Delay question

[kata]

I don't find any specific answer to this question : Is it possible to have delay (for a specific protocol maybe) less than 1mS (say 335 uS) or I have to use ASM code ? Thank you for your time !

[malc-c]

Try the PAUSEUS command.

Here are the details as listed in the manual - http://www.melabs.com/resources/pbpmanual/ - (which should of been the first place to look)



PAUSEUS

PAUSEUS Period

Pause the program for Period microseconds. Period is 16-bits, so delays can be up to 65,535 microseconds. Unlike the other delay functions (NAP and SLEEP), PAUSEUS doesn't put the microcontroller into low power mode. Thus, PAUSEUS consumes more power but is also much more accurate. It has the same accuracy as the system clock.

Because PAUSEUS takes a minimum number of cycles to operate, depending on the frequency of the oscillator, delays of less than a minimum number of microseconds are not possible using PAUSEUS. To obtain shorter delays, use an assembly language routine in an ASM..ENDASM construct.

OSC Minimum delay
3 (3.58) 20us
4 24us
8 12us
10 8us
12 7us
16 5us
20 3us
25* 2us
32* 2us
33* 2us
40** 2us

* PIC17Cxxx and PIC18Cxxx only.
** PIC18Cxxx only.

PAUSEUS assumes an oscillator frequency of 4MHz. If an oscillator other that 4MHz is used, PBP must be told using a DEFINE OSC command. See the section on speed for more information.

PAUSEUS 1000 ' Delay for 1 millisecond

[skimask]

Try the PAUSEUS command.

Here are the details as listed in the manual - http://www.melabs.com/resources/pbpmanual/ - (which should of been the first place to look)


That's amazing isn't it? You find the wierdest information in the oddest places. :) Especially if the information (the PBP book) came with the program package (bought directly from MeLabs).
JDG

[mister_e]

Come on Skimask ;)

[therian]

Can someone please explain me how to use assembly to make shorter delay than pauseus can make ?

[HenrikOlsson]

Hi,
The shortest possible delay is achieved by using the assembler command NOP (short for No OPeration):


@NOP 'Delay for OSC/4

The NOP instruction takes 1uS at 4Mhz and 250nS at 20Mhz etc. If you want longer delay add more NOP instructions or build a loop, but then it's probably easier to use PauseUs anyway.

/Henrik Olsson.

[therian]

thank you, as i understand 1us is minimum limit for 4mhz crystal

[HenrikOlsson]

Therian,
Yes, that is correct. With a 4Mhz X-tal one intruction cylce is 1uS and since the NOP command takes one cycle to execute the delay will be 1uS.

/Henrik Olsson.

[therian]

just curious, is 250nS enough to generate radio frequency directly from pic without any rf modules, it sound like cheap radio link

[HenrikOlsson]

Therian,
Well, let's see:


Start:
PortB.1 = 1 'Set pin high
@ NOP 'Wait....Two NOP's are neede if we want 50% dutycycle
@ NOP 'since the GOTO takes two instruction cylcles.
PortB.1 = 0 'Set pin low
Goto Start 'Do it again. The GOTO takes the same time as two NOP's

This, I think, will generate a pulsetrain with 50% duty cycle. The frequency would be ~833kHz. Not much of a radio frequency.... besides that, you can't do anything else in the code without slowing it down more...much more.

/Henrik Olsson.

[therian]

maybe overclocking might work ?

[therian]

wait 833kHz is in range of medium AM broadcast 300-3000kHz and 30-300kHz is longwave AM broadcast but as i know 3-30kHz is enough for submarine communication. so what wrong with using low frequency, huge antenna ?

[HenrikOlsson]

Therian,
First all, I've stated 250nS in my previos posts. That is of course wrong and should be 200nS. NOT 250.

With that being said you could go to a PIC capable of 40MHz, then a single cycle instruction (like NOP) is executed in 100nS instead of 200.

