SPBRG not programming with #use rs232 for F874

[jventerprises]

hi everyone,

i am trying to use the UART on an F874. everything seems to work except the SPBRG doesn't get programmed! (i have the same code working on a F877a)

the relevant bits from my header file are:

#include <16F874.h>

#device ICD=TRUE
#device adc=10
#use delay(clock=4000000)
#fuses NOWDT,XT, PUT, NOPROTECT, NOLVP, NOCPD, NOWRT, BROWNOUT, DEBUG
#use rs232(baud=4800,parity=N,xmit=PIN_C6,rcv=PIN_C7,bits=8)


i have to add the following line in main() for things to work:

SPBRG = 12;

and this in the include:

#byte SPBRG = 0x99


the 874 is operating at 4mhz, so this should give me 4800 baud. I have the same code (really!) for an 877a running at 20mhz and the SPBRG programs just fine. when i include the two lines above everything runs error free (for hours at least, that is as long as i have run the BER)

anybody seen this before? my compiler is 3.187. I checked the bug list, but didn't see anything that sounded like this.

thanks.
_________________
Jon

[PCM programmer]

[jventerprises]

hi!

well here is my list code for the init stuff

.................... void main()
.................... {
*
05DA: CLRF 30
05DB: MOVLW 59
05DC: MOVWF 2F
05DD: MOVLW 14
05DE: MOVWF 59
05DF: MOVLW 71
05E0: MOVWF 5A
05E1: MOVLW 00
05E2: MOVWF 5B
05E3: MOVLW 0B
05E4: MOVWF 71
05E5: MOVLW A0
05E6: MOVWF 72
05E7: MOVLW 00
05E8: MOVWF 73
05E9: MOVLW 45
05EA: BSF 03.5
05EB: MOVWF 20
05EC: MOVLW F1
05ED: MOVWF 21
05EE: MOVLW 00
05EF: MOVWF 22
05F0: MOVLW 0B
05F1: MOVWF 71
05F2: MOVLW 00
05F3: MOVWF 72
05F4: MOVWF 73
05F5: CLRF 04
05F6: MOVLW 1F
05F7: ANDWF 03,F
05F8: MOVLW 06
05F9: MOVWF 1F
05FA: MOVLW 0C
05FB: MOVWF 19
05FC: MOVLW 22
05FD: MOVWF 18
05FE: MOVLW 90
05FF: BCF 03.5
0600: MOVWF 18
.................... int1 M433_RX_IN; // a message has come in the 433 and is in the buffer
....................
.................... // SPBRG = 12;
.................... setup_adc_ports(RA0_RA1_RA3_ANALOG);


these two lines should do it...
05FA: MOVLW 0C
05FB: MOVWF 19

but they dont because the bank bit for bank1 was cleared by
05F6: MOVLW 1F
05F7: ANDWF 03,F

and i am missing your line that sets back to bank!

!!

therefore address 0x19 is updated, not 0x99. my init functions look like

int1 M433_RX_IN; // a message has come in the 433 and is in the buffer

// SPBRG = 12;
setup_adc_ports(RA0_RA1_RA3_ANALOG);
setup_adc(ADC_CLOCK_DIV_8);
setup_psp(PSP_DISABLED);
setup_spi(SPI_MASTER|SPI_H_TO_L|SPI_CLK_DIV_16);
setup_counters(RTCC_INTERNAL,RTCC_DIV_1);
setup_timer_1(T1_DISABLED);
setup_timer_2(T2_DIV_BY_4,125,4);
enable_interrupts(INT_TIMER2);

output_high(SPI_CS);

// make sure the XCVR is idle
XCVR_433_control(TRANSMIT_IDLE);
XCVR_433_control(RECEIVE_ACTIVE);

// allocate memory for buffer and clear
TXRX_buffer = malloc(60);
memset(TXRX_buffer, 0, 60);
.
.
.

here is the relevant part of my header file

#include <16F874.h>

#device ICD=TRUE
#device adc=10
#use delay(clock=4000000)
#fuses NOWDT,XT, PUT, NOPROTECT, NOLVP, NOCPD, NOWRT, BROWNOUT, DEBUG
#use rs232(baud=4800,parity=N,xmit=PIN_C6,rcv=PIN_C7,bits=8)

#include "STDLIBM.h"

#byte SPBRG = 0x99



something funny is going on.
_________________
Jon

[PCM programmer]

Post a test program that causes that output.
The problem is apparently caused by some #use statement
or #device statement that affects the startup code.

