Advanced Mixers
Transcription
Advanced Mixers
Berkeley Advanced Mixers Prof. Ali M. Niknejad U.C. Berkeley c 2014 by Ali M. Niknejad Copyright Niknejad Advanced IC’s for Comm Sampling Versus Mixing Note that the essential difference between a “mixer” and a “sampler” is the hold operation. Most samplers have a zero-order hold, so the output is held high by the capacitor. In a “mixer”, the output returns to zero due to the lack of energy storage elements. The conversion gain is simply the fundamental Fourier coefficient of the LO waveform. For a 50% square wave gc = 2 2π Z Pi sin(x)dx = 0 Niknejad 1 ≈ 0.318 (-10 dB) π Advanced IC’s for Comm Sampling For a ZOH sampler, we can model it as an ideal sampler followed by an a LPF with bandwidth 1/TD . Note that TD is the duty cycle over which the output is held. Assume TD is small so we can model the system as follows: yout (t) = (p(t) · x(t)) ∗ h(t) Yout (f ) = (P(f ) ∗ X (f ))H(f ) For a sinusoidal input, we have 1 X (f ) = (δ(f − fx ) + δ(f + fx )) 2 and ∞ 1 X P(f ) = δ(f − nfLO ) TLO n=−∞ Niknejad Advanced IC’s for Comm Sampling Taking the convolution of X and P, we note that things don’t overlap unless the input is shifted to the peaks of P ∞ X 1 (δ(f − nfLO − fx ) + δ(f − nfLO + fx )) X (f )∗P(f ) = 2TLO n=−∞ The hold action can be modeled by a filter with constant finite impulse response of duration TD . Taking the Laplace Transform, we get a sinc frequency response. H(f ) = sin(πfTD ) −jπfTD e = TD sinc(πfTD )e −jπfTD πf Niknejad Advanced IC’s for Comm Sampling Conversion Gain 1.0 0.8 0.6 0.4 0.2 -3 -2 -1 1 2 3 Finally, multiplying by the sinc with our impulse train, we get the overall transfer function. For a IF frequency, reducing the hold time improves the gain of the system. This is kind of a non-intuitive result ! There’s actually a trade-off between the bandwidth of the IF and the hold time, so there’s no free lunch. Niknejad Advanced IC’s for Comm Tayloe’s Sampling Receiver Product detector and method therefor, US Patent 6230000 B1 Note that the mixer is realized by four samplers, each sampling with 25% duty cycle. Such a sampler has I/Q quadrature down-conversion and is balanced. As we have seen, this bilateral mixer also creates a narrowband RF filtering effect which is beneficial for rejection of blocker signals. Niknejad Advanced IC’s for Comm Harmonic Reject (HR) Mixers Niknejad Advanced IC’s for Comm The Need for Harmonic Rejection If the RF bandwidth is wider than an octave, then harmonics of the LO are downconverter to the same IF and corrupt the desired signal. Imagine operating a receiver at 800 MHz. A PA operating at 3X this frequency or 2400 MHz will couple into the receive chain and get downconverter by the third harmonic of the LO. Even with an Rx filter, the coupled signal may be very strong (30 dB of isolation implies a 0 dBm 3rd harmonic blocker) Cable receivers are among the systems that process very wide swaths of bandwidth. Reconfigurable LTE transceivers operate from 500 MHz to 3600 MHz and have HR requirements. Wideband Software Defined Radios (SDR) also need it for similar reasons. Niknejad Advanced IC’s for Comm Why Mixers Like Harmonics Even if we drive the LO with a sine wave, as long as the amplitude exceeds the threshold voltage of the switching, then the output is steered more like a square wave, which is rich in harmonics. ∞ 4 X sin(2π(2n − 1)fLO t) s(t) = π 2n − 1 n=1 Niknejad Advanced IC’s for Comm … F … … … How to Synthesize a Sine Wave? -T/2 T/2 t T -7/T -5/T -3/T -1/T 0 1/T 3/T 5/T f 7/T -1 (b) |So(f)| so(t) 1 f … … -T/2 T/2 T … t … F -9/T -7/T -5/T -3/T -1/T 0 1/T 3/T 5/T 7/T 9/T -1 (c) Notice that sampling a sine wave produces a square wave. If Figure 4.10 Sample and hold operation on sinusoid in time and frequency domains for three