Question for Mr. Darrel Taylor and Mister-e Timer1 Reload

[mikebar]

Good morning,
I'm trying to understand the purpose of the slider "Reload (Instr. Cycle)" anyone can see into the fantastic tool named Pic MultiCalc made by Mister-e.
The reason why I'm trying to fully understand it is because I'm trying to setup Timer1 to interrupt each 100ms.
So, having a 20Mhz oscillator and using Pic MultiCalc - Timer Helper, with a "Reload Instr. Cycles" of (default) 7 and a Prescaler of 8, I've got the value of 3036 with an error of 0.001% to pre-load Timer1.
So, the Timer1 really interrupts each 100.001 ms.

I've used the very good Darrel's Instant Interrupt and it works like a charm.
BUT now here is the problem.

I need a very precise time-base, precise at least for 24 hours continuously.
Doing the math, in 24 hours are 60 * 60 * 24 = 86400 seconds.
With a 0.001% error, in 24 hours I'll have 86400 - 0.001% = 86399.136.
Doing integer rounding to show on a display, I'm loosing about 1 second per day.

I know that the "reload" is used to "compensate" the time "loose" to get to the interrupt and to reload the Timer1 register.
Playing with that slider, I've seen that I could land to 0.0% error if "Reload Instr. Cycle" is 2 or 8.

That said, the whole question is: how to estimate the correct value to be used as input for that slider? How to calculate the time needed for that task? Is the default value of 7 for what reason?

Also a little request to Mister-e: Pic MultiCalc on Windows7 have all the captions of the labels and text on the buttons that overflow the boundaries of the button or label (see picture).
The reason is of course Micro$oft bad habits. In Windows7 the standard font get bigger than XP.
Could you please increase the size of the labels and the buttons in order to fit it, or just use the Tahoma font (even a little smaller just to be sure...)
THANK YOU A LOT for your good tool and help.

Mike

[Demon]

I'll let Mr E know.

Robert

[Normnet]

I would step through the interrupt with MPLAB SIM to more accurately determine the time involved for interrupt reload.
I think interrupt reload is the time or instructions used on interrupt to again restart the interrupt.


Norm

[richard]

typical timer1 reload example in asm
; All interrupts are disabled
CLRF TMR1L ; Clear Low byte, Ensures no
; rollover into TMR1H
MOVLW HI_BYTE ; Value to load into TMR1H
MOVWF TMR1H, F ; Write High byte
MOVLW LO_BYTE ; Value to load into TMR1L
MOVWF TMR1H, F ; Write Low byte
; Re-enable the Interrupt (if required)
this is 7 instructions when you include the enable/disable interrupts

another thing to consider is that your typical xtal is only accurate to about 50 parts per million , so unless your xtal is in a temperature controlled oven your expecting too much, fudging the preload is not a proper solution.

[pedja089]

One instruction to jump to interrupt, one to stop timer, 2 to load TMR1H, 2 to load TMR1L, and 1 to enable time. So that is 7 instruction. So if you need 8, just add @ NOP.
But you say that you use DT_Int, so you probably use more than 7 instruction. You need to check dissembly of hex file to count how instruction is used between int vector, and timer reload. Also, this depending on clock source for that timer. This is case if clock source is from OSC/4. If you have 32768Hz crystal, then your Reload cycle is 0. Because TMR clock is much slower than interrupt. So probably 1 clock cycle for tmr last longer then all instruction to reload timer.

Your interrupt error is 0.001%, but crystal error are probably 1-5%, so that small error won't contribute much to overall time precision.
EDIT:
richard type faster than me

[mikebar]

Thank you all for your answers.
Really appreciated.

Richard, you said that "fudging the preload is not a proper solution". Could you please suggest what could be a better approach to the problem?
What I "really" need is not to show an increment or a loss of even 1 second while the clock is running for EXACTLY 24 hours continuously.
Of course, if the suggested approach give good results beyond the 24 hours, that's great!

Thank you again for your help.
Regards.
Mike.

[richard]

if using the appropriate reload and preload as calculated is not keeping good time then it means you xtal frequency is slightly off (as pedja089 suggested 1 to 5 % error is to be expected) , you might try trimming the capacitors on the crystal (might take a bit of experimenting if you haven't got a frequency counter) , changing the preload + or minus 1 will probably be too coarse an adjustment . you need to keep in mind that xtal ocillators aren't atomic clocks and that the osc frequency will be temperature dependant and will drift over time .
another possibility is adding or subtracting the "error" say once per day, this was used in a slicon chip magazine project (pic toc )many years ago and was suprisingly good at time keeping .

[mikebar]

Ok, thank you Richard.
I'll try to get a freq meter and see what exactly going to the oscillator.

The chip I'm using is PIC 18F4685. What about to connect a 32768Hz watch quartz in order to feed the Timer1?
I've had no experience yet with this approach so I don't know yet what exactly I'm talking about.
All I need is to interrupt each exactly 100ms. It is possible to have that using the Timer1 feeded by an external source? And if so, using what setup?

