SPECTRUM-COMPRESSING RECEIVER AND RECEPTION METHOD FOR NON-CONTIGUOUS CARRIER AGGREGATION

Abstract

A method for carrier aggregation receives a signal with multiple non-contiguous carrier bands. Frequency converting of the signal to a compressed single intermediate frequency band with a pseudonoise code applied to a local oscillation of each of the multiple non-contiguous carrier bands while maintaining separation of the multiple non-contiguous carrier bands permits reduced complexity digital signal processing to detect spectral power density and demodulate waveforms across multiple channels A receiver includes a pseudorandom noise generator applying a pseudo noise code to the local oscillator generator to produce a unique set of spectral tones in the output signal that sample-specific channels over the multiple non-contiguous carrier bands.

Claims

1. A method for carrier aggregation, the method comprising: receiving a signal with multiple non-contiguous carrier bands; frequency converting the signal to a compressed single intermediate frequency band with a pseudonoise code applied to a local oscillation of each of the multiple non-contiguous carrier bands while maintaining separation of the multiple non-contiguous carrier bands; analog to digital converting the compressed single intermediate frequency band to a combined digital signal; and conducting digital signal processing of the combined digital signal to detect spectral power density and waveform characteristics across multiple channels over non-contiguous carrier bands and/or obtain data from each of the multiple non-contiguous carrier bands from the combined digital signal.

2. The method of claim 1, wherein the frequency converting comprises modulating a local oscillator signal with the pseudonoise code followed by N-path filtering.

3. The method of claim 2, wherein the pseudonoise code is configured to produce a unique set of spectral tones that sample-specific channels over the multiple non-contiguous carrier bands.

3. The method of claim 1, wherein the conducting digital signal processing demodulates the data for each of the multiple non-contiguous carrier bands without altering a local oscillation frequency used in said converting.

4. The method of claim 1, wherein the conducting digital signal processing demodulates the data for each of the multiple non-contiguous carrier bands without re-tuning.

5. The method of claim 1, wherein the conducting digital signal processing demodulates the data for each of the multiple non-contiguous carrier bands without re-calibrating frequencies of the multiple non-contiguous carrier bands.

6. The method of claim 1, wherein the conducting digital signal processing completes an observation of the power spectral density and then changes modulated code to sample additional channels without a change in local oscillator frequency.

7. The method of claim 1, wherein the converting the signal maintains a minimal spacing between the multiple non-contiguous carrier bands in the compressed single intermediate frequency according to the bandwidth of the signals and spectral components of pseudonoise code.

8. The method of claim 1, wherein the converting the signal comprises first dividing the frequency into in-phase and quadrature components separately (I+, Q+, I− and Q−) and applying the pseudonoise code to each of the in-phase and quadrature components I+, Q+, I− and Q.

9. The method of claim 1, further comprising amplifying the compressed intermediate signal prior to the analog to digital converting.

10. The method of claim 9, wherein the amplifying the compressed intermediate signal comprises nonlinear amplification to generate sharp edges in the compressed intermediate signal.

11. A spectrum compressing receiver for carrier aggregation of multiple non-contiguous carrier bands, the receiver comprising: a local oscillator generator that receives a carrier aggregation signal and a fixed local oscillator signal and provides a modulated output signal; a pseudorandom noise generator applying a pseudo noise code to the local oscillator generator to produce a unique set of spectral tones in the output signal that sample-specific channels over the multiple non-contiguous carrier bands; an N-Path Filter to down convert output of the output signal to an aggregated intermediate frequency output signal with sufficient bandwidth for the carrier aggregation channels; an analog to digital converter to convert the aggregated intermediate frequency output signal to a digitized signal; and a digital signal processor to recover each of the multiple non-contiguous carrier bands from the digitized signal.

12. The receiver of claim 11, wherein the local oscillator generator comprises a non-linear output buffer to generate sharp edges in the modulated output signal.

13. The receiver of claim 11, comprising a frequency divider prior to the local oscillator generator to divide the carrier aggregation signal into in-phase and quadrature components separately (I+, Q+, I− and Q−).

