The Design of a Mobile Radio Receiver
`using a Direct Conversion Architecture
`
`Polly Estabrookt
`Jet Propulsion Laboratory
`California Institute of Technology
`4800 Oak Grove Drive
`Pasadena, CA 91109
`(313) 354-7066
`
`Abstract
`
`This paper outlines the advantages of building a mobile radio
`receiver using a direct. downwnversion architecture rather than
`the conventional heterodyne approach. Design problems partic~
`ular to mobile radio direct conversion receivers, such as achieving
`good sensitivity and dynamic range, are discussed and solutions
`proposed. Namely the integrated approach taken for the design
`of the RF mixers, the subsequent low frequency amplifier, and
`active filter is detailed. This aliows low conversion loss and high
`output impedance mixers to be designed, amplifiers with low
`noise performance own the dc to £32”: bandwidth to be con-
`structed, and, filters that set selectivity without impacting the
`signal-to-noise ratio at their output to be specified. Design ad-
`vantages that are a direct consequence of this choice of receiver
`architecture, such as high selectivity, small size, and low power
`consumption, are detailed. Finally we describe the receiver im—
`plementation and characterize its periormanw.
`
`Introduction
`
`The direct conversion receiver considered here is depicted in
`Figure 1. To illustrate its performance, consider the input signal,
`fflp, to have a center frequency fc and a bandwidth SW”.
`The input signal, after optional amplification, is power-Split and
`sent to the inputs of two broadband mixers. The local oscillator
`frequency, I“), of both the in-phase and quadrature—phase {I}r
`mixers is locked to In. Consequently, the signal at the output
`of each mixer, fa spans from dc to Bwflpfl. These signals are
`amplified, filtered, and finally processed together to reconstruct
`the transmitted signal.
`Luann
`mm WIT
`
`
`
`Figure l: The direct conversion receiver.
`
`1Tbis work was performed as part of Ms. Estabrook’s doc-
`toral research at Stanford University.
`
`Bruce B. Lusignan
`Stanford University
`Communication Satellite Planning Center
`ERL 202
`Palo Alto, CA 94305
`
`The receiver lends itself to work with many signal frequencies
`and formats. The low level signal processing, comprised of the
`RF mixers and the first stage of amplification and filtering, is ac-
`complished by low noise, high dynamic range, circuits. Although
`the input mixer will need to be designed for operation with a
`specific 3‘; and BWRF for best performance, the subsequent low
`level amplifier-filter stage may be designed in several different
`bandwidths (Bil/rm,r a BWRHQ) to serve a wide number of
`applications. Thus by choosing the correct mixer and amplifier-
`filter stage for a. specific application, a large range of signals with
`center frequency’s between 50 and 900 MHz, thus covering all
`mobile radio bands, and signal bandwidth‘s from a few KHz,
`for 553 voice applications, to a few hundred KHz, for digital
`applications can be accomodated. The second stage of amplifi-
`cation and filtering is more application—specific; however, stan-
`dard CMOS opamps and switched capacitor filters with variable
`cutofl‘ frequencies can be used as the signal power has now been
`boosted above the noise floor. Finally, signal reconstruction -
`unfolding and correct frequency placing of the baseband signal —
`and demodulation must take place; these circuits are also appli-
`cation dependent. These functions can be implemented by one
`circuit and realized using standard processing techniques.
`This paper discusses the application of direct conversion to
`the design of a mobile cellular radio at 900 MHz. This work
`has been inspired by Weaver‘s design of a SSH direct conver-
`sion receiver in 1956 [l] and by the more recent use of direct
`conversion receivers for VHF radio paging applications using
`BFSK modulation [2.3]. Translation of this work to higher sig-
`nal frequencies and to more complex signal sets has been slowed
`pending clarification of certain issues, the most important being
`the determination of achievable gain and phase balance between
`the ’I‘ and ‘Q‘ channel and the impact of this imbalance on sig-
`nal reception for several modulation types, the development of a
`receiver front-end that can provide the sensitivity and dynamic
`range required, and finally the design of modulation specific de-
`modulation circuits.
