Download Analog design in the information age

ANALOG DESIGN in the
INFORMATION AGE:
A Personal Overview
Barrie Gilbert
Northwest Labs
Portland, Oregon
A NA LO G
DEVICES
1
OUTLINE
A N A LO G
DEVICES
A Little Bit of Analog History
The View from ADI’s Ramparts
Old and New Technology Wars
Analog Circuits aren’t What They Seem
Some New Developments
Analog in the Post-Monolithic Age
2
ANALOG TECHNIQUES
EMERGED FROM THE
FIRST DAYS OF RADIO
3
IN THE CONTEXT OF
RECORDED HISTORY,
RADIO IS A “NEW” ART
We are only at the beginning of the
history of “wireless” systems, which
date back only to the beginning of
the Twentieth Century.
4
THE ROOTS OF RADIO
MY VIEWPOINT IS BASED ON WESTERN HISTORY
1774
Alessandro Volta of Como makes a chemical battery
1820’s Ampere, Oersted, Ohm, Henry .... electricity
1823 Baron Schilling perfects a signaling system using
five galvanometer needles; devised a coding scheme
5
1831
Michael Faraday discovers electromagnetic induction
1835
Morse develops his telegraph and a new code
1858
First Atlantic cable laid
1859
First American oil-wells (Pennsylvania)
A N A LO G
DEVICES
THE ROOTS OF RADIO ... cont
A N A LO G
DEVICES
1861
Johann Philipp Reis made a system to transmit tones
and coined the word “telephone”
Elisha Gray becomes founder of a company that will
later become Western Electric
6
1867
Nobel invents dynamite; later has regrets
1869
Mendeleyev makes periodic table of elements
THE ROOTS OF RADIO ... cont
A N A LO G
DEVICES
7
1873
Maxwell publishes his treatise on electromagnetism,
a mathematical validation of Faraday’s observations
interrelating magnetic and electrical phenomena
1876
Bell invents the voice telephone
1877
Edison invents the phonograph
1883
Edison effect noted (a crude thermionic diode)
1884
Paul Nipkow patented first TV system
THE ROOTS OF RADIO ... cont
A N A LO G
DEVICES
8
1887
Hertz demonstrates electromagnetic waves in Bonn
1888
Tesla’s first alternating current motors/generators
1892
Branly invents the coherer (a crude detector)
1895
Röntgen discovers X-rays
1896
Marconi demonstrates wireless telegraphy over 2km.
In the same year, Popov in Russia also sent a
wireless message
THE ROOTS OF RADIO ... cont
A N A LO G
DEVICES
9
1900
Planck proposes quantum nature of matter
1901
Marconi first transmits across the Atlantic; Bose
files patent on a strange microwave detector
1902
Caruso makes first phonograph record
1903
Wright brothers make first aircraft flight
1904
Henry Ford’s factory in first year of business;
Fleming invents an improved thermionic diode;
Bose’s patent issued (March, USP 755,840)
THE ROOTS OF RADIO ... cont
A N A LO G
DEVICES
10
1905
Einstein proposes theory of special relativity
1906
San Francisco earthquake; Rolls-Royce founded
1907
Discovery of blood types; cubism in Paris
1908
First Model-T rolls out of Ford
1909
Blériot makes first flight over English Channel
THE FIRST BROADCAST
On Christmas Eve, in 1906, Reginald A. Fesseden, at
Brant Rock, Massachusetts, modulated the 1kW output
of a 50kHz alternator, designed by Ernst Alexanderson,
by putting a microphone in series with the antenna!
Electronics 50th Anniversary issue
April 17, 1980, p. 75
In this same year, Lee De Forest invented the “Audion”,
the first vacuum-tube triode.
G. L. Archer, A History of Radio to 1926,
Stratford Press, N.Y., 1938, p. 11
11
A N A LO G
DEVICES
WIRELESS GETS SERIOUS
12
1912
Edwin Armstrong first demonstrates “regeneration”
1913
Neils Bohr writes “On the Constitution of Atoms
and Molecules”, Philos. Mag., 26, Vol. 1.
1914
Panama canal opens; World World I begins
1918
Armstrong devises the superheterodyne receiver
1920
Regular radio broadcasts begin in USA
A N A LO G
DEVICES
WIRELESS GETS SERIOUS
A N A LO G
DEVICES
1925
Lilienfeld proposes a triode device incorporating a
semiconductor layer (USP 1,745,175, filed 1928)
1926
Baird invents a practical TV system
1927
The word “Electronics” first appears in a paper by
Grondahl & Geiger: “A New Electronic Rectifier”
1930
Sam Weber launches Electronics magazine (April)
with contributions from Fleming, Millikan, Goldsmith.
Armstrong conceives of FM (frequency modulation)
13
INTERLUDE
14
A N A LO G
DEVICES
1935
Watson-Watt (UK) and Page (NRL) demonstrate RADAR
1938
First successful magnetron, later smuggled to MIT (1940)
1939
Haeff invents inductive output tube, a forerunner of the
klystron (100W CW, 450MHz, 35%, 10dB gain) for TV
1940
Hewlett & Packard open a garage operation in Palo Alto.
