View AN565_460125.PDF datasheet online --- IC-ON-LINE

Datasheet File OCR Text:

AN565 APPLICATION NOTE
MEDIAN-TIME-TO-FAILURE (MTF) OF AN L-BAND POWER TRANSISTOR UNDER RF CONDITIONS
W. E. Poole - L. G. Walshak
1. INTRODUCTION.1 At last year's Symposium (1973), two papers presented preliminary data from the first known comprehensive life-tests on any microwave power transistor operated under RF conditions. This paper will report the latest findings of RF life-testing on two 30-watt L-band power transistors reported by the author last year. The preliminary purpose of this work is to determine the median-time-to-failure (MTF) of the MSC-1330/A and a higher efficiency version known as the MSC-1330/B, under the actual RF conditions used in the several solid-state phased array radar systems under development which use these devices. A second objective is to identify the failure mode(s) and failure mechanism(s) with the goal of improving device reliability by possible design modifications. With the field history of microwave power transistors less than 10 years, the true long-term reliability of this class of devices is unknown. In the meantime, reliability predictions must be made on the basis of accelerated life-tests. In the past, however, accelerated life-tests have been run only under DC operating or high temperature reverse bias conditions. Unfortunately, the relationship between these DC tests and the real-world of RF operation has never been determined -- if indeed it even exists. 2. TEST PROGRAM. The MTF of the two L-band test devices are measured under two pulse conditions and under worse case system conditions: Frequency = 1.4GHz Pulse width / Duty Factor = 120s/30% 1500s/15% Collector Voltage = 28VDC To determine the MTF for each pulse condition within a reasonable length of time, life-tests were run under the RF conditions using elevated device junction temperature (Tj) as the accelerating factor. The test matrix consists of a total of nine (9) separate tests. For the MSC-1330/A device, the MTF was measured under two (2) pulse conditions and at three (3) junction temperatures each. The MTF of the MSC-1330/B was determined for one (1) pulse condition at three (3) junction temperatures.
1.
This paper was presented at the 12th Annual IEEE Reliability Physics Symposium April 1974
March 2001
1/16
AN565 - APPLICATION NOTE Specifically, the test matrix used is as in table 1. Table 1: Test Condition Matrix TEST T1 T2 T3 T4 T5 T6 T7 T8 T9 DEVICE MSC-1330/A MSC-1330/A MSC-1330/A MSC-1330/A MSC-1330/A MSC-1330/A MSC-1330/B MSC-1330/B MSC-1330/B RF PULSE 120s/30% 120s/30% 120s/30% 1500s/15% 1500s/15% 1500s/15% 120s/30% 120s/30% 120s/30% JUNCTION TEMPERATURE 340C 280C 250C 340C 280C 250C 340C 280C 250C
Thus, for each pulse condition and each device type, the measured MTF at three (3) values of Tj can then be plotted as a log-Normal function of reciprocal temperature, 1/Tj(K). The straight line drawn through these points (the so-called Arrhenius curve) can then be extrapolated to the lower junction temperatures found under actual operating RF conditions and thus, predict the MTF at normal Tj. Since normal operating Tj falls into the ranges 75-100C, extrapolation down from the 250-340C range used in these tests involves definite risks -- namely, the possibility that failure mechanism(s) and failure rates produced at elevated Tj may not be the same as those at normal Tj. On the other hand, the elevated Tj tests used in this matrix has given first order MTF data in a minimum length of time and early enough to be useful in system planning and design. Each test consists of from 8-11 devices for a total sample size of 81. Although a larger sample is obviously desirable, the cost and complexity of RF testing forces this compromise. It is for this very reason, in fact, that extensive, RF life-testing has not been attempted in the past. However, this sample size should be adequate for at least a first order MTF approximation. The mechanics of setting-up the various accelerated tests at elevated Tj are relatively complex. For example, before any one of the tests can be run, some reliable means of forcing each device to have the desired maximum Tj value must be found. In the specific case of T1, external heating of the device must be used to artificially raise Tj from its normal value of ~100C to 340C. To do this, the specially designed RF life-test racks incorporate a heated block beneath each test amplifier. The entire test circuit is heated up to a maximum of 175C and device dissipation must then raise Tj to the desired value of T1=340C. The actual determination of device junction temperature can not be left to calculated values based on nominal thermal resistance values. Instead, each device is tuned-up on the bench under the exact test conditions in its own particular test amplifier and Tj is measured directly using an Infrared microscope. After suitable calibration, the IR microscope can give instantaneous Tj as a function of time during the RF pulse. Thus, each device is tuned-up and the peak Tj during the pulse is set equal to the desired value of Tj (i.e. T1=340C, T2=280C, etc.) by minor circuit adjustment. In this way, each device test is accurately known to be operating at the desired Tj value. Without such a direct measurement, the variation of Tj between devices would seriously affect the data since MTF is a strong function of Tj.
