MSP430FR5959: ADC Leakage current

"They also need to be able to tolerate 34V through 22k"

I assume you mean *3.4*, not 34 here.

Hi Michael,

Any input voltage beyond VCC + 0.3V is outside of the maximum ratings, is capable of damaging the inputs along with any connected peripherals, and TI cannot guarantee expected operation with these conditions. ADC & REF pad logic can be seen in Figure 6-3 & 6-4 and the REF_A & ADC Block Diagrams are provided in Figure 33-1 & 34-1 of the User's Guide, respectively.

Can the customer provide any more details on there issue, like whether the internal temp readings are ever abnormal before over-voltage is applied? Are there any schematics or code snippets they are concerned about?

Regards,
Ryan

Hi Ryan,

Here is the 34V fault case we need to survive:

The red diode is the ESD diode structure described in figure 3 of the MSP ESD diode specification.

The document describes that over voltage conditions can be applied if the current is limited to less than 2mA.

"Sometimes signals on specific pins exceed the supply of the MSP MCU. In such a case, the device can handle this overvoltage condition through the ESD diodes, but the ESD diode specification must be considered during application design. The items to be considered are described in this document."

Reading through the whole document, I believe every "item to be considered" has been considered.  Let me know if I am missing something.

"To follow the specification, the application must protect the device pin externally so that a signal does not exceed the ±2-mA specification. In other words, voltages greater than the actual device supply (DVCC and AVCC) can be applied, but the current that flows through the ESD diodes must be controlled."

It is my understanding that although this condition is described as safe for the device, there are possible leakage paths in the analog MUX or elsewhere that might cause unintended behavior.  We believe this is manifesting as a internal temperature reading offset of 10degC.  When 3.5V gets applied to one ADC pin, we immediately see the internal temperature jump a fixed 10C.  When the 3.5V drops to 1.5V, we immediately see the internal temperature reading return to expected values.  There is no visible thermal time constant or supply current increase in this scenario, just a fixed step in the value.

First I want to confirm that my understanding is correct: this condition will not damage the device.

Second I want to see what other possible unintended leakage paths or other oddities may arise from such a condition.

Let me know if I can clarify anything further.

Thanks,

Dan

Hi Dietmar,

I guess I shouldn't have included the 5.8V annotation.  Consider the input to be 34V through a 22kohm resistor.  This is 1.54mA.

It sounds like you are confirming that this type of operation is safe if the 2mA spec is adhered to AND the supply can sink current.  This design meets both requirements.

Does the MSP team understand how this type of conduction path can lead to internal temperature sensor reading shifts? There is a very strong correlation between ESD diode conduction and internal temperature reading shifts of 10degC. I do not believe this is a real thermal rise.  1.54mA * 0.7V * 30degC/W (RHA package) is just 0.03C temp rise.

Is there any other possible unintended behavior we should be aware of while conducting through ESD diodes?

Thanks,

Dan

Hi Dan,

sorry for late . Normally it should be safe even if not the most recommended way to operate the part.

There are some critical items you need to consider. It is always important that the DVCC poweres up before you apply this scenario and you shut down this voltage before you shut down the DVCC. Otherwise you supply your device via backdoor which can lead to unpredictable behavior.

The shift in temperatue might come from a shift in th ereference voltage used for ADC conversion. Was this already checked?

Thanks Dietmar.

We haven't tried verifying the internal reference voltage that the temp sensor uses while in this condition.  I'll look into that but might not be able to get the data for some time.  It is more likely we will design this condition out in future revisions.

Last question: is this "backdoor" power state still a concern if AVCC and DVCC are clamped to 0V? When unpowered, the MSP430 rails get clamped to 0V.  The clamp is only released during the power up sequence, shortly before 3.3V is applied.