i.MX RT1064 TEMPMON Accuracy

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

i.MX RT1064 TEMPMON Accuracy

2,299 Views
tmeyer1
Contributor III

Hi,

We have a custom RT1064 design and we are noticing a very high temperature being reported from all of our boards.  For example, I'm running the "temperaturemonitor" project from the 2.7 SDK and it reports a ~46C (+24C above room temperature) even though the processor is clocked down to 75MHz. If i fix the clock to 600MHz, the temperature will rise above 50C. 

This is an industrial grade part, and the OCOTP->ANA1 reports the following:

  Hot Temp: 105

  Hot Count: 1213

  Room Count: 1337

What I find interesting is that top of part is only about 30C, despite it reporting 46C. (or 33C when 49C is reported). Running the same SDK example on a RT1060 EVK results in roughly 30C being reported, which seems reasonable.

Other tests running our custom firmware (@600MHz) shows a temperature reported consistently 25C-30C above the thermal chamber temperature. 

Could the OTP calibrated values be incorrect? Am I missing something?

Thanks,

Tim 

Labels (1)
Tags (1)
0 Kudos
11 Replies

2,120 Views
tmeyer1
Contributor III

I did a little more digging into some of the power management to see how it would effect the reported temperature. 

I merged the "power_mode_switch_bm" example with "temperaturemonitor" to print out the temperature upon exiting a low power mode.

When in "Full RUN" mode, the temperature is reported just around 50C.

After being in "Suspend" mode (option 'G') for 4 minutes, the temperature of the processor reports 44C upon exiting this mode. The PCB right next to the processor always seems to be lower than 33C. I was expect the RT1064 to be much closer to the PCB temp after 4 minutes.

I couldn't find any documented errata surrounding this, is anyone aware of any errata the RT1064 tempmon?

One difference between the RT1064 I have and those on the EVKs is that I have industrial grade parts, perhaps that is relevant?  Is the industrial grade temperature calibration different in anyway from that of commercial grade ones? For example, is the OCOTP_ANA1 ROOM_COUNT measured at 25C for both parts? 

Note: Even after powering off the entire board, allowing it to cool to room temperature (21C), and then re-running the termperaturemonitor example, the first temperature is about 31C, and reaches 40C within seconds. 

Thanks,

Tim

0 Kudos

2,120 Views
FelipeGarcia
NXP Employee
NXP Employee

Hello Tim,

 

The main difference between Industrial and Consumer product is the temperature operation range. You can check details in datasheet.

 

Have you tried using different board with Industrial RT1064? Just to discard the chip may be damaged. So far, we have not been reported for any errata similar to this.

 

Best regards,

Felipe

0 Kudos

2,120 Views
tmeyer1
Contributor III

Thanks Felipe,

Understood that the Industrial grade part is rated for a larger temperature range. This is one of the key reasons for choosing it.

The same temperaturemonitor produces comparable results on over a 12 boards. I've also ran the example on a earlier board revision which has a different industrial grade RT1064. Both the CVL5A & CVJ5A exhibit the same behavior. 

I captured the OCOTP_ANA1 from the 3 different boards/processors:

Board

COUNT @ 25C

High Temp

Count @ High temp

Rev1 RT1064 (CVJ5A)

1321

105

1179

Rev2 RT1064 (CVL5A)

1337

105

1213

Do you have any suggestions/guesses on why I could be seeing this behavior?

Thanks,

Tim

0 Kudos

2,120 Views
FelipeGarcia
NXP Employee
NXP Employee

Hi Tim,

 

I am going to check this with applications team as I do not have a way to test this myself with industrial device. Any update I will let you know.

 

Best regards,

Felipe

0 Kudos

2,120 Views
FelipeGarcia
NXP Employee
NXP Employee

Hi Tim,

Have you taken a look at the processor's power consumption on your board? I want to make sure there isn't actually an issue that is causing them to draw more current which would also cause the die to heat.

Regards,

Felipe

0 Kudos

2,120 Views
tmeyer1
Contributor III

Hi Felipe, we have not tested this.

Our board doesn't have any shunt resistors exposed (like the EVK does) in order to sample the power. Also, the 3v3 power rail is used for many other ICs on the board, so just measuring our 3v3 wouldn't give us much information. We may be able to measure 1.5V core power and the ADC power though. Would this add value to addressing your concerns?

Thanks,

Tim

0 Kudos

2,120 Views
FelipeGarcia
NXP Employee
NXP Employee

Hi Tim,

 

I would highly recommend to check power consumption properly as I was confirmed by applications team that Industrial processors do not have erratas in TEMPMON either.

 

Best regards,

Felipe

0 Kudos

2,117 Views
tmeyer1
Contributor III

Hi,

The best I could do to sample the power was to remove a reasonable amount of parts that share the 3v3 power rail. In doing this I was able to see by total 3v3 power consumption get down to ~400mW at 600MHz and 165mW in SNVS mode. Upon coming out of SNVS mode, the reported die temp was still above 35C. At this point, the board and RT1064 are cool to the touch ~22C-26C (very close to room temperature).

Also, by completely removing a RT1064 from the board the power consumption dropped about 300mW. 

These numbers seems slightly higher than indicated in the AN12245 app note, which was for a RT1060.

I was hoping that one of the removed ICs which was attached to the processor would have dropped the die temp but that was not the case. So, I don't know that test adds much value, but I figured I would share anyway. 

 

Tim

0 Kudos

2,120 Views
tmeyer1
Contributor III

Hi Felipe,

Thanks for asking the applications team. I will discuss with the team about modding the board to get some sort of power numbers from our MCU. This will be quite difficult, so I suspect we won't be able to get decent measurements if any.  

In the meantime if you have any other suggestions or ideas, I would be open to hear them. 

Thanks,

Tim

0 Kudos

2,120 Views
FelipeGarcia
NXP Employee
NXP Employee

Hi Tim,

 

Have you tried the example using MIMXRT1064-EVK? I tried to reproduce this and the temperature seems to be reasonable using temperaturemonitor SDK 2.8.0 example.

pastedImage_4.png

Best regards,

Felipe

0 Kudos

2,120 Views
tmeyer1
Contributor III

Hi,

Thanks for sharing your results with the RT1064 EVK and 2.8 SDK. 

I have run this test on a RT1060 EVK & SDK 2.7 get very comparable results to what you saw: ~30C.

I just re-ran temperaturemonitor on our custom board with the 2.8 SDK and still see over 50C.

Any thoughts?

Thanks,

Tim

0 Kudos