Research Article  Open Access
Apratim Roy, "HighGain PowerEfficient Front and BackEnd Designs for a 90 nm TransmitReference Receiver", International Scholarly Research Notices, vol. 2012, Article ID 435209, 15 pages, 2012. https://doi.org/10.5402/2012/435209
HighGain PowerEfficient Front and BackEnd Designs for a 90 nm TransmitReference Receiver
Abstract
A new microwave receiver configuration which transmits reference pulses embedded in data streams for synchronization is analyzed with a 90nm IBM CMOS standard. A twostage cascode lownoise amplifier (LNA) is proposed for the receiver frontend which is matched by a passive network to save on powerexpensive matching techniques. The amplifier exploits a doubledifferential topology and achieves a below 4 dB noise figure near the center frequency. The overall 3dB bandwidth is 3.3 GHz with peaking up to 20.5 dB in the band. The backend of the receiver is implemented through an adjustable analog windowdetection circuit. It avoids the use of control voltage generators and samplehold (S/H) blocks to save electronic overhead and is simulated with a 0.1~2.0 Gbps pulse stream. The achieved speedtopower ratio for the backend has a maximum limit of 266 GHz/W. When compared against simulated results of published literature, the proposed designs show improved performance in terms of smallsignal gain, noise, speed, and power dissipation.
1. Introduction
The ultrawideband technology, with benefits like high data rate, low cost, and low complexity, has been considered as a promising initiative for shortdistance wireless applications [1]. The main challenge of designing a wideband transceiver involves satisfying linearity, reverse isolation, gain, and noise requirements over a wide bandwidth. Different circuit techniques have been proposed in recent literature to achieve wideband operation for a radiofrequency (RF) frontend [2, 3]. Focus on the design of the backend section of the receiver, on the other hand, has received relatively less attention. The wide variety of modulation and multiplexing techniques employed by the ultrawideband (UWB) system allow it to achieve high bit rates over a wide frequency range. For example, recent developments in UWB standards have proposed to exploit OFDM (orthogonal frequency division multiplexing), IR (impulse radio), and TR (transmitreference) modulation techniques. Among them, IRUWB is suitable for low rate applications (e.g., sensor network) and OFDMUWB is more appropriate for high data rate transmission [4].
The idea of a transmitreference (TR) system is that by injecting a reference pulse over the same channel as the information signal estimation of the channel model (to be directly used for convolution by a mixer) can be avoided. It is recognized that a TR system may face some limitations for a bandlimited channel, but it allows data symbols to be decoded without direct calculation of multipath channel coefficients [5]. To realize this standard, efficient implementation is necessary for a millimeterwave frontend lownoise amplifier (LNA) which immediately follows the receiving antenna and an RF multiplier (mixer) which correlates the received pulses with their delayed version [6]. As design requirements, this lownoise amplifier needs to provide reduced power consumption, low reflection (return) losses at peripheral ports, and reasonable forward gain. We aim to incorporate a doubledifferential topology for the LNA to provide less sensitivity to bias supply or commonmode noise, improved power supply rejection ratio, and a more linear behavior [7]. To complete the backend section of the TRreceiver, a window comparator subcircuit with a current limiting mechanism is proposed which determines the final speed to power ratio at the receiver loadport. It avoids auxiliary electronic overheads in the form of control voltage generator and samplehold blocks, therefore improving the overall performance of the transceiver.
To initiate the design of the receiver frontend, a highgain twostage cascode amplifier using a differential configuration is presented on an IBM .09μm CMOS platform. The LNA operates in the frequency band with a 3.3 GHz bandwidth. Noise figure stays around 3.5 dB near the center point and the circuit is expected to be unconditionally stable according to Rollett criterion. Passive matching techniques are used at the interfacing ports to avoid area consuming transmission line matching. The amplifier is followed with a 90 nm threshold detection block using a separate control circuit to manipulate the ratio of integrated current mirrors. Its speed to power ratio ranges from 14 to 266 GHz/W as the input capacity is raised from 0.1 to 2.0 Gbps. The decision circuit dissipates only 7.5 mW of power (including bias circuits) and completes the implementation of the backend part of a TRreceiver.
The paper is organized as follows. Section 2 introduces the proposed front and backend architectures of a TRreceiver. The operation of the microwave amplifier circuit and the design of the window comparator are discussed in detail in Sections 3 and 4, respectively. The results from the frontend amplifier and the decision circuit in the backend occupy the focus in Section 5. Finally, Section 6 summarizes the performance of the designs and compares them with simulated examples from published literature.
2. Proposed TR Front and BackEnd Architectures
The microwave frontend including the feature of selfsynchronization and the backend with a decision block are presented in Figure 1(a) in the form of a traditional TRreceiver. For this architecture, a simplified transmitted TR frame typically consists of a “data pulse” modulated by signal message and a “reference pulse” necessary for synchronization. The two pulses are separated by a temporal displacement (delay) predefined during the design of the transceiver. This TRreceiver does not depend on direct estimation of the interchip channel and requires an architecture with a lower power demand. For Figure 1(a), demodulation is executed without obtaining channel state information and the scheme is resistant to distortion due to multipath propagation [6]. The transmitted signal passes through the channel for wireless interconnection and is received by the collecting antenna which relays it to a microwave frontend lownoise amplifier (LNA). The amplifier boosts the degraded signal while limiting the thermal noise accumulated in the frontend. The amplifier output (RF signal) and its delayed version (LO signal), produced by a specially designed delay unit (DU), are subjected as excitations to an RF mixer or correlator. This feature of selfsynchronization reduces the receiver's architectural complexity but introduces the challenge of designing analog delay lines to align the LO signal with the amplifier output [8]. At the end of the frontend, the RF mixer facilitates the process of decoding the input data form the pulse stream. So, it is evident that the microwave LNA is a critical component for the receiver frontend of a TR scheme. It is also observed that a majority of RF LNAs reported in literature have been implemented with singleended configurations [9–11].
