info-vax (12/05/82)
>From Fathi.ES@PARC-MAXC Sun Dec 5 08:54:48 1982
Mail-From: ARPANET host PARC-MAXC rcvd at 3-Dec-82 1659-PST
To: INFO-VAX@SRI-CSL
Cc: es@Fathi, es@BKAO
Remailed-Date: 4 Dec 1982 1047-PST
Remailed-From: the tty of Geoffrey S. Goodfellow <Geoff5 at SRI-CSL>
Remailed-To: Info-VAX@SRI-CSL: ;
**** SPICE2G5 AC ANALYSIS BUG ***
While running the latest version of SPICE2 (G5), we noticed that for the ac
analysis KCL is violated. As we know, a MOS transistor is a 4 terminal device,
which the algebraic sum of currents at Drain, Gate, Source and Bulk (Substrate)
should be zero.
1. This problem only exists in AC analysis (DC and Transient produce correct
results)
2. On the previous version of SPICE2 (version D) the AC analysis is correct
and KCL is not violated. (Please note that the capacitance model on SPICE2G5 is
Meyer's model which indeed is different from SPICE2D).
3. I have used the equivalent linearized, small-signal MOSFET model in
place of the tested transistor, and again KCL was correct.
I heard from different sources that Berkeley knows about this bug and they have
already fixed it.!!
Does anyone out there knows about these changes (They should be in overlay
ACAN and in routine ACLOAD were the complex coefficient matrix is loaded) ,
or where I can get a copy of Berkeley's enhancements.
I will appreciate any help or suggestions .
Thanks,
Nader Fathi
Fathi.ES at PARC-MAXC
VLSI CAD Development
Xerox Corporation, MS A1-85
701 S. Aviation Blvd.
El Segundo, CA 90245
The following SPICE2 run demonstrates such bug :
1*******11-NOV-82 ******* SPICE 2G.5 (10AUG81) *******12:28:34*****
0THIS IS THE KCL TEST
0**** INPUT LISTING TEMPERATURE = 27.000 DEG C
0*******************************************************************
****
.WIDTH OUT = 72
IIN 0 7 AC 1
*
*... THESE ARE DUMMY VOLTAGE SOURCES TO MEASURE CURRENTS
*
VG 7 1 0
VD 2 4 0
VS 3 4 0
VB 6 0 0
VE 4 5 0
R 7 0 1G
*
*... A VOLTAGE CONTROL VOLTAGE SOURCE WITH GAIN OF 1000
*
E1 5 0 7 0 -1000
*
*... AND HERE COMES THE TRANSISTOR AS A 4 TERMINAL DEVICE
*
M1 2 1 3 6 ENH W=70U L=120U
.MODEL ENH NMOS (LEVEL=2 NSUB=3E15 TOX=0.1375U UEXP=0.36
+ NSS=4E10 LD=0.1N XJ=1U CGSO=2.125E-10 CGDO=2.125E-10)
*
*... AS KCL STATES:
* A. I(VE)=I(VD)+I(VS)
* B. 1.0 AMP + I(VB) = I(VE)
.PRINT AC I(VG),I(VD),I(VS),I(VE),I(VB)
.AC DEC 10 10K 200K
.END
1*******11-NOV-82 ******* SPICE 2G.5 (10AUG81) *******12:28:34*****
0THIS IS THE KCL TEST
0**** MOSFET MODEL PARAMETERS TEMPERATURE = 27.000
DEG C
0*******************************************************************
****
ENH
0TYPE NMOS
0LEVEL 2.000
0VTO 0.504
0KP 1.51D-05
0GAMMA 1.257
0PHI 0.633
0CGSO 2.12D-10
0CGDO 2.12D-10
0CJ 1.76D-04
0TOX 1.38D-07
0NSUB 3.00D+15
0NSS 4.00D+10
0TPG 1.000
0XJ 1.00D-06
0LD 1.00D-10
0UO 600.000
0UEXP 0.360
1*******11-NOV-82 ******* SPICE 2G.5 (10AUG81) *******12:28:34*****
0THIS IS THE KCL TEST
0**** SMALL SIGNAL BIAS SOLUTION TEMPERATURE = 27.000 DEG
C
0*******************************************************************
****
NODE VOLTAGE NODE VOLTAGE NODE VOLTAGE NODE
VOLTAGE
( 1) 0.0000 ( 2) 0.0000 ( 3) 0.0000 ( 4) 0.0000
( 5) 0.0000 ( 6) 0.0000 ( 7) 0.0000
VOLTAGE SOURCE CURRENTS
NAME CURRENT
VG 0.000D+00
VD 0.000D+00
VS 0.000D+00
VB 0.000D+00
VE 0.000D+00
TOTAL POWER DISSIPATION 0.00D+00 WATTS
1*******11-NOV-82 ******* SPICE 2G.5 (10AUG81) *******12:28:34*****
0THIS IS THE KCL TEST
0**** OPERATING POINT INFORMATION TEMPERATURE = 27.000
DEG C
0*******************************************************************
****
0
0**** VOLTAGE-CONTROLLED VOLTAGE SOURCES
0 E1
V-SOURCE 0.000
I-SOURCE 0.00E+00
0
0**** MOSFETS
0 M1
0MODEL ENH
ID 0.00E+00
VGS 0.000
VDS 0.000
VBS 0.000
VTH 0.500
VDSAT 0.000
GM 0.00E+00
GDS 0.00E+00
GMB 0.00E+00
CBD 0.00E+00
CBS 0.00E+00
CGSOVL 1.49E-14
CGDOVL 1.49E-14
CGBOVL 0.00E+00
CGS 0.00E+00
CGD 0.00E+00
CGB 1.67E-12
1*******11-NOV-82 ******* SPICE 2G.5 (10AUG81) *******12:28:34*****
0THIS IS THE KCL TEST
0**** AC ANALYSIS TEMPERATURE = 27.000 DEG C
0*******************************************************************
****
FREQ I(VG) I(VD) I(VS) I(VE) I(VB)
X
1.000E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.905E-04
1.259E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.904E-04
1.585E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.903E-04
1.995E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
2.512E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
3.162E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
3.981E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
5.012E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
6.310E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
7.943E+04 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
1.000E+05 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
1.259E+05 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
1.585E+05 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
1.995E+05 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
2.512E+05 1.000E+00 8.848E-03 8.848E-03 1.770E-02 9.902E-04
This circuit measures the current at all these 4 nodes so that:
1.0 Amp+I(VB) = I(VE)=I(VD)+I(VS)=2*I(VD)=2*I(VS)
but we can see that : I(VB)=9.9E-4, I(VE)=1.77E-2
so 1+9.9E-4 is not equal to 1.77E-2 (PROBLEM !!?)
Observations:
1. This problem only exists in AC analysis (DC and Transient produce correct
results)
2. On the previous version of SPICE2 (version D) the AC analysis is correct
and KCL is not violated. (Please note that the capacitance model on SPICE2G5 is
Meyer's model which indeed is different from SPICE2D).
3. I have used the equivalent linearized, small-signal MOSFET model in
place of the tested transistor, and again KCL was correct.
I heard from different sources that Berkeley knows about this bug and they have
already fixed it.!!
Does anyone out there knows about these changes (They should be in overlay
ACAN and in routine ACLOAD were the complex coefficient matrix is loaded) ,
or where I can get a copy of Berkeley's enhancements.
I will appreciate any help or suggestions .
Thanks,
Nader Fathi
Fathi.ES at PARC-MAXC
VLSI CAD Development
Xerox Corporation, MS A1-85
701 S. Aviation Blvd.
El Segundo, CA 90245