Later grows to HP
INTERLUDE
A N A LO G
DEVICES
15
1943
Kompfner invents traveling-wave tube, later perfected
by Pierce and Field at BTL
1944
Analog computers important and in widespread use
1945
Arthur C. Clarke writes “The Space Station: Its Radio
Applications”; in October, Wireless World publishes his
“Can Rocket Stations give Worldwide Radio Coverage?”
1946
ENIAC developed by Eckert and Mauchly; no thoughts
yet of combining it with a portable radio, however
THE TRANSISTOR AGE
16
1947
(December 16) Brattain and Bardeen (with help)
accidentally make a point-contact transistor
1950
Shockley’s “Electrons and Holes in Semiconductors”
published by Van Nostrand; alloy junction transistors
1951
Simulated emission in lithium fluoride using RF
1953
First JFETs (Dacey and Ross, Proc. IRE, Vol.. 41)
A N A LO G
DEVICES
THE TRANSISTOR AGE
A N A LO G
DEVICES
1954
MASER invented at Columbia University; first
International Solid-State Circuits Conference;
first regular color TV service starts in USA
...... and I start work
17
1957
(May 7) First MOS patent (Ross, USP 2,791,760);
models of an IC presented at RRE by Dummer;
Bob Noyce joins Fairchild
1958
Jack Kilby joins TI; by September he’d fabricated
a monolithic germanium oscillator and a flip-flop
1959: A TRANSITION YEAR
A N A LO G
DEVICES
Kurt Lehovec files patent describing the concept of
junction isolation (USP 3,029,366, issued April 1962).
The silicon planar technology is developed by Hoerni.
Robert Noyce of Fairchild hears of Kilby’s successes.
Jack Kilby files (February 6) for patent “Miniaturized
Electronic Circuits”.
Noyce files (July 30) patent for “Semiconductor
Device-and-Lead Structure”.
Shockley having commercial difficulties with his twoterminal (pnpn) switching devices.
18
THE “SOARING SIXTIES”
A N A LO G
DEVICES
19
1960
Maiman demonstrated first LASER at Hughes Corp.
1961
(March) Fairchild announces its Micrologic family
1962
First semiconductor LASERs and red LED’s arrive
1963
Institute of Radio Engineers merges with the
American Institute of Electrical Engineers to become
the Institute of Electrical and Electronic Engineers
(IEEE); notice that “Radio” is dropped from title
1964
100 transistors integrated in a 5-by-3mm chip (SSI);
RCA introduces first production process for MOS.
THE “SOARING SIXTIES”
A N A LO G
DEVICES
20
1965
Ray Warner moves from Moto to TI, joins Kilby’s team;
Tektronix establishes its own in-house Si IC fab.
First (successful!) comms-sat (Intelsat-1) operational
1966
Monolithic active mixers, multipliers and current-mode
circuits; first issue of Journal of Solid-State Circuits;
double-implanted high-frequency pnp transistors
1967
Berkeley develops SPICE; superintegrated circuits
appear, leading to invention of I 2L; Carrier Domain
devices first demonstrated
THE IC COMES OF AGE
A N A LO G
DEVICES
21
1968
Gusev et al. in the Soviet Union fabricated the first
double-implanted high-frequency pnp transistors
1969
Ion-implantation used at Mostek (by Sevin); first
microwave transistors with Arsenic emitters (Toshiba);
1000-transistor logic chips (MSI) become available
1970
Boyle and Smith of Bell Labs announce CCDs;
CMOS process is developed at Philips by Shappir;
Jobs and Wozniak start Apple Computers
THE IC COMES OF AGE
A N A LO G
DEVICES
22
1974
First 8-bit microprocessor from Intel
1975
First 10,000-transistor products (VLSI) rolled out
1977
Three mass-market personal computers arrive: the
Apple II; Radio Shack's TRS-80; Commodore PET
1980
Apple Macintosh; first camcorders; IBM’s PS/2
1981
IBM unveils their much-awaited PC; HP goes 32-bit
Current
Status
23
THE “PRODUCTIVITY ENGINE”
A N A LO G
DEVICES
Growing IC Markets
Very high R&D and
Manufacturing
Investments
24
Exponential trends
in Technology
and Productivity
Chatterjee & Doering, Proc. IEEE, Jan 1998
THE “RADIO ROUNDABOUT”
A N A LO G
DEVICES
Growing utilization
of spectral window
Push to provide
ever higher levels
of IC functionality
25
Escalating number
of radio techniques
and new standards
RADIOS ARE FOREVER
They have become a ubiquitous and
indispensable aspect of contemporary
life, but we’ve only just started. Much
wider use of wireless communication
devices is yet to come. Marriage with
ENIAC appears to be consummated.
26
l
l
l
l
l
l
l
l
l
l
27
AM/FM Radio
Cell phone
Pager
Car Keys
Radar Detector
Watch
Garage key
Collision Avd.
GPS
TV in back
ANALOG TECHNIQUES
DEVELOPED FROM
1
2
3
4
5
28
TELEPHONE
RADIO
RADAR
INSTRUMENTS
FIRST COMPUTERS
ANALOG TECHNIQUES
ARE IN TRANSITION
The first one hundred years are now behind us.