`
`This paper addresses the first three issues as they are com-
`mon to all applications. The design of the CMOS active filter,
`the modulation specific circuits, the AFC circuits and the LO
`generation circuits are outside the scope of this work. First the
`motivation to reconsider this receiver is outlined. Second issues
`concerning phase and gain balance between channels are exam-
`ined. Third the functions performed by the front~end circuits
`are detailed and their design given. Finally the overall receiver
`specifications are calculated and compared with those of a typ-
`ical heterodyne receiver. For the purpose of this design, signals
`are taken to be voice signals modulated with amplitude com-
`pandered single sideband {ACSB}. The spectrum is assumed to
`have been divided up in 3.1 KHz channels, with 900 Hz guard
`bands between them.
`
`63
`
`CH2379-1f89f000030063 $1.00 9 1989 IEEE
`
`TCL EXHIBIT 1025
`TCL EXHIBIT 1025
`Page 1 of 10
`Page 1 of 10
`
`

`

`Advantgges of the Direct Conversion Receiver
`
`Evaluation of Mafl'tude of Phase and Gain Imbalance Problem
`
`The direct conversion architecture has several advantages
`over a standard heterodyne receiver:
`(1) the image frequency
`lf-‘mm = fw 1F fa Where far- = fro 3: f»). potentially a
`source of interference in heterodyne systems, is here a frequcy
`within the RF bandwidth — thus rejection of unwanted signals
`at jaw. is infinite; (2) it possesses good selectivity because of
`the use of low pass filters (LPF) rather than band pass filters
`(BPF) as used in a heterodyne receiVer, these can be designed
`with active op-amps and CMOS filters; (3) it has size and power
`consumption advantagu due to the lack of IF filters; (4) the
`majority of the signal processing is done at very low frequencies
`where many high performance, special purpose, circuits already
`exist — thus signal processing is simple, low cost and of high
`performance; (5) lastly the receiver‘s performance is as good as
`or better than a comparable heterodyne receiver. Ih'om a tech-
`nology development point-of-view, this receiver is ideal because
`it merges analog and digital signal processing, because it lends
`itself to monolithic realization — thereby permitting the con-
`struction of generic receiver building blocks, and, because its
`architecture, effectively an I/Q demodulator at RF, can work
`with many modulation schemes. The multipurpose nature of
`the direct conversion receiver’s components can be used to jus»
`tify the development of costly receiver LSI and VLSI building
`blocks.
`The ability of this receiver to work with many modulation
`schemes merits further discussion. Historically the direct con—
`version receiver was proposed by Weaver in 1956 as a method to
`generate and receive 353 signals
`To demodulate SSE two
`haseband mixers and a summer are necessary. Accurate recon-
`struction of the baseband signal, or suppression of the unwanted
`sidebaud within the wanted signal band, is determined by the
`gain and phase balance that can be achieved between the ’I’
`and 'Q’ channels. Weaver was able to keep the unwanted signal
`20 dB to 40 dB below the wanted signal with varying levels of
`dificulty. The direct conversion receiver’s lack of widespread
`acceptance at that time can probably be attributed to the un-
`familiarity of its design, the need for higher received signal-try
`noise (SfN) levels and the shift from SSH to PM as a preferred
`modulation.
`
`FM signals can be received with the direct conversion re—
`ceiver using the "sin—cookie“I demodulation technique which con—
`sists of difl'erentiator and mixer circuits. Here both gain and
`phase match between the channels is needed for correct demod-
`ulation. An FM receiver built with "sin-cosiue” demodulation
`circuits was develoPed in 1979 by Vance et al. for radio recep-
`tion at VHF; the demodulator performance is detailed in [4}.
`To receive'AM modulation the output of the ‘1’ channel can be
`tapped, the 'Q’ channel output is used to generate the control
`voltage for the L0. Fbr this application phase match between
`‘I’ and ‘Q’ channels is necessary to lock fly, to f.,. Finally the
`receiver works well with all types of digital modulation, such as
`QPSK and QAM, as it is efi'ectively an IIQ demodulator placed
`at the RF frequency rather than at some intermediate frequency
`(IF) as is typical in heterodyue receivers. The direct conversion
`receiver then performs the dual function of conversion and de-
`modulation in one process as opposed to heterodyne receivers
`which convert the RF to IF and then demodulate the digital
`signal with an IIQ demodulator. For these applications, phase
`match between channels is of primary importance to attain good
`QPSK reception as is amplitude match for QAM.
`
`Gain and phase match between the ‘I' and 'Q’ channels af-
`fects the quality of signal reception for most modulation formats.
`One way to gauge the gain and phase match is to measure the
`suppression of the unwanted sideband in a SSE receiver. A
`schematic of s. 383 direct conversion receiver is shown in Fig-
`ure 2. Its operation can be detailed as follows. The RF signal,
`of center frequency ft and bandwidth BWRF, is downconverted
`to DC by the input mixers of both channels. The signal at the
`output of these mixers now spans from DC to BWRs-fl. The
`downconverted signal spectrum is folded over itself and no longer
`pom the original RF spectral shape. The low frequency sig-
`nals are then filtered, to remove any adjacent channel signals,
`and amplified to set the noise floor. To reconstruct the original
`spectrum both ‘I’ and 'Q’ channel signals are mixed again with
`quadrature LO’s at a frequency, fut“ equal to Elfin-{2. At
`the output of web of the baseband mixers the sum and differ-
`ce signals created by beating the input spectrum, spanning
`DC to Biting-[2, by BWRFIZ At the mixer output the sum sig-
`nal from DC to 8W” and the difference signal from Blimp/2
`to DC therefore are present. For far 3 fm the sum signal
`at the output of the ‘1‘ channel will be in phase with the sum
`signal at the output of the 'Q’ channel and the difierence sig-
`nals will be 180° out-cf-phase. For f3;- < he the reverse will
`occur. Thus, for Ime- ) fw, when the outputs of the two
`baseband mixers are summed together only the sum signals will
`added and the difference signals subtract. The extent to which
`the unwanted sideband is suppressed is an indication of the level
`of gain and phase matching. For analog voice signals, the level
`of the unwanted sideband relative to the desired sideband sets
`the signal-to—distortion {S} D) ratio at the output of the receiver.
`The extent of gain and phase match required to produce a given
`level of suppression can easily be calculated. For example 20 dB
`of suppression can be achieved by a gain match of 1.5 dB and a.
`phase match of 5° or with a gain match of 0.5 dB and a phase
`match of 11°.
`
`
`
`Figure 2: Schematic of the 853 receiver.
`
`When the receiver is built with discrete parts, suppression
`levels similar to what was achieved by Weaver are expected.
`Hmver when the direct conversion receiver is implemented
`with integrated circuits, so that circuits required by both chan-
`nels can be built on one integrated circuit, then improved gain
`and phase matching between channels can be realised. Regard—
`less of this potential improvement, it is important to remember
`that the S," D or SIN achievable at the output of the direct con-
`version receiver is not necessarily the SIN of the received signal.
`Higher SfN’s are obtainable through the use of compandering
`for 8313 signals {A083} [5] and through the use of codes for
`digital signals.
`
`TCL EXHIBIT 1025
`
`TCL EXHIBIT 1025
`Page 2 of 10
`Page 2 of 10
`
`

`

`To determine the level of gain and phase match possible for
`a direct conversion receiver, the SSB receiver shown in Figure 3
`was built. Figure 3(a) shows the RF portion of the receiver;
`Figure 3(b) depicts the low-level processing circuits and the SSB
`demodulation circuits.
`
`For the purposes of this experiment the 5513 signal was as-
`sumed to have a 6 KHz bandwidth, its suppressed carrier was
`
`
`
`-- +..,...,a. -.._,..fl.._.
`
`.
`
`.
`
`{b} low-signal level amplifiers and SSB demodulation board
`
`Figure 3: The direct conversion receiver implemented with dis—
`crete elements.
`
`taken to be at 70 MHz. The RF signal could then be character-
`ized by:
`
`ft
`away
`
`70.003MH2
`GKHZ
`
`To illustrate the generation of unwanted sidehands and to mea-
`sure suppression in the direct conversion receiver, the reception
`of a SSB tone at
`
`Innis” = 70.005MH2.
`
`is examined. The outputs of the two UHF mixers, fa, and IN
`for the 'I’ and 'Q' channels, respectively. will then be:
`
`fo, 2 2KHz (0°
`
`foo
`
`= 2KHz £90“.
`
`The baseband mixers operate with
`
`B W
`2RF
`11,0? 2
`= 3KHZ
`
`so that the sum and difference signals produced are at 5 KHz
`and l KHz. The 5 KHz signals are in-phase with each other
`and the 1 KHz signals are a 180° out—of—phase with each other.
`When the spectrum at the output of the summer was measured,
`the l KHz signal was typically found to be 24 dB below the 5
`KHz signal. This can be calculated to translate to a gain balanoe
`of less than 1 dB and a. phase balance of better than 2.5”. With
`the low frequency amplifier and summer aligned for best channel
`match, a suppression of 40 dB was possible.
`
`
`Design of the Low—Level Signal Processing
`
`The low level signal processing circuits shown in Figure l are
`depicted in greater detail in Figure 4. Only one channel is shown.
`The RF mixer, the low frequency amplifier, and, filter comprise
`these low level circuits and form the front—end of the receiver.
`A typical RF spectra is shown at the input to the mixer,
`it
`consists of the wanted signal, W,, and two interferers, I, and [3,
`adjacent to W, and having higher power than W, The noise
`floor is illustrated by the darkened line. The spectrum at the
`output of each component is shown to elucidate their function
`and their design constraints.
`The front-end components are processing the input signal
`while it. is still at a low power level. These circuits contribute in
`setting the noise floor and dynamic range of the receiver contrary
`to the subsequent stages whose additional noise only contributes
`inconsequentially to the overall noise. To guarantee the greatest
`dynamic range possible, the low frequency amplifier possesses
`the minimum gain necessary to reduce filter noise components
`to an acceptable level.
`Receiver noise temperature is given at the receiver input as
`a function of frequency as:
`
`Trec(f)
`
`Tmirer(fl + CL I TMflP(f)
`Tjitlzr(f)
`Gishamp
`
`+ C -
`L
`
`(1}
`
`where Tmnn Twp. and, TN“, are the noise temperature of
`the mixer, amplifier. and, filter respectively, Gm.” is the low
`
`
`
`Figure 4: Front-end of one channel of the direct conversion re—
`oeiver with typical spectrum plots.
`
`65
`
`TCL EXHIBIT 1025
`TCL EXHIBIT 1025
`Page 3 of 10
`Page 3 of 10
`
`

`

`frequency amplifier gain (this amplifier may be made of several
`stages of gain G....,), and C; is the mixer‘s Conversion loss. All
`noise sources are a function of frequency due to the dominance
`excess noise term.
`
`Noise figure is determined from noise temperature according to:
`
`NF = lO-Iog(1+ Tm) -
`(2)
`When conjugate matching is used between components the gain
`of the front-end circuits can be written in terms of their indi-
`vidual gains as:
`
`_ Gtotnrmp ‘ Glitter
`GTE: _
`
`'
`
`(3}
`
`1. Mixer Design. Three design goals must be kept in mind
`while designing the direct conversion mixer: {1) eflicient trans-
`lation from the RF to 1'0; {2) low excess noise or lylf noise at
`the mixer output; (3) impedance transformation from R5” to
`Hg. The first objective listed is common to all mixer denign‘
`era especially for receivers with mixer front-ends. Translation of
`the input signal from the RF frequency at the mixer input port
`to some output frequency is specified by the mixer‘s conversion
`loss. Conversion loss is defined as:
`
`C __ Power Available from the Source
`" “
`Power Delivered to theLoad
`
`'
`
`Minimization of mier noise figure is the second criterion for
`a direct conversion mixer as the mixer generated noise, along
`with that generated by the amplifier and filter cascade, sets the
`noise floor of the receiver. The noise at the mixer output is due
`to shot noise, thermal noise and excess noise. Excess noise, or
`lff noise, is of particular importance in the direct conversion
`receiver: it is the dominant noise source at the low frequencies
`at which much of the receiver’s signal processing is carried out.
`Excess noise is generated by all devices, active and passive alike,
`such as diodes, transistors, resistors etc. This noise source can
`be modelled as a noise current of mean squared value:
`
`:1— = K - I; - Af
`
`where K, a, and, b are device dependent parameters, a as 0.5
`- 2.0 and b as 1.0. The noise spectrum of a typical device with
`shot and excess noise is shown in Figure 5. The ljf frequency
`knee, or frequency where the excess noise contribution is equal to
`that caused by shot and thermal noise, is marked on the graph.
