betaFeedback

WorksEngineering

Edge-Combining Subharmonic Mixer-First Receiver

Abstract

A 1-10 GHz subharmonic mixer-first receiver using edge combining to drive N-path switches from 50% duty-cycle clocks at half the local oscillator (LO) frequency avoids the gain penalty of harmonic mixing. Inverter-based differential baseband amplifiers with back-gate feedback linearization set their gain by process-defined capacitance ratios, using no additional components and consuming no additional power. A 22 nm fully depleted silicon on insulator (FD-SOI) prototype receiver achieves a conversion gain of 12.8-18.1 dB and a noise figure of 5.7-11 dB across the band. It achieves up to 15.4 dBm out-of-band IIP3 while consuming 16.7-22.3 mW.

Figure 1. Proposed subharmonic mixer-first receiver architecture with edge combining for 2× clock frequency multiplication.

Introduction

Integrated N-path mixer-first receivers can cover wide radio frequency (RF) bands and provide high-Q RF filtering around a digitally-controlled center frequency for use in software-defined radios, but two bottlenecks limit how far this approach scales in frequency. First, the switches require NN non-overlapping clock phases with 1/N1/N duty cycle at the local oscillator (LO) frequency fLOf_{\mathrm{LO}}, usually derived from a master clock at a multiple of fLOf_{\mathrm{LO}}. At multi-GHz LO frequencies, the divider and switch drivers dominate the receiver power [1, 2]. Driving the mixer at a subharmonic and downconverting with the nnth LO harmonic relaxes the clock frequency, but the conversion gain falls as sinc⁡(n/N)\operatorname{sinc}(n/N), which also degrades noise figure [3, 4]. Second, the baseband amplifiers, rather than the switches, often limit the linearity and bandwidth of passive mixer-first receivers [5, 6, 7].

This work addresses both clock power and baseband linearity bottlenecks in mixer-first receivers. Edge combining, previously used for frequency multiplication in RF digital-to-analog converters [8], is moved into the mixer switches: series-stacked switches driven by 50%-duty-cycle phases at fLO/2f_{\mathrm{LO}}/2 synthesizes the 25% switching function directly at the RF input, so the receiver downconverts at the fundamental of the effective LO without the harmonic-mixing loss. The baseband uses inverter-based amplifiers with back-gate feedback linearization [9], whose gain is set by a well-controlled front-to-back gate capacitance ratio in fully depleted silicon on insulator (FD-SOI) technology and costs no additional power nor components.

Section II provides a review of edge combining. Section III describes the receiver architecture, followed in Section IV by measurement results. Finally, Section V ends with conclusions.

Background

Edge combining is a frequency multiplication technique that works by interleaving multiple time-delayed versions of a lower-frequency clock signal with 50% duty cycle. The rising and falling edges of the delayed phases are combined by a logical AND operation, and the output signal switches at a multiple of the input frequency, determined by the number of phases. Edge combining in N-path filters and mixers extends the N-path principle to the clock section and relaxes the LO generation requirements. N-path mixers, which multiply the RF input by a square wave LO, downconvert LO harmonics with increasing loss. For an ideal N-path mixer driven by non-overlapping clocks with 1/N1/N duty cycle, the magnitude of the nnth harmonic voltage conversion gain from the RF input at a frequency fRF=nfLOf_{\mathrm{RF}} = n f_{\mathrm{LO}} to a single baseband path is

∣Gv,n∣=sinc⁡(nN)(1)\lvert G_{\mathrm{v},n} \rvert= \operatorname{sinc}\left(\frac{n}{N}\right) \tag*{(1)}

in normalized sinc notation. For example, subharmonic downconversion with the second harmonic of a four-path mixer gives ∣Gv,2∣=sinc⁡(1/2)≈0.64\lvert G_{\mathrm{v},2} \rvert= \operatorname{sinc}(1/2) \approx0.64 (−3.9 dB), compared with ∣Gv,2∣=sinc⁡(1/4)≈0.90\lvert G_{\mathrm{v},2} \rvert= \operatorname{sinc}(1/4) \approx0.90 (−0.9 dB) at the fundamental. Edge combining increases the fundamental LO harmonic presented to the RF input and avoids the lower conversion gain from using higher harmonics.

