A forum for the exchange of cirruH · .· .:\ /::::::.~'-··,.:and software for real-world signal JH'O('l'ssing
`•.·· ... '
`COMPLETE, LOW-DISTORTI . ·• _
`':FREQUENCIES UP TO 500 MHz (page 3)
`~.
`More about sigma-d_eltit;~n~rtef{f -:J ,-eir applications (see pages 6 & 24)
`Audio preaiitpfifiefct~~itfmf microphones ( page 12)
`·::::::\/j:;o~piete)iiit~nts on page 3
`
`.
`
`. .. ·- -
`
`-_-_··.- ---·----- ..
`
`-
`
`-
`
`- .. . ·/ •:·; -
`
`\
`
`~ - \_ - · . -. --~
`
`Volume 28. Number 2. 1994
`
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. Exhibit 1025 Page 1
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. v. Analog Devices, Inc. IPR2020-01559
`
`

`

`Editor's Notes
`NEW FELLOWS
`We are pleased to note that Richie
`Payne and Paul Ruggerio were
`introduced as Analog Devices
`Fellows at our 1984 General
`Technical Conference. Fellow, at
`Analog Devices, represents the
`highest level of advancement that a
`technical contributor can achieve,
`on a par with Vice President. The
`criteria for promotion to Fellow
`are very demanding. Fellows will have earned universal respect and
`recognition from the technical community for unusual talent and
`identifiable innovation at the state of the an; their significant technical
`contnbutions will have had a major impact on the company's revenues;
`they must have demonstrated superior creative ability in product- or
`process cechnology leading to commercial success..
`
`Other attributes include roles as mentor, consultant, organizational
`bridge, gatekeeper, entrepreneur, teacher, and ambassador. They
`must also be effective as leaders of teams and contributors co team
`effort and in understanding the needs of the customer. Our two
`new Fellows' accomplishments and technical abilities, as well as
`their personal qualities, well-qualify them for this appointment.
`They join Fellows Derek Bowers ( 1991 ), Paul Brokaw ( 1980),
`Lew Counts ( 1984), Barrie Gilbert (1980),Jody Lapham (J 988),
`Fred Mapplcbcck ( 1989), jack Memishian (1980), Mohammad
`Nasser (1993), Wyn Palmer (1991), Carl Robercs (1992), Brad
`Scharf (1993), Mike Timko (I 982), Bob Tsang (1988), Mike
`Tuthill (I 988), and Jim Wilson (1993).
`
`RICHIE PAYNE
`Dr. Richard Payne is best-known
`in recent years as leader of the
`team that developed the prize(cid:173)
`winning ADXL50 accelerometer,
`which brought surface-micro(cid:173)
`machining co commercial viabiliry.
`However, he has a long career of
`technical concributions to semi(cid:173)
`conductor processing, going back
`to the '70's, when he did pioneer(cid:173)
`ing work at Bell Labs, inventing the first successful process for
`fabricating high-frequency bipolar transistors using ion
`implantation.
`
`,
`·
`
`At Analog Devices, he has contributed to and managed process
`development for high-performance linear, digital, and mixed-signal
`devices. The strategies and people chat he put in place have
`significancly enhanced Analog's leadership position in high(cid:173)
`performance analog and mixed-signal technology. He also built
`and managed a major wafer-fabrication facility that put the results
`of process development to the severest test : the ability to
`manufacture and ship advanced ICs in significant quantiry.
`
`He has an A.B. degree from Dartmouth and received his Ph.D. in
`Physics from Yale in 1970. He is a Fellow of IEEE; in 1992, he
`received the IEEE Electron Device Society's J.J. Ebers Award for
`"engineering achievements in process architecture and device
`design for twin-tub CMOS integrated-circuit technology,
`contributions to bipolar technology, and advancement of these
`technologies in commercial utilization."
`
`PAUL RUGGERIO
`Paul Ruggerio has been a key
`contributor to Analog Devices
`semiconductor manufacturing, in
`ways too numerous to mention,
`from the day we acquired our fir-st
`fab in the early '70's. His expertise
`embraces the broad areas of lith(cid:173)
`ography, thin film, IC passivation,
`and metallization. His early
`contributions include our first Epi
`facility, maskmaking, projection aligning, chrome masking, and
`stepper lithography. He was one of the industry's first investigators
`in the area of ultraviolet resist stabilization.
