How can I count how long it takes for a pin to change states?

P!-Ro

I tried getting it to count the time and display the value, but it just freezes when it hits the waitfreq command.
code:
repeat··
····· lcd.cls
······lcd.str(string("wait...", 13))
····· RCTemp := cnt
······waitpeq(%100000, 8, 0)
····· waitpeq(%000000, 8, 0)
····· RCTemp := cnt - RCTemp

···· lcd.cls
···· lcd.dec(RCTemp)
···· waitcnt(clkfreq/1000 *30 + cnt)

However, if I use this in it's place, it will tell me the value with no problem:
· repeat
·· waitcnt(clkfreq/1000 *30 + cnt)
·· lcd.cls
··· if ina[noparse][[/noparse]8]
···· lcd.str(string("1"))
··· else
···· lcd.str(string("0"))

What am I doing wrong?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

mynet43

Here's some code that's used to count the time it takes a capacitor to discharge. It's used to sense the position of a pot. I think you can modify some of it to make your first method work.

' monitor volume pot and adj vol if needed
  vpin := 26                  ' set volume pot to pin 26
  vmsk := 1<<vpin

  outa[noparse][[/noparse]vpin]~~                ' initialize pin state to high
  dira[noparse][[/noparse]vpin]~~                ' set pin to output to charge capacitor
  waitcnt(clkfreq/4 + cnt)    ' wait for cap to charge to 3.3V

  c1 := cnt                   ' count when charged
  dira[noparse][[/noparse]vpin]~                 ' toggle pin to input to start cap discharge
  waitpne(vmsk,vmsk,0)        ' wait for pin vpin to go low
  c2 := cnt                   ' store count when discharged to 1.65V

  if c2 > c1                  ' skip display if time wraps

Hope this helps...

Ariba

WAITPEQ needs a pinmask as second parameter, not the pin number.
change your code to this:
...
waitpeq(%1_0000_0000, %1_0000_0000, 0) 'wait until Pin 8 high
waitpne(%1_0000_0000, %1_0000_0000, 0) 'wait until Pin 8 low
...

Andy

P!-Ro

Thanks, guys, I tried it and it works. The only problem is it doesn't seem to be very accurate. Is there a way to fix that?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

Cluso99

you are saving the counter before the pin toggles so move your statement
RCTemp := cnt
down 1 line

P!-Ro

Heres the updated code:

····· waitpeq(%1_0000_0000, %1_0000_0000, 0) 'wait until Pin 8 high
····· waitpne(%1_0000_0000, %1_0000_0000, 0) 'wait until Pin 8 low
····· RCTemp := cnt
····· waitpeq(%1_0000_0000, %1_0000_0000, 0) 'wait until Pin 8 high
····· RCTemp := cnt - RCTemp

It is meant to tell me what the time is between pulses of a spark plug, so that is why I wait for it to go high first, to improve accuracy. What I think part of the problem is, though, is when cnt restarts to zero, giving wrong results. Have any of you found a way to fix this?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

Cluso99

cnt is an endless 32 bit counter which just overflows back to 0.
What your code is doing is just timing between the pin going low and back to high again - not the whole cycle.

Paul Baker

Pi, there's a shorthand for making a pin mask, |<, the first line can be rewritten as
waitpeq(|<8,|<8,0)

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

P!-Ro

doesn't explain how my numbers can suddenly jump up by 10,000 units and back down almost emmediately after. It also does it in the opposite direction making it very low and back up again. It doesn't do it very often, but I don't really like it doing it period.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

Ariba

What is the frequency of the pulses? Perhaps Spin is to slow and loose out one pulse from time to time.

You can try it with a counter to detect the pulses. This code uses the POSEDGE DETECTOR mode. Every time a pulse occures, the PHSA register will be incremented:

   ctra := %01010 << 26 + 8   'Init PosEdge Detector on Pin 8
   frqa := 1
   ...
   repeat
      phsa~
      repeat until phsa    'wait until first pulse
      RCTemp := cnt
      phsa~
      repeat until phsa    'wait until second pulse
      RCTemp := cnt - RCTemp
      ...

Andy

mynet43

The problem may be that you didn't account for the CNT wrapping when it fills up the 32 bit register.

