Resonator 217
To further validate the high-Q of the DTL-SRR, EM simulation (Agilent
ADS momentum) is performed for STL-SRR and DTL-SRR structures shown
in Figure 9.3(a) and (b). The conductivity of the topmost metal layers M6
and M7 are 4.6 × 10
7
S/m, the metal layers M1∼M5 are 4.1 × 10
7
S/m and
the silicon substrate is 10 S/m according to the 65-nm CMOS process files.
The simulation res ult of reflection coefficient (Γ) is plotted against different
gap sizes in the Smith Chart as shown in Figure 9 .4. Note that the r esonance
happens when the imaginary part of Γ eq uals zero. One can observe that the
reflection coefficient of DTL-SRR at resonance frequency is much higher than
that of STL-SRR. More over, the reflection coefficient is increased for a smaller
gap size. For example, at the minimum gap of 2 µm that is allowed by the
design r ule , the reflection coefficient of the differential T-line is 10 .6% higher
than that of single-ended T-line. Since the minimum gap size is limited by
the design rule, the maximum r eflection coefficient one can obtain is around
0.9. The Q factor for both resonators are a lso compared by the reflection
coefficient as shown in Figure 9.5. As discussed, a high reflection coefficient
of DTL-SRR can be directly transferred into a high Q. One can observe that
the Q of DTL-SRR is around 20 ∼ 40% higher than that of STL-SRR with
the same gap size.
9.2.3 Comparison with Standing-Wave Resonator
The proposed DTL-SRR resonator is further compared with the standing-
wave r esonator using co planar strip line (CPS). As shown in Figure 9.2, they
are both designed under the same resonance frequency and are also provided
with floating metal shie lding to reduce substrate loss.
The optimization of the two structures is conducted with the full-wave
EM simulator (Agilent Momentum). As for DTL-SRR-bas ed metamaterial
resonator, the sta cked SRR unit-cell is designed with the top two metal layers
(M7, M6). M7 a nd M5 are used for the design of the host T-line and the
floating metal s trips for shielding of the two resonators, respectively. The sizes
of T-line, SRR and floating metal strips are carefully selected to obtain the
desired frequency. Moreover, for the CPS-based standing-wave resonator, its
Q factor also depends on the width a nd the s eparation of the T-line, the width
of the floating metal strip and the spacing between two adjacent floating metal
strips. Due to the parasitic capacitance of the cross-coupled NMOS tra nsistors
and the layout-dependent parasitic effect, the physical length of CPS is shorter
than the idea l length of λ/ 4. The detailed physical sizes are shown in Figure
9.2 and o ne can observe that the use o f SRR has 40% area reduction versus
the use of C PS.
Note that the Q of one resonator can b e described by
Q = ω
Average
energy stored
Energy loss/second
. (9.2)
As such, one can c ompare the Q fa ctors of the DTL-SRR with the standing-