Linearised bidirectional distributed amplifier with 20 dB IM3 distortion reduction

Abstract

Proposed is the design of a linearised CMOS bidirectional distributed amplifier (DA) that greatly suppresses the third-order intermodulation (IM3) distortion. The drain and gate transmission lines are staggered to purposely mismatch their time-delay and filter out the IM3 distortion. A modified CMOS cross-coupled cascomp circuit is proposed to enhance the linearity of the gain cell with a varactor-based active post nonlinear drain capacitance compensator for wider linearisation bandwidth. The proposed linearised CMOS bidirectional DA achieves a measured IM3 reduction of 20 dB in both RF directions, with ultrawideband 0.1 to 9.5 GHz frequency of operation.Introduction: In radio-over-fibre systems, a highly linear optical transmitter is necessary to achieve the required signal dynamic range. Nonlinearities cause harmonics and unwanted intermodulation products which degenerate the transmitter performance and must be minimised. Several linearised distributed amplifiers (DAs) have been reported in the literature [1 -6], however most of the published DA linearisation methods do not provide large third-order intermodulation (IM3) distortion reduction, and they either involve system-level DA linearisation with bulky components suffering from limited linearisation over broad bandwidth, or circuit-level DA linearisation that have narrow linearised bandwidth and apply only DC-based linearisation techniques.In this Letter, we propose a linearised CMOS bidirectional DA that combines device-level with circuit-level linearisation which greatly suppresses the IM3 distortion and achieves a measured IM3 distortion reduction of 20 dB over an ultra-wideband 0.1 to 9.5 GHz frequency of operation with a two-way amplification of 4.9 dB in both RF directions.

Analysis of circuit conditions for optimum intermodulation and gain in bipolar cascomp amplifiers with non‐ideal error correction

“…Mathematically, if we look at the output current for both the main and error amplifiers, we see the same gain components add together at the output of the complete Cascomp circuit [13]. Simple series expansions of the currents i 1 or i 2 are not possible as (8) and (11) are insoluble for these variables in terms of V IN(m) . Therefore we use an alternate method where the derivative of i 1 or i 2 with respect to V IN(m) is found by inverting the first derivatives of (8) and (11) with respect to the differential currents i 1 and i 2 .…”

“…Therefore we use an alternate method where the derivative of i 1 or i 2 with respect to V IN(m) is found by inverting the first derivatives of (8) and (11) with respect to the differential currents i 1 and i 2 . The following series expansions of the transfer functions (8) and 11, give (12) and 13which describe the main and non-ideal error amplifier gain components, A m and A e , respectively.…”

“…Many improvements to the topology were patented and used by Tektronix in the 1980s [3][4][5][6][7]. More recently, Cascomp style amplifiers, referred to as 'cross-coupled pairs', have been used in CMOS applications to eliminate third-order intermodulation (IM3), for example, see [8][9][10]. Other common distortion reduction methods, for example [11,12], do not trade-off as much gain as Cascomp topologies, suggesting that improvement of the fundamental gain is important in making a Cascomp topology more useful as an amplifier.…”

“…This paper provides a mathematical basis for an improved understanding of the Cascomp topology in BJTs, which can potentially be extended for use with technologies such as HBTs. Other authors have investigated related techniques to improve linearity in FET amplifiers [8][9][10]; however, a rigorous investigation and comparison in a FET technology is beyond the scope of this paper. In the following analyses, it is shown that when a simple BJT error amplifier is modelled as a non-ideal differential amplifier with its own emitter degeneration, a more effective bias point for maximising gain and minimising distortion can be found compared with that suggested by the original Cascomp theory.…”