Ok, for awhile I had some real trouble with your code, but I think I
have figured it out. I guess I have trouble doing 8 bit binary logic in
my head. Plus IOR and XOR sometimes confuse me. Please check my comments
within to see if I've got it.

I will likely be converting this to a timer based system, based on a 28
or 40 pin PIC. This would give me enough pins for one or two 8 channel
PWMs, and an 8 bit + control lines parallel port for input. The
controller will likely be a 28 pin PIC that receives data via UART, and
has an 8 bit data port common with all the slaves, and individiual
control (address) lines. We'll see, I haven't figured out the control
end quite yet. I will have to build in a failsafe so if one of the
slaves freezes or dies, the controller doesn't sit in a loop waiting for
a response. I figured on one line for a "Data Ready" and another one for
"Request to Send". DR would be output from controller, and RTS would be
input to controller. The controller would assert DR when it wanted to
update the slave, then wait for the slave to assert RTS. Then the
controller would put data on the bus, and either drop and reassert DR,
or perhaps assert another pin. Then the slave would aknowledge
transmission via RTS, and wait for the next value. The problem I see is
that if the slave dies, the controller may sit in an endless loop
waiting for a RTS from the slave. I hope that isn't too confusing.

The other thing I was thinking is that if I switch to a 16f628, and a
hacked up parallel comms between slave and master, I might be able to do
some interesting stuff. If I don't care about PWM frequency, could I
just run the PIC on the internal clock? Since PWM is more about a ratio
between on and off, if I don't care about the total period, I should be
able to get away with internal oscillator, and let it run however it
wishes, right? As long as I plan the parallel comms with enough leeway
for both extremes of course.

Anyway, see comments in the code.


Scott Dattalo wrote:
(I'll use the abbreviated version of the code here to save space)

>         CLRW                    ;Build the bit mask for turning
>                                 ;off the PWM outputs. Assume that
>                                 ;all of the outputs will be turned
>                                 ;off.
>                                 ;
>         DECFSZ  pwm0,F          ;If the first counter has not reached 0
>          IORLW  00000001b       ;then we don't want to turn it off.
>                                 ;
>         DECFSZ  pwm1,F          ;Same for the second one
>          IORLW  00000010b       ;

Create a mask of all the bits on the PWM port that have had their
counters expire. A 0 indicates an expired timer.

>
>         ANDWF   pwm_state,W     ; Clear all of those pwm outputs
>                                 ;that have reached zero.

AND the mask with the current state. If an output was 1 and the timer
expired, you get 0, if it's 1 and the timer hasn't expired, you get a 1.
So you end up with what the new state should be, assuming the main timer
hasn't expired.

>                                 ;
>         XORLW   00000011b       ;Toggle the current state.

Basically invert the mask, 1 becomes 0, 0 becomes 1.

>         INCFSZ  rising_edge,F   ;If the rising edge counter has not
>          XORLW  00000011b       ;rolled over then toggle them again.
>                                 ;Double toggle == no effect. However,
>                                 ;if the rising edge counter does roll
>                                 ;over then a single toggle will turn
>                                 ;the pwm bits on, unless of course the
>                                 ;pwm counter has just rolled over too.

If the main timer has expired, then you accept the inverted state, if it
has expired, you invert again, which ends up righting the output, and
then you can copy it right to the port.

*** What happens if you have a timer that expires while the mask was
originally a 0? When you invert it the first time, it becomes a 1, which
then would go and turn on the port. If it was a 1 though, then it
becomes a 0, and the port is turned off. That should only occur if the
value in the pwmX register was equal and in sync to the rising_edge main
timer register. Does that mean that to have a PWM port off, you then
need to reload the pwmX register with a 0x01 so it will time out
immediately? Also, should there be a subroutine that runs after the main
timer expires that will reload all the individual pwmX registers?

>         MOVWF   pwm_state       ;Save the state
>         MOVWF   PWM_PORT        ;update the outputs

Ok, super long post, but comments truly appreciated. I am just trying to
get my head around a lot of this.

Thanks,

Josh
--
A common mistake that people make when trying to design something
completely foolproof is to underestimate the ingenuity of complete
fools.
        -Douglas Adams

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics