The load life characteristics of thick-film resistors

M/crodectron/e: and Rd/ab///ty

Pergamon Press 1973. Vol. 12, pp. 61-71.

Printed in Great Britain

THE LOAD LIFE CHARACTERISTICS OF THICK-FILM RESISTORS P. J. HOLMZS

Royal Aircraft Eatablhhmant, Farnborough, Hanta, England ~mResulta of load life testa on thick-film remtora in different laboratories have ahown a variety of resimznce drifts for identical materlah under identical power loadings per unit area. Th~ paper reporta some teat results which indicate the factors which must be specified if the information obtained on laboratory teat samples is to be used to predict the stability of resistors in operational thick-film circuits. Principal among these are the voltage gradient end the actual working temperature of the resistor: the latter is compounded from the temperature rise due to Joule heating in the resistor and the background temperature provided by the ambient and any heat sources (resistors and active devices) on the sub,rate. 1. INTRODUCTION EXTSNSmN of the use of thick films into new areas has focused attention on their capability of operating under heavy electrical power or voltage loading, and the aim of this paper is to consider the implications of some load life tests conducted in our own laboratories and under contract in industrial laboratories. Comparison of the results leads to a recognition of the deficiencies of conventional load life testing as applied t o thick-film resistors, and it becomes clear that a uniform method of specifying operating conditions is required if the results of laboratory tests are to be applied directly to real-life operation.

(c) Especially in high resistivity materials, high voltage gradients may exist under d.c. conditions, causing temporary or permanent electrical breakdown effects in the less conducting parts of the matrix. (d) In circuits containing a number of resistors or active components, the thermal conductivity of the substrate leads to the operating temperature of each resistor being dependent on the heat dissipated in its neighbours. This is in addition to any background temperature provided by the ambient conditions. These effects are all superimposed, so that it is only in very simple cases of low loading that the drift of resistance can be attributed to one cause alone. There is a gradual onset of mechani~n (c) as higher resistivity materials are considered, while the likelihood of mechanism (a) decreases, according to the power relationship P ~ Va/R. The evidence for current flow effects Co) in the materials which are reported here is lacking, but there is some indication that there is a polarity dependence in the failure mechanisms of resistors which succumb to excessive Joule heatln~. Some work will be described below which gives clear evidence of case (d): this is a phenomenon which can, if the sign of the external heating effect is opposite to that of the internal heating (a), result in the anomaly of resistors carrying little or no

2. LOAD ~

lPlt~ w r r H TEIIOK-FILM l~818TOIm 2.1 Degradat/on m e c ~ When a potential is applied to a thick-film resistor, a number of effects can act to produce permanent changes: (a) Joule heating raises the temperature of the resistor and encourages diffusion and chemical reaction of constituents. This may be more pronounced in regions where the current density is increased by trimming or other constrictions. CO) D.c. current flow may cause drift of some of the component materials, e.g. electromigration at the contacts. 61

P. J. H O L M E S

62

3. LOAD ~

current changing more than: those which are appreciably loaded.

3.1 Dupont "Birox" D.c. load tests were run at 10, 20 and 4 0 W per square inch of resistor area (0.2, 0.4 and 0.8 W per resistor) (Table 1). The loading tests were carried out in an oven ambient of ~70°C and the measurements were done at room ambient of 20°C, T o keep within a substrate loading of 2 W/in. =. using a test pattern of four 0.2 by 0.1 in. resistors on a 1 in. by ½ in. substrate, all the resistors could be used at the 10 W level, two resistors for the 20 W level and one resistor for the 40 W level. Both untrimmed and trimmed resistors of each ink

2.2 The u~rking tonperature of a re~tor The actual working temperature of a resistor on its substrate is a function of the ambient temperature, the Joule heating within the resistor itself, and heat conducted through the substrate or encapsulating medium from other sources. When there are no other sources and no heat sink the temperature rise in a small resistor dissipating I.I W on a 0.025 in. thick substrate , I in. by 0.5 in., is about 600C above ambient. When four

so

0

I

20

40

60

EVALUATION

2

80

I00

tpo

"T'er~ rise,

140

160

180

200

220

"C

FIo. 1. Power-temperature relationship. Curves for 1, 2 and 4 resistors u n d e r power.

