the rf reader adi
TRANSCRIPT
The RF Reader
A collection of published articleson RF from Analog Devices
www.analog.com/RF
Smart Partitioning for WiMAX Radios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
AD8368: A Broadband RF/IF VGA with 34 dB Linear-in-dB Gain Control Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Measuring the RF Power in CDMA2000 and W-CDMA High Power Amplifiers (HPAs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Simulating the Effect of Blockers on Data Converter Performance in Wideband Receivers . . . . . . . . . . . . . . . . . . . . . . . . . 15
Converter Performance Approaches Software-Defined Radio Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Calibration and Temperature-Compensation Techniques Using an RMS-Responding RF Detector . . . . . . . . . . . . . . . . . . . . . 21
Improved DDS Devices Enable Advanced Comm Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
RF Standards for Short-Range Wireless Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
A Broadband I/Q Modulator for Broadband Radio Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Evaluating Linear Distortion in ADC Driver Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
IQ Modulators Advance Reconfigurable Radio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
How to Determine an Effective Damping Factor for a Third-Order PLL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Log Amps and Directional Couplers Enable VSWR Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Direct Digital Synthesis Enables Digital PLLs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A Direct-Conversion Transmitter for WiMAX and WiBro Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Converters for 3G Are Optimized for Cost, Size, and Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
RF Power Detection: Measuring WiMAX Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Measuring VSWR and Gain in Wireless Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Low Power Direct Digital Synthesizer Cores Enable High Level of Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
The RF ReaderA collection of published articles
on RF from Analog Devices
. . .
WWIIRREELLEESSSS TTEECCHHNNOOLLOOGGIIEESS
Figure 1 shows a block diagram of atraditional WiMAX system. An RF trans-ceiver is connected through a power am-plifier (PA) and RF switches to the anten-na on one side, and to a digital baseband(DBB) on the other. The interface be-tween the RF transceiver and the DBB iscomposed of analog signals, which canbe at intermediate frequency (IF) or base-band. Note that the ADCs and DACs inthis architecture can be discrete devices,or can be integrated on an ASIC.
In some applications, a two-chip solu-tion may have higher performance andlower cost than a single-chip solution.The key is to know how to divide thefunctions between the two chips to bestexploit both the circuit topology and theavailable manufacturing technologies.Smart partitioning does just this, allow-ing an RF system-on-a-chip (SoC) to pro-vide a complete RF-to-bits solution in-cluding all required automatic gain con-trol, transmit power control and RFcalibration loops. Including control loopson the radio front end enhances ease ofuse, provides for an easier mix-and-match capability with different DBBmodems and improves performance. The
T he need for broadband wireless ac-cess (BWA) has long been acknowl-edged as the next step in the evolu-
tion of Internet access. Unfortunately, thelack of robust technology at a competi-tive price has been a barrier to its imple-mentation. Today, though, momentum tocross the chasm is gathering—earlyadopters have endorsed the technologyin under-served rural areas of the world,while standardization efforts have re-duced costs enough that mainstreamusers can now consider WiMAX a viablealternative for broadband access with afuture promise of mobile access.
WiMAX, based on IEEE 802.16 specifi-cations, supports operation in multiple fre-quencies and multiple air standards. To en-sure interoperability between multipleWiMAX solutions, the WiMAX Forum, anindustry consortium, has developed pro-files that specify the operating frequency,bandwidths, air-interface and medium ac-cess protocols. These profiles are based ona 256-carrier orthogonal frequency divi-sion multiplexing (OFDM) air interface forfixed/nomadic operation, and scalable-OFDM-access (S-OFDMA) air interface forportable/mobile applications.
Smart Partitioningfor WiMAX Radios
NOMAN RANGWALA AND RICK MYERSAnalog Devices Inc., Norwood, MA
Reprinted with permission of MICROWAVE JOURNAL® from the November 2006 issue.©2006 Horizon House Publications, Inc.
1
accompanying reduction of real-time software control results insimpler system design. All analogand RF specific controls are inte-grated on the RF front-end IC.This is smart partitioning. Figure2 illustrates a block diagram for asystem using smart partitioning.
COST BENEFITS ENABLED BY SMART PARTITIONING
For communication systemssuch as WiMAX and BWA, con-sumer prices less than $100 areessential. In CPE equipment forasymmetric digital subscriberloop (ADSL) and 802.11g Wi-Fi($20 to $30), for example, volumesincreased dramatically as pricesdeclined. Emerging markets suchas WiMAX are also experiencingsimilar price pressures. End-userCPE prices are expected to beless than $100 by mid-2007. Toachieve these targets, chipsetpricing must fall to $20 or $25.Much lower than the current
cost, this reduction will requiresignificant improvements formarket prices to yield an accept-able profit.
Smart partitioning offers theopportunity to dramatically re-duce the total cost of a WiMAXsystem. Today’s traditional DBBsare mixed-signal ASICs, withover 90 percent of their area oc-cupied by digital gates and 5 to10 percent used for data convert-ers. The cost to manufacture sucha mixed-signal device is over 1.5times the cost of manufacturing adigital-only IC. The major con-tributors include higher waferprice (1.2 times), higher test cost(1.1 times), higher yield cost (1.1times) and larger die size (1.05times), totaling a 1.5 times in-crease in cost.
In addition to the tangible cost,there is a large opportunity costincurred. Data converters typical-ly lag behind by one generationof the process, and proven cores
for 90 nm or 65 nm are not avail-able for integration on today’sfine-line digital processes. Theopportunity cost for using a 130-nm process for the digital base-band instead of a state-of-the-art90-nm process can be up totwice. Data converters integratedon a DBB constrain the cost,keeping the IC from taking ad-vantage of Moore’s law.
EASE OF USE ENABLED BY SMART PARTITIONING
By itself, the integration ofdata converters is not sufficientfor smart partitioning. The dataconverters required for WiMAXare typically over-sampled, sohandling the raw data rate in andout of the transceiver would pre-sent implementation challenges.However, integrating decimationand interpolation filters on thetransceiver allows the interfacespeed to be reduced. The avail-ability of mature fine-line RFCMOS processes, coupled withadvances in analog and RF mod-eling capabilities, have now madeit possible to move data convert-ers and other mixed-signal blocksto the RFIC in WiMAX radio de-signs. For cost and power effi-cient implementation of the digi-tal blocks, fine-line CMOS is adefinite plus. This article exploresthe choice of digital interface andthe ease of use advantages intro-duced by simple RF drivers forreceivers and transmitters.
DIGITAL INTERFACE: CHOICES,ISSUES AND CHALLENGES
The evaluation board designand layout has a critical impacton the performance of the mixed-signal component of the DBB.The analog I/O on the referenceboard is sensitive to externalnoise, and the supply routes tothe mixed-signal portion of thedesign require high isolation.Eliminating the analog I/O mini-mizes these noise-coupling is-sues, and solves the problem ofinterfacing analog cores from dif-ferent vendors (such as RF chipand mixed-signal convertercores). For example, some ADCcores require a discrete 5 V driverop-amp to obtain specified data
WWIIRREELLEESSSS TTEECCHHNNOOLLOOGGIIEESS
Ref
PLL
ADC
LO ADC
DAC
DAC
Digital Modem/MAC
Slow SpeedControl Port
Mixed Signal is integrated ondigital ASIC or Standalone
Real Time Control Signals
Real Time Control Loops partitioned betweentwo separate chips and vendors
Memory
▲ Fig. 1 Block diagram of a traditionally partitioned WiMAX system.
Ref
PLL
ADC
LO ADC
DAC
DAC
Digital Modem/MACValidated by gatelevel simulation
Digital I/Q
Slow SpeedControl Port
Real Time Control Signals
Real Time Control Loops Integratedon RF Transciever
90/65nm digitalonly process
Memory
▲ Fig. 2 Block diagram of a WiMAX system using smart partitioning.
2
sheet performance. Modems us-ing smaller processes, such as130 nm or 90 nm, must reducethe signal swing and match thecommon-mode level to that of theRFIC. These considerations re-quire valuable engineering re-sources. For systems using smartpartitioning, the boundary be-tween the transceiver and theDBB is digital, simplifying theseissues.
Two basic options for selectinga digital interface are a highspeed serial data stream usinglow voltage digital swing (LVDS)signaling, or a slower speed par-allel bit stream. Variations ofthese schemes include embedded
clock, synchronous clocking, ornibble transfers. Each approachhas its advantages and disadvan-tages.
High speed serial links (seeFigure 3) have a lower pin count,reduced switching noise due tothe differential signaling andlarger separation (between DBBand RF transceiver). However,the high speed circuit design riskis one of the biggest implementa-tion challenges. Implementationof the serializer and de-serializeris complex and requires clock re-covery circuits and other customdesign blocks that are not readilyavailable in standard digital li-braries.
The parallel bit stream ap-proach offers lower data ratesand a standard CMOS I/O, but in-creases the pin count. To reducethe pin count to a manageablenumber, the data bus can usetime duplexing to multiplex be-tween receive and transmit data.Additionally, the I/O can be sin-gle-ended if the switching andhigh frequency noise are careful-ly managed and isolated from thehighly sensitive RF circuitry. Thedesign on the DBB is straightfor-ward, and can be implementedwith a standard hardware de-scription language (HDL)-baseddesign flow.
The JEDEC Committee (JC-61),formed in 2002, was chartered tocreate an open standard for digi-tal interface, enabling smart par-titioning and multi-vendor solu-tions. The published standard,JESD96, offers the high speedLVDS approach. A proposal andbasic configuration for the paral-lel interface have also been ac-cepted. Figure 4 shows an exam-ple of a parallel interface imple-mented on Analog Devices’AD935x family of smart parti-tioned transceivers. The ADI/Q™digital I/Q interface provides thebasis of the JC-61 parallel stan-dard.
AUTONOMOUS AUTOMATICGAIN CONTROL (AGC)
In addition to ADCs and deci-mation filters, smart partitionedtransceivers also integrate the au-tomatic gain control (AGC) circuit-ry on the transceiver. The AGCadjusts the gain of the receiverpath such that the input signal tothe ADC is maximized in scenar-ios with and without interference.The AD935x receiver signal chainis illustrated in Figure 5.
Time division duplexing (TDD),the preferred system for the fu-ture, supports framed waveforms(bursts). The media access con-troller (MAC) at the base stationgenerates a downlink frame,which starts with a preamble,and follows with a frame controlheader and multiple data frames.The duration of each frame isshort (1 to 2 ms). The input powerduring the burst varies by 3 dB
WWIIRREELLEESSSS TTEECCHHNNOOLLOOGGIIEESS
Rx Clock
Tx Clock
Data Field Data Field
Data Field
Frame Boundary must be a Multiple of M(User defined variable)
Register Field
Sync
Sync
Tx Data
R/W
Modem
ProsLow Pin Count
Differential Signaling
ConsComplex ImplementationHigh Speed (≈500MHz)
RF Transceiver
Sync
Header
Header
Header
I Q I Q Control Field CS Address Data CRC P
Rx Data
PNullP
▲ Fig. 3 Digital interface — Serial interface option.
D0:D9
SYNCSYNC_VALID
TXNRXEN_AGC
ENABLE
CTRL_OUT
SPI
CLK
MACModem
AD935x
ADI/Q™ — A Digital I/Q Interface
RF FE
2
3 4
2
GPO
RXTX
AUX_DAC
10
▲ Fig. 4 Parallel interface example from the AD935x family of smart partitioned transceivers.
3
for fixed systems and 10 to 12 dBfor mobile systems.
The transceiver uses the framepreamble to lock the gain of thereceiver (see Figure 6). The pre-amble is one or two OFDM sym-bols consisting of multiple toneswhose phases are aligned to cre-ate a waveform with a smallpeak-to-average power ratio. Thetones are also distributed such
that the waveform is repetitive inthe time domain.
To detect the minimum desiredsignal, the receiver gain is set tomaximum. The in-channel re-ceived power is measured at theoutputs of the ADCs and decima-tion filters. The peak detectorsdistributed along the receiverchain and the ADCs are alsomonitored for over-ranging. If
detected, the receiver gain is re-duced depending on the type ofover-ranging. If the basebandpeak detector before filters indi-cates clipping, for example, thenthe LNA gain is stepped down.The AGC algorithm cyclesthrough these iterations and con-verges on an optimum gain set-ting. The system then freezes thegain for the remainder of theframe. For the modem to syn-chronize and correlate to the sig-nal, the receiver gain must befixed. A fast AGC lock time al-lows the modem more time tosynchronize and make accuratechannel estimates, reducing theimplementation loss and improv-ing the system performance.
Traditional systems achievethis by distributing the AGCfunction on both modem andtransceiver. The dotted line in thereceiver architecture indicatesthe functional partitioning. In thisapproach, the DBB must monitorthe gain, detect peaks and set anew gain. The algorithm is gener-ally implemented in the RF soft-ware driver. Every time the gainis changed, the transceiver mustbe recalibrated and the DC-off-sets must be removed. The RFdriver must maintain accuratetiming and must respond to in-terrupts generated from thetransceiver, making optimizationa tedious, time-consuming task.In the case of multiple vendors,each RF driver must be cus-tomized for a specific transceiver.
Preamble Preamble
TransitionGap
UplinkDownlink
Signal Detection,
AGC,AntennaSelection
CoarseFrequencyOffset and
TimingSynchronization
Channeland Fine
FrequencyOffset
Estimation
FCH
VariableGain Fixed Gain
DLBurst
2
DLBurst
n
DLBurst
1
ULBurst
1
ULBurst
2
▲ Fig. 6 The 802.16 OFDM waveform.
ENV
ELO
PE
(V)
TIME (μs)
1.0
0.8
0.6
0.4
0.2
00 10 20 30 40
Input Waveform: 802.16 OFDM Downlink
SQRT (I2 + Q2) vs. time
Gain LockedAGC Lock time < 4 μs
▲ Fig. 7 Digital I/Q receive signalmeasured at the output of the AD935x.
CP
CP
CP
CP
CP
CPPreamble Symbol 2 Symbol n Symbol n+1 Symbol n+2Symbol 1
Beam-formed Signal
TIME
Fading within a burst
Zone 1PUSC
Zone 2MIMO
3 dB
9-12 dB
INP
UT
PO
WER
▲ Fig. 8 Power variations in a burst for advanced MIMO and beam-formed signals.
LNA
Mixer
Mixer VGA
VGA
OVR
OVROut ofBand In
Band
To Modem
To Modem
LNA Gain Setting
DC Offset DAC
LPF Gain SettingVGA Gain Setting
Calibration SettingsLPF Gain SettingLNA Gain SettingVGA Gain SettingDC Offset DAC
InBandOut of
Band
ADC
ADC
LPF
LPF
PeakDetector
AGC/Offset
ControlDA
C
DA
C
▲ Fig. 5 Receiver architecture on the AD935x family of smart partitioned transceivers.
WWIIRREELLEESSSS TTEECCHHNNOOLLOOGGIIEESS
4
WIRELESS TECHNOLOGIESsymbol-to-symbol AGC. Thepower variation within a burst inthis scenario is 9 to 12 dB. A typi-cal power variation versus timefor a beam-formed or space-timecoded waveform is shown in Fig-ure 8. To accommodate this pow-er step, one option is to increasethe ADC dynamic range and paythe corresponding cost of in-creasing a bit in performance.Another option is to reacquirelock on a symbol-by-symbol ba-sis. A fast AGC with short locktimes, coupled with accurate tim-ing control for starting and stop-ping the AGC loop, could enablesymbol-to-symbol AGC.
Figure 9 shows a plot of re-ceiver input vs. measured receiv-er error vector magnitude (EVM)for a WiMAX transceiver, usingsmart partitioning along with asoftware modem implementedwith Agilent VSA software. Theperformance curve exemplifiesthe ease of implementationachieved using the smart parti-tioning approach. For low inputpower, the receiver gain is auto-matically adjusted to accommo-date the small input signal; thegain is then backed off automati-cally until the EVM is limited bythe linearity of the transceiver.
TRANSMITTER POWERCONTROL
The 802.16 standard specifies aranging process that determinesthe correct output power radiatedby the terminal. This process en-sures that transmissions frommultiple terminals arrive at thebase station at the desired powerlevel (within a certain range thatcan be handled by the base sta-tion). The standard specifies a
transmit power control range of50 dB for the terminal. This willallow terminals to be distributedaround the cell site to meet theequal power criteria at the basestation.
During the ranging process,the base station requests the ter-minal to send out a ranging sig-nal. The base station will thencommand the terminal to in-crease or decrease its transmitpower. The WiMAX Forum iscurrently discussing require-ments for accuracy and numberof iterations.
In traditional architectures, ex-ternal attenuators and true rmspower detectors can be used toachieve the system specifications.Using smart partitioning, the in-tegrated ADCs and DACs offerthe transmitter the ability torapidly measure highly accurateburst output power. Figure 10shows the components of theunique transmit power controlscheme implemented on theAD935x transceiver. To utilize thepower detector, the transmittedsignal is sensed from an externalcoupler. It is then fed back to thereceiver, where it is down-con-verted to baseband. The receiveADCs digitize the signal, which isthen processed by a digital rmspower meter block. This takes ad-vantage of the half-duplex natureof the system to make an accuratemeasurement using the idle cali-brated receiver path. The detec-tor is capable of measuring pow-er on a TX burst-by-burst basis,providing the modem with near-real-time power information. Thefront-end mixer is designed to betemperature and frequency inde-pendent, and thus requires only a
In multiple instances, vendorshave struggled to achieve64QAM operation on their refer-ence designs because of thesecomplex interactions.
A transceiver using smart par-titioning integrates the completecontrol loop including monitor-ing and control algorithms on asingle device. The basic processto lock the gain remains thesame, but the responsibility forcontrol is transferred to the RFtransceiver. From the modem’sperspective, the loop is au-tonomous and does not requireany dynamic interactions. Themodem can still accurately startand stop the loop, and can stillmonitor the received signalstrength indicator (RSSI) andgain settings. All internal calibra-tions are now self-contained.
With well-managed timingconstraints, this smart partition-ing approach results in two ad-vantages: a simpler RF driver andshorter AGC locking time. Fig-ure 7 shows that the AGC locktime, when using the autono-mous AGC loop on the AD935x,is of the order of four microsec-onds for an 802.16 waveform.Further advanced techniquessuch as stronger signal detection,radar detection and interferenceback-offs can also be easily im-plemented.
Advanced systems, operatingin mobile environments with fad-ing channels, multiple antennasand beam-formed signals, re-quire new techniques such as
EVM
(dB
)
INPUT POWER LEVEL (dBm)
−20
−25
−30
−35
−40−80 −60 −40 −20 0
230025502700
MaximumSignal
MinimumSignal
▲ Fig. 9 Receiver performance of a smartpartitioned receiver implemented on theAD935x.
WWIIRREELLEESSSS TTEECCHHNNOOLLOOGGIIEESS
DAC Modulator
Decode
External AntennaInterface
LORMS
Control
txrssi
pdref
ADC
▲ Fig. 10 Smart partitioned transmitter block diagram.
5
WIRELESS TECHNOLOGIESone-point factory calibration.This feature saves test time andreduces calibration complexity.
The ease of use advantage isequally applicable to the trans-mitter. The modem can vary thepower on a burst-by-burst basisby simply writing to the registerbefore the burst. The transmitpower is automatically adjustedto the open loop accuracy specifi-cation of the device. If greater ac-curacy is required, the transmitpower control loop can be initiat-ed and a correction factor can beapplied to the next burst. Othernovel techniques such as self-generated short test signals canbe transmitted before the actualburst to calibrate the device inclose loop. These techniques canbe explored as emissions require-ments and system requirementsevolve.
