sd 16 not get accurate result

i am useing msp430f4270 controller.in it i attech channel 2 a opamp amplifier to amplify voltage.i use sd16 channel 2 to meassure milivolts . i use calibration technique for nullify offset effect and linear equation for converting counts to milivolts. here i attech my code and snap of schematic.

here is my code

#include <msp430.h>

//#define Num_of_Results 50

/* Arrays to store SD16_A conversion results */
/* NOTE: arrays need to be global to */
/* prevent removal by compiler */
#define a 0x80
#define b 0x40
#define c 0x20
#define d 0x10
#define e 0x08
#define f 0x04
#define g 0x02
#define h 0x01
unsigned char common[4],n;
int i=0,mode=0,display,number1,test,y1,offset,dif;
int PW[4],digit=3,done=0,value[4],tp,amb_offset,ambient,cal_amb;
char list_mode,pass=0,zero1=0,cal1=0,cpass1=0;
unsigned char gain=1,dp=0,val_u1,y=0,val_k1,flag;
unsigned int val[4],x1=0,x2=0,crt;
double temp=0;
char blink=0;
unsigned int packed=0,dp_val=0,inp=0,gain_val;
unsigned int hex=0xFFFF,results,oldresult;
int zero_val,span_val,outlow_val,outhi_val,packed11=0,offset_val,temp_amb;
unsigned int calz_tc,cals_tc,clol_val,cloh_val,calz_rtd,cals_rtd,calz_mv,cals_mv,z1,s1;
unsigned int pass_val,input_val,old_val=0,pinput_val,oinput_val,input21;
long int input,sample1;
long int input2;
unsigned int results12[8];
long double val_a,val_b,k2;
volatile unsigned int refresh;
//long int avg[5];

void keyscan(void);
void delay(int i);
void adc(void);
void hex2bcd(void);
void number(void);
const char char_gen[] =
{ // definitions for digits
g,//a+b+c+d+e+f, // Displays "0"
a+d+e+f+g,//b+c, // Displays "1"
c+f,//a+b+d+e+g, // Displays "2"
f+e,//a+b+c+d+g, // Displays "3"
a+d+e,//b+c+f+g, // Displays "4"
b+e,//a+c+d+f+g, // Displays "5"
b,//a+c+d+e+f+g, // Displays "6"
d+e+f+g,//a+b+c, // Displays "7"
0,//a+b+c+d+e+f+g, // Displays "8"
e//a+b+c+d+f+g // Displays "9"
};

void main(void)
{
volatile unsigned int i; // Use volatile to prevent removal
// by compiler optimization

WDTCTL = WDTPW + WDTHOLD; // Stop WDT
FLL_CTL0 |= XCAP14PF; // Configure load caps
for (i = 0; i < 10000; i++); // Delay for 32 kHz crystal to
P2DIR = 0XFF; //port initialization
P5DIR = 0X1E;
P2SEL = 0X00;
P5SEL = 0X00;
P6SEL = 0X0F;
P1SEL = 0xFF; // stabilize
CCTL0 |= CCIE; // CCR0 interrupt enabled
CCR0 = 300;
TACTL = TASSEL_2 + MC_2+ID_3;

_BIS_SR(GIE);

SD16CTL = SD16REFON + SD16SSEL0 ;
for (i = 0; i < 0x3000; i++);
SD16AE = SD16AE6+ SD16AE7;
SD16CCTL0 |= SD16UNI;
SD16INCTL0 = SD16INCH_2 + SD16INTDLY_0+SD16GAIN_2;
SD16CCTL0 |= SD16SC;

_EINT();

while(1)
{
if(flag)
adc();



}
}

void keyscan(void)
{
int key1=0xFF;
P6DIR = 0X0F;
P6SEL = 0X0F;
key1 = P6IN;
key1 = key1 & 0xF0;
key1 = key1 >> 4;

switch(key1)
{
case 0xE: //Mode key
delay(500);
key1 = P6IN;
key1 = key1 & 0xF0;
key1 = key1 >> 4;
if(key1==0xE)
{
delay(500);
while(key1==0xE)
{
key1 = P6IN;
key1 = key1 & 0xF0;
key1 = key1 >> 4;
}
calz_mv=input;
}
break;

case 0xD: // Enter key
delay(500);
key1 = P6IN;
key1 = key1 & 0xF0;
key1 = key1 >> 4;
if(key1==0xD)
{

while(key1==0xD)
{
key1 = P6IN;
key1 = key1 & 0xF0;
key1 = key1 >> 4;
}
cals_mv=input;
}


break;
}




}

void adc(void)
{
unsigned int i,sample=0;

kk: sample=0;

val_a=0;
val_b=0; // Use volatile to prevent removal
flag=0;
sample1=0;
input=0;
y1=0;
k2=0;


adc1: while ((SD16CCTL0 & SD16IFG)==0); // Poll interrupt flag
results = SD16MEM0;

results12[0]=results;

for(i=0;i<8;i++)
{
val_a=val_a+(results12[i]/8);

