TMP468: THERM2 output stuck low until power-cycle

Hello Carl,

If the 3V3_i power is always on, then then the temperature should be constantly on and can constantly receive THERM Status. There is not a setting that would cause the THERM2 output to get stuck low. I do not see an issue with turning off the SCL and SDA pull up if your are not wanting to communicate to the part during board shut down. When leaving the SCL and SDA pullups on, does the temperature measurement get stuck as well? What temperature range are you seeing this issue? You mentioned that some of your boards presented this problem. Were there boards that did not present this issue?

Best Regards,

Meredith McKean

There is one more observation related with Hysteresis register. If the hysteresis is 0 or 1C then it seems to be working for all TMP channels but if the same hysteresis is configured with 7C then it works only channel 4 and if thermal shut down occurs due to other channels then it doesn't works. It seem strange why common parameter behave differently per channel ? As of now seems to be only one board seems to be working with desired hysteresis which is 10C for one channel. Not sure what needs to be done to work with all channels ?

Hello Vinod,

Thank you for providing more details here. All your statements have been correct. Hysteresis of 10C should be able to work for all channels. Can you provide all the contents within all your registers? I can look as see if there is any other issues that could of been missed.  

Best Regards,
Meredith McKean

Hi Meredith,

Thank you for responding. Please find the below details. We are not using channel 2 and 3 it is only connected and has default vaule.

Profile - 28 Register entries - We have 100C shut down for most but only 0x63 is getting triggered and to simulate other channel i have reduced the thermal shutdown limits. Which also displayed the same issue.  

Thanks,

Vinod Gosavi

Hello Vinod,

Thank you for providing this information. I'm not seeing the issue with what was provided in this thread. What application is the TMP461 used for? What are the remote temperature sensors measuring?

Best Regards,

Meredith McKean

Hi Meredith,

It is TMP468 device not TMP461.

It is measuring three Agilex FPGA temp diode, 1 microchip switch, 1 phy and 1 thermal sense diode BC847BFA as recommended in the TMP468 datasheet. 

Not sure why THERM3 output stuck low ? only Hysteresis 0 or 1 seems working fine. anything beyond this number is not working.

Thanks,

Vinod

In our board we are using THERM pin as warning LED and THERM2 pin for shutdown. Therefore limit for THERM2 pin is higher than THERM pin. I don't see any issue with it but why does it work with some hysteresis value ? if circuitry to THERM2 is latching itself then it would latch all hysteresis value right ? 

Hello Vinod,

It is strange behavior that THERM works for 10C hysteresis but THERM2 does not. Have you attempted to keep all voltage supplies on? I'm curious if this would change the outcome of the THERM functionality. There could be noise within the SCL and SDA pins. There shouldn't be much current lost in SCL/SDA when idling high. Keep in mind the input leakage current spec. 

Best Regards,

Meredith McKean

Hi Meredith, Yes i tried to keep all voltages to TMP device On but still observe same results. I think you miss understood none of the pin works with 10C hysteresis but with 0C or 1C hysteresis both pins works fine.

Thanks,

Vinod

Hello Vinod,

Thank you for providing that clarity. I do not see any issues with the information that is presented in this thread. Can you provide I2C screen shots of the communication and provide the error that you see with the THERM pins. 

Best Regards,

Meredith McKean

Hi Meredith,

When i isolate the circuitry from THERM2 pin then i observe TMP device seems to be working fine.

Thanks,

Vinod

Hello Vinod,

What do you mean by isolate the circuitry from THERM2? Did you try and remove SN74LVC1G125YZPR from the THERM pins? How are you measuring failure in the THERM Pins. How are you comparing temperature? By stuck, do you mean the THERM pins are stuck low after 10C hysteresis? Have you reached out to our FAEs? I can set up a call with our FAE to help find a solution to your issue?

Best Regards,

Meredith McKean

Hi Meredith,

It means THERM2 pin is open (not connected to anything). When THERM2 pin is not connected then it works fine with 10C hysteresis. I can measure temperature using our scripts which gives temperature.

When THERM2 pin is connected on power-up - Boards starts fine. As soon as temperature reaches to it shutdown limit then it turn off the board using THERM2 pin signal which disables the 12V efuse (Therefore boards no longer gets power). Since board has no power it cools down within few seconds and now hysteresis should be in play but even temperature cools down. Observed THERM2 pin stuck low. 

THERM2 pin is driving BC846 NPN transistor. Now we will be replacing it with FET and see if that resolves the issue or not. 

I have not yet reach out to your FAE. Please set-up the call with FAE to discuss the issue.

Thanks,

Vinod

Hello Vinod,

I have created an email thread with FAE with this issue. The FAEs are located in Germany so we I need to arrange time to meet with them if needed.

Best Regards,

Meredith McKean

Hi Meredith,

It appears that what may be happening is that when the TMP468 device detects an over-temp condition and shuts the board down, the thermal sense diodes that reside on the die substrate have a different DC offset as a result of the power-loss. This creates a large feedback loop where the measured temp "increases" greatly because the voltage coming from the embedded sense diode has a higher DC bias when the measured device loses power, somehow. So upon power loss, all the carefully measured offsets and calibration settings go out the window. The smaller hysteresis values probably work better because they get triggered before the sense voltage has time to change when the measured part is powered-down. This was verified by watching the sense inputs drift upwards upon power-down. If I force the power-supplies to restart, the temp settings become accurate again and the TMP468 stops asserting THERM2.

Perhaps the only way to resolve this is to use discrete sense diodes/transistors located outside of the device, near enough to be thermally coupled. I'll let you know if I have any further data as we attempt to characterize this further.

Thanks,

Carl Thoma

Hi Carl,

Thank you for sharing this with me. Losing the offsets and calibration setting of TMP468 would imply that the device itself is going through a power cycle. I agree that testing with external diodes is a good approach and will help determine if the embedded sense diode is the issue. Please keep me updated with this issue.

Best Regards,

Meredith McKean

To be clear, the TMP468 never loses power, only the devices being measured lose power, and this corrupts the measurement.

Thanks for the clarity.

Best Regards,

Meredith McKean