Receiver Architecture

Edge Combining

Each of the four baseband paths uses two parallel stacks of two series-connected switches in place of a single switch (Fig. 1). The clock phase separation in the series switches is 180∘+360∘/(N⋅M)=225∘180^\circ+ 360^\circ/(N \cdot M) = 225^\circ where M=2M = 2 is the clock frequency multiplication factor, and N⋅M=8N \cdot M = 8 is the total number of clock phases. A series pair conducts only when both gate phases are high, implementing the logical AND of two 50%-duty-cycle phases taken from an eight-phase set at fclk=fLO/2f_{\mathrm{clk}} = f_{\mathrm{LO}}/2. The phase offset sets the overlap to TLO/4T_{\mathrm{LO}}/4. The parallel connection ORs two such pulses spaced by Tclk/2T_{\mathrm{clk}}/2, so each path conducts with a 25% duty cycle at twice the clock frequency (Figure 2). Because every clock net toggles at fLO/2f_{\mathrm{LO}}/2 with 50% duty cycle, the divider and buffers avoid generating narrow pulses at fLOf_{\mathrm{LO}}. The trade-offs are that stacked switches need approximately twice the width as a single switch for the same on-resistance, and the extra stacks add parasitic capacitance at the RF input node. That capacitance limits higher frequency multiplication factors and the upper end of the RF range.

Figure 2. Edge combining waveforms.

Clock Generation

An external single-ended sinusoidal clock signal at 2fLO2f_{\mathrm{LO}} is converted to differential by an on-chip transformer balun, squared up by input buffers matched to 50 Ω\Omega, and divided by a four-stage Johnson counter into eight 50%-duty-cycle phases (Figure 3). Each phase driver is an inverter chain with sizing optimized for jitter and power [10]. Clock phases are distributed to the switches through an H-tree network for equal propagation delay, which is critical because phase skew directly changes the edge-combined pulse width and the I/Q balance.

Figure 3. Multi-phase clock generation.

Baseband Amplifiers and Filters

At baseband frequencies, the large loop gain available makes negative feedback a suitable linearization technique to reduce distortion, particularly if the feedback factor depends on a well-matched ratio in the semiconductor process rather than absolute values. Feeding the transistor output to the back gate in FD-SOI technology exploits the linearizing effect of negative feedback with a closed-loop voltage gain

Av=A01+χA0=−gmro1−χgmro(2)A_{\mathrm{v}} = \frac{A_{0}}{1+\chi A_{0}} = -\frac{g_{\mathrm{m}}r_{\mathrm{o}}}{1-\chi g_{\mathrm{m}}r_{\mathrm{o}}} \tag*{(2)}

where A0=−gmroA_{0} = -g_{\mathrm{m}}r_{\mathrm{o}} is the intrinsic gain [9], and the feedback factor is the back-gate capacitive coupling ratio

χ=−gmbgm=−Cbox∥CSiCox(3)\chi= -\frac{g_{\mathrm{mb}}}{g_{\mathrm{m}}} = -\frac{C_{\mathrm{box}} \mathbin{\|} C_{\mathrm{Si}}}{C_{\mathrm{ox}}} \tag*{(3)}

where CboxC_{\mathrm{box}}, CSiC_{\mathrm{Si}}, and CoxC_{\mathrm{ox}} are the buried oxide, thin silicon channel, and gate oxide capacitances, respectively [11]. Considering the intrinsic gain as a first-order system

A0(s)=−gmro1+s/ωp(4)A_{0}(s) = -\frac{g_{\mathrm{m}}r_{\mathrm{o}}}{1+s/\omega_{\mathrm{p}}} \tag*{(4)}

with a dominant pole ωp\omega_{\mathrm{p}}, the root of the characteristic equation