2/16
AN565 - APPLICATION NOTE Following IR scanning, each device can then be hermetically sealed and proceed to process conditioning to weed-out infant mortalities. This condition consists of: - Visual pellet inspection; - High temperature storage (non-operating) 200C (168 hours); - Temperature cycling -65C to +200C (10 cycles); - Hermeticity test, Fine and Gross (10-8cc/sec); - RF burn-in (pulsed) Tj=175C 1500s/15% duty (168 hours); - DC test. 3. TEST RESULTS. At this time, seven (7) of the nine (9) programmed tests are complete with the remaining two (2) tests (T4 and T6) in progress. Analysis of this data allows the following conclusions to be drawn: 1) The extrapolated MTF of the MSC-1330/A and MSC-1330/B is of the order of 106 hours or 100 years at normal operating Tj under pulsed RF conditions with duty factors approximately 30%. 2) The first order MTF of these devices is inversely related to the RF pulse duty factor i.e. MTF is inversely related to the time that the device is "on". MTF (DF)-1. 3) The MTF is a function of pulse width -- the MTF decreasing with increasing pulse width. Over the range 120-2000s, this decrease is typically 50%. 4) MTF is also a weak function of the total temperature rise during the pulse, Tj. For typical values of
Tj, MTF can vary by 10-40%.
These results are based upon analysis of the following test data: 3.1 Short Pulse Tests (120s/30%). Short pulse test on the MSC-1330/A (T1, T2, T3), each with a sample size of eight (8) devices, were run with elevated Tj=340C, 280C, and 250C respectively. In each test, devices were run until "failure". Failure was arbitrarily defined as the point at which output power dropped to one-half its original value (3dB). To determine the MTF for each Tj value, the failure distribution was plotted, as in the example of T1 and T2 in figure 1, and the MTF value read-off as the time corresponding to 50% cumulative failure. Note also that the straight line relationship between time-to-failure and cumulative percent failure implies a logNormal failure distribution, characteristic of wear-out mechanisms. Along with similar log-Normal plot for test T3, the measured MTF of the three (3) short pulse tests on the MSC-1330/A devices are shown in table 2. Table 2: MTF Results of the 3 Short Pulse Test on MSC-1330/A Devices TEST T1 T2 T3 DEVICE MSC-1330/A MSC-1330/A MSC-1330/A RF PULSE 120s/30% 120s/30% 120s/30% Tj 340C 280C 250C MTF (-3dB) 72 hrs 480 hrs 660 hrs
3/16
AN565 - APPLICATION NOTE Figure 1: Failure Distribution: Test T1 and T2
1004 9 8 7 6 5 4 3 2
TIME-TO-FAILURE (HRS)
1003 9 8 7 6 5 4 3 2
T2=280C
MTF = 480 HRS
= 0.615
100 9 8 7 6 5 4 3 2 10 2%
T1=340C
MTF = 72 HRS
= 0.65
5
10 15 20 30 40 50 60 70 80 85 90
95 98%
CUMMULATIVE PERCENT FAILURE
4/16
AN565 - APPLICATION NOTE Having thus determined the MTF at three (3) elevated Tj values, MTF is then plotted as a function of 1/Tj (K) on log-Normal coordinates as shown in figure 2. Note temperature scale in C. Figure 2: Median-Time-To-Failure vs. Peak Junction Temperature
106
107
105
106
MTF MEDIAN-TIME-TO-FAILURE (HRS)
104
105
LONG PULSE 1500s/15%
SHORT PULSE 120s/30%
103
102
CONDITIONS FREQ=1.4GHz VCC=28V Sample Size=63 TOTAL =MSC-1330/A =MSC-1330/B
400 300 200 100
10
Tj (C)
5/16
MTF (HOURS)
