- Email: [email protected]
Radiation effects on JFETS, MOSFETS, and bipolar transistors, as related to SSC circuit design
452
Nuclear Instruments and Methods in Physics Research A307 (1991) 452-457 North-Holland
Radiation effects on JFETS, MOSFETS, and bipolar transistors, as related to SSC circuit design * E.J. Kennedy Department of Electrical and Computer Engineering, The University of Tennessee, Knoxville, TN 37996, USA
G.T. Alley and C.L. Britton, Jr. Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831-6006, USA
P.L. Skubic University of Oklahoma, Department of Physics and Astronomy, Norman, OK 73019, USA
B. Gray and A. Wu Department of Electrical and Computer Engineering, The University of Tenneessee, Knoxville, TN 37996, USA Received 12 February 1991 and in revised form 23 May 1991
Some results of radiation effects on selected junction field-effect transistors, MOS field-effect transistors, and bipolar junction transistors are presented. The evaluations include dc parameters, as well as capacitive variations and noise evaluations. The tests are made at the low current and voltage levels (in particular at currents _<1 mA) that are essential for the low-power regimes required by SSC circuitry. Detailed noise data are presented both before and after 5-Mrad (gamma) total-dose exposure. SPICE radiation models for three high-frequency bipolar processes are compared for a typical charge-sensitive preamplifier.
1. Introduction
The operating environment anticipated for the Superconducting Super Collider (SSC) is a very harsh operating regime. The proposed event period of 16 ns demands very high frequency performance from electronic circuits, yet the need for thousands of circuits absolutely requires a low power operating budget. Additionally, an absolute limitation to circuit performance is the radiation due primarily to gamma and neutron bombardments that can exceed a total dose of 1 Mrad (gamma) and 1013 n e u t r o n s / c m 2 in a typical year of operation, dependent upon the distance from the beam [1]. This article reports research conducted at the Oak Ridge National Laboratory ( O R N L ) =1 and The University of Tennessee, Knoxville, which was primarily concerned with gamma radiation total-dose effects on
* This research was funded by the DOE SSC Generic Detector Program. ~*l The Oak Ridge National Laboratory is operated for the U.S. Department of Energy by Martin Marietta Energy Systems under contract No. DE-AC05-84OR21400. Elsevier Science Publishers B.V.
semiconductor devices and circuits. The source was a Gammacell 200 6°Co irradiator. The circuits developed were primarily preamplifiers that could be used in large-volume calorimeters having detector capacitances in the range of 20-500 pF. The radiation properties of three high-frequency bipolar processes (from A T & T , Harris Semiconductor, and VTC, Inc.) were investigated, as well as several commercially available junction field-effect transistors (JFETs). Although it was understood that the use of M O S F E T s should require a special radiation-hardened process, we nevertheless decided to investigate the properties of a MOSIS array to see just how deleterious radiation effects would be for a typically processed C M O S circuit. Although the restricted length of this article prevents reporting all the data, the most essential radiation produced changes are stated.
2. Gamma radiation effects on J F E T s
The J F E T generally suffers the least damage in a radiation environment when compared with bipolar and MOS transistors. The reason is that a J F E T is a major-
453
E.J. Kennedy et al. / Radiation effects on JFETS, MOSFETS, and bipolar transistors 10 4 , 10'
~L
(')
4|
r
T
o~ I1"'1--.1
,.
Rn SMrsd
~
Rn
[--t [-
T" e~
....
°
I 0 ~t
tt ! i
10 3 ,
ee
[[ lOl
"
"''""1
•
1~
""'''!
. . . . . .
10:
1 0 :~
Freq.
"1
. . . . . .
10 4
,
. . . . . . .
10 s
10:
I
•
• -..,'-I
101
10'
. . . . .
101
.,'1
•
• ,, ,,,'1
Freq.
(Hz)
I
........
[
10 ~
IO s
l D =
100 ~A (VDs = 2.5 V).