SUMMARYThe smart partitioning of a
WiMAX system enables the low-est system cost and reduces thedependence of real-time controlfrom the DBB. Integration of theADCs and DACs by itself is notsufficient to achieve these advan-tages. To reduce the speed of thedigital interface, decimation andinterpolation filters are also inte-grated in the transceiver. Thesestages also include the channelfilters. A large portion of the RFdriver complexity is managingthe real-time signaling betweenthe modem and transceiver toachieve fast and accurate AGCand TPC. To reduce the process-ing load on the modem, the AGCand TPC algorithm blocks are in-tegrated on the transceiver. Othersmart features can be integratedon the transceiver such as auxil-
iary ADCs and DACs, RF gener-al-purpose outputs for RF switchand PA control. The AD935x fam-ily of transceivers exemplifies theRF system-on-chip features thatcan be implemented. ■
Noman Rangwala received his bachelorof engineering degree from VictoriaJubilee Technical Institute, Mumbai, in1992, his MSc degree from the Universityof New Mexico, Albuquerque, NM, in 1994,and his MBA degree from San Diego StateUniversity, San Diego, CA, in 2000. He is amarketing manager with responsibility forWiMAX transceiver marketing within theHigh Speed Signal Processing division ofAnalog Devices Inc. (ADI). He also hasover six years of experience in IC designand development.
Richard H. “Rick” Myers received his BSdegree from Old Dominion University in1983. He is a senior applications engineerfocused on wireless applications for theHigh Speed Signal Processing division ofAnalog Devices Inc. (ADI) and is based inRaleigh, NC. Before joining ADI, he wasCDMA hardware development managerfor handsets at Ericsson in ResearchTriangle Park, NC, and principal electricalengineer of RF and receivers at E-Systems(Raytheon) in Falls Church, VA.
WWIIRREELLEESSSS TTEECCHHNNOOLLOOGGIIEESS
6
FEATURE ARTICLE
AD8368: A Broadband RF/IF VGA with 34 dB Linear-in-dB Gain Control Rangeby Phillip Sher, Analog Devices
Introduction
Most wireless receiv-ers incorporate some form of variable gain.
Variable gain, which is usually implemented at Intermediate Frequency (IF), is used to ampli-fy or attenuate received sig-nals so that a constant signal level is presented to subsequent stages in the receiver. Usually implemented at Intermediate Frequency (IF), variable gain is used to amplify or attenuate received signals so that a con-stant level is presented to subse-quent stages in the receiver.
Variable Gain Amplifiers (VGAs) will either have digital gain control or analog gain con-trol and are generally optimized for either receive or transmit applications. While the choice of an analog controlled or digitally controlled VGA often comes down to the preferences of the equipment designer and the available control signals, there are solid technical reasons for choosing one over the other. This article will discuss those trade-offs and will go on to discuss the operation and appli-cation of Analog Devices’ new analog controlled VGA.
A Linear-in-dB IF Variable Gain AmplifierThe AD8368 is an IF VGA with an operational bandwidth from near DC to 800 MHz and a gain control range of 34 dB from -12 dB to +22 dB. It also includes an onboard square-law detector that can be used in an AGC loop with the VGA. Using Analog Devices’ patented X-AMP architecture, the AD8368 achieves accurate linear-in-dB gain control. The part is designed with 50 input and output impedances.
The main signal path consists of a variable input attenuator followed by an integrator and output buffer. Feedback around the integrator creates a fixed-gain amplifier. Figure 1 shows a block diagram of the VGA.
Input Attenuator and InterpolatorThe input attenuator is built from a resistor ladder with eigh-teen -2 dB tap points. Each of these tap points is fed into
separate variable transconduc-tance gm stages, whose out-puts are summed and fed into an integrator. Gain control is achieved by using the GAIN pin to control the interpolator. As GAIN is swept from 0 V to 1 V, the interpolator selects different tap points by varying the transconductance of the gmstages. For gains between two tap points, the interpolator var-ies the transconductance such that the weighted sum of several adjacent tap points are chosen. In this way, an accurate contin-uous linear-in-dB gain control response is produced.
Integrator and Output BufferThe current outputs of the gm
stages are summed and fed into an integrator. Resistive feed-back from the output of the integrator to the gm stages cre-ates a low-noise, high-linearity fixed-gain amplifier. The out-put of this amplifier is fed into the output buffer, which provides an active 50 output impedance and additional 6 dB of fixed gain.
Output DC Level and Offset CorrectionSince the AD8368 is single-end-ed, the DC levels at the input and output are regulated to VPSI/2 by an internal regulator. The output of this regulator is connected to the DECL line and requires an external decoupling
capacitor. Since the fixed-gain amplifier and output stage have an extremely large overall gain, small DC offsets at the input of the fixed-gain amplifier can lead to large output offsets. To correct for these Vgain, supply and temperature variations, a low-pass offset correction loop is used which senses and main-tains the output DC level at the voltage on the DECL pin. The low-pass corner frequency of this loop is controlled by the size of the capacitor on the HPFL pin.
The AD8368 contains an accurate stand-alone RMS detector that enables versatile AGC operation. To form a com-plete AGC loop, the MODE pin is pulled low and the detector output pin is directly connected to the GAIN pin and an inte-grating capacitor. Then, by con-necting the VGA output OUTP directly to the detector input DETI, the output is leveled to the set-point -11 dBm. This ref-erence level can be raised by dividing down the output signal before applying it to the detec-tor input, allowing for a range of AGC levels.
Input and Output ImpedancesThe input to the AD8368 should be externally AC cou-pled to prevent disrupting the DC levels on the chip. Thus, a large coupling capaci-tor should be used, so that the series impedance of the capacitor is negligible at the frequencies of interest. On the chip, the input is connect-ed directly to a resistor ladder network whose impedance is nominally 50 .
The output of the part should also be AC coupled to prevent disrupting the output DC level. As with the input, a sufficiently large value of capacitance should be used so that the series impedance of the capacitor is negligible at the frequen-cies of interest.
The output impedance is synthesized by the output buf-fer. The fixed gain of the out-put buffer combined with the resistive feedback from output to input provides a nominally 50 output impedance.
Figure 1: Simplified Block Diagram
Figure 2: Gain Control Transfer Function and Linearity at 70 MHz.
7
Accurate Linear-in-dB Gain ControlThe AD8368 has a linear-in-dB gain control interface that can be operated in either a gain-up (positive gain control sense) or gain-down (nega-tive gain sense) mode. With the MODE pin pulled high in the gain-up mode, the gain increases with increasing gain control voltages; in gain-down mode (MODE pulled low) gain decreases with increasing GAIN voltages (Figure 2).
With MODE pulled high, the ideal gain function is given by the equation:
Gain(db) = 37 VGAIN – 14
With MODE low, the ideal gain function is given by the equation:
Gain(db) = -37.5 VGAIN + 24.8
where VGAIN is expressed in Volts.In addition to showing the
gain-up and gain-down transfer functions, Figure 2 also shows the error plot that reflects the deviation between the ideal gain control transfer function and real performance. From Figure 2, it is clear that the gain control transfer function closely follows the ideal for well over 30 dB.
Figure 3 shows the effect of temperature on the gain con-trol function at 140 MHz. To more closely examine the per-formance over temperature, the slope and intercept of the 25º C data is calculated over the linear part of the gain control range. This provides a sim-ple linear model to compare with the actual response of the AD8368. The comparison
of the ideal linear model to the three transfer functions at 25º C, -40 º C, and +85º C yields the linear conformance error curves scaled in dB. This method of error calculation resembles the expected error from single temperature calibra-tion during system production.
Using the AD8368 VGAThe AD8368 is a general pur-pose VGA suitable for use in a wide variety of applications where voltage-control of gain is needed. While having 800 MHz bandwidth, its use is not limited to high frequency signal process-ing. Its accurate, temperature-stable and supply-stable linear-in-dB scaling will be valuable wherever it is important to have a more dependable response to the control voltage than is usu-ally offered by Voltage Variable Attenuators (VVAs).
The input (INPT) and output (OUTP) of the AD8368 should be externally AC coupled to prevent disrupting the DC lev-els on the chip.
As a function of the gain voltage, the noise and distor-tion characteristics are easily predicted since the AD8368 consists of a passive variable attenuator followed by a fixed gain amplifier. The input-re-ferred noise increases in pro-portion to the attenuation level. Figure 4 shows output noise floor, as a function of VGAIN for the MODE pin pulled high. In receiver applications, the minimum NF should occur at the maximum gain where the received signal presum-ably is weak. At higher levels, a lower gain is needed, and the increased NF becomes less important.
The input-referred distortion varies in a similar manner to the noise. Figure 4 illustrates how the third-order inter-cept point at the input, OIP3, behaves as a function of VGAIN.At lower levels, a degraded OIP3 is acceptable. Overall, the dynamic range, represented by the difference between IIP3 and NF, remains reasonably con-stant as a function of gain. The output distortion and compres-sion are essentially independent of the gain. At low gains, when the input level is high, input overload can occur causing premature distortion.
The MODE pin controls whether the gain of the part is an increasing or decreasing function of the GAIN voltage. The ENBL pin is used to enable or disable the part. When ENBL is high, the part is enabled. With ENBL low, the part is disabled and draws a fraction of the nor-mal supply current.
The DECL pin should be decoupled using a large capaci-tor so that DECL acts as an AC ground. The HPFL pin is used to control the low-pass corner frequency of the out-put offset correction loop. The high pass corner frequency is inversely proportional to the HPFL bypass capacitor.
AGC OperationAs already noted, the AD8368 may be used as an AGC ampli-fier, as shown in Figure 5. For this application, the accurate internal square-law detector is employed. The output of this detector is a current that varies in polarity depending on wheth-er the rms value of the output is greater or less than its internally determined “set-point” of -11
dBm. This is 178 mV pk-pk for sine-wave signals, but the peak amplitude for other sig-nals, such as Gaussian noise, or those carrying complex modu-lation, will invariably be some-what greater. However, for all waveforms having a reasonable crest factor (less than 13dB), the rms value will be correct-ly measured and delivered at VOUT. The output set-point may be adjusted upwards using an external resistive divider net-work as depicted in Figure 5.In this configuration, the rms output voltage will be equal to (1+n)63mVrms, where n=R1/R2. For the default set-point of 63mVrms, simply short R2 (direct connection from OUTP to DETI) and remove R1. Other setpoints may be implemented by calculating the ratio of R2/R1. For example, a 0 dBm AGC output corresponds to 222 mV rms. This is 3.5 times 63 mV rms. Therefore (1+n)=3.5 and n=2.5. If we start with R1=100
then R2=250 . These val-ues can be further adjusted as needed.
The AGC mode of operation requires that the correct gain direction is chosen. Specifically, the gain must fall as VAGCincreases to restore the needed balance against the set-point. Therefore, the MODE pin must be pulled low.
Received Signal Strength Indicator (RSSI)A valuable feature of using a square-law detector is that the RSSI voltage is a true reflec-tion of signal power, and may be converted to an absolute power measurement for any given source impedance. The AD8368 may be employed as a
FEATURE ARTICLE
Figure 4: Noise and OIP3 vs. Control Voltage at 140 MHzFigure 3: Gain Control Transfer Function vs. Temperature
8
true-power meter by monitor-ing the voltage present at the DETO/GAIN interface.
Figure 7 illustrates the measured error-vector-mag-nitude (EVM) performance
for a 16-QAM modulation at 10MSymbols/sec using CDETO=1000pF. At lower symbol rates the AGC loop could start to track the peak to peak transitions due to the
modulation. At lower symbol rates it may be necessary to slow down the response of the AGC loop by increasing the value of CDETO .
ConclusionThe AD8368 is a high-per-formance voltage-controlled variable-gain amplifier/attenu-ator with a 34 dB linear-in-dB gain control range up to 800 MHz. The AD8368 operates from a supply voltage of 4.5 to 5.5 V and consumes 54 mA of current. It can be fully powered down to <1mA by grounding the ENBL pin. The AD8368 is fabricated in Analog Devices’ proprietary SiGe SOI comple-mentary bipolar IC process. It is available in a 24-pin CSP and operates over the industri-al temperature range of –40º C to 85º C. An evaluation board is available upon request.
About the AuthorPhillip Sher is a Senior Applications Engineer with the RFW group at Analog Devices in Wilmington, MA. He received a BEng in Electrical Engineering from Lakehead University in Thunder Bay, Canada and has more than 9 years of Engineering experience.
ANALOG DEVICES
FEATURE ARTICLE
Figure 5: AGC Mode of Operation
Figure 6: Output Power vs. Input Power in AGC Mode at 140MHz.
Figure 7: EVM Performance vs. Input Power in AGC Mode For 16 QAM at 10 Msym/sec
9
Amplify performance as you simplify RF designs.
Across the signal chain, analog is everywhere.
www.analog.com/rfamps-ad
Reduce the power, size, cost, and time-to-market
of your designs
The Analog Devices family of RF amplifiers continues to grow. This meansyour budget and time-to-market can shrink, thanks to the industry’s bestcombination of power and performance for your 50 MHz to 6.0 GHzwireless designs.
All of our RF amplifiers, including the new ADL5321 RF driver amplifier, arefully specified over frequency, temperature, and supply voltage to minimizethe need for extensive device characterization. Every ADI RF amplifier offersunique performance advantages such as higher linearity, lower noise, lowersupply current, internal active bias, and Class 1C ESD protection. Most offeradditional features including internal matching and dual configurations.
Across the entire RF and IF signal chain, wherever you need moreperformance, greater integration, or lower solution costs, Analog Devicesis there. For more information about all of ADI’s solutions for RF designs, call 1-800-AnalogD or visit www.analog.com/rfamps-ad.
Low Noise Amplifiers (LNA)
Part
Number
Freq Range
(MHz)
Gain
(dB)
OIP3
(dBm)
P1dB
(dBm)
NF
(dB)
Current
(mA)
Specs @
(MHz)Price
ADL55211 400 to 4000 15.3 35.3 22.5 0.82 65 1950 2.15
ADL55231 400 to 4000 17.5 33.7 21.9 1.02 65 1950 2.15
Intermediate Frequency Amplifiers (IFA)
Part
Number
Freq Range
(MHz)
Gain
(dB)
OIP3
(dBm)
P1dB
(dBm)
NF
(dB)
Current
(mA)
Specs @
(MHz)Price
ADL55301 DC to 1000 16.8 37.0 21.8 3.0 110 190 1.56
ADL5531 20 to 500 20.9 41.0 20.4 2.5 101 70 2.25
ADL5534(Dual)
20 to 500 21 40.4 20.4 2.6 90 70 3.29
Gain Blocks
Part
Number
Freq Range
(MHz)
Gain
(dB)
OIP3
(dBm)
P1dB
(dBm)
NF
(dB)
Current
(mA)
Specs @
(MHz)Price
AD83531 1 to 2700 19.5 22.8 8.3 5.6 42 900 0.48
AD83541 1 to 2700 19.5 19.3 4.8 4.4 25 900 0.48
ADL5541 50 to 6000 14.7 39.2 16.3 3.8 90 2000 1.65
ADL5542 50 to 6000 18.7 39.0 18.0 3.2 93 2000 1.65
Driver Amplifiers
Part
Number
Freq Range
(MHz)
Gain
(dB)
OIP3
(dBm)
P1dB
(dBm)
NF
(dB)
Current
(mA)
Specs @
(MHz)Price
ADL5320 400 to 2700 13.7 42.0 25.6 4.2 104 2140 2.55
ADL5321 2300 to 4000 13.8 39.7 24.9 4.1 84 2600 2.55
ADL5322 700 to 1000 19.9 45.3 27.9 5.0 320 900 3.48
ADL5323 1700 to 2400 19.5 43.5 28.0 5.0 320 2140 3.481 3 V bias is also supported.2 Includes external input match
All prices shown are $U.S. in 1k quantities.
New
RF amplifiers available in 8-lead LFCSP, 16-lead LFCSP, and SOT-89 packages—from 6 mm2 to 16 mm2—with single/dual options
10
RF Bonus Feature
FEATURE ARTICLE
Measuring the RF Power in CDMA2000 and W-CDMA High Power Amplifiers (HPAs) By Larry Hawkins, Analog Devices, Inc.
Introduction
Designers of high-pow-er amplifiers (HPAs) used in CDMA2000
and W-CDMA base stations encounter many challenges in achieving accurate trans-mit power measurements. Complications include high peak-to-average ratios, peak-to-average ratios that change with base station call load-ing, large operating tempera-ture ranges, and large trans-mit power ranges. Taking advantage of accurate RMS output power measurement allows HPA manufacturers to reduce the maximum power for which they design. This article describes several ways to accurately measure and control RMS power over tem-perature.
Complex modulation schemes like CDMA2000 and W-CDMA have large peak-to-average ratios. For a given maximum average-output-power requirement, the max-imum designed-for power requirement will generally increase (or the linearization requirements increase) as the peak-to-average value increas-es, due to base station spectral mask and EVM requirements. If the peaks of the modulated signal are clipped, third order distortion will increase, caus-ing the base station to fail its spectral mask requirements. Clipping the peaks of a modu-lated signal can also cause a loss of data, making the system fail its EVM require-ments. Designing an HPA based on peak power trans-mission requirements is expen-sive but necessary. The added expenses are due to both an increased cost in the electrical components and a decreased efficiency of the HPA. There is always a $/Watt associated with the maximum designed for power of an HPA and running a HPA well below its saturation point is inefficient. A decrease in efficiency will increase the cost of an HPA module because it increases the cost, size, and weight of the mechanical structure used to dissipate the heat, reduc-es the HPA’s reliability, and
increases its operating cost. Reducing the maximum
designed-for power of an HPA is important to HPA manufac-turers. The closer an HPA’s saturation point can get to its average power, the more efficient and cost effective the HPA will be. There are many techniques used to get
a HPA’s saturation point as close as possible to the average transmitted power, but these techniques are all limited by the system’s ability to measure the output power. The maxi-mum designed-for power of the HPA needs to be increased by the RF power measurement tolerance (including variation
vs. temperature and peak-to-average ratio) to ensure spec-tral mask and EVM compli-ance. This makes the accuracy of the RF Power Measurement System critical to reducing the cost and efficiency of an HPA.
Not only do CDMA2000 and W-CDMA modulation schemes have large peak-to-av-erage values but also the peak-to-average value changes with the call volume of a particular base station. In CDMA2000 IS-95A, for example, the for-ward link crest factor is 6.6 dB for pilot only, and 12 dB for 64 channels (using no CF reduction techniques). Large peak-to-average values cause errors in non-RMS-responding RF power detectors. A modu-lation scheme’s large peak-to-average ratio can be calibrated out in production if it stays constant but variation in the peak-to-average ratio based on the amount of users is more difficult to handle. This requires keeping track of how many users are on the system, tight control of which Walsh codes are being used, and a very large look up table in order to know the peak-to-average ratio of the signal at a particular time. A better option is the use of a RMS-responding detector. Unlike diode detectors or log amps, RMS-responding detectors are largely immune to variations in crest factors. Figure 1 shows the error of a high performance Log Amp (AD8318) compared to that of an RMS-responding detector (AD8364) as a result of the crest factor change (user loading) in the TX section of a CDMA2000 IS-95A base sta-tion. Note that the AD8318’s output changes by 3.5 dB (or 86 mV) between CW and 64 channel CDMA2000 IS-95A and 2.4 dB between Pilot only and 64 channel CDMA2000 IS-95A, while the AD8364’s output only changes by 0.1 dB (or 5 mV). A diode detec-tor behaves similarly to a Log Amp, in that its output voltage changes with respect to the crest factor of the detected sig-nal. If a Log Amp were used for power detection in this system, the 2.4 dB variation
Figure 1: The errors in a RMS-responding RF detector (AD8364) vs. a non-RMS-responding RF detector show the effect of peak-to-average ratio on power detection. While a non-RMS-responding RF detector (AD8318) exhibits significant measurement error as the peak-to-average ratio of its input signal varies, an RMS-responding RF detector (AD8364) is largely immune to changes in peak-to- average ratio.