such resistors, uniformly spread over the substrate, are each~dissipating this power, the temperature rise of each resistor is 160°C above ambient. Under these conditions, there is little difference between the resistor temperature and the mean substrate temperature, so we can talk in terms of power density over the substrate*, moreover the relation between temperature rise and power is linear over a wide range, as shown in Fig. I. It is accordingly possible to estimate the approximate working temperature in those cases quoted in this report where no direct measurement was recorded during the test. • Fahley ¢1~ has employed the approximate rehttion TR ---- T a (1"3--0"3 exp (--2PMPs)) whore TR is the resistor workin Z temperature, 7"8 is the average substrste temperature and P a respectively are the power applied to t h e resistor in question and to the substrate as a whole. This gives TRool-26 T s for one res~tor and

TRNI'06 Ts for four equally powered resistors.

were studied. The substrates in this series of tests were mounted in edge connectors. Table 2 shows the results up to 1000 hr. F o r the three lower resistivity inks the percentage deviations were very small and general trends were hardly discernible, but for 1051 and I053, the deviations were significant, due to the large applied voltages. Usually the msximum individual deviations were roughly twice these mean values. In all cases, the trimmed resistors showed a more negative and greater deviation than the untrimmed ones. The greatest rate of chAr~ge in the resistance values always seemed to occur in the first 5 hr, a change in the slope of the deviation curve occurring between 5 and 10 hr. A further series of tests were undertaken at 40 W/in. = for 1021, 1031 and 1041 andat 10 W/in. = for the other two. These were put on load in laboratory ambient (approximately 20°C) conditions and measured at intervals from 15 rain. up

THE LOAD LIFE CHARACTERISTICS

OF T H I C K - F I L M

RESISTORS

63.

Table 1. Powor~amt ~ t a g e Ioadings for tests on Birox series Power W/in t resistor

Mean resistance

Ink

Voltage gradient V/in.

Voltage

area

1021

285 a

10 20 40

7.5 10 15

1031

2.95K~

10 20 40

25 33 50

1041

32.5 Kf~

10 20 40

1051

290 Kt~

i053

900 K o

Power on substrate e W

37.5 50 75

Approx. resisto~ temperature °C

0.8 0-8 0.8

98 106 110

125 165 250

0.8 0.8 0.8

98 106 110

80 107 160

400 535 800

0,8 0.8 0.8

98 106 110

10 40

245 490

1225 2450

0.8 0.8

98 110

10

450

2250

0.8

98

• 4, 2 and 1 resistor energized, respectively, at 10, 20 and 40 W/in t resistor loadins, on 1 in. by 0-5 in. substrate. Multiply power by 2 to obtain watts per square inch of substrate.

to 24 hr. (The lower ambient temperature was chosen to enable measurements to be taken more quickly after the load was removed.) Results indicated that the deviation rose fairly steadily with time, the tr/mmed resistors again showing more negative deviations than the untrimmed ones. For the two highest resistivity inks, the maximum deviations were considerably reduced by comparison with the runs in a 70°C ambient, e.g. 1051 approximately 0-I per cent after 5 hr instead of 0-25 per cent;

1053 approximately 0-4- per cent after 5 hr instead of 2.8 per cent. A test was also undertaken to find the power necessary to destroy a 0.050 in. square 1021 resistor when the 8ubstrate was mounted on an infinite heat sink. Resistance changes over a few minutes were less than 20 pe r cent up to 15 W (6 kW/in s of resistor area) when breakdown occurred and the resistance increased by several hundred per cent. RI2A

I¢)

~

_N~ 3A

2 - ~ 20 =S OMean k ~ I chonge 2 R54A 3 Io

20

50

~o

20o

50o

,ooo

2oo~

50o0

hr FIo. 2. Alloys Unlimited leries A, load tests. 100°C mnbient.