AN565 - APPLICATION NOTE Similar tests on the MSC-1330/B device were also conducted under the same 120s/30% pulse conditions and Tj. These three (3) tests (T7, T8, T9) also resulted in log-Normal failure distributions with measured MTF shown in table 3. Table 3: MTF Results of the 3 Short Pulse Test on MSC-1330/B Devices TEST T7 T8 T9 DEVICE MSC-1330/B MSC-1330/B MSC-1330/B RF PULSE 120s/30% 120s/30% 120s/30% Tj 340C 280C 250C MTF (-3dB) 53 hrs 450 hrs 820 hrs
Plotting these three (3) data points on the MTF vs. 1/Tj curve figure 2, we find that these points essentially coincide with those found for the MSC-1330/A. So, to a first approximation a single straightline can be drawn through all six (6) points. Note, however, that although the MTF vs. 1/Tj curve for the MSC-1330/B is essentially identical to that of the MSC-1330/A, the reliability of the MSC-1330/B device is actually 4 times higher at normal operating conditions. This results from the fact that the maximum value of Tj for the MSC-1330/B is typically 20 lower than the MSC-1330/A as a result of its higher collector efficiency. With this set of data, we can now make first order predictions of the MTF of these devices at normal operating condition where Tj75-100C (depending on duty factor, efficiency, heatsink temperature, etc.). Referring to figure 2, we find that at Tj75-100C the MTF under 120s/30% duty is of the order of 106 hours or 100 years. Of course, extrapolation of this sort involves definite risks as mentioned earlier. However, at this point there is simply no practical alternative for generating this kind of essential data short of putting the devices in the field and waiting a few hundred years. [To the best knowledge of the authors, the longest field application of microwave transistors in sizeable quantities are the approximately 900 MSC-3003 -- 3W/3GHz devices -- which have run without device failure under conditions CW operating for greater than three (3) years (24 million unit-hours). An additional 3,000 of these devices have also run greater than two (2) years without failure (54 million unit-hours)]. 3.2 Long Pulse Tests (1500s/15%). Three (3) long pulse tests (T4, T5, T6) using the MSC-1330/A were included in the original test matrix to evaluate the effect of pulse width on device MTF. To give direct comparison with the short pulse tests, T4, T5 and T6 are run at the same time Tj as the three (3) short pulse test series (T1, T2, T3) i.e. Tj=340C, 280C and 250C respectively. At this point, the 280C test, T5, is complete with T4 and T6 in progress. The log-Normal failure distribution of T5 gave a measured MTF of 1100 hours. By comparison, test T2 with the same device at the same Tj (280C) but under short pulse operation had a MTF of only 480 hours. At first thought this roughly two (2) times higher MTF for the long pulse case appears to be contrary to normal expectation. However, further thought will recall that the pulse duty factor for the long pulse was only 15%, while for the short pulse duty factor was 30%. Or, in other words, although the device sees greater stress during the long pulse, the total stress for the short pulse case is greater simply because the device is "on" twice as long as the long pulse case. This follows directly from the 30% duty factor for the short pulse vs. the 15% duty factor for the long pulse case.