Figure 2: The Analog Devices AD8364 output voltage and log conformance error vs. Pin (@ 450 MHz) shows very little change when temperature is cycled from –40°C to +85°C. This remains true for 30 devices taken from differ-ent production lots, even though the performance is slightly different over temperature.
11
in detected power would need to be removed through signal processing or added to the maximum designed-for power in the HPA.
The errors in a RMS-responding RF detector (AD8364) vs. a non-RMS-responding RF detector show the effect of peak-to-average ratio on power detection. While a non-RMS-responding RF detector (AD8318) exhib-its significant measurement error as the peak-to-average ratio of its input signal var-ies, an RMS-responding RF detector (AD8364) is largely immune to changes in peak-to-average ratio.
Being able to accurate-ly measure the RMS power across the operating temper-ature range of the HPA is also critical in determining the maximum designed-for power of the HPA. The accuracy of this measurement (or lack thereof) will need to be added directly to the maximum designed-for power, unless the difficult and costly process of calibration over temperature is performed. All components involved with the detection of the HPA’s output power (e.g. directional coupler, attenua-tor, etc.) can add errors over temperature, but most change very little over the HPA’s oper-ating temperature. Generally, the accuracy of measuring the HPA’s output power over tem-perature is directly related to the temperature performance of the detector. In recent years RF detection technology has made large strides in creating devices with responses that are very stable over temperature (better than ±.5dB from –40°C to +85°C). Figure 2 shows the temperature performance of the AD8364 dual RMS-responding power detector. This data was taken at +25°C (black), –40°C (blue), and +85°C (red) @ 450 MHz. It includes voltage and error over temperature (after ambient calibration) vs. input power for at least 30 devices from multiple production lots. Each part behaves slightly dif-ferently over temperature.
The Analog Devices AD8364 output voltage and log confor-mance error vs. Pin (@ 450 MHz) shows very little change when temperature is cycled from –40°C to +85°C. This remains true for 30 devices taken from different produc-tion lots, even though the per-
formance is slightly different over temperature.
It is not only essential to accurately measure an HPA’s maximum output power, but it is also necessary to measure the output power over the entire transmit power range of the HPA, though the accu-racy at lower power levels is sometimes not as critical. However, the accuracy of mea-suring over a large dynamic range is related to both the detector and the ADC resolu-tion. Figure 3 shows the out-put of two RMS-responding detectors, the AD8364 and ADL5500. The ADL5500 is linear in RMS volts to the input RF signal, and the AD8364 is linear in RMS power (dB) to the input RF signal. Based on the requirements for dynamic range and accuracy at lower power levels, the required res-olution of the ADC used with the ADL5500 could be much greater than it would be for the AD8364. System require-ments will dictate which detector/ADC will provide the most cost-effective and easiest-to-implement solution based on accuracy at lower power levels and dynamic range requirements.
A comparison of a detec-tor whose output is linear to the input RMS power in dBm (Analog Devices AD8364) to a detector whose output is linear to the input RMS volts (Analog Devices ADL5500) shows the difference in dynamic range and emphasizes the necessity of choosing an ADC with the proper resolution.
In some instances, accurate-ly controlling the power or gain of a system using an ana-log feedback loop can improve the performance and replace simple power detection. Many currently offered detectors can control power using an analog feedback loop (i.e. a detector used in Controller Mode) in addition to detecting it. If an RMS-responding detector is used in Controller Mode, power can be set very accurate-ly vs. input power, temperature, and crest factor. This power can be set very accurately, and its level can be changed using an analog voltage controlled by an ADC. Using a power detector in Controller Mode to accurately control the input or output power of an HPA would be an ideal application, as it would remove the need
FEATURE ARTICLE
Figure 3: A comparison of a detector whose output is linear to the input RMS power in dBm (Analog Devices AD8364) to a detector whose output is linear to the input RMS volts (Analog Devices ADL5500) shows the difference in dynamic range and emphasizes the necessity of choosing an ADC with the proper resolution.
Figure 4: In Controller Mode, the detector determines the power at its input and adjusts a VGA (or variable attenuator) until the detected power coincides with the level set by the power control input voltage (VSTA).
Figure 5: When one side of Analog Devices AD8364 dual RMS-responding detector is used to control the power of a sys-tem, the power at the input of the detector (and at Pout) stays constant vs. input power and temperature (less than ±.1 dB).
12
to detect the input or output power. In Controller Mode the detector determines the power at its input and adjusts a VGA (or variable attenuator) until the detected power coincides with the power set by the Power Control input voltage. Figure 4 shows a basic schematic of an RMS-responding detector (AD8364) used in Controller Mode to control the output power. Figure 5 shows the overall cir-cuit performance vs. input power and temperature when the VGA is controlled by one side of an AD8364 (dual RMS-responding log detector). Note that an HPA can be put between the VGA and coupler as long as the power level is set correctly into the AD8364, and that any VGA (or variable attenuator) can be used if the control voltages are set properly between it and the AD8364 (an op-amp may be needed to invert and/or level shift the control voltage). If the control levels between the detector and VGA are set properly and power levels are properly designed, the power control range/usable input power range will be close to the detectable power range of the detector
(60 dB, in the case of the AD8364). In Controller Mode, the detector deter-
mines the power at its input and adjusts a VGA (or variable attenuator) until the detected power coincides with the level set by the power control input voltage (VSTA).
When one side of Analog Devices AD8364 dual RMS-responding detector is used to control the power of a system, the power at the input of the detector (and at Pout) stays constant vs. input power and temperature (less than ±.1 dB).
A dual RMS-responding detector oper-ating in Controller Mode can also be used to control the gain of an HPA very accurately vs. input power, temperature, and crest factor. If the gain of an HPA module is controlled with enough accuracy over input power, temperature, and crest factor, the HPA module’s output power would not have to be reported but would be directly related to the power feeding it. If both inputs of a dual detector are put in Controller Mode, the detector determines the power at each input and adjusts the gain of a VGA until the power detected
on one input is equal to the power on the other. Figure 6 shows a basic schematic of the AD8364 (dual RMS detector) used to control the gain of a system. Figure7 shows the performance of this setup. Everything that needs to be accurately controlled should be included between the two couplers. Note that a VGA, variable attenuator, or even the bias of the HPA can be used to control the gain. If the control levels between the detector and VGA are set properly and power levels are properly designed for, the usable input power range will be close to the detectable power range of the detector (60 dB, in the case of the AD8364).
When both inputs of a dual detector are used in Controller Mode, the detector will control a VGA (or VVA, etc.) in such a way as to equalize the power it detects at both RF inputs. The gain of the system will be determined by the couplers and attenuators used to set the power being detected by the dual detector.
When both inputs of Analog Devices dual RMS detector (AD8364) are put in Controller Mode, the gain is controlled to better than ±.15 dB vs. temperature and input power, with a dynamic range almost equal to the dynamic range of the RMS detector.
Many of the challenges associated with RF power detection for HPAs used in CDMA2000 and W-CDMA systems can be solved using RMS-responding RF detectors. Variations in detected power due to large peak-to-average values that change with base station loading, large operating temperature ranges, and large transmit power ranges are now manage-able. New ways to control power and gain accurately enough to remove the need for detect power are now available. All of these things allow HPA manufac-turers to reduce the cost and improve the reliability of their HPAs.
ANALOG DEVICES
FEATURE ARTICLE
Figure 6: When both inputs of a dual detector are used in Controller Mode, the detector will control a VGA (or VVA, etc.) in such a way as to equalize the power it detects at both RF inputs. The gain of the system will be determined by the couplers and attenuators used to set the power being detected by the dual detector.
Figure 7: When both inputs of Analog Devices dual RMS detector (AD8364) are put in Controller Mode, the gain is controlled to better than ±.15 dB vs. tempera-ture and input power, with a dynamic range almost equal to the dynamic range of the RMS detector.
13
Part
Number
–3 dB
BW
(MHz)
Gain
Low
End
(dB)
Gain
High
End
(dB)
Number
of
Channels
Noise
Figure
(dB)
Supply
Current
@ 5 V
(mA)
Gain
Accuracy
(dB)
OIP3
(dBm)Package
Price
@ 1K
($U.S.)
AD8376 690 –4 20 2 8.7 260 0.2 50 LFCSP 6.49
AD8372 130 –9 32 2 7.8 212 0.2 35 LFCSP 6.50
AD8375 690 –4 20 1 8 130 0.2 50 LFCSP 4.49
AD8370 700 –25 34 1 7 78 – 35 TSSOP 4.20
AD8369 600 –5 40 1 7 37 0.5 19.5 TSSOP 4.20
www.analog.com/rf-vga
Industry’s first digitally controlled dual VGAs:
2 the functionality; 1/3 the space.
ADL5521 and ADL5523 Low Noise Amplifiers
400 MHz to 4 GHz low noise amps with the optimumamount of gain and current consumption.
ADF4360-9 Integrated PLL and VCO with
Programmable Output Divider
Built-in VCO saves space and cost; 30 mA typicalcurrent consumption.
ADL5541 and ADL5542 Broadband RF Gain Blocks
Operate from low frequencies up to 6 GHz; 50 internalmatching and bias circuitry reduces external components.
ADL5350 Low Frequency to 4 GHz High Linearity Mixer
Broadband RF, IF, and LO inputs allow it to be specified inboth receiver and transmitter signal paths.
AD6655 IF Diversity Receiver
Mixed-signal IF receiver comprising dual 14-bit, 80 MSPSto 150 MSPS ADCs, and a wideband downconverter.
ADI covers the entire RF signal chainAnalog Devices is the only vendor that offers acomplete portfolio of RF ICs. Optimize performanceand simplify your designs with:
Precise gain control, high IP3, low noise—for a wide
variety of receiver applications
Why use two individual VGAs when the Analog Devices AD8376 digital VGA
gives you dual channels in 67% less space? Analog Devices’ industry-
leading solution offers 50 dBm output IP3 on just 130 mA of current per
channel, providing unparalleled linearity and minimizing distortion. You
can maximize signal level prior to the ADC across a broad choice of IFs
up to 400 MHz, while reducing your footprint with the AD8376’s compact,
5 mm 5 mm 32-lead LFCSP package. For applications requiring
additional gain control range, such as upstream CMTS receivers, ADI’s
AD8372 VGA offers 41 dB of range in 1 dB steps.
For more information about Analog Devices RF VGAs, please call
800-AnalogD or visit www.analog.com/rf-vga.
Dual- and Single-Channel Digital VGAs
14
RF Bonus Feature
23S e m i c o n d u c t o r H i g h l i g h t
www.ECNmag.com • 09/2007
Many different standards forwireless communicationsequipment are in use today.
Narrowband communication standardsuse stronger transmission in a smallslice of bandwidth. Wideband stan-dards use lower transmission poweracross a larger bandwidth. Each stan-dard defines minimum performancecharacteristics for receivers, andincludes specifications such as band-width, maximum signal level, and sensitivity.GSM is one narrowband example; the channelbandwidth is 200 kHz. A GSM receiver musthave a minimum sensitivity of –104 dBm andbe able to tolerate a –13 dBm signal at theantenna. In contrast, CDMA2000 is a wide-band standard that uses a 1.25 MHz band-width. CDMA2000 receivers need to have aminimum sensitivity of –117 dBm/1.25 MHzand tolerate a maximum signal of –30 dBm at900 kHz offset.1
Wideband and narrowband communicationstandards can overlap in the crowded RF spec-trum, making it even more challenging todesign receivers that are immune to interfer-ence. Narrowband receivers can use narrowbandpass filters to remove interference, allow-ing them to amplify and then digitize only thechannel of interest. Unfortunately, designers ofwideband and multicarrier receivers cannot usethis solution, because channel selection is per-formed after the signal is digitized. Therefore,the entire spectrum must be digitized as cleanlyas possible to allow the digital signal processor(DSP) to remove the blockers in the digitalrealm. Wideband receivers must tolerate largenarrowband signals within the band of interestwithout degrading the sensitivity of the receiver.Although they cannot use filters to remove in-band interference, wideband and multicarrierreceivers will use Nyquist filters to remove out-of-band interference.
In the block diagram of a receiver (Figure 1),the analog portion of the receive signal chainconsists of:
• Antenna. Selective to the spectrum ofinterest — this provides some attenuation forout-of-band interference.
• Down-conversion. Translates the receivedsignal to a lower frequency that can be ampli-fied and converted to digital information. Thiscan be accomplished with one or two mixingstages. A single mixer stage can reduce the sig-nal’s center frequency from a few gigahertz to acouple hundred megahertz. A second mixerstage can translate the signal down to tens ofmegahertz. Using a single mixer stage savesboard space and the expense of the secondPLL, VCO, and filter. However, the higher IFrequires an amplifier and analog-to-digitalconverter (ADC) with very good performanceat high frequencies.
• VGA. A variable-gain amplifier is used toadjust the gain of the circuit depending on thereceived signal strength.
• Anti-aliasing filter. Attenuates signals out-side the band of interest.
Large-scale blockers interfere with convert-ing the desired signal in two ways:
1. An ADC’s input range is specified as someamplitude, typically 1V or 2V peak-to-peak orin terms of dBm. This input range limits howmuch gain the VGA can apply to the signalbefore the ADC starts to distort or clip. An rmsdetector or log-amp is frequently used todetermine the amplitude of the composite sig-nal, thus determining how much gain the VGAshould apply in order to use the ADC’s full
input range. Large-scale blockersincrease the amplitude of the receivedsignal, limiting the amount of gain thatcan be applied before saturating theinput of the ADC. In-band blockerseffectively limit the amplitude of thedesired signal. If no blocker was pres-ent more gain could be applied to thesignal to help overcome the noise floorof the ADC.
2) Nonlinearities within the ADCwill create intermodulation distortion productswithin the desired signal band, thus negativelyimpacting spurious free dynamic range (SFDR)performance. Intermodulation distortion occurswhen two frequencies mix together and createsignals at the sum and difference of the input
frequencies. For example, applying two tones tothe input of an ADC where F1=50 MHz andF2=52 MHz, will create new signals at 2 MHz(F1–F2), 102 MHz (F1+F2), 54 MHz (2F2–F1),48 MHz (2F1–F2), and so on. The amplitude ofthese new signals depends on the linearity of theADC and the amplitude of the input signals.The same phenomenon happens when a largein-band blocker mixes with another blocker orthe desired signal. New signals, which can bevery close to the signal of interest, are created.
Simulating the Effect of Blockerson Data Converter Performancein Wideband Receivers
ADCs
by Michael Sink and Pamela Aparo,Analog Devices
Figure 1. Receiver block diagram.
Figure 2. ADC performance comparison between CDMA2000 input andCDMA2000 input with blocker.
15
Knowing these blockers can cause problems inreceiver sensitivity, how does the engineer knowthat a design will meet the required specifications?There are a couple ways to do this (aside from trialand error). The ADC’s signal to noise ratio (SNR)and SFDR performance can be used to calculatethe effect of in-band blockers on a receiver. TheADC performance will be specified for differentsample rates and input frequencies, which maynot match your application. These specificationsare measured with pure sine wave inputs asopposed to the signals received from an antenna.Alternatively, a software package such as AnalogDevices’ (ADI) VisualAnalog can simulate ADCperformance with a real-world waveform and anysample rate.
The tool’s drag-and-drop GUI allows circuitcreation from functional blocks. It can createcomplex waveforms, import waveform data, andcustomize FFT analysis to quickly determine if aspecific ADC will meet the performance specs fora given communication standard. ADI’s ADCmodels account for the effects of clock jitter, tran-sitions between stages inside the part, the drooprate of the sample and hold capacitors of theADC input, and ultimately the SNR and SFDRperformance over frequency and sample rate.Using this tool, ADC performance can be simu-lated with any input signal — including a com-plex waveform like CDMA2000.
Figure 2 shows the results of using a
VisualAnalog canvas to evaluate the AD9246, a14-bit 125 Msps ADC. By creating a signal similarto a CDMA2000 waveform with and without anin-band blocker we can use the tool to analyze the ADC’s performance. This same canvas file(Figure 3) can be easily modified to accept anoth-er waveform (W-CDMA, GSM900) or evaluate adifferent ADC using another ADC model.
ConclusionWideband communication standards such asCDMA2000 require receivers that can toleratelarge in-band blockers. Although it is possible to
calculate the effect of the ADC’s SNR and SFDRon receiver sensitivity, the datasheet may notspecify ADC performance at the desired samplerate and input frequency. New simulation soft-ware makes it possible to quickly and easily eval-uate different ADCs with real-world signals, ulti-mately reducing design risk.
References1 AN-808 “Multicarrier CDMA2000 Feasibility”, BradBrannon and Bill Schofield. ©2006
“Correlating high-speed ADC performance to multi-carrier 3G requirements”, Brad Brannon, RF Design,June 2003, pp. 22-28.
Pamela Aparo is a marketing manager in theHigh Speed Converters Group at Analog Devices,Inc. Pam has worked in marketing and applicationsroles with ADI for seven years. Prior to ADI, sheapplied her engineering degree to medical applica-tions working for Baystate Medical Center and OECMedical Systems. Pam earned a BSEE from theUniversity of Connecticut. Michael Sink is an appli-cations engineer in the High Speed Converters Groupat Analog Devices. He has been with ADI for fiveyears. Michael earned a BSEE from North CarolinaState University. For more information, contactAnalog Devices, One Technology Way, Box 9106,Norwood, MA 02062-9106; (800) 262-5643;www.analog.com.
24 S e m i c o n d u c t o r H i g h l i g h t ADCs
Figure 3. VisualAnalog canvas used to analyze the effect of a blockeron ADC performance.
As seen in the September 2007 issue of ECN magazine
16
17
18
19
20
PartNumber
RF Frequency (MHz)
DynamicRange (dB)
OutputResponse
ResponseTime
Temp Stability (dB)
SupplyVoltage (V)
SupplyCurrent (mA)
Package Comments
AD8361 100 to 2500 30 Linear-in-volts 320 ns 0.25 2.7 to 5.5 1 2.9 mm 2.8 mm, 6-lead SOT-23
Low power, low cost rms detector
AD8362 >0 to 3800 60 Linear-in-dB 45 ns 1.0 4.5 to 5.5 20 5 mm 6.4 mm, 16-lead TSSOP
Wireless infrastructure
AD8363 >0 to 6000 >50 Linear-in-dB 8 s 0.5 4.5 to 5.5 66 4 mm 4 mm, 16-lead LFCSP
Broadband
AD8364 >0 to 2700 60 Linear-in-dB 45 ns 0.5 4.5 to 5.5 70 5 mm 5 mm, 32-lead LFCSP
Dual-channel rms detector
AD45102 50 to 3800 60 Linear-in-dB 45 ns 1.0 4.5 to 5.5 20 5 mm 6.4 mm, 16-lead TSSOP
Operation to 125°C
ADL5501 50 to 4000 30 Linear-in-volts 6 s 0.25 2.7 to 5.5 1.1 2 mm 2 mm, 6-lead SC70
Reduced size, improved temperature stability
TruPwr™ RMS Power Detectors
Features
Industry’s first patented RF true power detectors
Innovative line of TruPwr rms responding detectors with frequency ranges from >dc to 6 GHz
Immune to crest factor changes enabling modulation-independent true power detection
Single- and dual-channel rms detectors with linear-in-volts or linear-in-dB outputs
RF Bonus Feature
21
22
23
24
to diffract specific wavelengths oflight from the incoming spectrum.