`His supporting role in innovative photomask technology made
`possible the first complete single-chip AID converter, the AD57 I
`(Analog Dialogue 12-1, 1978). Later, he led a ceam that developed
`the universal double-level metallization process, now widely used
`within Analog Devices. Thin-film precision resistors have been a
`key to AD I's preeminence in high-performance analog I Cs; Paul
`has studied the chemistry, composition, and structure of thin-film
`materials and has invented masking schemes and optimized
`target compositions to obtain thin-film resistors with desired
`properties.
`
`Graduating in 1 963 from Rochester Institute of Technology, with
`a BS in Photographic Science and Engineering, Paul joined the
`GTE Development Laboratory in Bayside, NY, working in the
`new field of photolithography and mask-making for semiconductor
`manufacturing. After 5 years, he went to work at Sprague
`Electronics, in maskmaking and IC process development. He
`Cl
`joined Analog Devices in 1973.
`Dan Sheingold
`
`THE AUTHORS
`Rupert Baines (page 8) is a
`Senior Technical Marketing
`Engineer in Corporate Market(cid:173)
`ing, where he specializes in IC
`products for Communications.
`He has a BSEE from
`the
`University of Hull (United
`Kingdom) and an MBA from
`IESE in Ba reel o na (Spain).
`Hobbies include socializing,
`travelling, and crashing jet-skis.
`
`Mike Curtin (pp. 6-7) is a Senior Applications Engineer at our
`Limerick, Ireland, facility. His photo, and a biographical sketch,
`appeared in Analog Dialogue 27-1, I 993.
`
`(More authors on page 30)
`Cover: The cover illustration was designed and executed by
`Shelley Miles, of Design Encounters, Hingham M.A.
`
`Analog Dialogue
`
`One Technology Way, P.O. Box 9106, Norwood, MA 02062-9106
`Published by Analog Devices, Inc. and available ar no charge 10 engineers and
`scientists who use or think abou1 LC. or discrt1c analog, conversion, da1a handling
`and DSP circuits and systems. Corrc,ipondencc is welcome and should be addressed
`10 Editor, Analog Dialogue, at the above address. Analog Devices, Inc., has
`repm;cntativcs, sales offices, and distributors throughout the world. For information
`regarding our producis and their applicarions, you arc in,·i1cd 10 use the enclosed
`Business Reply card, write 10 the above address, or phone 617,937-1428,
`1-800-262-5643 (U.S.A. only) or fax 617-821-4273.
`
`2
`
`ISBN 0161-3626 ©Analog Devices, Inc. 1994
`
`Analog Dialogue 28-2 (19941
`
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. Exhibit 1025 Page 2
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. v. Analog Devices, Inc. IPR2020-01559
`
`

`

`Low-Distortio11 Mixer:
`+24 dBm 3rd-Order
`Intercept for Only
`-10 dBm LO Power
`500-MHz AD831 integrates a
`doubly balanced mixer, LO
`preamplifier, and output amplifier,
`all in a 20-pin PLCC
`by Bob Clarke and Barrie Gilbert
`The AD831.,. is a wide-dynamic-range monolithic mixer designed
`for use in demanding applications. Its low distortion features a
`+24-dBm third-order intercept point (IP3). In addition to the basic
`mixer circuit, it includes an output amplifier and a local-oscillator
`(LO) input amplifier (Figure !). The LO input amplifier is a
`limiting preamplifier, requiring only -10 dBm of local-oscillator
`drive. RF and LO frequencies up to 500 MHz can be used; at
`500 MHz, the third-order intercept exceeds +20 dBm.To simplify
`impedance matching, both the RF and LO ports provide
`differential high-impedance inputs that can be ac or de coupled
`and driven from either differential or single-ended sources.
`
`The AD83 I's low-distortion IF amplifier provides a low-impedance
`output voltage suitable for driving reverse-terminated filters. The
`gain of this amplifier, and thus the conversion gain of the mixer,
`can be increased using external resistors.