The recent marriage between the radio and the
computer has led to a new view of design.
New services for radio are in demand, with the
emphasis on providing data the most prominent.
Nature’s finite spectrum is now very crowded.
Integrated circuit capabilities and their limitations
are shaping many new analog techniques.
29
HOW WILL THEY EVOLVE?
A N A LO G
DEVICES
Trends in wireless systems:
l
Aggressive cost reduction
-- consumable radios?
l
l
Greater reliance on bits
More bandwidth - megadata
-- or maybe less: bit-dribblers
l
l
l
l
30
Prompt digitization in Rx
Total software flexibility
Increased use of CMOS
Spread spectrum
HOW WILL THEY EVOLVE?
A N A LO G
DEVICES
Many questions remain:
l
l
l
l
l
Do large mixed-signal analog-plus-digital
monolithic SoC’s make sense?
Is System-in-a-Package a better choice?
What changes will occur in the integrated
circuit industry affecting analog design?
Are bipolar transistors still needed?
What is the role of Silicon-Germanium?
- or GaAs HBTs, MESFETs, PHEMTs, etc.?
l
Spreading the spectrum further:
- is operation purely in the time domain practical?
31
SIGNAL FORMS
P(f)
Classical
Radio
t
f
P(f)
TimeDomain
(UWB)
t
Is Ultra-Wide-Band radio a viable alternative to CW-based radio?
Can UWB comfortably co-exist with CW, much like CDMA does?
32
f
Analog
Design
at ADI
33
TECHNOLOGIES at ADI
A N A LO G
DEVICES
l
LEADING-EDGE IN-HOUSE PROCESSES
- STRONG HF EMPHASIS inc SiGe
- COMPLEMENTARY BIPOLAR (CB)
- SILICON-ON-INSULATOR (SOI)
- THIN-FILM COMPONENTS
- LASER-TRIMMED CALIBRATION
l
CMOS. 0.6, 0.5. 0.35, 0.25 & 0.18µm
- ENHANCEMENTS FOR MIXED-SIGNAL
l
SEVERAL GENERATIONS OF BiCMOS
and even CBCMOS
34
ANALOG DESIGN TOOLS
A N A LO G
DEVICES
l
A FULL-TIME CAD TEAM SERVES THE
SPECIAL REQUIREMENTS OF THE
ANALOG DESIGN COMMUNITY
l PROPRIETARY SIMULATOR (ADICE5)
INCLUDES MANY NEW FEATURES
l POWERFUL POST-PROCESSING AND
PRESENTATION CAPABILITIES
l PROPRIETARY SCHEMATIC CAPTURE
-- all programs linked synergistically
35
ANALOG IC DEVELOPMENTS
A N A LO G
DEVICES
l
TWO KEY THRUSTS
MIXED-SIGNAL SYSTEMS-ON-A-CHIP
Large team efforts; strongly focused on major
system initiatives; aimed at providing low-cost
solutions; low-power consumption; high-volume
HIGH-PERFORMANCE COMPONENTS
Small-scale of integration; characterized by a
strong innovative content; new solutions to old
challenges; may use “boutique” technologies
36
SYSTEM-on-a-CHIP (SoC)
A N A LO G
DEVICES
The current working assumption is that the
cheapest solutions to highly integrated mixed
signal systems will always be fully monolithic.
This has been the most successful guiding
principle of the IC business.
37
SYSTEM-on-a-CHIP (SoC)
A N A LO G
DEVICES
The current working assumption is that the
cheapest solutions to highly integrated mixed
signal systems will always be fully monolithic.
This has been the most successful guiding
principle of the IC business.
However, this may not always be the most
efficient approach to mixed-signal products.
38
SYSTEM-on-a-CHIP (SoC)
A N A LO G
DEVICES
1.
In radios, and other specialized systems,
serious challenges stand in the way of
integrating sensitive analog sections with
“energetic” digital signal sections.
Spurious signal injection; VCO phasenoise disturbances; CMOS inverter-style
logic cells are especially troublesome.
39
SYSTEM-on-a-CHIP (SoC)
A N A LO G
DEVICES
2.
Disparate and conflicting demands are
placed on a common technology.
For example, while much has been done
to utilize submicron CMOS in radio, there
remain many aspects of performance that
lie beyond what can sensibly be delivered
by a single “one-size-fits-all” technology.
40
SYSTEM-on-a-CHIP (SoC)
A N A LO G
DEVICES
3.
High integration levels Reduce the overall die yield
- complex chips at the mercy of a few analog cells
Greatly complicate test development
and production throughput rate
Place immense demands on mixedsignal teams and therefore adversely
impact time to market
41
SYSTEM-on-a-HEADER (SoH)
A N A LO G
DEVICES
An alternative path may be the use of smaller die
combined into complete systems, using advanced
assembly technologies. Critical to the success of
this approach will be the use of highly-automated
(robot) handling of bare die.
This may be called the “System-on-a-Header”, or
“System-in-a-Package” concept.
This can be viewed as simply a natural extension of
the manufacturing techniques already widely used in
both the electronics and automobile industries.