`Because this noise occurs in the band to which the RF signal has
`been translated, minimization of its magnitude is particularly
`important for the direct conversion receiver. The signal and
`noise spectrum at the mixer output is depicted in Figure 4.
`
`
`
`L00“)
`
`u' lull ‘
`Figure 5: Noise spectrum, due to shot and excess noise, of a
`typical device.
`
`Lastly direct conversion mixers have the unusual require—
`ment that they must provide an impedance transformation from
`the low input impedance of the antenna [50 [l or 75 Q) to the
`high source impedance desired by the subsequent low frequency
`amplifier. As will be discussed, typical bipolar opp—amps pos—
`sess lowest noise temperature when they are driven by source
`impedances on the order of 10 K0.
`This problem of lff noise generated by all fronteend com-
`ponents and the problem of DC ofisets signals accumulating
`throughout the receive chain and becoming a source of receiver
`instability lead to the idea of using a DC notch filter and AC
`coupled op-amps in the ‘l‘ and ‘Q‘ channels. This DC notch fil-
`ter at baseband elfectively removes the signal power previously
`centered about 1;. Phase lock loop circuitry will need to be im»
`plernented upstream from the notch filter. Certain modulation
`types may be unafi‘ected by the filter’s presence, others may re-
`quire special generation circuits for use with this receiver. For
`example, techniques developed for fading channels such as trans-
`mit tone in band can be applied for generation of both analog
`and digital signals [6]. Determing the bandwidth required by
`this notch filter is one of the goals of this work.
`2. Low
`uen Am lifiers. Standard bipolar amps pos-
`sess the lowest noise temperature when operated with high source
`impedancm, on the order of 10 [(9. Their llf frequency knee
`occurs between 100 Hz and I KHz. As it. is difficult to design
`low 0;, RF mixers that can translate to such high 35, the ma-
`jor impetus of this work was to identify noise mechanisms and
`circuit requirements so that low noise amplifiers wuld be de-
`signed to work with source impedances of 3 K or less, ie. those
`typical of RF mixers outputs, or to modify op-amps to perform
`in this range. The noise performance of these amplifiers must
`be identified as a function of frequcy so that the DC notch
`filter can be constructed with a bandwidth sufficiently narrow
`to
`its impact on the signal spectrum yet large enough
`so that the remaining noise in the signal band was set at an
`acceptable level.
`Second the gain of the amplifier stage must be calculated to
`insure high receiver dynamic range; it cannot be too high lest the
`power in the adjacent channels, denoted in Fig. 4 by 11 and lg,
`saturate the filter stage. Third the lff frequency 1mm of these
`amplifiers must be identified so that the bandwidth required of
`the DC notch filters can be determined.
`
`3. Active Filter Design. These circuits serve to perform the
`first stage of filtering. Their bandwidth is that of the wanted
`signal (BWygm,
`Because they are cascaded by a sub-
`sequent filtering stage aa shown in Fig. 1 they do not set the
`receiver selectivity alone but act primarily to reduce the level of
`the out-of-band interfere-rs so that these signals do not saturate
`the next filter stage. Reduction of adjacent channel interferers
`900 Hz away from the edge of band of the wanted signal by 30
`dB and of the subsequent channel interferer by 50 dB is desired.
`To reduce filter noise temperature it is implemented with bipolar
`op-amps rather than with CMOS circuits. Receives selectivity
`is produced by both low level and high level filters.
`
`Receiver Component Design
`
`I
`I I.
`.
`a
`I L.
`A new time domain based technique for computer-aided mixer
`analysis and design has been developed in part to address prob—
`lems specific to the design of a direct conversion mixer and in
`part to provide mixer designers with a difierent view on mixer
`
`so
`
`TCL EXHIBIT 1025
`TCL EXHIBIT 1025
`Page 4 of 10
`Page 4 of 10
`
`

`

`optimization. Once the mixer diode parameters are given and
`the mixer circuit has been specified, this tool calculates the volt—
`age and current waveform at each of the circuit’s node. This is
`done by employing SPICE IT], a CirCuit simulator program, to
`arrive at the steady state solution for the network equations. To
`quantify conversion 1068 performance and to generate small sig-
`nal impedances. Fourier transforms are taken at various nodes.
