Servo - August 2008

the 1 MHz clock, or once every microsecond.

As you can see, the pwm_out <= (window_32ms < (servo_vector + minpulsewidth)); Verilog statement uses the 32.768 ms clock, the servo_vector value, and the minimumpulsewidth constant to create a complete PWM signal. The proof is in the pudding. One slice of servo driver pie is represented in Screenshot 1, which is the 1.5 ms pulse our Verilog code generated. I captured the servo rotor-centering pulse for you with a CleverScope. Screenshot 2 is another CleverScope shot of the complete 32.768 ms servo control pulse window, including the 1.5 ms servo position pulse.

With the XC2C64A code, we’ve put together so far, we can change the position of the hobby servo rotor by simply changing the value of the servo_vector parameter. Valid servo_vector values range from zero to 1000 decimal, with a zero value producing a 1 ms servo position pulse and a value of 1000 decimal pushing the hobby servo to its full opposite extent with a 2 ms pulse. We have easily taken control of one hobby servo using only our 64A CPLD. Why not two hobby servos??

One XC2C64A. Two Servos.

You already understand what it takes to move a hobby servo rotor with an XC2C64A CPLD. So, let’s not waste any time getting some double rotor code down and running. First, let’s add that second PWM output to our rcservo module:

module rcservo(
 input clk_1mhz,
 output reg pwm_out1,
 output reg pwm_out2
 );

We’re going to take a different turn here. Instead of loading the servo rotor position manually via the servo_vector parameter, we’re going to instantiate a pair of servo_vector registers to hold our desired servo rotor positions:

parameter

minpulsewidth = 1000;

reg
reg
reg
reg

[14:0] window_32ms;
[9:0]  servo_vector1;
[9:0]  servo_vector2;
servo_direction;

Recall that our servo_vector values can range from zero to 1000 decimal. That means we need 10-bits to hold the maximum servo_vector values. The single bit servo_direction register should give you a clue as to where that different turn will take us.

Nothing has changed about the way we generate the servo control pulse window:

The CPLD Servo Driver

//*******************************************************
//* GENERATE SERVO CONTROL PULSE WINDOW
//*******************************************************
always @(posedge clk_1mhz)
 window_32ms <= window_32ms + 1;

We will build upon our single servo code to add the second PWM output pulse. We laid the ground work for the extra PWM output earlier in the rcservo module:

//*******************************************************
//* GENERATE SERVO CONTROL PULSES
//*******************************************************
always @(posedge clk_1mhz)
begin

pwm_out1 <= (window_32ms <
 (servo_vector1 + minpulsewidth));
pwm_out2 <= (window_32ms <
 (servo_vector2 + minpulsewidth));

end

Notice that we replaced the servo_vector constant with our servo_vector register variables.

Up to now, we’ve been seeing double. Let’s drive down that road we made the different turn onto. I don’t think you’ll have any problem following the idea behind this Verilog code snippet:

//*******************************************************
//* DETERMINE SERVO ROTOR DIRECTION
//*******************************************************
always @(posedge clk_1mhz)
 begin
 if(window_32ms == 0)
 begin
 if(servo_vector1 == 0 || servo_vector1 == 1000)
 servo_direction = ~servo_direction;

 if(servo_direction)
 begin
 servo_vector1 <= servo_vector1 + 1;
 servo_vector2 <= servo_vector2 - 1;
 end
 else
 begin
 servo_vector1 <= servo_vector1 - 1;
 servo_vector2 <= servo_vector2 + 1;
 end
 end
 end
endmodule

Everything we do must synchronize to the beginning of the servo control pulse window. So, we must always check to see if the servo control pulse window counter register is at zero as this is the synchronization point we must adhere to. The idea is that if we have a new servo vector value to enter, it must be entered at the beginning of a new servo control pulse window.