sampling we oversample the sine (a) wave cycle, we frequencies. fs=2fi. eight (b) fs=4fi.times (c) fs=8fiper . produce a staircase approximation to the sine wave. As the sampling rate is pushed even higher, the resulting is SHS waveform The key observation is that this waveform free from will thehave 3rd harmonic! Can somehow use thisthewaveform even and fewer 5th harmonics. Intuitively thiswe makes sense by inspecting time-domain?signal Niknejad Advanced IC’s for Comm which is an SHS waveform, is the key element to the rejection of unwanted harmonics in Summing Four Square Waves the harmonic-rejection mixer presented in the following section. T/16 ss(t) … … -T/2 t=0 T/2 T T/2 T T/2 T T/2 T t p0(t) … … 2 -T/2 t=0 t p1(t) … … 1 -T/2 t=0 t p2(t) … -T/2 1 t=0 …t Figure 4.14 SHS waveform resulting from shifting of sampling position and the square waves which compose the SHS waveform. The staircase approximation to sine can be synthesized by summing three square waves of the right magnitude. The square waves illustrated in Figure 4.14 need to be scaled in amplitude and phase shifted appropriately to reject the third and fifth harmonics. Signal p0(t) is scaled Niknejad Advanced IC’s for Comm Fourier Series of Square Waves The wave p0 (t) is a 50% duty cycle square wave √ 1 1 1 4 2 (cos(ωt)− cos(3ωt)+ cos(5ωt)− cos(7ωt)+· · · ) p0 (t) = π 3 5 7 The waveform p1 (t) is just a time and amplitude shifted version of p0 (t) √ 2 2 1 1 1 p1 (t) = (cos(ωt) + cos(3ωt) − cos(5ωt) − cos(7ωt) + · · · )+ π 3 5 7 √ 2 2 1 1 1 (sin(ωt) − sin(3ωt) − sin(5ωt) + sin(7ωt) + · · · ) π 3 5 7 Niknejad Advanced IC’s for Comm Synthesizing a Sine from Squares (II) Similarly, p2 (t) is also a time shifted copy √ 2 2 1 1 1 p2 (t) = (cos(ωt) + cos(3ωt) − cos(5ωt) − cos(7ωt) + · · · )+ π 3 5 7 √ 2 2 1 1 1 (− sin(ωt) + sin(3ωt) + sin(5ωt) − sin(7ωt) + · · · ) π 3 5 7 The desired waveform, a sum of the three, is free from 3ω and 5ω √ 8 2 1 p(t) = p0 (t) + p1 (t) + p2 (t) = (cos(ωt) − cos(7ωt) + · · · ) π 7 Niknejad Advanced IC’s for Comm 99 4.5 Harmonic-Rejection Mixer Weldon/Berkeley Architecture RL RL Vout (W/L)sw I (W/L)sw 2 (W/L)sw LO-45 Vin (W/L)in I 2 (W/L)sw (W/L)sw I LO0 (W/L)in 2 (W/L)in Ibias (W/L)sw LO45 2 (W/L)in (W/L)in I 2 Ibias (W/L)in Ibias Figure 4.16 Simplified circuit diagram of the harmonic-rejection mixer. Three parallel mixers are used, each driven by p0 (t), p1 (t), and p2 (t).potentially We must scale the achieved devicesbyappropriately totherealize The scaling could have √ been scaling the load of three the right amplitude (need 2) [Weldon] mixers. However, in this case the current from each mixer is summed and an output Niknejad Advanced IC’s for Comm is multiplied by the SHS waveform. Matching Requirements Therefore comparing the output of the HRM to square wave represents the relative improvement in the suppression of the harmonics. T 2 (1+ )p0(t+ T/2 ) 2 (1+ ) t square wave HD3 1 t p1(t) HD5 HR3 HR5 t p2(t) 1 fLO 3fLO 5fLO 7fLO f Figure 4.17 Graphical representation of HR3 and HR5. Time mismatch and/or amplitude mismatch leads to finite 3rd and 5th. Based on the harmonics of a square wave from Equation (4.5), the relationship between and HD3 in dB is given by HDNiknejad HR3 3 1 20Advanced log IC’s for Comm Matching Requirements (3rd) 40 35 =0.01 30 HR3 (dB) =0.05 25 20 15 10 =0.1 0 0.5 1 1.5 2 2.5 3 3.5 4 (degrees) GoodFigure HR requires good matching ! of gain and 4.18 Thirdvery harmonic rejection as a function Niknejad Advanced IC’s for Comm phase error. 