DDS devices are excellent RF driversfor exciting crystals within an AOTF,due to the high resolutions attain-able. Precise frequency switching im-plemented by the DDS involves verylittle latency and no settling time.
An on-chip clock multiplier offersphase noise performance sufficientfor AOTF applications, and easy syn-chronization of additional multi-channel chips is possible. Operatingat a maximum system rate of 500Msamples/s, a four-channel DDS de-vice produces output frequenciesfrom dc to 250 MHz in steps of about116 mHz, with each DDS channelconsuming less than 165 mW.
Military systems applicationMilitary communications, electroniccountermeasures, and communica-
tions surveillance applications de-mand exceedingly fast switchingspeeds. Traditionally, the challengehas been to achieve this withoutcompromising frequency resolution.
The AD9910, from Analog Devices,is a 1-Gsample/s DDS with an inte-grated 14-bit DAC. 32-bit frequencytuning translates to 0.23-Hz tuningresolution.
The incredibly fast parallel dataport, used for writing frequency tun-
C O M M U N I C AT I O N S S P E C I A LC O M M U N I C AT I O N S S P E C I A L
D irect digital synthesis (DDS)has come a long way sincethe days when implementa-
tions were limited to complex, cost-ly designs that consumed largeamounts of power. Today’s DDS ICsoffer drastically reduced power dissi-pation and size—as well as greatlyimproved phase noise, spurious per-formance, and ease of use—makingthem attractive alternatives to ana-log synthesis in many communica-tions systems. DDS devices now of-fer many levels of integration,performance, packaging, and costbenefits, allowing customized andeasy selection into new designs.
The primary advantages of using adigital synthesizer over traditionalanalog solutions are fast frequencyhopping, extremely fast switchingspeeds with virtually no settlingtime, finer tuning resolution, andchannel flexibility. Designers requir-ing agile frequency sources with ex-cellent phase noise and low spuriousperformance often choose DDS.
Available DDS ICs now support in-ternal speeds of 1 Gsample/s, pro-vide outputs up to 400 MHz, and in-clude a high-performance 10-, 12-,or 14-bit DAC. With 48-bit frequen-cy tuning, 14-bit phase tuning, and10-bit amplitude scaling, extremelyfine resolution (1 μHz, 0.022°, and 3mV) is attainable.
Once viewed as power-hungrymonsters, some DDS cores nowboast power consumption less than165 mW at 400 Msamples/s. A phasenoise floor below –150 dBc/Hz at upto 100 MHz is possible, and im-provements in wideband SFDR (>55dBc) and narrowband SFDR (80 to
90 dBc) make DDS a more attractivesolution.
I/Q generationFor applications requiring in-phaseand quadrature (I and Q) generation,DDS offers unparalleled matching ofI and Q outputs. By using DDS tosend quadrature signals to the I/Qinputs of an analog modulator, theredundant sideband can be attenuat-ed significantly, relaxing the nor-mally tight filtering requirements.
To further simplify this approach,new two-channel DDS chips provideinherent synchronization acrosschannels while maintaining excellentisolation between them. Independentfrequency, phase, and amplitude con-trol on each channel provides flexibil-ity to correct imbalances betweenquadrature signals. In Fig. 1, two syn-chronous DDS channels generate digi-tal I and Qdata which isthen upcon-verted using amodulator.
OpticalsystemsWith channelbandwidth ata premium inoptical net-working sys-tems, packingmore functionality into a smallerspace is increasingly desirable and asingle-chip, four-channel DDS IC of-fers clear benefits.
Acousto-optics, a technique oftenapplied in optical communicationsystems, uses the relationship be-tween vibrations and light to targetan optical wavelength. During opti-cal transmission, devices such as theacousto-optic tunable filter (AOTF)use a slight adjustment in frequency
Today’s DDS ICs offer drastically reduced power dissipation and size
Improved DDS devices enableadvanced comm systems
69ELECTRONIC PRODUCTS http://electronicproducts.com SEPTEMBER 2006
2-ChannelDDS
AD9958
CH0
AD8349ADL537x
0º/90º +
Data(I)
Data(Q)
Modular
RF out
CH1 LPF
LPF
LO
BY VALOREE YOUNGAnalog DevicesGreensboro, NChttp://www.analog.com
Fig. 1. A two-channel DDS provides precise, synchronous control for in-phase and quadrature data generation.
25
ing word changes, allows the DDS ICto take a 16-bit word every 4 ns. Thedevice offers a 14-bit phase offsetword, a 1,024-element RAM, a linear-sweep block, and other features.
In systems needing to synthesizespecific frequencies with very lowspurious components, a recentlydeveloped implementation usesauxiliary DDS channels to reducethe magnitude of harmonic spurs inthe output spectrum. This spur re-duction approach extends the im-plementation to designs demand-ing better wideband SFDR (spuriousfree dynamic range) performancethan was previously attainable by aDDS.
In fact, a 12-dB improvement of asingle spur can be seen at 200 MHz.The technique is beneficial to cur-rent DDS users in reducing the tire-some task of frequency planningand in eliminating the need for acomplex filter design.
The AD9912 showcases this spurreducing concept, and includes anon-chip comparator for generating asquare wave output for clock genera-tion applications. Extremely finetuning resolution of 3.5 μHz on out-puts up to 400 MHz is made possibleby a 48-bit frequency tuning word.
Cellular base station benefitsIn base station transmit architecturesthat employ a direct-IF approach toupconversion, DDS-based modula-tors offer an alternative to traditionaldual-DAC direct-conversion solu-tions. A quadrature digital upcon-verter (QDUC) integrates a 1-Gsam-
ple/s DDS core, a highperformance 14-bitDAC, clock multipliercircuitry, digital filters,and other DSP func-tions onto a singlechip.
This solution offerscost, power, and spacesavings, while provid-ing excellent dynamicperformance. Figure 2shows architecture op-tions for single analogupconversion usingtwo different ap-proaches.
Implementing aQDUC in the trans-mission path avoids
quadrature phase and amplitude im-balance issues. In addition, only onemixer and one DAC are required, sothe overall noise will be reduced.
The QDUC supports real outputs upto 400 MHz. A 250-MHz 18-bit-widedata input port supports interleaveddata rates as high as 125 Msymbols/s,and innovations in low-power DDScores have brought power down toabout 1 W at 1 Gsample/s.
Portable devicesDesigners of portable and handheldwireless systems should note thedrastic reduction in power consump-tion. DDS ICs consuming less than50 mW are typically offered at fre-quencies of 50 MHz or lower.
Some higher-speed DDS devicesrunning at up to 250 Msamples/s arenow approaching 50-mW power lev-els. These cores are even being usedas auxiliary components, performingtasks such as multitone generationand spur reduction within highly in-tegrated complete DDS devices.
Other communication appsClock generation applications alsoreap the benefits of multipurposeDDS functionality, since the abilityto generate a clean, stable clock is vi-tal in many systems.
As advances in SFDR performanceand power consumption continue,DDS will expand into markets thatwere previously off limits. Its roleas a standalone function willevolve to recognize its potential asa building block in other morecomplex chips. �
COMMUNICATIONS SPECIAL - IMPROVED DDS DEVICES ENABLE ADVANCED COMM SYSTEMS
70 ELECTRONIC PRODUCTS http://electronicproducts.com SEPTEMBER 2006
DSPD/A
DUCfor
multiplecarriers
Antenna
DigitalFiltering
D/A
AmplifierComponents
ClockDistribution PLL/VCO
AD8349ADL537xAD9779
0/90
I
Q
+
DSP
Antenna
D/A AmplifierComponents
ClockDistribution PLL/VCO
AD9857AD9957
0/90
I
Q
+DEMUX Power
RampProfile
Fig. 2. Upconversion schemes can use a dual TxDAC(a) or DDS-based modulator (b).
(a)
(b)
26
RF/MICROWAVE SPECIALRF/MICROWAVE SPECIAL
The term “short-range device” (SRD) refers to a device capable of wireless communications
over a relatively short distance—from just a few centimeters up to a few kilometers. Several wireless stan-dards defining the communications processes for such devices exist, with new standards continuing to evolve.
ConstraintsGovernments impose restrictions on the use of the frequency spectrum. Figure 1 shows the UHF (300-MHz to 3-GHz) ISM unlicensed frequency bands available in different parts of the world.
As the figure shows, no common unlicensed ISM band is available be-low 2.4 GHz, although some RF transceivers will support operation across several of these sub-1-GHz bands. The Analog Devices ADF7020/-1, for example, supports operation from 135 to 950 MHz. Many designs use proprietary com-munications protocols in this fre-quency range.
Despite the inherent range advan-tage of the lower-frequency bands, the global nature of the 2.4-GHz band makes it attractive for many SRD communications protocols such as Bluetooth, WLAN and ZigBee. Ir-respective of the communications protocol used, countries apply addi-tional constraints driven by factors such as safety and quality of perfor-mance; these constraints are required to limit interference between differ-ent equipment. Some examples rele-vant for the 2.4-GHz band are shown in Table 1.
Existing standardsThe various communications proto-cols each offer advantages and disad-vantages, with the optimum choice
BY MARY O’KEEFFEAnalog Devices, Norwood, MAhttp://www.analog.com
New standards provide optimized solutions for the differing needs and priorities of different applications
RF standards for short-range
ELECTRONIC PRODUCTS http://electronicproducts.com NOVEMBER 200727
RF STANDARDS FOR SHORT-RANGE WIRELESS CONNECTIVITY
depending on the application. Blue-tooth, for example, offers data rates up to 3 Mbits/s, whereas 802.11g en-ables data rates as high as 54 Mbits/s and ZigBee is limited to 250 kbits/s.
While 802.11g has the higher data rate, it also has higher cost and high-er power consumption. ZigBee has the advantage of low power con-sumption. It can also support a high number of nodes. For example, Blue-tooth’s maximum of 8 nodes in a net could be a limiting factor in an in-dustrial application.
For sensor applications, where only a limited amount of data must
be transferred and low power con-sumption is of significant value, Zig-Bee appears to have an advantage over Bluetooth or WLAN. For appli-cations such as wireless headsets, the data rate offered by Bluetooth meets the requirements while maintaining a relatively low cost.
The Wibree standard currently in development can operate in a stand-alone mode or as a complement to Bluetooth, offering a lower-power so-lution than Bluetooth and a maxi-mum data rate of 1 Mbit/s. This is a higher data rate than ZigBee, but its range would be shorter than a low-
power ZigBee device. Other factors that differentiate the various stan-dards include latency and resilience.
Developing standardsThe industrial environment is one where resilience is of particular value. The recently ratified Wireless Hart standard is targeted at the industrial space. The SP100 group is also look-ing at a standard for industrial appli-cations.
Both the Wireless Hart standard and the ISA-SP100.11a standard, which is in development, indicate the use of an 802.15.4-compatible ra-
dio. The 802.15.4 radio also consti-tutes the physical layer for the Zig-Bee standard.
While several short-range device standards are already in existence, new standards are continuing to evolve. These are driven by and tar-geted at providing optimized solu-tions for the differing needs and pri-orities of different applications.
Table 1. Worldwide communication standardsRegion Standard Relevant Link
Europe ETSI EN 300 328ETSI EN 300 440
http://www.etsi.org/WebSite/homepage.aspx
USA FCC CFR47 part 15 http://www.access.gpo.gov/nara/cfr/waisidx_04/47cfr15_04.html
Japan ARIB STD-T66 http://www.arib.or.jp/english/
Fig. 1. Several UHF (300-MHz to 3-GHz) ISM unlicensed frequency bands are available in different parts of the world.
For more on RF ICs and RF standards, visit
www.electronicproducts.com/linear.asp
wireless connectivity
28
A BROADBANDI/Q MODULATORFOR BROADBANDRADIO DESIGNS
Modern digital radio transmitter designposes increasing challenges forequipment designers. The ongoing
trend towards increased throughput of data isincreasing the modulation density and carrierbandwidths of transmitted signals. Peak-to-av-erage ratios increase with higher order modu-lation schemes and to maintain good adjacentchannel power ratio (ACPR) while transmit-ting the same rms power level, componentswith lower intermodulation distortion andlower noise must be used.
Baseband, IF and RF bandwidth must beflat across the channel to maintain the spectral
shape of the modulated carrier.Furthermore, if digital pre-dis-tortion techniques are beingused then the higher order har-monics need to be passedthrough the baseband inputsand gain flatness must be main-tained up to the higher orderharmonics of the baseband sig-nal. When a radio transmitterdesign calls for operation over avery wide range of RF frequen-cies the RF gain flatness of theoverall signal chain becomes
critical. Minimizing gain variations in the sig-nal chain over frequency eases the burden ofsignal chain planning and budgeting. This arti-cle focuses on I/Q modulators, which are acritical component in modern transmitters.
A BROADBAND I/Q MODULATORI/Q modulators perform the frequency
translation that mixes the baseband signal tothe desired location in the RF spectrum. AnI/Q modulator consists of a local oscillator(LO) input that is split into in-phase (I) andquadrature (Q) components that are separatedby 90°. These two signals drive separate mixersthat are also driven by I and Q baseband sig-nals. The outputs from both mixers are thensummed to provide a modulated carrier eitherat RF or IF. The ADL5385 contains these ba-sic blocks (see Figure 1) and is able toachieve a wide tuning range that spans five oc-taves (50 MHz to 2.2 GHz) through the use ofan active divide-by-two LO splitter instead ofthe more traditional passive polyphase filter.Its wideband performance can be seen in
ANALOG DEVICES INC.Norwood, MA
IBBN
LOIP
LOIN
QBBP
QBBN
TEMP
Divide-by-2Quadrature
PhaseSplitter
VOUT
IBBP
ENBL
TemperatureSensorBIAS
Fig. 1 The ADL5385 I/Qmodulator’s basic blockdiagram. ▼
Reprinted with permission of MICROWAVE JOURNAL® from the September 2006 issue.©2006 Horizon House Publications, Inc.
29
Figure 2, where the output powerhas a very flat response over the en-tire output frequency range, with a 1dB bandwidth of 1300 MHz. Thisnew modulator is designed to directlydrive a 50 Ω load and also includes anintegrated temperature sensor.
GAUGING SIGNAL QUALITYUSING ERROR VECTORMAGNITUDE
Error vector magnitude (EVM) is ameasure of the quality of modulationof a signal and it is directly affected bythe quadrature and amplitude errorswithin the modulator. The amount ofquadrature and amplitude errors can
be gauged by observing the level ofsideband suppression in a single-side-band spectrum. Figure 2 shows thatthe native uncompensated sidebandsuppression of the ADL5385 I/Q mod-ulator is better than –38 dBc out to900 MHz. This level of sideband sup-pression typically yields EVM perfor-mance that is more than acceptable formost communication standards. Ifhigher performance is required, side-band suppression can be optimized byadjusting the relative gain and phase ofthe baseband signals.
The 64QAM constellation, eye-di-agram and spectrum, shown in Fig-ure 3, was generated using randomdata at 5.056941 MSym/s with a filteralpha of 0.18. This closely mimics thedata rate and modulation for a typicalcable modem head end application.It can be seen that the EVM for thissignal is 0.33 percent rms with aquadrature error of 0.27° and a gainerror of 0.003 dB.
SIGNAL QUALITY VS. POWER LEVEL
Figure 4 shows how ACPR varieswith output power for the same64QAM modulated carrier. The highOIP3 of the ADL5385 I/Q modulatorenables it to achieve high outputpower levels with minimal adjacentchannel leakage. This results in lessgain required in the subsequentstages of the radio.
The displayed performance wasobtained without digital compensa-tion of the baseband data. This,along with the wide RF tuning
range, allows the modulator to beused without factory calibration.This can significantly reduce thetime and effort required for designand manufacturing.
DIVIDE-BY-TWO SPLITTERENABLES BROADBANDOPERATION
Systems such as cable modem headend equipment must be able to dy-namically place carriers anywhere inthe the 40 to 900 MHz range. Tradi-tional modulators that use a passive re-sistor-capacitor polyphase network tosplit the LO into quadrature compo-nents have generally been unable tospan such a wide frequency range.This is because the resistor-capacitornetworks are tuned for a particularcenter frequency and typically have auseful range of just over two octaves.Traditional cable modem headendequipment designs use a two-stage up-conversion. The baseband signal is up-converted using an I/Q modulator toan IF frequency above the cable band,typically around 1100 MHz. This IFsignal is then mixed down into the ca-ble band using a mixer. These solutionsrequire more components and thecomplexity associated with such de-
PRODUCT FEATUREO
utpu
t Po
wer
and
Car
rier
Fee
dthr
ough
(dB
m)
Und
esir
ed S
ideb
and
Supr
essi
on (
dBc)
Output Frequency (MHz)
10
0
−10
−20
−30
−40
−50
−600 400 800 1200 1600 2000
−40°C Output Power+25°C Output Power+85°C Output Power−40°C Carrier Leakage+25°C Carrier Leakage+85°C Carrier Leakage−40°C Undesired Sideband Suppression+25°C Undesired Sideband Suppression+85°C Undesired Sideband Suppression
▲ Fig. 2 Single-sideband performance vs.output frequency from –40° to +85°C.
LO
I
I
I
Q
D
D
▲ Fig. 5 The ADL5385 modulator’s divide-by-two phase splitter showing theapplied LO at twice that of the desired LOfrequency in the mixer.
AC
PR
(dB
c)
CHANNEL POWER (dBm)
−55
−60
−65
−70
−75
−80−18 −14 −10 −6 −2 2
▲ Fig. 4 64QAM ACPR vs. output power(symbol rate = 5.056941 MSym/s with a filteralpha of 0.18, adjacent channel bandwidth =5.25 MHz).
▲ Fig. 3 The spectrum, constellation and eye-diagram of a 64QAM carrier at 350 MHz.
30
signs increases the design time and ef-fort. System cost and complexity canclearly be reduced if this signal chaincould be simplified to a single-stage di-rect launch architecture.
The ADL5385 overcomes the two-octave limitation of traditional I/Qmodulators by utilizing a divide-by-two LO splitter. This architecture isillustrated in Figure 5, where two D-flip-flops are clocked by an LO signal
and its inversion. In the ADL5385the inversion is achieved by crossingthe polarities of the inputs on one ofthe differential D-flip-flops. The Iand Q signals that drive the mixercores shown in the ADL5385 blockdiagram are generated through thealternate clocking of the D-flip-flopsby the two LO input signals. Close in-spection of the timing diagram on theleft of the figure will reveal that it isimperative that the applied LO signalbe at twice the desired RF outputfrequency and that the duty-cycle ofthat LO signal be exactly 50 percent.Any deviation from 50 percent willdegrade the 90° split and this will inturn degrade sideband suppression.