(a)
(b)
The proof of the operation of the TR transceiver has been reported in detail in literature [12–16]. According to the model presented in [12], the signal of a TR scheme is grouped in individual message blocks. The length of each block is where the number of “data pulses” and “reference pulses” are denoted by and , respectively. If is the frame period of the standard, is a gaussian white noise function, is the received pulse function, and stands for the message stream, the received signal collected by the antenna in a TR scheme can be modeled with Here the noise function has a spectral density of and is calculated using the channel impulse response and the transmitted pulse function. If is the delay suffered by a cluster , is the delay of multipath relative to the arrival of cluster , and is the multipath gain factor, the channel impulse response in this case can be modeled with (in the absence of lognormal shadowing) [17] where and are the numbers of clusters and multipath for the distribution of clusters and signals. Here, the frame period has to be larger than the temporal spread introduced by the channel and the channel is assumed to be stationary according to the block fading assumption. At the end of the frontend, each data pulse is correlated with a template generated by the delay unit and the signal is passed to the backend. A traditional approach to build the backend section following the RF mixer would include the use of filters, a sampleandhold block (S/H), and a separate decision circuit for threshold detection. This would necessitate the inclusion of control voltage generators for providing regulatory signals for integrator and samplehold (S/H) sections. A separate circuit will also be needed to provide a calibrated reference voltage for driving the output decision circuit. If the output of the RF mixer does not have sufficient strength to drive the threshold detection section, further processing will be required before feeding the signal to the backend. The aim of this paper is to propose an architecture which exploits a differential frontend and simplifies the circuit overhead of the receiver by implementing its backend as an integrated comparator block.
Figure 1(b) shows the diagram of the proposed architecture for a transmitreference receiver which employs a modified backend section after the RF mixer. The frontend in this design is realized with a twostage differential lownoise amplifier (LNA) with passive matching circuits. It makes the frontend compatible with differential topologies of the following RF mixer and improves the overall noise sensitivity of the receiver. The differential amplifier can also support direct connection to a doublebalanced correlator and offers benefits like commonmode noise rejection with respect to substrate and power rail. Additionally, it has been reported that the differential architecture can influence the reduction of the second intermodulation product (IP_{2}) [18]. On the other hand, the proposed backend uses a window comparator and reduces the receiver's electronic burden by substituting S/H blocks with logic and buffer sections. The comparator is built with a transconductance amplifier core with integrated bias circuits to deliver the comparator reference voltage and the gate voltage for a tail current source. It also includes a multiplication factor control block to regulate the width of the detection window and a logic section which delivers the output signal of the window comparator. The buffer block at the output of the backend is implemented with a pair of cascaded inverters. The proposed TRreceiver front and backend circuits are discussed in detail in the next section.
3. Matched FrontEnd LowNoise Amplifier
The proposed doublestage highgain lownoise amplifier, which immediately follows the antenna in the receiver frontend (see Figure 1(b)), is presented in Figure 2. The circuit uses two backtoback connected fully differential amplifier stages which allow it to exploit the advantages of the differential topology for a multistage amplifier. It employs a passive reactive Yshaped matching circuit to provide it with input (), interstage, and load port matching. With the inclusion of these matching circuits, the amplifier achieves very low reflection losses (minimum return loss is better than −18 dB at the interfacing ports.) Matching also contributes to the improvement of the amplifier's noise figure following a mechanism described in [19]. The fully symmetric structure is expected to make the architecture more resilient against process variation for the 90 nm technology. Using an LC tuning circuit as the resonance bank provides the amplifier with the ability to fine tune its center frequency. The inclusion of multiple integrated bias circuits leads the LNA to be driven by a single supply voltage (1.2V). It manages a high small signal forward gain with the doublestage core and careful selection of bias currents provides it with simultaneous power and noise matching. The architecture will be compatible with a fully differential TR frontend where the RF mixer can be built with a doublebalanced topology. The amplifier includes features like improved linearity (because of cancelation of intermodulation products), is more resistant to power supply noise, and supports the use of image rejection schemes, dipole antenna, and Gilbert mixers.