113 4.6 Matching in HRM Matching Requirements (5th) 40 35 =0.01 30 =0.05 25 HR5 (dB) 20 15 =0.1 10 5 0 0.5 1 1.5 2 2.5 3 3.5 4 (degrees) Figure 4.19 Fifth harmonic rejection as a function of gain and phase error. Niknejad Advanced IC’s for Comm labeled for the sakeSynthesis of simplicity. LO Waveform D D1M __ D Q __ clk Q Q1M D D1S __ D Q __ clk Q D1 Q1 D D2M __ D Q __ clk Q Q2M D D2S __ D Q Q2 __ clk Q D2 LO LO Q1M Q1 Q2M Q2 Figure 4.22 Expanded view of divide-by-four and timing diagram. Requires an LO to be provided at 4 times higher frequency Niknejad Advanced IC’s for Comm Harmonic Reject Transmitter 135 5.2 Harmonic Rejection Transmitter Architecture IBB IIF DAC fRF 0 90 ` QBB PA RF Filter QIF DAC = HRM LO2 8 phase generator LO1 Figure 5.1 Block diagram of a Harmonic Rejection Transmitter. Similar concept applies for the receiver. The basic function of the HRT is as follows. The baseband is identical to the homodyne and heterodyne transmitter which were discussed in Chapter 3 as digital inphase and quadrature baseband signals pass through a DAC then a low-pass filter. Niknejad Advanced IC’s and for Comm Forming Gains at Baseband Prof. Al Molnar’s MS thesis had this 8-phase passive mixer with HR suppression [Molnar]. Niknejad Advanced IC’s for Comm Sub-Harmonic Mixers Niknejad Advanced IC’s for Comm Sub-Harmonic Mixer Design Sub-Harmonic Mixer fRF fLO = 0.5fRF The subharmonic mixer is driven by an LO signal that is an integer fraction, or subharmonic, of the desired LO frequency. For example, if the RF signal is 2GHz, and the desired LO is 2GHz for direct conversion, a subharmonic mixer will be driven by a 1GHz LO signal. The advantages are: Lower LO re-radiation through the antenna (LO leakage) Lower LO self mixing (lower DC offset at IF) Relaxed requirement on the device switching speed. Lower LO buffer current Niknejad Advanced IC’s for Comm 25% LO Waveform The idea is to generate an LO signal that is half the desired frequency, yet is rich with a second harmonic, then somehow use the second harmonic, along with the RF, to get desired IF. A 50% duty cycle LO has no second harmonic. However a 25% duty cycle LO has the desired harmonic. VLO TLO /4 TLO The Fourier series expansion of the above square wave is: VΦ1 = " # √ √ 1 √ 2 2 cos(3ω0 t) − cos(5ω0 t) + ... 2 cos(ω0 t) + cos(2ω0 t) + π 3 5 Where ω0 is the subharmonic LO frequency, which in this case is half the desired LO frequency. Niknejad Advanced IC’s for Comm Sub-Harmonic Mixer LO1 + RF IF LO2 The subharmonic mixer topology uses two identical mixers excited by two phases of the 25% duty cycles. The RF signal is multiplied by these two delayed 25% LO signals and the IF is added in phase at the output, as shown in the above figure. Niknejad Advanced IC’s for Comm Sub-Harmonic Mixer Waveforms The two LO signals, VF 1 and VF 2 , are T /2 delayed relative to each other. VLO TLO /4 TLO TLO /2 Niknejad Advanced IC’s for Comm Sub-Harmonic Waveforms VΦ2 " # √ √ √ 1 2 2 = − 2 cos(ω0 t) + cos(2ω0 t) − cos(3ω0 t) + cos(5ω0 t) + · · · π 3 5 Looking at the combined IF output of the subharmonic mixer we can write: 2 VIF = VRF VΦ1 + VRF VΦ2 = VRF cos(2ω0 t) π As seen, the resulting IF output is the product of the RF signal and an effective LO that has twice the subharmonic LO frequency, which is in fact the desired LO. The above analysis was for only one mixer. The question then, how can one build quadrature subharmonic mixer for both I and Q channels? The answer comes in the way to generate a sin(2ω0 t) LO signal using similar 25% duty cycle signals as follows. Niknejad Advanced IC’s for Comm I/Q Sub-Harmonic Mixer VLO TLO /8 TLO 5 TLO /8 The LO signal LOF 3 is delayed T /8 relative to LOF 1 . LOF 4 is delayed by T /2 relative to LOF 3 . The Fourier series expansion of the above square waves is: VΦ3 = VΦ4 = 1 " √ π 1 π 2 cos(ω0 t − π 2 √ ) + sin(2ω0 t) + 2 3 sin(3ω0 t − √ 2 " π 2 √ )+ 2 5 cos(5ω0 t − √ 2 π 2 # ) + ... √ π π π − 2 cos(ω0 t − ) + sin(2ω0 t) − sin(3ω0 t − ) − cos(5ω0 t − ) + ... 