STM32 Series Microcontroller - Calculations of Timer Variables

Question

This is a subject that seems to have a lot of confusion (I do not fully understand). How to calculate the Variables:

• Period
• Prescaler
• Duration

if only a few other variables are known. At higher frequencies, for example 1000000Hz simple rules work. At lower Frequencies like 1281Hz things get a little more complicated.

As an example: if we aim for an output frequency of 1281Hz, we need to set the ARR Register to: 32785. We need to set the PSC Register to: 3 and if we want a 50% DutyCycle, we need to set the CCRx Register to: 16393. Where the x in CCRx is the specified Channel one is going to use. 1, 2, 3 or 4.

The Registers need to see values of 16 BITs, Integer Values(65535 or less).

Equations exist so calculations can be made, for example:

\$TIM_{Update Frequency(Hz)} = \frac{Clock}{(PSC - 1) * (Period - 1)}\$

\1281ドル.11273792094 = \frac{84000000}{3 * 32785} \$

now, the problem becomes more complicated, for a desired frequency of 840Hz:

\425ドル.946209078739 = \frac{84000000}{167 * 1189}\$

Here in lies the problem, how does one find the value of PSC when one knows only the values for "Clock" and "Period"? The value "TIM_Update_Frequency(Hz)" is not always the actual Desired Frequency.

Other equations must exist so as to calculate the "PSC" Value, or to calculate the "TIM_Update_Frequency(Hz)" so as to deduce the other values.

Guess work is just not a logical approach here.

Can anyone help with more equations for solving for the other variables?

Thank you

Chris

• gbulmer: A couple of points. 1. Numbers usually have a character every 3 digits, to make it easier to parse. For example, 1000000Hz, using the GB/USA-English convention is easier to read written as 1,000,000Hz, or elsewhere 1 000 000Hz (or even 1_000_000). 2. There are SI unit Metric prefixes so 1000000Hz is easier to understand when it is written as 1MHz. 3. A typical crystal, used as the frequency source of an STM32, is accurate to 20 parts in 1 million or worse (2 in 100,000), so quoting numbers like 1281.11273792094, is misleading; 1281.11 is enough.
• Rusty Nail: @gblumer - Yes I understand what your points are. However, many Function Generators do have 4 or 6 digit resolutions after the decimal point, expensive ones can have more. A nice example of this: ebay.com.au/itm/… put the numbers in after the decimal points really just to show that I was not messing with the numbers generated. Thank you, helpful post: +1.
• gbulmer: That link leads to a ebay.com.au/itm/… device which says "Accuracy:±2ppm". It matters less where the decimal point is, and more how many significant figures are used. So "±2ppm" is only one more digit than a typical crystal, i.e. not 1281.11273792094 but only 1281.113.
• EmbeddedSoftwareEngineer: I don’t understand the question, how can you know the period register value without knowing the prescalar value? The prescaler is to help achieve longer timer periods. Normally you start with a known system/peripheral clock and a desired timer frequency you then need to deduce the prescaler and timer period register values, is this what you are asking?

Answer 1

I have deleted my last answer because it was not correct all of the time. I have managed to use this equation to get correct results:

Prescaler = ((((ClockSpeed) / ((period) / (1 / frequency))) + 0.5) - 1);

Factoring Numbers was not the answer.

Here is a simple method I am now using to work out the other variables, Period and Duration.

private void CalculateVariables()
 {
 // Set the Bus Speed (APB1 or APB2), all timers are APB1 except TIM1 and TIM8...
 if (PWMPin == CPU.Pin.Pin7_X3 || PWMPin == CPU.Pin.Pin9_X3 || PWMPin == CPU.Pin.Pin7_X4)
 {
 ClockSpeed = CPU.Clock.APB2;
 }
 else
 {
 ClockSpeed = CPU.Clock.APB1;
 }
 #region Fields...
 uint ONE_MHZ = 1000000;
 uint prescaler = 0;
 uint period = m_period;
 uint duration = m_duration;
 uint scale = ((uint)m_scale);
 // Scale in MHz... 
 uint scaleMHz = scale / ONE_MHZ;
 uint clk = ((uint)ClockSpeed / ONE_MHZ);
 if (duration > period) duration = period;
 #endregion
 #region Calculate Variables (This is a C# implimentation of GHI Source ("PWM_ApplyConfiguration"))...
 prescaler = clk / scale;
 if (prescaler == 0)
 {
 if (period > (0xFFFFFFFF / clk))
 {
 // Avoid Overflow... 
 prescaler = clk;
 period /= scaleMHz;
 duration /= scaleMHz;
 }
 else
 {
 prescaler = 1;
 period = period * clk / scaleMHz;
 duration = duration * clk / scaleMHz;
 }
 }
 else
 {
 while (prescaler > 0x10000)
 {
 // Prescaler too Large... 
 if (period >= 0x80000000) break;
 prescaler >>= 1;
 period <<= 1;
 duration <<= 1;
 }
 }
 bool TimerIs16Bit = true;
 // Were not using Timer 2 of Timer 5...
 if (TimerIs16Bit)
 {
 // All Timers are 16 bit Timer's except 2 and 5... 
 while (period >= 0x10000)
 {
 // period too large 
 if (prescaler > 0x8000) break;
 prescaler <<= 1;
 period >>= 1;
 duration >>= 1;
 }
 }
 #endregion
 if (debug)
 {
 Debug.Print("*** Scale *******");
 // Print to Debug window new config...
 Debug.Print("Period = " + period);
 Debug.Print("Prescaler = " + PrescalerFromPeriodFrequencyAndClock(period, this.Frequency));
 Debug.Print("Duration = " + duration);
 }
 ARRValue = (period - 1);
 PSCValue = PrescalerFromPeriodFrequencyAndClock(period, this.Frequency);
 CCRxValue = duration;
 }

a Simple static Class CPU and I have added:

public enum Pin : int
 {
 Pin7_X3 = 1,
 Pin8_X3 = 2,
 Pin9_X3 = 3,
 Pin7_X4 = 4,
 Pin8_X4 = 5,
 Pin9_X4 = 6
 }

Answer 2

Depending on the clock, prescaler and auto reload, repetition counter (if present) parameters, the 16-bit timer can generate an update event from a nanosecond to a few minutes.

For example, to generate an event two times per second, choose Clock frequency(TIM_CLK) = 168 Mhz, Prescalar(PSC) = 16799, Period(ARR) = 4999, repetition counter(RCR) = 0,

then the 16-bit timer will generate an event two times per second.

enter image description here

For more information ;

STM32 Timer Overview

Answer 3

There is a tool written in python to do all the calculations: https://github.com/v0idv0id/STM32-Scaler