6/16
AN565 - APPLICATION NOTE This argument can be seen more quantitatively from the following analysis of the actual time spent at or near the peak Tj during both pulse conditions. Figure 3a and 3b show the measured transient Tj vs. time during both the long and short pulse peak Tj of 280C in both cases. Assuming the device failure mechanism is a relatively strong function of temperature, its rate of reaction at the higher values of Tj during the pulse must be significantly greater than the rate at lower Tj. For example, referring to the MTF vs. 1/Tj data of figure 2, we see that the difference in MTF between Tj = 280C and Tj = 260C is about a factor of 2. Thus, to a first approximation, assume that the net effect during the time 260C Tj 280C is significantly greater than the net effect during the time Tj 260C. Figure 3: Peak Junction Temperature vs. Time
280C
280C
PEAK JUNCTION TEMP.
PEAK JUNCTION TEMP.
260C
260C
Tj
RF PULSE
0
0.5
1.0
1.5
0
50
100
150
TIME (ms) a) T5 - LONG PULSE (1500s / 15%)
TIME (s) b) T2 -SHORT PULSE (120s / 30%)
As shown in figure 3a and 3b 260C Tj 280C for 646s during each long (1500s) pulse and 54.4s during each short (120s) pulse. Since the given duty factor is 15% for the long pulse and 30% for the short pulse case, there are 100 long pulses/second and 2500 short pulses/second respectively. Consequently, 260C Tj 280C only 6.46% of the time under long pulse and 13.62% of the time for short pulse.Thus, the ratio of the times during which 260C Tj 280C is 13.62 / 6.46 = 2.1. Or, the time during which "significant stress" is seen by the device during the short pulse case is 2.1 times longer than that experienced under the long pulse case. The ratio is within 10% of the ratio of actual measured MTF for these two cases: MTF(T5) long pulse 1100hrs = 2.3 = MTF(T2) short pulse 480hrs Similar agreement is seen with the assumption of larger and smaller values of the significant Tj range as seen in table 4. Table 4: Short/Long Pulse MTF Ratio for Different Tj Ranges Time Within Tj Range Per pulse Long Pulse 373s 537s 646s 740s Short Pulse 34.5s 45.0s 54.5s 65.5s % Time Within Tj Range A Long Pulse 3.73% 5.37% 6.46% 7.40% B Short Pulse 8.62% 11.20% 13.62% 16.37%
Tj Range
Ratio B/A
% Difference from Measured MTF Ratio 1.3 12 11 6
7/16
270-280C 265-280C 260-280C 255-280C
2.31 2.09 2.11 2.21
AN565 - APPLICATION NOTE 4. MTF CALCULATIONS. As stated earlier, the primary objective of this work is to determine the MTF of the MSC-1330/A and MSC-1330/B transistors under actual RF operating conditions. If these conditions are at either the 120s/30% or 1500s/15% pulse conditions used in these life-tests, MTF calculations for normal operations reduce simply to extrapolation of the MTF vs. 1/Tj curve generated at elevated Tj values down to the Tj value under normal conditions. However, in general, it would be preferable to be able to extrapolate the measured MTF data to any pulse duty factor or pulse width. With this goal in mind, a separate series of tests were run to determine the relationship between MTF, duty factor and pulse width. As a result of these tests and the original series of nine (9) tests, the following parameters (shown in table 5) were found to be most significant in determining MTF (in decreasing order of importance). Table 5: Major Parameters Effecting MTF PARAMETER 1. Junction temperature 2. Duty factor 3. Pulse width 4. Tj during pulse RELATIONSHIP MTF Exp (O / kTj) MTF 1/DF MTF 1/PW MTF Tj < 200X < 30X < 2X < 1.4X TYPICAL MAGNITUDE