WIDE BASEBAND BANDWIDTHINCREASES DATA CAPACITY
In single-channel modulation sys-tems, data capacity can be increasedby either using a higher order modu-lation scheme or by using more band-width. Figure 6 shows the normal-ized baseband frequency response ofthe ADL5385. With wider carrierbandwidths, the challenge is to main-tain a flat gain across the bandwidthof the carrier. This ensures that thespectrum is not distorted by gain rip-ple. If the gain ripple is too greatthen precompensation might be re-quired in the digital domain. Thisprocess will require the characteriza-tion of the frequency response ofevery radio and will increase thecomplexity of the design and drive upthe cost to manufacture the radio.
The ADL5385 offers a 0.1 dB base-band gain flatness out to 85 MHz.This means that for most applica-tions, there should be no need to per-form any sort of precompensation.
A SEAMLESS INTERFACE TO BASEBAND I/Q DACS
The ADL5385 is designed to in-terface seamlessly with Analog De-vices’ family of transmit digital-to-analog converters (TxDAC). The in-terface between the two devicestypically involves six resistors and asimple LC filter (see Figure 7). Thefour 50 Ω resistors shunting toground from each of the DAC out-puts provide the 500 mV DC bias forthe ADL5385 baseband inputs whilethe 100 Ω resistor in shunt betweeneach differential pair sets the ACswing of the baseband inputs. Withthis simple interface the need for sin-gle-ended-to-differential or level-shifting amplifiers is eliminated.
PACKAGE, AVAILABILITY,EVALUATION BOARDS
The ADL5385 is packaged in aRoHS-compliant 24-lead LFCSPwith exposed paddle. Samples andevaluation boards are currently avail-able and may be ordered on the com-pany’s web site.
Analog Devices Inc., Norwood, MA (781) 329-4700,www.analog.com.
PRODUCT FEATUREN
OR
MA
LIZE
D B
ASE
BA
ND
FREQ
UEN
CY
RES
PO
NSE
(dB
) 2
1
0
−1
−2
−3
−4
−56 126 246 366 486 606 726 846
BASEBAND FREQUENCY (MHz)
▲ Fig. 6 The ADL5385 baseband section’snormalized frequency response.
AD9779 ADL5385
IBBP
IBBN
QBBN
QBBP
OptionalLow Pass
Filter
OptionalLow Pass
Filter
IOUTA1
IOUTA2
IOUTB1
IOUTB2
▲ Fig. 7 AD9779 and ADL5385 interfaceschematic.
31
32
Part Number
RF Frequency (MHz)
Gain(dB)
OutputP1dB (dBm)
OutputIP3 (dBm)
NoiseFigure (dB)
SupplyVoltage (V)
SupplyCurrent (mA)
Specs@ (MHz)
Package Comments
AD83531 1 to 2700 19.5 9 22.8 5.6 2.7 to 5.5 42 900 2 mm 3 mm, 8-lead LFCSP Gain block
AD83541 1 to 2700 20 5 19.3 4 2.7 to 5.5 23 900 2 mm 3 mm, 8-lead LFCSP Gain block
ADL5320 400 to 2700 17 25.4 45 4 4.75 to 5.25 104 880 4.25 mm 4.6 mm, 3-lead SOT-89
Driver
ADL5321 2300 to 4000 14 25 40 4 4.75 to 5.25 84 2600 4.25 mm 4.6 mm, 3-lead SOT-89
Driver
ADL5322 700 to 1000 20 28 45 5 4.75 to 5.25 320 880 3 mm 3 mm, 8-lead LFCSP Matched driver
ADL5323 1700 to 2400 19.5 28 43.5 5 4.75 to 5.25 320 2140 3 mm 3 mm, 8-lead LFCSP Matched driver
ADL55211 400 to 4000 15 22.5 35 0.82 3.0 to 5.25 65 1950 3 mm 3 mm, 8-lead LFCSP Single LNA
ADL55231 400 to 4000 17.5 22 36 12 3.0 to 5.25 65 1950 3 mm 3 mm, 8-lead LFCSP Single LNA
ADL55301 DC to 1000 17 22 37 3 3.0 to 5.5 110 190 3 mm 2 mm, 8-lead LFCSP Matched IF amplifier
ADL5531 20 to 500 21 20.4 41 2.5 4.75 to 5.25 100 70 3 mm 3 mm, 8-lead LFCSP Matched IF amplifier
ADL5534 20 to 500 19.4 20 40 2.7 4.75 to 5.25 180 190 5 mm 5 mm, 8-lead LFCSP Dual ADL5531
ADL5541 50 to 6000 15 20 44 3.5 4.5 to 5.5 90 500 3 mm 3 mm, 8-lead LFCSPBroadband matched
gain block
ADL5542 50 to 6000 19.5 20.6 46 2.8 4.5 to 5.5 93 500 3 mm 3 mm, 8-lead LFCSPBroadband matched
gain block
ADL55701 2300 to 2400 29 31 — — 3.2 to 4.2 130 2350 4 mm 4 mm, 16-lead LFCSP WiMAX power amplifier
ADL55711 2500 to 2700 29 31 — — 3.2 to 4.2 130 2600 4 mm 4 mm, 16-lead LFCSP WiMAX power amplifier
13 V bias supported.
2NF includes external input match.
RF/IF Amplifiers
Features
Broadband and narrow-band RF/IF amplifiers
High linearity and output power
Fully characterizedover frequency range, temperature and powersupply variation
Internal active bias
Most amplifiers internally matched to 50
Low power consumption
Small footprint packages
RF Bonus Feature
33
34
35
36
37
38
C Drivers
PartNumber
Bandwidth @–3 dB (MHz)
Gain(dB)
Distortion2nd (dBc)3rd (dBc)
Output IP3(dBm)
NoiseFigure(dB)
Input Noise(nV/√Hz)
SupplyVoltage (V)
SupplyCurrent
(mA)Package Comments
AD8350-15 900 15 –65/–66(50 MHz)
28(50 MHz) 6.8 1.7 4 to 11 28 3.1 mm 5.05 mm,
8-lead SOIC/MSOP
AD8350-20 700 20 –66/–66(50 MHz)
28(50 MHz) 5.6 1.7 4 to 11 28 3.1 mm 5.05 mm,
8-lead SOIC/MSOP
AD8351 2200(AV = 12 dB) 0 to 26 –79/–81
(70 MHz)31
(70 MHz) 10 2.7 3 to 5.5 28 3 mm 4.9 mm,10-lead MSOP
Gain adjustable withexternal resistor
AD8352 2000 0 to 24 –85/–85(100 MHz)
41(180 MHz) 10 2.6 3 to 5.5 37 3 mm 3 mm,
16-lead LFCSP
Gain adjustable withexternal resistor/ultralow
distortion
AD8375 700 –4 to +20 –85/–88(200 MHz)
50(70 MHz) 8.5 — 4.5 to 5.5 130 4 mm 4 mm,
24-lead LFCSPDifferential input/output
digital gain amplifier
AD8376 700 –4 to +20 –85/–87(200 MHz)
50(70 MHz) 8.5 — 4.5 to 5.5 260 5 mm 5 mm,
32-lead LFCSP
Differential input/output, dual-channel,digital gain amplifier
ADA4937-1/ADA4937-2 1900 Adjustable –84/–91
(70 MHz) — 15 2.2 3.3 to 5 40
Single: 3 mm 3 mm,16-lead LFCSP
Dual: 4 mm 4 mm,24-lead LFCSP
Single-ended input/differential output or
differential input/output;adjustable VOCM
ADA4938-1/ADA4938-2 1000 Adjustable –82/–82
(50 MHz) — 15.8 2.6 5 to 5 37
Single: 3 mm 3 mm,16-lead LFCSP
Dual: 4 mm 4 mm,24-lead LFCSP
Single-ended input/differential output or
differential input/output;adjustable VOCM
Features
Low distortion, fully differential amplifiers
Range of adjustable and fixed gain devices
Drive high resolution (12-bit to 16-bit) analog-to-digitalconverters (ADCs) at high speed (100 MHz)
Small footprint packages, single- and/or dual-supply
RF Bonus Feature
High Frequency AD
39
40
41
42
43
5
PartNumber
RF Frequency(MHz)
VGA Range (dB)
I/Q Frequency(MHz)
Phase Error(deg)
AmplitudeError (dB)
NoiseFigure (dB)
P1dB(dBm)
Input IP3(dBm)
SupplyVoltage (V)
Supply Current (mA)
Package
AD8333 DC to 50 N/A N/A 0.1 0.05 7.8 14.5 30 5.0 +44/–20 5 mm 5 mm,32-lead LFCSP
AD8347 800 to 2700 69.5 65 1 0.3 11 –2 11.5 2.7 to 5.5 64 9.8 mm 6.5 mm,28-lead TSSOP
AD8348 50 to 1000 45 75 0.5 0.25 11 13 28 2.7 to 5.5 48 9.8 mm 6.4 mm,28-lead TSSOP
ADL5382 700 to 2700 N/A 500 0.5 0.25 14 13 30 4.75 to 5.25 195 4 mm 4 mm,24-lead LFCSP
ADL5387 50 to 2000 — 240 0.5 0.25 15 14 30 4.75 to 5.25 180 4 mm 4 mm,24-lead LFCSP
Demodulators
Features
Wide frequency range from 50 MHz to 2.7 GHz
On-chip RF and baseband amplifiers
Wide demodulation bandwidth enables most high order modulation formats, including QAM, QPSK, and 8 PSK
Small footprint packages, single supply
Splitters
Features
Ideal for distribution of CATV signals
Differential inputs and outputs
1 dB gain flatness to 865 MHz
25 dB isolation between channels
Part Number I/O Configuration Input:Outputs 1 dB Bandwidth (MHz) Max Gain (dB) CSO (dBc) CTB (dBc) Noise Figure (dB) Package
ADA4302-4 Differential 1:4 900 5.7 –73 –66 4.4 3 mm 3 mm, 20-lead LFCSP
ADA4303-2 Single-ended 1:2 1200 4 –62 –72 4.4 3 mm 3 mm, 12-lead LFCSP
ADA4304-2 Single-ended 1:2 1000 3 –62 –72 4.6 3 mm 3 mm, 16-lead LFCSP
ADA4304-3 Single-ended 1:3 1000 3 –62 –72 4.6 3 mm 3 mm, 16-lead LFCSP
ADA4304-4 Single-ended 1:4 1000 3 –62 –72 4.6 3 mm 3 mm, 16-lead LFCSP
44
RF Bonus Feature
45
46
47
48
49
50
51
52
53
54
55
56
Features
RFCMOS mobile TV Tuners and receivers
Low power, low NF, good sensitivity
Smaller package sizes (SoC and SiP solutions)
Suitable for mobile/portable applications
Covers worldwide mobile TV standards:
CMMB/DTMB (China): UHF band
ISDB-T 1seg (Japan, Brazil)
DAB/DAB-IP/T-DMB (Europe, China, Korea)
S-DMB (Korea)
DVB-H (Europe and U.S.)
Mobile TV Tuners
Part Number
Mobile TV Configuration Frequency (MHz)Noise Figure
(dB)1
Sensitivity(dBm)1
Power Consumption (mW)
Package
ADMTV300
T-DMB/DAB/FM
RF88 to 108 (FM),
168 to 245 (Band-III),1450 to 1492 (L-Band)
2 70/70/78(FM/band-III/L-band) 5 mm 5 mm, 32-lead LFCSP
ADMTV315 RF+DemodSoC
88 to 108 (FM),168 to 245 (Band-III),
1450 to 1490 (L-Band)–102 90/90/100
(FM/band-III/L-band)7 mm 7 mm, 144-lead FBGA;
5 mm 5 mm, 75-lead WLCSP (optional)
ADMTV316 RF+DemodDual SoC
88 to 108 (FM),174 to 245 (Band-III) –103 90/180
(FM/T-DMB dual) 9 mm 9 mm, 192-lead FBGA
ITD3020 RF88 to 108 (FM),
174 to 245 (Band-III),1450 to 1592 (L-Band)
2.5/3.5(Band-III/L-Band)
100/100/160(FM/band-III/L-band) 6 mm 6 mm, 40-lead LFCSP
ITD3010 RF 88 to 108 (FM),174 to 245 (Band-III) 2.5 100/100
(FM/band-III) 6 mm 6 mm, 40-lead LFCSP
MTV330 S-/T-DMB/FM Dual RF88 to 108 (FM),
174 to 245 (Band-III),2605 to 2655 (S-Band)
2.5/3.5(T-DMB/S-DMB)
100/100/140(FM/T-DMB/S-DMB) 6 mm 6 mm, 48-lead LFCSP
MTV320 S-DMB RF 2605 to 2655 (S-Band) 3.5 140 5 mm 5 mm, 36-lead LFCSPADMTV340 CMMB RF 2635 to 2660 (S-Band) 6.7 150 5 mm 5 mm, 32-lead LFCSP
ADMTV102 DVB-H/DVB-TCMMB/DTMB RF 171 to 245 (VHF),
470 to 862 (UHF) 2.8 180/200(VHF/UHF) 5 mm 5 mm, 32-lead LFCSP
ADMTV202
ISDB-T
RF 470 to 806 (UHF) 2.5 75 2.1 mm 2.1 mm, 23-lead WLCSP
ADMTV203 RF 90 to 222 (VHF),470 to 806 (UHF) 2.5 75 2.1 mm 2.1 mm, 23-lead WLCSP
ITD2010 RF 90 to 222 (VHF),470 to 770 (UHF) 4 100 6 mm 6 mm, 40-lead LFCSP
MTV211 RF + Demod SiP 470 to 806 (UHF) –109 140 9 mm 9 mm, 46-lead LGA
1Numbers for typical value in center frequency.
RF Bonus Feature
57
58
59
60
Logarithmic Amplifiers
Part Number
RF Frequency (MHz)
DynamicRange (dB)
Temp Stability (dB)
ResponseTime (ns)
SupplyVoltage (V)
SupplyCurrent (mA)
Package Comments
AD8302 >0 to 2700 60 1.0 60 2.7 to 5.5 19 5 mm 6.4 mm, 14-lead TSSOP Dual-channel gain and phase detector
AD8306 5 to 400 100 1.0 73 2.7 to 5.5 16 10 mm 6.2 mm, 16-lead SOP Military-specified part available
AD8307 DC to 500 92 1.0 400 2.7 to 5.5 8 5 mm 6.2 mm, 8-lead SOIC/DIP High dynamic range
AD8309 5 to 500 100 1.0 67 2.7 to 6.5 16 5.1 mm 6.5 mm, 16-lead TSSOP Amplitude and limiter outputs
AD8310 DC to 440 100 1.0 15 2.7 to 5.5 8 3.1 mm 4.9 mm, 8-lead SOIC Low cost, high dynamic range
AD8313 100 to 2500 70 1.25 40 2.7 to 5.5 14 3 mm 4.9 mm, 8-lead SOIC Industry standard
AD8314 100 to 2700 45 1.0 70 2.7 to 5.5 4.5 2 mm 3 mm, 8-lead SOIC/LFCSP Industry standard
AD8317 1 to 10,000 55 0.5 6 3.0 to 5.5 22 2 mm 3 mm, 8-lead LFCSP Smaller package, lower cost version of the AD8318
AD8318 1 to 8000 70 0.5 10 4.5 to 5.5 68 4 mm 4 mm, 16-lead LFCSP High accuracy, fast responding
AD8319 1 to 10,000 45 0.5 6 3.0 to 5.5 22 2 mm 3 mm, 8-lead LFCSP Pin-compatible with AD8317
ADL5513 1 to 4000 75 0.5 <20 3.0 to 5.5 30 3 mm 3 mm, 16-lead LFCSP Next-generation AD8313
ADL5519 1 to 10,000 62 0.5 6 3.3 to 5.5 56 5 mm 5 mm, 24-lead LFCSP Dual-channel version of the AD8317
Features
Wide frequency range from dc to 10 GHz
Superior accuracy, better than 0.2 dB and wide dynamic range up to 100 dB
Available with both log and limiter outputs, single and dual channel
Wide range of package options to save board space in RF designs
RF Bonus Feature
FEATURE
14 • www.PlanetAnalog.com March 27, 2006
bbyy CChhrriiss CClloonniinnggeerr,, Analog Devices
CCoonnvveerrtteerrss ffoorr 33GG aarree ooppttiimmiizzeeddffoorr ccoosstt,, ssiizzee aanndd ppoowweerr
Industry demand
for lower-cost
systems and an
ever-increasing
array of features
have caused
designers to shift
their priorities.
A s communications systems haveevolved from second-generation (2G)standards to third-generation (3G)
standards, the performance requirements ofthe analog-to-digital converters (ADCs) anddigital-to-analog converters (DACs) used havealso evolved. In original 2G basestations,designers often selected the highest-perform-ance converters available to provide as muchsystem margin as possible and improve man-ufacturability. This also allowed the manufac-turer to differentiate its basestation with aperformance advantage over its competitors—for example, a higher receive sensitivity.
Focus on costThis is not usually the case with 3G designs,however. Industry demand for lower-costsystems and an ever-increasing array of fea-tures have caused designers to shift theirpriorities. Instead of focusing solely on per-formance, they are now focusing first onother system-enhancing elements, such ascost, power, size and integration, and sec-ondarily on performance. This article willexplore the transition from the highest-per-formance converters in 2G systems to thesmall, low-cost, low-power, highly integratedconverters needed for 3G platforms.
As early 2G/2.5G systems were designed
and deployed in the late '90s and early2000s, many designers needed state-of-the-art converters to address the performancerequirements for these systems. At that time,the highest available performance was foundin the AD6645 14-bit, 80-MSPS pipelinedADC, which was designed on a bipolarprocess and consumed 1.5 watts. Similarly,the best-performing 14-bit, 100-MSPS DACswere based on bipolar or BiCMOS technolo-gies and consumed more than 500 mW.Although this level of performance wasrequired for early-generation systems, thedevices represented leading-edge technology,so they tended to be somewhat costly.
As time has passed, the cost of the con-verter technology has decreased. Since thattime, there are at least three 14-bit, 125-MSPS ADCs that dissipate less than 800 mW.With the introduction of these new competi-tive devices, the price of high-performancehas decreased by at least 50 percent.Although these new devices do not equal theperformance of the higher-power ADCs, theydo offer extremely good performance at lessthan half the power of the first-generationdevice. The latest CMOS DACs, such as theAD9707, dissipate a mere 60 mW at 175MSPS and outperform their BiCMOS cousins.
The first-generation ADCs were designedon a bipolar process, and thefirst-generation DACs weredesigned using 0.6-micronBiCMOS technologies. Thecurrent generation of convert-ers is designed using 0.18-micron CMOS technology.This allows for much smallerdie size, increased feature setsand lower power dissipation,all of which allow semiconduc-tor manufacturers to supportan overall lower cost struc-ture. This benefit gets passedon to the end customer.
Power dissipationAt first glance, power dissipa-
Figure 1: ADC powerconsumption has
decreaseddramatically during
the last decade.
The transition from 2G to 3G standards also means that ADC power andperformance specifications have changed as well.
61
FEATURE
16 • www.PlanetAnalog.com March 27, 2006
tion would seem somewhat insignificant in acellular basestation, especially when consid-ering that the power amplifiers (PAs) canoutput 40 watts. As basestation OEMs arelooking to reduce cost, however, they arelooking everywhere. In particular, somebasestation deployments are now occurringat the top of towers directly connected to thePA itself. The advantage to this type ofbasestation is that it does not need to com-pensate for the cabling losses normally asso-ciated with putting the transceiver at thebottom of the cellular tower.