}

input=val_a;

keyscan();

val_a=cals_mv-calz_mv;
val_b=input-calz_mv;
val_b=24000*val_b;
val_a=val_b/val_a;
val_a=1000+val_a;
input=val_a;
input=input+offset_val;


if(input<=9999)
{
dp_val=3;
}
if(input>9999)
{
input=input/10;
dp_val=2;
}


hex2bcd();
number();
for(i=7;i>0;i--)
{
results12[i]=results12[i-1];
}



}
void delay(int i)
{
while(i!=0)
{
i--;
}
}

void hex2bcd(void)
{
unsigned int dummy;
temp=input;
if(temp<0)
{
hex=input*(-1);
}
else
{
hex=input;
}
if( (mode==0 && input > 0x270F) || (cal1==1 || cal1==2) )
{
dummy=hex/0x2710;
hex=hex%0x2710;
dummy = dummy<<12;
input=dummy;

dummy=hex/0x3E8;
hex=hex%0x3E8;
dummy = dummy<<8;
input=input + dummy;

dummy=hex/0x64;
hex=hex%0x64;
dummy = dummy<<4;
input = input + dummy;

dummy=hex/0xA;
hex=hex%0xA;
dummy = dummy<<0;
input = input + dummy;
}

else
{
dummy=hex/0x3E8;
hex=hex%0x3E8;
dummy = dummy<<12;
input=dummy;

dummy=hex/0x64;
hex=hex%0x64;
dummy = dummy<<8;
input = input + dummy;

dummy=hex/0xA;
hex=hex%0xA;
dummy = dummy<<4;
input = input + dummy;

dummy=hex;
input = input + dummy;
}
if(temp<0)
{
input=input*(-1);
}
}

void number(void)
{
unsigned int j;
//hex2bcd();

if(input>=0)
{
j= input & 0x000F; // Last character fatch
common[3] = char_gen[j];

j= input & 0x00F0; // Last character fatch
j= j>>4;
common[2] = char_gen[j];

j= input & 0x0F00; // Last character fatch
j= j>>8;
common[1] = char_gen[j];

j= input & 0xF000; // Last character fatch
j= j>>12;
common[0] = char_gen[j];
}
if(input<0)
{
input=input*(-1);

j= input & 0x000F; // Last character fatch
common[3] = char_gen[j];

j= input & 0x00F0; // Last character fatch
j= j>>4;
common[2] = char_gen[j];

j= input & 0x0F00; // Last character fatch
j= j>>8;
common[1] = char_gen[j];

common[0]=g;
}

}

// Timer A0 interrupt service routine
#pragma vector=TIMERA0_VECTOR
__interrupt void Timer_A (void)
{
int j;

refresh++;
if(refresh==600)
{
flag=1;
refresh=0;
}

if(display!=3 && display > 3)
{
display=0;
}

P5OUT = 0XFF;
switch(display)
{
case 0:
j=0; // Last character fatch
if(dp_val==3)
{
P2OUT = common[j];
}
else
{
P2OUT = common[j]+h;
}
P5OUT = 0X10;
display++;


break;

case 1:
j=1; // 3rd character fatch
if(dp_val==2)
{
P2OUT = common[j];
}
else
{
P2OUT = common[j]+h;
}
P5OUT = 0X08;
display++;


break;

case 2:
j=2; // 2nd character fatch
if(dp_val==1)
{
P2OUT = common[j];
}
else
{
P2OUT = common[j]+h;
}
P5OUT = 0X04;
display++;


break;
case 3:
j=3; // Last character fatch
if(dp_val==0)
{
P2OUT = common[j];
}
else
{
P2OUT = common[j]+h;
}
P5OUT = 0X02;
display++;


break;
}
CCR0 += 250;


}

i attech 100nf capactior to vref pin and giving voltage useing presice yokogava calibration source

problem is that when i calculate milivolts from counts not get reliable results.

i calibrate for zer0=1milivolts

                      span=25milivolts;

as you can see from code.

i does this practice because i want to display temperature from thermocouple milivolts.

i spent 2-3 days and gone through various posts but unable to solve my problem........ please help me i am tired from this sd16 because its counts is so fluctuing so i have to implement running average methode

here i not attech r24 and shorted -tca to agand

now i am giving 5.260mv and reading on my led display is 5.217mv