42
SYSTEM-on-a-HEADER (SoH)
A N A LO G
DEVICES
SYNERGY:
ISOLATION:
FLEXIBILITY:
VARIETY:
TEST/YIELD:
TIME-to-MARKET:
MATERIAL RE-USE:
LOWER RISK:
LOWER COST:
43
Optimal IC technologies can be employed
Critical sections are more readily decoupled
Systems may be easily configured on demand
Diverse passive components can be included
The ICs are simpler, may be tested separately
Can be much quicker than VLSI development
Numerous existing, proven die may be utilized
Transferred from large, complex die to simple
assembly; mistakes can quickly be corrected
For many systems, total development cost
can be lower than that of a VLSI-SoC
SYSTEM-on-a-HEADER (SoH)
A N A LO G
DEVICES
The use of highly-automated assembly techniques,
in which a broad variety of optimal technologies are
synergistically combined, will herald a new phase in
the evolution of mixed-signal electronics. It will soon
be common to expect that the highest performance,
quickest time to market, and lowest development
cost will be achieved through multi-chip methods.
We are at the beginning of the
“POST-MONOLITHIC ERA”
44
But Technology
Wars are Still
Being Fought!
45
TECHNOLOGY WARS
A N A LO G
DEVICES
DIGITAL!
versus
ANALOG?
46
TECHNOLOGY WARS
A N A LO G
DEVICES
N
!
w
e
Bette
r!
Adv
ance
d!
DIGITAL
!
re!
u
t
u
F
e
h
T
versus
Sup
erio
r!
Tired
Obsolete
ANALOG?
Sta
le
47
Unstable
Old
TECHNOLOGY WARS
A N A LO G
DEVICES
Everyone knows that analog is “old stuff ”.
It’s unreliable, prone to drift and noise, it
goes out of calibration. It’s where all the
troubles are found in any equipment.
Communications systems will very soon
be entirely digital.
48
TECHNOLOGY WARS
A N A LO G
DEVICES
Everyone knows that analog is “old stuff ”.
It’s unreliable, prone to drift and noise, it
goes out of calibration. It’s where all the
troubles are found in any equipment.
Communications systems will very soon
be entirely digital.
Nothing could be farther from the truth.
49
BENEFITS of “DIGITAL” RADIO
A N A LO G
DEVICES
1. Digital modulation has immense advantages
(data compression and spectral efficiency; use
of error-correcting codes; ubiquity of digital data)
2. Digitization at the earliest point in a receiver
allows the use of very powerful DSP filtering
3. Some bulky analog filters can be eliminated
4. Similar kinds of benefits arise in transmitters
5. Software control of channel frequency, gain,
power control and other critical parameters
makes radio a “computer peripheral”
50
BENEFITS of “DIGITAL” RADIO
A N A LO G
DEVICES
6. Leads to the development of fully-programmable
transceivers (the “universal” or “software” radio)
7. Makes possible the use of sophisticated techniques
such as CDMA, even pure time-domain methods,
in which complex coding and decoding is needed
8. The overall management of cellular, satellite, cable
and fibre networks would be impossible without
complex digital supervision
9.
51
Leads to more innovative use of deep sub-micron
CMOS technologies and “digitally-friendly” circuit
techniques (such as sigma-delta converters) in
what were formerly strictly analog domains
NEVERTHELESS ...
A N A LO G
DEVICES
52
A
There is little likelihood of finding ways to totally
dispense with analog techniques in many areas
of modern electronic systems of all kinds.
B
Examples include: power sources; amplification;
many kinds of HF filtering; microwave mixers and
modulators; HF power generation; high-accuracy
measurement devices in medical instruments as
well as in radio; numerous kinds of VCOs for
carrier and reference frequencies; new photonic
interfaces and support of optical elements; etc.
C
Analog signal processing can often provide a
significant reduction in complexity (it is “elegant”).
SO.....
WHAT IS THE ROLE OF ANALOG
in the INFORMATION AGE?
It is a matter of common knowledge that electronics
has radically changed in character since the advent
of the microprocessor.
Analog techniques really have become a less central
aspect of modern electronics, even to the point where
it is believed by many they are no longer important.
Few young people pursue analog design as a hobby;
simple, basic materials are often hard to acquire.
53
A SENSIBLE SYNERGY
A N A LO G
DEVICES
DIGITAL!
teaming with
ANALOG!
54
Why is
Analog Design
“So Hard”?
55
QUESTIONS
A N A LO G
DEVICES
?
?
?
56
Why is analog design so often regarded
as especially difficult by new graduates?
Will there be enough analog designers in
the work force when the “first generation”
(those trained during World War II years)
have all gone? Where will they come from?
What can be done in universities to ensure
that classical concepts in electronics and
the linear signal-processing topics are not
neglected in the curriculum?
