Measurely

Measurely is a senior capstone project completed over two semesters. It is an assistive baking device designed to restore independence in the kitchen for people living with arthritis, Parkinson's, or similar mobility conditions. View the project website at measurely.co.

Baking requires a demanding combination of gross motor strength, such as pouring heavy ingredients, and fine motor precision, measuring spices, leveling spoons. For millions of people, those fine motor requirements present an insurmountable barrier. Managing physical tremors while simultaneously following complex recipe steps creates what researchers call Cognitive-Motor Interference (CMI), which significantly increases error rates and safety risks in the kitchen.

Existing assistive devices address only isolated problems. Large-button scales exist but don't guide the user; automatic dispensers exist but don't handle bulk ingredients. Measurely combines both in an automated dispenser that handles precise small-ingredient measurements, paired with an integrated scale and guided display that walks the user through each step.

System Components

• Dry ingredient dispenser, an auger/screw mechanism driven by a DC motor and linear actuator
• Integrated scale, an HX711 load cell with NeoPixel LED ring feedback
• Guided display, an Arduino Giga screen with tactile mechanical keyboard navigation
• Electronics & integration, covering circuit design, power management, CAD for electronics housing, and the universal key mechanism

The final dispenser uses an auger/screw mechanism driven by a DC motor. A single motor-driven key inserts into each spice slot's keyhole to rotate the auger, dispensing a precise amount. The key is positioned and engaged using an H-Bridge motor driver controlling a linear actuator, compact, reliable, and quiet compared to earlier approaches.

Linear actuator extending and retracting to position the DC motor key into the auger slot. Testing engagement timing before the universal key mechanism was finalized.

When we first loaded the screw with spice and tested, the jam force broke the key. The walls of the attachment are thin by necessity (they must fit through the center of the slip ring), leaving little material at the point of highest stress. I made three structural improvements in CAD:

• Added a chamfer at the stress concentration point
• Increased print infill and made the center column solid past the attachment joint
• Added a reinforcement ring at the base to prevent cracking under load

Tolerances were 0.005 in. Even printing the same file multiple times produced small inconsistencies, so I printed three copies and test-fit each to find the best match.

I also resolved a stepper motor issue during this phase. The motor was moving erratically on startup because it received a small pulse of current before the Arduino fully initialized. Adding a pull-down resistor on the step pin absorbed that initial current, eliminating the spurious movement, a small detail critical to dispensing accuracy.

Universal key engaging the auger screw, dispensing, and retracting. The dog clutch teeth self-align regardless of the auger's resting rotation angle — no manual positioning needed.

The scale uses a 5 kg HX711 load cell. To convert torque into weight accurately, I designed thin spacers using calipers-measured M4 and M5 hardware. The surface is a three-layer acrylic stack. The first layer mounts the load cell, the second spaces the screw head from the shank, and the third (slightly translucent) serves as the decorative top surface and diffuses the LED light evenly. The diameter is 7 inches, with the top plate extending 0.2 inches larger to accommodate the NeoPixel strip underneath.

One challenge was the gray metal screw was visible through the translucent top plate. I tried opaque tape, paper inserts, and surface sanding, all of which looked unprofessional. The fix was painting the screw white; the top layer's light-diffusing properties blurred the slight texture irregularities from the paint, producing a clean result.

Based on mentor feedback from Dr. Zarske, I upgraded from a simple on/off LED strip to an Adafruit NeoPixel ring (94 pixels) that fills progressively as weight approaches the target and changes color based on accuracy, yellow for underweight, green for on-target (±5 g), and red for overweight. An engraving on the top surface guides users to center their bowl.

#include <Adafruit_NeoPixel.h>
#include "HX711.h"

#define PIN               6
#define NUMPIXELS        94
#define DOUT              3
#define CLK               2
#define CALIBRATION_FACTOR  -205.0
#define TARGET_WEIGHT_G   272.155
#define TOLERANCE_G         5.0

Adafruit_NeoPixel pixels(NUMPIXELS, PIN, NEO_GRB + NEO_KHZ800);
HX711 scale;

void setup() {
  scale.begin(DOUT, CLK);
  scale.set_scale(CALIBRATION_FACTOR);
  scale.tare();
  pixels.begin();
  pixels.clear();
  pixels.show();
}

void loop() {
  float weightG = scale.get_units(3);  // average 3 readings
  if (weightG < 0) weightG = 0;

  uint32_t color;
  if      (weightG > TARGET_WEIGHT_G + TOLERANCE_G) color = pixels.Color(255, 0, 0);    // Red
  else if (weightG >= TARGET_WEIGHT_G - TOLERANCE_G) color = pixels.Color(0, 255, 0);   // Green
  else                                                color = pixels.Color(255, 150, 0); // Yellow

  // Fill arc proportional to weight
  int numLit = (int)((weightG / TARGET_WEIGHT_G) * NUMPIXELS);
  if (numLit > NUMPIXELS) numLit = NUMPIXELS;

  pixels.clear();
  for (int i = 0; i < numLit; i++) pixels.setPixelColor(i, color);
  pixels.show();
  delay(200);
}

Calibration was determined through systematic trial and error using known reference weights (my phone, then a friend's phone to verify independently). The final calibration factor of -205.0 accurately converted the load cell's raw torque readings into grams across multiple test weights.

Scale calibration test using a known reference weight. The NeoPixel ring fills progressively as weight approaches the target, shifting from yellow to green when within ±5 g.