TLV760: Thermal limit

Hi Kai,

We had two boards, each with 30 devices. About 20 on each board died on startup.

Toby

Hi Eric,

What we have seen when testing a single component is that the regulator limits the current to keep the dissipation and die temperature constant at some value above 150°C. Looking at the block diagram in section 7.2 of the datasheet, we had expected this to be true irrespective of the ambient temperature. We thought the current would always be reduced to a level at which the die temperature is constant at the protection level. So indeed our test board has closely spaced TLV760s and heat from adjacent devices will have a significant impact on the temperature of a device.

I have done some additional tests on a single device and found that the current is regulated to a minimum of about 45-50mA at high dissipation. This is also shown in figure 16 of the datasheet. When the dissipation is increased further by increasing the input voltage, the current drops slower than the voltage is increased. So the die gets hotter and hotter until it finally breaks down for temperatures above 200°C.

Because in our appication, we need to consider relatively long periods for the output to be shorted, I think we need a stronger temperature feedback that regulates the current to whatever value required to keep the junction temperature constant.

Thanks for your help, it is much appreciated.

Best regards,

Toby

Hi Toby,

you should not expect too much from this ultra tiny SOT23 package. The ground plane to which the regulator must be connected to, is part of the cooling management. Do you have such a ground plane in your application? The connection to ground plane changes the thermal flow within the chip. This has also influence on the thermal shutdown, because at these huge heats you get considerable temperature gradients, even within the chip. This can also be seen with many other chips: If the recommended cooling (heat sink, ground plane,...) is omitted the chip can overheat or the thermal shutdown can start delayed in spite of having a working thermal shutdown. So, it's extremely important to use the recommended cooling management, which is the mounting on a ground plane here.

It's also important, that the temperature in the direct neighbourhood of regulator is normal, as normal as in a standard application. So, there should not sit other chips with a chip temperature of 170°C. It's very unusual to have this in a standard application. But if you have to expect this, you must improve your cooling management.

You wrote:

"When testing its short-circuit behavior with a 13V input voltage, we noticed temperatures around 170°C on the outside of the package. ... I have tested several devices seperately and that worked well, although indeed there was significant spread on the steady-state temperatures.."

So, it worked as promised and expected! The failures occured only when you carried out your "flamethrower" test. :-) Right?

Kai

Hi Kai,

I had missed your message, hence my late reply.

The problem we have with this IC is that it is not short-circuit proof under all conditions. If you short circuit the device while having a medium supply voltage the thermal protection works and the device becomes at bit warmer than 150°C. However, if the supply voltage is increased, the device becomes hotter and hotter, because the current is not scaled back as fast as the voltage is ramped up. This is also shown in figure 16 of the datasheet. If the voltage is ramped up futher, the device will fail before reaching the specified maximum supply voltage of 30V. This test was done at room temperature. We have products that have an ambient temperature of 80°C, so there isn't much thermal room to play with.

The regulator is used to supply an interface to other products. In case there is a short-cicuit fault in the system, we don't want our product (of many hundreds of dollars) to be returned because a $0,13 LDO failed. We want the interface to be robust. 

Regards,

Toby