2 3 2 5 2 Niknejad Advanced IC’s for Comm # Generation of LO Looking at the combined IF output of the subharmonic mixer we can write: 2 sin(2ω0 t) VIF Q = VRF VΦ3 + VRF VΦ4 = VRF π In order to generate all 4 phases of the subharmonic LO signals, a divide by 4 prescalar needs to be used. This means the VCO needs to run at 4X the subharmonic LO, 2X the RF signal for direct conversion. IFI LO1 LO2 RF 2f0 LO3 V CO ÷4 4f0 LO4 IFQ Niknejad Advanced IC’s for Comm Sub-Harmonic Realization IF (+) IF (−) LO(+) LO(+) LO(−) LO(−) RF(+) RF(−) The above is a simplified implementation of one subharmonic mixer. Since a subharmonic mixer is two mixers in one driven by the same RF input but has two different LO phases, a single common Gm stage is used. Niknejad Advanced IC’s for Comm Mixer First Architecture Niknejad Advanced IC’s for Comm Why Not Mixer First Getting rid of the LNA seems like a really bad idea. Isn’t an LNA a must ? The LNA input impedance is matched to the antenna (or filter) for optimal power transfer into the receiver. Furthermore, the impedance match simplified board design since transmission lines can be used to bring in the signal from the antenna without worrying about any potential impedance transformation (except length should be minimized to reduce losses). Niknejad Advanced IC’s for Comm Why Not (Noise and LO Leakage) Recall that the LNA is there to minimize the noise contribution of sub-sequent stages, so it should be designed with low noise and as high of a gain as possible. Also the LO signal leakage to the antenna port is minimized, an issue if we remove the LNA. Niknejad Advanced IC’s for Comm Maybe ? In practice the gain is limited by linearity considerations, so a trade-off is made. The first potential benefit of a mixer first architecture is therefore to realize much higher linearity. If the linearity can be made good enough to reduce or eliminate off-chip filters (LC, SAW, or other bulky components), then there’s a potential win. But this means mixer noise cannot be too high. Since the filters introduce about 1-2dB of loss at the input of the Rx, perhaps a mixer first Rx can be about 1-2 dB noisier than an LNA first counterpart. Niknejad Advanced IC’s for Comm Cook’s Mixer First Receiver Fig. 1. Transceiver block diagram. Fig. 2. Simplified transceiver front-end schematic. Ben Cook, a Berkeley grad working with Prof. Pister, proposed a mixer first receiver and was among the first of recent IC designers to realize the benefits of n-path filtering offered by passive bilateral mixers. [Cook] Fig. 3. Ben Cook and Axel Berny started a company based on this concept (Passif) that was acquired by Apple. Quadrature VCO utilizing back gate coupling. and DC level of the VCO signal driving the switches. The mixer outputs differentially drive a bandpass filterNiknejad comprising Fig. 4. Circuit model for tapped-capacitor resonator. Advancedcoupled IC’s forVCO Comm back-gate architecture was used here because it Mixer First Motivation and Benefits Use a passive mixer to achieve very good linearity. Take advantage of the bilateral nature to realize a bandpass response to improve out-of-band linearity. Make the system tunable by controlling the LO frequency, rather than tuning capacitors or inductors, potentially realizing a very broadband receiver that is also linear and robust to out-of-band interference. Niknejad Advanced IC’s for Comm Andrew/Molnar Mixer First Proposal Fig. 9. Equivalent baseband amplifier noise models. Fig. 11. Simulated and analyt Fig. 10. Proposed passive-mixer first receiver with resistive feedback amplifiers. the squared noise by ) a ), yield Use a multi-phase voltage mode passive mixer driventiplying by by sistive feedback. We find the new effective present on each delayed copies of the LO. As we shall see, we can combine the branch by applying the Miller effect to