ECE319K Lab 7 — ADC Sensing and LCD Display System

Overview

This project builds a real-time position sensing system on the MSPM0G3507. A slide potentiometer connected to PB18 gets sampled by a 12-bit ADC at 30 Hz using a hardware timer interrupt. The raw ADC value gets converted to a fixed-point distance measurement (0.000 to 2.000 cm) using a calibrated linear equation, then displayed on an ST7735 LCD over SPI. The system also plots a scrolling history of position values on the LCD screen.

This lab was the foundation for Lab 8, where the same ADC and Convert pipeline got extended into a full distributed UART communication system across two microcontrollers.

How It Works

ADC Sampling

The slide potentiometer outputs a voltage between 0 and 3.3V on pin PB18. ADC1 channel 5 samples this voltage and returns a 12-bit result (0 to 4095). The ADC runs at 40-48 MHz with a divide-by-8 clock, software triggered, no hardware averaging.

Calibration and Convert Function

Five physical measurements were taken at known potentiometer positions and recorded in Calibration.xls. A linear regression fit produced two constants k1 and k2 that map raw ADC values to distance in units of 0.001 cm: Position = ((k1 * Data) >> 12) + k2 Position = ((1739 * Data) >> 12) + 148 The right shift by 12 keeps everything in integer arithmetic without floating point. This matters on the Cortex M0+ because floating point operations take roughly 100x longer than integer operations. FloatConvert was implemented alongside Convert specifically to benchmark this difference using SysTick timing.

Fixed Point Display

OutFix takes the integer position value (0 to 2000) and formats it as X.XXX cm for the LCD. No floating point involved. The whole and fractional parts are split with integer division and modulo:

whole = n / 1000;   // ones digit
part  = n % 1000;   // three decimal digits
printf("%u.%03u cm", whole, part);

Interrupt Architecture

TimerG12 fires every 33.3ms (30 Hz) and sets a semaphore flag. The main loop spins waiting on the flag, clears it when set, then runs the Convert and OutFix pipeline. This keeps the ADC sampling precisely timed regardless of how long the LCD update takes. TimerG12 ISR (30 Hz): Main loop: Sample ADC then Wait for flag Set Flag = 1 then Clear flag Return then Convert raw ADC to distance then Display on LCD then Plot scrolling graph

Scrolling Plot

Every 15 samples (about 2 Hz) the system plots a point on the ST7735 LCD and advances the cursor, creating a scrolling position history graph. ST7735_PlotNextErase handles the scrolling behavior.

Hardware Configuration

Component	Detail
Microcontroller	TI MSPM0G3507 (80 MHz)
ADC	ADC1, 12-bit, channel 5, PB18
Display	ST7735 LCD via SPI
Potentiometer	Slide pot, pin 2 to PB18
Sampling rate	30 Hz via TimerG12
Timer period	2,666,667 cycles (80 MHz / 30 Hz)

Calibration

Calibration was performed by measuring raw ADC output at 5 known physical positions across the slide potentiometer range. A linear fit through these points produced: k1 = 1739 k2 = 148 Position (0.001 cm) = ((1739 * ADC) >> 12) + 148 Full range: ADC = 0 → Position ≈ 148 (0.148 cm)
Full range: ADC = 4095 → Position ≈ 1881 (1.881 cm)

Performance Measurements

Measured using SysTick cycle counter at 80 MHz:

Operation	Time
ADCin() execution	~1 µs
Convert() execution	< 1 µs (integer math)
FloatConvert() execution	~100x slower than Convert
OutFix() execution	~6 ms (LCD SPI transfer)
Sampling period	33.3 ms (30 Hz)

The FloatConvert vs Convert comparison was the most interesting result. The Cortex M0+ has no hardware floating point unit so every float operation gets emulated in software. The integer fixed-point approach is dramatically faster and good enough for this application.

Files

File	Description
ADC1.c	ADC init, sampling, Convert, FloatConvert
Lab7Main.c	Multiple test mains, final interrupt system
Calibration.xls	Raw calibration data and linear fit

Test Structure

The lab was built incrementally through six test mains:

• main1 -- verified potentiometer voltage range using TExaS logic analyzer
• main2 -- tested ADCinit and ADCin, measured execution time with SysTick
• main3 -- tested Convert accuracy against physical measurements
• main4 -- tested OutFix display timing on LCD
• main5 -- final system with TimerG12 interrupt and scrolling plot
• main6 -- Central Limit Theorem study using histogram averaging

What I Learned

The biggest takeaway from this lab was understanding the real cost of floating point on embedded hardware. I implemented both a fixed-point and floating point version of the same conversion and measured the difference directly with the SysTick timer. Seeing that gap in actual clock cycles made the case for fixed-point arithmetic much more concrete than any textbook explanation could.

Getting the semaphore pattern right between the ISR and main loop was also a good exercise. The ISR just sets a flag and returns immediately. It does the minimum possible work. Main does the heavy lifting. That separation of responsibilities carried directly into Lab 8 where the same pattern scaled up to a full producer-consumer UART system.