The specific relationship of each of the above parameters is as follows: 4.1 MTF vs. Tj. Test T1, T2, T3 along with T7, T8, T9 give MTF vs. 1/Tj for the MSC-1330/A and MSC-1330/B by measuring MTF at three (3) different values of Tj for a given pulse width and duty. Thus, to predict the MTF of these devices at any other value of Tj for the same pulse width and duty, simply project the straight line drawn between the data down to the desired Tj as shown in figure 2 and read-off the corresponding MTF from the graph. Mathematically, this relationship can be expressed in the form: MTF = C exp (O / kTj) where MTF = median-time-to-failure (hrs) C = constant (empirically derived) O = activation energy for the mechanism(s) (eV) k = Boltzman's constant Tj = junction temperature (K) In this case: MTF = 2.425 x 10-5 (0.7875/kTj) 4.2 MTF vs. Pulse Duty Factor. It was shown earlier that MTF was inversely proportional to the device "on" time. Similarly, it has also been shown experimentally that MTF is inversely proportional to pulse duty factor. In separate tests, a series of devices were life tested with the various duty factors but with Tj and pulse width constant (Tj=360C, PW=120s). As shown in figure 4, the measured MTF of these devices plotted as a function of 1/DF shows a straightline relationship. Normalizing the measured MTF to that under 30% duty:
30 Normalized M TF = -------
DF (%)
8/16
AN565 - APPLICATION NOTE Thus, the MTF of a device operated at 15% duty has approximately two times the MTF of one operated at 30% duty (all other parameters being equal) simply because the "on" time is one-half. Figure 4: Median-Time-To-Failure vs. Pulse Duty Factor
1.6
50 1.4
1.2 40
NORMALIZED MTF
1.0
0.8
MEASURED THEORETICAL
0.6 20
NORMALIZED MTF = 30 / Duty Factor (%)
0.4 10 0.2
100 / DUTY FACTOR (%)
0 2 4 6 8 10 0
DUTY FACTOR (%)
100 50 30 20
MTF (HRS)
30
9/16
AN565 - APPLICATION NOTE 4.3 MTF vs. Pulse Width. To fully evaluate the MTF vs. pulse width relationship, one approach would be to place a minimal sample of devices on life test at various pulse widths and determine their MTF as done in the original nine-part test matrix. However, a simpler and considerably less time consuming approach was used which is based upon the experimentally confirmed relationship between MTF and device "on" time as illustrated in figures 3a and 3b as well as in the MTF vs. duty factor experiments. A series of devices were run at six different pulse widths between 120s and 2000s with Tj and Tj held constant. Note, Tj=280C (peak) and Tj=115C where Tj is defined as the total junction temperature rise during the RF pulse. At each pulse width, the transient Tj during the pulse was carefully measured using the IR microscope. Then, using the same technique used in explaining the two times higher MTF of the long pulse test, T5, the "on" time or time within an arbitrary 20C of Tj (peak) for each pulse width is computed. Let this "on" time within the pulse be expressed as some fraction of the pulse width:
--------t = ---- pulse width
"on" time during pulse
Since the number of pulses per unit time is DF/PW, the total "on" time per second is:
DF t ( TOTA L ) = -------- ( t ) ( P W ) = DF ( t ) PW
For any given pulse width, the MTF is inversely proportional to the total "on" time:
1 ----M TF ------------------ --t ( TOTAL )
Consequently, the ratio of MTF between any given pulse width PW1 and PW2 must be given by:
( MTF )1 t 2 ( TOTAL ) ( DF ) 2 x t 2 -------------------------- = --------------- -------- = -------- ------------t 1 ( TOTAL ) ( DF ) 1 x t 1 ( MTF )2 ( MTF ) t2 --------- ----1 = ------- ( MTF )2 t1
and for the case of constant duty factor (DF1=DF2):
That is, at constant duty factor, the ratio of MTF between any two pulse widths is simply the ratio of their fractional "on" times. Thus, the relative MTF between the six different pulse widths of this experiment can be calculated by measuring the fraction of the time within each pulse when 260C Tj 280C and computing the ratio given above. Doing so for pulse widths of 120s, 250s, 500s, 1000s, 1500s, and 2000s, the following values are found in table 6. Table 6: MTF vs. Pulse Width Pulse width (s) 120 250 500 1000 1500 2000 "ON" time per pulse (s) 46 98 204 465 732 1155 t (%) 38.3 39.2 40.8 46.5 48.8 57.8 Normalized MTF 1.00 0.98 0.94 0.82 0.79 0.66