14. The receiver of claim 13, wherein the frequency divider comprises D flip-flops (DFFs) from thin-oxide 1V transistor devices.

15. The receiver of claim 11, further comprising an amplifier between the N-Path filter and the analog to digital converter.

16. The receiver of claim 15, wherein the amplifier comprises separate I and Q amplifiers.

17. The receiver of claim 11, wherein the N-Path filter comprises a voltage mode filter that provides separate output ports for down conversion and sampling.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The invention will be explained in greater detail hereinafter on the basis of exemplary embodiments illustrated in the drawings, in which:

[0012] FIG. 1A (prior art) is a schematic diagram of a conventional current carrier aggregation receiver; FIG. 1B is a schematic diagram of a preferred embodiment spectrum sensing carrier aggregation receiver;

[0013] FIG. 2 shows the approximated frequency components of Walsh family of length 16;

[0014] FIG. 3 plots the probability of error (Pe) as a function of signal-to-noise ratio (SNR) for an additive white Gaussian noise (AWGN) channel and a fading wireless channel;

[0015] FIG. 4A illustrates a preferred pulse shaping technique; FIG. 4B shows the measured spectrum of generated waveform with pulse shaping;

[0016] FIG. 5 is a block diagram of the preferred embodiment spectrum sensing carrier aggregation receiver;

[0017] FIG. 6 shows a comparison between NMOS and PMOS R.sub.onC.sub.off product (simulation);

[0018] FIG. 7A shows a simulated comparison of S-parameters S.sub.21 and S.sub.11 for an N-Path filter of FIG. 5 based on NMOS and PMOS switches; FIG. 7B illustrates the configuration of the NMOS and PMOS switches;

[0019] FIG. 8 shows a preferred transistor implementation of the preferred embodiment spectrum sensing carrier aggregation receiver;

[0020] FIG. 9 shows the simulated impact of the nonlinear output buffer in FIG. 8 on a modulated LO signal with W.sub.16,

[0021] FIG. 10 illustrates that clock gating self-cleans the phase noise of frequency divider and LO modulation of FIG. 5 and FIG. 8;

[0022] FIG. 11 shows the simulated LO spectrum before modulation showing a peak at f.sub.LO of 1 GHz;

[0023] FIG. 12 provides a transistor-level schematic of an amplifier/buffer that samples the PMOS N-path filter output of FIG. 5.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0024] The present inventors have determined that a carrier aggregation (CA) receiver should also support spectrum sensing (SS) over a wide bandwidth (BW) with low sensing latency. In addition, the receiver should tolerate strong blockers within the RF band during the SS or CA operation.

[0025] Preferred embodiment receivers and reception methods provides a receiver that has a single receiver chain and single ADC to capture multiple non-contiguous bands simultaneously. The preferred embodiment provides a sub-linear increase in the power consumption of the receiver and realizes improved energy for carrier aggregation. Digital processing applied on the RF side before ADC conversion compresses the spectrum while maintaining sufficient separation between non-contiguous bands such that signals from each of the non-contiguous bands can be recovered after conversion of the compressed spectrum to digital.

[0026] A preferred embodiment receiver mixes several channels across a wide range of RF frequency bands to adjacent channels at an intermediate frequency (IF)(IF>BW*N where N is number of signals to simultaneously capture) chosen to receive the total signal bandwidth (BW). The compression in the bandwidth results in a reduction in the processing bandwidth of the ADC. Furthermore, the channel selection does not require phase locked loop (PLL) tuning and an example receiver according to the invention has achieved channel switching times of 25 ns. The prototype receiver chip tunes over a bandwidth of 100 MHz to 1.4 GHz and demonstrates out-of-band (008) rejection of 49 dB and measured sensitivity of receiver is −86 dBm, which is the lowest measured for N-path-based receivers. The chip performance has been verified with over-the air (OTA) measurement of two transmitted signals 80 MHz apart that are received simultaneously with a 10 MHz ADC with error-vector magnitude (EVM) around −20 dB. The measured dynamic range for EVM<−20 dB is better than 86 dB for a QPSK constellation. The example chip was fabricated in a Global Foundries 45 nm RF SOI process and occupies an active area of 0.55 mm-sq while the power consumption is 23.1 mW when operating at 1 GHz.