the feedback resistor : output phases to form an I/Q signal with desirable properties. This multi-phase idea was inspired by the Weldon (31) sub-harmonic mixer and a similar structure using 8-phases was Substituting the new into the impedance matching LTI modelM.S. shows thesis. that we can perform impedance matching using in Molnar’s the amplifier feedback resistors. Once we have added the feedback amplifiers to implement , the noise performance changes as well. Whereas most of Niknejad Advanced IC’s for Comm We have verified this lation using periodic ste , Equivalent Model REWS AND MOLNAR: IMPLICATIONS OF PASSIVE MIXER TRANSPARENCY 3093 1. (a) Simplified circuit model of 4-phase passive mixer. (b) LO driving waveforms. (c) Equivalent model to (a), with ng nature of the waveforms in the middle. lumped with based on nonoverAssume non-overlapping 25% clock drive to the mixers. Assuming the switches turn on instantly, and have an ng range. These advantages imply that passive mixer-first with fundamental frequency and time varying on-resistance Rsw .the flex- phase and amplitude , which capture both modulation ivers will likely provide the next step in of improving ty and performance of highly integrated wireless receivers. and offset frequency of the received signal. If the amplitude and Note that at any given time,phase oneoffset switch is always by can thebe approxthey change slowly relative seen to PASSIVE MIXER TRANSPARENCY: FIRST-ORDER ANALYSIS imated as constant over a given LO period, and the input can be source, so we can move switch resistance and lump it with the as he passive mixer analyzed here contains four switches (tran- approximated antenna resistance Ra to form Ra0 = Ra + Rsw . ors) which are successively turned on in four nonoverlap- (2) g, 25% duty-cycle phases over the course of one local ostor (LO) period [1], [10], [15]–[17]. These nonoverlapping To compute the input impedance presented by the mixer to es are necessary for preventing the I-Q crosstalk described the antenna, we start by computing the voltage across each of Advanced IC’s for Comm 12]. The input port of the mixer is connected directlyNiknejad to the Charge Balance Assume that the baseband acts like a sampler and holds the voltage over the LO cycle. In other words, the time constant RB CL TLO . Using this assumption, we can invoke charge balance and derive the magnitude of the baseband voltage on stage VC ,m (m = 0, · · · , 3) Qm = VC ,m TLO = charge flowing into RB RB Z = (m+1)TLO 4 mTLO 4 + T − LO 8 Niknejad TLO 8 VRF − VC ,m dt Ra0 Advanced IC’s for Comm Charge Balance for Sinusoidal Drive If the input VRF is assumed to be a tone at ω0 , we have VRF = A cos(ωLO t + φ) VC ,m TLO A sin(ωLO t + φ) T2 VC ,m TLO = − R0 RB ωLO Ra0 4 a T1 1 1 A (sin(ωLO T2 + φ) − sin(ωLO T1 + φ)) VC ,m TLO + = 0 RB 4Ra ωLO Ra0 √ 2 2 RB mπ VC ,m = A cos(φ + ) π RB + 4Ra0 2 Niknejad Advanced IC’s for Comm Up-Convert Baseband Voltage The baseband voltage will induce an RF current to flow from the antenna VRF (t) − Vx (t) Ia (t) = Ra0 To find Vx (t), we note that it’s a staircase function due to the periodic rotation of the switches with a fundamental period of TLO . Four successive outputs in one cycle are given by {cos(φ), − sin(φ), − cos(φ), sin(φ)} = {i, −q, −i, q} If we find the fundamental coefficient of the Fourier series, we can find Vx,ωLO : Vx,ωLO = A 8 RB 8 RB cos(ωLO t + φ) = VRF 2 2 0 π RB + 4Ra π RB + 4Ra0 Niknejad Advanced IC’s for Comm RF Current We can now calculate the RF current flowing from the antenna, and hence the impedance seen by the antenna Ia,ωLO = VRF − Vx,ωLO Ra0 4Ra0 + 1−8 R π2 B = VRF (t) 0 Ra RB + 4Ra02 Note that as RB → 0, Ia = VRF /Ra0 . RB is in series with Ra0 . Note that as RB → ∞, Ia = VRF (1 − 8/π 2 )/Ra0 . This means that there’s