If CNT < RCTemp then skip this loop and reset RCTemp.

P!-Ro

I·have an attachment of my code using mynet's and Pauls suggestions. It still doesn't seem to work quite the way I want it to, though, skipping from 24000 to 1300 every once in a while. I'll try Andy's method later and see if it works.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

mynet43

You didn't make it clear what's triggering pin 8, whether it's a constant frequency and what that frequency is...

It would be easier to help if we knew exactly what you're trying to accomplish. The latest code looks like you're trying to display the half-cycle time of pin 8, every 500 ms.

Need more info...

mynet43

Noticed one more thing.

You have:

      if T1 > T2
         counter

"counter" is the name of the routine you're in. Generally, it's not a good idea to call yourself unless you really know what you're doing.

It would be better to reverse the logic:

     if T2 > T1
       time := T2 - T1

       lcd.cls                                             
       lcd.dec(time)

     waitcnt(clkfreq/1000 *500 + cnt)

P!-Ro

mynet43 said...
You didn't make it clear what's triggering pin 8, whether it's a constant frequency and what that frequency is...

It would be easier to help if we knew exactly what you're trying to accomplish. The latest code looks like you're trying to display the half-cycle time of pin 8, every 500 ms.

Need more info...

The only reason I made it every 500ms is so I could read the results on the lcd. I was hoping to eventually use the time to find out the rpm, but right now I'm wondering if counting the pulses over a set period of time to find the RPM would be better. As for what is triggering the pulse is a wire wrapped around the sparkplug and attached to an npn transistor (on the breadboard). The wire doing this is the green one in the picture, attached to a breadboard with an alligator clip. Every time there is a pulse, the blue led on the breadboard flashes and pin 8 is connected to base.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

Cluso99

Please describe your circuit to pin 8 better. Am I correct in assuming you have connected pin 8 to the base of the transistor? While not a specialist in this area, I would think that the input probe (your wire to the sparkplug) should be connected to the base.

Further, what protection do you have for the prop input pin (and the transistor for that matter)? I am concerned at what voltage the spike is. You need some form of voltage limiter here. Maybe a resistor (high value) from the input probe to the base of a transistor, and the emitter (presume npn) to ground on the pcb. Place a capacitor and a zener across the base - emitter. I suggest a 0.1uF and 5v zener (or 3v3). You may not even require the transistor at all.

One of the forum members may care to comment on this circuit as it is not my area. Beau - you there???

P!-Ro

I described the schematic better in a different post on a robot I am building, but I posted this question here so I could get better answers on what to do for the programming. You can read this post here. http://forums.parallax.com/showthread.php?p=756088

I actually stole this picture from parsko's post, but I have been using it to show basically how it is set up.

Besides what is on this picture, I have a 10k ohm resistor connected to 3.3v going to the pin allong with the output of the transistor.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

Post Edited (I LIKE PI) : 11/10/2008 4:43:28 AM GMT

Cluso99

OK, the circuit looks fine to my way of thinking.

So, presume your engine can do a max 6,000 rpm, that makes the pulses coming at max 100 rps = 10mS. Therefore, with an 80MHz xtal you should be counting a cycle maximum of 80,000,000 / 100 =800,000. If you have 1000 rpm then 800,000 / 6 = 120,000 clocks approx. This should help you work out what you expect to see.

Now for your code: As mentioned above, you are only counting half the cyce and probably the spike width is varying quite· deal (not very accurate). So try this..

waitpeq(|<8,|<8,0) 'wait until Pin 8 high
waitpne(|<8,|<8,0) 'wait until Pin 8 low 

 
repeat   
      T1 := cnt
      waitpne(|<8,|<8,0) 'wait until Pin 8 low 
      waitpeq(|<8,|<8,0) 'wait until Pin 8 high 

      T2 := cnt
'I think the following will work
      time := t2 - t1
'but if it does not work when t1 > t2 (wraps around) then use this
'      if T1 > T2
'          time := t1 - t2
'      else
'          time := T2 - T1           
           
'     lcd.cls                 '\ <== this may be causing you to miss the next cycle                           
'     lcd.dec(time)           '/ you may have to use another cog to display the data?? but try this
 
      lcd.dec(time)
      lcd.out($0D)            ' or whatever the display char routine is called for carriage return/line feed
 

 
'remove the following line
'     waitcnt(clkfreq/1000 *500 + cnt)

P!-Ro

Sorry I haven't posted here in a while but I've been too busy to mess with code lately. Since I've had a break for·Thanksgiving·I've been reading through the Propeller Manual to understand a little more about the propeller. I'm still confused though, and I'll tell you about that later. First, though, I'll tell you what I now would like·to do with the code.

