Soldering station

[HenrikOlsson]

Hi,
A fullblown PID is probably overkill but, for the record, incPID_mc is the multichannel version of the PID-filter. It works fine with a single channel but it takes a lot more codespace than the original incPID routine.

You can find v1.5 of the single channel version in this post. And with THAT said it still won't fit the 12F675 due to the limited amount of RAM available in that device.

What I'd do is probably something in the line of what Jerson showed but I'd also add a bias or feedforward based on the setpoint (not the error). If you know that in order to maintain a temperature of say 250°C you need a dutycycle of 40% then, once you're within the window of the regulator you enable the P(I) regulator and add the bias/feedforward to the output.

You'll need to work out a suitable bias/feedforward "gain" since it's not going to be perfectly linear. Since the system (probably) heats up faster than it cools down and you can't have "negative drive" you should err on the low side for the bias/feedforward so that IT alone can never drive the temp OVER the setpoint. If it does the regulator can't bring it down since it can never go "below" 0% PWM.

/Henrik.

EDIT: The original version of incPID, which can be found i post 1 in the thread above, compiles to 314 words on the 12F675 and uses 29 bytes of RAM. Depending on how many bytes of RAM your application uses it MIGHT fit but again, it's probably not worth it.

[Archangel]

Depending on how many bytes of RAM your application uses it MIGHT fit but again, it's probably not worth it.

Hello Henrik,
Probably not in Sweden or the USA, but in Romania, it might be very well worth it. Import tax is high there, wages not so much . . . Thanks for helping him.
JS

[fratello]

Thanks to all for support !!! Glad to see people helping others...
I will take attention to inc_PID, but ... now my code is 1002 words . The best (so far) solution it's to "upgrade" the PIC (maybe 18F...) but, for moment, I will try to keep this "low-cost variant".

[Ioannis]

Try to compact your code. Maybe repeated code can be a subroutine etc.

Also if you use flags, then declare them as BIT and not as BYTE or even worse as WORD.

If you need a BYTE variable, do not declare as WORD.

Ioannis

[fratello]

Whitout "graphics elements" the code it's 659 words. Maybe it's a good start for use PID ?! (+314 words)

' version 2
@ DEVICE PIC12F675, intrc_osc_noclkout, wdt_off, mclr_off
DEFINE OSC  4
DEFINE ADC_BITS 10
DEFINE ADC_CLOCK 3
DEFINE ADC_SAMPLEUS 50

CMCON    = 7
TRISIO   = %001001
INTCON   = 0 
GPIO     = 0

ANSEL    = %00010001
ADCON0   = %10000001
wpu      = %00000010
 
            D_C      var     GPIO.2       
            SDIN     var     GPIO.4
            SCK      var     GPIO.5
            MOSFET   var     GPIO.1
          buton    var     GPIO.3

            LcdReg          var  byte        
            x               var  byte
            y               var  byte
            z               var  byte
            a               var  byte
            PosX            var  byte
            PosY            var  byte
            Chr             var  byte
            LcdData         var  byte
            Offset          var  byte
            Char            var  byte
            adval           var  word
            w1              var  word
          vt              var  word
          cnt             var  byte
            grup            var  word
            mode               var  byte
          calibra         var  word
          
             
            DATA @0,$3E,$51,$49,$45,$3E,_       ';// 0                   
                    $00,$42,$7F,$40,$00,_       ';// 1
                    $42,$61,$51,$49,$46,_       ';// 2
                    $21,$41,$45,$4B,$31,_       ';// 3
                    $18,$14,$12,$7F,$10,_       ';// 4
                    $27,$45,$45,$45,$39,_       ';// 5
                    $3C,$4A,$49,$49,$30,_       ';// 6
                    $01,$71,$09,$05,$03,_       ';// 7
                    $36,$49,$49,$49,$36,_       ';// 8
                    $06,$49,$49,$29,$1E,_       ';// 9
                    $08,$08,$3E,$08,$08,_       ';// +
                    $08,$08,$08,$08,$08,_       ';// -
                    $00,$60,$60,$00,$00,_       ';// .
                    $00,$02,$05,$02,$00,_       ' grad
                    $3E,$41,$41,$41,$22,_       ' C
                    $7F,$02,$0C,$02,$7F,_       ' M     
                    $00,$00,$00,$00,$00,_       ' space
                 $08,$04,$7F,$04,$08,_       ' arrow up
                    $20,$40,$7F,$40,$20         ' arrow down
        