`
`The AD831 supplants diode-ring or FET-ring mixers, which could
`require as much as l watt of LO power for high-intercept
`performance. Such mixers require an external LO preamplifier to
`generate sufficient drive co switch the diodes or FETs into hard
`limiting-and shielding to minimize electromagnetic interference
`(EMD from the high-level LO. An output diplexer may also be
`needed for proper 50-ohm termination.
`
`The AD83 l, shown in simplified detail in Figure 2, comprises
`three functional blocks: a mixer, a LO preamplifier, and an output
`amplifier. The mixer is a doubly balanced Gilbert multiplier circuit,
`supported by a very linear, low-noise VII converter at the RF port,
`z
`z
`!!:
`
`<(
`
`18 COM}
`
`17 VFO
`
`OUTPUT
`AMPLIFIER
`
`OUT
`
`VN
`
`BIAS
`
`0
`Z
`C)
`
`Q.
`>
`
`Q. %
`Q.
`>
`0
`0
`....
`....
`~
`LOCALOSC.
`INPUT
`
`Figure 1. The AD831 consists of a mixer, LO preamplifier,
`and output amplifier in a 20-pin PLCC.
`*For technical d3t3 , use the reply card. Circle l
`
`Analog Dialogue 28-2 (19941
`
`and a high-gain, high-speed driver stage in the LO path. In the
`mixer core, the RF signal is multiplied by a square wave at the LO
`frequency. For sinusoidal RF signals, the mixer produces the sum
`and difference frequencies, plus a series of higher harmonics. Since
`the output will be filtered to remove all but the.sum or difference
`frequency, and only half the signal power appears in each, it would
`suffer a loss of about 6 dB. However, the amplitude of the
`fundamental in the multiplying square wave is 4/Jt (i .e., 2 .1 dB).
`Thus the net conversion loss is only 3.9 dB; the AD831 's output
`amplifier provides the gain required to restore the overall
`conversion gain to O dB.
`
`The LO preamplifier is a high-gain limiting amplifier, it accepts a
`low-amplitude input, which may be either sinusoidal or square(cid:173)
`wave-and converts it into a balanced square-wave output that
`drives the mixer core. The amplitude of a sine LO input can be
`reduced to -20 dBm (an amplitude of 32 mV) at local oscillator
`frequencies below 100 MHz, and to -10 dBm (100 mV) at
`500 MHz, without substantially reducing conversion efficiency.
`The IF oucput from the mixer core is available as either a differential
`low-level output at pins IFN and IFP, or as a single-ended high(cid:173)
`level voltage output at pin OUT. The differential outputs at IFN
`and IFP are from nominal 50-!l source impedances, and they
`provide I.he full 500-MHz output bandwidth when ac-coupled to
`a load via a transformer or capacitors. In down-conversion
`(difference-frequency) applications, a single capacitor connected
`between these terminals can implement a low-pass filter to reduce
`
`IN THIS ISSUE
`Volume 28, Number 2, 1994, 32 Pages
`Editor's Notes (New Fellows: Richie Payne & Paul Ruggcrio),Authors
`2
`Conipku, kM-dis1,mwn miur for fa.V/luma"es up lb 500 MHz (AD83 I) . . . . . 3
`lnstrummtation Appli,a1iJms of 1he New I'-L1 Co11vmers (AD7714 & 7716)
`. 6
`GSM chip ut implenima oD bauband /11111:rionJ (AD20msp4 I 0) . . . . . . . . . 8
`Trco new LVDT inu·rfac~ IC, with a11alog or digital owput (AD2S93, AD698) 9
`S.-nal-inpr,1 16-bit curw11-/oop DAG has choice of rang,:j (AD420) . . . . . . . I 0
`800-MHz i,p amp has high P<rfonna11u , Imo /J(llver, Imo ,osi (AD800 I)
`. . . 11
`Micropl,o,,e pnamplifim for audo--a brief guuk for 1mn . . . . . • • • • . . . . . 12
`New-Product Briefs:
`What's new in ND conveners, multiplexers, codecs . . . . . . . . . . . . . . 19
`What's new in DIA com·crcers, analog switches
`. . . . . . . . . . . . . . . . . 20
`What's new in op amps and analog multipliers . ... .... .... ...... 21
`What's new in interfaca, regulators, and temperature sensors
`. . . . . . 22
`A rea der notes-Comments on high-frequency signal contamination . . . 23
`Ask The Applications Enginc..-r-16: Usi11g Sigma-Delta cumxrurs, pan 11 24
`, . . . .. .. . . . .. . . . . .. . .. . . .. . . . . .. .. . . 30
`Wonh Reading, Authors
`Potpourri . . . . . . . . . . . . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . . . . . . 31
`
`3
`
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. Exhibit 1025 Page 3
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. v. Analog Devices, Inc. IPR2020-01559
`
`

`

`VP
`
`l50!l
`
`Figure 2. Simplified schematic of the AD831
`
`the amplitudes of the unwanted sum component, for decreased
`intermodulation in the IF amplifier.