Post a test program that is small, but shows the problem.

[jventerprises]

but it might not be until friday. thanks for your help!
_________________
Jon

[jventerprises]

it seems to be the inclusion of STDLIBM.h that does it. here are the stripped files.

sprofiler.c::

#include "sprofiler.h"

#int_TIMER2
TIMER2_isr()
{

}


/************************************/

/************************************/

void main()
{
// SPBRG = 12;
setup_adc_ports(RA0_RA1_RA3_ANALOG);
setup_adc(ADC_CLOCK_DIV_8);
setup_psp(PSP_DISABLED);
setup_spi(SPI_MASTER|SPI_H_TO_L|SPI_CLK_DIV_16);
setup_counters(RTCC_INTERNAL,RTCC_DIV_1);
setup_timer_1(T1_DISABLED);
setup_timer_2(T2_DIV_BY_4,125,4);
enable_interrupts(INT_TIMER2);

}

and sprofiler.h::

#include <16F874.h>

#device ICD=TRUE
#device adc=10
#use delay(clock=4000000)
#fuses NOWDT,XT, PUT, NOPROTECT, NOLVP, NOCPD, NOWRT, BROWNOUT, DEBUG
#use rs232(baud=4800,parity=N,xmit=PIN_C6,rcv=PIN_C7,bits=8)


#include "STDLIBM.h"


if you build without the STDLIBM.H include, everything is ok. if you build with it, you get the bank fault as before....


maybe the include is in the wrong place?
_________________
Jon

[jventerprises]

its #include "STDLIBM.h" thats doing it
Posted: Wed Nov 24, 2004 6:51 pm

--------------------------------------------------------------------------------

i posted this again since i posted it right before thanksgiving and people were busy. i really would like to get this fixed....



it seems to be the inclusion of STDLIBM.h that does it. here are the stripped files.

sprofiler.c::

#include "sprofiler.h"

#int_TIMER2
TIMER2_isr()
{

}


/************************************/

/************************************/

void main()
{
// SPBRG = 12;
setup_adc_ports(RA0_RA1_RA3_ANALOG);
setup_adc(ADC_CLOCK_DIV_8);
setup_psp(PSP_DISABLED);
setup_spi(SPI_MASTER|SPI_H_TO_L|SPI_CLK_DIV_16);
setup_counters(RTCC_INTERNAL,RTCC_DIV_1);
setup_timer_1(T1_DISABLED);
setup_timer_2(T2_DIV_BY_4,125,4);
enable_interrupts(INT_TIMER2);

}

and sprofiler.h::

#include <16F874.h>

#device ICD=TRUE
#device adc=10
#use delay(clock=4000000)
#fuses NOWDT,XT, PUT, NOPROTECT, NOLVP, NOCPD, NOWRT, BROWNOUT, DEBUG
#use rs232(baud=4800,parity=N,xmit=PIN_C6,rcv=PIN_C7,bits=8)


#include "STDLIBM.h"


if you build without the STDLIBM.H include, everything is ok. if you build with it, you get the bank fault as before....


maybe the include is in the wrong place?
_________________
Jon

[PCM programmer]

The problem is caused by this line, in stdlibm.h:

#USE DYNAMIC_MEMORY

That line causes the compiler to put in startup code for malloc, etc.

Apparently the compiler does not realize that it has cleared
the bank select bits when it puts in these two lines:

MOVLW 1F
ANDWF STATUS,F

So it doesn't re-select Bank 1, as it should.

There may well be some trick that can cause the compiler
to do it correctly. I didn't find one in the little time that
I played with it. Even so, you'd still have to document it
if you did find one. The best solution may be to do what
you're doing. Set the SPBRG manually, and document why
you did it.