storage to minimize annealing. The n-channel JFET units evaluated were the NJ3600, NJ1800, NJ903, 2N6451 and 2N4858 from InterFet, Inc.; the 2N5432 and IMF5912 from Intersil; the 2SK152 from Sony; and the 2N4416A from Siliconix. Since space is limited, only data for the largest area (N J3600) and smallest area (2N4416A) devices are presented in this article. All devices were characterized at three drain currents of 1 mA, 100 p~A and 10 ~tA, and with a drain-source voltage of 2.5 V, which are operating regimes anticipated for the low power requirements of the SSC. The equivalent noise voltage (ENV, or "series" noise) of the N J3600 and 2N4416A are shown in figs. 1 and 2 for I D of 1 mA and 100 p~A, both before and after 5-Mrad gamma radiation. The noise is expressed as an equivalent noise resistance. The lower straight line on the graphs indicates the theoretical midband noise resistance anticipated ( R n = 4 k T ( O . 6 6 / g ~ ) ) . The "error bars" indicate the range of measurements (min, max) for the five units of each device tested. As is apparent in figs. 1 and 2, the region of midband noise is not achieved until > 100 kHz after radiation. The lowfrequency "noise corner", fc, at which the series noise resistance is twice that of the midband noise resistance,
ity carrier device and thus does not suffer from the minority carrier recombination problems that degrade a bipolar transistor. Although the MOS transistor is also a majority carrier device, the MOS transistor (unlike the JFET) is controlled by the electric field due to both the insulating gate oxide and the oxide-semiconductor interface states, and thus is quite sensitive to excess charge accumulation at the oxide and interface produced by ionizing radiation. The most significant radiation produced degradation in a JFET is the increase of reverse gate leakage current ( I o), which typically results in an increase from a few picoamps to a few nanoamps after several megarads of gamma dose. When a JFET is used as the front-end device in a charge-sensitive (CS) preamplifier, the parameters of most interest are series noise, transconductance (gm), capacitance, and output resistance (usually expressed as the ratio of the Early voltage (VA) to the dc drain current (ID)). Generally, five samples of each device were characterized both before and after a 5-Mrad 6°Co irradiation. The devices were irradiated in a passive mode (all leads shorted together), and postirradiation testing was accomplished within typically 2-3 days after irradiation. All devices were kept at - I ° C during
I 0~
(,) ....
~"
' • .....
(Hz)
Fig. 1. Series noise resistance versus frequency for the InterFet NJ3600L at (a) ID = 1 mA and (b)
Io'-[--D..L,.L,
•
10 4
I0 ~
~o' [" [" "t...L"
Rn SMrsd
10 s
(b) 4~ I
E
E
@
@ l 0 '~
10 s
J'''g"
" ~
°" • ""
-m -- ~,. .
i
Rn
i
~
5Mrad
Rn
IL~r.
e10 ~ .
.
.
.
.
.
.
.
.
.
.
""
....
.
.
.
.
.
~.
_~..._
•
' ......
-
_~.
1o~t . . . . . . .
_
10 2 ........
10'
I
10Z
•
I
........
10 5
I
10 ¢
I
lO s
.......
I
i0'
10~"11"
......
10t
-- -~_ -~--~-_-- .7--- - ~ --- = - ~
.....
I0 ~
Freq. (Hz) Fig. 2. Series noise resistance versus frequency for the Siliconix 2N4416A JFET at (a)
!
........
10 ~
|
10 4
........
I
lO s
•
• .-,,-I
10 6
Freq. (Hz) I o =
1 rrlA and (b)
I D =
100 p.A (VDs = 2.5 V).
E.J. Kennedy et aL / Radiation effects on JFETS, MOSFETS, and bipolar transistors
454 700 -
(a)
(b)
240
f~O°
~" .qle.
400
]
300. -2
-3
.............
'-]Hlllt
-
140..iLLL.L&
" -1
-nr
I!1111 i
-4
-$
0
I~ksl~R~.~ -1
-2
-3
V~v
-4
-5
Va~ v
Fig. 3. Capacitance variations for the NJ3600L JFET, both before and after a 5-Mrad dose. (a) Input capacitance Ciss and (b) gate-to-drain capacitance Cr~~.
is shifted from 15 Hz (before radiation) to - 70 kHz after radiation in the N J3600, versus - 2 kHz (prerad) to - 500 kHz (postrad) for the 2N4416A, when I D = 1 mA. The multiple peaks in the postrad data are characteristic of recombination-generation noise due to trapping centers introduced by possible dislocations in the crystal lattice structure. The measured input and output capacitance (Cis ~ and Crs~) for the JFETs generally indicated very slight changes in the postrad data relative to the initial values. Fig. 3 illustrates data comparisons for the N J3600, while fig. 4 shows a shght ( - 0.2 pF) increase of capacitance for the 2N4416A device after radiation. The initial decrease of the C,~s values for Vr~..... < 1 V for the N J3600 transistor after radiation is puzzling. Several other parameters of the tested JFETs indicated very little change with radiation, with pinch-off voltages changing < 1%, l o s s values typically unchanged, and gm values indicating perhaps a slight ( < 1.8%) decrease after radiation. The Early voltage values, however, increased from +18% for the NJ3600 to + 160% for the 2N4416A devices. As anticipated, the gate leakage currents increased because of radiation from < 10- a3 A to 0.4 nA for the 2N4416A, and for the largest area device (N J3600) the change was from 0.3 to 6 nA after 5 Mrad. -
3. G a m m a r a d i a t i o n e f f e c t s o n b i p o l a r t r a n s i s t o r s
The most significant effect of 6°Co gamma radiation on the high-frequency bipolar transistors tested was a decrease in the dc current gain, Beta (fl) or hFE. The Hewlett Packard Semiconductor Parameter Analyzer, Model 4145B, was used extensively for all dc measurements. Since radiation causes a decrease in minority carrier lifetime as well as a decrease in effective doping, the most radiation-resistant bipolar transistors will be those that have a relatively heavy base doping as well as a very narrow base width. Both of these criteria are present in high-frequency transistors having large gain-bandwidth products (fT). The processes evaluated were those of the leading bipolar analog VHF semicustom integrated circuit silicon foundries, namely AT &T, Harris Semiconductor, and VTC, Inc. The A T & T ALA200 array is typically a 4.5 GHz npn and 3.8 GHz pnp fT family. The Harris analog F A S T R A C K VHF process utilizes 1.2 GHz npns and 1 GHz pnps, while the VTC V J900 process provides a typical maximum f T of 6 G H z (npn) and 1.5 G H z (pnp). All three manufacturers provide complete SPICE models for their transistors. Fig. 5 compares the prerad and postrad current gains
r
(')
4.0
:~
x.s -
(b)
1.6.3.0
,oJ
2.0-
-".----
1.5
o
-;
.i
0
-1
VaEv
-2
-3 VI~ v
Fig. 4. Capacitance variations for the 2N4416A JFET.