64

P. J. H O L M E S

3.2 Alloys Unlia6ted wri~ " A " A n extensive evaluation of the reliability of series " A " resistors included d.c. load life testing and step-stress tests, as well as some a.c. tests. Samples for this work had four 0.2 in. b y 0.1 in.

trimmed resistor* o f the following resistivities on 1 in. by 0-5 in. mbstrates. R12A (100 t~/sq.), R13A (1 KO/sq.) and R54A (50 Kf]/sq.). Contacts i n this series of tests were soldered with high melting point solder.

Table 2. Percentage det~tion in resistance values for Birox re~.stors undergoing d.c. load tests at 70°C a m b / ~ t temperature

Ink Code

Trimming state

1021

Untrimmed

1021

Trimmed

Load level watta per (in.)* resistor area

5 hr

50 hr

500 hr

1000 hr

0 10 20 40

+0"01 --0.01 +0"02 +0.01

0 +0"01 +0"02 +0"01

+0.01 +0.03 +0.05 +0.03

+0"01 +0"05 +0"05 +0"05

0 10 20

--0-02 --0.09 --0-08

--0"04 --0-11 --0.08

--0"02 --0"12 --0"08

40

--0.09

--0"03 --0-11 --0.07 --0-08

--0.08

--0.06

% Resistance deviation

1031

Untrimmed

0 10 20 40

+0"03 --0-02 --0"03 --0.04

+0"03 +0"05 0-04 +0"02

+0.04 +0.06 +0.06 +0-04

+0.01 +0"02 +0"04 +0"02

1031

Trimmed

0 10 20 40

--0-02 --0"10 --0'07 --0"10

--0"07 --0"14 --0"13 --0"12

--0"08 --0"13 --0-09 --0.12

--0"05 --0-11 --0.06 --0"08

1041

Untrimmed

0 10 20 40

0 --0.01 --0.06 --0.06

0 --0.02 --0.01 --0.05

+0-07 --0.05 --0.03 --0.03

+0-02 --0.06 --0.02 --0"03

1041

Trimmed

0 10 20 40

--0-01 --0.12 --0.07 --0-06

--0.02 --0.09 --0.12 --0.14

--0.04 --0.21 --0.11 --0.12

--0.05 -0.20 -0-12 --0-12

1051

Untrimmed

0 10 40

+0.06 --0.04 --0.84

+0.04 --0.01 -- 1.20

0 --0.15 -- 1"61

--0.04 --0.23 -- 1.83

1051

Trimmed

0 10 40

--0.07 --0-15

--0.11 --0"22

--0.16 --0.38

--0"09 --0.30

--0"89

-- 1"48

--2"03

-- 2"03

1053

Untrimmed

0 10

+0"03 --1"55

+0.01 --2"80

0 --3.74

--0"01 --4"16

1053

Trimmed

0 10

--0"05 --2.30

--0"12 --3.93

--0.17 --4.68

--0"12 --4"76

T H E L O A D L I F E C H A R A C T E R I S T I C S OF T H I C K - F I L M R E S I S T O R S Power loadings were ealcadated from mean trimmed areas, and the R12A and R13A samples were stressed at 10 W/in s at a 100°C ambient, while the RS¢A was operated at 5 W/in s at the same temgmmture (Table 3). Results are shown in Fig. 2, A notable feature of the short-term drifts was that they were of the same shmlute value, rather than percentage value, for the different starting

yT•

2

!° |

Fro. 3.