To maintain design symmetry for the first block of the design, with (60 μ/0.1 μ) play the role of driving transistors and with (50 μ/0.1 μ) insulate the second stage from the driving input. The central operating point of the frontend built with this cascode configuration is determined by and in the first block, which form a pair of resonance tanks in conjunction with associated capacitances. The resistive elements in the tanks ( and ) do not represent independently inserted components but model for parasitic contributions from adjoining tuning inductors. To include a provision of fine tuning, parallel capacitors are also included in the tanks in the form and . A source degenerating port matching network (built with and ) adjusts the input impedance offered by while works as a coupling capacitor. An LC matching branch is also placed at the input port of . To ensure a smooth transition of signal and to save on power penalty and noise contribution, a secondary passive linking network joins the intermediate nodes between the two stages. It is made with , , and for the lefthalf segment of the amplifier. , , and provide a similar service for the second set of intermediate differential nodes located on the righthalf of the design. The linking circuits ensure that additional input matching networks are not necessary for the second stage of the differential amplifier, leading to a boost in the overall amplifier gain. The two cascode trees in the second stage are made with (50 μ/0.1 μ and 40 μ/0.1 μ) and (similar dimensions). They closely follow the size of the devices in the first block except for process variation. The reactive elements in the secondary resonating banks (, , and ) are carefully matched with their counterparts in the first stage to maintain a stable operating point. Finally, Tshaped passive matching circuits (previously placed with intermediate nodes) are also utilized for the differential load ports ( and ) to interface them with typical microwave loads that may follow. The components in the load matching circuits are designated as , , , and like. A tail transistor () is added with the bottom rail of the first stage to exert control over the overall bias current which is split between the differential half circuits. The tail current controlling arrangement is not repeated for the second block for easier manipulation of the bias current. The gate bias supply for the control device () and the gate signals for the driving transistors ( and ) necessitate inclusion of three bias circuits to push the concerned transistors in a region of saturation. This threeway supply rail is also included in Figure 2, built with diodeconnected devices (2 μ/1 μ), (2 μ/1 μ), (4.7 μ/1 μ), and bootstrapping resistors (). For a 1.2V 90nm process, , , and are set to 0.851, 0.851, and 0.699 V, respectively. An interfacing balun (balancedunbalanced) circuit is placed at the ports during simulation to interface the frontend with the antenna that feeds it. Three ideal coils () in a transformer formation facilitate this balanced to unbalanced conversion with coupling parameters . If we define a quality factor for the input network [20] appended with (excluding coupling capacitors), it will be a function of source impedance () and operating frequency () while relating to the gatesource capacitance () of the transistor : This quality factor has to be optimized to balance gain and bandwidth achieved by the first stage of the amplifier. In the resonance tank formed on top of , the quality factor of the reactive element () will ultimately determine the inductor's series resistance () which can be approximated by Therefore, this undesirable resistive element can be minimized by selecting a reasonable value for the drain inductor. The designer also has to be careful about large quality factors achieved at microwave frequencies, which can push up the resistive value of [21]. For the 90nm CMOS process (with the presence of parasitic junction capacitors) the output resistance offered by the cascode device () has low to moderate values like The effective output resistance () seen from the load port of the first stage amplifier can be estimated using the adjoining resistive components with The quality factor seen by a load connected to the first stage can, therefore, be simplified as Tuning of the controlling parameters of this quality factor would influence the overall gain achieved by the first block of the doubledifferential amplifier. A very small would increase and, as a consequence, message bandwidth and gain of the LNA may suffer [20]. To set the operating point in the band, the bank inductor () resonates with and an output capacitive element which can be calculated with where is the cascode capacitance seen from the drain of , represents extraneous parasitic elements existing between that node and ground, is the capacitance offered by a following stage (load), and stands for the output coupling capacitance. So, all these considerations have to be taken into account during the design of the lownoise amplifier. After careful selection of the reactive elements for gainbandwidth optimization, the input impedance of the amplifier [7] looking from the gate terminal of the transistor can be derived as where is the transconductance of the transistor . The dimensions of , , , and the aspect ratio of are manipulated to eliminate the imaginary part of the expression (as part of input matching) which reduces the equation and matches it with characteristic source or antenna resistance (),
For the output matching circuit located on the left segment of the amplifier [22] (see Figure 2), the impedance can be derived as where , , and represent measures of reactance and is offered by the circuit block which follows the lownoise amplifier in the receiver. Now the expression can be reduced to if the parameters are manipulated to equalize , and . The design equation which makes equal to the firststage output resistance () will be where is the real element of the port impedance . As an example, when and near the center frequency for the proposed design, if we set Ω, then following (13) should result in being equal to and being matched with (see Figure 2). A similar matching mechanism is also executed through a reactive network connected with the node which is made with , , and .
Noise figure contributed by the driving transistors ( and for the first stage) of a differential amplifier has been modeled in literature [23] and is expressed in terms of quality factor of the input network , transit point (), and center frequency (): where is defined by zerobias drainconductance and device transconductance, correlates drain and gate noise functions, models channel noise, and models noise induced at gate terminal. If it is possible to partially cancel out the third and fourth components in (14), the expression for minimum noise figure (applicable for FET amplifiers) can be approximated by If the effect of the cascode device in the first stage () is considered, the cascode noisefactor () will be proportional to device transconductance () [24] where denotes the parasitic capacitance linking the two stages of the amplifier. Finally, an output balun block will be necessary to interface the amplifier if the following mixer or another frontend component is singleended. The three reactive elements for the RF_{out} port balun are denoted by , supported with carefully selected linking parameters (). In this arrangement, the bias supply is set to 1.2 V for process requirements.
4. BackEnd Architecture
The proposed decision section for a TRreceiver backend is driven by a window detector with output logic and buffer gates, thus relaxing the overhead requirements for the receiver design. The symmetric nature of the decision circuit makes it resilient with respect to variation of process. The controlling voltage of the tail current source in the design and the reference voltage of the comparator are provided with integrated diode connected bias circuits. The inclusion of multiple bias arrangements allows the backend to be driven by a single supply rail (1.2V) which eliminates the need for external gate supply voltages. Optimization of the bias currents for the comparator reduces their power requirement with respect to comparable analog detection circuits. The architecture of the receiver backend is simplified by avoiding samplehold blocks and using logic sections and buffer gates. This reduces the electronic burden on the comparator's transconductance amplifier core and pushes down its power demand even further. The window of detection for the decision circuit can be controlled by multiple design parameters, one of which is influenced by a multiplication factor control block with external switching signals.
Analog window comparators have primary applications in testing of controlled systems and they typically capture multiple analog inputs to produce the output in a decoded format. They are called window detectors because they respond when the difference between the two excitations of the decision circuit is within the range of [], with being the detection threshold for the comparator [25]. The proposed decision section, shown as a functional block diagram in Figure 3(a), is built on a differential input core, a number of current mirrors and a tail current source (the window decision circuit in Figure 3(b)), a mirror factor controlling block (Figure 3(c)), two bias supplies (Figure 3(d)), and an output logic section (Figure 3(e)). Transistors and (20 μ/0.1 μ) constitute the input differential pair where represents the input signal (driving excitation) coming from an RF multiplier in the TRreceiver and denotes a calibrated reference voltage for the decision circuit. So, the input network of the window comparator resembles a differential transconductance amplifier core. If the input voltage () falls in the vicinity of the threshold level (i.e., it is within the range ), the logic section implemented by transistors , (20 μ/0.1 μ), and , (40 μ/0.1 μ) produces a positive response in terms of output voltage. As a result, the comparator is able to detect whether the input signal falls within a particular window around the reference voltage or not [26]. If the tail bias current flowing through the design is , when both comparator excitations are sufficiently at the same level ( equals or resides within the range ), transistors , , , , and (20 μ/0.1 μ) remain in a region of saturation. A split current (/2) flows through the transistors which depends on the dimensions of current mirrors. The tail current source is implemented by a device (, 20 μ/0.1 μ) attached to the bottom power rail and powered by a gate bias supply. The two bias circuits in the design (see Figure 3(c)) are developed by diode connected transistors (, 0.1 μ/1 μ and , 5.45 μ/1 μ) which deliver calibrated dc voltages for and for . represent low to moderate values of impedance which ensure that remains at a value near the bias rail () and the reference level () can be fine tuned within the 0.6–1.2 V range using device dimensions (1–10 μm as width of ). Four ptype current mirrors are implemented by transistors , all following a unity mirror factor. The ntype current mirrors introduce a regulating mirror factor () as a way of controlling the width of the detection window. The first type mirror is formed by (20 μ/0.1 μ) and (variable dimensions) which split the bottom current path in five ways so that can be adjusted by the designer. Inclusion of switching signals () for the devices turns it into a programmable current mirror which can be controlled by an external driving circuit. A similar extended mirror circuit is realized by the devices and , also driven by a 5bit input signal. The factor can be compounded (raised above 5) by adjusting the dimensions of the control transistors () as multipliers of the width of . When the input voltage and the reference level are not in close proximity, currents flowing through devices and become and or vice versa with depending on the difference between the two inputs. The expression for the detection window width of the comparator ultimately takes the form of [25]: where is the carrier mobility, is the tail current, denotes the gate capacitance per unit area, is the current mirror factor, and indicates the aspect ratio of or . Because of the resemblance of the input pair of the comparator section with a differential transconductance () amplifier, its gainbandwidth product () will depend on the input pair's effective load capacitance () where is a constant depending on load capacitance and is the overdrive voltage defined by Here and are the input signal and the threshold voltage for the input differential pair, respectively. The factor for the decision circuit is optimized to fine tune the detection window. The bias rail () may be connected with at least two of the switching combinations () to maintain a nominal value for (=2). Here, it is assumed that the relation between the output current and the input voltage of the two input transistors follows the standard equations of saturation region.
(a)
(b)
(c)
(d)
(e)
As indicated by the threshold expression, the detection window can be adjusted by changing the bias current which, in turn, is controlled by the voltage and the transistor . As carrier mobility and gate capacitance are intrinsic device parameters, the other controlling factors of the error threshold are the multiplying factor of the tail current mirrors () and the size of the input differential pair . The gate biasing voltages in the comparator are provided by biasing arrangements already discussed in relation to Figure 3(d) where voltages are collected from the drain terminals of . To evaluate the performance of the backend, characterization of the window detector [26] is performed in detail in the next section and the responses from the decision block for TR inputs are discussed.
5. Results and Discussion
5.1. FrontEnd LNA
The proposed design is analyzed on an RF simulator including layout parasites generated by the 90nm circuit elements to enable accurate prediction of microwave parameters during an RF analysis. In the first phase, the design parameters of the wideband lownoise amplifier are obtained on the 0.09μm CMOS platform to evaluate power efficiency and gain performance of the proposed frontend.
5.1.1. Noise Performance
The projected noise figure () and theoretical noise figure contributed by the frontend amplifier are documented to be around 3.5 and 2.93 dB, respectively, near the center frequency of the message bandwidth, as shown in Figure 4(a). In total, the driving transistors (, in the front stage, , in the second block) are responsible for about 23% of the total accumulated noise and the percentage contribution is slightly lower for the insulating devices.
(a)
(b)
5.1.2. Propensity of Oscillation
The lownoise amplifier is expected to be unconditionally stable within the bandwidth as the Rollette stability factor (), plotted versus frequency in Figure 4(a), is always greater than 23 within the bandwidth, hence satisfying the criteria for nonoscillatory behavior [7]. This factor is defined in terms of parameters: Therefore, the LNA is not prone to oscillating in the presence of unwanted noise or interference.
5.1.3. Power and NoLoad Gains
Available gain (), transducer gain () and average power gain () coincide around the center frequency at 20.5 dB in Figure 4(b). This coincidence can be attributed to the matching networks included in the design. Available gain is an estimation of the power gain provided by the core amplifier without any matching, which is then lowered down to transducer gain after the addition of passive linking networks at intermediate and interfacing ports of the frontend. The noload voltage gain of the amplifier reaches a higher limit with its peak touching 28 dB, as shown in the same figure.
5.1.4. and Stability Factor
Figure 5(a) shows the maximum smallsignal forward gain () with a peak of 20.5 dB at the central operating point, managing a 3.3 GHz threedecibel bandwidth. To verify a second criterion for stability, the stability factor is also calculated for the amplifier and it resides within a range denoted by the inequality 0.3 1. The standard definition of the secondary factor () is given with
(a)
(b)
5.1.5. Input/Output Return Loss
Figure 5(b) verifies that the parameter representing reverse isolation () is always better than −55 dB within the bandwidth. As a measure of return losses incurred at the frontend ports, the same frame plots the input and output reflecting parameters ( & ) showing dips at −18 dB and −25 dB, respectively, and maintaining a separation of 30 dB with the peak of the amplifier's reverse isolation.