-4
-5
455
E.J. Kennedy et al. / Radiation effects on JFETS, M O S F E T S , and bipolar transistors
Table 2 SPICE simulation comparisons for a CS folded cascode bipolar transistor preamplifier with a calorimeter detector capacitance of 112 pF, peaking-time (CR-RC) of 50 ns.
180 HARRIS 160
AT&T
140 120
v
~
80
m
60
I
~
_
__~.~
.-'""~
40
..-" "
. . - " " " HARRIS
10 -5
10 -4
Rise time Ins]
Pdiss [mW]
Noise ENC e [rms electrons]
Preradiation
At&T Harris VTC
4.9 5.5 4.9
19.1 20.0 19.3
4098 5001 4026
Postradiation (after 5 Mrad, 6°Co)
AT&T Harris VTC
7.0 8.0 7.0
18.3 17.7 18.5
4553 7821 5075
',
0 10-6
Process
10 .3
10 .2
Collector Current ( Amp. )
Fig. 5. Beta vs. I c for npn high-frequency transistors (VcE = 2.5
v).
It is apparent that the lower fT pnp devices suffer the most degradation due to the ionizing gamma radiation. There were only slight changes ( < 5%) in the Cbc and Cbe depletion layer capacitances produced by the irradiations. The Early voltages for npn devices typically decreased because of irradiation, from 3% to 10%, while VA increased by 15% for the Harris pnps, but decreased by 5% for the A T & T pnps and by 25% for the VTC pnps. The noise comparisons are shown in table 1, comparing the series noise (i.e., the equivalent noise voltage or the equivalent noise resistance) for the three processes, both before and after 5 Mrad radiation. The low-frequency " n o i s e c o m e r " , fc, for both a minimum size (1 x 1 emitter, or 1 x device) and a large size (16 × ) device are compared for both npn and pnp devices. Measurements of the equivalent noise currents (ENI) for the devices have not yet been made. Using the HP4145B parameter analyzer, we were able to obtain an accurate S P I C E model for each bipolar process, both before and after radiation. A comparison was made of each process to simulate a basic low-power, 20 mW, folded-cascode CS preamp [2] for use with a 2 cm × 2 cm silicon calorimeter detector (Cde t - - 1 1 2 pF). The results for the rise time, power dissipation, and anticipated equivalent noise charge (ENCe) for each pre-
(Beta) for the smallest area npn devices (an emitter size of 1 x 1, approximately 10 Ixm x 10 Fm), while fig. 6 compares fl for the small area (1 x 1) pnps, although the VTC unit tested was available in only a 1 x 9 area.
120
-Pre-radiatien (solid lines) Post-radiation, 5 Mrad (dashed lines)
100
< F-U.I m
HARRIS
60 40 .
. ........
~%i
.......
__------_- .. .-.-. ". .-. . . . . . . 10-6
10-5
.......
10-4
"'"
VTC
'
10-3
10-2
Collector Current ( Amp. )
Fig. 6. Beta vs I c for pnp high-frequency transistors. (VcE = -2.5 V).
Table 1 Series noise comparisons for bipolar processes [Ic = 1 mA, I VEcI = 2.5 V]. Process
Noise corner frequency fc [kHz] (lx) Prerad
NPNs
AT&T Harris VTC
2.4 0.6 0.35
PNPs
AT&T Harris VTC
5.0 6.0 -
R n
(16×) Postrad 2.4 2.4 0.55 6.0 10.0 -
(midband) [£]
(1×)
(16×)
Prerad
Postrad
6.0 2.0 2.5
10.0 2.8 3.2
76 550 455
55 62 52
5.0 1.7 (8 x ) 1.0
5.5 2.4 (8 X ) 1.5
51 148 (8 × ) 78
16 26 (8 x )78
456
E.J. Kennedy et al. / Radiation effects on JFETS, MOSFETS, and bipolar transistors IE4
amplifier are compared in table 2, both before and after radiation.
