key: cord-0972305-c63h85sz
authors: Benali, Saif; Trabelsi, Hatem
title: Analysis of 25% Duty-Cycle Up-Conversion Passive Mixer for UWB Transmitter
date: 2020-06-01
journal: Int J Electron Commun
DOI: 10.1016/j.aeue.2020.153295
sha: 06ea24f161a286746fa8833d230c6e74d937293a
doc_id: 972305
cord_uid: c63h85sz

The performance of differential up-conversion passive mixer operating in the 3–5 GHz band for UWB transmitter driven by 25% duty-cycle clocks is studied and analyzed. A reasonably accurate LTI equivalent circuit accounting for harmonics is derived. We have demonstrated that the conversion gain, input and output impedances of the proposed LTI equivalent circuit matches those of the LTV up-conversion passive mixer. The LTI model can be used to define the reachable design space. The mixer input and output impedance can be tuned by varying resistor at the mixer output allowing for impedance match to connected circuits. We have shown how each design parameters impacts critical performance of the up-conversion passive mixer. Simulations of the proposed up-conversion passive mixer show -3.68 dB of conversion gain, 15.9 dBm of OIP3, 116.7dBm of OIP2 and a NF as low as 5.2 dB while consuming 1.5 pW. Furthermore linearity and ports isolation performances due to voltage threshold mismatch variation of ±60mV shows an OIP3 greater than 15.85dBm and an LO_RF isolation smaller than -51.9dB.

The standard Wireless Body Area Networks (WBAN) is known to provide efficient, low power and optimized wireless link between implantable and/or wearable devices for continuous and real time physiological signal monitoring [1] [2] . In particular WBAN wearable devices can be used to wirelessly monitor Coronavirus patients without exposing medical staff to the virus, therefore limiting the spread of the virus within the hospital.

The Ultra-wideband (UWB) offers robust performance for WBAN with low complexity transmitter and ultra low power operation. The UWB power spectral density emission limit of -41.3 dBm/MHz provide safe power levels for the human body [3] [4] [5] . The wireless sensor node should be able to operate until ten years under battery energy supply. Passive mixerbased UWB transmitter allows high performance and low power wearable devices implementation, therefore increasing battery life. Unlike passive mixers for receivers, there are only a few works that deal with passive mixers for UWB transmitters. Active mixers have the advantage of having high conversion gain and good baseband-RF ports isolation. Various circuit techniques have been proposed to increase the conversion gain and improve the linearity of active mixers over a wide bandwidth [6] [7] .

On the other hand, passive mixers have the following performances which do not exist in active mixers [8] [9] [10] [11] : extremely low power consumption because passive mixer dissipates no DC power, robust linearity, good noise figure performance, the passive mixer-based transmitter occupies a very small chip area, better LO-RF and LO-baseband ports isolation performance [12] .

The 50% duty cycle CMOS up-conversion mixer in [13] is based on a chain of switched CMOS inverters for 2.45 GHz and 400 MHz bands. The circuit has similarities with the H-Bridge ring mixer. The up-conversion mixer achieves a conversion gain of -5dB and consumes 23.1mW from 3.3V supply voltage. In [14] authors studied and analyzed conversion gain and noise performances of 25% duty-cycle passive mixer for IQ direct-conversion transmitters.

They made Linear Time Variant (LTV) as well as Linear Time Invariant (LTI) analysis to derive the input/output transfer functions at LO and its harmonics. The transmitter proposed in [15] employs a pseudo double-balanced up-conversion passive mixer with more than 22dB LO leakage rejection. The power consumption is 14.4mW under 1.2V supply.

The double-balanced IQ passive mixers with 50% duty cycle are widely used in receivers as down converter [9] . To be able to use this structure in transmitter as up-converter it is necessary to use a duty cycle less than or equal to 25%. Indeed for a duty cycle of 50% the two channels I and Q on the RF sides cannot be shorted together because at any given instant there is a switch from I channel and another from Q channel at the ON state. On the other hand, for a duty cycle less than or equal to 25% there is a single switch in the ON state which eliminates any possible interaction between I and Q channels.

