Synchronous pulses

[Mike, K8LH]

Hi Andrew,

The disadvantage of this particular method is that you need a PIC with lots of RAM, something like an 18F2620, for the 1024 byte toggle array. If this is a one-off project then perhaps that may not be a big deal.

The PrepArray() routine builds the following (abbreviated) toggle[] array from the pulse[] array values and the Output() routine uses those values to update LATB as shown below;

 interval                                ^ LATB
0000-usecs  toggle[   0] = 0b00001111  0b00001111  RB3..RB0 toggled on
0001-usecs  toggle[   1] = 0b00000000  0b00001111
....
0043-usecs  toggle[  43] = 0b00000000  0b00001111
0044-usecs  toggle[  44] = 0b00000010  0b00001101  RB1 toggled off
0045-usecs  toggle[  45] = 0b00000100  0b00001001  RB2 toggled off
0046-usecs  toggle[  46] = 0b00000000  0b00001001
....
0119-usecs  toggle[ 119] = 0b00000000  0b00001001
0120-usecs  toggle[ 120] = 0b00000001  0b00001000  RB0 toggled off
0121-usecs  toggle[ 121] = 0b00000000  0b00001000
....
1000-usecs  toggle[1000] = 0b00000000  0b00001000
1001-usecs  toggle[1001] = 0b00001000  0b00000000  RB3 toggled off
....

The advantage of using a "toggle" data array and XOR'ing the values with LATB over using an "output" data array and simply writing the values to LATB is that the "toggle" array only requires inserting eight "toggle" bits into the array. An "output" data array would require inserting many many "output" bits into the array.

The Do-While loop in the Output() routine compiles to the following 5-cycle instruction sequence (by the BoostC compiler) for perfect 1-usec intervals (with a 20-MHz clock);

loop:
        movf    _postinc0,W     ;
        xorwf   _latb,F         ;
        btfss   _fsr0h,3        ;
        bra     loop            ;

In your program you would edit the pulse array values for the outputs and then call the PrepArray() routine to build the toggle array. Then I suspect you would wait for some "trigger" event before calling the Output() routine.

Any of this make sense Sir? I'm rather pressed for time and that tends to make my rushed explanations rather useless to most people. Sorry...

Regards, Mike

[AndrewC]

More or less makes sense.
As I'm only juggling three outputs, I think I'll just sort the outputs into increasing time, calculate the incremental time between trailing edges, put them all on together and then pull them back down sequentially with PAUSEUS delays.

Thanks for taking the time to answer

Andrew

[Mike, K8LH]

You're welcome.

Good luck on your project.

Regards, Mike

[Jerson]

This last post made me to put on my thinking cap I do this stuff within an ISR usually, but your case doesn't need it.

How about something like this. BEWARE : untested code

Out1    var   PORTB.0
Out2    var   PORTB.1
Out3    var   PORTB.2
AllOn    con  %00000111      ' all three outputs on

Delay1 var   word                 ' turn off delay for output1
Delay2 var   word                 ' and 2
Delay3 var   word                 ' surely this has to be for 3

RunDelay var word               ' this is the running timer

'  set all outputs on together
TRISB = 0
Delay1 = 1000                     ' in units of PAUSE time
Delay2 = 2000
Delay3 = 3000
Loop:
    PORTB = AllOn

    for RunDelay = 0 to 65535  '  or whatever is the maximum you want
                                               ' this would be your cycle  / repeat time
      if RunDelay > Delay1 then
         Out1 = 0
      endif

      if RunDelay > Delay2 then
         Out2 = 0
      endif

      if RunDelay > Delay3 then
         Out3 = 0
      endif
   
      pauseus 10                        ' this is the resolution of your delays
    next
goto Loop

[Mike, K8LH]

Hi Jerson,

I agree that ISR code isn't necessary or even desirable in this case because of the high resolution required and the ISR context save/restore 'overhead'.

Andrew mentioned in his first post that the pulse width range could be several microseconds up to about one millisecond.

I believe the pulse width resolution for your example would be the loop code 'overhead' plus the 16-bit compare code 'overhead' plus the delay. The 16-bit compare code 'overhead' may well be the biggest limiting factor affecting resolution and the delay may not be necessary. Have you checked to see how many cycles are used for 1 loop cycle in your example Sir?

Regards, Mike

[Mike, K8LH]

Hey Andrew,

Once I figure out how to force an address for the toggle array in Swordfish BASIC I suspect my method might look something like this BASIC example, which I hope makes more sense than my previous pseudo C example.

Kind regards, Mike

Dim Tgl(1024) As Byte                   ' ??? force to $400..$7FF ???
Dim Pulse(3) As Word
Dim LastElement As FSR0H.booleans(3)
'
    Pulse(0) = 3                        ' RB0 pulse, 0..1023 usecs
    Pulse(1) = 11                       ' RB1 pulse, 0..1023 usecs
    Pulse(2) = 12                       ' RB2 pulse, 0..1023 usecs
'
'  clear the 1024 element toggle array
'
    FSR0 = AddressOf(Tgl)               ' indirect for Tgl(0)
    Repeat                              ' clear toggle array
        POSTINC0 = 0                    ' clear 0x400..0x7FF
    Until LastElement                   '
'
'  insert 3 output toggle bits into our 1024 interval toggle array
'
    Tgl(Pulse(0)) = Tgl(Pulse(0)) Or 1  ' insert RB0 toggle bit
    Tgl(Pulse(1)) = Tgl(Pulse(1)) Or 2  ' insert RB1 toggle bit
    Tgl(Pulse(2)) = Tgl(Pulse(2)) Or 4  ' insert RB2 toggle bit
    Tgl(0) = Tgl(0) Xor $FF             ' prep Tgl(0) element
'
'  implement a button press or some other trigger before falling into
'  the output routine (1024 1-usec intervals with 20-MHz clock)
'
    FSR0 = AddressOf(Tgl)               ' Tgl(0..1023), 0x400..0x7FF
    Repeat                              ' toggle outputs from toggle array
        LATB = LATB Xor POSTINC0        ' at 1-usec (5 cycle) intervals
    Until LastElement                   ' until end-of-array (0x800)