How to determine the reliability of FETs?

[ooiben]

Good day,
I have an amplifier that has a large current swing. At P1dB, the current swing is 100% of its original value. How do I determine the reliability of this part? The information I have at hand is the activation energy and lifetest of a smaller FET from the same technology. Can these information be used? i.e. will the activation energy and lifetest change with FET size and same power density through the FET?

Thanks.

 

[jiripolivka]

In all semiconductor devices, the reliability defined by MTBF is mainly determined by junction temperature. Component manufacturers publish their data on MTBF.
If not available by manufacturers, an user would have to run an accelerated test on a big number of components of interest to get the data.
Driving a semiconductor component close to breakdown like in the case you mention can present a grave effect on MTBF, too.

 

[biff44]

The best way I have seen is you make up a number of amplifiers, power them up (DC and RF), put them in an oven to run at an elevated junction temp, like 120 deg C, and pull them out every week and measure the gain and output power. If you get a few failures, you can then plot statistically the mtbf.

 

[BigBoss]

Normally reliability tests are done by observing the device for a long term and some critical parameters are re-measured after this observation period.
In fact there are tenths of device coming from the same wafer and they are re-measured and MTBF figure of merit of reliability is calculated.
You can not do that by yourself but you may take a thermal photograph of the die if you are able to..
It will give you a perspective..

 

[Orson Cart]

Ideally first calculate the average die temp based on the heaksinking used and the power into (or through) the fet, if the average die temp (accurately calculated - and based on tab measurements) is much above 120 degC then you are on the knee point of the reliability curve for your average commercial mosfet. Next calculate the peak die temp based on the lowest freq of operation and the thermal characteristics of the die/package, if the peak die temp is much above 140C or the temp amplitude is much bigger than +/- 25 degC at low freqs then again you are entering territory where peak temp and/or thermal cycling is going to impact on a normal commercial mosfet.
Regards, Orson Cart.