In this paper we present an analysis of a 25% duty cycle up-conversion passive mixer intended to be used in UWB transmitter for WBAN applications. Section 2 presents transmitter block diagram, modeling details of the 25% duty-cycle up-conversion passive mixer and derives the LTI equivalent circuit. A comparative study between LTI equivalent circuit accounting for harmonics and the LTV up-conversion passive mixer is carried out in Section 3. Implication of LTI equivalent circuit for design of passive mixer-based transmitter is discussed in Section 4. Section 5 concludes the paper.

The proposed transmitter block diagram front-end is shown in Fig. 1 . The differential baseband (BB) analog signal from the DAC is first filtered with a Low Pass Filter (LPF) to remove image components and quantization noise. The filtered baseband signal are upconverted to UWB RF using a differential double-balanced up-conversion passive mixer. The differential output UWB RF signal is buffered with a high input impedance buffer then 4 converted to single ended signal. The output signal of the driver is typically amplified by a power amplifier (PA). Differential double balanced up-conversion passive mixer can eliminate BB and LO feed through. Voltage mode passive mixer operation was adopted because designing of a voltage buffer is simpler than the current buffer at UWB RF frequency. Currentmode passive mixer is typically used in receivers [14] . the mixer. R lpf is the LPF output resistance. In the following sections we will study and analyze in time domain the proposed time-variant up-conversion mixer circuit.

To analyze the differential double balanced up-conversion passive mixer we begin by presenting its basic operation in the time domain. We assume that the switches are ideal, except that they have non zero ON resistance R on . Fig. 3 shows the proposed half circuit used to analyze the up-conversion passive mixer. C par models the input parasitic capacitor of the NMOS switch.

The parasitic capacitors of the switches can be lumped into the baseband impedance. These capacitors are ignored in the following analysis. The differential baseband input voltage from LPF, V bb (t) is given by:

where , ( and is the magnitude. Let us define the following = -ω bb ≪ ω LO )

periodic functions corresponding to the four LOs.

(2)

The output voltage shown in Fig. 3 is given by: 1 ( ) when =1 and when

In the ON phases, the charge flow Qc into RF capacitor C RF during can be written as:

and . In the OFF phases the capacitor C RF discharges = 2 = + + 1 through R RF and the voltage is given by:

( 1 -) 4 < < and the discharge of C RF through R RF correspond to equation (8) .

Conservation of charge implies from which V M can be computed by: h = Up-conversion passives mixers simultaneously down-converts the RF signals to baseband and up-converts those to RF (transparency property) as reported in [11] . Since the LO phases are not overlapping in time, the down-converted voltage can be written as (10) . ( )

LO signals can be represented by their Fourier Series.

Then we can derive (16) . For , the voltage is given by (17) 

Replacing by its expression (9), the component of the down-converted voltage at the baseband frequency is obtained (18) . 

At baseband frequency the total input current of the up-conversion passive mixer , (Fig.2) is given by (20) . According to (19) and (20) the input current relation can be determined as shown in the equation (21). ( ) Fig. 4 , the input impedance can be determined :

We can notice that in equation (26) the input impedance can be tuned by changing R RF . The range is limited as shown in (27).

This analysis will be confirmed by simulation in the next section.

In order to set the optimal size of NMOS switches that set the switching ON resistance R on , we made S-parameter simulation of the circuit of Fig. 4 using ADS (Advanced Design System) tool. Fig. 5 Fig. 5(b) . We can see that the optimum point OIP2=116.6dBm

is obtained for V GS =0.76V, this corresponds to R on =38.3Ω. =8.5GHz). So the 25% duty-cycle differential double balanced passive mixer did the upconversion correctly. Each transistor drain is biased at V p =300mV and without oscillator circuit the up-conversion passive mixer consumes only 5pA.

Now we will compare input and output impedances of the LTI equivalent circuit shown in Fig. 4 and the LTV up-conversion passive mixer at LO duty cycle of 25% shown in Fig. 2 .