5.1.6. Region of Linear Behavior
The input referred 1dB compression point (1 dB) is crossed by the LNA at −15 d of input power which pushes the amplifier into a region of nonlinearity (see Figure 6). The 3rd order intersecting point for intermodulation products () settles near −4 d with respect to input power (theoretically, it should have a difference of 10 dB with 1 dB). At this cross point, the frontend would deliver 10 d power to a matched load connected with the output port.
5.2. Window Detector in the BackEnd
The results from the decision section (window detector with buffer gates, as discussed in Section 4) implementing the backend of the TRreceiver are also obtained with 90nm process parameters on a CMOS platform.
5.2.1. Dynamic Response
In an effort to study the dynamic behavior of the proposed backend, the input at the reference port of the comparator block is tied with a zerophase sinusoid and the signal port () is exited with a pulse of variable phase (0~360 degrees). The corresponding response in Figure 7(a) detects a driving signal positively (high) only when the two phases follow each other. With an increase in phase difference, the circuit response starts to resemble nonuniform pulse peaks.
(a)
(b)
5.2.2. Behavior for Variable Amplitude
Both input ports of the detection subcircuit are tested with a number of combinations of frequency and magnitude for the input sinusoids. With a fixed reference level (settled at 1000 mV), the signal port is tied to a sinusoid whose amplitude is raised in steps from 500 to 2500 mV. Figure 7(b) demonstrates that, with an increase in the difference of magnitude between the two excitations or for signals existing outside the detection window, the magnitude of the amplitude response fails to reach the level which is achieved when the strengths of and are matched.
5.2.3. Setting the Reference Level
As the first step of performing static characterization for the detection unit, the operating point of the window detector () is manipulated by setting the reference level with variable dc voltages (within 0.5~1.2 V). The port is fed with a ramp (linearly increasing) source and corresponding output voltages are plotted for different combinations of the two inputs. The shape of the response in Figure 8(a) suggests a dependence of the detection window on the strength of reference level. For example, when is tied to 0.69 V the detectable range for the input is uniform and approximated by 0.55~.78 V. In this case, the detection window is not perfectly symmetric around the reference level and the asymmetry may increase for wider ranges for the window. As for V, the window covers the domain of 0.76–1.12 V and for V, the window becomes open ended at one side. This would indicate the necessity of a suitable choice for the detection width with regard to maintaining symmetry. The next section would test the influence of the controlling parameters obtained from the expression of the decision threshold.
(a)
(b)
5.2.4. Mirror Factor
Using the control circuit implemented by , the factor can be set from 1 to 5 if these devices have the same sizes as and . Out of the parameters obtained from (17), exerts a greater influence in manipulating the detection window. By doubling the size of as compared to the dimensions of or , can be extended to a value of 10. Within this coverage, the window width can be adjusted between ±80 to ±750 mV around the reference voltage, as shown in Figure 8(b). A high value of can lead to a very wide window, causing spurious detection in the receiver. So, a suitable value for the mirror factor is chosen as 23 to ensure a symmetric window. This can be achieved by tying two or three of the switching signals () in Figure 3(c) to the bias supply rail ().
5.2.5. Final Response from the BackEnd
The backend of a TRreceiver starts with an RF mixer (multiplier) which is followed by the threshold detection block (see Figure 1(b)). The RF mixer response, duely processed, is turned into a driving signal for the window detector. Rather than using separate integrator, samplehold, and threshold detection blocks, we have proposed a buffered comparator core as the basis of the detector circuit. In the most simple form, each frame in a received UWB pulse train consists of a pair of pulses or a doublet. The first pulse in the frame is employed as a “synchronizing pulse” or “reference pulse” and it is followed by a “data pulse” modulated according to the message signal. We have seen in Figure 8(a) that the proposed decision circuit provides a quasisymmetric detection window around a reference voltage of 0.58~0.78 V. Therefore, the mixer response is processed by a wideband amplifier and a shift in its dc base is introduced by a simple resistive adder. The shifted and modified RF mixer response is fed to the decision section as shown in Figure 9(a), and the reference level is carefully selected to vary between 0.65 to 0.705 V in minute steps of 0.01 V to optimize the reference signal. The corresponding response from the comparator becomes bipolar and quasisymmetric with respect to its dc base when approaches 0.67 V. So the iteration of is repeated, this time in even smaller steps, and the final reference limit is set to the vicinity of 0.672 V. The factor is regulated in Figure 9(b) to assess the effect of a variable mirror factor after the reference signal is already optimized. The sharp degradation of the output signal suggests that the controlling parameters in (17) should be employed in a mutually exclusive fashion. And finally, a buffer gate should be placed at the output of the threshold detection circuit to recover a regular decoded data stream.
(a)
(b)
Figure 10(a) presents the response from a buffered window detector for an input bit rate of 0.1 Gbps. The two driving signals for the RF mixer (RF and LO excitations) are plotted in the same frame to visualize the overlapping of the two inputs at instances when the decision circuits produce a positive detection. The stream corresponds to an arbitrary fourbit sequence of 1 0 1 0, with the magnitude of the Gaussian pulses falling within the range of ±300 mV. The amount of delay between the two multiplier inputs is equal to the separation between the “reference pulse” and the “data pulse”. The reference pulses of the second mixer input, after being delayed by a wideband element, become synchronized in time with the data pulses of the driving RF signal and there is no visible dispersion effect on the final responses. The signals at different stages of the proposed backend for a TRreceiver are drawn separately in Figure 10(b), which include the input pulse streams (RF and LO) fed to a radiofrequency mixer, the baseshifted response from the mixer, the bipolar signal from the comparator (window detector), and the final response from an output buffer section. The buffer gate is implemented with a pair of cascaded inverters, producing a decoded stream from the response of the window comparator. The template data rate of the backend circuits is tested up to 2 Gbps and at higher data rates (after 1 Gbps) the window detector performs with an inverse response, as shown in Figures 11(a) and 11(b). A certain amount of timeshift is also introduced by the logic gates existing in the design. The performance metrics of the TR front and backend components and their comparative standing with respect to published results are summarized in the next section.