1+χA0(s)=01+\chi A_{0}(s) = 0
s=−(1+gmbro)ωp(5)s = -(1+g_{\mathrm{mb}}r_{\mathrm{o}})\omega_{\mathrm{p}} \tag*{(5)}

is real and negative. The closed-loop pole is in the left-half plane, so the system is stable with back-gate feedback. When the loop gain magnitude ∣χgmro∣≫1\lvert\chi g_{\mathrm{m}}r_{\mathrm{o}}\rvert\gg1, the closed-loop gain is approximately 1/χ=−gm/gmb1/\chi= -g_{\mathrm{m}}/g_{\mathrm{mb}} and is set by the transistor’s front-to-back gate oxide capacitance ratio in the semiconductor process.

Back-gate feedback makes the baseband amplifier gain approximately independent of the load resistance. The small-signal voltage gain of the complementary common-source amplifier with back-gate feedback

Av=−(gmn+gmp)(ron∥rop)1+(gmbn+gmbp)(ron∥rop)≈−gmn+gmpgmbn+gmbp(6)A_{\mathrm{v}} = -\frac{(g_{\mathrm{mn}}+g_{\mathrm{mp}})(r_{\mathrm{on}} \mathbin{\|} r_{\mathrm{op}})}{1+(g_{\mathrm{mbn}}+g_{\mathrm{mbp}})(r_{\mathrm{on}} \mathbin{\|} r_{\mathrm{op}})} \approx-\frac{g_{\mathrm{mn}}+g_{\mathrm{mp}}}{g_{\mathrm{mbn}}+g_{\mathrm{mbp}}} \tag*{(6)}

shows the same intrinsic gain −(gmn+gmp)(ron∥rop)-(g_{\mathrm{mn}}+g_{\mathrm{mp}})(r_{\mathrm{on}} \mathbin{\|} r_{\mathrm{op}}) through the front gate as in open loop (Figure 4). Adding feedback reduces the output resistance ron∥ropr_{\mathrm{on}} \mathbin{\|} r_{\mathrm{op}} by the loop gain (gmbn+gmbp)(ron∥rop)(g_{\mathrm{mbn}}+g_{\mathrm{mbp}})(r_{\mathrm{on}} \mathbin{\|} r_{\mathrm{op}}), so the overall gain is approximately the front- and back-gate conductance ratio 1/χ1/\chi from (3) in saturation while requiring no additional components and consuming no additional power compared to without feedback [9].

Figure 4. Dual common-mode feedback differential amplifier with back-gate feedback linearization [9].

Measurement Results

A prototype mixer-first receiver front-end was implemented in GlobalFoundries’ 22 nm FD-SOI technology using 0.8 V thin-oxide devices for the clock generation circuitry and the core switches, with 1.8 V thick-oxide I/O devices for the baseband amplifiers to extend the linear output operating range (Figure 5). Wire bonding pads on the top and bottom (not visible) supply power and connect the differential baseband I/Q outputs to external buffers capable of driving a 50 Ω\Omega load. The test circuit occupies a total area of 0.335 mm2^2 including pads. The active or functional block area is 0.0979 mm2^2 when including inductors and 0.0345 mm2^2 without inductors. It is noted that one of the inductors is used to receive the clock from off chip and would be obviated in a system on chip implementation.

Die micrograph of the 22 nm FD-SOI prototype receiver front-end

Figure 5. Die micrograph of the 22 nm FD-SOI prototype receiver front-end.

A linear fit y=mx+by = mx + b of the clock power consumption has a slope m=0.674m = 0.674 mW/GHz and an intercept b=5.28b = 5.28 mW. The mean baseband power is 10.0 mW, and the total power consumption ranges from 16.7 mW at 1 GHz to 22.3 mW at 10 GHz.

