

  • Battery seated to First inrush.
  • First inrush to Microcontroller starts.
  • Microcontroller starts to Switch enabled.
  • Switch enabled to Second pulse.
  • Second pulse to Running.
  • Running to Sleep.

What I did: Switch enabled, Sleep.

One large inrush becomes two smaller ones. The second passes through the MOSFET, after the battery is already seated.

Power MOSFET: the switch

The switch between the bus and the switched capacitors. I chose it and the other parts, and added the soft capacitor-enable circuit to the schematic.

The second pulse

When it turns on, the rest of the bank charges through it in a second, short pulse, with the battery already seated.

Checked against its limits

I checked the second pulse's measured current, voltage and duration against the safe operating area in the MOSFET's datasheet: the limits within which it survives a pulse. It stayed inside.

How I checked it

Blocks: Board under test (with and without the disconnect); Oscilloscope (current and voltage); Inrush at insertion (compared); LTspice model; Second pulse (current, voltage, width); Safe operating area (MOSFET datasheet).

  • Board under test to Oscilloscope: battery insertion.
  • Oscilloscope to Inrush at insertion.
  • Oscilloscope to Second pulse: measured.
  • LTspice model to Second pulse: simulated.
  • Second pulse to Safe operating area.

What I did: Oscilloscope, Inrush at insertion, LTspice model, Safe operating area.

The oscilloscope gave the inrush at insertion and the second pulse. The second pulse, measured and simulated, was checked against the MOSFET's safe operating area.

Gate drive

A gate-⁠drive stage between the microcontroller's enable line and the MOSFET's gate.

What it does

It level-shifts the microcontroller's enable line to drive the MOSFET's gate.

Microcontroller: when to switch

The board's STM32 microcontroller drives the enable line.

My firmware

I wrote the firmware that sequences it, on the STM32 in IAR: a timer interrupt turns the switch on once the microcontroller has started.

In sleep

The switch is off again, which also cuts the sleep current.

Design decisions

With my advisor I worked out the circuit, then chose the MOSFET and the other parts. The choice that shaped it was to switch only part of the bank, not the whole power path. I also wrote the firmware that sequences it.

Bring-⁠up

I simulated the circuit in LTspice first. Then I had it laid out on the board, and once the board was built I tested it with an oscilloscope, with and without the disconnect. How I checked the MOSFET is in its panel.

Results

With the disconnect, the inrush at insertion measured 113 to 172 A, typically 135 A: about 46 % lower than the same board without it. I compared three boards: this one with and without the disconnect, and a rotary hammer’s board with it. On the rotary hammer’s board, the fix stopped the arcing at its terminals.

Credit

I did this as an intern in an engineering team at Milwaukee Tool, on a board that is the company’s design. I designed the circuit with my advisor’s guidance, wrote the firmware that sequences it, had it laid out, and tested the built board.