(a)
(b)
(a)
(b)
6. Comparison of Performance
6.1. FrontEnd
Table 1 documents the performance of the proposed frontend doublestage 90 nm amplifier and compares it with simulated results from reported millimeterwave frontend circuits [20, 27–38]. Simulated data have been collected from the references or concerned designs have been analyzed in the authors' environment to make a fair comparison. To evaluate designs built on different platforms, a composite figureofmerit parameter () is defined with the following equation: As indicated by a high (), the proposed design of the frontend delivers relatively high forwardgain, lower noise contribution, and a high peak gain frequency in the band, facilitating its interfacing with the following mixer in the TRreceiver.
 
^{
/}Including probe pads *For entire frontend. 
6.2. BackEnd
The performance of the buffered window comparator in the backend is tabulated in Table 2, illustrating its relative merits with respect to data rate, power dissipation, speed/power ratio, and supply voltage. To quantify the relative simulated performance of the proposed decision block with respect to published detection circuits [39–49], a second figureofmerit () is defined with The table shows the contrast in the figuresofmerit achieved by pipelined analogtodigital converters and a window comparator. These findings verify the merits of the proposed TRreceiver backend in terms of speed and powerefficiency.

7. Conclusions
The subject of this paper is the implementation of front and backend blocks for a transmitreference (TR) receiver using noise and power efficient architectures. A 90nm wideband amplifier is proposed for the frontend which achieves a forward gain 20.5 dB at the centerpoint with a 3.3 GHz bandwidth. The peak port matching parameters stand at −18 dB and −24.8 dB whereas the amplifier behaves linearly up to a power limit of −4 d at the driving port. To support the twostage differential structure, the frontend consumes 27.2 mW while the noise performance (NF) approaches a minimum level of 3.5 dB near the center frequency (in band). The backend of the TRreceiver is completed with a buffered detection circuit with a regulated decision window. Static and dynamic behavior for the detector block are tested with a variable transmitreference pulse stream (0.1~2 Gbps). For each unit of consumed power, the achieved speed of the window detector covers a range of 14 to 265.4 GHz. The proposed architectures fare better when compared with published designs of their counterparts and should help the onchip realization of a TRreceiver.
Acknowledgments
The author would like to thank S. Rashid, Dr. H. Rashid, and the authority of Bangladesh University of Engineering and Technology.
References
 B. Park, K. Lee, and S. Choi, “A receiver frontend design in 0.13 μm CMOS for multiband OFDM UWB system,” in Proceedings of the Asia Pacific Microwave Conference (APMC '09), pp. 241–244, December 2009. View at: Publisher Site  Google Scholar
 C. W. Kim, M. S. Kang, P. T. Anh, H. T. Kim, and S. G. Lee, “An ultrawideband CMOS low noise amplifier for 35GHz UWB system,” IEEE Journal of SolidState Circuits, vol. 40, no. 2, pp. 544–547, 2005. View at: Publisher Site  Google Scholar
 A. Ismail and A. A. Abidi, “A 3.1 To 8.2GHz zeroIF receiver and direct frequency synthesizer in 0.18μm SiGe BiCMOS for mode2 MBOFDM UWB communication,” IEEE Journal of SolidState Circuits, vol. 40, no. 12, pp. 2573–2582, 2005. View at: Publisher Site  Google Scholar
 M. C. Ha, Y. J. Park, and Y. S. Eo, “A 3–5 GHz noncoherent IRUWB receiver,” Journal of Semiconductor Technology and Science, vol. 8, no. 4, pp. 277–282, 2008. View at: Google Scholar
 Q. H. Dang, A. Trindade, A. J. Van Der Veen, and G. Leus, “Signal model and receiver algorithms for a transmitreference ultrawideband communication system,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 4, pp. 773–779, 2006. View at: Publisher Site  Google Scholar
 M. R. Casu and G. Durisi, “Implementation aspects of a transmittedreference UWB receiver,” Wireless Communications and Mobile Computing, vol. 5, no. 5, pp. 537–549, 2005. View at: Publisher Site  Google Scholar