[0027] Preferred embodiments of the invention will now be discussed with respect to the drawings. The drawings may include schematic representations, which will be understood by artisans in view of the general knowledge in the art and the description that follows. Features may be exaggerated in the drawings for emphasis, and features may not be to scale.

[0028] FIG. 1B shows a preferred embodiment spectrum-compressing receiver architecture 100. A pseudonoise (PN) (also known as pseudorandom) code generator 102 provides a PN code that modulates a single local oscillator 104 to produce a unique set of spectral tones that sample-specific channels over a wide RF band received via an antenna 106. The individual channels are down-converted via the PN-modulated local oscillator (LO) 104 to an aggregated intermediate frequency with sufficient bandwidth for the carrier aggregation channels where each channel is separated by a desired channel spacing and digitized with a single analog-to-digital (ADC) 108 operating at a sample rate high enough to process all three signals corresponding to three separate bands. Generally, the minimal spacing is determined by the bandwidth of the signals and the PN spectral components that are related to the chip rate. The selection of the chip rate determines the IF separation. A digital signal processor (DSP) 110 filters and demodulates the carrier aggregation signals or detects spectral power density over multiple channels through spectrum sensing algorithms in DSP 110. Once the observation is complete, the DSP 110 can change the modulated code to sample other channels without any change in the LO frequency for rapid reconfiguration. In other words, the DSP can first detect occupied signal bands, and then simultaneously demodulate a plurality of bands. The PN generator modulates the LO with a PN sequence and does not require re-tuning the PLL or re-calibration at each frequency channel, thereby significantly reducing the scan time.

[0029] Regarding the PN code, a random code C(t) of length M can be decomposed into Fourier coefficients based on the sequence values,

[00001]$\begin{array}{cc}C\left(t\right)=\underset{m=-\infty }{\overset{\infty }{.Math.}}\underset{k=1}{\overset{M}{.Math.}}a\left[k\right]{\left(-j\right)}^{\frac{4m\left(k-1\right)}{M}}{a}_m\text{?}\mathrm{where}{a}_m=\frac{\sin m\frac{\pi }{M}}{m\pi }{e}^{\frac{-j\pi m}{M}}.& \left(1\right)\end{array}$$\text{?}\text{indicates text missing or illegible when filed}$

[0030] Here, ω.sub.c=2πf.sub.c, where f.sub.c is the chip rate of the sequence. The Fourier domain representation of the periodic code gives insight into the ability to tailor the LO spectrum with weights at the different frequency taps. To simplify (1), we consider only the spectral components up to the chip rate:

[00002]$\begin{array}{cc}C\left(t\right)\approx \underset{m=-M/2}{\overset{+M/2}{.Math.}}{a}_m\text{?}\underset{k=1}{\overset{M}{.Math.}}a\left[k\right]{\left(-j\right)}^{\frac{4m\left(k-1\right)}{M}}.& \left(2\right)\end{array}$$\text{?}\text{indicates text missing or illegible when filed}$

[0031] In other words, each M-length code has up to M frequency components spaced by f.sub.c/M as well as a component at dc. The amplitude at each frequency component is:

[00003]$\begin{array}{cc}{c}_m=a_m\underset{k=1}{\overset{M}{.Math.}}a\left[k\right]{\left(-j\right)}^{\frac{4m\left(k-1\right)}{M}}.& \left(3\right)\end{array}$

[0032] Therefore, the Fourier series is

[00004]$C\left(t\right)\approx {.Math.}_{m=-2}^{+M/2}{c}_m{e}^{{j\left(\frac{m{\omega }_c}{M}\right)}^t}.$

For example, a Walsh code of length 8 has eight potential codes. The second Walsh code (W.sub.2) has a sequence of α[k]=[1 0 1 0 1 0 1 0]. After applying (2) to W.sub.2, each frequency is

