TY - GEN
T1 - Design of well-behaved low-loss millimetre-wave CMOS transmission lines
AU - Amakawa, S.
AU - Orii, A.
AU - Katayama, K.
AU - Takano, K.
AU - Motoyoshi, M.
AU - Yoshida, T.
AU - Fujishima, M.
PY - 2014
Y1 - 2014
N2 - It is a challenge to design single-mode transmission lines for above 100 GHz following strict design rules of modern CMOS processes. This paper reports characteristics of three types of microstrip lines in 65 nm CMOS up to 325 GHz, designed with or without using an auto-dummy exclusion layer. The lowest-loss design among the three is a shielded microstrip protected with an exclusion layer. The metal density requirement is met, as is commonly done, by placing sidewalls as far from the signal line as allowed by the design rules. The other two designs are microstrips without sidewalls or the exclusion layer. One of them has high-density auto dummy fill inserted by the foundry and shows significantly higher attenuation than the shielded microstrip. The other is filled with low-density fill that prevents auto dummy fill from being inserted. It is only marginally lossier than the shielded microstrip. The microstrips without sidewalls are found to exhibit more well-behaved attenuation especially above 100 GHz. The frequency dependence of the attenuation of the shielded microstrip, on the other hand, exhibits ripples, indicating possible presence of spurious modes. Attenuation constants estimated by multiline TRL (thru-reflect-line) from lines of various lengths indicate that the longest line measured should be very long, perhaps 2mm or longer, for the estimates to be reliable.
AB - It is a challenge to design single-mode transmission lines for above 100 GHz following strict design rules of modern CMOS processes. This paper reports characteristics of three types of microstrip lines in 65 nm CMOS up to 325 GHz, designed with or without using an auto-dummy exclusion layer. The lowest-loss design among the three is a shielded microstrip protected with an exclusion layer. The metal density requirement is met, as is commonly done, by placing sidewalls as far from the signal line as allowed by the design rules. The other two designs are microstrips without sidewalls or the exclusion layer. One of them has high-density auto dummy fill inserted by the foundry and shows significantly higher attenuation than the shielded microstrip. The other is filled with low-density fill that prevents auto dummy fill from being inserted. It is only marginally lossier than the shielded microstrip. The microstrips without sidewalls are found to exhibit more well-behaved attenuation especially above 100 GHz. The frequency dependence of the attenuation of the shielded microstrip, on the other hand, exhibits ripples, indicating possible presence of spurious modes. Attenuation constants estimated by multiline TRL (thru-reflect-line) from lines of various lengths indicate that the longest line measured should be very long, perhaps 2mm or longer, for the estimates to be reliable.
UR - http://www.scopus.com/inward/record.url?scp=84904512282&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84904512282&partnerID=8YFLogxK
U2 - 10.1109/SaPIW.2014.6844526
DO - 10.1109/SaPIW.2014.6844526
M3 - Conference contribution
AN - SCOPUS:84904512282
SN - 9781479935994
T3 - 2014 18th IEEE Workshop on Signal and Power Integrity, SPI 2014 - Proceedings
BT - 2014 18th IEEE Workshop on Signal and Power Integrity, SPI 2014 - Proceedings
PB - IEEE Computer Society
T2 - 18th IEEE Workshop on Signal and Power Integrity, SPI 2014
Y2 - 11 May 2014 through 14 May 2014
ER -