3D Hall Sensor for Robotic Hands KTH5701AQ2DNE vs MLX90393 Parameter Comparison

Engineers choosing a position sensor for a dexterous-hand joint eventually land on the same question: should this part move from an imported device to a domestic one. Rather than reading two parameter tables end to end, filter the choice through the four gates that actually stall a design — range, compute, button input, and power. The two parts are CONNTEK's KTH5701AQ2DNE and Melexis's MLX90393, one domestic and one imported, with noticeably different parameter orientations. The gates come first, and the full comparison table sits at the end for item-by-item checking.

Gate 1 Magnetic field range

To keep finger joints thin, the gap between magnet and chip is usually squeezed to 1–3mm, and the small neodymium magnets used in dexterous hands easily exceed 50mT at the surface.

The MLX90393 limits all three linear ranges to ±50mT, which locks magnet selection down tightly — a slightly larger gap forces a stronger magnet, and assembly tolerance cannot be relaxed. The KTH5701AQ2DNE offers ±130mT on the XY axes and ±80mT on Z, leaving far more headroom, which widens the magnet and gap selection window and directly helps production-line consistency.

Gate 1 verdict: the KTH5701AQ2DNE gives clearly more freedom on magnet and gap selection.

Gate 2 Who computes the angle

The KTH5701AQ2DNE integrates three independent Hall sensors for the X, Y and Z axes, feeding a high-precision amplifier into a 16-bit ADC. A built-in CORDIC engine directly outputs planar angles for the XY, XZ and YZ planes, so the TABZ data the host reads back already contains angle and magnitude. Supply voltage is 2.8V to 5.5V, IO supply goes down to 1.8V, and the package is a QFN3×3-16L with a DFN2×2.5-8L option.

The MLX90393 is also a digital-output part measuring all three axes, but it has no hardware angle-computation block — it only emits raw TXYZ data, and the angle has to be computed in the MCU. One angle calculation per finger adds up: ten fingers means a dozen or more solver channels running on the main controller, and a raised sample rate easily eats into the budget for motor FOC.

Accuracy orients differently on the two sides: the KTH5701AQ2DNE specifies ±1°@40mT in its high-filter mode, while the MLX90393 reaches a typical ±0.5° with factory calibration, more gain steps and a maximum sensitivity of about 6211 LSB/mT — an advantage in small-field, high-resolution applications.

Gate 2 verdict: if the MCU budget is tight and hardware angles are preferred, take the KTH5701AQ2DNE; for small-field high-resolution work, the MLX90393 still has value.

Gate 3 Independent button pin

The KTH5701AQ2DNE has a dedicated BUTT_OUT pin that detects press events from the changing distance between magnet and chip, and it can also be configured as a Trigger output for single-shot measurements. When a hand needs tactile or press feedback at the fingertip, this one part covers both joint angle and fingertip press without an extra sensor. It also supports independent XY angle wake-up thresholds and Z field wake-up thresholds, so a single chip handles knob and button detection at the same time.

The MLX90393 shares INT and TRIG on one pin and has no dedicated button output, so that function needs another approach.

Gate 3 verdict: for a dedicated button or touch-detection pin, the KTH5701AQ2DNE is the more direct choice.

Gate 4 Power and temperature range

On power the two parts each win somewhere. Wake-up standby current on the KTH5701AQ2DNE is only 2.4μA — it wakes on a change in the magnetic field, takes one measurement, and returns to the low-power state otherwise. The MLX90393 draws 43μA, nearly a twenty-fold difference, which matters a great deal for battery-powered portable hands and robotic grippers during long idle periods. On the other side, MLX90393 conversion current is lower: 2.29mA for a single XY axis versus 4.89mA. If the system does continuous high-rate measurement and never cares about standby draw, that has to be counted.

Temperature and ESD form another pair of numbers. The KTH5701AQ2DNE is graded industrial -40°C to +105°C, while the MLX90393 is Grade E -40°C to +85°C with no 105°C option. ESD is ±5kV HBM versus 2.5kV HBM — twice the margin. On communication, the MLX90393 SPI tops out at 10MHz, twice the 5MHz of the domestic part, and both do I2C 100k/400k. The MLX90393 can also calibrate its zero point through an OFFSET register and offers four independent change thresholds for X, Y, Z and T.

Gate 4 verdict: for standby-sensitive designs, 105°C operation or more ESD margin, take the KTH5701AQ2DNE; for continuous high-rate measurement and higher SPI bandwidth, the MLX90393 fits better.

How to decide between KTH5701AQ2DNE and MLX90393

If any entry in the left column applies, take the part in the right column:

One reminder: the KTH5701AQ2DNE is not a drop-in replacement for the MLX90393. Pin definitions and register maps differ, so a new board and a rewritten driver are required, and the angle moves from MCU computation to direct readback.

KTH5701AQ2DNE vs MLX90393 full parameter comparison

Samples and volume supply

CONNTEK is a domestic magnetic sensor manufacturer. The KTH5701 series comes in both QFN3×3-16L and DFN2×2.5-8L, and the smaller DFN suits the space-constrained structure of a dexterous hand. Compared with imported parts, the domestic supply chain gives better control over lead time and price volatility, and volume production is less likely to be blocked by a single channel. For samples, datasheets or comparison test data, the distributor channel is the fastest route — Shenzhen Manduic Technology Co., Ltd. is a CONNTEK distributor and its contact details are at the bottom of the page.

For samples, product datasheets, detailed adaptation solutions or technical support, please contact us:

Tel: +86-150-1290-5940

Email: sales@manduic.com

Website: www.manduic.com