Is there room for improvement in this polling routine?

[jonnybgood]

Hi,

I wrote this code for measuring period of an input squarewave with a PIC24F. First I implemented an interrupt routine but then I realize that I would be better off by polling, as the operation would be executed only once for one cycle. I am measuring from rising edge to rising edge. My clock frequency would be 16Mhz with a crystal of 32Mhz (fosc/2). My input period would vary from 2µs to 200µs.

I have 32cycles available. With this polling, I am using 14 clock cycle and I what to know I can lower them more as I will be doing an ADC conversion (conversion only - no sampling - requires 12 clock cycles) and some 4 or 5 simple tasks like mov etc... my code is below:

.text
.global _asmFunction

_asmFunction:
mov.w #0x0, w4
mov.w w4, 0x2CA

clr.w w1			   ;clear WREG1
clr.w w3			   ;clear WREG3
clr.w TMR1		   ;clear TMR1

mov.w #0x8000, w1 ;move 0x8000 to w for t1con
                                   

start: BTSS PORTB, #9  
GOTO start   

mov.w w1, T1CON 	;timer 2 OFF   

falledge: BTSC 0x2CA, #9 	;test RA7 if not set skip next(low) 
GOTO falledge

stop: BTSS 0x2CA, #9 	;test RA7 if set skip(complete cycle) 
GOTO stop

mov.w w3, T1CON 	;timer 2 OFF

mov.w TMR1, w5       ;save 
mov.w w5, 0x800	;timer value 

return
 
.end

 

[bepobalote]

wouldn't using timers and Input Capture be a better option to measure the period of your square wave?

 

[jonnybgood]

That's what I just did yesterday, but with with the latency of ISR it cost me 16 clock cycles, now it costs me 14, not a lot of improvement but not bad. Can I use input capture with out going to ISR? Just polling?

 

[FvM]

Can I use input capture with out going to ISR? Just polling?

Yes, of course.

Utilizing the capture buffer, you can also start a measurement and read it later. No time-critical code is involved.

 

[jonnybgood]

So I can eliminate the ISR and the Latency that come with it?

Can you show me an example with just simple pseudocode please?

 

[FvM]

Referring to 15.6 CAPTURE BUFFER OPERATION in the PIC24F family refernce manual:
- reset the input capture module
- setup input capture and associated tímer to the intended mode
- enable input capture
- wait for two consecutive capture events by polling ICBNE and read their results

 

[jonnybgood]

thanks, therefore, for what I need to do, I can first wait for the first rising edge by polling the ICBNE and then wait for the second rising edge by polling the ICIF am I right? Because unless I read the first capture value in the IC1BUF, the ICBNE will not turn 0 and therefore I need to relay on another flag.

Regarding the mnemonic "BTSS.b", is it the most efficient mnemonic I can use, as it takes 3 machine cycles...