MSP430F5419A Internal Temperature Calibration Issue

sandesh mukartihal said:

ADC2_DEG() holds the temperature in 10's degree C. The temperature values I am getting makes no sense to real temperature readings.

You can get better responses if you give all the information we need to do the math. So please show all data needed to calculate result: calibration values, ADC reading and room/chip temperature you measure

ADC Reading: ADC12MEM6 = 1864

Calibration values: ADC_2V5T30CAL = 1320, ADC_2V5T85CAL = 1532

temperature(C) = { (1864-1320) * [(85-30)/(1532-1320)] } + 30 = 171.13

If those calibration parameters are correct, and your reference is setup properly, then the math is correct.

But I would question the one of those first two assumptions. Are you sure you have the calibration values for 2.5V reference? Are you sure that the 2.5V reference is properly selected and turned on?

You don't post the part of the code that shows when you sample the temperature, but be advised that when using the reference module, it has a settling time that you have to wait before it is stable and can be used for making measurements. Check the User's Guide and datasheets for your device.

sandesh mukartihal said:

ADC12CTL0 &= ~ADC12ENC, ADC12CTL0 |= ADC12ENC; /* trigger ADC */

Also, you use ADC12SHS_3, so CCR1 will trigger the ADC. But right in the ISR whcih is triggered at the same tiem the ADC starts the sequence, you disable and re-enable ENC, which interrupts the sequence and leaves unpredictable results.

So let the timer trigger the ADC sequence start , or do it manually, but not both.

sandesh mukartihal said:

   if (TBIV & TB0IV_TB0IFG) 

If TB0IFG is the only interrupt that can happen, you don't need the IF at all (as it is the only situation in which the ISR could have been called). To clear the interrupt, you can do "TB0CTL &= ~TBIFG;" Or (this is a destructive read too) "(TB0IV);"

And if you have more than one interrupt, this IF will prevent you from properly detecting which one it was if it wans't the first one you checked.

I am able to use the following code and make it work from the temerature range -40 to +85degC in MSP430F5419A micro.  I used on time sensor value read.  The only problem i faced is i have explictly do ADC12CTL0 &= ~(ADC12SC); to stop the conversion.  Otherwise, the conversion never stopping and it is always in continuous BUSY. Also the ADC value in ADC12MEM4 is not available immediatley even after IFG flag is set.  I have to give a delay to read the ADC12MEM4 value after bit in IFG1 is set.

#define ADC_2V5T30CAL 0x1A22
#define ADC_2V5T85CAL 0x1A24

void ReadOnChipTemperatureSensorValue()
{
  // Allow ADC_12 to control everything
  REFCTL0 = 0;
  ADC12MEM4 = 0;
  ADC12CTL0 = ADC12SHT0_1 | ADC12SHT1_1 | ADC12ON | ADC12REF2_5V | ADC12REFON; 
  ADC12CTL1 = ADC12CSTARTADD_4 | ADC12SHS_0 | ADC12CONSEQ_0 | ADC12DIV_0;
 
  ADC12CTL2 = ADC12RES_2 | ADC12SR;
  ADC12MCTL4 = ADC12SREF_1 | ADC12INCH_10;                 /* A10 = THERM */
 
  ADC12MCTL4 |= ADC12EOS; /* End of sequence */
  ADC12CTL0 |= (ADC12ENC | ADC12SC); /* trigger ADC */

  // wait for the conversion to happen
  for (volatile int j=0; j<500; j++);

  ADC12CTL0 &= ~(ADC12SC); /* Stop conversion */
 
  // wait for the data to be updated
  while (ADC12MEM4 == 0);
 
  uint16 adc85CalVal = *((uint16 *)ADC_2V5T85CAL);
  uint16 adc30CalVal = *((uint16 *)ADC_2V5T30CAL);
 
  int32 g_unTemperature = (int32)(8500 - 3000)/((int32)adc85CalVal-(int32)adc30CalVal);
  g_unTemperature = g_unTemperature * ((int32)ADC12MEM4-(int32)adc30CalVal);
  g_unTemperature += (int32)3000;
  g_fTemperature = ((fp32)g_unTemperature)/100.0; 
}

prasad samudrala said:

i have explictly do ADC12CTL0 &= ~(ADC12SC); to stop the conversion.

