Simplified RF noise de-embedding method for on-wafer CMOS FET

Abstract

A simplified RF noise de-embedding method for an on-wafer CMOS device using only one dummy structure for simplicity and area reduction is presented. The method describes the use of a 'thru' test structure to subtract completely the parasitic effects of pads and in/out interconnections from the device under test. A comparison of the de-embedding results among the simplified method and existing de-embedding methods is given, which proves that the new method is effective, accurate and time efficient.

Full text

Page 1

Copyright © 2007 Year IEEE. Reprinted from ELECTRONICS LETTERS 31st August
2007 Vol. 43 No. 18. This material is posted here with permission of the IEEE. Such
permission of the IEEE does not in any way imply IEEE endorsement of any of Institute
of Microelectronics’ products or services. Internal or personal use of this material is
permitted. However, permission to reprint/republish this material for advertising or
promotional purposes or for creating new collective works for resale or redistribution
must be obtained from the IEEE by writing to pubs-permission@ieee.org.

Page 2

Simplified RF noise de-embedding method
for on-wafer CMOS FET
Y.-Z. Xiong, A. Issaoun, L. Nan, J. Shi and K. Mouthaan
A simplified RF noise de-embedding method for an on-wafer CMOS
device using only one dummy structure for simplicity and area
reduction is presented. The method describes the use of a ‘thru’ test
structure to subtract completely the parasitic effects of pads and in=out
interconnections from the device under test. A comparison of the
de-embedding results among the simplified method and existing
de-embedding methods is given, which proves that the new method
is effective, accurate and time efficient.
Introduction: AssiliconCMOStechnologiesarepushedfurthertowards
nanoscale dimensions, new device noise sources emerge, such as random
telegraph signal (RTS) noise at low frequencies and thermal noise and
induced gate noise at high frequencies. As a result, the accuracy of noise
models used by circuit designers in the prediction of high speed, RFand
microwave circuit performances becomes questionable and noise model
improvement is called for. Getting accurate and reliable device noise
measurement data is the first challenge faced by modelling engineers.
Various noise de-embedding methods have been developed by different
research groups [1–4]. The majority of the proposed noise de-embedding
methods use more dummy test structures and a device under test (DUT)
for transforming the measured noise raw data to the device noise.
Moreover, the calculations involved are complicated and time consum-
ing. The method reported in [4] uses only two test structures (THRU and
DUT), but the involved extraction procedure of the transmission line
parameters is too complicated if not impossible for a short ‘thru’ line. In
this Letter, a simplified RF noise de-embedding method is introduced
using only one dummy structure, THRU line.
Fig. 1 DUT and de-embedding structures – THRU
a DUT
b THRU
c THRU schematic
d THRUþDUT schematic
Theory and procedure: The simplified noise method is shown inFig. 1
and it consists of two test structures, device and thru line, along with
their equivalent-circuit representations used for noise de-embedding
purpose. The DUT test structure shows the device connected to
the input and output symmetrical interconnections and two signal-
ground-signal (GSG) pads. The DUT’s source port is connected to
ground with metals having short length and large width. Therefore, the
source leg inductance and resistance can be ignored. The thru line is
constructed from the DUT test structure by deleting the device and
ensuring that the input and output interconnections are electrically equiva-
lent such that the representation of Fig. 1c applies. Hence, the dummy
structure THRU can be easily modelled as input and output equivalent-
circuit sections and their circuit element values are given in [5]:
Y1¼ Y11þ Y12
Z2¼ ?ð1=Y12Þ=2 ¼ ?ð1=Y21Þ=2
Y3¼ Y22þ Y21
ð1Þ
ð2Þ
ð3Þ
The proposed noise de-embedding procedure uses simple two-port and
noise correlation matrix transformations [6], and is described by:
1 Measure the scattering parameters {S dut}, {S thru} and the noise
parameters NFmin
matrix.
2 Owing to the symmetry of the THRU line structure, in order to
minimise measurement uncertainty errors, perform data averaging on
the {S thru}
bycomputing{S thru C}
(S11þS22)=2 and S12
elements of {S thru} and {S thru C}, respectively.
3 Convert {S thru C} to its admittance matrix {Y thru C}, then use
(1)–(3) to calculate Y1, Y3(here Y1¼Y3) and Z2.
4 Compute the noise correlation chain matrix {CA
the noise correlation admittance matrix {CY
5 Calculate CY
6 Calculate the correlation matrix {CY
and convert it to {Cz
de Y1 Y3}?Cz2
8 Convert {Cz} to {CA} and compute the de-embedded noise para-
meters NFmin, Goptand RN.
DUT, Gopt
DUTand RN
DUTof the DUT; where, {} denotes a
using
S11
0¼
S22
0¼
0¼ S21
0¼ (S12þS21)=2, where Siiand Sii
0are the
DUT} and convert it to
DUT}.
Y1¼CY
Y3¼2kT<(Y1).
de Y1 Y3¼{CY
????where Cz
DUT}?CY1
Y
0
0
CY3
Y
?????
?????
de Y1 Y3}.
7 Calculate {Cz}¼{Cz
z
0
0
Cz2
z
????
z2¼2kT<(Z2).
Fig. 2 Measured and de-embedded RF noise data for 12-finger device at
Vgs¼0.7 V and Vds¼1.8 V bias points
a Minimum noise figure NFmin
b Noise resistance RN
c Magnitude of optimised source reflection coefficient jGoptj
d Phase of optimised source reflection coefficientffGopt
s measured raw data
u de-embedded data using thru only
D de-embedded data using open-short
, de-embedding data using two-line
Experiment and discussion: To assess the present noise de-embedding
algorithm and the accuracy of the proposed method, the required
dummy structures, including OPEN, SHORT, THRU, THRU1
(line1) and THRU2 (line2) [2, 7], were fabricated using a standard
0.18 mm CMOS technology. As a test sample, a 12-finger device with
gate length of 0.18 mm and gate width of 5 mm was selected. The input
and output interconnection width and length were set to 10 and
50 mm, respectively. The GSG pads were constructed using three
ELECTRONICS LETTERS31st August 2007Vol. 43 No. 18