`
`The gain of the AD83 l's output amplifier can be programmed
`using a pair of feedback resistors placed between the pins OUT,
`VFB, and COM. When OUT is tied directly to VFB, the small(cid:173)
`signal bandwidth, using the amplifier, exceeds 250 MHz. The
`amplifier's low de offset allows it to be direct coupled in such
`applications, and in some cases the AD83 l may be used in direct(cid:173)
`to-baseband applications.
`
`APPLICATIONS
`The AD831, with its low distortion and wide dynamic range,
`provides an ideal solution in receivers for multi-signal
`environments, in which the mixer must cope with an abundance
`of high-level desired and undesired inputs. In RF modems> two
`AD83ls can be used to make a quadrature (1/Q) modulator or
`demodulator. The -10-dB LO drive requirement and high input
`impedance allows the AD83 l to be driven from passive phase
`shifters or digital phase shifters constructed from ECL gates, with
`the additional advantage that small LO inputs minimize the risk
`offeedthrough and EMI.
`
`Since the AD83 l is de-coupled throughout, it can be used as a
`modulator. In a suppressed-carrier application, the carrier is
`applied to the LO port and the modulation signal is applied to the
`RF pore. The nominal full-scale signal level at the RF input is
`± 1 V ( + 10 dBm), but some overrange can be tolerated. As a simple
`amplitude modulator (AM), a baseband signal of0-to-1 Vapplied
`to one of the RF inputs controls the RF output. In these
`applications, the built-in gain of the output amplifier will result in
`3.9-dB more gain than in the mixing mode.
`The AD83 I is the newest of a growing family of linear parts for
`RF and IF applications from Analog Devices. Others include the
`AD600, AD602, and AD603 amplifiers for low-noise linear-in(cid:173)
`dB gain conaol, the AD734 and AD834 linear multipliers, the
`AD640 precision wideband log amp, and the AD606 logarithmic/
`limiting amplifier that provides both hard-limited IF and received(cid:173)
`signal-strength-indicator (RSSI) outputs.
`The ADS 31 requires ± 5-volt supplies or a 9-to-11-volt single
`supply, drawing 125 mA maximum (programmable) quiescent
`current. Housed in a 20-lead PLCC, it operates over a supply
`range of -40 to +85°C. An evaluation board is available. Pricing
`Cl
`in 1000s is $12.
`
`ABOUT MIXERS
`No one active in "RF-land" needs to be told what a "mixer" is
`or of its pivotal importance in determining the overall
`performance of an RF circuit, especially a receiver. However, a
`brief discussion of what mix.ers do, a couple of popular
`approaches, and some of the terminology will be helpful to
`readers who specialize in other disciplines but have a healthy
`curiosity about this ubiquitous function for which analog circuits
`are still of vital imponance.
`A mixer, as defined in the Dictionary of Electronics, by S. W.
`Amos, (Butterworths, 2nd edition, 1987), is "a device which
`accepts two inputs at different frequencies and generates an
`output at the combination frequencies. In particular, the RF
`mixer in a superheterodyne receiver accepts an input from the
`antenna (or an RF amplifier) and from the local oscillator and
`generates an output at the sum or difference frequency which
`is equal to the intermediate frequency."
`
`A mixer is a member of the class of nonlinear elements used
`for frequency translation. The most productive of these elements
`form a subclass sharing the property of signal multiplication(cid:173)
`in one way or another their output is the product of two inputs.
`For example, an analog multiplier-such as the Analog Devices
`AD834-produces an output precisely proponional to the
`instant-by-instant product of its two inputs.