resistances in each group, e.g. R12A (1000hr at 10W/inS); mean drift 2.4ta -----+1.15 per cent or leas; R13A (1000hr at 10W/ins); mean drift 14 ta = +0.67 per cent or lees; R54A (1000 hr at 5 W/ins); mean drift 1.4ta ----1.36 per cent or less. The obvious inference that the termination region is responsible has yet to be fully substantiated. Beyond I000 hr the spread of drifts widened rapidly for ~ e h of the three groups. Tests at 100°C without load showed that little change took place after about the first 3 hr, when mean drifts of +0-24 and 4-0.12 t~ were recorded for R12A and R13A (+0.115 and 4-0.006 per cent). It appeared that some extraneous factors were influencing the early life of some of the R13A samples, since some went slightly positive and some equally slightly negative at this temperature. This was probably an effect of the abrasive trimming. Temperature step-stress tests (20 hr at each 30 ° temperature increment from 20°C upwards) on the three resistor batches showed changes to increase rapidly above 170°C (Fig. 3). T h e soldered connections showed obvious signs of deterioration above 230°C, and were probably affecting the observed changes slightly at the 200°C level. Power step--stre~ test~ were carried out at intervals of 10 W/in s of resistor area on d.c. each for 20 hr (Fig. 4), with further RSCA samples

Table 3. Power and ~oltage loadingsfor Alloys A series resistors

Ink

Mean resistance

Power W/in s resistor area

65

Voltage (approx.) 6 6"5

Voltage gradient

Power on substrate s

Approx. resistor temperature

V/in.

W

°C

30 32

0.72 0-72

128 128

R12A

205 240

10 10

R13A

2"05K 2"40K

10 10

19 21

95 105

0.72 0"72

128 128

R54A

103K 130K 155K

5 5 5

97 108 118

485 540 590

0"36 0"36 0"36

114 114 114-

• Four rmistors energized on each 1 in. by 0"5 in. substrate. Multiply power by 2 to give watts per square inch of substntte.

P. J. H O L M E S

66

to a resistor temperature of less than 170°C, there was good correSnion for R12A between the drifts observed o n t h e two types of stress, (Above this teraperamre the drifts were larger on the temperature stress @~n on equivalent d,c, load.) However, the same analysis applied to the R13A resistors showedthe d.c. load effect to be much ~ r and in the opposite direction to the equivalent temperature drifts, indicating that a second mechanism, other than the pure temperature effect shown by R12A, was operating. This was even more marked when R54A was tested, although here both temperature and d.c. load dr/fts were in the same n~ative direction. An a.c. test was also applied to lR54A samples, and the effect of a given power level was found to be larger than for d.c. Examination 'of the results revealed that the drifts were similar when compared on the basis of maximum voltage gradient, i.e. peak a.c. voltage, rather thanlr.m.s. voltage, is the 'important factor. The necessity for a statement of the total power appl/ed to the substrate is brought home by the results on R12A, where at 60 W/in = the resistor temperature was 76°C if it is the only one energized, but 182°C if all four were energized. (These conditions are approximately 2.2 and 8 . 8 W respectively, per square inch of substrate.) Different working temperatures may well account for some of the apparent discrepancies b~fween manufacturers' claims for stability and the figures obtained by some users.

tested on a.c. loading at 5 and 10 W/in s steps. In all of these tests, all four resistors on the 1 in. by 0.5 in. substrate were energized, and the resistor working temperatures are indicated in the figure. It was found that as long as the stress gave rise

RI2A

(4s) (75) (~a (12~ (J~.,,.,.~ i0 20 30 ~ so

(209) (23¢) 70 so

~.2

R54A

FIG. 4. Alloys Unlimited series A, step stre~ teats.

Table 4. Power and ~oltage loadingsfor Alloys B series resistors

Ink

Mean resistance

Power W/in = resistor area

RllB

32

20

3.2

40

4.5

20 40

24.7 35.0

123.5 175.0

R13B R54B R54B

1.92 K 111 K 137 K

Voltage (approx.)

Power on substrate • W

Approx. resistor temperature °C

15.9

1.36

106

22.5

2.72

135

1.28 2.32

135

Voltage gradient V/in.

106

20

180

900

1.16

106

40

253

1265

2.32

135

20

203

1015

1.2

106

40

286

1430

2.4

135

• Four resistors energized on each 1/n. square substrate.

