RTWDOG Clock Source

[dasfaas]

Hi,

I'm using Watchdog 3 (RTWDOG) on an imxrt1024 and don't have a 32kHz external xtal on the product. The internal 32kHz RCOSC is not sufficient for this, so this rules out source 1 & 2. The reference manual mentions that clock source 3 (ERCLK, external clock) is a 1MHz RCOSC based on the 24MHz RC oscillator is an option, but it's not clear how accurate this is or how to configure this? I've selected it, but it seems to tick at 500kHz.

The application calls for a 10s watchdog and as the RTWDOG clock source 0 (ipg_clock_root) is at 99MHz, this is not suitable to use to get a 10s without changing the ipg_clock_root to something quite a bit lower.

Thanks!

[kerryzhou]

“The 24 MHz RC oscillator is a self-tuning circuit that will output the programmed” , which means output clock is automatically trimmed.

So according to design spec , the accuracy is +/- 100KHz.

[kerryzhou]

Hi @dasfaas 

 1MHZ should can be used directly, and the accuracy is determined by the internal RC 24OSC, to the RC24MOSC accuracy, our internal side also check it now.

  If you directly use this 1MHZ, whether it matches your requirement or not when you test it?

Best Regards,

Kerry

[kerryzhou]

Hi @dasfaas ,

    Thank you for your interest in the NXP MIMXRT product！

    To the RTWDOG, or any other WDOG,  I highly recommend you add the external 32KHz crystal. Otherwise, the time maybe not very stable in the mass products.

    Please note, it is not you select the source resolve it, the 32Khz is very important to the wdog stable work, we highly recommend you add it.

Best Regards,

Kerry

[dasfaas]

Hi Kerry,

Thanks for getting back to me!

I understand your point about the 32kHz crystal, but I am looking at using a different clock source for this, wdog3_ext_clk which is ref_1m_clk (see attached). I am trying to find information on this clock source. It's a 1MHz clock sourced from a 24MHz RCOSC, but I can't find much information on this clocks accuracy.

Thanks again!

[kerryzhou]

Hi @dasfaas ,

    About the ref_1m_clk, it's really lack the information, I will check internally.

   But, from my support experience, no matter which source you are using, from the IP internal consideration, the 32khz is the must, as we already have some customer remove it, and cause some devices wdog is abnormal when do the mass product. It is caused by the 32khz removed. This is also confrimed by our internal R&D team.

   For 1MHz clock, I will help you check it internally, then give you reply later.

Best Regards,

Kerry

[dasfaas]

Hi @kerryzhou,

Thanks for checking this. I appreciate that. I don't need it to be very accurate, just need a watchdog set for ~10s to make sure the software doesn't hang. 

The reason the 32kHz RC OSC is not suitable is I have found it to be very unreliable and sometimes tick significantly faster than expected and cause a watchdog in ~1s instead of 10s. I was able to repeat this by freezing the MCU. In normal operation the product would appear to function as expected, but it would randomly watchdog for no apparent reason.

Any information you can provide on the 1MHz reference clock/RC OSC and the 24MHz RC OSC would be very useful.

Thanks again!

[kerryzhou]

Hi @dasfaas ,

   32kHz RC OSC need the external 32Khz crystal, then it will be precise.

   To the 1MHz, please keep patient, I am waiting for the internal reply, thanks.

Best Regards,

Kerry

[salvalatorre]

Hi @kerryzhou,

After facing the same problem, we switched to using the clock that was supposedly running at 1 MHz but seems to actually ticking at 500 kHz. The timeout was adjusted assuming that its frequency was 500 kHz.

Unfortunately, we are experiencing watchdog resets, which leads us to wonder whether the clock might switch to 1 MHz in certain cases, thereby reducing the expiration time to half of what was expected.

Did you manage to clarify the actual operating frequency of that clock through internal consultations?

Best Regards,
Salvador.