API~R 1.3 RRad A - - A Sample 2 O i00 ~A
4. Gamma radiation effects on M O S F E T S
Although a special radiation-hard C M O S process will be required for SSC circuits, we wanted to see how a typical MOSIS C M O S process would survive a g a m m a dose. Special N M O S and PMOS dual arrays having w i d t h / l e n g t h ratios of 100 p.m/10 ~ m and 1000 ~tm/10 t~m were designed using the M A G I C layout program. The chips were fabricated under the standard 2 ~tm p-well Orbit Semiconductor process, using the MOSIS fabrication program. All transistors were fabricated in a " q u a d " configuration, with diagonal transistors connected to obtain an overall well-matched dual transistor configuration, which is indicative of the construction required for low offset-voltage differential front-end devices. The dual transistors were biased in a normal conductive state (VDs = 2.5 V, I D = 1 mm, 100 p.A and 10 ~A) during the irradiation. The threshold voltage VT was monitored, and devices were removed when the observed threshold voltage shift for the P M O S units had shifted to approximately 2V-r (prerad), which occurred at a total dose of 1.3 Mrad. It was found that the P M O S devices had an average shift in VT from - 0 . 7 to - 1.5 V, while N M O S devices had VT shifts from 0.9 to 0.6 V, although undoubtedly the worst-case negative shift from the initial 0.9-V value was much larger due to the fixed oxide charge (Not), before the interface-trapped charge (Nit) effect produced a positive Vv trend. An example of the large increase ( - 5 × ) in the lowfrequency noise due to radiation is shown in fig. 7 for an N M O S device with W / L = 100 ~tm/10 p.m. Of most significance was the large increase in the offset voltage Vos of the dual units, with prerad values of < 1 mV for both n- and p-channel units observed for 1000 ~ m / 1 0 ~tm devices, while in a postrad measurement Vos values of > 100 mV were obtained. Further, the shift in gm (at 1 mA) was significantly higher than that observed in JFETs, approximately - 1 2 % for P M O S and - 1 9 % for N M O S units after radiation. The Early voltages, similar to those for JFETs, increased after radiation from - 20 V to - 60 V for P M O S units, and from - 30 V to - 150 V for N M O S units.
.~ O Z
100
• - - ~ S ~ p l e 2 0 lObot 10 hot
i00
i000
Frequency
~ o
IE4
Fig. 7. Noise spectra for MOSIS NMOS transistor with W / L =100 ~,m/10 p.m, both before and after a 1.3-Mrad (6°Co) total dose.
noise spectrum which was then measured by the HP3400A rms meter. After the rms value of the noise for a particular frequency was measured, a test signal having the same frequency as the center frequency of the band-pass filter was then applied to the system thus allowing calculation of the system noise gain. F r o m this information the data acquisition system calculated the series noise resistance ( R n ) of the J F E T being tested and stored the result. The process was repeated for all measured points in the spectrum. The error in the J F E T noise measurement due to the preamplifier noise contribution was calculated and verified with PSPICE to be on the order of _+ 2%, worst case. The accuracy of the variable band-pass filters was confirmed by using a calibrated noise source and was found to be repeatable and accurate with less than + 1% worst case error. The integration time of the rms meter was extended by converting the rms voltage measurement to rms power and taking the average of several
5. N o i ~ m e ~ e m e n t s
A block diagram of the noise measurement system used to acquire the drain current noise of the J F E T s is shown in fig. 8. The preamplifier was carefully designed to have low noise and sufficient gain to dominate the noise of the subsequent gain stage. The variable bandpass filter amplified only a narrow bandwidth in the
IE5
(Hz)
Fig. 8. Noise measurement system.
E.J, Kennedy et al. / Radiation effects on JFETS, MOSFETS, and bipolar transistors
samples. The number of samples required to produce measurements that were repeatable with an accuracy of greater than 95% was empirically found for each center frequency of the variable band-pass filter. This required as many as 10000 samples for the 10 Hz and 20 Hz center frequencies. A similar procedure was used for the MOSFET measurement except that fewer samples were used for a given frequency. The estimated error for these measurements was approximately + 15% worst case at the lowest frequencies.
457
when compared with an FET and shows promise as a useful technology for the short peaking times that will be required for the SSC.
Acknowledgements The authors acknowledge the various semiconductor companies, namely InterFet, AT&T, Harris, and VTC, Inc., for providing sample transistors and arrays for testing. We also thank M. Kemper for assistance in measurements and testing.
6. Conclusions Radiation using a 6°Co gamma source has shown a significant increase in the low-frequency noise for both JFETs and MOSFETs. JFETs are decidedly superior to MOSFETs where constancy of gm and threshold voltage are concerned. Although the Beta degradation is severe, particularly for pnp transistors, the highfrequency bipolar transistor offers very low series noise
References [1] D.E. Groom (ed.), Radiation Levels in the SSC Interaction Regions, SSC-SR-1033 (1988). [2] C.L. Britton, Jr. et al., Bipolar Monolithic Preamplifiers for SSC Silicon Calorimetry, presented at the Symp. on Detector Research and Development for the SSC, Fort Worth, TX, October 17, 1990.