However, when a system is mounted atthe top of the tower, no active cooling is pos-sible, while the temperature must be kept aslow as possible because there is a log-linearrelationship between temperature and MTBF(mean time between failures). In addition to
the cost of active cooling, OEMs are attempt-ing to save money in the mechanicals of thebasestation. Higher-power converters with-out active cooling require the mechanicalunit in which the transceiver is housed to bemore complex and more expensive in orderto dissipate the heat efficiently.
Converter manufacturers have addressedpower concerns through advances in processtechnology and architectural changes. Asmentioned previously, the latest ADCs andDACs are being designed in fine-line CMOStechnologies. This greatly reduces switchingtransients in the logic cells. In some cases,
dynamic logic elements are employed, fur-ther reducing switching transients by usinga parasitic capacitance to store the logicstate. In the case of ADCs, the architecturehas transitioned from pure flash to pipelinedsingle- and multibit converters, greatlyreducing the number of comparatorsrequired and hence the overall power. Thedata in Figure 1 demonstrates the dramaticdecrease in power consumption for 12-bitand 14-bit ADCs during the last decade.
SizeLike power dissipation, size has become amajor factor in determining the best convert-er for an application. In deployments suchas the tower-mounted system mentionedabove, OEMs are not only limited on power,but are also trying to implement more carri-
ers on the same form factor that orig-inally supported a single sector of abasestation. This increased densityallows operators to deploy more car-riers and thus support more users inheavily populated areas.
In addition to simply reducing the actual size of the converters,another trend is increasing integra-tion. Digital-to-analog converters nowinclude extensive digital signal pro-cessing, such as interpolating filters,complex modulation, inverse sinc filters and numerically controlledoscillators (NCOs). For ADCs, integra-tion has included adding NCOs anddecimating filters to improve per-formance. Both interpolating filtersfor DACs and decimating filters forADCs have the added benefit ofreducing the high-speed switchingnoise between the converter and thedigital logic.
While it may not be obvious, new-generation converters have also
reduced size by enabling simpler architec-tures. In particular, new low-power, high-performance DACs now make direct-conver-sion architectures feasible. In early genera-tions, a low intermediate frequency (IF) wasused, and then multiple mixer stages wereused to achieve the proper RF signal, Figure2a. By removing multiple mixers and ampli-fiers from this signal chain, as shown in Figure 2b, a significant size and costadvantage can be realized.
Although direct conversion has not yetbeen realized for basestation receivers, theIF-sampling capability of many A/D convert-
Figure 2a: Traditionalsuperheterodyne
architecture
62
FEATURE
18 • www.PlanetAnalog.com March 27, 2006
ers has been greatly improved. It is now pos-sible to IF sample at up to 450 MHz andmaintain very good performance. That willallow the removal of a single mixer stage andagain reduce overall system cost.
PerformanceAlthough performance is still an importantfactor, it is no longer the first priority formost designers. As the number of competi-tors in the converter space has increasedover the last five to seven years, the numberof basestation suppliers has also increased.This, in conjunction with the high price paidby operators for 3G licenses, has led toincreased cost pressure for all basestationOEMs. It is this cost pressure that hasforced many designers to look at convertersthat just meet the system requirementsinstead of paying the price premium for thehighest performance converters. The goodnews for many designers, though, is that itis easier to meet many of the 3G (W-CDMA)converter requirements for the 2,100-MHzband than those of the GSM 900-MHz band.For example, a 12-bit, 65-MSPS ADC and14-bit, 65-MSPS DAC can meet the require-ments for a two-carrier W-CDMA system,and a 14-bit, 65-MSPS ADC and 14-bit, 125-MSPS DAC can meet the requirements of athree- to four-carrier system. On the otherhand, only a small number of ADCs andDACs can meet the requirements for themulticarrier GSM 900-MHz band. These, asnoted, represent leading-edge technologyand thus carry a price premium. Multicarrierand multistandard architectures utilizing
leading-edge converters can lower overalldevelopment and manufacturing costs byreducing the number of design variantsrequired, but that must be weighed againstthe higher bill-of-materials cost.
ConclusionIt's clear that the driving force behind choos-ing a converter for a basestation design is toremove cost from the system. Cost is mani-fested in many subtle ways, including powerdissipation, size and performance. Theseattributes affect choices such as systemarchitectures, number of carriers and evendeployment locations. This shift in thinkingis driving many converter manufacturers tointroduce parts that meet as many of thesedimensions as possible. Since it's very diffi-cult to meet every attribute with a singledevice, a broad product portfolio offers thebest chance of providing the right converterfor the specific radio architecture and per-formance requirements. ■
Chris Cloninger joined Analog Devices Inc. in 1995 after receivinga bachelor’s degree in computer engineering from ClemsonUniversity. He is a mixed-signal marketing/systems engineer forhigh-speed analog/digital converters for wireless infrastructureand can be reached at [email protected].
Figure 2b: A direct-conversion architec-ture reduces compo-nent count and cost
compared with a tra-ditional superhetero-
dyne architecture.
For more informationPlanet Analog: monthly in print, plus muchmore on the Web at www.planetanalog.com,covering all things "analog" (includingpower, too)ON
THE
WEB
63
64
Variable Gain Amplifiers
Features
Wide selection of VGAs with frequencies ranging from dc to 3 GHz
Wide dynamic range up to 60 dB
Analog or digital, choice of gain control suited for your applications
Differential or single-ended I/O
Part Number
ControlType
Bandwidth (MHz)
Gain(dB)
GainAccuracy
(dB)
Output IP3
(dBm)
NoiseFigure (dB)
InputNoise
(nV/√Hz)
Supply Voltage
(V)
Supply Current
(mA)Package Comments
AD603 Analog DC to 90 –11 to +31, +9 to +51 0.5
15 (40 MHz) 8.8 1.3 4.75 to
6.3 12.5 5 mm 6.2 mm,8-lead SOIC/DIP Single-ended input/output
AD604 Analog DC to 40 0 to +48, +6 to +54 0.3
35 (10 MHz) 8.4 1.8 5 64 8.2 mm 7.8 mm,
24-lead SSOP/SOIC/DIP Single-ended input/output
AD605 Analog LF to 40 –14 to +34, 0 to +48 0.2
33 (10 MHz) 8.4 1.8 5 36 10 mm 6.2 mm,
16-lead SOIC/DIP Single-ended input/output
AD8260 Digital LF to 180 –6 to +24 1.2542
(10 MHz) — 2.4 3.3 to 5
28.2 5 mm 5 mm, 32-lead LFCSP Single-channel transceiver
AD8330 Analog DC to 150 0 to +50 0.527
(10 MHz) — 5.0 2.7 to 6 20 3 mm 3 mm, 16-lead LFCSP/QSOP Differential input/output
AD8331 Analog LF to 120 –5 to +43, +7 to +55 0.3
33 (10 MHz) 4.2 0.8 4.5 to 5.5 25 9 mm 8.75 mm,
20-lead QSOPSingle-ended input LNA/
differential output
AD8332 Analog LF to 100 –5 to +43, +7 to +55 0.3
32 (10 MHz) 4.2 0.8 4.5 to 5.5 58 5 mm 5 mm, 28-lead
TSSOP, 32-lead LFCSPSingle-ended input LNA/
differential output
AD8334 Analog LF to 100 –5 to +43,+7 to +55 0.3
32 (10 MHz) 4.2 0.8 4.5 to 5.5 116 9 mm 9 mm,
64-lead LFCSPSingle-ended input LNA/
differential output
AD8335 Analog DC to 85 –10 to +38, –2 to +46 0.2
31 (10 MHz) 7.0 1.3 4.5 to 5.5 76 9 mm 9 mm,
64-lead LFCSPSingle-ended input/differential output
AD8336 Analog DC to 100 –14 to +46,0 to +60 0.3 — — 3.1 3 to
12 150 4 mm 4 mm, 16-lead LFCSP Single-ended input/output
AD8337 Analog DC to 280 0 to +24 0.2528
(45 MHz) 14 2.2 2.5 to 5
15 3 mm 3 mm,8-lead LFCSP Single-ended input/output
AD8367 Analog DC to 500 –2.5 to +42.5 0.2
36.5 (70 MHz) 6.2 1.9 2.7 to 5.5 26 5.1 mm 6.5 mm,
14-lead TSSOPSingle-ended input/output
VGA/AGC operation
AD8368 Analog LF to 800 –12 to +22 0.433.7
(70 MHz) 9.5 1.3 4.5 to 5.5 60 4 mm 4 mm, 24-lead LFCSP
Single-ended input/outputVGA/AGC operation
AD8369 Digital 0.001 to 600 –5 to +40 0.5
19.5 (70 MHz) 7.0 2.0 3 to 5.5 37 5.1 mm 6.4 mm,
16-lead TSSOP Differential input/output
AD8370 Digital 0.001 to 700
–11 to +17, +6 to +34 0.5
31 (70 MHz) 7.4 2.1 2.7 to 5.5 78 5.1 mm 6.4 mm,
16-lead TSSOP Differential input/output
AD8372 Digital 1 to 130 –9 to +32 — 35 (65 MHz) 7.9 1.7 4.5 to 5.5 106 5 mm 5 mm,
32-lead LFCSPDifferential input/output,
dual-channel
AD8375 Digital 700 –4 to +20 — 50 (70 MHz) 8.5 1.9 4.5 to 5.5 130 4 mm 4 mm,
24-lead LFCSP Differential input/output
AD8376 Digital 700 –4 to +20 — 50 (70 MHz) 8.5 2.0 4.5 to 5.5 260 5 mm 5 mm,
32-lead LFCSP Dual-channel AD8375
ADL5330 Analog 1 to 3000 –34 to +22 1.531
(900 MHz) 7.8 1.3 4.75 to 6 240 4 mm 4 mm, 24-lead LFCSP Differential input/output
ADL5334 Digital 150 to 2400 60 0.2530.5
(1900 MHz) 5.7 — 4.75 to 5.25 250 6 mm 6 mm,
40-lead LFCSPSingle-ended dual 30 dB
attenuators
Supply
RF Bonus Feature
WIMAX STANDARDSIEEE 802.16, the formal specification of
WiMAX, is targeted at providing broadbandwireless access beyond that which is currentlyavailable using IEEE 802.11x (WLAN).802.16-2004 (sometimes called 802.16d), thelatest full revision of the WiMAX standard, fo-cuses primarily on fixed position point-to-point or point-to-multipoint networks. Thestandard defines OFDM modulation, a fre-quency range of 2 to 11 GHz, and data ratesup to 70 Mbps. OFDM modulation in 802.16dutilizes up to 256 subcarriers with bandwidthsfrom 1.25 to 28 MHz. The subcarriers arespaced such that they are orthogonal to eachother, thus reducing signal interference. Thechoice of signal bandwidth can be determinedin multiple ways. The base station can changethe signal bandwidth based on transmissiondistance and signal environment, or networkproviders may determine the bandwidth avail-able to a user, based on various pricing plans.Figure 1 shows a basic diagram of the signalstructure for an 802.16d network.
CARLOS CALVOAND MATTHEW PILOTTEAnalog Devices Inc.Norwood, MA
WiMAX, which offers high speed dataaccess to large geographical areas —covering distances up to 30 km — is
defined by the IEEE 802.16 family of stan-dards. It uses orthogonal frequency divisionmultiplexing (OFDM) to attain the high datarates. Consequently, the composite signal en-velopes of the data bursts have significantpeaks, which cause large modulation crest fac-tors. Accurate measurement and control ofWiMAX signals is challenging, due to the typi-cally high WiMAX crest factors of 12 dB andto its susceptibility to the varying modulationpatterns. As WiMAX users move towards oraway from the base station, the transmitterchanges the waveform composition (or modu-lation) to optimize data speed and receptionreliability. As the waveform changes, the cor-responding crest factor variation introducesRF power measurement errors.
This article describes several methods toaccurately measure and control the power ofWiMAX transmitters. WiMAX transmit signalpaths can employ high dynamic range loga-rithmic amplifiers and accurate rms detectorsto ensure accurate control of the transmittedsignal across changing modulation types andover temperature. Some of the highlightedtopics will cover the difficulties in dealing withchanging crest factors and rapid envelopechanges.
RF POWER DETECTION:MEASURING WIMAXSIGNALS
Reprinted with permission of MICROWAVE JOURNAL® from the September 2006 issue.©2006 Horizon House Publications, Inc.
65
A recent amendment to 802.16-2004 focuses on the OFDMA physi-cal layer. This updated standard,802.16e-2005 (or Mobile WiMAX),introduces specifications that allowfor mobility in a WiMAX network atspeeds up to 75 MPH (120 kPH). Inorder to accomplish this, 802.16e in-creases the number of available carri-ers from 256 to 2048, with the BPSKpilot tones no longer at fixed intervalsduring each data burst. The band-width for each data burst includes1.25, 5, 10 and 20 MHz. It is also pos-sible that additional bands at 3.5, 5.5and 7 MHz will be made available foruse in Europe. While the 802.16dstandard includes specifications forthe entire 2 to 11 GHz band, 802.16efocuses on licensed bands below 4GHz. Figure 2 shows how the databursts for an OFDMA network over-lap in time, as opposed to the individ-ual bursts of OFDM. This addedcomplexity allows for a larger numberof users and handles the necessarycomplexity for a multi-path mobileenvironment. 802.16e also mergesWiBro under this IEEE standard.WiBro, a Korean system for wireless
broadband access, was introduced inFebruary 2002 when the Korean gov-ernment allocated 100 MHz of spec-trum from 2.3 to 2.4 GHz. This bandwas then standardized by the KoreanTelecommunications Technology As-sociation (TTA) in late 2004. WiBrobase stations offer a theoretical datarate up to 50 Mbps and cover a radiusof 1 to 5 km, allowing the use ofportable and mobile Internet deviceswithin the range of a base station.Under 802.16e, WiBro is defined asone of the available system profilesand consists of 1024 subcarriers with-in an 8.75 MHz bandwidth.
As shown in the figures, WiMAX istransmitted using OFDM and is madeup of 256 to 2048 subcarriers, each ofwhich is modulated using BPSK,QPSK, 16QAM or 64QAM. The com-bination of all these subcarriers and dif-ferent modulation schemes results inthe potential for large peaks andtroughs during each signal burst. Intheory, it is possible for these extremesto fall on top of each other causing alarge peak-to-average ratio (PAR) orcrest factor (CF). Variations in data rateand burst length will affect the overallsignal crest factor. This can cause issueseven in the simplestof WiMAX systems.For example, a sys-tem which utilizes256 subcarriers willhave a theoreticalcrest factor of 10log(256) = 24 dB. Inpractice, it is morelikely to only have apeak crest factor of12 dB because theprobability that thephase of every sub-
carrier will add together is very low. Acrest factor of 12 dB still poses signifi-cant design considerations with regardsto the selection of high linearity devices(mixers, modulators, power amplifiers,etc.) in the RF signal chain. BecauseWiMAX systems can be used for non-line-of-sight applications, gain controlof the transmitter is necessary to adjustthe output TX level depending on thechannel quality. It is also necessary toaccurately control the power amplifier’soutput in order to avoid signal clippingand increased distortion. Some systemsemploy crest factor reduction schemes,typically in the digital baseband pro-cessing, to minimize these effects.
RF POWER DETECTION IN THETRANSMIT SIGNAL CHAIN
Figure 3 shows the block diagramof a typical WiMAX transmit signalchain. The transmit signal path con-sists of three consecutive stages: digi-tal baseband processor or digital sig-nal processor, radio and power ampli-fier. A portion of the transmittedsignal is sampled by the directionalcoupler before it reaches the anten-na. The sampled RF power is deliv-ered to the power detector where it isconverted to a DC voltage. The out-put voltage of the power detector isdigitized and fed to the digital signalprocessor (DSP). Once the powermeasurement is available as a digitallevel, a decision is made based on themeasured output power versus thedesired output power. The DSP willadjust the output power using a digi-tal-to-analog converter to drive thesignal path power control, either atthe baseband, radio or power amplifi-er. The RF power management loopwill reach a steady-state once themeasured output power and the de-sired output power are balanced. Atemperature sensor can also be intro-
TECHNICAL FEATURE
Burst bandwidth=1.25 to 28 MHz200 Subcarriers-5.6 to 90 kHz spacing
Frame ControlHeader
8 Fixed PositionBPSK Pilots
Preamble Data Burst #1 Data Burst #2 Data Burst #N
Modulated SubcarriersBPSK, QPSK, 16QAM, 64QAM
▲ Fig. 1 806.16d OFDM data burst and signal spectrum.
Symbol Number
Time
Subc
hann
el N
umbe
r
FCH
Pre
ambl
e
Dow
nlin
k M
ap
Upl
ink
Map
Downlink Burst #4
Downlink Burst #1
Downlink Burst #2
Downlink Burst #3
Downlink Burst #5
Downlink Burst #6
Downlink Burst #7
▲ Fig. 2 806.16e OFDMA data burst.
RF Detector(Log Amp or
TruPWR)
TempSensor
DSP
DIRECTIONALCOUPLER
ANTENNA
PA Radio
▲ Fig. 3 WiMAX transit signal chain with RF power management.
66
duced as an input to the DSP to addtemperature compensation capabili-ties. This RF power managementconfiguration is not limited to a par-ticular application. Both base stationsand subscriber stations alike may in-corporate variations of this same RFpower control system.
There are two basic methods bywhich the RF power detector and theDSP can interact to control the pow-er of the WiMAX burst. The firstmethod, which is similar to the tech-nique used in envelope ramping inGSM applications, shapes the RFburst instantaneously. It uses thefeedback of the detector to shape theenvelope of the RF burst, made up ofthe preamble, frame control headerand data. This envelope shapingmethod requires high speed detec-tors and fast feedback paths. A morecommonly adopted method is that ofoutput power monitoring. Thismethod takes a power measurementduring a burst and adjusts the RFpower accordingly during the subse-quent burst. The power adjustmentsare highly dependent on the linearityof the radio components to scale andshape the RF burst. A single mea-surement of RF output power can beaffected by the high frequency com-ponents in the measured signal,which can manifest themselves as
noise or AC-resid-ual on the detector’sDC output. To miti-gate this effect,multiple measure-ment points can betaken throughoutthe burst to averageout the AC-residualerror.
DETECTOR BACKGROUNDHistorically, diode detectors have
been used in RF power control cir-cuitry to regulate transmitted power.The simple diode circuitry offers asmall dynamic range with poor tem-perature stability. Even with tempera-ture compensation circuitry, a diodedetector can only offer a small detec-tion range with worsening tempera-ture performance at low input powers.
A popular alternative to the diodedetector is the demodulating logarith-mic amplifier (log amp). The log ampoffers an easy to use linear-in-dB RFpower detection response, a wide dy-namic range, temperature stability andnanosecond response times. Thenewest RF power measurement alter-native is the TruPWR rms-respondingdetector, which offers wide dynamicranges and temperature stability. Inaddition, rms detectors are insensitiveto changes in the peak-to-average ra-tios, whereas diodes and log amps areboth waveform dependent.
