On the use of body biasing to improve linearity in low LO-power CMOS active mixers
Introduction
Down-conversion mixers are one of the important building blocks of wireless communication systems. These mixers are typically used to translate the frequency spectrum of the radio-frequency (RF) received signal of the antenna to a lower or zero intermediate frequency (IF). Important design specifications of a down-conversion mixer include gain, noise figure, linearity, power, and port-to-port isolation. Similar to any other block in a receiver, there is a trade-off among these design parameters. For the building blocks of the receive path, as the blocks get farther from the antenna, their contribution to the noise figure would be lowered by the aggregated gain of their preceding stages (Friis׳ equation [1]). Thus, the higher the gain of the down-conversion mixer, the lower the noise contribution of the stages after the mixer (e.g., baseband amplifiers in a generic integrated receiver) will be. However, the higher the gain of the mixer, the more pronounced the nonlinearity distortions of the subsequent stages would be. This is due to the fact that as the blocks get farther from the antenna, their contribution to the nonlinearity of the receiver would be boosted by the aggregated gain of their preceding stages [2].
Due to the relatively high noise figure of active mixers, the gain of the LNA is typically chosen to be high enough to keep the overall noise figure at an acceptable level. However, this leads to more restrictions on the linearity of the mixer as its IIP3 is scaled down by the gain of the LNA. Thus, it is important to enhance the linearity of the mixer such that the high gain of LNA does not limit the linearity of the whole receiver. Unfortunately, the scaling of CMOS technology which results in lower supply voltages, makes the design of highly linear active mixers challenging.
Power consumption is another performance metric which is of great importance in low-power as well as portable wireless receivers. Thus, in low-power designs, it is desirable to reduce, as much as possible, the power consumption of each single block in the receiver including down-conversion mixers. However, it is challenging to decrease the power without compromising the noise figure and linearity. To keep the power consumption of the receiver to the minimum, it is also desired that the mixer operates with small LO signal levels. This is due to the fact that the lower the amplitude of LO signal, the less power would be required in the VCO or frequency synthesizer to generate the LO signal.
In this paper, we build on our previous work in [3] and propose the idea of enhancing the linearity of down-conversion mixers through incorporating body biasing in the LO-stage transistors. Conventionally, for large LO systems, linearization techniques are commonly applied to the RF stage of the mixer (e.g., see [4], [5], [6], [7], [8]) due to the fact that with the abrupt switching of LO stage, most of the distortion of the mixer is from the RF stage [6], [9], [10]. However, if the LO power is low and the transistors in the LO stage are not instantaneously switched on and off, the LO stage remains in active region for a considerable amount of time during the LO period, and thus the nonlinearity contribution of the LO stage on the mixer distortion should be investigated. To do this, we first overview the structure of the conventional current-commutating mixers in Section 2. Section 3 discusses the linearity of mixers with low-power LO which in turn have non-ideal switching in their LO stage. 5 Using body biasing for linearity enhancement, 6 Effect of body biasing on conversion gain present the application of body biasing to improve linearity and conversion gain in this type of mixers. In order to validate the proposed idea, two proof-of-concept prototypes have been designed and fabricated in 0.13-µm CMOS; one is a standard double-balanced mixer and the other one is a current-bleeding double-balanced mixer. Measurement results of these two prototypes are presented in Section 7. Conclusion remarks as well as comparison with the state-of-the-art mixers are presented in Section 8. Note that while the focus of this work is on the linearity improvement of down-conversion mixers; the results can also be extended to up-conversion mixers as well.
Section snippets
Current-commutating active down-conversion mixers
The most commonly used structure for active CMOS mixers is the current-commutating architecture [2], [11]. The simplified single-balanced version of such mixer is shown in Fig. 1(a). In this mixer, transistor M1, which is typically referred to as the RF stage transistor, operates as a transconductance stage (gm stage) which converts the RF input voltage signal to a current signal. Assuming that LO signal has a large enough amplitude, transistors M2 and M3 act as differential current switches
Non-ideal switching in active down-conversion mixers
For the mixer shown in Fig. 1, ideal current commutation happens if the LO signal is large enough to fully divert the tail current between the branches. Under this condition, LO-stage transistors operate in either off or triode region and produce a current with a square-shape waveform. However, in practice in many applications, the LO signal does not abruptly switch the entire tail current between the differential branches.
For example, in low-power wireless transceivers where the power level of
Nonlinearity analysis in active mixers with non-ideal switching
As mentioned before, in many practical cases, especially in mixers with low LO power, the LO-stage transistors do not switch abruptly and are operating in their active region for a considerable portion of the LO period. The larger the value of Δ in Fig. 2, the longer the LO-stage transistors operate in their active region and the lower will be the conversion gain of the mixer due to the saturation of conversion gain imposed by the nonlinearities shown in Eq. (1). During the time that LO-stage
Using body biasing for linearity enhancement
From Eq. (9), to decrease the nonlinearity in a mixer with a realistic (non-ideal) switching signal, the third-order nonlinear term of the current of transistors in both RF and LO stages, namely, gm3,RF and β3, should be minimized. Moreover, as the contribution of the nonlinearity of the LO stage is magnified by the gain of the RF stage, special attention should be paid to the ratio of third-order and first-order terms of the current of the transistors of the LO stage, namely, . Note that
Effect of body biasing on conversion gain
In addition to linearity enhancement, depending on the type of the mixer, body biasing at the LO stage can also improve the conversion gain of the mixer. To further explain this potential conversion-gain enhancement, we first review one of the gain degradation sources in mixers, namely, the total capacitance seen at the drain of the RF stage transistor(s). Fig. 4 depicts this capacitance, namely, CX, which is the sum of gate-source capacitance (Cgs) and source-bulk capacitance (Csb) of LO-stage
Experimental results
To verify the validity of the proposed body-biasing technique, two proof-of-concept double-balanced mixers are designed and implemented in IBM CMRF8SF 0.13-µm CMOS process (8 metal layers). The micrograph of the fabricated body-biased mixers are shown in Fig. 5. The schematic of the first prototype mixer (Fig. 5(a)) is presented in Fig. 6. The mixer uses degeneration inductors (Ls) in order to improve the linearity of the RF stage. Note that the mixer is intended to operate with a low LO power.