Nuclear Instruments and Methods in Physics Research A307 (1991) 452-457 North-Holland
Radiation effects on JFETS, MOSFETS, and bipolar transistors, as related to SSC circuit design * E.J. Kennedy Department of Electrical and Computer Engineering, The University of Tennessee, Knoxville, TN 37996, USA
G.T. Alley and C.L. Britton, Jr. Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831-6006, USA
P.L. Skubic University of Oklahoma, Department of Physics and Astronomy, Norman, OK 73019, USA
B. Gray and A. Wu Department of Electrical and Computer Engineering, The University of Tenneessee, Knoxville, TN 37996, USA Received 12 February 1991 and in revised form 23 May 1991
Some results of radiation effects on selected junction field-effect transistors, MOS field-effect transistors, and bipolar junction transistors are presented. The evaluations include dc parameters, as well as capacitive variations and noise evaluations. The tests are made at the low current and voltage levels (in particular at currents _<1 mA) that are essential for the low-power regimes required by SSC circuitry. Detailed noise data are presented both before and after 5-Mrad (gamma) total-dose exposure. SPICE radiation models for three high-frequency bipolar processes are compared for a typical charge-sensitive preamplifier.
1. Introduction
The operating environment anticipated for the Superconducting Super Collider (SSC) is a very harsh operating regime. The proposed event period of 16 ns demands very high frequency performance from electronic circuits, yet the need for thousands of circuits absolutely requires a low power operating budget. Additionally, an absolute limitation to circuit performance is the radiation due primarily to gamma and neutron bombardments that can exceed a total dose of 1 Mrad (gamma) and 1013 n e u t r o n s / c m 2 in a typical year of operation, dependent upon the distance from the beam [1]. This article reports research conducted at the Oak Ridge National Laboratory ( O R N L ) =1 and The University of Tennessee, Knoxville, which was primarily concerned with gamma radiation total-dose effects on
* This research was funded by the DOE SSC Generic Detector Program. ~*l The Oak Ridge National Laboratory is operated for the U.S. Department of Energy by Martin Marietta Energy Systems under contract No. DE-AC05-84OR21400. Elsevier Science Publishers B.V.
semiconductor devices and circuits. The source was a Gammacell 200 6°Co irradiator. The circuits developed were primarily preamplifiers that could be used in large-volume calorimeters having detector capacitances in the range of 20-500 pF. The radiation properties of three high-frequency bipolar processes (from A T & T , Harris Semiconductor, and VTC, Inc.) were investigated, as well as several commercially available junction field-effect transistors (JFETs). Although it was understood that the use of M O S F E T s should require a special radiation-hardened process, we nevertheless decided to investigate the properties of a MOSIS array to see just how deleterious radiation effects would be for a typically processed C M O S circuit. Although the restricted length of this article prevents reporting all the data, the most essential radiation produced changes are stated.
2. Gamma radiation effects on J F E T s
The J F E T generally suffers the least damage in a radiation environment when compared with bipolar and MOS transistors. The reason is that a J F E T is a major-
453
E.J. Kennedy et al. / Radiation effects on JFETS, MOSFETS, and bipolar transistors 10 4 , 10'
~L
(')
4|
r
T
o~ I1"'1--.1
,.
Rn SMrsd
~
Rn
[--t [-
T" e~
....
°
I 0 ~t
tt ! i
10 3 ,
ee
[[ lOl
"
"''""1
•
1~
""'''!
. . . . . .
10:
1 0 :~
Freq.
"1
. . . . . .
10 4
,
. . . . . . .
10 s
10:
I
•
• -..,'-I
101
10'
. . . . .
101
.,'1
•
• ,, ,,,'1
Freq.
(Hz)
I
........
[
10 ~
IO s
l D =
100 ~A (VDs = 2.5 V).