ASPECTS
ASPECTS of
of ANALOG
ANALOG DESIGN
DESIGN
Analog circuits are more intimately a part of the
physical world than are digital processors, since
they are concerned with manipulation of signals
having DIMENSION. Their physical attributes
are traceable directly to fundamental quantities
and constants, governing all real systems:
LENGTH:
MASS:
TIME:
CHARGE:
57
[L]
[M]
[T]
[Q]
Meter
Kilogram
Second
Coulomb
LENGTH:
MASS:
TIME:
CHARGE:
[L]
[M]
[T]
[Q]
Meter
Kilogram
Second
Coulomb
[MLT2L-2Q-1]
[QT-1]
[ML2T-3]
[T-1]
Volt
Ampere
Watt
Hertz
[ML2T-1Q-2]
[M-1L-2T2Q2]
[ML2Q-2]
Ohm
Farad
Henry
Dimensional signals:
VOLTAGE:
CURRENT:
POWER:
FREQUENCY:
Dimensional elements:
RESISTANCE:
CAPACITANCE:
INDUCTANCE:
58
THE “BIG DIFFERENCE”
A N A LO G
DEVICES
ANALOG DESIGN DIFFERS FUNDAMENTALLY
FROM DIGITAL DESIGN MAINLY BECAUSE IT
IS AN INSEPARABLE PART OF THE PHYSICAL
WORLD. IT IS “NEWTONIAN”.
DIGITAL LOGIC USES ELECTRONICS BECAUSE
SILICON ALLOWS THE FABRICATION OF FAST
AND COMPLEX ALGORITHMIC ENGINES AT
VERY LOW COST. IT IS ONLY INCIDENTALLY
ELECTRONIC. DIGITAL DESIGN IS NOT OFTEN
CONCERNED WITH CRITICAL CIRCUIT ISSUES.
59
THE “BIG DIFFERENCE”
A N A LO G
DEVICES
ANALOG CIRCUIT DESIGN INVOLVES VERY
FEW COMPONENTS, YET THESE CAN LEAD
TO VERY COMPLEX MATHEMATICS. THIS IS
ESPECIALLY TRUE WHEN ALL NONLINEAR
EFFECTS ARE CORRECTLY INCLUDED.
NUMEROUS SUBTLE MECHANISMS MATTER
BUT THEY ARE OFTEN NOT ADDRESSED IN
TEXTBOOKS: FOR EXAMPLE, THE EFFECTS
OF VARACTOR PARASITIC CAPACITANCES
IN RF AMPLIFIERS.
60
ANALOG ARRAY PROCESSOR
A N A LO G
DEVICES
Iw1
Ix1
Iw2
Ix2
Iwk
IwN
Ixk
IxN
Iy
Iwk =
61
Ixk
Σk Ixk
Iy
VECTOR SUMMATION
A N A LO G
DEVICES
IX
IZ
(OUTPUT)
(INPUT)
IY
(INPUT)
IZ = √ IX2 + IY2
THE “BIG DIFFERENCE”
A N A LO G
DEVICES
TOPOLOGICAL (COMBINATORIAL) RICHNESS
OF SIMPLE ANALOG CIRCUITS QUICKLY LEADS
TO EXPLOSIVE VARIETY.
FOR EXAMPLE, THERE ARE TWENTY-FOUR
DISTINCTLY DIFFERENT CIRCUITS THAT CAN
BE DEVISED USING ONLY TWO TRANSISTORS!
BY THE TIME THREE OR FOUR TRANSISTORS
ARE USED, THE NUMBER OF COMBINATIONS
BECOMES HUNDREDS.
63
THE “BIG DIFFERENCE”
A N A LO G
DEVICES
TOPOLOGICAL (COMBINATORIAL) RICHNESS
... cont
NOT ONLY ARE THE NUMBER OF DIFFERENT
CIRCUITS VERY LARGE BUT THE BEHAVIOUR
QUICKLY BECOMES COMPLICATED AND MAY
NOT BE READILY APPARENT BY INSPECTION
64
KERMIT_6
A N A LO G
DEVICES
IOUT = I1 − I6
I1
Q1
VIN+
2e
1
I6
+
Q1
14e
2.9
Q3
4.7
Q4
33e
33e
2.9
Q5
14e
1
Q6
VIN−
2e
IT
This is a very linear transconductance
whose magnitude is proportional to IT
Output Currents
I1
I6
gm
2% of IT
VIN
VERY LINEAR CLASS-AB BEHAVIOUR
gm constant to within ±0.05dB to VIN = ±0.65V
A N A LO G
DEVICES
PARAMETRIC SLAVERY
A N A LO G
DEVICES
The deeply-physical nature of analog circuits places
quite different demands on a manufacturing process
to those needed for the fabrication of digital circuits,
primarily due to the unavoidable dependence on the
absolute values of numerous parameters.
Achieving design robustness – which requires the
systematic elimination (or at least, the minimization)
of sensitivities to all process parameters – poses
unique challenges. Design for Manufacture in the
analog world demands unrelenting attention to a
very large number of small details.
67
PARAMETRIC SLAVERY
A N A LO G
DEVICES
This is commonly known as the PVT challenge –
Process, Voltage and Temperature must all be
made to have minimum impact on performance.
Sound design practice, including careful choices
of product architecture and technology, and cell
topology, can lower the risk of parametric failures.
But thorough simulation of the full range of PVT
conditions is essential in analog practice, and
invariably this work is extremely time-consuming.