Each WiMAX application has di-verse power control and RF detec-tion needs. Subscriber stations can bedesigned with dynamic ranges assmall as 30 dB, but are susceptible tosupply power consumption. Base sta-tions have more allowance for powerconsumption, but need to control dy-namic ranges of up to 60 dB. Both,similarly, require temperature stabili-ty for improved accuracy. Only logamps and rms detectors are able tomeet those needs.
LOGARITHMIC AMPLIFIERSThe first detection method to be
looked at is a peak-detecting device,the logarithmic amplifier. A wide va-riety of log amps is available with de-tection ranges from 40 to 100 dB, andfrequencies from DC to 10 GHz. Atypical block diagram of a log amp isshown in Figure 4.
The core architecture of a log ampis a cascaded chain of linear ampli-
fiers. Each amplifier has a gain of 5 to20 dB. The combination of gain andnumber of amplifiers determines thedetection range of the log amp. Theoutput of each amplifier stage is fedinto a full wave rectifier (markedDET). The outputs of each rectifierare summed together, and the sum-mer’s output is applied to a low passfilter to remove the ripple of the rec-tified signal. This yields the logarith-mic output (often referred to as the“video” output), which will be asteady-state DC output for a steady-state AC input signal. It is the band-width of this video output that is par-ticularly important in a WiMAX sys-tem. The wider the video bandwidth,the faster the log amp is able to re-spond to changes in the peak voltage,or amplitude, of the input signal. Thismakes the log amp particularly suitedto accurately keep up with the enve-lope of the WiMAX burst. With re-sponse times as low as 8 ns, log ampscan easily keep up and measure smallperiods of the RF burst, such as thepreamble, which last about 26 μs.
Using a peak-detecting device likea log amp is advantageous when mea-suring the signal power of a waveformat an exact point in time. Because thelog amp is able to track the envelopeof its input, provided the modulationrate is lower than the video band-width of the log amp, the DC outputwill be an instant-by-instant measure-ment of the peak amplitude of the in-put signal. This kind of measurementis useful in a WiMAX system to detecthigh crest factor signals during a burstand make the appropriate adjust-ments in power amplifier biasing orimplement a crest factor reductionscheme in the next burst. Figure 5shows the output voltage and linearityerror of a log amp at various 2.35GHz OFDM modulations, 256 sub-carrier signals with 10 MHz band-width. The error, normalized toQPSK with 3/4 encoding rate, isgraphed on the secondary y-axis,scaled in dB. While the log amp isable to maintain approximately 50 dBof measurement range within ±1 dBor error for each modulation, there isan obvious shift in the intercept of thetransfer function. The intercept is thepoint on the x-axis through which thetransfer function would pass if theoutput voltage could go to 0 V. Thisintercept shift is a byproduct of the
TECHNICAL FEATURE
INHI
INLO
V I
I V
SETPOINT
OUTPUT
DET DETDETDETΣ
▲ Fig. 4 Basic block diagram of a logarithmic amplifier.
OU
TPU
T (V
)
LINEA
RITY ER
RO
R (dB
)
INPUT (dBm)
2.0
1.6
1.2
0.8
0.4
0
5
3
1
−1−3−5−60 −50 −40 −30 −20 −10 0
CW16QAM 1/2BPSK16QAM 3/4
QPSK 1/264QAM 2/3QPSK 3/464QAM 3/4
▲ Fig. 5 Log amp output voltage andlinearity error at various 2.35 GHz OFDMmodulations.
67
successive detection architecture of alog amp. The amount of interceptshift is based on the crest factor of thesignal. Because the log amp behavioris repeatable over manufacturingprocess variations, the intercept shiftof the log amp for a sine wave versus amodulated input signal can be easilycharacterized. The DSP can then usean offset correction to compensate forthe detector’s output voltage and yieldaccurate RF power measurement.
The low power consumption, of theorder of 15 to 30 mA, makes log ampsviable in both base stations and sub-scriber stations alike. The well-estab-lished log amp architecture offers ex-cellent temperature stability acrosslarge dynamic ranges as well as fast re-sponse times for burst tracking andpeak sampling. However, as the peak-
to-average ratio of the RF signal varies,the output response of a log amp willalso vary. This introduces an uncertain-ty that in many cases must be compen-sated for by the DSP.
RMS-RESPONDING DETECTORSUnlike diodes and log amps, mean
power detectors (or rms detectors)have responses which are indepen-dent of waveform. The waveform-in-dependence is particularly useful asWiMAX systems optimize the qualityof the link by dynamically adjustingthe signal modulation. The compositesignal envelopes of the data burstsmay have significant peaks that candrastically change over time andthrow off measurement accuracy. Us-ing log amps in the RF power controlsystem require some method of com-
pensation; however,rms detectors sim-plify the complexityof the system by re-ducing and in somecases eliminatingc o m p e n s a t i o nschemes.
Figure 6 showsthe block diagramof a 30 dB rms de-
tector. It achieves independencefrom peak-to-average ratios by com-puting the square, mean and rootfunctions of an rms calculation. TheRF input is fed to one of two identi-cal squaring-cells. The squared signalis then averaged through a low passfilter network. The signal is fed to ahigh gain error amplifier that has thesecond squaring-cell in its feedbackpath. This feedback loop performsthe square-root function, thus com-pleting the rms calculation. The out-put is a linear-responding DC voltagewhose conversion gain has units ofVDC/Vrms. The linear-in-volts rms de-tector is able to operate at frequen-
cies as high as 6GHz. The rms-re-sponding detectorallows the RF pow-er control system tomonitor and dy-namically adjust thetransmitter’s outputpower even as thepeak-to-average ra-tio of the transmit-ted signal changes.Figure 7 illustrates
the accuracy in measuring variousOFDM waveform types. The methodused to calculate the error is similarin nature to that used in the log amperror calculation. The linearity errorof the detector is within ±0.5 dBacross the dynamic range of the de-vice. The various waveforms lie ontop of each other with a deviation of acouple tenths of a dB. This 30 dB dy-namic range and low 1 mA powerconsumption is useful for subscriberapplications. The slower responsetime, in the range of 25 μs, limits therms detector use to output powermonitoring. Figure 8 shows theblock diagram of a 60 dB rms detec-tor, which is appropriate for wider dy-namic range base station applications.The input signal is applied to a 12-step, continuously variable gain am-plifier, which is controlled by the set-point, a logarithmic control voltage.The output of the VGA is fed to anaccurate squaring-cell. The fluctuat-ing output is filtered and comparedwith the output of an identical squar-er. At this point, the square and meanoperations of the rms calculation arecomplete. The output is fed back tothe VGA setpoint, making the outputproportional to the logarithm of therms value of the input. The detectorresponse is linear-in-dB, allowing thedevice to measure RF signals in a 60dB dynamic range. The final step ofperforming the square-root functionis not needed for accurate rms detec-tion. Figure 9 shows the perfor-mance of the 60 dB rms detectorwhen measuring various OFDMmodulated input signals. Again, thevarious waveforms lie on top of eachother with negligible deviation. The
TECHNICAL FEATURE
RFINi
OUTPUT
X2
iX2 Buffer
ErrorAmp
TRANS-CONDUCTANCECELLS
+
−
▲ Fig. 6 Block diagram of a TruPWR rms detector with 3 dB linear-in-volts response.
INHI
INLOOUTPUT
SETPOINT
+
− Σ
X2
X2VTGT
▲ Fig. 8 Block diagram of a linear-in-dB TruPWR rms detector with60 dB dynamic range.
LIN
EAR
ITY
ERR
OR
(dB
)
OU
TPU
T (V)
INPUT (dBm)
3
2
1
0
−1−2−3
10
1
0.1
0.01−25 −20 −15 −10 −5 0 5 10
BPSK16QAM 1/2QPSK 1/264QAM 2/3
QPSK 3/464QAM 3/416QAM 3/4
▲ Fig. 7 Linear-in-volts rms detectoroutput voltage and linearity error at various2.35 GHz OFDM modulations.
LIN
EAR
ITY
ERR
OR
(dB
)
OU
TPU
T (V)
INPUT (dBm)
2.0
1.5
1.0
0.5
0
−0.5
−1.0
−1.5
−2.0
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0−60 −50 −40 −30 −20 −10 0 10
CW16QAM 3/4BPSK16QAM 1/2
QPSK 1/264QAM 2/3QPSK 3/464QAM 3/4
▲ Fig. 9 Linear-in-dB rms detector outputvoltage and linearity error at various 2.35GHz OFDM modulations.
68
humps in the linearity error curvecorrespond to the steps of the loga-rithmic VGA. Still, the linearity erroracross the dynamic range stays wellwithin ±0.5 dB. As in the case of thelinear-in-volts detector, the 70 μs re-sponse time of this 60 dB detectoralso limits this detector application tooutput power monitoring.
CONCLUSIONThe emerging WiMAX standard
has great potential to offer wide areacoverage and mobile access to high
speed networks. The large crest fac-tors associated with the OFDM mod-ulation scheme require accurate trans-mit power measurements for PA con-trol and the implementation of crestfactor reduction algorithms. Designsrequiring fast response to the OFDMenvelope should consider the accuracyof log amps. The waveform-indepen-dence provided by rms detectors canreduce, or even eliminate, the needfor compensation schemes in thesenetworks, simplifying the overall de-sign of the transmit chain. ■
Carlos Calvo received his BS and MS degreesin electrical engineering from WorcesterPolytechnic Institute. He is an applicationsengineer in the Advanced Linear ProductsDivision at Analog Devices Inc.
Matthew Pilotte received his BS degree inelectrical engineering from WorcesterPolytechnic Institute. He is an applicationsengineer in the Advanced Linear ProductsDivision at Analog Devices Inc.
TECHNICAL FEATURE
69
70
Mixers/Multipliers
Part Number
RF Frequency (MHz)
IF Frequency (MHz)
LO Frequency
(MHz)
LODrive(dBm)
ConversionGain (dB)
IP3(dBm)
P1dB(dBm)
NoiseFigure(dB)
SupplyVoltage
(V)
SupplyCurrent
(mA)Package Comments
AD831 400 200 400 0 0 24 10 10 5 100 10.02 mm 8.38 mm,20-lead PLCC
IF active mixer
AD8342 LF to 500 DC to 350 LF to 850 0 3 24 8.5 12 4.75 to 5.25 97 3 mm 3 mm,
16-lead LFCSPRF/IF active
mixer
AD8343 DC to 2500 DC to 2500 DC to 2500 –10 7 16.5 2.8 14 5 50 5.1 mm 6.5 mm,14-lead TSSOP
RF/IF active mixer
AD8344 400 to 1200 70 to 400 470 to 1600 0 4 24 8.0 11 4.75 to 5.25 90 3 mm 3 mm,
16-lead LFCSPRF/IF active
mixer
ADL5350 LF to 4000 LF to 4000 LF to 4000 4 –6 26 20 6 2.7 to 3.5 16 3 mm 2 mm,8-lead LFCSP
RF/IF passive mixer
ADL5390 20 to 2400 20 to 2400 DC to 230 N/A 5 24 11 21 4.75 to 5.25 135 4 mm 4 mm,
24-lead LFCSPVector multiplier
ADL5391 DC to 2000 DC to 2000 DC to 2000 N/A Variable 29 15.1 — 4.5 to 5.5 130 3 mm 3 mm,16-lead LFCSP
RF/IF multiplier
Features
High linearity active mixers provide conversion gain
Broadband family portfolio with operation up to 4 GHz
Integrated LO driver on-chip
Small footprint packages, single supply
RF Bonus Feature
WIRELESS TECHNOLOGIESWIRELESS TECHNOLOGIES
used to measure gain and VSWR can re-duce overall component count. This articlewill focus on techniques that can be usedto perform these in-situ measurements inwireless transmitters.
A TYPICAL WIRELESS TRANSMITTERFigure 1 shows a typical wireless
transmitter. It consists of mixed-signalbase band circuitry, an up-converter(which generally includes one or moreintermediate frequencies or IFs), ampli-fiers, filters and a power amplifier. Thesecomponents may be located on differentPCBs or may even be physically separat-ed. In the example shown, an indoor unitis connected to an outdoor unit with a ca-ble. In such a configuration, both unitsmay be expected to have well defined,temperature-stable gains. Alternatively,each unit might be expected to deliver awell-defined output power. There are twodifferent approaches to the ultimate goalof delivering a known power level to theantenna: power control or gain control.
With power control, the system relies onbeing able to precisely measure the outputpower (using detector D in this example).Once the output power has been mea-sured, the gain of some component in the
Measurement and control of gain andreflected power in wireless trans-mitters are critical auxiliary func-
tions that are often overlooked. The powerreflected back from an antenna is speci-fied using either the voltage standingwave ratio (VSWR) or the reflection coeffi-cient (also referred to as return loss). PoorVSWR can cause shadowing in a TVbroadcast system as the signal reflectedoff the antenna reflects again off the pow-er amplifier and is then rebroadcast. Inwireless communications systems, shad-owing will produce multi-path-like phe-nomena. While poor VSWR can degradetransmission quality, the catastrophicVSWR that results from damage to coaxialcable or to an antenna can, at its worst,destroy the transmitter. The gain of a sig-nal chain is measured and controlled aspart of the overall effort to regulate thetransmitted power level. If too much ortoo little power is transmitted, the resultwill be either violation of emissions regu-lations or a poor quality link. The reflec-tion coefficient is calculated by measuringthe ratio between forward and reversepower. Gain, on the other hand, is calcu-lated by measuring input and output pow-er. The high commonality of hardware
Measuring VSWRand Gain in Wireless
SystemsEAMON NASH
Analog Devices, Wilmington, MA
Reprinted with permission of MICROWAVE JOURNAL® from the November 2005 issue.©2005 Horizon House Publications, Inc.
71
WIRELESS TECHNOLOGIES
system (in this case, it might be theIF VGA) is varied until the correctoutput power is measured at theantenna. It is not necessary toknow the gain of the circuit or theexact input signal amplitude; it isjust a matter of varying the gain orinput signal until the output poweris correct. This approach is often(incorrectly) referred to as auto-matic gain control or AGC. To becorrect, it should be referred to asautomatic power control or APCsince it is power not gain that is be-ing precisely regulated.
Gain control takes a differentapproach. Here, at least two powerdetectors are used to precisely reg-ulate the gain of the complete sig-nal chain or a part thereof. A pre-cise input signal is then applied tothe signal chain. A number of fac-tors ultimately determine whichapproach is used. Power controlrequires only one power detectorand makes sense in a non-config-urable transmitter whose compo-nents are fixed. For example, pow-er could be measured at the outputof the RF HPA but adjustments
would be made using the IF VGA.Gain control, on the other hand,may make more sense in a recon-figurable system whose compo-nents come from different vendors.In the example, the input powerand output power of the HPA arebeing measured (using detectors Cand D) so the gain can be regulat-ed independent of the other blocksin the circuit. Note that the pow-er/gain control loops can be allanalog or microprocessor based.Gain control would be less practi-cal in the example since the two re-quired detector signals (detectorsA and D) are physically remotefrom one another. A more practicalapproach would be to indepen-dently control the gain of the in-door and outdoor units.
RF DETECTORSUntil recently, most RF power
detectors were built using a tem-perature-compensated half-waverectifying diode circuit. These de-vices deliver an output voltagethat is proportional to the inputvoltage over a limited dynamic
range (typically 20 to 30 dB). As aresult, the relationship betweenoutput voltage and input powerin dBm is exponential (see Fig-ure 2). While the temperaturestability of a temperature-com-pensated diode detector is excel-lent at high input powers (+10 to+15 dBm), it degrades significant-ly as the input drive is reduced. Alog detector, on the other hand,delivers an output voltage pro-portional to the log of the inputsignal over a large dynamicrange (up to 100 dB). The temper-ature stability is usually constantover the complete dynamicrange. A log-responding deviceoffers a key advantage in gainand VSWR measurement applica-tions. In order to compute thegain or the reflection loss, the ra-tio of the two signal powers(either OUTPUT/INPUT orREVERSE/FORWARD) must becalculated (see Figure 3). Ananalog divider must be used toperform this calculation with alinear-responding diode detector,but only simple subtraction is re-quired when using a log-responding detector (since log(A/B) = log (A) – log (B)). A dualRF detector has an additional ad-vantage compared to a discreteimplementation. There is a natur-al tendency for two devices (RFdetectors in this case) to behavesimilarly when they are fabricat-ed on the same silicon wafer.Both devices will have similartemperature drift characteristics,for example. At the summingnode, this drift will cancel to yielda more temperature-stable result.
GAIN MEASUREMENTEXAMPLE
Figure 4 shows a transmitterwhose gain is regulated using adual power detector. The simplifiedtransmit signal chain shown con-sists of a high performance IF-syn-thesizing DAC, a VGA, a mixer/up-converter and a high power ampli-fier. High performance DACs, suchas the AD9786 and AD9779 thatrun at sampling frequencies up to500 MSPS and beyond, are capableof synthesizing intermediate fre-quency outputs (100 MHz in thisexample). The output of the DAC is
Mixed Signal (ADCs/DACs) and μProcessor/DSP
IFVGA
IFAMP
RFAMP HPA
Σ
DAC
SAW BPF
DETA
DETB
DETC
DETD
INDOOR UNIT OUTDOOR UNIT
▲ Fig. 1 Power control versus gain control.
OU
TPU
T V
OLT
AG
E (V
)
INPUT POWER (dBm)
2.5
2.0
1.5
1.0
0.5
0 -70 -60 -50 -40 -30 -20 -10 0 10 20
Diode DetectorLog Detector
▲ Fig. 2 Transfer functions of diodeand log detectors.
DIODE DETECTOR
Pin A Pin B
Vout = Vin A/Vin B
= Gain(V/V)
LOG DETECTOR
Pin A+ -
Pin B
Vout = log(Pin A) - log(Pin B)
= log(Pin A/Pin B)= Gain(dB)
� �Σ
▲ Fig. 3 Calculating the gain usingdiode and log detectors.
72
WIRELESS TECHNOLOGIESNyquist filtered using a bandpassfilter before being applied to anADL5330 variable gain amplifier.Conveniently, the amplifier acceptsa differential input that can be tieddirectly to the output of the differ-ential filter. This, in turn, is tied tothe DAC output. The VGA outputis converted from differential tosingle-ended using a balun trans-former, and is then applied to theADL5350 mixer. After appropriatefiltering (not shown), the signal isamplified and transmitted at amaximum output power level of 30 W (approximately +45 dBm).