storage to minimize annealing. The n-channel JFET units evaluated were the NJ3600, NJ1800, NJ903, 2N6451 and 2N4858 from InterFet, Inc.; the 2N5432 and IMF5912 from Intersil; the 2SK152 from Sony; and the 2N4416A from Siliconix. Since space is limited, only data for the largest area (N J3600) and smallest area (2N4416A) devices are presented in this article. All devices were characterized at three drain currents of 1 mA, 100 p~A and 10 ~tA, and with a drain-source voltage of 2.5 V, which are operating regimes anticipated for the low power requirements of the SSC. The equivalent noise voltage (ENV, or "series" noise) of the N J3600 and 2N4416A are shown in figs. 1 and 2 for I D of 1 mA and 100 p~A, both before and after 5-Mrad gamma radiation. The noise is expressed as an equivalent noise resistance. The lower straight line on the graphs indicates the theoretical midband noise resistance anticipated ( R n = 4 k T ( O . 6 6 / g ~ ) ) . The "error bars" indicate the range of measurements (min, max) for the five units of each device tested. As is apparent in figs. 1 and 2, the region of midband noise is not achieved until > 100 kHz after radiation. The lowfrequency "noise corner", fc, at which the series noise resistance is twice that of the midband noise resistance,
ity carrier device and thus does not suffer from the minority carrier recombination problems that degrade a bipolar transistor. Although the MOS transistor is also a majority carrier device, the MOS transistor (unlike the JFET) is controlled by the electric field due to both the insulating gate oxide and the oxide-semiconductor interface states, and thus is quite sensitive to excess charge accumulation at the oxide and interface produced by ionizing radiation. The most significant radiation produced degradation in a JFET is the increase of reverse gate leakage current ( I o), which typically results in an increase from a few picoamps to a few nanoamps after several megarads of gamma dose. When a JFET is used as the front-end device in a charge-sensitive (CS) preamplifier, the parameters of most interest are series noise, transconductance (gm), capacitance, and output resistance (usually expressed as the ratio of the Early voltage (VA) to the dc drain current (ID)). Generally, five samples of each device were characterized both before and after a 5-Mrad 6°Co irradiation. The devices were irradiated in a passive mode (all leads shorted together), and postirradiation testing was accomplished within typically 2-3 days after irradiation. All devices were kept at - I ° C during
I 0~
(,) ....
~"
' • .....
(Hz)
Fig. 1. Series noise resistance versus frequency for the InterFet NJ3600L at (a) ID = 1 mA and (b)
Io'-[--D..L,.L,
•
10 4
I0 ~
~o' [" [" "t...L"
Rn SMrsd
10 s
(b) 4~ I
E
E
@
@ l 0 '~
10 s
J'''g"
" ~
°" • ""
-m -- ~,. .
i
Rn
i
~
5Mrad
Rn
IL~r.
e10 ~ .
.
.
.
.
.
.
.
.
.
.
""
....
.
.
.
.
.
~.
_~..._
•
' ......
-
_~.
1o~t . . . . . . .
_
10 2 ........
10'
I
10Z
•
I
........
10 5
I
10 ¢
I
lO s
.......
I
i0'
10~"11"
......
10t
-- -~_ -~--~-_-- .7--- - ~ --- = - ~
.....
I0 ~
Freq. (Hz) Fig. 2. Series noise resistance versus frequency for the Siliconix 2N4416A JFET at (a)
!
........
10 ~
|
10 4
........
I
lO s
•
• .-,,-I
10 6
Freq. (Hz) I o =
1 rrlA and (b)
I D =
100 p.A (VDs = 2.5 V).
E.J. Kennedy et aL / Radiation effects on JFETS, MOSFETS, and bipolar transistors
454 700 -
(a)
(b)
240
f~O°
~" .qle.
400
]
300. -2
-3
.............
'-]Hlllt
-
140..iLLL.L&
" -1
-nr
I!1111 i
-4
-$
0
I~ksl~R~.~ -1
-2
-3
V~v
-4
-5
Va~ v
Fig. 3. Capacitance variations for the NJ3600L JFET, both before and after a 5-Mrad dose. (a) Input capacitance Ciss and (b) gate-to-drain capacitance Cr~~.
is shifted from 15 Hz (before radiation) to - 70 kHz after radiation in the N J3600, versus - 2 kHz (prerad) to - 500 kHz (postrad) for the 2N4416A, when I D = 1 mA. The multiple peaks in the postrad data are characteristic of recombination-generation noise due to trapping centers introduced by possible dislocations in the crystal lattice structure. The measured input and output capacitance (Cis ~ and Crs~) for the JFETs generally indicated very slight changes in the postrad data relative to the initial values. Fig. 3 illustrates data comparisons for the N J3600, while fig. 4 shows a shght ( - 0.2 pF) increase of capacitance for the 2N4416A device after radiation. The initial decrease of the C,~s values for Vr~..... < 1 V for the N J3600 transistor after radiation is puzzling. Several other parameters of the tested JFETs indicated very little change with radiation, with pinch-off voltages changing < 1%, l o s s values typically unchanged, and gm values indicating perhaps a slight ( < 1.8%) decrease after radiation. The Early voltage values, however, increased from +18% for the NJ3600 to + 160% for the 2N4416A devices. As anticipated, the gate leakage currents increased because of radiation from < 10- a3 A to 0.4 nA for the 2N4416A, and for the largest area device (N J3600) the change was from 0.3 to 6 nA after 5 Mrad. -
3. G a m m a r a d i a t i o n e f f e c t s o n b i p o l a r t r a n s i s t o r s
The most significant effect of 6°Co gamma radiation on the high-frequency bipolar transistors tested was a decrease in the dc current gain, Beta (fl) or hFE. The Hewlett Packard Semiconductor Parameter Analyzer, Model 4145B, was used extensively for all dc measurements. Since radiation causes a decrease in minority carrier lifetime as well as a decrease in effective doping, the most radiation-resistant bipolar transistors will be those that have a relatively heavy base doping as well as a very narrow base width. Both of these criteria are present in high-frequency transistors having large gain-bandwidth products (fT). The processes evaluated were those of the leading bipolar analog VHF semicustom integrated circuit silicon foundries, namely AT &T, Harris Semiconductor, and VTC, Inc. The A T & T ALA200 array is typically a 4.5 GHz npn and 3.8 GHz pnp fT family. The Harris analog F A S T R A C K VHF process utilizes 1.2 GHz npns and 1 GHz pnps, while the VTC V J900 process provides a typical maximum f T of 6 G H z (npn) and 1.5 G H z (pnp). All three manufacturers provide complete SPICE models for their transistors. Fig. 5 compares the prerad and postrad current gains
r
(')
4.0
:~
x.s -
(b)
1.6.3.0
,oJ
2.0-
-".----
1.5
o
-;
.i
0
-1
VaEv
-2
-3 VI~ v
Fig. 4. Capacitance variations for the 2N4416A JFET.