Note: Duty factor = 10% and Tj = 115C
10/16
AN565 - APPLICATION NOTE This relationship is also presented graphically in figure 5 along with parametric values of Tj. The relationship of MTF vs. Tj will be discussed in the following section. In any case, we note from figure 5 that as pulse width increases, normalized MTF decreases (all other parameters being equal). Although not as significant a factor as duty factor, variations in pulse width over the 120s -- 2ms range can result in a 2:1 change in MTF. Figure 5: Relative MTF vs. Pulse Width for Different Tj During Pulse
1.0
RELATIVE MTF
0.8 0.7 0.6 0.5 0.4 0.2 0 500 1000 1500
DATA FOR Tj = 115C ALL CURVES RELATIVE TO PULSEWIDTH = 120s AT GIVEN Tj
120 115 110 100 90 80 70 60
2000
PULSE WIDTH (SEC)
Tj (C)
11/16
0.9
AN565 - APPLICATION NOTE 4.4 MTF vs.
Tj.
The fourth and least significant parameter affecting MTF has been found to be the junction temperature rise during the pulse, Tj. The relative magnitude of this term has been measured at two pulse widths
(120s & 1500s) over the Tj range 60C -- 200C. Again, the technique used to determine relative MTF was to measure the "on" time for each case from the measure transient Tj response during the pusle (using an IR microscope). Normalizing to the MTF at Tj=65C for the MSC-1330/A and Tj=95C for the MSC-1330/B, the relative MTF vs. Tj at 120s pulse width is plotted in figure 6. As seen in this plot, the relative MTF varies by less than 30% -- 40% over the complete Tj range. Using the MTF vs.
Tj data at both 120s and 1500s, the effect of Tj was included in the MTF vs. Pulse Width curve of
figure 5. Figure 6: Relative MTF vs. Tj During Pulse (120s)
1.5
1.4
1.3
1330/A
1.2
1.1
RELATIVE MTF
1330/B
1.0
0.9
0.8
0.7
0.6
0.5 20 40 60 80 100 120 140 160
Tj = Tj MAX - Tj MIN (C)
12/16
AN565 - APPLICATION NOTE 5. SAMPLE MTF CALCULATION. The above relationships can be used to extrapolate the MTF data for the MSC-1330/A and MSC-1330/B measured under 120s/30% and 1500s/15% pulse conditions to any other pulse width and duty factor. Such calculations could be used by system designers to: 1. Make reliability predictions for any given pulse conditions. 2. Evaluate RF performance/reliability trade-offs. The variation of device reliability with pulse duty factor and pulse width has not been generally considered in system reliability studies up to now. To illustrate such an extrapolation the following sample calculation will compute the MTF of a MSC-1330/ B operating under typical L-band conditions. 1. Given: Pulse width (PW) = 500ms Duty factor (DF) = 15% RF output (Po) = 30W (peak) RF input (Pin) = 5W (peak) Efficiency (Eff) = 60% Peak thermal resistance (jc) = 2.6C/W Case/flange temperature (Tc) = 25C 2. First determine Tj (max). Measure this directly using IR techniques or calculate:
Tj ( max ) = Tc + Tj = Tc + P diss jc
= 25 + 25 2.6 = 25 + 65 = 90C
3. Referring to the MTF vs. 1/Tj curve of figure 2, the extrapolated MTF of the MSC-1330/B at Tj=90C~4x106 hours. This data, however, is under the 120s/30% duty life-test conditions with Tj=95C. Thus, this MTF value must be corrected to account for the different pulse duty factor, pulse
width, and Tj of the given application. Making use of the empirically derived curves described earlier, a scale factor for each of the three term can be determined. Then, the MTF of the given condition relative to the life-test conditions can be computed as follows: pulse MTF (500s / 15% / 65C) DF (120s / 30%) = x [ width ] x [Tj Factor] MTF (120s / 30% / 95C) DF (500s / 15%) factor
[
]
Where the 65C and 95C notation refers to the value of Tj. Computing each term individually: 4.
DF (120s / 30%) 30 = DF (500s / 15%) 15
5. Pulse width factor
=2
MTF ( 500s 65C ) --------- --------------- = 0.88 --= --- -------------- ----MTF ( 120s 65C )
(from figure 5) at
Tj=65C
13/16
AN565 - APPLICATION NOTE 6. Tj Factor
M TF ( 120s 65C ) -------------------- --------------------- = 0.89 ---= ----M TF ( 120s 95C )
(from Figure 6)
7. Thus,
=
MTF (500s / 15% / 65C) = (2) (0.88) (0.89) = 1.6 MTF (120s / 30% / 95C)
8. Therefore, the MTF under the actual given conditions at a value based on life-test conditions and given by figure 2. Or:
Tj=65C
is 1.6 times the extrapolated
MTF (500s / 15% / 65C) = (1.6) (4 x 106) hrs MTF (500s / 15% / 65C) = 6.4 x 106 hrs (700 years)
As this sample calculation clearly shows, the reliability of the transistor under pulsed RF is a direct function of many non-device related parameters. As a result, the system designer now has the tools available to evaluate possible design trade-offs between RF pulse conditions and device reliability. Such trade-offs have not been generally considered until now. 6. FAILURE ANALYSIS. Detailed failure has shown one dominant failure mode for both long and short pulse tests -- degradation or shorting of the emitter-base junction. A possible model to explain this observation has been developed and described in detail in last year's paper. Consequently, the following will only summarize the basic features of the model. Figure 7: Cross-sectional View of Base and Emitter Contact Structure
Pt Si
BASE FINGER ALUMINUM
Si3N4 SiO2 EMITTER
EMITTER
X
Pt Si
X
BASE
A. BASE CONTACT EMITTER FINGER ALUMINUM
Si3N4 SiO2
EMITTER
BASE
X
X
B. EMITTER CONTACT
14/16
AN565 - APPLICATION NOTE The model developed is based upon scanning electron microscope (SEM), Energy Dispersive X-Ray Analysis (EDAX), and optical analysis of each of the many device dielectric and metal layers. The model explains the localized shorting of the emitter-base junction as the result of the reaction between the aluminum metallization and the silicon dioxide passivating layer in the contact openings as shown in the device cross-sectional view of figure 7. This aluminum -- SiO2 reaction is accelerated with temperature. As this reaction proceeds, the SiO2 in contact with the aluminum emitter metal makes a low resistance contact to the base doped area of the device. The low resistance leakage path thus formed, causes a localized current which raises the local temperature which, in turn, reduces the path resistance further in the classic run-away sequence events. As a larger and larger fraction of the device emitter periphery is effectively short-circuited, RF output must decrease until it is only supplying a fraction of its original output power. This proposed failure mechanism is not new. In fact, such emitter base degradation in the early days of UHF transistors was in large responsible for the development of improved device geometries such as the overlay and matrix designs as well as improved dielectric isolation techniques. In contrast to the matrix structure used in the devices in this test program, early devices used interdigitated geometries with "washed emitters" which gave only about 1/10 the junction protection found in the matrix. In a matrix design, the distance, x, (see figure 7) from the edge of the contact or the metal to the emitter-base junction can be made about 10 times greater than that possible with "washed emitters" where the emitter is diffused through the same contact opening. In that case, the protection, x, is equal only to the amount of lateral emitter diffusion which can be less than 1,000 angstroms in which case, failure often occurs in a matter of minutes at 400C. Thus, the 106 hours (100 years) measured MTF of the MSC-1330 