`With this technique the interaction of diode and mixer circuit
`can be understh and conversion loss performance optimized
`for realistic mixer circuits. (See [8] for a description of the mixer
`design tool.)
`Previous mixer analysis techniques have been better suited
`to analysis of ideal circuits where the output load at each mixer
`produced frequency (m far-in.fm) are specified
`These tech-
`niques are ideally suited to the design of waveguide mixers or
`mixers where fflp differs greatly from fun so that each mister
`spur can be easily separated and matched by a simple circuit
`that are not necessarily very frequency selective.
`The time domain based mixer dign tool has been used to
`optimize mixer performance, both conversion loss and impedance
`transformation, with practical mixer diode having both series
`resistance and capacitance. Medium barrier p- type Si Schottky
`diodes are used as they are thought to possess low lff noise.
`Mixers have been designed with devices whose barrier capac—
`itance at zero bias, C3,. is as great as 1.0 pF; the maximum
`typically encountered for UHF mixer diodes. Two mixer types
`have been optimized: single-ended and single-balanced mixers.
`Figure 6(a) depicts the optimized single—ended mixer. The lo
`cal oscillator source is applied to the circuit via a directional
`coupler. The optimized single-balanced mixer is illustrated in
`Fig. 6(b). Balanced mixers differ from singleended mixers in
`several ways: first the output signal has fewer spectral compo-
`nents as it is isolated from the L0 source and its harmonics
`(therfore providing greater isolation of the RF and L0 signals
`at the output port); second the mixer has fewer noise compo-
`nents at its output as the amplitude modulated component of
`the L0 noise is cancelled internally by the balanced circuitry;
`third good circuit VSWR can be obtained. For these reasons
`the added circuit elements and active components of balanced
`mixers are often tolerated.
`
`Table 1 lists the optimized performance of two single-ended
`mixers With Cy. of 0.25pF and LUpF and the performance of a
`single-balanced mixer with a C59 of 0.25pF. Single—ended mixers
`have a phasing circuit (L1, Cg) to achieve low 0;; balanced
`mixers have both phasing circuit {L1, C1) for RF signal and
`driving waveform circuit (L3, 04) for LO signal to achieve low
`CL.
`
`Noise temperature timates at the mixer output frequency
`are based on mixer noise analysis for shot and thermal noise pro—
`cesses as detailed by Mass [10]. Only single—ended mixer noise
`temperature: can be calculated. The embedding network is as-
`sumed to be noiseless. The analysis makes use of the Fourier
`coefficients of the diode’s small signal barrier resistance, Rafi},
`and barrier capacitance, (IE-(t), to calculate the noise power
`at the output termination. This method for evaluating mixer
`noise temperature does not include lff noise. The balanced
`mixer noise temperature has been estimated from Alpha man-
`ufacturer’s characterization of lji diode noise1 and from esti-
`mates of circuit effects on mixer noise. The latter is expected
`to be low as the mixer circuit parametrically pumps the barrier
`capacitance to achieve low conversion loss. The worst—case noise
`temperature for the balanced mixer is believed to be so 500 K.
`
`Better characterization of mixer noise over frequcy remains to
`be done.
`From Table 1 the conversion loss of the balanced mixer can
`
`It is
`be seen to be lower than that of the singleended mixer.
`also capable of translating from a lower input impedance {730)
`to a higher output load impedance {2K9}.
`
`I'M”
`
`zwun
`
`
`
`
`(a) singleended mixer
`
`(b) singlebalanced mixer
`
`Figure 6: Schematics of the two mixer types simulated.
`
`2. Low {Enough low noise amplifier.
`The goal of this work was to identify transistor noise mechanisms
`and circuit characteristics that determine the optimum source
`impedance for minimum amplifier noise, Rm,“ and to find the
`variation of noise figure and RN”. over frequency. The noise
`model for an amplifier consists of a current noise source, 17?,
`across the input and a voltage noise source, 3?, in series with
`the input to an ideal noiseless amplifier. The noise temperature
`of the amplifier,I FM“... is then defined as:
`
`
`+ ER‘
`E?