T H E LOAD L I F E C H A R A C T E R I S T I C S OF T H I C K - F I L M R E S I S T O R S 3.3 Alloys Unlimited series " B " ! Most results given here for this series refer to samples coated with Scotchcast 281 to a depth of about 0.11 in. Comparison with uncoated samples showed, however, that there was no chemical interaction with Scotchcast, so that the degradation observed differed from that on coated samples only as a result of the altered conditions of heat dissipation: the thermal resistance of the coating means that the resistor and substrates inside get slightly hotter. Substrate temperature measurements showed an almost linear rise of 20°C per watt dissipated on the 1 in. square substrate, up t o a mean temperature of 144°C at 6.8W/in I of substrate. Three resistivities ( R l l B (10~/sq.), R13B (1 K~/sq.) and R54B (50 Kfl/sq.)) were studied, all in a test pattern of eight resistors 0.100 in. wide and 0.200 in. long, on a 1-in. square substrate. Each resistor was trimmed by up to 20 per cent of its width to achieve a fixed value.

:!

i

67

from those of R l l B and R13B, in that more than half of the drift on R54B had already occurred by the time of the first measurement at 4 hr. This was also true of some unencapsulated samples which were measured for the first 24 hr. The results in Table 5 can clearly be used (although perhaps naively) to assess the contributions to drift from the ambient temperature, the working temperature and the powerpassed into the resistor. For example, R l l B gives Drift due to 70°C ambient +0.27% Drift due to 1.36 W/in s on substrate +0.78% (36° temperature rise: resistors at 106°C) Drift due to 20 W/in S power q-0.24% Drift due to 2.72 W/in S on substrate +1.87% (65 ° temperature rise: resistors at 135°C) Drift due to 40 W/in S power +0-47% Temperature step-stress tests were carried out at 10° steps each of 20 hr, from 150° up to 2500C on each of the batches, as well as on some R12B (100f~/sq.) resistors (Fig. 6). Unlike the "A" RI3B

•e•,•"

RIIB

-6 u

•~c

I

2O

I

5O

T1 ~ ~ l i

I00

200

~00

|000

a54e

2000

hr

FIo. 5. Alloys UnlL,nited series B, load tests, 70°C ambient. D.c. power !oadings were calculated from the mean trimmed areas. Table 4 shows the powers and voltages applied to achieve the two standard life test levels of ~0 W and 40 W/in m of resistor area. Contacts were soldered to the terminations. D.c. load life tests (l~lg. 5) were conducted on an ambient of 70°C. Results are illustrated in Fig. 5 for 40 W/in s and also tabulated iniTable 5 for 1600 hr. Resistor temperatures estimated from Fahley's formula are also shown. The drifts on RS4B diifi~ed most significantly

series, these samples were unsolddred and did not show any drastic worsening of drift above 170°C. Power step-stress tests were operated in 10W/in s steps, each of 20hr, from 40 to 100 W/in I of resistor area except for R54B which was taken in 5 W steps from 10 to 40 W/in I (Fig. 7). Once again, the drifts were ~ m e a ~ e d on the unloaded resistors on the same substrate. R l l B showed an initial tendency to drift very slightly negative, turning positive at the 50 W level and then rising steadily, The unloaded

•

P. J. HOLMES

68

resistors also moved the mane way, turning podtive after the @08 W/ins subetrate level (60W/in' of energ/zed resistor) and reaching -/-0.7 per cent at the final step. The difference between loaded and unloaded RI3B resistors was the most marked, the latter increasing in value

'~

.2l~~.._ w/in z

l /

Io

Olltt~

|

~

'

20

3o

J

I

" ~

R.~UB F,~ ~ / " ~l . . . , . ; a ' ~ 70

40 t'~

,,

llll I J.~ T

(193) eo

i [

I

(135) (150) (1~1,) (I79).RI3B

(OOkD.)

1.5

RI3B RIIB

I

I'0~

5

t

0-5 -

FIG. 7. Alloys Ur,!im~t~ ~Mes B, step s~'eu tests.

