How to cool SMT power packages and measure its temperature?

[technix]

Let's say I have a SMT power transistor in DPAK. How do I cool it and allow me to take temperature measurement of the heatsink?

[SeanB]

Do it the recommended way the manufacturers use. Take the heatsink and drill a 1mm hole through it to contact the middle of the rear surface of the package, mount the device and use a ( preferably calibrated) welded tip thermocouple and a blob of heatsink compound to measure the temperature. Caution is that the thermocouple will be in electrical contact with the device, so will have the same voltage excursions as the device.

Less accurate is to place the thermocouple on the heatsink next to the device, insulated by 2 thicknesses of Kapton tape from the device and heatsink, and taped down so it is firmly in contact with the side of the device but is still insulated. Will have a constant temperature difference to true temperature of the device, but at least will give you an indication within 5C or so of actual base of device temperature.

[technix]

Do it the recommended way the manufacturers use. Take the heatsink and drill a 1mm hole through it to contact the middle of the rear surface of the package, mount the device and use a ( preferably calibrated) welded tip thermocouple and a blob of heatsink compound to measure the temperature. Caution is that the thermocouple will be in electrical contact with the device, so will have the same voltage excursions as the device.

Less accurate is to place the thermocouple on the heatsink next to the device, insulated by 2 thicknesses of Kapton tape from the device and heatsink, and taped down so it is firmly in contact with the side of the device but is still insulated. Will have a constant temperature difference to true temperature of the device, but at least will give you an indication within 5C or so of actual base of device temperature.

I don’t really need the accurate temperature here, just a rough ballpark reading for implementing an overheat protection. Is there any simpler solution like place a temperature sensor chip somewhere and thermally bond it to the same heat sink as the power MOSFET?

[langwadt]

Do it the recommended way the manufacturers use. Take the heatsink and drill a 1mm hole through it to contact the middle of the rear surface of the package, mount the device and use a ( preferably calibrated) welded tip thermocouple and a blob of heatsink compound to measure the temperature. Caution is that the thermocouple will be in electrical contact with the device, so will have the same voltage excursions as the device.

Less accurate is to place the thermocouple on the heatsink next to the device, insulated by 2 thicknesses of Kapton tape from the device and heatsink, and taped down so it is firmly in contact with the side of the device but is still insulated. Will have a constant temperature difference to true temperature of the device, but at least will give you an indication within 5C or so of actual base of device temperature.

I don’t really need the accurate temperature here, just a rough ballpark reading for implementing an overheat protection. Is there any simpler solution like place a temperature sensor chip somewhere and thermally bond it to the same heat sink as the power MOSFET?

can't use a protected mosfet?

[technix]

Do it the recommended way the manufacturers use. Take the heatsink and drill a 1mm hole through it to contact the middle of the rear surface of the package, mount the device and use a ( preferably calibrated) welded tip thermocouple and a blob of heatsink compound to measure the temperature. Caution is that the thermocouple will be in electrical contact with the device, so will have the same voltage excursions as the device.

Less accurate is to place the thermocouple on the heatsink next to the device, insulated by 2 thicknesses of Kapton tape from the device and heatsink, and taped down so it is firmly in contact with the side of the device but is still insulated. Will have a constant temperature difference to true temperature of the device, but at least will give you an indication within 5C or so of actual base of device temperature.

I don’t really need the accurate temperature here, just a rough ballpark reading for implementing an overheat protection. Is there any simpler solution like place a temperature sensor chip somewhere and thermally bond it to the same heat sink as the power MOSFET?

can't use a protected mosfet?

The device in question has a USB uplink. Sure I can use a protected MOSFET but I also need to let the main MCU know that the thermal protection has tripped. So instead of a protected MOSFET I would prefer a temperature sensor and a normal MOSFET, and use the MCU to implement the thermal protection with uplink. It is cheaper and serves my purpose.

[langwadt]

Do it the recommended way the manufacturers use. Take the heatsink and drill a 1mm hole through it to contact the middle of the rear surface of the package, mount the device and use a ( preferably calibrated) welded tip thermocouple and a blob of heatsink compound to measure the temperature. Caution is that the thermocouple will be in electrical contact with the device, so will have the same voltage excursions as the device.

Less accurate is to place the thermocouple on the heatsink next to the device, insulated by 2 thicknesses of Kapton tape from the device and heatsink, and taped down so it is firmly in contact with the side of the device but is still insulated. Will have a constant temperature difference to true temperature of the device, but at least will give you an indication within 5C or so of actual base of device temperature.

I don’t really need the accurate temperature here, just a rough ballpark reading for implementing an overheat protection. Is there any simpler solution like place a temperature sensor chip somewhere and thermally bond it to the same heat sink as the power MOSFET?

can't use a protected mosfet?

The device in question has a USB uplink. Sure I can use a protected MOSFET but I also need to let the main MCU know that the thermal protection has tripped. So instead of a protected MOSFET I would prefer a temperature sensor and a normal MOSFET, and use the MCU to implement the thermal protection with uplink. It is cheaper and serves my purpose.

there are plenty of cheap switches with temp/current protection and error output, and you can toast a fet faster than an external temperature sensor can react
so calculation power dissipation would probably be a safer option

[technix]

Do it the recommended way the manufacturers use. Take the heatsink and drill a 1mm hole through it to contact the middle of the rear surface of the package, mount the device and use a ( preferably calibrated) welded tip thermocouple and a blob of heatsink compound to measure the temperature. Caution is that the thermocouple will be in electrical contact with the device, so will have the same voltage excursions as the device.

Less accurate is to place the thermocouple on the heatsink next to the device, insulated by 2 thicknesses of Kapton tape from the device and heatsink, and taped down so it is firmly in contact with the side of the device but is still insulated. Will have a constant temperature difference to true temperature of the device, but at least will give you an indication within 5C or so of actual base of device temperature.

I don’t really need the accurate temperature here, just a rough ballpark reading for implementing an overheat protection. Is there any simpler solution like place a temperature sensor chip somewhere and thermally bond it to the same heat sink as the power MOSFET?

can't use a protected mosfet?

The device in question has a USB uplink. Sure I can use a protected MOSFET but I also need to let the main MCU know that the thermal protection has tripped. So instead of a protected MOSFET I would prefer a temperature sensor and a normal MOSFET, and use the MCU to implement the thermal protection with uplink. It is cheaper and serves my purpose.

there are plenty of cheap switches with temp/current protection and error output, and you can toast a fet faster than an external temperature sensor can react
so calculation power dissipation would probably be a safer option

How do I do that in a constant current load? Just rely on the input current and input voltage sense lines?