But then again if you use software to generate the carrier frequency you'd be hard pressed to have any instruction cycles left to be able modulate it. I really believe you should look at a dedicated TX/RX chipset or module.

Anyway sorry for the confusion. One instruction cylcle is 200nS at 20Mhz, nothing else.

/Henrik Olsson.

[therian]

im really thankful for you reply.
I know about cheap and good modules out there, but dont anyone try to make some from scratch, it easy to fing FM bugs circuits in google but searching for rf module give really few results

[paul borgmeier]

Not to muddy the waters because it sounds like you have decided on another route, but for completeness ...


Start:
PortB.1 = 1 'Set pin high
@ NOP 'Wait....Two NOP's are neede if we want 50% dutycycle
@ NOP 'since the GOTO takes two instruction cylcles.
PortB.1 = 0 'Set pin low
Goto Start 'Do it again. The GOTO takes the same time as two NOP's

at 4 MHz, 3 instructions on and 3 instruction off = 6 instructions per cycle, therefore f=1000000/6=166666.667 Hz (twice that noted above)

Interestingly enough the above code does 3 instructions on and 4 instructions off on 16F devices and 3 on 3 off on 18F (at least on an 18F877A and 18F452, respectively). To combat this on the 16F's, you could do something like this in ASM


ASM
here
bsf PORTB,1
NOP
NOP
bcf PORTB, 1
goto here
ENDASM

Good Luck

[HenrikOlsson]

Hi Paul,
Thanks for filling in and for correcting me. My head seems to be completely messed up today - can't seem to get one thing right.....

How come it differs between 16F and 18F parts? I haven't looked at the generated code but would've thought that the compiler generated just what you wrote in the 'pure asm' example.

Anyway, I'm off to bed now. Not doing much good anyway.

Thanks
/Henrik Olsson.

[paul borgmeier]

How come it differs between 16F and 18F parts?


The 16F code inserts a "clrf PCLATH" command before the goto - the 18F uses the bra command (instead of goto) and does not - why PBP inserts the clrf PCLATH, I would guess because "Start:" is in page 0 where PCLATH = 0 (EDIT: and does not check to see if we are already in page 0, which we are)

[HenrikOlsson]

Hi Paul,
New day, fresh head...
You said:


at 4 MHz, 3 instructions on and 3 instruction off = 6 instructions per cycle, therefore f=1000000/6=166666.667 Hz (twice that noted above)

That is of course correct. I based my calculation on a 20Mhz so 5000000/6 = 833kHz which is what I stated in my post which is also correct (but not half of 166kHz).

Thanks for the 16F vs 18F explanation. I guess it would take up more code to check which bank we're in before each goto instead of just setting the correct one?

Thanks!
/Henrik Olsson.

[paul borgmeier]

Henrik,

My apologies - you are and were indeed correct (166.667 / 2 != 833.33).

I am guessing the 16F PBP compiler code is not optimized to check to see if it is in the same page as the goto. I bet it sets the PCLATH every time it encounters a goto just to be safe. I do not think it would add PIC code - I think it would take more compiler code to do this check - MELABS must not have thought it was not worth it to potentially save one instruction (100% guess by me)?

[mister_e]

I think it will use the CHK?RP macro simply for safety sake. sure the code is not optimized in any sort, but it's always solid as rock. When you need something tight, use ASM. Hence, for safety sake and for none 18Xxxxx your asm example could be...


ASM
CHK?RP ; make sure we are on the right BANK
here
bsf PORTB,1
NOP
NOP
bcf PORTB, 1
goto here
ENDASM

PORTB.1=0 call MOVE?CT macro. wich is


MOVE?CT macro Cin, Regout, Bitout
CHK?RP Regout
if (((Cin) & 1) == 1)
bsf Regout, Bitout
else
bcf Regout, Bitout
endif
endm

there you discover the extra cycle

[therian]

separate coil-capacitor Oscillator and use pic only to turn on and of this oscillation we will get Pulse Modulation

[therian]

it sound too simple to be true