Page 3

pads of dimensions 50?50 mm. The scattering and noise parameters
were measured using the noise measurement system WinNoise from 2
to 12 GHz. Because the frequency of investigation is up to 12 GHz
only, the distributed effects of the interconnections can be ignored.
Fig. 2 shows the measured and the de-embedded noise parameters
using three approaches for a bias point of Vgs¼0.7 Vand Vds¼1.8 V.
In the frequency range 2–12 GHz, the de-embedded minimum noise
figures NFmin are very close. The noise resistance RN using the
proposed method (thru only) is in between those of the method
called two-line described in [2] and the method called open-short
described in [7]. The differences of de-embedded noise resistance in
thru only and two-line, thru only and open-short are within 3.3 and
1.2%, respectively. The slight differences of noise resistances at
12 GHz between thru only and open-short, thru only and two-line
are 2.7 and 1.0 O, respectively, which is by ignoring source resistance
and inductance. The de-embedding magnitude and phase of the
optimised source reflection coefficient jGoptj and ffGopt are almost
the same as open-short and slightly different from two-line in the
performed frequency range.
Conclusions: An effective, accurate and time efficient RF noise
de-embedding method has been presented. Only one dummy test
structure is needed for the purpose of de-embedding. The results show
that the accuracy of the proposed method is comparable with the
methods of open-short and two-line, and prove the usefulness of the
simplified method for de-embedding noise parameters of a CMOS
FET.
Acknowledgment: The authors thank F. Lin, J. Brinkhoff, W. G. Yeoh
and R. Singh for the in assistance and support.
# The Institution of Engineering and Technology 2007
17 May 2007
Electronics Letters online no: 20071442
doi: 10.1049/el:20071442
Y.-Z. Xiong, A. Issaoun, L. Nan and J. Shi (Institute of Micro-
electronics, 11 Science Park Road, Singapore Science Park-II,
Singapore)
E-mail: yongzhong@ime.a-star.edu.sg
K. Mouthaan (Department of Electrical and Computer Engineering,
National University of Singapore)
References
1Deen, M.J., Chen, C.H., Asgaran, S., Rezvani, G.A., Tao, J., and
Kiyota, Y.: ‘High-frequency noise of modern MOSFETs: compact
modeling and measurement issues’, IEEE Trans. Electron Devices,
2006, 53, (9), pp. 2062–2081
Chen, C.H., and Deen, M.J.: ‘A general noise and S-parameter
de-embeddingprocedurefor
measurements of MOSFETs’, IEEE Trans. Microw. Theory Tech.,
2001, 49, (5), pp. 1004–1004
Guo, J.-C., and Lin, Y.-M.: ‘A new lossy substrate de-embedding method
for Sub-100 nm RF CMOS noise extraction and modeling’, IEEE Trans.
Electron Devices, 2006, 53, (2), pp. 339–347
Cho, M.H., Huang, G.W., Chen, K.M., and Peng, A.S.: ‘A novel cascade
based de-embedding method for on-wafer microwave characterization
and automatic measurement’, IEEE MTT-S Int. Microw. Symp. Dig.,
2004, pp. 1237–1240
Song, J., Ling, F., Flynn, G., Blood, W., and Demircan, E.: ‘A
de-embeddingtechniquefor
Performance of Electronic Packaging Conf. (EPEP), Cambridge, MA,
USA, October 2001, pp. 129–132
Hillbrand, H., and Russer, P.H.: ‘An efficient method for computer-aided
noise analysis of linear amplifier networks’, IEEE Trans. Circuits Syst.,
1976, CAS-23, pp. 235–238
Aufinger, K., and Bock, J.: ‘A straightforward noise de-embedding
method and its application to high-speed silicon bipolar transistors’.
Proc. ESSDERC, 1996, pp. 957–960
2
on-wafer high-frequencynoise
3
4
5
interconnects’.Proc.Electrical
6
7
ELECTRONICS LETTERS 31st August 2007Vol. 43No. 18