III. Gm-C Filtering - Epublications - Université de Limoges

III.4 Comparison of the filters
III.4.a Comparison of the Performances
Table 33 summarizes the performances obtained with the different filter and a
compares them with [III.1].
Table 15. Performances Comparison
Filter [III.1] Filter 1 Filter 2 Units
Linearization tech. DSD* DSD* MGTR -
Tuning range 50 - 300 45 - 385 45 - 450 MHz
Q factor 6 3.5 3.7 -
Gain 6 10 6 dB
NF 20 18 16.3 dB
IIP3 2 to 6 -4 to 2 8 to 10 dBm
Power
consumption
7.6 16.5 30 mW
Supply 1.2 2.5 2.5 V
CMOS Technology 130 65 65 nm
- 115 -
*DSD stands for Dynamic Source Degeneration
Filter 1 and Filter 2 present a larger tuning range than the filter reported in [III.1]. The
high-end of the band is different between Filter 1 and Filter 2 due to a difference in the gm
value. Filter 1 uses a 2.5mS transconductance whereas Filter 2 uses a 10mS gm. This
difference is also visible on the NF value, which is higher when using smaller gm, despite a
higher filter gain.
As already mentioned, the linearity of Filter 1 is not high enough to compete with the
other two filters. Besides, the filter gain is 4dB larger, which would involve tighter
constraints on the LNA stage. Compared to reference [III.1], Filter 2 presents a higher
dynamic range due both to its lower selectivity and to its particular linearization technique.
This dynamic range is obtained at the cost of a higher power consumption, which also
enables a larger input voltage swing (2.5V supply).
Compared to reference [III.1], Filters 1 and 2 linearity is smaller. A first possible
explanation is that here, simulations have been performed with 1MHz spaced tones whereas
measure in [III.1] used 10MHz spaced ones. In the latter case, the third order intermodulation
products may be partially filtered out, which finally increases the intercept point value.
Furthermore, in the present case, NF is lower. Hence, by means of the use of higher
transconductances than in [III.1], the trade-off linearity versus noise is a bit shifted towards
lower noise.