is a direct result of the improved junction protection made possible by the matrix geometry and improved processing. Similar failure mechanisms exist with other metal systems. Gold, for example, immediately forms an eutectic solution with silicon at approximately 400C. Consequently, gold metallization must have barrier layers (such as tungsten) between it and the silicon service. Thus, device reliability is determined primary by the integrity of the barrier layer. Preliminary reliability data on gold systems are now being generated particularly at very high temperatures (>300C). Meaningful comparison, however, will have to wait for the results of longer-term tests at lower temperatures. For example, accelerated life-tests by Pitetti of Bell Laboratories have shown that although gold appears to have improved reliability at very high temperatures, the margin of improvement decreased at lower temperatures and actually reached a cross-over point at about 150C. 7. CONCLUSION. Data from the first known comprehensive life-test of a microwave power transistor under RF conditions allows the following conclusions to be drawn: 1. MTF of the MSC-1330/A and the MSC-1330/B L-band devices is of the order of 106 hours of 100 years under 120s/30% duty, pulsed RF conditions; 2. MTF is an exponential function of temperature, inverse function of pulse duty factor and pulse width, and a relatively weak function of temperature rise during the pulse; 3. Detailed failure analysis has shown one dominant failure mode -- degradation of the emitter-base junction; 4. Probable failure mechanism is the dissolution of the SiO2 dielectric layer by the aluminum metallization in the vicinity of the metal contact openings.
15/16
AN565 - APPLICATION NOTE AKNOWLEDGEMENT. This work was partially supported by the Naval Research Laboratory and U.S. Marine Corp. under Contract No. N00014-17-C-0472 under the technical direction of Mr. Brooks Dodson. The contributions of Messrs. B. Dodson, W. Weisenberger of NRL, and P. Kalfus, W. Klatskin, and F. Stallone of Microwave Semiconductor Corp. are gratefully acknowledged. REFERENCES. 1. W.E. Poole, "Median-Time-T o-Failure (MTF) of Microwave Power Transistors Under RF Conditions," 1973 IEEE Reliability Physics Symposium, P.287. 2. K. Fisher, "Pulsed RF Life of an L-Band Power Transistor," 1973 IEEE Reliability Physics Symposium, P.293. 3. R.C. Pitetti, "Electromigration of Ti/Pd/Au Conductors," 1972 IEEE Reliability Physics Symposium, P.171.
Information furnished is believed to be accurate and reliable. However, STMicroelectronics assumes no responsibility for the consequences of use of such information nor for any infringement of patents or other rights of third parties which may result from its use. No license is granted by implication or otherwise under any patent or patent rights of STMicroelectronics. Specification mentioned in this publication are subject to change without notice. This publication supersedes and replaces all information previously supplied. STMicroelectronics products are not authorized for use as critical components in life support devices or systems without express written approval of STMicroelectronics.
The ST logo is a registered trademark of STMicroelectronics (c) 2001 STMicroelectronics - Printed in Italy - All rights reserved STMicroelectronics GROUP OF COMPANIES Australia - Brazil - China - Finland - France - Germany - Hong Kong - India - Italy - Japan - Malaysia - Malta - Morocco Singapore - Spain - Sweden - Switzerland - United Kingdom - U.S.A. http://w ww.st.com
16/16

▲Up To Search▲

Price & Availability of AN565

	To Download AN565 Datasheet File
If you can't view the Datasheet, Please click here to try to view without PDF Reader .