[00005]$c_m=\left[\frac{1}{\pi }000.Math.0.Math.000\frac{1}{\pi }\right]$

at frequency components

[00006]$\left[\begin{array}{ccccc}-f_c& -\frac{3f_c}{4}& .Math.& +\frac{3f_c}{4}& +f_c\end{array}\right].$

[0033] When the code is multiplied by an LO frequency f.sub.LO, the spectral components of the code are modulated around the LO, for example

[00007]$\left[\begin{array}{ccccc}{f}_{\mathrm{LO}}-f_c& f_{\mathrm{LO}}-\frac{3f_c}{4}& .Math.& f_{\mathrm{LO}}+\frac{3f_c}{4}& f_{\mathrm{LO}}+f_c\end{array}\right].$

[0034] Therefore, the modulated LO has two components with 5 dB of loss relative to the local oscillator, which potentially degrades the sensitivity. Since the dc component of a Walsh codes is always 0, e.g., cm=0, the RX can select different frequency components for carrier aggregation without interference from the blockers near the local oscillator frequency.

[0035] To aggregate two carriers at f.sub.1 and f.sub.2 without interference to an IF location, f.sub.LO and f.sub.c need to be tuned. The two RF carriers are down converted to IF frequencies IF.sub.1 and IF.sub.2

[00008]$\begin{array}{cc}{f}_{\mathrm{LO}}=\frac{f_1+f_2+{\mathrm{IF}}_2+{\mathrm{IF}}_1}{2}& \left(4\right)\end{array}$$\begin{array}{cc}{f}_c=f_2-f_1+{\mathrm{IF}}_2+{\mathrm{IF}}_1.& \left(5\right)\end{array}$

[0036] To determine minimum spacing, assuming that the two desired carriers have a bandwidth BW, the receiver should allow that IF.sub.2=IF.sub.1+BW. Therefore,

[00009]$\begin{array}{cc}{f}_{\mathrm{LO}}=\frac{f_1+f_2+\mathrm{BW}}{2}& \left(6\right)\end{array}$$\begin{array}{cc}{f}_c=f_2-f_1+2{\mathrm{IF}}_1+\mathrm{BW}.& \left(7\right)\end{array}$

[0037] To capture example existing bands of 2100 and 3500 MHz, a Walsh code with a respective local oscillator frequency and chip rate of 2801 and 1406 MHz can be used.

[0038] If more than two carriers are aggregated, a different Walsh code can be selected. For Walsh 6 (W.sub.6), [1 0 1 0 0 1 0 1] and the Fourier coefficients are [0 0.256 0 0.132 0 0 0.132 0 0.256 0] allowing the aggregation at four RF channels.

[0039] A comprehensive list of the frequency components and relative power levels of Walsh codes of length 16 are shown in FIG. 2. The relative amplitudes of the Fourier coefficients are indicated with darker contrast. The first Walsh code (W.sub.1) is a sequence of ones, so it only has a dc component and can select only one channel. The W.sub.2, W.sub.3, and W.sub.4 codes select two channels simultaneously. Note that W.sub.3 and W.sub.4 are degenerate. The W.sub.5-W.sub.8 codes select four channels simultaneously. The W.sub.9-W.sub.16 codes aggregate up to six channels. Given a fixed local oscillator, changing the Walsh code can select any channel across the band in FIG. 2.

[0040] FIG. 2 suggests that reduced local oscillator amplitude at each spectral component leads to higher spot noise figure (NF) due to the reduced conversion gain. Nonetheless, the higher NF is offset by a significant advantage from the NC-CA diversity technique. FIG. 3 plots the probability of error (Pe) as a function of signal-to-noise ratio (SNR) for an additive white Gaussian noise (AWGN) channel and a fading wireless channel [D. Tse and P. Viswanath, Fundamentals of Wireless Communication. Cambridge, U.K.: Cambridge Univ. Press, 2013]. In the AWGN channel, Pe improves exponentially with SNR, suggesting that higher NF is undesirable.