-4
-5
455
E.J. Kennedy et al. / Radiation effects on JFETS, M O S F E T S , and bipolar transistors
Table 2 SPICE simulation comparisons for a CS folded cascode bipolar transistor preamplifier with a calorimeter detector capacitance of 112 pF, peaking-time (CR-RC) of 50 ns.
180 HARRIS 160
AT&T
140 120
v
~
80
m
60
I
~
_
__~.~
.-'""~
40
..-" "
. . - " " " HARRIS
10 -5
10 -4
Rise time Ins]
Pdiss [mW]
Noise ENC e [rms electrons]
Preradiation
At&T Harris VTC
4.9 5.5 4.9
19.1 20.0 19.3
4098 5001 4026
Postradiation (after 5 Mrad, 6°Co)
AT&T Harris VTC
7.0 8.0 7.0
18.3 17.7 18.5
4553 7821 5075
',
0 10-6
Process
10 .3
10 .2
Collector Current ( Amp. )
Fig. 5. Beta vs. I c for npn high-frequency transistors (VcE = 2.5
v).
It is apparent that the lower fT pnp devices suffer the most degradation due to the ionizing gamma radiation. There were only slight changes ( < 5%) in the Cbc and Cbe depletion layer capacitances produced by the irradiations. The Early voltages for npn devices typically decreased because of irradiation, from 3% to 10%, while VA increased by 15% for the Harris pnps, but decreased by 5% for the A T & T pnps and by 25% for the VTC pnps. The noise comparisons are shown in table 1, comparing the series noise (i.e., the equivalent noise voltage or the equivalent noise resistance) for the three processes, both before and after 5 Mrad radiation. The low-frequency " n o i s e c o m e r " , fc, for both a minimum size (1 x 1 emitter, or 1 x device) and a large size (16 × ) device are compared for both npn and pnp devices. Measurements of the equivalent noise currents (ENI) for the devices have not yet been made. Using the HP4145B parameter analyzer, we were able to obtain an accurate S P I C E model for each bipolar process, both before and after radiation. A comparison was made of each process to simulate a basic low-power, 20 mW, folded-cascode CS preamp [2] for use with a 2 cm × 2 cm silicon calorimeter detector (Cde t - - 1 1 2 pF). The results for the rise time, power dissipation, and anticipated equivalent noise charge (ENCe) for each pre-
(Beta) for the smallest area npn devices (an emitter size of 1 x 1, approximately 10 Ixm x 10 Fm), while fig. 6 compares fl for the small area (1 x 1) pnps, although the VTC unit tested was available in only a 1 x 9 area.
120
-Pre-radiatien (solid lines) Post-radiation, 5 Mrad (dashed lines)
100
< F-U.I m
HARRIS
60 40 .
. ........
~%i
.......
__------_- .. .-.-. ". .-. . . . . . . 10-6
10-5
.......
10-4
"'"
VTC
'
10-3
10-2
Collector Current ( Amp. )
Fig. 6. Beta vs I c for pnp high-frequency transistors. (VcE = -2.5 V).
Table 1 Series noise comparisons for bipolar processes [Ic = 1 mA, I VEcI = 2.5 V]. Process
Noise corner frequency fc [kHz] (lx) Prerad
NPNs
AT&T Harris VTC
2.4 0.6 0.35
PNPs
AT&T Harris VTC
5.0 6.0 -
R n
(16×) Postrad 2.4 2.4 0.55 6.0 10.0 -
(midband) [£]
(1×)
(16×)
Prerad
Postrad
6.0 2.0 2.5
10.0 2.8 3.2
76 550 455
55 62 52
5.0 1.7 (8 x ) 1.0
5.5 2.4 (8 X ) 1.5
51 148 (8 × ) 78
16 26 (8 x )78
456
E.J. Kennedy et al. / Radiation effects on JFETS, MOSFETS, and bipolar transistors IE4
amplifier are compared in table 2, both before and after radiation.
