ESPCLOCK4 - A Rethink On Tick Pulse Generation

Having dabbled recently with PWM (Pulse Width Modulation) for motor control, it struck me that maybe the same technique can be applied for pulsing the Lavet stepper motor in the analog clock.

Just a recap. The Lavet stepper motor within the analog clock is design to function at ~1.5V. However, the output on the ESP8266 and ESP32 is 3.3V, and in ESPCLOCK3, the ATtiny85 is powered at ~3V. For reasons I don't fully understand, the pulse width needs to be as long as 100ms in this case, otherwise the clock will not tick reliably. This also leads to high power draw, because higher voltage + longer pulse width. In contrast, the original clock signal is 32ms pulse width at ~1.5V (which can go as low as ~1.1V since it is running on a normal AA battery).

So in ESPCLOCK3, a diode bridge is introduced, where 2 diodes in each direction drops the 3V output by the ATtiny85 down to ~1.8V (3V - 0.6V x 2), which worked remarkably well in reducing the power draw and making the clock tick with 100% reliability.

Much like PWM in DC motor control, the idea is to eliminate the diode bridge and instead output a PWM modulated pulse signal, which should reduce the effective voltage to the clock.

With some experimentation, the ideal duty cycle of the PWM was found to be 62.5%. The output is 1.25ms on, 0.75ms off, repeated 20 times (for a total of 40ms). Both 62.5% and 40ms were found to be lower bounds for reliable operation.

The 62.5% duty cycle surprised me, because from my naive calculation, for an effective voltage of say, 1.5V, the duty cycle should be 1.5 / 3.3, which is 45%. 62.5% duty cycle implies an effective voltage of ~2V, which is much higher than I expected. I suspect my naive formula for calculating the duty cycle is wrong.

It was quite easy to modify the test program to output PWM pulses. This macro was added:

// Pulse given tickpin - 40ms duration = 20 x (1.25/2ms) PWM => 62.5% duty cycle ~ 2V
#define X_TICK_PIN(pin) \
    I_MOVI(R0, 20), \
  M_LABEL(_PULSE_##pin), \
    X_GPIO_SET(pin, 1), I_DELAY(10000), \
    X_GPIO_SET(pin, 0), I_DELAY(16000-10000), \
    I_SUBI(R0, R0, 1), \
    M_BGE(_PULSE_##pin, 1)

The macro is then used in the ticking logic:

    // Decide which pin to tick
    X_RTC_LOAD(TICK_PIN, R0),                           
    M_BGE(_HANDLE_TICKPIN2, 2), 
    // Pulse tick pin 1
    X_TICK_PIN(TICKPIN1),
    X_RTC_SAVEI(TICK_PIN, 2),
    I_HALT(),
  M_LABEL(_HANDLE_TICKPIN2),
    // Pulse tick pin 2
    X_TICK_PIN(TICKPIN2),
    X_RTC_SAVEI(TICK_PIN, 1),
    I_HALT()

I was pleasantly surprised this worked very well. It made the clock tick reliably without the diode bridge. Even better, this approach brought the power draw from 1.37mA down to 1.13mA, so it's a keeper!

ESPCLOCK1 / ESPCLOCK2 / ESPCLOCK3 / ESPCLOCK4