Started on 03-06-2011
Update on 12-12-2014: the pendulum STOPs swinging after 3,5 years with batteries now at 1.4V


The second useless machine ever
The first, the real "MOST Useless Machine Ever" can be found at:

This project is almost useless, I want just to realize a pendulum that swings endless. Nothing more. No use at all. This is not an original project but my version of a project of the past century, already seen in many different ways on the Net.

#include <htc.h>

#define _XTAL_FREQ 4000000


volatile unsigned char control TRIS @ 0x06; // it is not declared in PICC 9.81

void main()


char Temp;

// Enable Wake-up on change on GP0-1-3, disable pull-up, prescaler 128 to WDT


ADCON0bits.ANS1=0;   // GP1= digital

TRIS=0b11111011;     // Set Only bit GP2=OUTPUT, GP1=INPUT



//    if the program starts because of interrupt or because of watchdog

//    it kicks the pendulum

//    __delay_ms(1); // wait for the best moment to kick

      GP2=0;           // ON, because of the PNP transistor it is inverted

      __delay_ms(10); // kick time

      GP2=1;           // OFF

      __delay_ms(30); // wait for the end


      // read input pin before entering sleep, read datasheet for info





One transistor and two coils or two transistors and one coil?

The same coil can be used both as a sensor and an actuator.

No! A transistor and a PIC. The smallest one (just six pins) but still an MCU.
This is my first time with a so small PIC. It's very interesting. It can be used for a lot of different situations, substituting transistors, logic ports, NE555 and so on with a lot more flexibility.
To the left, the signal at TP1 (see schematic diagram), at the coil. Choosing the right magnet and coil, the voltage is enough to trigger a digital port of a PIC (a peak of 3V in this picture), no ADC is used. In this way the PIC can go sleeping for a long time between pulses, it works just for 10ms for each semi-cycle. The "Wake on input change" feature restarts the program every time the magnet passes over the coil.
For a small swinging pendulum, the period can be calculated with the formula:
where: L = pendulum length and g = acceleration of gravity (9.81 m/s^2).
In practice the PIC is awake for 10ms every 500ms for a 25cm long pendulum (2% duty cycle), 10ms every 1000ms for 1m long pendulum (1% duty cycle).
With a small pendulum the pulse can also be shorted because less energy is required, with an even less power consumption.
It is interesting to see how the shape and the amplitude of the signal varies using different type of magnets.
To the right, the signals at TP2 (trace A) and TP3 (trace B).
At TP2 the signal, compared to TP1 one, is rectified and amplitude limited by the Resistor-Diode-Zener network.
At TP3 we have the pulse that triggers the transistor for a new kick to the pendulum. I have used a PNP transistor to have a high side driving for the coil, so the pulse must be negative.
The PIC10F222 works fine in the 2 to 5.5V power supply range. The high level threshold for a digital input is related to the power supply. With 3V power supply we can see that the trigger starts when the voltage at GP1 rise above 1V. With a 5V power supply it needs an higher lever and therefore more time to start.
I have found the right delay for the kick and the right duration experimentally.

So... once again I've linked mechanics, physics, a little bit of maths, electronics, programming and photography too to realize and share this project.
The goal #1 is fully achieved. I hope some one else can have fun too with this example.
There is a direct relationship between the amplitude of the input signal and the amplitude of the swing. Because the period of the pendulum is constant, for a larger swing (more distance to travel) it needs to run faster (V = S/t). The induced electromotive force (EMF) in any closed circuit is equal to the time rate of change of the magnetic flux through the circuit (Faraday's Law of Induction). So, you can know if the kick is correct looking at the amplitude of the signal at TP1, it increases slightly at every cycle until it reaches the equilibrium.
There are a lot of different schematics on the Net. Someone uses a sensor (Hall effect or optical) to reveal when the pendulum is approaching the coil, someone else uses two coils, one as a sensor and one as an actuator. There are also some circuits with MCUs (Atmel or Microchip) with additional sensors or using an analog port to read the coil voltage. In my optimization mania I want to Keep It Simple as much as possible. Furthermore (in order to satisfy goal #3) the MCU must go sleep most of the time to save batteries.

As the picture shows the circuit is very simple even on a perfboard. No quartz, very few external components. Most of the wires below the board are to connect ICSP and test plugs. Once programmed and tested they can be removed from the schematic saving even more space.

Do you remember the first kind of electromechanical clocks? Have you ever "deeply examined" one of them? I did. I have destroyed... hem... examined someone of them during my childhood.
The basic theory is the same I've used. The balance wheel has permanent magnets and it is driven by a coil with a very simple electronic circuit. The electric pulse to the coil is synchronized with the balance wheel period by sensing the magnets. In this way we have an oscillating system that requires a very little amount of energy to keep moving. The power is provided by a battery instead of a spring. The rest of the gear mechanism is the same as a traditional clock.

I have always been fascinated by this kind of extremely simple mechanism. I'm collecting magnets, coils and everything that can be useful to realize it since a long time.

My interest was re-awakened by an article published on Nuts & Volts of September 2009 by Tony Gasparovic about a pendulum driven by two transistors:
A quick googling shows up how many similar projects there are:
I'm not alone! Which circuit may I use? None of them of course. I want my own system, using something more sophisticated like an MCU. Using one or two transistors doesn't add anything to my experience.
My goals are:
1) to have fun
2) to learn something more about PICs
3) to have a circuit that can be powered with just two AA batteries for a very long time.
4) to review some of the knowledge about physics I'm going to forget after so long time.
So, it is no so useless.
Also the eyes must be satisfied. Using a nice wooden box allows us to keep the pendulum exposed also out of the laboratory.
I was lucky on finding a box with a hole on top to allocate the coil as much as close to the path of the magnet.
The magnet acts as the massive bob of the pendulum and the active part of the circuit too.
Using a telescoping antenna for the rod I'm free to fix the fulcrum everywhere and trim the exact length for the right position.
A little bit of theory and practice
The software is so simple.
I've used MPLAB IDE with HI-TECH C compiler for PIC10/12/16 MCUs,
The configuration fuses sets internal clock to 4MHz, it is more than enough for our purposes and save some power compared to 8MHz clock. It also enables the watch-dog to increase reliability of the circuit.
The TRIS declaration is because of a bug in the include files of version 9.81 of the compiler.
With OPTION register it sets the "wake on input change" feature to restart the program from the sleep status when the magnet passes over the coil. The 1:128 prescaler is applied to watch-dog timer to have a 2.3s period for it.
When the program starts, for the first time, when input changes or for a WDT timeout, it reset the GP2 output enabling the PNP transistor that powers up the coil for 10ms. Then waits for the magnet passing away, to avoid another trigger of GP1, resets WDT, resets inputs for the next event and go to sleep.
That's all folk!
The software
The mechanics
The fulcrum must oblige the pendulum to oscillate in only one direction, a flexible rod is not allowed, otherwise the coil could kick the magnet in strange directions.
For this reason I've used a "knife" style fulcrum that adds very low friction still maintaining the right oscillating path.
The movement shown with a picture in ambient light and a slow shutter speed for the camera
The pendulum movement frozen by the flash light.