an additional equivalent shunt impedance in parallel to RB ! Niknejad Advanced IC’s for Comm S ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 57, NO. 12, DECEMBER 2010 Mixer LTI Equivalent Circuit ause ed to step ding hase y the o the Fig. 3. LTI equivalent circuit for passive mixer with due to harmonics and The first Ra0 term is. obvious, but the next two terms are not impedance-transformed at all obvious. The term γRB is the baseband equivalent resistance seen by the antenna. The Rsh term is in shunt and represents additional power loss. It will be shown that this is actually the power loss due to the harmonics of the harmonics of the LO. Up to now it was derived using charge balance but if we go through each harmonic and calculate it’s loss as seen from the input port, we arrive at the same result. Niknejad Advanced IC’s for Comm Equivalent Circuit Elements From the model we can calculate the equivalent impedance Ia (ωLO ) = VRF (ωLO ) Ra0 γRB γRB + Rsh + Ra0 Rsh + Rsh γRB Equate this to the expression derived from charge balance to get 2 γ = 2 ≈ 0.203 π 4γ Rsh = Ra0 ≈ 4.3Ra0 1 − 4γ Note that Rsh was derived assuming Ra0 is constant over all frequencies, which is not a good assumption for most antennas or filters that may precede the antenna. Niknejad Advanced IC’s for Comm Impedance Matching Given that the input impedance has a real part that depends on the switch resistance (naturally) but also on the baseband impedance, we can turn the baseband impedance to realize a match Rin = Rsw + γRB ||Rsh Rsw < Rin < Rsw + Rsh We’d like to make Rsw Ra but we’ll return to this point later. For now, solve for RB RB = 1 Rsw Ra − Rsw Rsh γ Rsw + Rsh − Ra Niknejad Advanced IC’s for Comm Antenna Impedance Variation In the above charge balance derivation, it was assumed that the antenna impedance was constant, even at LO harmonics. Since the voltage Vx has harmonics of the LO, the current draw from the antenna at these harmonics will differ. If we write Vx as a Fourier series and account for the antenna impedance variation by defining Ra0 (nωLO ) = Ra (nωLO ) + Rsw Then the current through the switches is given by ∞ X Vx,n (t) VRF (t) Ia (t) = 0 − Ra (ωLO ) Ra0 (nωLO ) n=1,3,··· Invoking charge balance, we arrive at −1 ∞ X 1 Rsh = n2 Ra0 (nωLO ) n=3,··· Niknejad Advanced IC’s for Comm Rsh Harmonic Interpretation We now have an alternative expression for Rsh Rsh = ∞ X n=3,5,··· −1 1 n2 Ra0 (nωLO ) Let’s verify that this agrees with our earlier calculation when Ra is constant −1 ∞ X 1 Rsh = n2 (Ra + Rsw ) n=3,5,··· Rsh = π2 −1 8 1 (Ra + Rsw ) Niknejad −1 = Ra0 Advanced IC’s for Comm 4γ 1 − 4γ Non-Zero IF Excite the circuit with a frequency above or below the LO (to produce a non-zero IF) VRF (t) = A cos((ωLO + ωIF )t) Assume baseband impedance is an RC but excited at non-zero ZB (ωIF ) = RB || 1 RB = jωIF CL 1 + jωIF CL RB Repeating the math and assuming ωIF ωLO , an essentially identical expression is derived (now with ZB ): Vx (ωRF ) = VRF (ωRF ) Niknejad ZB 8 π 2 ZB + 4Ra0 Advanced IC’s for Comm ANDREWS AND MOLNAR: IMPLICATIONS OF PASSIVE MIXER TRANSPAREN General LTI Model Given above result, it’sfor now easy to show that the . Fig. 6. New LTIthe equivalent circuit frequency-dependent antenna port is loaded by an impedance given by to (6) γRB ||Rsh , we findZinan (ωIFexpression ) = Rsw + becomes essentially identic 1 + jωIF CL γRB ||Rsh From the RF port, this is transformed into a bandpass characteristic similar to what we have seen before (due to mixer transparency, a baseband low-pass is converted into a bandpass response). This then produces the same derivation of the current Niknejad Advanced IC’s for Comm (28 a Noise Model IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 57, NO. 12, DECEMBER . Equivalent noise models