Not to stop, but to start! You don't set the ADC12SHP bit. So ADC12SC directly controls the sample&hold gate. When you set SC, charging the sampling capacitor starts, and when you clear it, sampling ends and the conversion is started. This mode is mainly used in conjunction with a timer signal,controlling the ADC timing.
If you set ADC12SHP, then ADC12SC is only used as trigger, and the smapling timing is done by the internal sample timer, controlled by the ADC12SHTx bits.

Thanks Michael, it was my misunderstanding of ADC12SC bit.

@Prasad:

I tried this code, i get values in tens of degrees. For example, g_fTemperature=182 which makes more sense as 18.2 C.

Any ideas why?

The result shoud be in 1/100 C, but 1.82C seems wrong. It could be that
1) the reference voltage hasn't settled )the code doesn't wait fo rthe reference, so depending on CPU speed, the first conversion is probably wrong)
2) the calibration values stored in the chip are wrong. It has happened in the past that one of the values was wrong for a whole batch of devices. A rare case, but possible. On my own devices (based on MSP430F1611 without these calibration values), I usually do my own 2-point calibration.
It'S a good idea to make a probability check: take the two values and ensure that the 85C value is significantly larger than the 30C value. If not, complain.

Thanks Michael for your time.

According to User Guide, section "1.13.4.3 temperature Sensor Calibration", the Temp (in C) is calculated as below:

Temp = ( ADC(raw) - CAL_ADC_15T30) x  ( (85-30)/(CAL_ADC_15T85-CAL_ADC_15T30) ) + 30

So, How would the result be in 1/100C?

I read the temperature value using "ADC2_10DEG()" at regular intervals of 1 second.

#define ADC2_10DEG() ((uint16) (((ADC12MEM6-ADC_2V5T30CAL)*(850L-300L)/(ADC_2V5T85CAL-ADC_2V5T30CAL)+300L)))

The values I get are as below:

ADC Reading: ADC12MEM6 = 1864

Calibration values: ADC_2V5T30CAL = 1320, ADC_2V5T85CAL = 1532

temperature(C) = { (1864-1320) * [(85-30)/(1532-1320)] } + 30 = 171.13.

Do you think the above calibration values are wrong?

I use the same ADC settings to read the system voltage through another channel (A14) and it works perfectly for me:

/* Internal reference */
#define REF_VOLTAGE				2500L	

/* 12 bit */
#define ADC_RES					4096L	

/* VSYS voltage divider circuit */
#define VSYS_R1					68L
#define VSYS_R2					10L

#define VSYS2VOLTS() ((uint16) (ADC12MEM4 * REF_VOLTAGE *(VSYS_R1+VSYS_R2)/(VSYS_R2*ADC_RES)))

use 7.12

http://www.mikrokontroler.pl/content/7-mikrokontrolery-msp430-obsluga-adc/strona/0/4

making mul and div, convert values into float,

The first formula you cited assumes 85C-30C = 55 steps difference between the calibration values.
The #define you posted uses 850-300 =550 steps (which means, it gives a result in 1/10C)

And the formula in the Prasad’s post used 8500-3000 = 5500 steps, giving centigrades.

Thanks Michael, I was waiting to receive next batch of boards to do the testing.

I am still getting the temperature reading in tens of degrees. I have changed the #define now to use 85 and 30 instead of 850 and 300.

I believe it is working properly, but only the reading I get are in tens of degrees.

Do you think the values I posted(Calibration values at 85C and 30C, and the ADCMEM readings) in my previous post are correct?

Jens-Michael Gross said:

You (or whoever defined the addresses) was apparently using the TLV structure for the non-A version of the 5419.

This is where the correct way is to use a "TLV walking" function to get the data rather than a hard-coded address. (MSP430Ware has examples).

Do this once during startup and store the calibration values in a local static structure.

Downside is it takes a little more code and time. Upside is you do it once and it is portable across devices in the same family.

@Brian: Can you please provide the exact location of the example that uses the TLV walking?

I am referring to example: msp430x54xA_adc12_10.c in slac375e and this doesn't use TLV walking.

examples\driverlib\MSP430F5xx_6xx\tlv