About the calibration method of RTC in STM32

One of the core functions of RTC calibration is void BKP_SetRTCCalibrationValue (uint8_t CalibrationValue) function in library file Stm321f0x_bkp.c. The reference documents related to RTC calibration include AN2604.pdf, AN2821.pdf and AN2821.zip. All three documents can be downloaded from the official website of STM32.

According to the principle described in AN2604.pdf, the calibration value of RTC should be between 0-127. The achievable calibration error corresponds to 0-121ppm. This is equivalent to 0-314s for every 30 days.

A key point to note here is that RTC can only calibrate fast running, not slow running. If the nominal frequency of the watch crystal is 32768Hz, and the possible error range is ±2Hz, the actual frequency will be between 32766Hz-32770Hz. If the internal frequency division coefficient of the RTC is set to 32768, then 32768Hz is a frequency that does not require calibration, and 32768Hz-32770Hz is a frequency that can be calibrated (the maximum calibration capability is about 32772Hz). However, the slow running frequency range of 32766Hz-32768Hz cannot be calibrated. For this reason, in the recommended calibration method, 32766 is used instead of 32768 as the frequency division coefficient. In this way, 32766Hz is a frequency that does not require calibration, and 32766Hz-32770Hz is a frequency range that can be calibrated. The

remaining question is how to measure the error and use it to derive the calibration value. Generally speaking, there are two methods. One is to measure the frequency value of TamperPin and then calculate the ppm error; the other is to actually run for a certain number of days, compare it with the standard clock, first get the number of seconds that run fast every 30 days, and then calculate the ppm error. The

first method is described in detail in AN2604.pdf and AN2821.pdf. AN2821.zip uses timer T2 to automatically measure the frequency value of TamperPin and realizes automatic calibration. Automatic calibration does simplify user operation, but it depends on the accuracy of the 8MHz master clock. Automatic calibration cannot achieve a result with higher accuracy than the 8MHz master clock. Therefore, it is still a foolproof policy to leave a manual calibration interface for users. Even with automatic calibration, manual and automatic functions can be superimposed.

On the other hand, using the first method for calibration requires accurate measurement of the frequency value of TamperPin, such as reaching an accuracy of 511.xxxHz. Ordinary oscilloscopes cannot do this, nor can ordinary frequency meters. Only high-precision frequency meters can do this. Only professionals engaged in measurement have such equipment. As a person engaged in control systems, it is still difficult to use the first method to make a clock with non-measurement accuracy.

Whether it is the first method or the second method, the core is to calculate the ppm error. Let's first look at how the first method calculates the ppm error. Since 32766 is used as the frequency division coefficient, 32766Hz is a reference frequency that does not require calibration. Don't take 32768Hz too seriously. It is nothing now. 32766Hz can be regarded as the new nominal frequency. The frequency of TamperPin should be 32766Hz/64=511.968Hz. This is the reference frequency used repeatedly in the document to calculate the error. According to the example given in the document, if the measured frequency of TamperPin is 511.982Hz, the error is 27.35ppm. The calculation process is (511.982Hz-511.968Hz)/511.968Hz*10^6 = 27.35ppm. The document finally gives the closest calibration value of 28. Note that this is the final calibration value of 28, which is obtained by looking up 27 ppm in the table, not the misunderstanding in some posts that 27.35ppm is approximated to 28ppm.

In fact, the calculation formula for ppm error is: ppm error = deviation/reference value*10 to the sixth power. Based on this, when using the second method, the number of seconds running fast every 30 days is first obtained. The number of seconds that are faster is the deviation, and 30 days is the reference value. So ppm error = number of seconds that are faster every 30 days / (30 days * 24 hours * 3600 seconds) * 10 to the 6th power. This formula can easily explain the "0.65ppm is about 1.7 seconds per month error" mentioned in the document AN2604.pdf. Because: 1.7/(30*24*3600)*10^6 = 0.65ppm. After

calculating the ppm error, you still need to solve the table lookup. Don't pay much attention to the table given in the document. After figuring out where this table comes from, you can use a simple calculation formula instead of looking up the table. AN2604.pdf says that if the calibration value is 1, then when calibrating the RTC, 1 clock pulse is deducted for every 2 to the 20th power clock cycle. This is equivalent to 0.954ppm (1/2^20*10^6 = 0.954). The maximum calibration value is 127, so the maximum slowdown is 121ppm (0.954ppm*127 = 121). So this calibration table is obtained by simple multiplication and division operations. Of course, floating point operations are required to get accurate results.

The following is the RTC calibration program implemented using the second method.
First, two constants are defined. One is PPM_PER_STEP, which is accurate to the precision of 0.9536743ppm that can be represented by floating point numbers. The other is PPM_PER_SEC, which is the ppm error corresponding to one second faster every 30 days, which is accurate to the precision of 0.3858025ppm that can be represented by floating point numbers.   

#define PPM_PER_STEP   0.9536743 //10^6/2^20.
#define PPM_PER_SEC   0.3858025 //10^6/(30d*24h*3600s).

Then define the global variable FastSecPer30days. Set through the user menu and pass to the RTC calibration program.

u16 FastSecPer30days = 117; //Menu input. 117 is only used for demonstration.

The implemented calibration function is:

void RTC_Calibration(void)
{
  float Deviation = 0.0;
  u8 CalibStep = 0;
  
  Deviation = FastSecPer30days * PPM_PER_SEC; //Get ppm error
  Deviation /= PPM_PER_STEP; //Get the floating point number of the calibration value
  CalibStep = (u8)Deviation; //Get the integer number of the calibration value
  if(Deviation >= (CalibStep + 0.5))
   CalibStep += 1; //Round
  if(CalibStep > 127)
   CalibStep = 127; //The calibration value should be between 0 and 127
  
  BKP_SetRTCCalibrationValue(CalibStep); //Call library function
   
}
//End of function RTC_Calibration