MIMXRT1021 characteristics of internal RTC oscillator

[alex_conway]

I am using the watchdog, and I have found that it's source clock is the RTC clock.
I am not using a crystal or external RTC clock source so I'm relying on the internal ring-oscillator.
The characteristics of the ring-oscillator are inconsistently specified:

In the Hardware User Guide it is specifed as "...on-chip 40-kHz oscillator. The tolerance is ±50 %" (table 4)
In the Techincal Datasheet it has a similar specification in section 4.1.4 but in section 4.2.4.2 it says "The frequency range of this block is approximately 10–45 kHz."

Choosing a watchdog timeout time is sometimes a delicate balancing act, and a +/-50% tolerance on the source clock will have to be accommodated.

The questions I have are:

1. What is the actual nominal frequency of oscillation? If not 32kHz then why not?

2. If the nominal Frequency is not 32kHz, wouldn't that mean that the watchdog timeout is already wrong?

3. How does it vary with temperature? ie does it speed up or slow down when it gets hot?

4. Does the +/-50% include temperature variations? 

Thanks

Alex

[alex_conway]

Hi Habib,

Thanks for the answer, I have a few follow up questions:

Since you state that the nominal frequency is 32kHz, are the watchdog timeout times referenced to a nominal 32.786kHz, regardless of what frequency the oscillator is running at?

How much of the +/-50% is initial value tolerance and how much is temperature variation?

Thanks

Alex

[Habib_MS]

Hello @alex_conway, 
Answering your questions:

Since you state that the nominal frequency is 32kHz, are the watchdog timeout times referenced to a nominal 32.786kHz, regardless of what frequency the oscillator is running at?

The WDOG's clock frequency is derived from the oscillator. If the oscillator operates at a frequency other than 32 kHz, this deviation will directly affect the WDOG's timing.  That is why the internal oscillator is helpful for faster startup and tamper protection, but it should not be the only source for the 32 kHz clock.

How much of the +/-50% is initial value tolerance and how much is temperature variation?

As I mentioned in my previous post, frequency deviation due to temperature is affected by multiple factors and can be unpredictable. However, within the operating temperature range of -40°C to 105°C, the variation will remain within ±50%.

BR

Habib

[alex_conway]

Hi Habib,
So, to be clear, you are saying that the +/-50% includes all possible variations including initial value and temperature from -40C to 105C?

Is this tested in production? - i.e. are all chips tested from -40C to 105C to ensure that the oscillator frequency is within +/-50% or 32kHz?

Thanks
Alex

[Habib_MS]

Hello @alex_conway,

So, to be clear, you are saying that the +/-50% includes all possible variations including initial value and temperature from -40C to 105C?

Yes. initial value and temperature are included in this percentage

Is this tested in production? - i.e. are all chips tested from -40C to 105C to ensure that the oscillator frequency is within +/-50% or 32kHz?

All information of the datasheet is tested for production.

BR
Habib

[alex_conway]

Hi Habib,

All information of the datasheet is tested for production.

Which information in which datasheet? I refer you back to my original post:

"In the Hardware User Guide it is specifed as "...on-chip 40-kHz oscillator. The tolerance is ±50 %" (table 4)
In the Techincal Datasheet it has a similar specification in section 4.1.4 but in section 4.2.4.2 it says "The frequency range of this block is approximately 10–45 kHz."

None of these says 32kHz +/- 50% from  -40C to 105C, and none of these is in a table, which is the usual place for specifications that the manufacturer will stand by.

Thanks

Alex

[Habib_MS]

Hello @alex_conway,

I asked internally in order to obtain more information regarding the nominal value and the impact of temperature on this oscillator. Thank you for your patience.

In the meantime, I have a question: could you share more details about how you are currently testing the WDOG timeout? If you are using the RT1020-EVK, please note that it includes an external 32kHz oscillator, which means it does not rely on the internal oscillator:

BR
Habib

[alex_conway]

Hi Habib,
I'm not using the RT1020-EVK, I have a board with no RTC clock or crystal. The RTC_XTALI pin is connected to GND and the RTC_XTALO pin is open.
Thanks
Alex

[Habib_MS]

Hello @alex_conway,

Sorry for the late reply. Answering your questions:

1. What is the actual nominal frequency of oscillation? If not 32kHz then why not?

The nominal frequency is 32 kHz, but it can vary up to ±50%, Depending on process variations and temperature conditions.

2. If the nominal Frequency is not 32kHz, wouldn't that mean that the watchdog timeout is already wrong?

Could you provide me more details about how you are currently testing the WDOG timeout?

3. How does it vary with temperature? ie does it speed up or slow down when it gets hot?

The frequency deviation due to temperature is influenced by multiple factors and can be unpredictable. However, within the operating temperature range of -40°C to 105°C, the variation will remain within ±50%.

4. Does the +/-50% include temperature variations?
Yes, the ±50% tolerance accounts for variations due to temperature, as well as process variations.

BR
Habib