Documente Academic
Documente Profesional
Documente Cultură
www.elsevier.com/locate/microrel
Research note
Analog circuit fault diagnosis based on noise measurement
Yisong Dai*, Jiansheng Xu
Department of Electronics Engineering, Jilin University of Technology, Changchun 130025, People's Republic of China
Received 19 January 1999; received in revised form 6 March 1999
Abstract
In this paper, an analog circuit fault diagnosis method using a noise measurement and analysis approach is
suggested. Compared to the conventional circuit fault diagnosis methods, this method can discover hidden and early
circuit fault caused by the device defects. Since circuit fault diagnosis is more dicult than device-defect detection,
in this paper the circuit output noise calculation, the comparison between the normal and failure conditions and the
circuit fault diagnosis method have been discussed. Finally, an example of an active ®lter circuit fault diagnosis has
been given by using this method. # 1999 Elsevier Science Ltd. All rights reserved.
0026-2714/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 6 - 2 7 1 4 ( 9 9 ) 0 0 0 2 9 - 3
1294 Y. Dai, J. Xu / Microelectronics Reliability 39 (1999) 1293±1298
ment results.
2Re K m _m
u
f K i
f Sem im
f
2. Circuit noise calculation using the noise superposition where Sej
f is the noise spectrum of thermal noise ej
theorem in jth impedance Zj , i.e., Sej
f 4kTRe
Zj Sem
f
and Sim
f are the noise spectra of the voltage noise
As mentioned above, the accurate output noise spec- em and the current noise im of mth semiconductor,
trum must be calculated for analog circuit fault diagno- Sem im
f is the cross spectrum of em and im , these
sis. The noise superposition theorem is the fundamental noise spectra can be calculated only in the medium fre-
theorem in linear circuit analysis. Using this theorem, quency range. At low frequency they must be obtained
we can obtain the noise contribution of every noise by measurement [17]. If there are p impedances and q
Y. Dai, J. Xu / Microelectronics Reliability 39 (1999) 1293±1298 1295
Fig. 2. The diagram of circuit: (a) the active ®lter circuit and (b) the En ÿ In noise model of operation ampli®er Op-27.
semiconductors, the noise superposition theorem at the To simplify the calculation, all transfer functions
output terminal of a linear circuit can be written as from every noise source to the output terminal and the
overall noise contribution of all impedances in the cir-
X
p
cuit can be calculated directly by Pspice program.
Suo
f Sej
f jK ju
f j2
j1
However, the noise contribution of every semiconduc-
X
q ÿ tor must be calculated using Eq. (2), because the En ÿ
Sem
f jK m 2 m
u
f j Sim
f jK i
f j
2 In noise model of a semiconductor must be obtained
m1 by the noise measurement setup.
Xq
ÿ If the analog circuit consists of an interconnection
2Re K m m
u
f K i
f Sem im
f of a two-port network, we can use the noise matrix
m1
superposition expression to simplify the noise calcu-
Since the correlation term Sem im
f between em and im lation. More details can be found in Ref. [15,16].
is included in Eq. (2), comparing with the conventional
noise power superposition method, Eq. (2) is more cor-
rect. If the correlation term Sem im
f between em and im
is negligible, Eq. (2) is equal to the conventional noise 3. Analog circuit fault diagnosis by noise measurement
power superposition, i.e., the noise power at the
measurement point is the sum of the noise power of The noise measurement method used to diagnose the
the individual noise source. analog circuit fault is executed as follows:
1296 Y. Dai, J. Xu / Microelectronics Reliability 39 (1999) 1293±1298
Table 1
Noise models of Op-27
the transistor one, three problems have been discussed reliability in electronic devices. Advances in Electronics
in this paper: and Electron Physics 1993;87:205±57.
[6] Vandamme LKJ. Noise as a diagnostic tool for quality
1. A noise spectrum superposition theory has been and reliability of electron devices. IEEE Trans
developed for calculating the output noise value of Electronics Devices 1994;41:2176±8.
a complex analog circuit. Since the correct En ÿ In [7] Jevtic MM. Noise as a diagnostic and predication tool in
noise model of each transistor in the circuit is reliability physics. Microelectron Reliab 1995;35(3):455±
obtained by a noise measurement setup, this method 77.
can obtain the correct output noise spectrum. [8] Tsong-Ming Chen, Abdullab M Yassine. Electrical noise
2. Through the output noise measurement of the ana- and VLSI interconnect reliability. IEEE Trans Electron
Devices 1994;41(11):2165±72.
log circuit, if the output noise spectrum deviation
[9] Jones BK, Xu YZ. Excess noise as an indicator of digital
from the normal value is larger than tolerance of
integrated circuit reliability. Microelectron Reliab
the normal noise value, we may conclude that this 1991;31(2/3):351±61.
analog circuit is a failure. [10] Ursutiu D, Jones BK. Low frequency noise used as a
3. For diagnosing and locating the failure component lifetime test of LEDs. Semicond Sci Technol
of the analog circuit, we must select several output 1996;11:1133±6.
noise measurement terminals or only parts of circuit [11] Zeynep Celik-Butler. Prediction of electromigation failure
to perform noise measurements and comparisons in W/Al±Cu multilayered metallizations by 1/f noise
until the correct diagnosis and location of the failure measurements. Solid-State Electronics 1992;35(9):1209±
component is obtained. 12.
[12] Vossen JL. Screening of metal ®lm defects by current
noise measurements. Appl Phys Lett 1973;23(4):287±9.
[13] Yisong Dai. A dynamic noise measurement technique
References used to estimate activation energies for failure mechan-
isms of a transistor. Microelectron Reliab
[1] Liu RW. Testing and diagnosis of analog circuits and 1993;33(15):2207±15.
system. New York: Van Nostrand Reinhold, 1991. [14] Motchenbacher CD, Fitchen FC. Low-noise electronic
[2] Lin PM, Elecherif YS. Analog circuits fault dictionaryÐ design. New York: Wiley, 1973.
new approaches and implementation. Int J Circuit [15] 9999 Dai. Noise performance analysis of bipolar oper-
Theory and Appl 1985;13:149±72. ational ampli®er based on the noise matrix superposition
[3] Bander JW, et al. A nonlinear L1 optimization algorithm expression. IEE Proc-Circuits Devices Syst
for design modeling and diagnosis of network. IEEE 1998;145(5):343±8.
Trans CAS 1987;34(2):174±81. [16] Yisong Dai, Zhihong Quin, Yaqin Li. The noise matrix
[4] Somayajula SS, Sanchez-Sinencio E, Pineda de Gyvez J. theory of two-port network. Chinese Sci Bull
Analog fault diagnosis based on ramping power supply 1997;42(19):1665±9.
current signature clusters. IEEE Trans Circuits and [17] Yisong Dai. A precision noise measurement and analysis
Systems II: Analog and Digital Signal Processing method used to estimate activation energies for failure
1996;43(10):703±12. mechanisms of a transistor. Microelectron Reliab
[5] Jones BK. Electrical noise as a measure of quality and 1997;33(15):2207±15.