Almost Cycle-accurate measurements

You can use SYSCLK to get a 32-bit timer count for cycles from the system clock, so you don't have to worry about overflows
You can use the AWU IRQ Handler to get interrupts when the AWU triggers.

This lets you capture the data very quickly, and with much more precision than currently available.

#include "ch32v003fun.h"
#include <stdio.h>

#define ITERATIONS 1
#define SERIALPLOT

volatile uint32_t ldelta;
uint32_t llast;

void AWU_IRQHandler( void ) __attribute__((section(".text.vector_handler")))  __attribute__((interrupt));

void AWU_IRQHandler( void )
{
    uint32_t l = SysTick->CNT;
    ldelta = (l - llast);
    llast = l;
    EXTI->INTFR = EXTI_Line9;
}

int main()
{
    SystemInit();

    // enable power interface module clock
    RCC->APB1PCENR |= RCC_APB1Periph_PWR | RCC_APB1Periph_WWDG;

    // enable low speed oscillator (LSI)
    RCC->RSTSCKR |= RCC_LSION;
    while ((RCC->RSTSCKR & RCC_LSIRDY) == 0) {}

    // enable AutoWakeUp event
    EXTI->FTENR |= EXTI_Line9;
    EXTI->INTENR |= EXTI_Line9;

    // configure AWU prescaler and configure AWU window comparison value
    // /16 /(63+1) ~= 120Hz
    PWR->AWUPSC = PWR_AWU_Prescaler_8;
    PWR->AWUWR = 63;
    PWR->AWUCSR = (1 << 1); // AWUEN

    // enable interrupt
    NVIC_EnableIRQ( AWU_IRQn );

    uint32_t t;
    uint32_t lastt = TIM1->CNT;

    while(1)
    {
        // Wait for data to become available.   
        uint32_t printtime = 0;
        while( !( printtime = ldelta ) );
        ldelta = 0;
        printf("%d\n", printtime ); //raw_to_temperature(count));
    }
}

Just tested it at my desk here.

Also, I am seeing significant variation per part.

Part Test	Count at 25C	Count at -19 C	Counts Per Deg C
1	200300	203100	63
2	211000	215000	90
3	189500	192100	59

Additionally flex your PCB

Quescent at 25 C: 189900 Diagonal Torsion: 1888537 Inverse Torsion: 191050

Just using my little PCBs I had at my desk applying about 1 N*cm of force to the PCB.

eeucalyptus / ch32v003-temperature

Almost Cycle-accurate measurements #2