`F — 1 +
`)
`(
`dkTR.Af emf
`'
`From the above equation the optimum R, for minimum noise
`temperaturefiym, can be found to be:
`
`4
`
`values for two popular op—amps, the Ana—
`and
`Using the
`log Devices (JP-07 and OP-27, an idea of the noise figure — R,
`trade-off may be obtained by using Eqn. 4 to calculate amplifier
`noise figure. The results are shown in Figure 7 for operation at
`10 Hz and 1 KHz. The results shown correspond to the mini-
`mum op—amp noise figure attainable as no feedback around the
`
`amp is assumed. (Actual op-amps would have slightly higher
`
`1Corrupondance with Stanley Howe of Alpha Industries Inc.
`lQBT.
`
`in March
`
`:57
`
`TCL EXHIBIT 1025
`TCL EXHIBIT 1025
`Page 5 of 10
`Page 5 of 10
`
`

`

`
`
`is the dominant noise source, at
`noise figures.) At large R,
`low it. E? dominates. From Fig. 7 RN”, can be seen to be 30K
`at Hills and 75K at lKHz for the OP-UT with corresponding
`amplifier noise figures of 1.6 dB and 0.6 dB, respectively. The
`CUP—27 is a ultra—low noise precision op—amp; it possesses a. RN”.
`of 2.2K at 10 Hz and 8K at lKHz with corresponding noise fig-
`ures of 2.6dB and 0.64dB. To understand if better noise figures
`are possible for R. < 8K with other bias and circuit designs,
`as will be necessary to optimize the loading of the low conver-
`sion loss direct conversion mixer, the noise mechanism that set
`amplifier performance must be examined.
`
`Table 1: Direct Conversion Mixer Performance
`
`SingleEnded
`Mixer
`
`Single-Balanced
`Mixer
`
`C,‘
`Bias and L0 Drive:
`
`0.25 pF
`
`1.0 pF
`
`0.25 pF
`
`13m 20m
`13m 30m
`
`l3mA
`-U.5 V
`
`197 Q
`3 5 pH
`
`66 [l
`1.2,uH
`
`73 Q
`0.13 pH
`
`U 11 pH 0.04 pH
`0.03 pF'
`0.12 pF
`
`[1.07 pH
`0.04 pF
`
`
`IcNolanFlgtmtdB}
`
`Ibo Bias
`'
`HF source
`impedance:
`
`RF matching
`
`LO matching
`
`impedance:
`
`none
`none
`
`1.5 K9
`3 pF
`
`3.1dB
`7.6
`
`135K
`
`.
`
`0.08 ,uH
`0.04 pF
`
`2 K9
`3pF
`
`2.] dB
`27.4
`
`$3500K
`
`—I—- 09-07 son:
`---I--- user—1m:
`
`——*—- oesr1ma
`"'+" {JP-21M KHZ
`
`10
`
`“30
`
`1000
`
`10000
`
`100000
`
`Source “It! Illne- (OM!)
`
`Figure 1': Best case noise figure of the Analog Devices 0P-07
`and 0P-27.
`
`The difl'erential amplifier is chosen as the basic building block
`for the low frequency amplifier because of its rejection of signals
`which appear on both input ports. Differential amplifiers are
`commonly used as input stages for op-amps and hence will pro»
`vide an understanding of their performance. Lastly, because
`their symmetrical design is less succeptible to thermal drifts
`causing performance changes.
`these amplifiers are often used
`where operation at low frequencies (and DC} is desired [11].
`The three stage low frequency amplifier is illustrated in Figure 8.
`The difierential amplifier gain stage is outlined. Transistors I91
`and Q; are biased by a transistor current source, created by Q;
`and 04. This is the common form of op—amp bias as transistor
`current sources provide increased circuit performance insensi-
`tivity to temperature and power supply variations than emitter
`resistors and require less space on the integrated circuit [12].
`The voltage noise source at the input of the differential am-
`plifier, a. can be written as:
`
`=
`
`+
`
`[Emsuflawsiflsf-s
`Rr+rs+fs+Rsl
`e,-
`+ —.a.—“—
`
`(R.+n+r.+R‘g (fi.+1))’ T]
`fl
`4R:
`(
`2r,r
`)
`2n+R.
`'1
`+E+(rs+Rg,) «*3:
`(rh‘l-rx'i'Rfia): T
`'l‘ —"‘"fi— ‘1”
`+ (rb+rr+::§{flo+1))2_;r
`
`.