Thanks in advance.
Regards,
Mike.

[pedja089]

10Hz interrupt is problem to derive from 32768Hz. 16Hz or 32Hz etc isn't problem. To get 1sec just count 16 interrupts, and that is all.
I posted RTC code, with external crystal, 32 interrupts per second, and it consumes less than 10uA if you put pic in sleep mode.

[mikebar]

NICE!
Where I can see that code?

Thank you.

[pedja089]

I can't find that topic...
Here is cut/paste from my code for PIC18LF14K50

        DEFINE  OSC 4         
        Define  INTHAND myint    ' Define interrupt handler
        DEFINE  NO_CLRWDT 1

        OSCCON=%01010110
        OSCCON2.2=0
        OSCTUNE.7=1
        OSCTUNE.6=1
        REFCON0=0
        REFCON1=0
        ADCON1=0
        ADCON2=0
        ANSEL =0           
        ANSELH=0
        CM1CON0=0
        CM2CON0=0
        SRCON0=0
        SRCON1=0
        WPUA=0
        WPUB=0
        WDTCON=0   
        SLRCON=0           
        RCON.7=0
        RCON.6=0   
        UCON=0  
        IOCA=0
        IOCB=0
        T0CON=0
        T1CON=%00001111  
        T2CON=0
        T3CON=%00000110
        INTCON=192  
        INTCON2=0  
        INTCON3=0 
        PIR1=0
        PIR2=0
        PIE1=1
        PIE2=0
        IPR1=0
        IPR2=0
        porta=0
        portb=0
        portc=0
        trisa=255
        trisb=255
        trisc=255




'RTC_________________________
Seconds        VAR BYTE bankA  
Minutes        VAR Byte bankA       
Hours          var byte bankA
   



 
 clear

GoTo OverInt  ' jump over the interrupt handler and sub
ASM                  
myint 
 movlb .0
 ;If PIR1.0 =1 then
 BTFSS PIR1,0 
 GOTO End_RTC 
 MOVLW .252 ; TMR1H=252
 MOVWF TMR1H
 CLRF PIR1 ; Clear TMR1 int flag
 INCF _Ticks,F 
 ;if ticks>31 then
 MOVLW .31
 CPFSGT _Ticks  
 GOTO End_RTC   
    INCF _Seconds,F  
    CLRF _Ticks
    ; if seconds > 59 then
    MOVLW .59         
    CPFSGT _Seconds   
    GOTO End_RTC   
       CLRF _Seconds   
       INCF _Minutes,F
       ;if minutes > 59 then
       MOVLW .59         
       CPFSGT _Minutes   
       GOTO End_RTC      
          CLRF _Minutes
          INCF _Hours,F
          ;if hours > 23 then 
          MOVLW .23       
          CPFSGT _Hours   
          GOTO End_RTC    
              CLRF _Hours
End_RTC                                      
;Do your other stuff here
 retfie  FAST  
ENDASM  
OverInt:

Main:
@ Sleep
'program starts here 32 times in sec
goto Main

This is example for PIC's that have fast return stack, and saves and restores WREG, STATUS and BSR registers...

[mikebar]

Thank you a lot.

[pedja089]

Just one more thing that I had to deal with...
Selection of crystal and capacitors from OSC pins to ground is critical to get accurate clock. I tried one by one from 10 to 47pF, and best results are with 47pF. But that will depend from board layout and used crystal.
My suggestion is that you buy a lot of crystals, and select appropriate capacitor for it. And use that combination from now.

[mikebar]

Thank you pedja089,
looks to be a challenge...

[pedja089]

I'm glad that I could help.
Good luck

[Normnet]

Mike

You may wish to use a 32.768 kHz watch crystal.

Following from the PIC18F/LF1XK50 datasheet:

11.6 Using Timer1 as a Real-Time Clock
Adding an external LP oscillator to Timer1 (such as the
one described in Section 11.3 “Timer1 Oscillator”
above) gives users the option to include RTC functionality
to their applications. This is accomplished with an
inexpensive watch crystal to provide an accurate time
base and several lines of application code to calculate
the time. When operating in Sleep mode and using a
battery or supercapacitor as a power source, it can
completely eliminate the need for a separate RTC
device and battery backup.

Norm

[mikebar]

Thank you Norm, that looks interesting.

What about NEVER stop Timer1 into the ISR while setting TMRL=<some preload value> and TMRH=<some preload value> in that order?
This approach will be evil or could work?

Thanks.

[Normnet]

mikebar

I'm not sure about the interrupt question.
Also see Run-Time Calibration of Watch Crystals and Software Real-Time Clock and Calendar.

Norm

[mikebar]

Thank you Normnet,
both very interesting reading.

This will put some light. I hope :-)

Thanks again.
Mike.