[0041] However, the multi-path fading channel does not provide a significant benefit due to the SNR. Even when the RX SNR is large, the multiple paths can destructively interfere and degrade Pe inversely proportional to SNR. The NC-CA diversity improves the Pe of the wireless channel by adding an independent degree of freedom [D. Tse and P. Viswanath, Fundamentals of Wireless Communication. Cambridge, U.K.: Cambridge Univ. Press, 2013]. To achieve a Pe of 10.sup.−3 with a fading channel, the SNR needs to exceed 30 dB in the absence of diversity. With the diversity-2 channel, Pe of 10.sup.−3 is achieved with the SNR of 13 dB. Considering an SNR degradation of 3 dB with the diversity channel, Pe of 10.sup.−7 is reached for 30 dB. In other words, CA targets increased the capacity of the wireless channel for high SNR where NF is not usually the limitation. In commercial products, the low noise amplifier (LNA) gain is reduced or the LNA might be bypassed in the carrier aggregation mode to improve the linearity of the receiver to realize the diversity. See, D. Whitefield and S. Shah, “Integrated LAA/Wi-Fi RF front-end solutions meet new requirements in 5G smartphones,” in Proc. Skyworks Tech. Article, 2018. [Online] Available: http://www.skyworksinc.com/downloads/press_room/published_articles/MWJ_%5G_WiFi_RF_FEM_201902.pdf.

[0042] To eliminate aggregation from the out-of-band harmonics, a pulse shaping technique can be added to FIG. 4A. A suitable pulse shaping technique is disclosed in Hill et. al, “Watt-Level, Direct RF Modulation in CMOS SOI With Pulse-Encoded Transitions for Adjacent Channel Leakage Reduction,” IEEE Transactions on Microwave Theory and Techniques Volume: 67, Issue: 12 (2019). Pulsewidth modulation at the edges mimics a raised cosine filter and chosen to suppress the third-order harmonic. This comes at the cost of reducing the power of fundamental and increasing the power of higher order harmonics while increasing the chip rate by a factor of 8. The measured spectrum of generated waveform is shown in FIG. 4B.

[0043] FIG. 5 is a block diagram of the preferred embodiment spectrum sensing carrier aggregation receiver 100. With the receiver 100, an RF signal is balanced into a differential signal and a PMOS N-path filter 502 downconverts the signals to sample onto baseband amplifiers 508. The spectrum sensing carrier aggregation receiver 100 can be tapped from a primary RX that might have higher NF allowing for usage of the spectrum sensing carrier aggregation receiver 100 as a diversity path. The 75% duty-cycle local oscillators are generated and modulated on chip in a preferred implementation.

[0044] The N-Path Filter 502 is preferably implemented in PMOS, while NMOS switches can also be used. N=4 is the simplest N path filter. N is a general number of paths that are consecutively sampled. N is typically 4 or 8. A local modulator 504 receive the PN code that is configured to produce a unique set of spectral tones that sample-specific channels over the multiple non-contiguous carrier bands. In the preferred embodiment, a frequency divider 506 first separates the received band into four components, namely 0-89°, 90-179°, 180-269°, and 270-359° to process in-phase and quadrature components separately (I+, Q+, I− and Q−). With the four separate components of the band, the modulator 504 applies the PN code to each of the in-phase and quadrature components I+, Q+, I− and Q−. The N-Path Filter 502 then outputs a compress intermediate frequency in differential form to in-phase and quadrature baseband amplifiers 508. ADCs 510 convert the in-phase and quadrature signals to digital, which are then processed by a digital signal processor 512. While two ADCs 510 are shown, they could be replaced with single ADC to sample both I and Q channels. The pair of ADCs 510 are used to sample I and Q channels separately. The DSP 512 conducts standard signal processing, including determining which channel is optimal from the standpoint of signal-to-noise ratio, demodulating and sampling the data stream, and performing error correction of data. The DSP 512 can conduct reduced complexity digital signal processing, compared to prior techniques, to detect spectral power density and demodulate waveforms across multiple channels as a result of the frequency converting the signal to a compressed single intermediate frequency band with a pseudonoise code applied to a local oscillation of each of the multiple non-contiguous carrier bands conducted by the frequency divider 506, LO modulation 504 and N-path filed 502.