 B. Leung, VLSI for Wireless Communication, Prentice Hall, India, New Delhi, 1st edition, 2002.
 A. Roy, S. M. S. Rashid, M. A. Arafat, and A. B. M. H. Rashid, “Design of a wideband delay element for transmitted reference UWB receivers,” in Proceedings of the 6th International Conference on Electrical and Computer Engineering (ICECE '10), pp. 97–100, December 2010. View at: Publisher Site  Google Scholar
 Z.Y. Huang, “A Kaband CMOS lownoise amplifier for Kaband communication system,” in Proceedings of the World Congress on Engineering and Computer Science, pp. 1–4, October 2010. View at: Google Scholar
 Y.C. Chen, C. H. Wang, and Y.S. Lin, “Lowpower 24 GHz CMOS receiver frontend using isolation enhancement technique for automatic radar systems,” Microwave and Optical Technology Letters, vol. 54, no. 6, pp. 1471–1476, 2012. View at: Google Scholar
 T. P. Wang, “A lowvoltage lowpower Kband CMOS LNA using DCcurrentpath split technology,” IEEE Microwave and Wireless Components Letters, vol. 20, no. 9, pp. 519–521, 2010. View at: Publisher Site  Google Scholar
 F. S. Mitra, “On optimal data detection for UWB transmitted reference system,” in Proceedings of the Global Telecommunications Conference, pp. 764–768, San Francisco, Calif, USA, December 2003. View at: Google Scholar
 R. T. Hoctor and H. W. Tomlinson, “Delayhopped transmitted reference RF communications,” in Proceedings of the IEEE Conference on Ultra Wideband Systems and Technologies, pp. 265–270, Baltimore, Md, USA, 2002. View at: Google Scholar
 A. Schranzhofer, Y. Wang, and A. J. Van Der Veen, “Acquisition for a transmitted reference UWB receiver,” in Proceedings of the IEEE International Conference on UltraWideband (ICUWB '08), pp. 149–152, September 2008. View at: Publisher Site  Google Scholar
 Q. H. Dang and A. J. van der Veen, “A decorrelating multiuser receiver for transmitreference UWB systems,” IEEE Journal on Selected Topics in Signal Processing, vol. 1, no. 3, pp. 431–442, 2007. View at: Publisher Site  Google Scholar
 Q. H. Dang, A. Trindade, A. J. Van Der Veen, and G. Leus, “Signal model and receiver algorithms for a transmitreference ultrawideband communication system,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 4, pp. 773–779, 2006. View at: Publisher Site  Google Scholar
 J. Foerster, “Report on channel modeling,” IEEE P802.15 standard for Wireless Personal Area Networks, 2005. View at: Google Scholar
 H. K. Chiou, H. Y. Liao, and K. C. Liang, “Compact and low power consumption Kband differential lownoise amplifier design using transformer feedback technique,” IET Microwaves, Antennas and Propagation, vol. 2, no. 8, pp. 871–879, 2008. View at: Publisher Site  Google Scholar
 M. ElNozahi, A. A. Helmy, E. SánchezSinencio, and K. Entesari, “An inductorless noisecancelling broadband low noise amplifier with composite transistor pair in 90 nm CMOS technology,” IEEE Journal of SolidState Circuits, vol. 46, no. 5, pp. 1111–1122, 2011. View at: Publisher Site  Google Scholar
 X. Guo and K. K. O, “A power efficient differential 20GHz low noise amplifier with 5.3GHz 3dB bandwidth,” IEEE Microwave and Wireless Components Letters, vol. 15, no. 9, pp. 603–605, 2005. View at: Publisher Site  Google Scholar
 K. W. Yu, Y. Lu, D. Huang, D. C. Chang, V. Liang, and M. F. Chang, “A 26 GHz lownoise amplifier in $0.18\mu m$ CMOS technology,” in Proceedings of the IEEE International Symposium on Electron Devices Microwave Optoelectronic Applications, pp. 93–98, 2003. View at: Google Scholar
 S. M. Shahriar Rashid, A. Roy, S. N. Ali, and A. B. M. H. Rashid, “Design of A 21 GHz UWB differential low noise amplifier using .13μm CMOS process,” in Proceedings of the 12th International Symposium on Integrated Circuits (ISIC '09), pp. 538–541, December 2009. View at: Google Scholar
 D. K. Shaeffer and T. H. Lee, “A 1.5V, 1.5GHz CMOS low noise amplifier,” IEEE Journal of SolidState Circuits, vol. 32, no. 5, pp. 745–759, 1997. View at: Google Scholar
 H. Samavati, H. R. Rategh, and H. Lee Thomas, “A 5GHz CMOS wireless LAN receiver front end,” IEEE Journal of SolidState Circuits, vol. 35, no. 5, pp. 765–772, 2000. View at: Google Scholar
 A. Laknaur, R. Xiao, and H. Wang, “A programmable window comparator for analog online testing,” in Proceedings of the 25th IEEE VLSI Test Symposium (VTS '07), pp. 119–124, May 2007. View at: Publisher Site  Google Scholar
 A. Roy, “A window detection technique with adjustable threshold for transmitted reference receivers,” submitted to Central European Journal of Engineering. View at: Google Scholar
 B. A. Floyd, L. Shi, Y. Taur, I. Lagnado, and K. K. O, “A 15GHz wireless interconnect implemented in a 0.18μm CMOS technology using integrated transmitters, receivers, and antennas,” in Proceedings of the VLSI Circuits Symposium, pp. 155–158, June 2001. View at: Google Scholar
 Y. H. Yu, Y. J. E. Chen, and D. Heo, “A 0.6V low power UWB CMOS LNA,” IEEE Microwave and Wireless Components Letters, vol. 17, no. 3, pp. 229–231, 2007. View at: Publisher Site  Google Scholar
 F. Zhang and P. R. Kinget, “Lowpower programmable gain CMOS distributed LNA,” IEEE Journal of SolidState Circuits, vol. 41, no. 6, pp. 1333–1343, 2006. View at: Publisher Site  Google Scholar