Since there are obviously many problems with finding the exact time·between sparks, I would instead like to count them·over a certain amount of time.·I would like to do this by running two cogs, one of them counting the sparks, and a different one timing it then finding the number of sparks fired after the time is up. The only problem is I can't figure out how to do this over the cogs. I know to share variables you must place it in the pub line like this: :variable but other than that I really do not know what else to do. If you could clarify this and maybe give me suggestions for doing·this with the code I would appreciate it.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Pi Guy

StefanL38

Hello Pi,

PUB NameOfPubMethod : variablename

is a construction if you want a method to give back a value

As long as variables are in ONE objectfile (your mainfile or any other objectfile
the variables that were defined in the VAR-section are global even across COGs

below there is a short democode to demonstrate this

MyVar is defined in the VAR-section
then TWO other cogs where started with different PUB-methods of THE SAME *.SPIN-file
and they both can change the value of MyVar


If you want to use variables across OBJECTS (instead of cogs) you have to hand over
the RAM-adress as a pointer to the other-objects

you can do this by using the "@"-operator

VAR
long MyVar

PUB give_Other_Object_the_pointer (pointer)

call of this PUB

give_Other_Object_the_pointer (@MyVar)

access of this var

long[noparse][[/noparse]pointer] := ....

democode to show access of global variables across cogs

CON
  _clkmode = xtal1 + pll16x
  _xinfreq = 5_000_000

  
VAR
  long CogStack1[noparse][[/noparse] 20]
  long CogStack2[noparse][[/noparse] 20]
  
  long MyTestVar

OBJ
  debug : "FullDuplexSerial"


PUB Main
'the FIRST PUB-Method inside a spinfile is ALWAYS the startpoint where the program starts to run
  debug.start(31, 30, 0, 9600)
  
  MyTestVar := 100
  debug.str(string("Start MyTestVar="))
  debug.dec(MyTestVar)
  debug.Tx(13)

  cognew(M1,@CogStack1)
  cognew(M2,@CogStack2)
  
  repeat
    waitcnt(clkfreq + cnt)
    debug.str(string("MyTestVar="))
    debug.dec(MyTestVar)
    debug.Tx(13)


PUB M1
  repeat
    waitcnt(ClkFreq * 3 + cnt) 'wait 3 seconds then set MyVar to value 1
    MyTestVar := 1


PUB M2
  repeat
    waitcnt(ClkFreq * 5 + cnt) 'wait 5 seconds then set MyVar to value 2
    MyTestVar := 2
      

from this democode you should get an output to a serial-terminalsoftware like PST.EXE
that looks like this

MyTestVar=100
MyTestVar=100
MyTestVar=1
MyTestVar=1
MyTestVar=2
MyTestVar=1
MyTestVar=1
MyTestVar=1
MyTestVar=1
MyTestVar=2
MyTestVar=2
MyTestVar=1
MyTestVar=1
MyTestVar=1
MyTestVar=2
MyTestVar=2
MyTestVar=2

by the way: the debugoutput included can make a lot of things clear whats going on in a program
it uses the same pins as your USB-cable to download the program to the prop. This means it needs
no additional hardware no LCD, no tv, no vga-monitor, no LED

best regards

Stefan

P!-Ro

Thanks, Stefan. I understand it much better now. I guess I'll work on the code for it tomorrow, then.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
PG

P!-Ro

I now have accurate code to detect the rpm of my engine. The attachment is below. What I changed in it is I now have a cog doing the counting, and a cog displaying the results every second. Also, I've added an extra pause in it to prevent errors in the counting process. This has made it much more accurate. At 10,000 rpm, I have 6ms to work with between pulses, so my 2ms pause does nothing to lower the results. The current code I have has an accuracy of +-120rpm (1 spark), and I hope to make it even more accurate in the future.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
PG