API~R 1.3 RRad A - - A Sample 2 O i00 ~A
4. Gamma radiation effects on M O S F E T S
Although a special radiation-hard C M O S process will be required for SSC circuits, we wanted to see how a typical MOSIS C M O S process would survive a g a m m a dose. Special N M O S and PMOS dual arrays having w i d t h / l e n g t h ratios of 100 p.m/10 ~ m and 1000 ~tm/10 t~m were designed using the M A G I C layout program. The chips were fabricated under the standard 2 ~tm p-well Orbit Semiconductor process, using the MOSIS fabrication program. All transistors were fabricated in a " q u a d " configuration, with diagonal transistors connected to obtain an overall well-matched dual transistor configuration, which is indicative of the construction required for low offset-voltage differential front-end devices. The dual transistors were biased in a normal conductive state (VDs = 2.5 V, I D = 1 mm, 100 p.A and 10 ~A) during the irradiation. The threshold voltage VT was monitored, and devices were removed when the observed threshold voltage shift for the P M O S units had shifted to approximately 2V-r (prerad), which occurred at a total dose of 1.3 Mrad. It was found that the P M O S devices had an average shift in VT from - 0 . 7 to - 1.5 V, while N M O S devices had VT shifts from 0.9 to 0.6 V, although undoubtedly the worst-case negative shift from the initial 0.9-V value was much larger due to the fixed oxide charge (Not), before the interface-trapped charge (Nit) effect produced a positive Vv trend. An example of the large increase ( - 5 × ) in the lowfrequency noise due to radiation is shown in fig. 7 for an N M O S device with W / L = 100 ~tm/10 p.m. Of most significance was the large increase in the offset voltage Vos of the dual units, with prerad values of < 1 mV for both n- and p-channel units observed for 1000 ~ m / 1 0 ~tm devices, while in a postrad measurement Vos values of > 100 mV were obtained. Further, the shift in gm (at 1 mA) was significantly higher than that observed in JFETs, approximately - 1 2 % for P M O S and - 1 9 % for N M O S units after radiation. The Early voltages, similar to those for JFETs, increased after radiation from - 20 V to - 60 V for P M O S units, and from - 30 V to - 150 V for N M O S units.
.~ O Z
100
• - - ~ S ~ p l e 2 0 lObot 10 hot
i00
i000
Frequency
~ o
IE4
Fig. 7. Noise spectra for MOSIS NMOS transistor with W / L =100 ~,m/10 p.m, both before and after a 1.3-Mrad (6°Co) total dose.
noise spectrum which was then measured by the HP3400A rms meter. After the rms value of the noise for a particular frequency was measured, a test signal having the same frequency as the center frequency of the band-pass filter was then applied to the system thus allowing calculation of the system noise gain. F r o m this information the data acquisition system calculated the series noise resistance ( R n ) of the J F E T being tested and stored the result. The process was repeated for all measured points in the spectrum. The error in the J F E T noise measurement due to the preamplifier noise contribution was calculated and verified with PSPICE to be on the order of _+ 2%, worst case. The accuracy of the variable band-pass filters was confirmed by using a calibrated noise source and was found to be repeatable and accurate with less than + 1% worst case error. The integration time of the rms meter was extended by converting the rms voltage measurement to rms power and taking the average of several
5. N o i ~ m e ~ e m e n t s
A block diagram of the noise measurement system used to acquire the drain current noise of the J F E T s is shown in fig. 8. The preamplifier was carefully designed to have low noise and sufficient gain to dominate the noise of the subsequent gain stage. The variable bandpass filter amplified only a narrow bandwidth in the
IE5
(Hz)
Fig. 8. Noise measurement system.
E.J, Kennedy et al. / Radiation effects on JFETS, MOSFETS, and bipolar transistors
samples. The number of samples required to produce measurements that were repeatable with an accuracy of greater than 95% was empirically found for each center frequency of the variable band-pass filter. This required as many as 10000 samples for the 10 Hz and 20 Hz center frequencies. A similar procedure was used for the MOSFET measurement except that fewer samples were used for a given frequency. The estimated error for these measurements was approximately + 15% worst case at the lowest frequencies.
457
when compared with an FET and shows promise as a useful technology for the short peaking times that will be required for the SSC.
Acknowledgements The authors acknowledge the various semiconductor companies, namely InterFet, AT&T, Harris, and VTC, Inc., for providing sample transistors and arrays for testing. We also thank M. Kemper for assistance in measurements and testing.
6. Conclusions Radiation using a 6°Co gamma source has shown a significant increase in the low-frequency noise for both JFETs and MOSFETs. JFETs are decidedly superior to MOSFETs where constancy of gm and threshold voltage are concerned. Although the Beta degradation is severe, particularly for pnp transistors, the highfrequency bipolar transistor offers very low series noise
References [1] D.E. Groom (ed.), Radiation Levels in the SSC Interaction Regions, SSC-SR-1033 (1988). [2] C.L. Britton, Jr. et al., Bipolar Monolithic Preamplifiers for SSC Silicon Calorimetry, presented at the Symp. on Detector Research and Development for the SSC, Fort Worth, TX, October 17, 1990.