Not quite got my finger on the pulse

[SteveB]

I sounds like you want something like:

 400-1500Hz | 10-150us |  2-50us  |  10-50us |  2-50us  |  10-50us | 400-1500Hz
            +----------+          +----------+          +----------+
 ___________|          |__________|          |__________|          |___
              1st Pin               2nd Pin               3rd Pin

Does this look correct?

With a PIC (I'll assume an 18Fxxxx) there are some things you will need to keep in mind. Each instruction cycle is Fosc/4. So at 40Mhz it will complete most instructions in .1us. Some instructions use 2 instruction cycles to comeplete (like goto's). So you should be able to get close to .25us resolution, at least once the pulses start. It also looks like a 10/2/10/2/10 pulse string would not allow any other things to be done between starting and completing the series of 3 pulses if you need high accuracy.

Have you tried something like this to get the pulses? It should be close to the .25us accuracy you are looking for.

Pulse_1	VAR BYTE
Pulse_2	VAR BYTE
Pulse_3	VAR BYTE
Delay_1	VAR BYTE
Delay_2	VAR BYTE
'---------------------------
' Set the variables as needed
Pulse_1	= 10
Pulse_2	= 2
Pulse_3	= 25
Delay_1	= 10
Delay_2	= 50
'---------------------------
' Run this series of instructions
' at the appropriate interval

     portb.1 = 1
     pauseus Pulse_1	
     portb.1 = 0

     pauseus Delay_1	

     portb.2 = 1
     pauseus Pulse_2
     portb.2 = 0

     pauseus Delay_2

     portb.3 = 1
     pauseus Pulse_2
     portb.3 = 0

Now, what are you using to time the 400-1500Hz initial pulse? Does this also need to have a .25us accuracy? If it does, you may have trouble, since even using ASM and Interrupts, there is a 3-4 instruction cycle latency for the interrupt to start running after the trigger (so .4us). And this does not count any context saving that may be needed. Some more info may assist in getting more help for you problem.

SteveB

[Sean_Goddard]

Hi Steve, Thanks for the reply, but..... I think I need to use the timer as I said, the PIC has to do other things too (such as scan a front panel full of pots and switches) and this pulse train needs to be "automatic" and invisible to the rest of the processes.

I did try something along the lines of what you suggested, but I was leaning towards setting up a timer then having it time out and reloaded to give another delay, or use one timer to generate the delay, then trigger timer 2 to provide the pulse width. When this timer times out, it would triger another to give the second delay. I've yet to sit down and figure out the fine bits, but I think I need to learn about interrupts and timers on the PIC. Any help would be appreciated.

Thanks.

[SteveB]

1) A timer could be used as a source for the 400Hz(2250us)-1500Hz(667us) timing. This would certainly be best with interrupts. The unavoidable latency involved could be dealt with by adjusting the value loaded into the timer by an offset.

As a primer, have a look at Darrel Taylor's Instant Interrupts. It's a long thread, but the key stuff is in the first 8 posts, and also on his web site. Also, he has a little calculator to help come up with the values to add to the timer registers. This sould help to get the 400-1500Hz timing sorted out.

2) If you want to get an accurate 2us pause (only 20 instruction at 40 Mhz) between pulses, you are not likely to do this with timers and interrupts, and still have an accuracy of .25us, and get anything else accomplished. It would take most of 20 instructions cylcles just to vector to the interrupt routine, save context, toggle a pin, restore context, and return to the code (the 18F does have some context saving features that help here, but not that much).

So, use the interrupts to trigger the start of the pulse string, then control the pulse width/timing actively using PBP or ASM routines. Worst case you would have a 350us interruption in you main code to send out the pulse string followed by about 317us for the main code to run before the next interrupt. Not ideal, but consider that 317us is 3170 instruction cycles, and a fair bit can be accomplised it that time. Also, as the freqency decreases, much more time becomes available between interrupts.

SteveB

[SteveB]

Something that should also be mentioned (but is probably obvious) towards the timing accuracy you seek. It will only be as accurate as the oscillator you use. Additionally, the most of the PIC with high frequencies achieve this from PLL. All this will affect that .25us goal your aiming at.