Conclusion
In this paper, a method to enhance the linearity of active mixers through body biasing of LO stage transistors is presented. The technique is particularly suitable for low-power mixers with low power LO drive signals and gradual LO switching. To verify the effectiveness of the proposed technique, two proof-of-concept prototypes are fabricated in a 0.13-µm CMOS technology. The first prototype is a standard double-balanced mixer structure while the second one is a mixer that uses a
References (21)
Noise gures of radio receivers
Proc.Inst. Radio Eng. (IRE)
(1944)RF Microelectronics
(2011)- H. Rashtian, A.H. Masnadi Shirazi, S. Mirabbasi, Improving linearity of CMOS Gilbert-cell mixers using body biasing,...
- K. Wang, K. Ma, W. Ye, K.S. Yeo, H. Zhang, Z. Wang, A low voltage low power highly linear CMOS quadrature mixer using...
- M.-F. Huang, C.J. Kuo, A CMOS dual class-AB technique for highly linear even harmonic mixer, in: Proceedings of IEEE...
- et al.
A new linearization technique for CMOS RF mixer using third-order transconductance cancellation
IEEE Microw. Wirel. Compon. Lett.
(2008) - et al.
Improved Gilbert mixer with high-precision automatic sweet-spot biasing and active-inductor-based harmonic suppression
Electron. Lett.
(2012) - et al.
An ultra-wideband high-linearity CMOS mixer with new wideband active baluns
IEEE Trans. Microw. Theory Tech.
(2009) The micromixer: a highly linear variant of the Gilbert mixer using a bisymmetric class-AB input stage
IEEE J. Solid-State Circuits
(1997)- et al.
A new RF CMOS Gilbert mixer with improved noise figure and linearity
IEEE Trans. Microw. Theory Tech.
(2008)
Cited by (10)
UWB down-conversion mixer using an IM3 cancellation modified technique for zero and low IF applications
2021, Microelectronics JournalCitation Excerpt :Afterwards, the produced signal from the paralleled transistors is injected into the tail current source of the transconductance stage and then is multiplied by RF input signals. Two third-order intermodulation signals with opposite sign can cancel each other, resulting in IIP3 improvement [15]. Mainly, both of two adjacent channels can distort side-channel due to third-order intermodulation mechanism in the entire operating bands [16].
Linearity improvement of RF mixer using double-linearization for 5 GHz applications
2019, AEU - International Journal of Electronics and CommunicationsCitation Excerpt :In [10,15], conversion gain is enhanced while the IIP3 is compromised. The body biasing [6], self-biasing current reuse [12], and dynamic current injection [14] achieve moderate performance for conversion gain, NF, linearity, and low power consumption. In harmonic reinjection [13] and dynamic current injection [14], additional circuitry is required for injecting the harmonic current resulting in more noise contributed to the IF frequency.
A 261-µW ultra-low power RF mixer with 26-dBm IIP3 using complementary pre-distortion technique for IEEE 802.15.4 applications
2019, AEU - International Journal of Electronics and CommunicationsCitation Excerpt :Thus, the current challenge in the wireless industry is to develop circuit techniques that consume low power and low silicon area without compromising the system’s linearity performance. Some of the techniques for improving linearity of the RF mixer include derivative super-position [11], multi-gate transistors [12], linear duplication [13], resistive current reuse [14], feedforward compensation [15], body effect in LO switch [16], self-biasing current reuse [17], second harmonic reinjection [18,19], dynamic current injection [20]. Other techniques for improving the linearity of the transconductance stage include harmonic termination [21], optimum biasing [22], feedforward [23], derivative superposition [24], complementary derivative superposition [25], differential derivative superposition [26], modified derivative superposition [27], IM2 injection [28], noise and distortion cancellation [29], post-distortion cancellation [30], and pre-distortion cancellation [31].
Design of Low Voltage Integrated CMOS RF front-end based LNA and Mixer for GPS Application
2017, Materials Today: ProceedingsA zero charge-pump mismatch current tracking loop for reference spur reduction in PLLs
2015, Microelectronics JournalCitation Excerpt :Charge-pump architectures with negative feedback using high gain OPAMPs to minimize the mismatch current are presented in [22–26], but the non-ideal effects of OPAMPs such as stability and offset voltage limit the performance of CP. Bulk-driven circuit techniques are highly useful in the design of ultra low voltage analog circuits [27–30] and to compensate for variation effects (PVT, die-to-die and within-die), and to reduce the leakage power in digital circuits [31,32]. In this paper, we present a charge-pump mismatch current calibration technique utilizing an adaptive body bias tuning of its current source transistors which provide a very fine resolution in the mismatch current calibration and a zero CP mismatch current tracking PLL architecture to minimize the static phase error and reference spurs.
A High-Gain Low-Power Low-Noise CMOS Transconductance Amplifier
2023, 2023 IEEE 5th International Conference on Power, Intelligent Computing and Systems, ICPICS 2023