It also requires excellent device modeling, to a far
greater extent than needed for digital design.
68
PARAMETRIC SLAVERY
A N A LO G
DEVICES
Thus, using the CAD model libraries for the “fast”,
“slow”, and “nominal” device models (of all kinds,
not only transistors), and operating temperatures
of, say, –60°C, 30°C and 120°C one generates a
set of nine results; repeating at a supply voltage
of, say, 2.6V, 3V and 3.6V requires twenty-seven
simulation runs, for each of perhaps fifty or more
key performance aspects. Any time a change is
made in the design – even a small one – these
must be repeated all over again.
69
DESIGN COMPRESSION
A N A LO G
DEVICES
The impact of this tedious and time-consuming
verification work on the circuit design schedule
can be catastrophic, if not fully anticipated.
This leads to the realization that the bulk of the
design must be compressed into the first week
or two of a development schedule.
The latter two weeks must be reserved to prepare
for a Design Review and hand-over to Layout.
70
BACK EXTRACTION
A N A LO G
DEVICES
Only after the IC layout has been completed can
the capacitances of interconnections, and some
of their resistances, be extracted and added back
into a new set of simulation studies, which need to
be just as thorough as those during pre-layout.
These parasitics (along with others, for example,
the package reactances and ESD protection) are
of pivotal importance in RF-IC design. Experience
teaches that the behavior of almost any circuit will
be radically different when they are included. Time
is thus needed to pursue further remedial action.
71
NEW ANALOG FRONTIERS
A N A LO G
DEVICES
Fiber systems have heralded a return to many basic
physical principles and device development, being
essentially “Newtonian”, as are RF systems:
l
l
l
l
l
l
l
72
Dispersion compensation in optical fibers
All-optical erbium-doped fiber amplifiers
Multi-wavelength channel splitters
Power-sampling partial splitters
Laser-diode drivers and modulators
Micromechanical (MEM) multiplexers
Supervisory functions of many kinds
ANOTHER “WAR”
A N A LO G
DEVICES
squea
k
BIPOLAR?
versus
fizz
CMOS!
73
XFCB BIPOLAR TECHNOLOGIES
A N A LO G
DEVICES
HIGH-PERFORMANCE COMPLEMENTARY BIPOLAR
l Ultra-Low Inertia:
Small transit time and low capacitances
Substrate capacitance not a varactor
l Trench-Isolated; Bonded Wafers (SOI)
l Three processes in full production
l Laser-trimmable thin-film resistors
l High-quality metal-plate capacitors
l Excellent device modeling
74
WHY CONTINUE TO BE CONCERNED
WITH BJT’S − WHEN AN INCREASING
NUMBER OF MODERN SYSTEMS ARE
TURNING TO CMOS TECHNOLOGIES
?
FOR RF SIGNAL PROCESSING
75
A N A LO G
DEVICES
Highest Performance Devices will
Continue to use BJT Technologies
A N A LO G
DEVICES
BSC/
BSC/
PSTN
PSTN
Audio Interface/
Processing
Baseband
RF/IF
Codecs
DSP
MixedSignal
A/D & D/A
Converters
Amplifiers
Mixers
Modulators
LNAs, VGAs
VCO/Synth
Log Amps &
other RF M&C
76
A N A LO G
DEVICES
Highest Performance Devices will
Continue to use BJT Technologies
A N A LO G
DEVICES
BSC/
BSC/
PSTN
PSTN
Audio Interface/
Processing
Baseband
RF/IF
Codecs
DSP
MixedSignal
A/D & D/A
Converters
Amplifiers
Mixers
Modulators
LNAs, VGAs
VCO/Synth
Log Amps &
other RF M&C
CMOS territory
77
BJT, HBT,
MESFETs
A N A LO G
DEVICES
Highest Performance Devices will
Continue to use BJT Technologies
A N A LO G
DEVICES
BSC/
BSC/
PSTN
PSTN
Audio Interface/
Processing
Baseband
RF/IF
Codecs
DSP
MixedSignal
A/D & D/A
Converters
Amplifiers
Mixers
Modulators
LNAs, VGAs
VCO/Synth
LOGICAL,
ALGORITHMS
78
Log Amps &
other RF M&C
NEWTONIAN
A N A LO G
DEVICES
BJT MODELING
CONTRARY TO SOME VIEWPOINTS, THE BJT
IS FAR EASIER TO DESCRIBE IN MODELING
EQUATIONS, COMPARED TO CMOS. WHILE
THE EQUATIONS ARE STILL COMPLEX, THEY
ARE FUNDAMENTALLY SOUND, OVER A VERY
WIDE RANGE. USING SOI SUBSTRATES, THE
BJT IS A PURE, THREE-TERMINAL DEVICE; ITS
TINY COLLECTOR-SUBSTRATE CAPACITANCE
IS NOT VOLTAGE-DEPENDENT. THERE ARE
NO OTHER SUBSTRATE SENSITIVITIES.