[0045] FIG. 6 shows a comparison between NMOS and PMOS R.sub.onC.sub.off product (simulation) for different devices in a 45-nm RF SOI process. At the shortest channel length, the PMOS switches are only 14% slower than NMOS switches. The PMOS devices potentially offer higher breakdown and improved linearity and are preferred for those benefits.

[0046] FIG. 7A compares the simulated S-parameters of a shunt NMOS and PMOS-based N-path filter each with the same device geometry illustrated in FIG. 7B. The out-of-band (OOB) rejection is degraded by less than 1 dB, while the insertion loss (IL) is slightly improved due to higher switch resistance and higher equivalent resistances (R.sub.SH). Another advantage of using PMOS switches is the relatively high common-mode voltage (V.sub.dc≈1V) applied at the source and drain to keep the switch OFF when the LO is high. The high common-mode voltage allows an NMOS transconductance (G.sub.m) stage to directly sample the IF capacitors on the N-path filter. The simulated input P1 dB of PMOS N-path is 0.95 dB higher than that of NMOS N-path.

[0047] FIG. 8 shows a preferred transistor implementation of the FIG. 5 spectrum sensing carrier aggregation receiver 100. The FIG. 8 PMOS N-path filter shown in FIG. 8 consists of four paths. The size of the switches is an example implementation was 106.4 μm/40 nm, while the baseband capacitors are based on a hybrid capacitor bank [C. K. Luo, P. S. Gudem, and J. F. Buckwalter, “A 0.4-6-GHz 17-dBm B1 dB 36-dBm IIP3 channel-selecting low-noise amplifier for SAW-less 3G/4G FDD diversity receivers,” IEEE Transactions on Microwave Theory and Techniques, vol. 64, no. 4, pp. 1110-1121, April 2016] with high-density capacitors and shunt capacitors to improve the OOB rejection and CM noise generated on chip from LO leakage, and the equivalent capacitance value is 124.9 pF.

[0048] The PN-modulated LO generator 504 allows the adjustment of LO phase overlap with the pseudo noise sequence. Digitally XORing the LO phase with the pseudo sequence value produces the spectral components of the LO that shifts the bandpass impedance of the N-path filter 502 to the desired frequency channel. The frequency divider 506 is preferably implemented with D flip-flops (DFFs) based on thin-oxide devices with 1-V supply for speed, and thus, the common voltage on the RF and BB nodes is 1 V. An input buffer 802 (part of the frequency divider 506 in FIG. 5) sharpens the clock transitions at twice the LO frequency before the clock divider and provides a wideband matching block. An output buffer 804 (such as a standard inverter amplifier) to LO generator 504 can be nonlinear by design because the buffer 804 should generate sharp transition edges. When modulated with a W.sub.16 code, the LO signal is spread to different locations resembling the profile shown in FIG. 2.

[0049] FIG. 9 shows the simulated impact of the nonlinear output buffer 804 on a modulated LO signal with W.sub.16. The impact of nonlinear buffers on the modulated LO signal is minimal as the harmonic components of the modulated LO signal passed through two inverters show no change in the amplitude of the LO components. The even harmonics show an undesirable increase of around 30 dB but remain around −100 dBc.

[0050] While the frequency divider and LO modulator might be expected to degrade the LO phase noise, a gating technique prevents unnecessary jitter accumulation from the divider and modulator. FIG. 10 illustrates that clock gating self-cleans the phase noise of frequency divider and LO modulation. The code is sampled with the DFF clocked with the LO to ensure that the modulation begins at the beginning of the LO waveform. At the code transition, the only vulnerable edge occurs when the rising edge of the 2 f.sub.LO overlaps the jittered LO to transfer phase noise of the divider and modulator to the LO.

[0051] FIG. 11 shows the simulated LO spectrum before modulation showing a peak at f.sub.LO of 1 GHz, which compares the theoretical conversion gain for the PN-modulated N-path filter when modulated with W.sub.16. The simulation indicates close agreement with theory at six distinct peaks (−20-dB conversion gain) and some small discrepancy at the weaker harmonics. The conversion gain is normalized to the unmodulated LO.

