Risks and Precautions to take care while using On-Chip temperature sensors in Safety critical automotive applications

Ashish Kumar Gupta, Shubhra Singh (Freescale Semiconductors)

Automotive electronics have to sustain harsh conditions like temperature and noisy Environment. The aim these days is to place the electronics as near to engine and transmission control unit as possible so that their noise effect is minimal.

This in turn creates another challenge that electronics have to survive very high temperature as engine and transmission control unit can raise the temperature significantly.

Temperature becomes a first class design constraint in integrated circuits because high junction temperatures adversely affect circuit reliability and degrade the performance.

Designers face new challenge of providing thermal-efficient systems that balance or equally distribute possible on-chip hot spots. In this scenario, Dynamic Temperature Management (DTM) techniques arise as a promising solution. DTM relies on accurately sensing and managing on-chip temperature, both in space and time, by optimally allocating smart temperature sensors in the silicon.

Many on-chip safety features are usually implemented so as to deal with high/low temperature conditions (beyond which SoC’s operation is not reliable).

But at the same time these features should be used very intelligently so that it won’t affect normal functioning of chip or give any false indications regarding out of range temperature.

Before summarizing the issues that can happen if these features are not used correctly, lets look in brief, actual implementation of analog temperature sensor.

On-Chip Temperature sensors utilize a solid-state technique to determine the temperature. They use the fact that, as temperature increases, the voltage across a diode increases at a known rate.

By precisely amplifying the voltage change, it is easy to generate an analog signal that is directly proportional to temperature.

Temperature sensors provide the digital output signals when the junction temperature drifts below the low temperature threshold or above one of two high temperature thresholds.

These signals notify the device to take action to appropriately adjust the device temperature in response to an out of specification low or high temperature operating condition.

These indications are mostly used to reset the Chip as further heating may even burn the chip other than not guaranting the operation.

When calibration/trim values for over & under temperature flags are loaded/Changed there is possibility of Glitches in temperature sensor outputs.

There is also a possibility that before actual trims are loaded the default trim values corresponds to temperature that are within the operational frequency.

These are also effected by PVT conditions so may vary from Chip to Chip also.

Let’s discuss the issues that can occur in case precautions are not taken while integrating the temperature sensors in a SoC.

1. Temperature sensor is enabled by default for the safety intended devices and un-calibrated threshold may lie within the operating temperature range.

There can be issue as the chip can stay in reset or can stuck in reset loop, thus not allowing chip to boot and execute application program.

2. Temperature sensor is enabled when calibration data is changing.

There can be wrong indication for sometime as after trims are changed tsens takes sometime to align to new trim values and the status cannot be guaranteed during that time.

The proposed solution for problem number 1 &2 is to issue “Enable” temperature sensor only after calibration is done. This way safety requirement is also honored as the temperature sensor is getting enabled before the device gets out of RESET.

3. Temperature sensor fails to operate completely.

In case temperature sensor fails to operate completely (false indication or no indication). The Situation can be fatal as there may be possibility of chip being stopped unnecessarily or continues operation that can be unreliable.

To solve the issue temperature sensors output voltage should be continuously monitored by an ADC or redundancy can be introduced (i.e use of 2 temperature sensors) and better calibration can be done by comparing outputs of 2 temperature sensors.

4. Temperature Sensor Placement not correct in an SoC.

If not properly placed in SoC there can be the case in which some portion of chip actually at out of specification temperature but still chip continues operation that can can be unreliable.

To solve the issue temperature sensor should be placed closer to the region where the activity is expected to be maximum and during place and route care must be taken to make sure that there won’t be local hot spots exist in SoC.

Scenario 1. During power-up

Fig 1. Glitch in Tsens output during power-up

PS: The glitch in the undertemp and overtemp flag is absorbed before reset is enabled and chip comes out of reset

Over/Under temperature flag set during trim loading at power-up

Scenario 2: After chip out of reset and user programs trims through software

Fig 2. Glitch in Tsens output on destructive reset during RUN mode

PS: functional reset glitch on tsens boundary leads to destructive reset. Also after destructive reset, the power up scenario follows and hence chip again comes out of reset.

Fig 3. Flow chart to decide sensor location

Abbreviations:-