For the LTI equivalent circuit, input and output impedances are denoted by Zinequ and Zoutequ respectively. For the LTV up-conversion passive mixer, input and output impedances are denoted by Zinupmix and Zoutupmix respectively. Fig. 7(a) shows simulation results comparing the magnitude of input and output impedances for swept R RF from 1 Ω to 40 KΩ. We can observe that input and output impedances of the LTI equivalent circuit matches the input and output impedances of the LTV up-conversion passive mixer. This simulation proves that the LTI equivalent circuit can be used to analyze the 25% duty cycle differential upconversion passive mixer for UWB transmitter. We will study in section 3 the LTI equivalent circuit accounting for harmonics.

In the previous analysis, we defined the down-converted voltage V d only at baseband frequency f bb . However, Fig. 6 (b) indicates that V d contains harmonics at (2f LO ± f bb ), (2f LO ± 3f bb ), 3f bb ,...as well as its fundamental. Therefore, the input current I bb can be computed:

The first term at f bb represents the current coming from the LPF and the other terms represent the return current due to V d . Equation (27) show that the current I bb decreases compared to the fundamental current used in the previous analysis. Thus, the input impedance Zinupmix will increases. In order to account for the dissipation of power due to these harmonics we multiply the impedances by a factor h as shown in Fig. 8 . This model presents a reasonably accurate LTI equivalent circuit of the up-conversion passive mixer accounting for harmonics.

(a) (b)

A comparison between conversion gain of the 25% up-conversion mixer and its LTI equivalent circuit implies that h=2 and where . 1 = 0.9cosφ cosφ ≠ 0

In the context of the above LTI equivalent model, Fig. 9 

In order to reduce noise we need to reduce R on and increase R RF as shown in Fig. 11(a) and Fig. 11(b) . But this will increasing the size of the transistors and thus will increase the parasitic input capacitor of the NMOS switches. In addition it is recommended to have an impedance matching at the input and output of the up-conversion passive mixer. For this we have made S parameters simulation for several values of R on as shown in Fig. 11 (c) for S11 and Fig. 11 (d) for S22. R RF was set to 3KΩ for a noise factor of 3.36. From Fig. 11(c, d) , we deduce that R on must be less than 46 Ω to get impedances matching. In order to reduce parasitic input capacitor of the NMOS switches we choose the same dimensions of the NMOS (W=42.2µm and L=0.18µm) used in the previous simulation taking into account fundamental components only. We can see also whether these dimensions hold or not if we account for harmonics.

Now we will evaluate the performance of the up-conversion passive mixer using the dimensions of the transistors defined in the previous section. To do this, we start by evaluating OIP2 for different widths of the transistors. In the Fig. 12(a) we can see that the maximum of OIP2 corresponds to W = 41µm. This value is very close to the value found in the previous section (W = 42.2µm). This confirms that the LTI equivalent circuit model accounting for harmonics in Fig. 8 can be used to find the up-conversion passive mixer design parameters. When we studied the up-conversion passive mixer circuit we assumed that all switches exhibit identical characteristics and bias. In reality NMOS transistors used in this up-conversion mixer are not perfectly matched or balanced. They exhibit voltage threshold (V th ) and W/L ratio mismatch caused by wire bonding variation and process tolerance. V th depends on the doping levels in the channel and the gate, and these levels vary randomly from one transistor to another [16] . V th mismatch lead to DC offset at the circuit nodes. First this may drive the stages of the circuit into nonlinear operation or saturation. Second this will produce some LO to RF, LO to BB and BB to RF feedthrough. Since this up-conversion mixer is based on identical NMOS transistors connected to achieve a double balanced differential structure, it is important to evaluate its inter modulation distortion performance as well as its isolation between ports in case of voltage threshold mismatch. 