'L/ 1,50,,=,,,f 180 ".

210

240

Temp. °C

FIG. 6. from the 40 W level (2.58 W/in z of substrate) onwards, to no lees than -)-3.18 per cent at the 100 W level (6-44 W/in I of subat~ate) while the loaded first negative and then much lees than 1 per cent positive (Fig. 8). R54B when unenergized never drifted more than --0.11 per cent from their initial value on e/ther the 80 KO or the 136 K a batch, while under load the L~vohatches changed by different amounts, viz.: (80) I ~ --0.60 per cent at 10 W/ins to --3~)8 per cent at 40 W/in.s; (136)KQ --1.81 per cent at 10W/in s to --7.20 per cent at 40 W/in z These fggures make it quite clear that the loadL~ effect is completely over-riding the temperature effect, but that the loading effect is not purely a matter of power. It is also not proportional to the voltage applied, as might have been anticipated if

resistorsdrifted resistors

2--

t

1--

o--~

11,

~ I IJ'i t 40"=""~a.u._~lO~.,,,~-e.~,-~v. ~~ 90 iO0 t,~,a " ~ T o . e~__. U : ~ (,?:22) ..35 (15o) (is4) (i79)(m) W/in =

FIG. 8. Comparison of loaded and unloaded resistors (RI3B).

T H E L O A D L I F E C H A R A C T E R I S T I C S OF T H I C K - F I L M

Table 5. Percentage deviation of resistance mO.uafor Alloys U ~

RESISTORS

69

m~m B rai:tor: on d.c. load tests

at 70°C amb/ent temperature

Ink R11B

R13B

RS4B (111 Kt3)

R54.B (137 Kfl)

Mean drift (%)

Loading of resistors

Loading of substntte

20 Wlin. I No lead, on same substrate 40 W/in. s No load, on same substrate No load, 70°C ambient

+ 1-29 + 1.05 +2"61 +2.14 +0-27

~

20 W/in? No load, on same substrate 40 W/in? No load, on same subetrate No load, 70°C ambient

+0.89 + 1"2 +2-71 + 2"93 +0-003

"~

20 W/in. s No load, same substrate 40 W/in.s No load, same substrate No load, 70°C ambient

--3-49 --0.98 -- 8"2 --0"18 --0.011

"~

20 W/in." No load, same substrate 40 W/in? No load, same substrate No load, 70°C ambient

- 5"77 +0"87 -- 16"2 +0-031 +0.035

"~

it was a field-dependent phenomenon. T h e nearest fit to the two 40 W/in s results would be a drift proportional to the square of the voltage. T h e effect is worthy of further investigation.

4. CONCLUSIONS 4.1 Feature: ettablit~d by the load life tests Although the results am.-mbled here leave many questions unanswered, a number of condusione can be drawn. (1) T h e deterioration of thick-film r eAjators of lower resistivities (up to about 10ill/square) is primarily a function of the workinf temperature. In some cases the correlation with resistors held at the same temperature for the same time is almost exact.

1"36 W/in s 2.72 W/in s 1-29 W/in s 2"58 W/ins 1.16 W/in* 2.32 W/inS 1"2 W/in' 2.4 W/ins

(2) T h e working temperature of a resistor is a combination of three factors: the ambient temperature, the Joule heat generated in the resistor itself and the heat conducted from elsewhere on the substrate. (3) High resistivity materials deteriorate mainly as a result of the high potential gradients which have to be applied to achieve a significant power dissipation. There is evidence that the amount of drift is not necessarily proportional to voltage gradient or to power dissipated. This may reflect the ilxhomogeneity of thick-~lm resistors, in which local voltage gradients and current densities may be higher than average and produce disproportionate effects. (4) The mechanisms of degradation for a given