79
A N A LO G
DEVICES
ADVANCED MODELS
FOR MANY MICROWAVE APPLICATIONS IT IS
NECESSARY TO INCLUDE SPECIAL ASPECTS
OF BJT DEVICE BEHAVIOR. EXCELLENCE IN
THIS ARENA, CENTERED ON IN-HOUSE CAD
TEAMS, PROVIDES A MAJOR COMPETITIVE
ADVANTAGE. FOR EXAMPLE, THE BJT SUPERMODELS INCLUDE DISTRIBUTED ASPECTS
OF DEVICE BEHAVIOR. NUMEROUS OTHER
SUBTLE DETAILS ARE GENERALLY NOT OF
IMPORTANCE AT LOWER FREQUENCIES.
80
A N A LO G
DEVICES
HOWEVER . .
DEEP SUB-MICRON CMOS PROMISES
TO PROVIDE SOME PERFORMANCE
ADVANTAGES
(BETTER LINEARITY,
ABSENCE OF SHOT NOISE) THOUGH
NOT YET FULLY PROVEN; DIE COST
IS RARELY OF MATERIAL CONCERN
IN SSI AND EVEN LSI RF PRODUCTS
81
A N A LO G
DEVICES
HOWEVER . .
MANY OF THESE BJT CELL DESIGNS
HAVE ALREADY BEEN ADAPTED TO
CMOS FORM, AND IN GENERAL THE
THEORY IS DIRECTLY APPLICABLE;
THE “GAP” BETWEEN BJT AND CMOS
DESIGN PARADIGMS IS NARROWING
82
A N A LO G
DEVICES
HOWEVER . .
IN ULTRA-SHORT-CHANNEL CMOS
TRANSISTORS (BELOW 0.2µm) THE
SUBTHRESHOLD REGION EXTENDS
UP TO CURRENTS AT WHICH QUITE
HIGH fT’s ARE POSSIBLE, WHEN
CMOS è BJT
83
A N A LO G
DEVICES
CMOS SHORTFALL
l
l
l
PRESENT-DAY (0.35-0.18µm) CMOS
DEVICES STILL HAVE LOWER gm/C
RATIOS THAN COMPARABLE BJTs
LOW-FREQUENCY NOISE AN ISSUE
NONLINEAR OUTPUT CONDUCTANCE
- can quickly lose that nice gm linearity at the load
l
l
l
84
MANY SUBSTRATE SENSITIVITIES
EXTREMELY COMPLEX BEHAVIOUR
CONSEQUENTLY, POOR MODELING
A N A LO G
DEVICES
A BIT
About
the BJT
85
50 YEARS AGO
... this was the commonplace view of how the transistor works
iC=βiB
R
BASE
BATTERY
iB
COLLECTO
R
BATTERY
A N A LO G
DEVICES
50 YEARS AGO
... this was the commonplace view of how the transistor works
iC=βiB
R
BASE
BATTERY
iB
COLLECTO
R
BATTERY
The Beta View
A N A LO G
DEVICES
BUT LOOK CLOSER ........
?
A N A LO G
DEVICES
THE
THE BJT
BJT AT
AT HIGH
HIGH MAGNIFICATION
MAGNIFICATION
A N A LO G
DEVICES
THE HEART OF THE BJT
l
IC = IS(T) exp VBE / VT
l
VBE = VT log IC / IS(T)
l
ONE EQUATION; TWO FORMS
A N A LO G
DEVICES
TRANSCONDUCTANCE
l
IC = IS(T) exp VBE / VT
l
dIC / dVBE = { IS(T) exp VBE / VT
l
} / VT
gm = IC / VT
l WHICH IS LITTLE SHORT OF MIRACULOUS
!
A N A LO G
DEVICES
MULTIPLICATION
USING A SLIDE-RULE
Length ∝ 100
1
10
100
1000
SCALE ‘A’
x
1
10
100
1000
SCALE ‘B’
y
20
´
5
A N A LO G
DEVICES
MULTIPLICATION
USING TRANSISTORS
Ix
Iy
Vx ∝ log Ix
Vz ∝ log IxIy
Vy ∝ log Iy
A N A LO G
DEVICES
THE ‘‘BAND-GAP’
BAND-GAP’
EQUATION FOR VF(IF,T)
where
- H{VEN H
- η log H )}
is the ‘Hotness factor T/ TN
VEN = EGE - VN
{
VF = EGE
IF
VTN (log
IN
a temperature-independent
constant for a given device
A N A LO G
DEVICES
VFF versus TEMPERATURE
EGE
~ 1.2V
THE SLOPE IS DETERMINED
BY THE CURRENT DENSITY
~ 0.6V
VF
0
T (K)
0
300K
A N A LO G
DEVICES
TRANSLINEAR DESIGN
A N A LO G
DEVICES
>
Exploits a fundamental and unique
property of the bipolar transistor 
Transconductance (gm) is reliably a
linear function of collector current:
gm = IC / VT
>
>
96
where VT = kT/q
The basis of almost all analog multipliers
and numerous VGAs since the late ’60’s
Current-mode signal processing is often used
A N A LO G
DEVICES
TRANSLINEAR DESIGN
A N A LO G
DEVICES
> In short, the bipolar transistor retains certain unique
properties not matched by any CMOS technology, in
particular, the remarkable benefits of “translinearity”.