The conversion gain ranges from 12.8 dB at 10 GHz to 18.1 dB at 4 GHz (Figure 6). The baseband −3-3 dB bandwidth is 10 MHz on either side of the LO frequency for a 20 MHz total channel bandwidth (Figure 7). Input matching is quantified by S-parameters. S11S_{11} reaches a minimum of −35-35 dB at 6 GHz because of the series RF input inductor used to improve impedance matching, and S11S_{11} increases towards the upper and lower bounds of the operating frequency range. The noise figure was measured using the Y-factor method and ranges from 5.73 dB at 1 GHz to 11.0 dB at 10 GHz (Figure 8).

Figure 6. Measured conversion gain and S11S_{11}.

Measured conversion gain and S11 centered at fLO = 6 GHz

Figure 7. Measured conversion gain and S11 centered at fLO=6f_{\mathrm{LO}} = 6 GHz.

Measured noise figure

Figure 8. Measured noise figure.

Linearity is characterized by the input third-order intercept point (IIP3) and was measured with two tones f1=fLO+Δf+fIM3f_1 = f_{\mathrm{LO}} + \Delta f + f_{\mathrm{IM3}} and f2=fLO+2Δf+fIM3f_2 = f_{\mathrm{LO}} + 2\Delta f + f_{\mathrm{IM3}}. The third-order intermodulation products have a constant baseband frequency (2f1−f2)−fLO=fIM3=0.5(2f_1 - f_2) - f_{\mathrm{LO}} = f_{\mathrm{IM3}} = 0.5 MHz. IIP3 increases with offset frequency Δf\Delta f, from 3.74–10.8 dB in-band to 9.84–15.4 dB out-of-band over the full operating range (Figure 9 and Figure 10).

Figure 9. Measured in-band IIP3, out-of-band IIP3, and adjacent channel B1dB with a blocker offset Δf/f−3dB=1\Delta f/f_{-3\mathrm{dB}} = 1.

Figure 10. Measured IIP3 centered at fLO=6f_{\mathrm{LO}} = 6 GHz.

full operating range (Figure 9 and Figure 10).

Conclusion

A subharmonic mixer-first receiver in 22 nm FD-SOI moves edge combining into the N-path switches. The clock path runs at half the LO frequency with 50% duty cycle, while the RF input sees a fundamental 25% switching function, so the receiver avoids the conversion gain and noise penalty of harmonic mixing. Back-gate-feedback inverter amplifiers provide a process-ratio-defined baseband gain at no extra power. The prototype operates from 1–10 GHz with 12.8–18.1 dB

This WorkZijlma et al. [12]Huang and Molnar [13]Lien et al. [7]
Technology22 nm FD-SOI22 nm FD-SOI45 nm SOI45 nm SOI
RF InputSingle-endedSingle-endedSingle-endedDifferential
Frequency (GHz)1–101–103.7–6.50.2–8
BB BW (MHz)1010—10
Gain (dB)12.8–18.110–13—21
NF (dB)5.7–114.7–7.02.4–4.72.3–7.1
IB IIP3 (dBm)3.7–10.81–5−32*0*
OOB IIP3 (dBm) (@Δf/f−3dB\Delta f/f_{-3\mathrm{dB}})9.8–15.4 (10)21–23 (50)2839 (8)
Supply (V)0.8, 1.80.9—1.2
Static Power (mW)15.303650
Dynamic Power (mW/GHz)0.6713.116.430
Active Area (mm²)0.10.05—0.8

Table I. Performance Comparison

*Estimated from plots

conversion gain, 5.7–11 dB noise figure, and 16.7–22.3 mW total power in 0.0345 mm² of active area excluding inductors.

Acknowledgment

We would like to acknowledge GlobalFoundries’ University Partnership Program for providing process development kits.