The above analysis aims to offer an equivalent LTI circuit as close as possible to a real differential up-conversion passive mixer with LO duty cycle=25% for UWB transmitter. The accuracy verification of the proposed LTI model was done using the simulation results of input (a) (b) and output impedances as well as the conversion gain shown in Fig. 7, 9 and 10 . The resulting curves matches those found with LTV up-conversion passive mixer which clearly shows that the LTI model is equivalent to the actual circuit. The model proposed in Fig. 8 [14] , the authors used a purely capacitive output load, they neglected the effect of input AC coupling capacitors and they assumed the LO clocks are ideal with 25% duty-cycles and zero rise and fall times. They find a maximum conversion gain of -13dB. One implication of having this LTI equivalent model is that designers will quickly define the reachable design space from the constraints on the performance of the mixer. Thus allowing to optimize components sizes of the up-conversion passive mixer. Using the simple LTI equivalent circuit, we can see how each design parameters impacts critical performance of the up-conversion passive mixer like input/output matching, linearity, noise and power consumption.

Another implication is that both LTI equivalent circuit of the 25% duty cycle up-conversion passive mixer in Fig. 4 and Fig. 8 can be used for different values of BB_freq, R RF , C RF , R lpf and R on . For medical applications using devices worn by patients and powered by battery, the transmitter must be low power. For this, the proposed mixer is the best candidate thanks to its negligible power consumption. Table 1 presents the key performances of the proposed differential 25% duty-cycle up-conversion passive mixer and compare them with published works. The proposed up-conversion passive mixer performs higher OIP3 than works in [17] [18] [19] [20] [21] while consuming only 1.5pW. Regarding conversion gain this work obtains -3.68 dB. This value is greater than the gains from passive mixers published in [13] and [19] . This upconversion passive mixer achieves the best ports isolation performances.

We have presented analysis and simulation of a differential up-conversion passive mixer with 25% duty cycle intended to be used in UWB transmitter for WBAN sensor node. We

proposed an equivalent LTI model of this mixer then we studied the implication of this model in the design. After that we analyzed the influence of the different design parameters on linearity, conversion gain and noise figure. We studied the impact of voltage threshold mismatch on linearity and ports isolation performances. Using the results found in this paper, (2)): 2012.

Ultra-wideband transmitter for low-power wireless body area networks design and evaluation

A fully-integrated, short-range, low data rate FM-UWB transmitter in 90 nm CMOS

A 3-5 GHz FSK-UWB Transmitter For Wireless Personal Healthcare Applications

One Mbps 1 nJ/b 3.5-4 GHz Fully Integrated FM-UWB Transmitter for WBAN Applications

An RF-LO current-bleeding doubly balanced mixer for IEEE 802.15.3a UWB MB-OFDM standard receivers

Design of an active CMOS subharmonic mixer with enhanced transconductance

A passive-mixer-first receiver with baseband-controlled RF impedance matching, 6 dB nf, and 27 dBm wideband IIP3

Analysis and optimization of current-driven passive mixers in narrowband direct-conversion receivers

Analysis and optimization of direct-conversion receivers with 25 duty-cycle current-driven passive mixers

Implications of passive mixer transparency for impedance matching and noise figure in passive mixer-first receivers

A low-power, low-EVM, SAW-less WCDMA transmitter using direct quadrature voltage modulation

CMOS upconversion mixer with filterless carrier feedthrough cancelation and output power tuning

Analysis of Direct-Conversion IQ Transmitters With 25% Duty-Cycle Passive Mixers

A 1-Gb/s 6-10-GHz, Filterless, Pulsed UWB Transmitter With Symmetrical Waveform Analysis and Generation

Design of Analog CMOS Integrated Circuits. McGraw-Hill Series in Electrical and Computer Engineering

A low-power up-conversion CMOS mixer for 22-29-GHz ultra-wideband applications

2.4/5.7 GHz dual-band high linearity Gilbert up-converter utilizing bias offset TCA and LC current combiner

CMOS DSB Transmitter With Low TX Noise for UHF RFID Reader System-on-Chip

Harmonic-based nonlinearity factorization of switching behavior in upconversion mixers

Low power upconversion mixer for medical remote sensing

An all-digital frequency tunable IR-UWB transmitter with an approximate 15th derivative Gaussian pulse generator

☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: