Electrostatic Brake Driver

Why I started it

I wanted to explore whether an electrostatic clutch could make force feedback in a haptic glove smaller and lighter. A high voltage between two thin conductors pulls them together, and the friction between them resists motion. That makes the clutch an interesting alternative to a bulky mechanical brake, but it also means a glove needs a compact, controllable high-voltage power supply.

I designed and built a PCB around a USB input and a high-voltage output stage designed for up to 1200 V. It brings the supply, H-bridge, voltage measurement, and microcontroller interface onto one board so I can change the voltage and switching frequency during experiments.

Modeling Feasibility

I started with an ideal parallel-plate force model to see whether a small clutch was worth pursuing:

$$F = \frac{1}{2} \cdot \varepsilon_r \cdot \varepsilon_0 \cdot \frac{A \cdot V^2}{d^2}$$

Here $A$ is the plate overlap area, $d$ is the dielectric thickness, $V$ is the applied voltage, and $\varepsilon_r$ and $\varepsilon_0$ are the relative and vacuum permittivities. For a first estimate, I used a $5\,\text{cm}^2$ overlap, a 10 µm Mylar film, and $\varepsilon_r \approx 3.2$. The thin film keeps the gap small, which matters because force falls with the square of that gap.

With $A = 5 \times 10^{-4} \,\text{m}^2$ and $d = 1 \times 10^{-5} \,\text{m}$, the ideal model gives:

At $V = 500 \,\text{V}$:

$$F = \frac{1}{2} \cdot (3.2)(8.85 \times 10^{-12}) \cdot \frac{5 \times 10^{-4} \cdot (500)^2}{(1 \times 10^{-5})^2} \approx 18 \,\text{N}$$

At $V = 1000 \,\text{V}$:

$$F \approx 70 \,\text{N}$$

These are ideal attraction-force estimates; the actual force will likely be lower because of surface roughness and material properties. Still, they suggested the concept was worth testing.

Voltage generation

Getting a high-voltage output from 12 V without a bulky transformer was the main design challenge. It took some searching, but I eventually found the LT8365 boost converter, and took the charge-pump approach shown in its datasheet further. The reference circuit reaches 420 V using two pump stages while keeping the converter’s switch voltage much lower than the output.

I laid out eight charge-pump stages, targeting roughly 150 V per stage for a 1,200 V output. That kept the switch voltage below its rated maximum and gave me some margin for high-voltage experiments.

Bipolar output

The clutch also needs a way to release promptly. Leaving a DC voltage across the plates can saturate charge in the dielectric, so I designed an H-bridge to allow for bipolar switching. The switching frequency is adjustable in software, with most literature suggesting a range of 10–20 Hz for optimal performance.

The H-bridge uses four high-voltage MOSFETs in a standard configuration, with additional gate-drive components shown below.

Isolated gate drive

Driving the high-side N-channel MOSFETs was less straightforward. Their gates must sit above their source voltage to turn on, which would require a bulky high-voltage control supply. I also wanted to avoid bootstrapping to prevent issues with arbitrary control. Instead, I used APV1121S photovoltaic gate drivers for isolation. An LED on the control side drives a photovoltaic output that can float with each MOSFET’s source, providing its gate drive without a separate high-voltage control supply.

This is the H-bridge I built:

Software-adjustable voltage

Because the board is for experiments, I wanted to change voltage without swapping components. The LTC7106 current DAC injects a small current (on the order of microamps) into the boost converter’s feedback network so the microcontroller can control the output. Its digital output can also disable the high-voltage stage.

I added a high-voltage divider and a buffered measurement path so the controller can read the output voltage. Together with the current DAC and H-bridge, this allows the firmware to adjust both voltage and switching frequency using accurate closed-loop control.

Result

I built the driver and have confirmed that it works as designed. The board gives me a working platform for testing electrostatic clutches and exploring their potential for haptic gloves or other uses. I have built preliminary clutches and confirmed that they can generate a measurable resistive force, and I am continuing to explore their performance and potential applications.