I am learning C51 and am writing a C function to control RC servos with AT89S52. The pulses that control the servo shaft are generated by Timer 1.

The function:

SetServoPo(X);

where 0 <= X <= 180, in degree, sends a control pulse to the servo.

The segment of the program that assigns values to TH1 and TL1 is:

float tmp;
unsigned char TH,TL;

--------------------
tmp = 65536 - (10 * AxisAngle + 460);   //with a 12MHz crystal
TH = tmp / 256;                         //Derive the upper 8-bit from "tmp"
TL = tmp - TH * 256;                    //Derive the lower 8-bit from "tmp"
--------------------

and the program worked fine with a 12MHz crystal.
After that I modified the program to work with a 22.1184MHz crystal, so the segment above was replaced with:

--------------------
tmp = 65536 - (10 * AxisAngle + 460) * 22118400 / 12000000;
TH = tmp / 256;
TL = tmp - TH * 256;
--------------------

...and the servos started trembling fiercely (which indicated Timer 1 was refilled with random values).

It happened to work correctly with this segment:

--------------------
tmp = 65536 - (10 * AxisAngle + 460) * 221184 / 120000; //that is, by removing the two zeros
TH = tmp / 256;
TL = tmp - TH * 256;
--------------------

Since I used to program in assembly I have no idea what was going wrong with the code in the past few days.

I have no access to the Keil C51 software after school but I don't think the difference of the EVAL and the FULL versions caused the problem because the code size is within the 2k limit.


Any guidance will be appreciated.


Stanley