STM32F4 HSI Accuracy/drift

[Carl Tappan]

Posted on January 31, 2018 at 17:42

I am looking at rev 8.0 of DS8597, the datasheet for the STM32F417 micro and trying to reconcile two competing specifications for the absolute accuracy of the 16 MHz internal RC oscillator. Section 2.2.12 plainly says ''The 16 MHz internal RC oscillator is factory-trimmed to offer 1% accuracy over the full temperature range.'' In 5.3.9, Table 34, however, Accuracy of the HSI oscillator is given as +/- 1% at 25 *C and +/- 4% over the range of -10 *C to +85 *C.

Is there any interpretation in which the part meets the accuracy claimed in 2.2.12, or is this simply an error?

Best,

Carl

[Tesla DeLorean]

Posted on January 31, 2018 at 18:18

Would agree this doesn't make sense as stated. Would like to see documentation and graphs supporting this.

Not sure it drifts as much as it jumps around due to instantaneous voltage, current and temperature fluctuations.

I also don't think the trim is subject to some forth order control loop, or achieved with calibration up/down the temperature curves.

[Szymon PANECKI]

Posted on January 31, 2018 at 19:13

Hello,

HSI16 by default provides accuracy +/-1% in the temperature 25°C. For lower and higher temperature this accuracy is lower. In order to achieve 

+/-1% for HSI16 in given temperature, trimming process has to be done. It is described in

http://www.st.com/resource/en/application_note/cd00221839.pdf

. A nice example you can find on

https://ucexperiment.wordpress.com/2015/11/20/stm32f411-rc-calibration-using-a-ds3231-1hz-signal/

.

Regards

Szymon

[Tesla DeLorean]

Posted on January 31, 2018 at 19:41

The complaint would then be that the data sheet over promises.

http://www.st.com/content/ccc/resource/technical/document/datasheet/98/9f/89/73/01/b1/48/98/DM00035129.pdf/files/DM00035129.pdf/jcr:content/translations/en.DM00035129.pdf

 

The app note really doesn't address are way to measure the temperature, and calibration over temperature, or if the trim can be changed dynamically.

The other chart posted basically constrains the +/-1% to 25C and 3.3V operation

[Carl Tappan]

Posted on January 31, 2018 at 21:13

The method in that app note should work well enough at a given operating point, provided there is another stable frequency in the system and the external timer connection required to calibrate the internal oscillator off of the LSE [!] is present. To Clive's point, that is not what the datasheet says, and the language in 2.2.12 is at best misleading.

Is the 'HSI user trimming step' the change in HSI frequency per lsb of the HSITRIM value in RCC_CR, or the maximum adjustment range of the oscillator frequency? In the former case, with a 1% maximum resolution, how can the RC frequency accuracy after trim be better than 0.5% as suggested by table 4.1 in the referenced app note?