`If the two inputs are ac sinusoids at frequencies /RF and /w, a
`trigonometric identity shows that their product will contain
`two sinusoids, at frequencies !RF + fw, and /RF - fr.o, with
`amplitudes equal to one-half the product of their individual
`amplitudes. For example, if /RF = 1.5 MHz and
`fw = 1.955 MHz, the sum and difference frequencies will be
`3.455 MHz and 455 kHz. A bandpass filter selects the signal
`at the desired frequency, for example, the AM broadcast band's
`455-kHz intermediate frequency (IF) .
`
`AMPUTUDE
`
`'
`
`p
`
`,p
`
`,_
`
`-
`
`f
`
`In general, the RF signal will not be a simple sinusoid; it will
`represent the sum of sinusoids having a range of frequencies in
`the neighborhood of a carrier, fRp0-for example,
`1.5 MHz ± 10 kHz. Each frequency term, when multiplied by
`the/w term, will produce sum-and-difference terms; the result
`will be two bands of frequencies at 3.455 ± 0.01 MHz and
`455 ± IO kHz. Again, a bandpass filter can be used to select
`the desired frequency band. Note that if fw is equal to the
`carrier,/RFO, the sum term will be centered at twice the carrier
`
`4
`
`Anal~ DialO£Ue 28-211994)
`
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. Exhibit 1025 Page 4
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. v. Analog Devices, Inc. IPR2020-01559
`
`

`

`AMPUTUDE
`
`I
`
`,, '
`
`II
`
`f01F
`
`-
`
`frequency and the difference term will reproduce the ± 10-kHz
`baseband signal associated with the carrier.
`Multiplication by a sinusoidal LO signal produces a fairly simple
`frequency spectrum, but the output amplitude (signal gain) is
`sensitive to the amplitude of the LO. For better precision,
`modulators that amplify and clip the LO signal-such as the
`AD83 l - are used to set the amplitude of the output (IF signal)
`to a value that is proportional only to the RF signal. However,
`the spectrum of a clipped waveform (square wave) contains a
`family of odd harmonics of fw; therefore, the multiplication
`produces the sum and difference of each of these harmonic
`components with each frequency present in the RF signal.
`Although they will be suppressed if the difference signal is
`chosen, the large number of frequencies present requires a highly
`linear operation, because nonlinearities can produce additional
`harmonics and cause intermodulation products to appear in the
`difference band.
`
`Key objectives in the design of a high-perfo~ance active mixer
`are to achieve a very linear RF input section, followed by a near(cid:173)
`ideal polarity-switching stage, followed by a very linear IF output
`amplifier prior to the first filter.
`
`AMPUT1JDE
`
`fRF - flo
`
`flo
`
`fRF • f1.0
`'RF
`3fl o-fRF
`
`3flo
`
`MIXER TERMINOLOGY
`Intermodulation distortion: If the input to an ideally linear
`circuit is a pair of sinusoids at frequencies / 1 and 12, the output
`will contain only those frequencies. The output of a circuit with
`nonlinear distortion, viewed with a spectrum analyzer, can
`contain such frequencies as 2f1 and 2h (second harmonic
`distortion), 3J1 and 3h (third harmonics), etc., as well as
`
`I,, + !2 I and 111 - Ii I (second-order intermodulation products),
`111 + 2/il, If, - 2h I, 12/1 +hi, and I 2/1 -hi (third order
`intermodulation), and so on. An ideal mixer's output will contain
`only those frequencies that are expected according to theory.
`A mixer with distortion will also contain harmonics and
`intermodulation-products of input and output frequencies. For
`small signals, distortion is generally negligible, but as the
`amplitude approaches saturation, distortion products increase
`rapidly.