70

P. J. HOLMES

For the purpose of providing the system reliresistor range are not always the same with different combinations of resistance, power and ability analyst with a working figure of mean time temperature. Several different modes can be before failure (MTBF) or failure rate per unit recognized among the examples reported in time, long life tests are unavoidable, and if the addition to the primary modes mentioned in (1) thick-film circuit has been well designed, the "failure" will be a deterioration beyond preand (3): determined limits, rather than a catastrophic (a) reproducible changes at the terminations, failure. From the results given it is concluded: presumably metallurgical; (1) that load life tests should be conducted for a (b) early life changes, probably attributable to minimum of 5000 hr, with measurements at the relaxation of stress incorporated during logarithmically-spaced time intervals to defiring and trimming; tect whether the drift occurs mainly over (c) random changes due to disturbance of short or long terms (e.g. 1, 3,10, 30 hr, etc.); soldered contacts well below their actual (2) that life tests require to be conducted on melting point; standard sized substrates (1 in. square or (d) there may be differences in behaviour smaller), and that the number of resistors between samples, initially of the same resisdissipating at the specified power level on tance, which have been trimmed to final each substrate should be fixed, with addivalue in different ways (e.g. by cutting a tional unloaded resistors on the substrate lateral notch, by shaving along one edge, also monitored as the test proceeds; etc.), owing to the varying conditions of (3) that a realistic standard test ambient temcurrent crowding (hot spots) and exposed perature should be defined; 70°C is a level edges; this point needs further investigation. which, in the absence of power, will not cause significant resistance changes;

4.2 R e c ~ t / o ~ f o r

s ~ f s ~ / o o a kye

The tests described have been carded out at a variety of ambient temperatures, and have embraced working temperatures of 200°C and above, at which levels the question of the solder used to make contact becomes important. The resistor and substrate sizes have been standardized in most of the work reported here, and no heat sinks have been used. The load levels of 40 W/in a have proved too high for most high-valued resistors, because of the high voltages involved, while for lower valued resistors the effects of such loading have varied from catastrophic to tolerable over 1000-hr periods. There is no uniformity of behaviour in time: some materials drift rapidly and then stabilize, while at the other extreme some materials are hardly affected by 1000-hr operation but start to deteriorate rapidly at a later stage. From this summary of be~viour, it is difficult to define the ideal life test for disclosing every likely problem, and it could well be argued that if the aim is to do just this, the combination of power and temperature step-stress tests is a better and less time-consuming approach.

(4) that the actual working temperature of each resistor must be determined; (5) that tests should normally be carried out on unsoldered substrates, fitted with edge conhectors, with additional tests to determine how far the particular solder used in the assembled system worsens the upper limit of operating temperature. (To fulfil requirement (5) it is desirable for all the connections to be brought out to one edge of the test substrate, which means providing a return conductor track from the opposite edge.) Since it is not possible to devise a single test suitable for the full range of thick-film resistors, the following parameters should always accompany the experimental results: Resistor and substrate dimensions Watts per unit area of resistor Volts per unit length of resistor Watts on the substrate Ambient temperature The actual or estimated resistor working temperature.

T H E L O A D L I F E C H A R A C T E R I S T I C S OF T H I C K - F I L M R E S I S T O R S

71

The results should be displayed graphically, as a function of time or stress level with maximum and minimum drifts shown, plns a tabular statement of mean drift and standard deviation at the test time (e.g. 1000 hr) used in equipment reliability calculations and in both cases the effect on the unenergized resistors should be shown.

tropics) Ltd. and Messrs. R. Britton and B. Kranse of the Plessey Company. The help of Miss P. L. Shove, of the Admiralty Surface Weapons Establishment, is also acknowledged. This paper is published by permi~on of the Controller, H.M. Stationery Office. Crown copyright reserved.

A d m o c v / e d g ~ T h e tests reported here have been conducted by Mrs. J. B. McCloghrie of the Royal Aircraft Establishment, Mr. J. Barron of Tectonic (Elec-

1. W. A. F ~ ,

~ C E A Stamtardi~ed Hybrid Microcircuit

De~,n Gu/de. Proceeding1 of the 1970 IEEE Components Co4derence, p. 503.