The gain of the signal chain ismeasured by detecting the powerat the DAC output and at the out-put of the HPA. The gain is thenregulated by adjusting the gain of aVGA. At the DAC and PA outputs,a sample of the signal is taken andfed to the detectors. At the HPAoutput, a directional coupler isused to tap off some of the powergoing to the antenna. The transferfunction of the AD8364 dual detec-tor (see Figure 5) shows that at theoutput frequency used (2140 MHzin this case), the detector has thebest linearity and the most stabletemperature drift at power levels
below –10 dBm. Thus, the powercoming from the directional cou-pler (+25 dBm max) must be atten-uated before being applied to thedetector. If maximizing the detec-tor dynamic range is not critical tothe application, the attenuation canbe conservatively set at 41 dB sothat the detector sees a maximuminput power of –16 dBm. This stillleaves about 34 dB of useful dy-namic range over which the gaincan be controlled. To detect the in-put power level at the DAC output,a directional coupler is impracticalat this low frequency. In addition,directional coupling is not neces-sary since there will be little or noreflected signal at this point in thecircuit. Furthermore, the powerbeing delivered to the VGA is –10dBm, so the power to be deliveredto the detector is only 6 dB lower.Since the detector has an input im-pedance of 200 Ω and the VGA hasan input impedance of 50 Ω, itquickly becomes clear that the twodevices can simply be connected inparallel. With the same voltagepresent at both inputs, the 50 to200 Ω impedance ratio will result ina convenient 6 dB power differ-ence. Where high measurement
precision is required, care must bepaid to the temperature stability ofthe power detectors. This issue isfurther complicated if the tempera-ture drift characteristics of the de-tectors change with frequency. Thedual detector shown provides tem-perature compensation nodes. Thetemperature compensation is acti-vated by connecting a voltage tothe ADJ pins of each detector (thisvoltage can be conveniently de-rived using a resistor divider fromthe 2.5 V on-chip reference). Nocompensation is required for thelow frequency input (ADJB isgrounded), while a 1 V compensa-tion voltage is required at ADJA tominimize temperature drift at 2.1GHz. While the focus of the appli-cation circuit is gain measurement,it should be noted that input powerand output power can also be mea-sured. The outputs of the individ-ual detectors are available and canbe separately sampled. Becausethe detectors are log responding,their outputs can be simply sub-tracted to yield gain. This subtrac-tion is performed on chip and thegain result is delivered as a differ-ential voltage. The full-scale differ-ential voltage is approximately ±4V (biased up to 2.5 V) with a slopeof 100 mV/dB. Digitizing with a 10-bit ADC with an LSB size of ~10mV (±5 V full scale), 0.1 dB mea-surement resolution is achievable.
VSWR MEASUREMENTEXAMPLE
A dual log detector can also beused to measure the reflection co-efficient of an antenna. In Figure6, two directional couplers areused, one to measure the forward
IF Synthesizing
DAC60dBVGA
1nF 1.1
Mixer
DirectionalCoupler
+45dBm
41dB
20dB
LO
1nF
1nF
1nF100MHz
100 MHz-10 dBm
50Ω
50Ω
50Ω
DAC
ADC
ADC
μPro
cess
or/D
SP
ADC
OutputPower
InputPower
VSTB
CLPF
VGA Control
VGA Control
ITGT2
ISIG2
IERR
ITGT2
ISIG2
VSTA
CLPF
OUTA
FBKAChannel A
TruPwrTM
Channel BTruPwrTM
FBKB
OUTPOUTAOUTB
Vref
INLA
ADJA
VREF
ADJB
INLBINHB
0.1uF 0.1uF
-16 dBm(max)
0.1uF
1:4
0.1uF
INHA
OUTB
OUTN
Gain
HPA
▲ Fig. 4 Gain control using a dual rms-responding log detector.
GA
IN O
UTP
UT
(V)
CHANNEL A INPUT POWER (dBm)(CHANNEL B = -25 dBm)
543210
-1-2-3-4-5
2.52.01.51.00.50-0.5-1.0-1.5-2.0-2.5
-60 -50 -40 -30 -20 -10 0 10 20
GA
IN E
RR
OR
(dB
)
+85° +25° -40°
▲ Fig. 5 Gain transfer function of adual rms-responding log detector.
73
WIRELESS TECHNOLOGIES
power and one to measure the re-verse power. As in the previous ex-ample, additional attenuation is re-quired before applying these sig-nals to the detectors. The AD8302dual detector has a measurementrange of ±30 dB. The level planningused in this example is graphicallydepicted in Figure 7. In this exam-ple, the expected output powerrange from the HPA is 30 dB, from+20 to +50 dBm. Over this powerrange, reflection coefficients from0 dB (short or open load) up to –20dB should be able to be accuratelymeasured. Each of the AD8302’sdetectors has a nominal inputrange from 0 to –60 dBm. In thisexample, the maximum forwardpower of +50 dBm is padded downto –10 dBm at the detector input.When the HPA is transmitting at itslowest power level of +20 dBm, thedetector sees a power of –40 dBm,still well within its input range.
The power from the reversepath is padded down by the sameamount. This means that the sys-tem is capable of measuring re-flected power up to 0 dB. Thismay not be necessary if the sys-tem is designed to shut downwhen the reflection coefficientdegrades below a certain mini-mum (such as 10 dB), but it ispermissible because the detectorhas so much dynamic range. Forexample, when the HPA is trans-mitting +20 dBm, the reversepath detector will see an inputpower of –60 dBm if the antennahas a return loss of 20 dB. Theapplication circuit provides a di-rect reading of return loss, but noinformation is provided about theabsolute forward or reverse pow-er. If this information is required,the dual detector used in the gaincontrol would be more useful be-cause it would provide a measure
of absolute for-ward and reflect-ed power alongwith the reflec-tion coefficient.The dual log de-tector used in thereturn loss mea-surement alsoprovides a phaseoutput. Becauseof the large gainin the main signalpath of a progres-sive compressionlog amp, a limited(amplitude satu-rated) version ofthe input signal is
a natural by product. These lim-iter outputs are multiplied togeth-er to yield a phase-detected out-put with a range of 180° centeredaround an ideal operating pointof 90°. In a VSWR application, thisinformation constitutes the phaseangle of the reflected signal (withrespect to the incident signal) andmay be of use in optimizing thepower delivered to the antenna.
AMPLIFIER GAINMEASUREMENT USING ASINGLE LOG DETECTOR ANDAN RF SWITCH
Figure 8 shows an alternativeapproach to gain measurement,which is also applicable to VSWRmeasurement. In this application,measuring and controlling thegain of a PA is desired. The PA inthe example is running at 8 GHzand has an output power rangefrom +20 to +50 dBm. This is afixed-gain PA, so the output pow-er is adjusted by changing inputpower. Two directional couplersare used to detect input and out-put power. However, there is onlya single log detector so the twosignals are alternately connectedto the detector using a single-pole, double-throw RF switch.The AD8317 detector has a 0 to–50 dBm input range at this fre-quency. To measure the gain, theinput and output powers are al-ternately measured and digitized.The results are then simply sub-tracted to yield gain. Once thegain is known, the digital controlloop is completed by making anynecessary adjustments to the gainof the PA via a bias adjustment.The level planning for this exam-ple is shown in Figure 9. Attenu-ation is used so that the two inputpower levels at the RF switch areclose together and within the in-put range of the detector.
PRECISE GAIN MEASUREMENTWITHOUT FACTORYCALIBRATION
In addition to reducing compo-nent count, this gain measure-ment method has a number of in-teresting features. Because thesame circuit is being used tomeasure input and output power,it is possible to make precise,
+50
+40
+30
+20
+10
0
-10
-20
-30
-40
-50
-60
60 dBAttenuation
60 dBAttenuation
ForwardPowerRange
ReversePowerRange
Detector A/BInputRange
Powerat Input A Power
at Input B
P
OW
ER (
dBm
)
▲ Fig. 7 Level planning for VSWR measurement using a duallog detector.
μPro
cess
or/D
SP
ADC
ADC
60dB Log Amps(7 Detectors)
60dB Log Amps(7 Detectors)
20 dB 20 dB
40 dB 40 dB
0.1μF
0.1μF
INPA
INPB Log AMP B
Log AMP A
MFLT ForwardPower
ReversePower
MSET
PSET
VPHS
VMAG
PFLT
+
+ -
- +
- Σ Σ
Σ
Pin = -10 to -40 dBm
Pin = -10 to -60 dBm
Pout = +20 to +50 dBm
HPA
▲ Fig. 6 Return loss measurement using a dual log detector.
74
WIRELESS TECHNOLOGIESVOUT1 = SLOPE
• (PIN1 – INTERCEPT)
To figure out the unknown, PIN, theequation can be rewritten as
PIN1 = (VOUT1/SLOPE) – INTERCEPT
Since gain is the difference in themeasured input powers (the differ-ent attenuation levels of the twopaths still have to be factored in), itcan be written as
GAIN = (VOUT1 – VOUT2)/SLOPE
Therefore, the intercept of the detec-tor is not required to calculate thegain. Even though the slope of a de-tector will change from device to de-
vice and over temperature, if Vout1 and Vout2 are closeto each other (it can be done with good level plan-ning and because of the finite input range of the de-tector), a typical value for the slope can be taken di-rectly from the datasheet and used in the above cal-culation.
OUTPUT POWER MONITORINGIn the gain measurement using a single log detec-
tor, the power is measured in order to calculate gain,so the system shown can also be used to monitor theoutput power. However, this cannot be done pre-cisely without factory calibration. To calibrate thecircuit, the antenna must be temporarily replaced bya power meter. The output power and detector volt-ages are then measured at two points within the lin-ear range of the detector. These numbers wouldthen be used to calculate the slope and intercept ofthe detector. For optimum precision, the detector in-cludes a temperature compensation pin. A resistor isconnected between this pin and ground to reducethe temperature drift to approximately ±0.5 dB at thefrequency of operation (8 GHz in the exampleshown). As a result, it is not necessary to do any ad-ditional calibration over temperature.
CONCLUSIONBecause of their linear-in-dB transfer function, log
amplifiers can be easily used to measure gain and re-turn loss. When dual devices are used, very high mea-surement precision is achievable. In some cases, thiscan be achieved without factory calibration. In all cas-es, careful power level planning is necessary so thatthe power detectors are driven at power levels that of-fer good linearity and temperature stability. ■
Eamon Nash holds a BEng degree in electronics from theUniversity of Limerick, Ireland. He has worked at Analog Devices for15 years, first as a field applications engineer, based in Germany,covering mixed signal and DSP products, then as a product lineapplications engineer specializing in RF building block componentsfor wireless applications. He is now applications engineeringmanager for RF Standard Products at Analog Devices.
temperature-stable gain measurements withoutever calibrating the circuit. A look at the nominaltransfer function of a log detector will help in un-derstanding why (see Figure 10).
HPA40 dB
PABIAS
CONTROL
20 dB 20 dB
A
B
40 dB
DAC
μPro
cess
or/D
SP
ADCVOUT
500R100pF 0.1uF
VPOS
+5V
VSETV 1
CLPF
COMM
INHI
INLO1nF
1nF220pF
Select Input/Output
DET
Slope
DET DET DETΣ
Gain Bias
Pin = -20 to +10 dBm
Pout = +20 to +50 dBm@ 8 GHz
0.7pF
▲ Fig. 8 Gain measurement using a single log detector.
+50
+40
+30
+20
+10
0
-10
-20
-30
-40
-50
20 dBDirectional
Coupler
20 dBCoupler
+40 dB
Attenuation
PAInputPowerRange
PAOutputPowerRange
DetectorInputRange
Powerat SwitchInput A
Powerat SwitchInput B
P
OW
ER (
dBm
)
▲ Fig. 9 Level planning for gain measurement using a singlelog detector.
V out
(V
)
Erro
r (d
B)
Pin (dBm)
2.00
1.75
1.50
1.25
1.00
0.75
0.50
0.25
0
2.0
1.5
1.0
0.5
0
-0.5
-1.0
-1.5
-65 -55 -45 -35 -25 -15 -5 5 15
-40°C +85°C+25°C
Vout = Slope x (Pin-Intercept)Slope (mV/dB) = (Vout2-Vout1)/(Pin2-Pin1)Intercept (dBm) = Pin1- (Vout1/Slope)
Vout2
Vout1
Pin2Intercept
Pin1
▲ Fig. 10 Calibrating a log detector.
75
76
Part Number
RFFrequency
(MHz)
I/Q 3 dB Bandwidth
(MHz)
PhaseError(deg)
AmplitudeError(dB)
CarrierSuppress
(dBm)
SidebandSuppress
(dBc)
Noise Floor (dBm/Hz)
P1dB(dBm)
OutputIP3
(dBm)
SupplyVoltage (V)
SupplyCurrent
(mA)Package
AD8340 700 to 1000 230 — — –30 –32 –148 11 24 4.75 to 5.25 130 4 mm 4 mm,24-lead LFCSP
AD8341 1500 to 2400 230 — — –25 –34 –150.5 9 18 4.75 to 5.25 125 4 mm 4 mm,
24-lead LFCSP
AD8345 140 to 1000 80 0.5 0.2 –42 –42 –155 2.5 25 2.7 to 5.5 65 5.1 mm 6.5 mm,16-lead TSSOP
AD8346 800 to 2500 70 1 0.2 –42 –36 –147 –3 20 2.7 to 5.5 45 6.5 mm 5.1 mm,16-lead TSSOP
AD8349 700 to 2700 160 0.3 0.1 –42 –43 –156 6 19 4.75 to 5.5 135 5.1 mm 6.4 mm,16-lead TSSOP
ADL5370 300 to 1000 >500 0.76 0.03 –50 –41 –160 11 24 4.75 to 5.25 205 4 mm 4 mm,24-lead LFCSP
ADL5371 500 to 1500 >500 0.1 0.03 –50 –55 –158.6 14 27 4.75 to 5.25 175 4 mm 4 mm,24-lead LFCSP
ADL5372 1500 to 2500 >500 0.21 0.09 –45 –45 –158 14 27 4.75 to 5.25 165 4 mm 4 mm,
24-lead LFCSP
ADL5373 2300 to 3000 >500 1 0.13 –39 –39 –157 14 25 4.75 to 5.25 166 4 mm 4 mm,
24-lead LFCSP
ADL5374 2800 to 4000 >500 0.2 0.02 –32 –50 –158 12 22 4.75 to 5.25 175 4 mm 4 mm,
24-lead LFCSP
ADL5375 400 to 6000 >500 –0.05 –0.07 –46 –50 –160 9.4 23 4.75 to 5.25 131 4 mm 4 mm,24-lead LFCSP
ADL5385 50 to 2200 >500 0.39 0.03 –46 –50 –159 11 26 4.75 to 5.25 215 4 mm 4 mm,24-lead LFCSP_VQ
ADL5390 20 to 2400 230 N/A 0.5 N/A N/A –148 11 24 4.75 to 5.25 135 4 mm 4 mm,24-lead LFCSP
Analog I/Q and Vector Modulators
Features
Wide range of modulators covering frequencies of operation from 20 MHz to 4 GHz
Vector modulators provide simultaneous control of both gain and phase within a single IC
Pin compatibility to cover all cellular frequency bands
High dynamic range direct conversion
RF Bonus Feature
77
78
/1...../32
/1...../32
/1...../32
REF A
REF B
AD9549
DDS/DAC
REF MONITORSAND SWITCHING
SYSTEM CLOCKMULTIPLIER
DIGITAL PLLR, S DIVIDERS
HOLDOVER
LOW-PASSFILTER
AD9514
Δt LVDS/CMOS
LVPECL
LVPECL
10
–1000 500
SIG
NA
L P
OW
ER
(dB
m)
FREQUENCY (MHz)
0
–10
–20
–30
–40
–50
–60
–70
–80
–90
100 200 300 400
CANCELLEDSPUR
CANCELLEDSPUR
Figure 1. The plot shows the synthesizer’s output spectrum for a 399 MHz sine wave, with a 1 GHz sampling rate.
Figure 2. Shown is a complete clock synchronization, generation, jitter clean-up, and distribution circuit with the DDS at the heart of the system.
CARRIER = 399MHzSFDR WITHOUT SPURKILLER = –63.7dBcSFDR WITH SPURKILLER = –69.3dBcFREQUENCY SPAN = 500MHzRESOLUTION BW = 3kHzVIDEO BW = 30kHz
79
Direct Digital Synthesizers (DDS)
Features
AD9913: 250 MSPS DDS consumes only 50 mW power
Low power DDS moves into portable/handheld markets
Low cost DDS with integrated DAC replaces standalone DAC required after FPGA
Programmable modulus mode available for exact rational frequency synthesis
Part Number
MasterClock(MHz)
TuningWord Width
(Bits)
DAC Resolution
(Bits)
SFDR(dBc) to Nyquist
Narrow-BandSFDR (dB)/AOUT (MHz)/
Window (MHz)
Power Dissipation
(mW)Package
SupplyVoltage
(V)I/O Interface
REFCLKMultiplier
On-BoardComparator
Comments
AD9850 125 32 10 54 80/40.1/0.5 480 28-lead SSOP 3.3 to 5.0
Serial or parallel
AD9851 180 32 10 53 85/40.1/0.5 650 28-lead SSOP 3.3 to 5.0
Serial or parallel
AD9852 300 48 12 48 83/10/1 2200 80-leadLQFP/TQFP_EP 3.3 Serial or
parallel Chirp function
AD9854 300 48 12 48 83/10/1 2200 80-leadLQFP/TQFP_EP 3.3 Serial or
parallelQuadrature outputs,
chirp function
AD9858 1000 32 10 58 80/40/1 1900 100-leadTQFP_EP 3.3 Serial or
parallel
Integrated charge pump, phase detector,
analog multiplierAD9859 400 32 10 56 80/160/0.1 200 48-lead TQFP_EP 1.8 SerialAD9951 400 32 14 56 80/160/0.1 200 48-lead TQFP_EP 1.8 SerialAD9952 400 32 14 56 80/160/0.1 200 48-lead TQFP_EP 1.8 Serial
AD9953 400 32 14 56 80/160/0.1 200 48-lead TQFP_EP 1.8 Serial Programmable RAMLUT
AD9954 400 32 14 56 80/160/0.1 200 48-lead TQFP_EP 1.8 SerialProgrammable RAM
LUT, automaticfrequency sweep
AD9956 400 48 14 56 80/160/0.1 400 48-lead LFCSP 1.8 Serial On-board 2.7 GHz PLLAD9958 500 32 10 53 81/200/1 420 56-lead LFCSP 3.3/1.8 Serial 2 complete channelsAD9959 500 32 10 53 81/200/1 680 56-lead LFCSP 3.3/1.8 Serial 4 complete channels
AD9910 1000 32 14 53 86/300/0.5 800 100-leadTQFP_EP 3.3/1.8 Serial or
16-bit parallel
RAM, polar modulation, phase/
frequency/amp ramp
AD9911 500 32 10 53 81/200/1 275 56-leadLFCSP 3.3/1.8 Serial
Multimode modula-tion, targeted spur
reduction
AD9912 1000 48 14 58 86/398.7/0.5 800 64-leadLFCSP 3.3/1.8 Serial Spur reduction
AD9913 250 32 10 58 88/99.7/0.03 50 32-lead LFCSP 1.8 Serial or parallel
RF Bonus Feature
80
Analog Devices, Inc.
Worldwide Headquarters
Analog Devices, Inc.
One Technology Way
P.O. Box 9106
Norwood, MA 02062-9106
U.S.A.
Tel: 781.329.4700
(800.262.5643,
U.S.A. only)
Fax: 781.461.3113
Analog Devices, Inc.
Europe Headquarters
Analog Devices, Inc.
Wilhelm-Wagenfeld-Str. 6
80807 Munich
Germany
Tel: 49.89.76903.0
Fax: 49.89.76903.157
Analog Devices, Inc.
Japan Headquarters
Analog Devices, KK
New Pier Takeshiba
South Tower Building
1-16-1 Kaigan, Minato-ku,
Tokyo, 105-6891
Japan
Tel: 813.5402.8200
Fax: 813.5402.1064
Analog Devices, Inc.
Southeast Asia
Headquarters
Analog Devices
22/F One Corporate Avenue
222 Hu Bin Road
Shanghai, 200021
China
Tel: 86.21.2320.8000
Fax: 86.21.2320.8222
©2008 Analog Devices, Inc. All rights reserved.
Trademarks and registered trademarks are the property
of their respective owners.
Printed in the U.S.A. G07603-.55-8/08 www.analog.com/RF