`Third-order intercept point: A frequently used figure of merit
`for device linearity. The third-order intermodulation difference
`products (J1 - 212 and 2/1 - h) are those most likely to interfere
`with the desired signals. If device output power (unity slope)
`and third-order intermodulation (slope of 3) are both plotted
`against input power on a logarithmic scale, and the plots are
`then extrapolated until the two functions meet, the output power
`(dBm) at which the intersection occurs is called the third(cid:173)
`order
`intercept point. The larger this number, the
`better the device's overall linearity is presumed to be. The
`analysis on which this criterion is based assumes that distortion
`for a particular device can be modeled with a power(cid:173)
`series expansion.*
`
`THIRD-ORDER
`INTERCEPT POINT
`IS AT +2,t dBm
`=:J ldB WHERE 2 LINES CROSS
`
`E
`m
`1J
`
`~ 0
`
`-'18dBm
`INPUT-dBm
`
`Diode-ring mixer: A commonly used mixer topology for many
`years, employing a ring of four diodes, driven across one diagonal
`from the local oscillator, and across the other from the RF input.
`Both sources are transformer coupled, and the output is taken
`between the center taps of the transformer secondaries. The
`diodes may be silicon junction, silicon Schottky barrier, or
`gallium arsenide types. A major disadvantage of this circuit is
`the need for high drive power at the LO input, in order to ensure
`that the diode conduction is strong enough to achieve low noise
`and to allow large signals to be converted without excessive
`spurious nonlinearity. Other disadvantages include difficulty in
`impedance matching, coupling between ports, and insertion loss
`of this passive circuit.
`l·dB gain compression: The level of input power at which
`the output drops to l dB below the theoretical value-the "knee"
`of the output-input plot (see above figure).
`Conversion gain: The gain from the RF pon to· the IF port.
`Usually <O dB for passive mixers. Active mixers, such as the
`AD83 1, provide higher conversion gain (nominally O dB, but
`adjustable by choice of external resistors) and better port-to(cid:173)
`port isolation.
`
`"For a fuller explanation, see Analog~ 27-1 (I 993), "Ask the Applica(cid:173)
`tions Engineer-13:.Amplilicr distortion specs", by Walt Kester, pp. 27-29.
`
`Anal02 Dialoeue 28-2 11994)
`
`5
`
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. Exhibit 1025 Page 5
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. v. Analog Devices, Inc. IPR2020-01559
`
`

`

`using a divider with resistance values of24 kD. and 15 kn, as shown,
`for an AD77 l 4 reference voltage of 1. 9 2 V. Sioce the reference
`vole.age for the AD77 l 4 is derived from the same supply as the
`excitation voltage for the bridge, drifts of the bridge excitation
`voltage do not introduce scale-factor errors. With a programmed
`gain of 128, the full scale input span of the AD77 l 4 is 15 m V,
`which corresponds to the output voltage from the transducer for
`full-scale pressure of 300 mm Hg.
`
`EXCffATION VOLTAGE • •SV
`
`,sv - - - - - .
`AYco
`r:Noo
`
`our.
`
`AINI
`
`ADTT14
`
`--.... ------------------ -,
`l B ~·M)
`:
`• ..--~~-,r---, •
`•
`'
`CONVERTER
`PGA:
`AllfO-
`:
`ZEROED
`OtGITAL
`1
`!:-.,-
`Fl.TEA
`: MOOUUTOR
`
`,
`
`t-129 l Cl.OCK
`il~~ IIR~RI :
`
`A (cid:127)
`
`' - - -·- - - · - - - - - - - - - - - - - - - 1
`
`GENERATION
`r - -- - --- --- - --- - --- -- - - -
`SERIAL l~FACE
`:
`
`·-- --- ---- - - -- --- - ----..II
`
`IICU( . .
`MCU(OIIT
`
`SClJ(
`
`cs
`DIN
`D0UT
`
`lnstrun1entation
`Applications of
`New Sigma-Delta
`AID Converters
`Single-supply, low-power 24-bit AID
`converter digitizes low-level signals in
`industrial-control applications
`Mike Byrne
`The AD77l4 is the newest member of the AD77 Ix family of 24-
`bit sigma-delta converters (Aruwg Dialogue 26-1, pp. 7-9). Like
`other members of this high-resolution family, the part provides a
`programmable-gain front end and a programmable digital filter.
`It features very low noise (<300 nV rms at a gain of 128) and an
`input channel arrangement that accepts 3 fully differential-or 5
`pseudo-differential-inputs.
`
`The AD7714 is optimized for use in remote process-control and
`industrial-control applications. One important aspect of such
`applications is isolation, generally provided at the digital interface
`by opto-isolators. The smaller the number of digital interface
`connections, the easier it is to isolate the device, since each
`connection requires an opto-isolator. Designed with this in mind,
`the serial interface on a remotely powered AD77 I 4 can operate
`with just 3 wires: a serial clock, data ow, and data in lines.
