MSP430F47187: Interruption period change in low temperature

What trigger the SD16A to sampling? I think you need to double check for the trigger source.

The SD16A has the clock source set to SMCLK and OSR set to 512.
Also, SMCLK uses ACLK (32.768Hz), and we have confirmed that ACLK does not change at low temperatures.

Are you using continues conversion mode? Or single conversion mode? 

How do you know the interrupt frequency changes? Do you use oscilloscope with GPIO to test it?

Can you directly test the sampling and conversion time of ADC, instead of test the frequency?

Can you post more information of your findings?

I am using continuous conversion mode.
Changes in interrupt frequency are tested using an oscilloscope on the GPIO.
The ADC (SD16A) sampling time is 1.67ms (600Hz), and at low temperatures the sampling time becomes shorter as the temperature drops.

When the SD16A clock source was changed from SMCLK to MCLK or ACLK, it was normal even at low temperatures.
Even when the SMCLK clock source was set to MCLK, it was normal at low temperatures.

Register setting value is as follows.
SD16CTL：0x00D4
SD16CCTL0：0x0810
SD16INCTL0：0x00
SD16PRE0：0x00

The clock settings are as follows.
ACLK: 32768Hz (XT1 external crystal oscillator) ← Confirmed to be normal even at low temperatures
SMCLK: 2457600Hz ((74+1) x 32768Hz)
SD16A: 600Hz (2457600Hz/8/512) ← The lower the temperature, the more the cycle changes (about 614Hz at -10°C)

So, thank you for your test.

It seems that the problem lies on the SMCLK. What is the clock source of SMCLK. External oscillator or internal oscillator?

This is takayuki miura.
I couldn't log in with my previous account, so I'm replying with another account.

The clock source for SMCLK is DCOCLK.
DCOCLK = (74+1) × ACLK = 2457.6kHz
ACLK uses a 32.768 kHz external oscillator (model MC-306).

When the FN_X of the SCFI0 register was changed from 001x (2 to 17.9 MHz) to 0000 (0.65 to 6.1 MHz), the DMA interrupt cycle was no longer changed at low temperatures.

However, since the frequency of DCO is 2.4576MHz, there should be no problem with 001x.
It is unknown why the DMA interrupt cycle changes at low temperature.

If you check the DCO on page 40 of the data sheet, if the fn_x is 001x, it will be MAX 3.5MHz in 3V.
Therefore, is it more than 2.4576MHz, and is there a problem?

Hi,

This would be a problem. The DCO has temperature drift problem

I would suggest you use ACLK to give source to timer, and in timer interrupt, you can use software to trigger ADC sampling.

Is it better to use a timer interrupt without changing FN_x in the SCFI0 register?
For example, if you change the value of FN_x to 0000, it will be 1.3 to 2.7MHz even considering MIN MAX under 3V condition.
Therefore, 2.4576MHz should be no problem.

Ok another solution would be use timer(sourced with SMCLK) to capture the ACLK(32KHz), to calculate the right SMCLK frequency.

I would like to change the value of FN_x to 0000 as a countermeasure due to the electric circuit.
Do you have any problems?

Eason Zhou Thank you for your answers so far.
I changed the value of FN_x to 0000 and confirmed it on the actual machine, but there was no problem in operation.

I'm still unable to log in with my previous account, but this solves my question.