References

  1. [1]V. K. Purushothaman, E. A. M. Klumperink, B. T. Clavera, and B. Nauta, “A fully passive rf front end with 13-db gain exploiting implicit capacitive stacking in a bottom-plate n-path filter/mixer,” IEEE J. Solid-State Circuits, vol. 55, no. 5, pp. 1139–1150, 2020.DOI
  2. [2]F. Lin, P.-I. Mak, and R. P. Martins, “An rf-to-bb-current-reuse wideband receiver with parallel n-path active/passive mixers and a single-mos pole-zero lp,” IEEE J. Solid-State Circuits, vol. 49, no. 11, pp. 2547–2559, 2014.DOI
  3. [3]A. Flete, C. Viallon, P. Cathelin, and T. Parra, “A ×3 Sub-Sampling Mixer for a 77 GHz Automotive Radar Receiver in 28 nm FD-SOI CMOS Technology,” in 2022 IEEE 22nd Topical Meeting on Silicon Monolithic Integrated Circuits in RF Systems (SiRF), Jan. 2022, pp. 52–54.DOI
  4. [4]R. V. Rena, R. Kammari, and V. S. Pasupureddi, “A 0.4–1.8-GHz Quarter-Rate Subsampling Mixer-First Direct Down-Conversion RF Front-End,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 32, no. 3, pp. 552–563, Mar. 2024.DOI
  5. [5]S. Krishnamurthy, F. Maksimovic, L. Lotti, and A. M. Niknejad, “Analysis and design of submilliwatt interference-tolerant receivers leveraging n-path filter-based translational positive feedback,” IEEE Transactions on Microwave Theory and Techniques, vol. 69, no. 7, pp. 3496–3509, 2021.DOI
  6. [6]H. Razavi and B. Razavi, “A 0.4–6 ghz receiver for cellular and wifi applications,” IEEE J. Solid-State Circuits, vol. 57, no. 9, pp. 2640–2657, 2022.DOI
  7. [7]Y.-C. Lien, E. A. M. Klumperink, B. Tenbroek, J. Strange, and B. Nauta, “Enhanced-Selectivity High-Linearity Low-Noise Mixer-First Receiver With Complex Pole Pair Due to Capacitive Positive Feedback,” IEEE J. Solid-State Circuits, vol. 53, no. 5, pp. 1348–1360, May 2018.DOI
  8. [8]H. M. Nguyen, J. S. Walling, A. Zhu, and R. B. Staszewski, “A mm-Wave Switched-Capacitor RFDAC,” IEEE J. Solid-State Circuits, vol. 57, no. 4, pp. 1224–1238, Apr. 2022.
  9. [9]E. Danson and J. Walling, “Inverter-Based Differential Amplifiers With Back-Gate Feedback Linearization,” in 2026 24th IEEE Interregional NEWCAS Conference (NEWCAS), Jun. 2026, pp. 1–4.arxiv.org/abs/2601.17129
  10. [10]W. T. Overeem, M. S. O. Alink, and B. Nauta, “Inverter Chain Buffer Optimization for N-path Filter Switch Drivers and Validation through Simulations in 22nm FD-SOI Technology,” in 2023 IEEE International Symposium on Circuits and Systems (ISCAS), May 2023, pp. 1–5.DOI
  11. [11]H. Agarwal et al., “BSIM-IMG: Advanced Model for FDSOI Transistors with Back Channel Inversion,” in 2020 4th IEEE Electron Devices Technology and Manufacturing Conference (EDTM), Apr. 2020, pp. 1–4.DOI
  12. [12]E. Zijlma, S. van Zanten, R. Plompen, E. A. M. Klumperink, R. A. R. van der Zee, and B. Nauta, “Analysis and Design of a Low-Loss 1–10 GHz Capacitive Stacking N-Path Filter/Mixer,” IEEE J. Solid-State Circuits, vol. 60, no. 2, pp. 367–381, Feb. 2025.DOI
  13. [13]S. Huang and A. Molnar, “A 3.7-6.5GHz 8-Phase N-Path Mixer-First Receiver with LO Overlap Suppression Achieving <5dB NF and >5dBm OOB B1dB,” in 2021 IEEE Radio Frequency Integrated Circuits Symposium (RFIC), Jun. 2021, pp. 87–90.DOI

Paper details

vlnc:W_s9456tzm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Engineering

Not recorded for this paper

View statement
Looking for related papers…
Contents