`
`Industrial control environments are generally quite noisy and,
`since, the bandwidths of the signals to be measured are generally
`very low, the input is heavily filtered. The filter often includes a
`large decoupling capacitor on the input to the ND converter. Since
`the input to an unbuffered sigma delta converter is effectively a
`sampled capacitor syscem, source impedance associated with
`decoupling capacitors results in a gain error. The AD77 l 4
`on-chip buffer avoids this problem.
`
`Power requirements are also of concern. In many cases, the power
`to the remote transducer site is provided to devices in series in a
`4-20-mA analog data loop. In such applications, devices in the
`transducer circuit muse be able to operate on less than 4 mA. In
`other cases, with isolated digital oucputs, remote power may be
`furnished by batteries. Meeting both of these low-power
`requirements, the AD77 I 4 can operate from a 3-V power supply,
`consuming only 500 µA of supply current. In its power-down
`mode, standby current is typically less than IO µA.
`The programmable-gain front-end on the AD7714 allows it to
`handle unipolar analog inputs with ranges from 0 to +20 mV to
`Oto +2.5 V and bipolar inputs of±20 mV to ±2.5 V. Because the
`part operates from a single supply, the differential bipolar ranges
`are centered about an offset common-mode input.
`
`Figure l shows the analog connections for a typical application of
`the AD77 I 4 in pressure measurement, reading the differential
`output voltage from a Sensym BP0 I pressure transducer. With
`rated full-scale pressure (in this case 300 mm Hg) on the
`transducer, the full-scale differential output voltage from the bridge
`is 3 m V for each volt of excitation, or J 5 m V with a 5-V supply.
`The reference voltage is derived from the 5-V excicarion voltage,
`
`DONO BUffER
`
`POL
`
`ORDY RESET
`
`Figure 1. Pressure measurement using the AD7714.
`The AD77 I 4 is housed in 24-pin plastic & hermetic DIPs & SOI Cs
`and 28-pin SSOPs and is available for the -40 to +85°C and -55
`to +125°C temperarure ranges. Prices start at 813 (100s) .
`TheAD7714 was designed at the Analog De-vices facility in Limerick,
`Cl
`Ireland, by a team led by Pat Hickey.
`
`Sigma-delta techniques reduce hardware
`count and power consumption in
`biomedical analog front ends
`Mike Curtin
`Electrocardiograph (ECG) and electroencephalograph (EEG)
`equipment- and system designers have traditionally needed
`hardware-intensive input signal conditioning. The front-end
`circuits interface directly to signal transducers-electrodes on a
`patient's body-to measure heart or brain activity. The input signal
`cypically includes a variable de component of about 300 mV and
`a much smaller ac signal ofup to 10 mV pk-pk.A syscem's front(cid:173)
`end typically includes a low-noise instrumentation amplifier circuit
`(co scrip out the de component and amplify the ac signal), a 2nd(cid:173)
`order low-pass filter, and an AID converter. Single-channel EEG
`or ECG machines are rare; a typical system would have 8 channels,
`with 8 IA's, 8 filcers, and an 8-channel multiplexer, plus a 12-bit
`ADC. Such a system is hardware- and space-intensive, power(cid:173)
`hungry, and costly.
`
`An alternative to the traditional method of front-end signal
`processing is to use a low-cost, high-resolution ADC (about 5 or
`6 extra bits) to digitize the entire signal, including the variable
`offset. Then use the power of digital processing to extract the
`relevant information. For such an approach, the sigma delta
`conversion technique is very suitable.
`
`• The wide dynamic range and low noise performance of the sigma
`delta converter obviate the need for a high-performance
`instrumentation amplifier.
`• The analog low-pass filter can be reduced to simple anti-alias
`filtering for a sigma delta convener with suitable filtering.
`
`6
`
`Analog Dialogue 28-2 119941
`
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. Exhibit 1025 Page 6
`Xilinx, Inc. and Xilinx Asia Pacific Pte. Ltd. v. Analog Devices, Inc. IPR2020-01559
`
`

`

`• It is much easier to achieve a low-power front-end using sigma(cid:173)
`delta converters. Typical power consumption is much lower than
`for traditional AD Cs. This is very important for front-ends which
`operate from ao isolated supply.