                          
           pause 500                 
           LcdReg  =  %00100001         'LCD Extended Commands.
           call    PrintCtrlLcd
           LcdReg  =  $E6               'Set LCD Vop (Contrast).Initial $C5
           call    PrintCtrlLcd
           LcdReg  =  %00000110         'Set Temp coefficent.
           call    PrintCtrlLcd
           LcdReg  =  %00010011         'LCD bias mode 1:48.
           call    PrintCtrlLcd
           LcdReg  =  %00100000         'LCD Standard Commands
           call    PrintCtrlLcd
           LcdReg  =  $0C               'LCD in normal mode.  $0D inverse
           call    PrintCtrlLcd                 
Pause 100
Call CursorHome
Pause 100               
Gosub LCD_Clear
mode = 0
Pause 500                ' time to give-up :) 
'===================== READING VOLTAGE ==============================
Main:
low mosfet
 vt=0
 grup=0
     for cnt = 1 to 20
 ADCON0.1 = 1
 WHILE ADCON0.1 = 1 : WEND
 Pause 10
 adval.HighByte = ADRESH
 adval.LowByte = ADRESL
 PAUSE 10
     vt=adval * 500
     vt=div32 1023
     grup=vt+grup
 PAUSE 10
     next cnt
 w1=grup / 20

'~~~~~~~~~~~~~~~~~ setting cursor for display the voltage
                LcdReg  =  %10000000 + 25        ' cursor  X
                call    PrintCtrlLcd
                LcdReg  =  %01000011             ' cursor  Y
                call    PrintCtrlLcd
'~~~~~~~~~~~~~~~~~ now display results 
 if (w1 dig 2) <1 then
                char = 16  
                call Printchar
                else         
                Char = (w1 dig 2)            
                call    PrintChar 
 endif      


                Char = (w1 dig 1)           
                call    PrintChar

 if (w1 dig 0) < 5 then            
                Char = 0          
                call    PrintChar  
        else
                char = 5
                call PrintChar
 endif                         
                Char = 16                 
                call    PrintChar
            
                Char = 13
                call  PrintChar
               
                Char = 14
                call  PrintChar

                
if buton = 0 then 
mode = mode + 1
gosub memo
endif


gosub check
Pause 100                
Goto Main

'================== L C D ===========================================                

CursorHome:     LcdReg  =  %10000000            ' cursor Home
                call    PrintCtrlLcd
                LcdReg  =  %01000000            ' cursor Home
                call    PrintCtrlLcd
                return

PrintChar:      offset = Char * 5          
            
                for a = 1 to 5                
                    read offset, LcdReg
                    call  PrintDataLcd
                    offset = offset + 1
                next a
                LcdReg = 0
                call    PrintDataLcd
                return
                                    
PrintCtrlLcd:   D_C = 0                             

PrintDataLcd:   for x = 1 to 8                
                    SDIN = LcdReg.7              
                    SCK = 1
                    LcdReg = LcdReg << 1
                    SCK = 0
                next x               
                D_C = 1
                return               
                
LCD_Clear:
                PosX=0:PosY=0:Gosub LCD_GotoXY
                D_C = 1
                FOR Chr=1 TO 252
                    LcdData=0:Gosub LCD_ByteOut
                    LcdData=0:Gosub LCD_ByteOut
                next Chr
                return

LCD_GotoXY:
                D_C = 0
                LcdData=%01000000 | PosY :Gosub LCD_ByteOut
                LcdData=%10000000 | PosX :Gosub LCD_ByteOut
                return

LCD_ByteOut:
                SHIFTOUT SDIN,SCK,1,[LcdData]
                return  