 R. L. Wang, M. C. Lin, C. F. Yang, and C. C. Lin, “A IV 3.110.6 GHz fullband cascoded UWB LNA with resistive feedback,” in Proceedings of the IEEE Conference on Electron Devices and SolidState Circuits (EDSSC '07), pp. 1021–1023, December 2007. View at: Publisher Site  Google Scholar
 H. Zhang, X. Fan, and E. S. Sinencio, “A lowpower, linearized, ultrawideband LNA design technique,” IEEE Journal of SolidState Circuits, vol. 44, no. 2, pp. 320–330, 2009. View at: Publisher Site  Google Scholar
 S. C. Blaakmeer, E. A. M. Klumperink, D. M. W. Leenaerts, and B. Nauta, “Wideband balunLNA with simultaneous output balancing, noisecanceling and distortioncanceling,” IEEE Journal of SolidState Circuits, vol. 43, no. 6, pp. 1341–1350, 2008. View at: Publisher Site  Google Scholar
 D. Im, I. Nam, H. T. Kim, and K. Lee, “A wideband CMOS low noise amplifier employing noise and IM2 distortion cancellation for a digital TV tuner,” IEEE Journal of SolidState Circuits, vol. 44, no. 3, pp. 686–698, 2009. View at: Publisher Site  Google Scholar
 W. H. Chen, G. Liu, B. Zdravko, and A. M. Niknejad, “A highly linear broadband CMOS LNA employing noise and distortion cancellation,” IEEE Journal of SolidState Circuits, vol. 43, no. 5, pp. 1164–1175, 2008. View at: Publisher Site  Google Scholar
 G. Zare Fatin, Z. D. Koozehkanani, and H. Sjöland, “A 90 nm CMOS + 11 dbm IIP3 4 mW dualband LNA for cellular handsets,” IEEE Microwave and Wireless Components Letters, vol. 20, no. 9, pp. 513–515, 2010. View at: Publisher Site  Google Scholar
 H.C. Lee, C. Shiun Wang, and C. Kuang Wang, “A 0.22.6 GHz wideband noisereduction Gmboosted LNA,” IEEE Microwave and Wireless Components Letters, vol. 22, pp. 269–271, 2012. View at: Google Scholar
 X. Wang, J. Sturm, and N. Yan, “0.63GHz wideband receiver RF frontend with a feedforward noise and distortion cancellation resistivefeedback LNA,” IEEE Tranactions on Microwave Theory and Techniques, vol. 60, pp. 387–392, 2012. View at: Google Scholar
 K. H. Chen and S. I. Liu, “Inductorless wideband CMOS lownoise amplifiers using noisecanceling technique,” IEEE Transactions on Circuits and Systems I, vol. 59, pp. 305–314, 2012. View at: Google Scholar
 A. Ismail and M. Elmasry, “A 6bit 1.6GS/s lowpower wideband flash ADC converter in 0.13μm CMOS technology,” IEEE Journal of SolidState Circuits, vol. 43, no. 9, pp. 1982–1990, 2008. View at: Publisher Site  Google Scholar
 D. Choi, D. Kim, K. Cho, D. Kim, and M. Song, “A low noise 65nm 1.2V 7bit 1GSPS CMOS folding A/D converter with a digital selfcalibration technique,” in Proceedings of the International SoC Design Conference (ISOCC '10), pp. 194–197, November 2010. View at: Publisher Site  Google Scholar
 J. Lee, B. C. Michael, H. J. Park, and B. H. Park, “A 7b 1GS/s 60mW folding ADC in 65nm CMOS,” in Proceedings of the International SoC Design Conference (ISOCC '10), pp. 338–341, November 2010. View at: Publisher Site  Google Scholar
 I.N. Ku, Z. Xu, Y.C. Kuan, Y.H. Wang, and M.C. Frank Chang, “A 40mW 7bit 2.2GS/s timeinterleaved subranging ADC for lowpower gigabit wireless communications in 65nm CMOS,” in Proceedings of the IEEE Custom Integrated Circuits Conference, pp. 1–4, September 2011. View at: Google Scholar
 C. C. Hsu, C. C. Huang, Y. H. Lin et al., “A 7b 1.1GS/s reconfigurable timeinterleaved ADC in 90nm CMOS,” in Proceedings of the Symposium on VLSI Circuits (VLSIC '07), pp. 66–67, June 2007. View at: Publisher Site  Google Scholar
 E. Alpman, H. Lakdawala, L. R. Carley, and K. Soumyanath, “A 1.1V 50mW 2.5GS/s 7b timeinterleaved C2C SAR ADC in 45nm LP digital CMOS,” in Proceedings of the IEEE International SolidState Circuits Conference (ISSCC '09), pp. 76–78, February 2009. View at: Publisher Site  Google Scholar
 J. Proesel, G. Keskin, J. O. Plouchart, and L. Pileggi, “An 8bit 1.5GS/s flash ADC using postmanufacturing statistical selection,” in Proceedings of the 32nd Annual Custom Integrated Circuits Conference (CICC '10), pp. 1–4, September 2010. View at: Publisher Site  Google Scholar
 J. Yang, T. L. Naing, and B. Brodersen, “A 1GS/s 6bit 6.7mW ADC in 65nm CMOS,” in Proceedings of the IEEE Custom Integrated Circuits Conference (CICC '09), pp. 287–290, September 2009. View at: Publisher Site  Google Scholar
 R. C. Taft, P. A. Francese, M. R. Tursi et al., “A 1.8 V 1.0 GS/s 10b selfcalibrating unifiedfoldinginterpolating ADC With 9.1 ENOB at nyquist frequency,” IEEE Journal of SolidState Circuits, vol. 44, no. 12, pp. 3294–3304, 2009. View at: Publisher Site  Google Scholar
 Y. Z. Lin, C. W. Lin, and S. J. Chang, “A 5bit 3.2GS/s flash ADC with a digital offset calibration scheme,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 18, no. 3, pp. 509–513, 2010. View at: Publisher Site  Google Scholar
 S. Park, Y. Palaskas, A. Ravi, R. E. Bishop, and M. P. Flynn, “A 3.5 GS/s 5b flash ADC in 90 nm CMOS,” in Proceedings of the IEEE 2006 Custom Integrated Circuits Conference (CICC '06), pp. 489–492, September 2006. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2012 Apratim Roy. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.