Simple Joystick control of a servo motor with RTAI/Linux

By Pramode C.E.

The analog joystick which plugs onto the PC game port is a cool little device - you don't need to be a hardware wizard to learn how it works, and you can make it do fun and absolutely useless stuff like turn a stepper or servo motor. This article describes an experiment which I did with a joystick and a Futaba S2003 servo motor, both controlled by the real time operating environment for Linux called RTAI. Readers who are not familiar with RTAI might like to refer to the introductory article I had written earlier before continuing.

The Joystick

If you are game-crazy, you are sure to have used one. As far as electronics is concerned, it is a very primitive device - so is the game port which it plugs into. You will find adequate hardware descriptions here. The Linux IO-Port Programming Mini-HOWTO also provides sufficient information to get started with hacking the joystick.

Reading the buttons

Your first joystick programming assignment should be reading the state of the buttons. For this, you have to know the ISA port address which the gameport uses. Loading the standard Linux joystick driver (you will have to modprobe three modules - joydev, ns558 and analog) and doing `cat /dev/ioports' on my ASUS A7N266VM motherboard showed this to be 0x200.

The state of two of my joystick buttons is encoded in bits D4 and D5 of the 8 bit value returned by an `inb' on 0x200 (D0 is least significant bit and D7 is most significant bit). If the value is 1, the button is in the `released' state and if it is 0, the button is in the `pressed' state. Here is a small program which tests this out:

Listing 1

#include <asm/io.h>
#define JS_PORT 0x200

main()
{
    iopl(3);
    while(1)
        printf("%x\n", (inb(JS_PORT) >> 4)&1);
		
}

Reading the X and Y positions

Moving the joystick results in a potentiometer turning - the potentiometer is connected to the game port, which contains a 555 timer based circuit. A simple `outb' to 0x200 (the value written doesn't matter) will result in the circuit getting `reset' - now a read (ie, an `inb') from 0x200 will yield a bit pattern whose D0 and D1 bits are 1's. Keep on reading - after a short time these bits become zero. Measure the time it takes for the bits to become zero. Take the measurements with the joystick at the extreme left, middle and extreme right endpoints of the X axis as well as the top, middle and bottom points of the Y axis as part of a `calibration' process. Now, whatever be the position of the joystick along the X-Y axes, measuring the time it takes for the D0 (X axis) and D1 (Y axis) bits to become zero's after a `reset' (note, we `reset' by writing something to 0x200) should help us find it out (assuming that time varies linearly with distance between the middle and left/top as well as middle and right/bottom endpoints - which I really haven't verified).

Here is a program which measures the time it takes for the X-axis bit to become zero after a reset. It uses the `time stamp counter' which is a 64 bit counter available on all machines with, I believe, a Pentium and above CPU. If you have a 1GHz CPU, the timer gets incremented at a rate of 1,000,000,000 per second. My Athlon XP CPU runs at a clock speed of 1462.904 MHz (read from /proc/cpuinfo) - with this information, it is easy to compute the time elapsed between any two points in your program. The time stamp counter (TSC) can be read using a macro called `rdtsc' defined in the file /usr/src/linux/include/asm/msr.h.

Listing 2

#include <asm/io.h>
#include <asm/msr.h>

#define JS_PORT 0x200
#define CPU_HZ 1462904000

void  trigger(void)
{
    outb(0x0, JS_PORT);
}

main()
{
    unsigned int low1, high1, low2, high2;
    iopl(3);
    trigger();
    rdtsc(low1, high1);
    while(inb(JS_PORT) & 1);
    rdtsc(low2, high2);
    printf("low1=%u, high1=%u, low2=%u, high2=%u\n", low1, high1, low2, high2);
    printf("low2 - low1 = %u\n", low2 - low1);
    printf("in ms = %f\n", (((double)(low2-low1))/CPU_HZ)*1000);
}

The program should be compiled like this:

cc -I/usr/src/linux/include -O2 xmeasure.c

I got readings of 0.0262ms, 0.68ms and 1.60ms for the left, middle and right positions.

One trouble with this crude form of analog-to-digital conversion is that you have to sit in a loop waiting for the bits to drop to zero - this burns up CPU cycles. A better design would have been for the joystick hardware itself to perform the A-D conversion and send the resulting numbers to the PC - thus avoiding lots of software overhead.

Using a periodic RTAI task to sense joystick position

My experiment is this: I have a servo motor connected to the parallel port. The servo is not capable of rotating the full 360 degrees - it describes an arc of about 180 degrees. When I turn the joystick left, the servo also moves to the left end of the arc. When the joystick is in the `middle' position, the servo positions itself near the centre of the 180 degree arc and when the joystick moves towards the right end, the servo also moves towards the right end of the arc. Note that I try to sense only three joystick positions - left, middle and right.

The picture above shows two servo motors - the one at the bottom serving to rotate the platform resting on it - it is this servo which I will be moving with the joystick.

The idea is simple. A periodic task (period .33 milliseconds) monitors the joystick. At the first activation of this task (lets say at time 0), we `trigger' the game port (by writing to it) and assume that joystick position is `LEFT'. The next activation of the task would be at 0.33 milliseconds. If reading from the game port tells us that the X axis bit (D0) is still set, we assume that the joystick is in the `MIDDLE' position. The next activation of the task would be at 0.66 milliseconds - but we are not interested in checking the X axis bit at this point - we will take it that the joystick is in the `MIDDLE' position if the bit stays high for a period between 0.33 and 0.99 milliseconds (note that the `bit-high' times obtained experimentally were 0.026, 0.68 and 1.60 milliseconds respectively for extreme left, middle and extreme right positions). The next activation would be at 0.99 milliseconds - if bit D0 still stays high, we assume that the joystick is in the `RIGHT' position. Only at this point are we sure of the actual position of the joystick - we shall set a global variable, `joystick_position' to LEFT, RIGHT or MIDDLE.

Now we come to the servo motor control part - which is fairly simple. A hobby servo motor requires a `control pulse' on its white wire. The total on-off time of the pulse should be around 20 milliseconds - it is the 'on' time which actually controls the servo's position. My servo moves to one end of a 180 degree arc for an 'on' time of about 0.5 millisecond and moves to the other end for an 'on' time of about 2.2 seconds. A separate RTAI task controls the generation of this signal. A global array called `on_time' maintains the three 'on'-time values which would move the servo to the left, middle and right points of its arc. The servo task makes pin 3 of the parallel port (to which the servo's control wire is connected) high for a period of on_time[LEFT] if the current joystick position is `LEFT' - similarly for MIDDLE and RIGHT also. The 'off' time of the control pulse is stored in a variable `off_time' and is computed in such a way that the total 'on' plus 'off' time is around 20 milliseconds.

[Listing 3]

static void 
pwm_servo(int t)
{
    /* Servo is controlled by
     * signal on pin 3 of LPT1
    */
    while(1) {
        outb(2, LPT1_BASE); /* Pin 3 high */
        rt_sleep(on_time[joystick_position]);
        outb(~2, LPT1_BASE);
        rt_sleep(off_time);
    }
}

Conclusion

It has been fun playing with the joystick. I would like to know if there is a good method to monitor the joystick position continuously without loading RTAI too much (by increasing the timer frequency or resorting to busy loops) - let me know if you come across anything interesting. I can be contacted via my home page at pramode.net.

 

---

I am an instructor working for IC Software in Kerala, India. I would have loved becoming an organic chemist, but I do the second best thing possible, which is play with Linux and teach programming!