for LTI model from Fig. 3. . Given that this is an LTV system, we must account for all the noise folding that occurs from every harmonic of the LO. It is fortunate that this can be simply modeled by an equivalent Rsh . In fact, we have already derived that Rsh models the power loss due to the harmonics, so this is not completely Equivalent baseband amplifier noise models. surprising With this insight, we have a very simple LTI model. Niknejad Advanced IC’s for Comm Noise Analysis Each noise term is summed to calculate the total output noise. This expression is then converted to a NF by normalizing to the source noise Rsh Rsw + F =1+ Ra Ra Ra + Rsw Rsh 2 RB + Ra Ra + Rsw γZB 2 In practice, the baseband resistance is realized as a shunt feedback resistance with an amplifier. In this situation, the baseband amplifier contributes noise Rsh Ra + Rsw 2 γRF Ra + Rsw 2 Rsw + + + F =1 + Ra Ra Rsh Ra γRF va2 Ra + Rsw + Rsh 2 Ra + Rsw γ + 4kTRa γRF Rsh Niknejad Advanced IC’s for Comm Noise Optimization Zsh should be maximized to minimize contributions of harmonics to noise and loss. Zin = Rsh + γZB ||Zsh Also, this will maximize the tuning range of the system (by varying RB ). But note that maximizing Zsh requires one to maximize the antenna impedance at the harmonics of the LO, opposite to an LC tank or a resonant antenna. Niknejad Advanced IC’s for Comm Optimum Switch Size Since the switch adds parasitic capacitance in parallel to Zsh , we cannot arbitrarily make the switch large. Za (ω) → 0 at ω → ∞ Za0 (ω) → Rsw for ω → ∞ Reducing Rsw reduces Zsh at higher frequencies, which increases the noise! There’s an intermediate width for optimal performance. If we have a very high Q antenna such that the antenna impedance is zero at harmonics, then Rsh = 4.3Rsw and so F =1+ Rsw (Ra + Rsw )2 (1 − 4γ) + Ra 4γRa Rsw For Fmin ≥ 4.05 dB, choose Rsw p Rsw = Ra 1 − 4γ Niknejad Advanced IC’s for Comm d 8-Phase Architecture for varying showing shift in center frequency ase passive mixer. e repeat this analysis for the case where Noise figure can be improved if we can minimize the harmonic contributions to Rsh . One way to do this is to build a harmonic reject mixer. By using an 8-phase mixer with 12.5% duty cycle, the number of harmonics is cut in half compared to a 4-phase mixer (eliminate 3rd, 5th, 11th, 13th, and so on)! This boosts the value of Zsh and also changes the value of γ √ 2 γ8ph = 2 (2 − 2) π 8γ8ph (Rsw + Za ) ≈ 18.9(Rsw + Za ) Zsh,8ph = 1 − 8γ8ph The optimum switch size is also modified and the minimum achievable noise figure is close to 2 dB ! Niknejad Advanced IC’s for Comm References Weldon Jeffrey Arthur Weldon, High Performance CMOS Transmitters for Wireless Communications, Ph.D. Dissertation, 2005. Cook B.W. Cook, A. D. Berny, A. Molnar, S. Lanzisera, K.S.J. Pister, “An Ultra-Low Power 2.4GHz RF Transceiver for Wireless Sensor Networks in 0.13/spl mu/m CMOS with 400mV Supply and an Integrated Passive RX Front-End,” IEEE Internationa Solid-State Circuits Conference(ISSCC), Digest of Technical Papers, pp.1460,1469, 6-9 Feb. 2006. Molnar A. Molnar, B. Lu, S. Lanzisera, B.W. Cook, K.S.J. Pister, “An ultra-low power 900 MHz RF transceiver for wireless sensor networks,” Custom Integrated Circuits Conference, 2004. Proceedings of the IEEE 2004 , vol., no., pp.401,404, 3-6 Oct. 2004. AndrewsTCAS C. Andrews and A. Molnar, “Implications of Passive Mixer Transparency for Impedance Matching and Noise Figure in Passive Mixer-First Receivers,” IEEE Trans. on. Circuits and Systems-I: Regular Papers, vol. 57, no. 12, Dec. 2010. AndrewsJSSC C. Andrews and A. Molnar, “A Passive Mixer-First Receiver With Digitally Controlled and Widely Tunable RF Interface,” IEEE J. of Solid-State Circuits, vol. 45, no. 12, Dec. 2010. Niknejad Advanced IC’s for Comm