how to generate two signals from two ports at the same time by using p16f84

[bikxsici]

sory i am writing fast and i forget something. there is no deadline but i must submit as soon as possible.

i am going to review the program, and i will send the results , i must be careful. thanks a lot for your advices

best regards

[bikxsici]

hi cncmachineguy

i think there are two choices which depend on whether the codes ( decrement freq1 &freq2, IF !freq1&&freq2 then) are in interrupt loop or main loop.

for the first one, they are in the interrupt loop, these codes takes min 10 mic. sec. ( with adjusting option_reg, and loading tmr0 for min. time ) , therefore i can not obtain a period increment below 10 mic. seconds.

for the second one, i have tried to put those codes into main loop, but then i have problem to control tmr0 to obtain exact values.

after so much efforts i have decided to build this system is impossible by using pic16f84 with 10 mhz and i think we need at least two timer interrupts to obtain two different precise frequencies.

thank you a lots for advices.

[cncmachineguy]

I understand your frustration, but don't give up yet. Your teacher has it working so it is not impossible.

New question, you say you need 2uSec percision, but then you talk about using an 8uSec time base. this will not be able to provide 2uSec percision. So is the 8uSec ok?

lets look at some math for a minute. you have a 10Mhz clock, so thats 2.5MIPS. each instruction will take .4uSec, so for 8uSec time base, you have 20 inscrutions worth of time to decide what to do.

If you don't want to give up, lets keep working on this.

[bikxsici]

yes 8u sec. is also ok. let's look at it closely again ( by the way i hope i am a good student ) maybe this time we can do it. i think there is no problem about timing by using tmr0 and prescaler, it works well. the main problem is that we need a loop including two variables depending on tmr0 and to determine intended two periods time.
i think putting this loop into interrupt is better since if we have had to use a keyboard to select frequencies, we would use it in the main loop, so no waste time for our periods.
but on the other hand when we use below codes in the interrupt loop, it needs 22-25usec.. i am writing codes below how i calculate these times. i get 22-25 usec. by measuring pulse width of porta.1 via oscilloscope.

int_loop:
high porta.1 ; start of the pulse i measure

fre1=fre1+1 ; increment for period 1 depending on tmr0 interrupt time
fre2=fre2+1 ; increment for period 2 depending on tmr0 interrupt time

if fre1>=5 then ; 5 random number if i suceed it will be fraction of x8
toggle porta.0
fre1=0
else
endif

if fre2>=3 then ;3 random number if i suceed it will be fraction of x8
toggle porta.0 ; this must be porta.1 but now not important
fre2=0
else
endif
INTCON.2=0 ; setting up timer
TMR0 =254 ; reload tmr0

low porta.1 ; end of pulse

pauseus 100
goto int_loop


and this is the shotest loop i know for our aim, and it requires at least 20usec bigger than 8usec. (

thanks a lot

[cncmachineguy]

OK, Good start. First I think this will have to be your main and grabbing input will have to be your interrupt. I say this because even if you used pure ASM to get in the ISR, it will take at least 8 ASM instructions for the context save and restore. This leaves you almost no room to do anything. On the other hand, if the signalk generation is the main, with an interrupt call for the keypress, maybe it will be ok. you will get a glitch or 2, but only when you change freq.

No, lets see how to speed this up a bit.
First remove the else statements. you are not using them so you can get rid of them. I don't think this will speed it up, but lets see.

After you try that, change the counters to count down instead of up. Checking if freq = 0 should be much faster then if freq >= something. The reason for this is in the compiled code, it should compile to be something like this:

DECF freq    'decrenment freq
BTFSS freq  'test if freq is zero, if not skip the next step
JMP notzero
MOVLW 5
MOVF freq
notzero:
reutrn

In order to test for equal or greater then any number other then zero, first the number must be subtracted from the counter, then status.zero must be tested. If not zero, then borrow or carry must be tested and the program will branch accordingly. Does this make sense?

After you try these 2 things, report back with the results. There is at least 1 more thing we can do, but I want you to try each thing 1 at a time and understand why it worked or didn't work.

[cncmachineguy]

Forgot to mention, No need for a prescaler, 8uSec is WAY shorter then TMR0 just running. In fact, we may find the loop can be tuned to take EXACTLY 8uSec, so no need for the TMR at all!

But first we must get through timing tests as you are doing now.

Yes, you are a fine student, just don't give up.

1 more thing, does the language used matter? As a last resort we will try this with ASM. What type of class is this for?

[bikxsici]

yes you are right maybe we can do it without using TMR. for a language, normally we must use ASM but also can use PBP or C, but i prefer PBP since i have learned it before. it is a summer course, introduction to microprecessor, in my university and homework will not collected but i want to do it since firstly it seemed very easy but after struggling i think infact it is not easy and makes us more careful about timing.

now, i want to use PBP and we can put ASM codes and by using PEEK nad POKE we can change and get ASM variables in PBP, i hope there will not be a problem.

but before i want to write it via MPLAB to see exact timing and then put it in PBP. By the way the issue about if=0, you are right, i did not recognize it. i am going to try some codes and share results here.

best regards