`
`Some aspects of biomedical measurement-system design present
`problems similar to those found in industrial/process control. For
`example, ECG machines measure very small input signals in the
`presence of large de offset and common mode signals. The
`traditional solution has been to use expensive front-end circuitry
`including instrumentation amplifiers and low-pass anaJog filtering
`to condition these inputs.
`
`In addition, the ECG machine must process many channels. The
`standard 12-lead ECG requires various combinations of the signals
`from rune electrodes on the patient's body. The traditional way of
`doing chis has been to compute the differentials between electrodes
`by means of hardware, as shown in Figure 2: this requires a set of
`differential amplifiers in the front end.
`
`FAST
`12-8fT
`AOC
`
`OSP 0A
`IIICROCONTROU.£R
`
`9EOOLEAD$
`
`LA
`RA
`
`V3
`V4
`
`GROUNO
`REFERENCE
`
`VlRTUAl
`NEGATIVE
`
`Figure 2. Conventional 12-lead ECG system.
`
`The net result is a system that is hardware-intensive, expensive,
`and relatively inflexible. In a new approach, employing sigma-delta
`techniques, the AD7716«, described here, was created to satisfy
`the specific needs of ECG system design.
`Resolution requirements: If a sigma-delta solution is considered,
`it is possible to reconfigure and simplify the front end, compared
`to the depth of hardware required to implement the traditional
`12-lead ECG. However, by utilizing the high resolution of sigma
`delta converters, it is possible to consider a system whereby there
`is no need to take out the offset and common-mode signals in
`hardware. Instead, digitizing the complete signal to the required
`resolution will allow the ECG signal to be inteUigently separated
`by the system processor from the variable de baseline and other
`normal-mode noise.
`
`What resolution would be required?The maximum ECG signal is
`10 mV, and it must be resolved to 5 µV for an acceptable ECG
`plot. The typical offset voltage is 300 m V and so is the typical
`common mode voltage.Thus the total de component could exceed
`600 mV. Resolving a 600-mV signal to within 5 µV-a ratio of
`12 0, 000-requires a converter with a usable dynamic range
`corresponding to 1 7 bits. This is easily possible with sigma-delta
`techniques; the AD77 l 6 fu\fiUs this requirement.
`Bandwidth requirements: The bandwidth requirements for
`ECG machines depend on the application. There are two broad
`categories: diagnostic and monitor. The toughest requirement is
`for diagnostic systems, used in patient analysis and diagnosis. The
`
`American Heart Association standard requires that a diagnostic
`ECG have a flat 0.14 to 50 Hz frequency response, with maximum
`anenuation of3 dB at 0.05 and 100 Hz.
`
`The bandwidth requirements are not as strict for monitors. Here,
`accuracy to within a few dB in the 5-25-Hz band is sufficient.
`Reduced bandwidth helps the monitors resist disturbances caused
`by electrical noise, body movements etc., but it also restricts the
`amount of information in the ECG. Manufacturers choose the
`bandwidth to meet specific requirements for their machines, often
`a compromise between diagnostic and monitor standards.
`
`The AD7716 flexibly permits adjustment of the digital filter high(cid:173)
`frequency cut-off, from 584 Hz to 36 Hz in binary steps. More(cid:173)
`precise frequency setting, if needed, can be achieved by adjusting
`the device clock, since the cutoff frequencies scale in proportion.
`Easing multi-channel applications: The number of ECG
`channels can range from three electrodes, in the most basic ECG
`machine, to systems whkh convert signals from more than 16
`inputs. The AD7716 solution simplifies assembly of a multi(cid:173)
`channel system. Traditionally, designers use a multiplexer and fast
`ADC to deal with multiple inputs. However, multiplexing sigma(cid:173)
`delta converters is not practical; each time a channel is switched,
`the next conversion must wait for the digital filter to settle. This
`can require hundreds of milliseconds.
`
`The AD77 l 6 addresses this problem in two ways. First, each device
`has 4 modulators and 4 digital filters. Consequently, it can convert
`4 channels in parallel