SRC Forum - Message Replies

Forum: Reliability & Maintainability Questions and Answers

Topic: Reliability & Maintainability Questions and Answers

Topic Posted by: Reliability & Maintainability Forum ( )
Organization: System Reliability Center
Date Posted: Mon Aug 31 12:47:36 US/Eastern 1998

Back to message list Show all replies Topics List About this forum
Original Message:

Posted by: Sheila Prather ( )
Organization:Northrop Grumman Corporation
Date posted: Thu Jan 13 13:51:21 US/Eastern 2000
Subject: 217 Model for GaAs FET for high power, high frequency
MIL-HDBK-217 currently provides the high frequency GaAs FET model in section 6.8, but for that device type operating in the 1GHz range with power dissipations above 0.1W as well as for any other frequency level with high power dissipations, the model renders unrealistic results. As well, the piT factor for the devices addressed in section 6.8 have a great impact on the resulting failure rate, even for those items operating at a power level within the range of the model. These results are unrealistic since these devices typically have rated junction/channel temperatures around 175degC. To compare, the piT values for both the low noise/high frequency (section 6.6) and high power/high frequency (section 6.7) bipolar transistors are extremely smaller in comparison; despite comparable maximum junction/channel ratings. Using these models, as is, the high power/high frequency GaAs FET f.r. is unusable and obviously worst than that of the RF microwave bipolar devices. I furthered reviewed the results published in EPRD97 for the following device types. The published f.r.s for the GaAs devices were lower than those for the bipolar RF transistors for comparable packaging and type; which is totally contradictory to the results if using the model. So, my questions follow; a) what is the recommended model for the high power + high frequency GaAs devices (vs. just high frequency GaAs), and b) what, if any, modifications should be made to the piT values when using the model in section 6.8 where the frequency and power level is within the acceptable range as dictated. Thanks for your response.


Subject: GaAs FET Reliability Models
Reply Posted by: Bruce Dudley ( )
Organization: Reliability Analysis Center
Date Posted: Tue Jan 18 14:07:44 US/Eastern 2000
The reliability model for GaAs FET devices was developed in 1988 by IIT Research Institute for the Rome Air Development Center and was based on failure data collected from the early 1980's to 1988. GaAs FET devices at the time of data collection experienced a lot of failures (3,000 or more in the database) and as result the model reflects that time period. The acceleration factor for temperature also was developed from this failure prone data set and has a relative high activation energy of 0.4 electron volts when compared to bipolar devices at 0.2 to 0.25 electron volts. The activation energy level dictates the temperature multiplying factor as shown in the model which ranges from 1 to 150. This model will not be updated as the sponsoring agencies for MIL-HDBK-217 do not support the effort anymore. I do not recommend changing the temperature factor for the GaAs FET by using the bipolar FET rate as the data bases are different as well device technologies and the failure modes. The model temperature factors can only be changed if new data are introduced, evaluated and regressed. A model that could be exercised in place of this one is the BELLCORE TR-332 GaAs FET model. This model utilizes a lower temperature acceleration factor but still has relatively high base failure rate. Other limitations are the number of environmental conditions available. The TR332 copy has three environments which are: Controlled Ground Fixed, Uncontrolled Ground Fixed and Ground Mobile. This model can be obtained from the Bellcore customer service center (Telephone 1-800-521-2673). The GaAs data that you indicated was from EPRD-97 (Electronic Parts Reliability Data ) is probably for a later time frame and as a result is better than the GaAs FET data used in the 217 model. The use of this newer failure information or the use of recent vendor failure test data could be considered instead of the pessimistic 217 result. The best information is usually the latest data but is often difficult to obtain in timely manner.

Reply to this message