Pulse generation how to?

[Sean_Goddard]

Can someone please give me an english "How to" create pulses from a pic driven PI?

I have read the ITMD code and I think it is using interrupt to reload the timers but I'm not sure I understand how the whole thing does what it does.

My understanding is this;

Set TIMER1 to be the Tx pulse generator, then set the 16 bit timer (running as fast as possible) to be the pulse generator MASTER timer.

1) On trigger of the Tx pulse set the Tx output HIGH (or whatever) and trigger the 16 bit timer with the value to give the required pulse width of Tx pulse and point ISR to deal with the 16 bit timer.

2) When the 16 bit counter rolls over, shut off the Tx and load 16 Bit timer with the value for the delay of Tx(OFF) to Sample)_1.

3) When 16 bit counter rolls over set Sample_1 pin to HIGH and load the 16 bit timer with the value required to give the pulse width of Sample_1.

4) When 16 bit counter rolls over set Sample_1 pin to LOW and load the 16 bit counter with the value to give delay to Sample_2.

5) When 16 bit counter rolls over set Sample_2 pin to HIGH and load the 16 bit counter with the value to give the width of Sample_2.

6) When 16 bit counter rolls over set Sample_2 pin to LOW and load the 16 bit counter with the value to give delay to Sample_3.

7) When 16 bit counter rolls over set Sample_3 pin to HIGH and load the 16 bit counter with the value to give the width of Sample_3.

When 16 bit counter rolls over set Sample_2 pin to LOW and point the ISR to deal with interrupts from TIMER1 (looking for a roll over to signify "Next Tx pulse is due").

9) MEANWHILE, Do other stuff such as read pots, update display, change audio PWM, flash LED's etc.

Have I go the gist of this right or am I totally wrong about how this is done? I program PICS using Protobasic, but don't do much with timers etc a I don't need to but if I'm going to do PI stuff then I need to learn.

[eclipse]

Do you know this page http://www.geotech1.com/forums/conte...57-source-code
It is a goldmine. Lots of things to learn and its all in C.

Drawbacks - you cant have overlapping pulses. Precision is limited to at least few microseconds.

[waltr]

Check my HH2 thread that has the PIC code:
http://www.geotech1.com/forums/showt...ake-on-the-HH2
My code is in PIC assembler so READ the Comments for what it does. I have a few version of the code in my thread in ZIP files.

You have the basics for using TMR1 to generate timing. It is a sequence of TMR1 values, Polling the TMR1 Interrupt flag to know whne it rolls over.

I do the entire TX and Sampling timing within the ISR.
The ISR runs on TMR0 at about 665us intervals.

All slow operations such as reading ADCs (adjustment pots) etc happen in the MAIN loop but only every 20 Timing cycles or 13ms which is Real Time to human. I also only do one ADC or other operation per cycle since there are 20 cycles to get this stuf done.
The reason is any of this slow stuff MUST be finished before TMR0 interrupts to start the next Timing sequence.

This is a 'real time OS' technique for microprocessors explained in the Microchip App Note AN585.

[Carl-NC]

The method used in ITMD PI-2 (for example) is the bit-bang out the timing in an interrupt-driven loop. It works like this:

Timer0 is the main loop timer, set to interrupt at ~600Hz.
Timer1 is used to time the remaining delays.

Interrupt ISR
{
  turn on the TX pulse
  wait 75us
  turn off the TX pulse
  
  wait 15us (user variable delay via pot)

  turn on the Sample pulse
  wait 10us
  turn off the Sample pulse

  wait 100us

  turn on the EF pulse
  wait 10us
  turn off the EF pulse

  Read the delay pot
  Do other stuff
}

This approach wastes a lot of CPU time with delays, but is an acceptable way to do timing if (a) you are running an unaggressive pulse rate and/or (b) you are just creating simple timing and not much else. As you try to do more you will quickly run out of time for the "Do other stuff" part. You can stretch this a bit by doing round-robin processing, whereby you split the "Do other stuff" into parts A, B, and C and call them sequentially through each pass of the ISR. This increases the 'do more stuff', just not as often.

Overall, I consider this a poor way to do things but I wanted to use a small cheap PIC and resources are limited. As you move in to 16-bit PIC (dsPIC and PIC24) you will find more timers with more capability and you can program them as PWMs and even gate them off each other, so as to produce all the required timing autonomously, with absolutely no load on the main program. Other processors (like Cortex) have even more powerful timing capability, plus aren't power hogs like the PICs. I have personally switched to STM's family and run all code under an RTOS, something especially worthwhile as you try to incorporate more 'do other stuff' like wireless, graphics, live controls, and so forth.

[waltr]

Yep, that is how I do the timing on my HH2. It does work well using a simple processor and keeps the timing accurate cycle to cycle.

[eclipse]

I have made a development board based on PIC32MZ specially designed for PI metal detector test environment - it has got lots of pins for output signals, 4 leds, couple of binary switches,
couple of 3 way switches and lots of onboard trimpots.

So far I've managed to get pulse timing scheme (precision 5uS), ADC In + PWM Out, written on C with MPLABX and Harmony FW. Took 1 year to get to this point

Right now I'm working on adding Delay & Freq control pot.

I've got few boards of these (blank ones), I can donate them if anyone interested.

[Sean_Goddard]

THANKS ALL!!

Carl, I thought of the bitbang method. forming the pulse train in memory then simply clocking the data out serially. I did this for a MF IB, used the DSP to generate the IFT in a 8K DP RAM then clocked the data out. It's crude, but again, IT WORKS and it frees up the DSP for other stuff.

BTW, does ANYONE here have experience of RTOS systems because if you do I have a suggestion which you might like to have a go at which is metal detector related. Hush, hush until I can find someone who is a bit of a wizard with that sort of thing.

[Carl-NC]

BTW, does ANYONE here have experience of RTOS systems because if you do I have a suggestion which you might like to have a go at which is metal detector related. Hush, hush until I can find someone who is a bit of a wizard with that sort of thing.

As I mentioned above, my current efforts are using an RTOS. It's all work stuff so I can't share much of that.

[waltr]

I have made a development board based on PIC32MZ specially designed for PI metal detector test environment - it has got lots of pins for output signals, 4 leds, couple of binary switches,
couple of 3 way switches and lots of onboard trimpots.

So far I've managed to get pulse timing scheme (precision 5uS), ADC In + PWM Out, written on C with MPLABX and Harmony FW. Took 1 year to get to this point

Right now I'm working on adding Delay & Freq control pot.

I've got few boards of these (blank ones), I can donate them if anyone interested.

Nice PCB. I recently started a thread on PIC32 pulse generation here:
https://www.geotech1.com/forums/show...ng-in-hardware
PIC32's have hardware that the mid-level PIC's do not that makes creating pulse easy and very accurate.