[0052] FIG. 12 provides a transistor-level schematic of a preferred baseband amplifier for the I and Q amps 508 of FIG. 5. Unlike earlier work that realized the N-path as a mixer-first structure such that the BB impedance (Z.sub.BB) is low at dc for impedance matching (current-mode N-path) requiring a transimpedance amplifier (TIA)-based BB amplifier, the FIG. 12 circuit presents a high impedance such that the RF signal power passes through the filter and matches to the output port, i.e., voltage mode with high Z.sub.BB) N-path. One advantage of the voltage-mode N-path filter is providing both RF.sub.out and BB.sub.out ports for down-conversion and sampling. However, the disadvantage of both ports is an increased insertion loss by adding mismatch that might be tuned out with a matching network at the expense of lower frequency range. The FIG. 12 BB amplifier stage that samples the IF capacitors with a Gm-cell and a TIA (transimpedance amplifier). The baseband amplifier is implemented with thick-oxide devices operating at 1.5 V. Adding a G.sub.m cell between the baseband capacitors and the TIA serves to convert the sampled voltage across the caps to current for the high-linearity TIA. The dc level on the PMOS switches sets the common-mode voltage for the G.sub.m-cell. The input common-mode voltage range is given by:

V.sub.th1,2+V.sub.od1,2+V.sub.sat5≤V.sub.IN,CM≤V.sub.DD,A−|V.sub.sat3,4|.  (8)

[0053] Typical values of V.sub.th1,2=0.35 V, V.sub.od1,2=0.15 V, V.sub.DD,A=1.5V, and V.sub.sat5=|V.sub.sat3,4|=0.2 V suggest a range 0.7 V≤V.sub.IN,CM≤1.3 V. Consequently, choosing V.sub.IN,CM=1 V provides the maximum input voltage swing and improves linearity. Considering a G.sub.m-cell with PMOS input devices, the common-mode voltage requirement is between 0.2 V≤V.sub.IN,CM≤0.8 V. However, a common-mode approach to bias the NMOS switches at a voltage slightly higher than 0 V (around 0.5 V) to set the common-mode level of the PMOS input devices of G.sub.m-cell has two issues: 1) decreasing the ON resistance of the switch lowers the out of band rejection and 2) PMOS devices have lower mobility and thus lower transconductance for the same current compared to NMOS devices, which results in lower gain and higher NF (noise figure). Consequently, PMOS-based switches with an NMOS-based baseband amplifier are preferred.

[0054] The main amplifier FIG. 12 (508 in FIG. 5) has high transconductance (gm) to lower the NF. The CMFB is sized to be ⅛ smaller than the main amplifier. To stabilize the CMFB, a load capacitor is added to the feedback node and two Miller capacitors are added across the second gain stage of the feedback. The TIA is similar to the G.sub.m-cell with a feedback resistor and capacitor to set the BB BW. The gain of each of the BB amplifiers is approximately 36 dB in the example prototype that was fabricated. Testing of the prototype revealed significant reductions in power consumption and active area to realize a wideband RX capable of extremely high dynamic range. Switching times of less than 25 nanoseconds were demonstrated. Additional performance attributes of the prototype are given in the following table:

TABLE-US-00001 Process GF 45 nm RF SOI Frequency (GHz) 0.1-1.4 CA Span (GHz) 1 Limited by PLL settling time? No Switch Time (ns) <25 CA Sensitivity (dBm)/ −86/OPSK 1 Mbps Modulation type −77/QAM16 1 Mbps Power Consumption (mW) 23.1 BB Gain (dB) 36 NF (dB)  9.8-17.1 IB IIP3 (dBm) >−14.5 Minimum Detectable Signal (dBm)/ −861 Bitrate (Mbps) OOB Rejection (dB) 49 Active Area (mm.sup.2) 0.55

[0055] While specific embodiments of the present invention have been shown and described, it should be understood that other modifications, substitutions and alternatives are apparent to one of ordinary skill in the art. Such modifications, substitutions and alternatives can be made without departing from the spirit and scope of the invention, which should be determined from the appended claims.

[0056] Various features of the invention are set forth in the appended claims.