'===============================================================================
memo:

if mode = 1 then
 gosub poz
 char = 17
 call    PrintChar
endif

if mode = 2 then  
 gosub poz
 char = 15
 call    PrintChar
 calibra = vt
endif

if mode = 3 then 
 gosub poz
 char = 18
 call printchar
endif

if mode > 3 then 
  mode = 0
 gosub poz
 char = 16
 call printchar
endif

return
'===============================================================================
check:
SELECT CASE  MODE
CASE 1
PWM MOSFET, 255, 300
CASE 2
    if w1 < (calibra-1) then 
    pwm mosfet, 155, 200       'era 500
    else
    low mosfet 
    endif
CASE 3
LOW MOSFET
END SELECT    
return
'===============================================================================
Poz:
                LcdReg  =  %10000000 + 67        ' cursor  X
                call    PrintCtrlLcd
                LcdReg  =  %01000011             ' cursor  Y
                call    PrintCtrlLcd
return
'===============================================================================               

End             'of program

[mister_e]

You can still shave some bytes...

- you don't need 2 spi routines LCD_ByteOut & PrintDataLcd,
- replace the For-To-Next loop with Repeat-Until ones,
- if at all possible, try to skip the div32.
- use internal timer for delays (not sure it will that much though
- pwm with a high/low loop maybe?
- select case for a couple a if-then

couples of ideas like that. sounds like a nice challenge

[mister_e]

Another thing about the graphic part, I've messed with it a lot recently and I saved a load of code space by skipping the whole SELCT-CASE + Lookup thing by using something like

SearchChar:
        read (index), char
        if char == TheCharYoureLookingFor then
            Gosub PrintIt
            index=endtable
            endif
        index=index+6
        if (index)<endtable then SearchChar

But you need to modify a little bit the way your DATA lines are done.

            DATA @0,"0",$3E,$51,$49,$45,$3E,_       ';// 0                   
                    "1",$00,$42,$7F,$40,$00,_       ';// 1
                    "2",$42,$61,$51,$49,$46,_       ';// 2
                    "3",$21,$41,$45,$4B,$31,_       ';// 3
                    "4",$18,$14,$12,$7F,$10,_       ';// 4
                    "5",$27,$45,$45,$45,$39,_       ';// 5
                    "6",$3C,$4A,$49,$49,$30,_       ';// 6
                    "7",$01,$71,$09,$05,$03,_       ';// 7
                    "8",$36,$49,$49,$49,$36,_       ';// 8
                    "9",$06,$49,$49,$29,$1E,_       ';// 9
                    "+",$08,$08,$3E,$08,$08,_       ';// +
                    "-",$08,$08,$08,$08,$08,_       ';// -
                    ".",$00,$60,$60,$00,$00,_       ';// .
                    "G",$00,$02,$05,$02,$00,_       ' grad
                    "C",$3E,$41,$41,$41,$22,_       ' C
                    "M",$7F,$02,$0C,$02,$7F,_       ' M     
                    " ",$00,$00,$00,$00,$00,_       ' space
                    "^",$08,$04,$7F,$04,$08,_       ' arrow up
                    "_",$20,$40,$7F,$40,$20         ' arrow down

so when you want to display your grad character, just need to send G to that routine above. Simple... if it fits the EEPROM ...

[HenrikOlsson]

Hi,
314 words is the size when compiled by itself. Depending on what functions your code uses it might already have one or more of the PBP library routines included meaning the total size will be less than whatever + 314 words.

But like I said, the problem in this case is likely the amount of available RAM. If my count is correct your code uses 23 bytes and the original incPID uses 29 bytes, then, if I'm not mistaken, PBP adds ~24 bytes for its internal use (might change depending on what library routines are used, I don't know for sure). All in all this adds up to more than the 64 bytes of RAM available on the 12F675.

What I meant with not being worth it wasn't that you should throw a "bigger" PIC at the issue. I meant that I don't think the effort of implementing incPID is worth it considering the performance gain you might get compared to the aproaches already discussed. Sure, it it fits it's wort a try, if it doesn't fit I'd try to make do with what's available.

Finally incPID was written in order to be versatile and easy to use. It's probably possible to reduce the size and amount of RAM needed further if adapting it to a specific application - but that's up to anyone using it to do.

/Henrik.