`
`(5)
`
`where
`
`
`
`These equations strs the importance of choosing a device
`with a. low series resistance, n, and a high ,6 to minimise
`Device parameters for a super beta differential transistor pair,
`the National Semiconductor LM 394, are used for transistors Q;
`to Q‘. The collector current dependence of r. and ,3 is included
`
`
`
`Figure 8: Lowr frequency amplifier implemented by a cascade of
`three dilferential amplifiers. The outlined input dilferential pair
`with transistor current source is fundamental gain block studied.
`68
`
`TCL EXHIBIT 1025
`
`TCL EXHIBIT 1025
`Page 6 of 10
`Page 6 of 10
`
`

`

`
`
`in the device model. Using this transistor in the circuit depicted
`in Fig. 8,
`the noiSe figures shown in Figures 9 are obtained:
`Fig. 9(a) gives the performance over R, at 10 Hz, Fig. 9(b)
`charts performance at l KHz.
`The frmuency dependence of few and RN”. for minimum
`noise figure is apparent from Fig.9. At 10 Hz, NF...“ is 2.0 dB
`with Riva" between 15 K and 20 K and a In,” of 10 ,uA. At 1
`KHz, NF..."I is 0.4 dB with Ryan between 20 K and 25 K and
`a few, of 25 Jim.
`_
`Op-amps are typically biased at an I: of 5 put for lower power
`consumption; they are generally intended for operation at f 3
`100 Hz, thus requiring a. high R, to minimize noise figure per-
`formance. To operate at a lower R. without suffering a large
`penalty in increased noise figure, Fig. 9 indicates that One can
`operate the transistors at a higher bias. For example, at 10 Hz
`NF is 2.3 dB when a R, of 5K is used with an In of 25 uA (only
`0.25dB higher than Nan), or, at 1 KHz NF is 2.3 dB when a.
`R, of SR is used with an 16 of 25 wk [0.GdB higher).
`
`10
`
`100
`
`1000
`
`10000
`
`100000
`
`{as}
`
`AmpllflarNolanFlgure
`(dB)
`
`
`AmpllfiarNoiseFlguro
`
`Scum Hulntance (ohms)
`
`(a) Performance at 10 Hz
`
`fa1KHl
`
`100
`
`1000
`
`1 0000
`
`100000
`
`Source Hoslllanco (ohm!)
`
`(b) Performance at l KHz
`
`Figure 9: Noise figure vs. R. performance for the differential
`amplifier shown above.
`
`69
`
`To optimize operation with the optimized direct conversion
`mixers whose design is listed in Table I. the differential amplifier
`can be analyzed for R, equal to 500 (I, 1.5 K and 2.0 K, the
`optimum load impedances for the two singloended mixers and
`the balanced mixer. The required 1., for optimum noise figure are
`tabulated in Table 2 along with the noise temperature results.
`
`Table 2: Gain and Noise Performance of the Gain Stage
`
`
`
`
`I = 100 Hz:
`500 n
`1.5 KQ
`2.0 K!)
`
`The large difference in Optimum I, for minimum noise figure
`is visible as the frequency of operation is varied. L9,, is 25 {1A at
`[0113 and 03 ,uA at lKHz. Thus if the notch filter is to he very
`narrow to minimize signal distortion, eg.
`0 - 10 Hz, then the
`differential amplifier would have to be biased at low currents w
`25 mi. However if a filter from 0 — 100 He can be tolerated then
`the appropriate bias current to minimize noise would probably
`be A: 100 yA. Minimizing current is nemsary to conserve power
`and maintain good performance stability over temperature and
`power supply fluctuations. The dependance of optimum bias
`and minimum noise figure on R, is also illustrated in Table 2.
`At any of the frequencies the following two trends are observed:
`['1] more current is required to achieve N Fmpm." as R. l; (2)
`NFL",me 1‘ as R. 1.
`As the balanced mixer has the best conversion loss, the high-
`est impedancc transformation ratio, and, produces the lOWest
`fum, in addition to the inherent benefits of balanced structures
`it will be chosen as the prototypical direct conversion mixer.
`Table 3 gives the performance of the dificrential amplifier when
`driven by a 2K!) source resistor, such as will be