> In many difficult areas of wireless systems, the BJT
(or its more sophisticated cousin, the HBT) provides
significant performance advantages over CMOS, for
example, in low-noise amplifiers and power amplifiers
(where GaAs HBTs are proving valuable).
> The highly predictable and mathematical nature of the
bipolar transistor can be relied on to provide very high
levels of accuracy and temperature stability.
97
A N A LO G
DEVICES
ANALOG DESIGN in the
INFORMATION AGE must use
PARTNERSHIPS of TECHNOLOGY
ANALOG teaming with DIGITAL
BIPOLAR teaming with CMOS
SOC’s teaming with SOH’s
98
A N A LO G
DEVICES
Some BJT
“Tricks”
99
CLASSIC
CLASSIC MIXER
MIXER
IF
LO
Gm
RF
BIAS
CLASSIC MIXER
DEVICE
SIZES ?
IF
ZOUT ?
LO FEEDTHROUGH ?
LO
DC LEVELS ?
NOISE ?
RF
ZIN ?
LINEARITY ?
BIAS?
CLASSIC MIXER
IF
LO
RF
LINEARITY ?
(it is good enough?
how good is good enough?
how is it improved?)
The ‘MULTI-tanh’ TECHNIQUE
A N A LO G
DEVICES
>
>
>
>
103
A special class of translinear circuit, which
does not invoke current-mode techniques
Achieves reductions in distortion through
the use of a multiplicity of tanh functions
Very effective in certain circumstances
Once an academic curiosity, multi-tanh
cells are now widely used
MT-DOUBLET in a MIXER
A N A LO G
DEVICES
LO INPUT
Q1
RF INPUT
Ae
Q3
Q2 e
e
104
Q4
Ae
SUMMATION OF THE
TWO tanh SECTIONS
IOUT
1+2
+ IT
-VTlog Α
1
2
VTlog Α
- IT
VRF
INCREMENTAL GAIN of DOUBLET
A N A LO G
DEVICES
À+Á
À
VIN= -6VT
106
Á
VIN= 0
Example, for Α = 4
VIN= +6VT
VIN
Doublet: Relative Gain and HD3 vs. Α
dB
0
-1
-2
-3
-4
-5
-6
-7
-90
-100
-110
dB -120
1mV input
-130
-140
1
2
3
4
Α
5
6
7
Harmonic Signature for Doublet, Α = 4
0
H1
-50
H3
dBc
HD3
-100
Null at VT
-150
-50dBm
(1mV)
-25dBm
17.8mV
INPUT LEVEL
0dBm
0.32V
ACHIEVING LINEAR-in-dB GAIN
A N A LO G
DEVICES
IOUT
gm CELL
VG = 2.98mV/dB @ 300K
+
(PTAT, typ. 100µA)
Q2
QD
VG
IT
MIP exp
109
PRIMARY BIAS
IP
Q1
RF/IF
INPUT
-VG
VT
-
QB
Me
RB
+
RG
QE
e
IG
GAIN-CONTROL BIAS
(PTAT, typ. 0 to 50µA)
MULTI-tanh TRIPLET
A N A LO G
DEVICES
IOUT
Q1
Ae
Q2
e
e
RF/IF
INPUT
I
110
Q3
Q5
e
Q4
e
KI
Q6
Ae
I
INCREMENTAL
INCREMENTAL ggm
m
of
of aa typical
typical TRIPLET
TRIPLET
1+3+2
1
VX= -100mV
3
±0.043dB
@ ±56mV
2
A = 13, K = 3/4 VX= +100mV
Harmonic Signature for Triplet, A=13, K=3/4
0
ω
-50
3ω
dBc
-100
-150
-50dBm
(1mV)
-25dBm
17.8mV
0dBm
0.32V
A N A LO G
DEVICES
TRIPLUS: MULTI-tanh TRIPLET+
A N A LO G
DEVICES
IOUT
MULTI-tanh
TRIPLET
Q1
Ae
RF/IF
INPUT
CDEC
Q3
Q2
e
e
Q5
Q4
e
Ae
QB
QC
Ne
Me
Ne
RE
IP
Q6
e
QA
PRIMARY
BIAS
+
RG
GAIN
QE BIAS
e
RE
RB
IG
113 THIS SPECIAL BIASING SCHEME EXTENDS DYNAMIC RANGE
BIAS CURRENTS in INNER/OUTER gm CELLS
A N A LO G
DEVICES
3mA
2mA
1mA
INCREASING GAIN
CURRENT IN INNER gm PAIR
- PERFECTLY EXPONENTIAL
CURRENTS IN OUTER gm
PAIRS; NOT EXPONENTIAL
RATIO
= N/M
114
IG=0
INCREASING IG
TRIPLUS Incremental Gain vs. Instantaneous VIN
IN
30
25
MAX. GAIN
20
15
10 dB
5
MIN. GAIN
0
-5
-100 -80
-60
-40
-20
0
20
VIN in mVP
115
40
60
80
100
Harmonic Signature for an Optimized Quadlet
0
ω
-50
3ω
dBc
HD3 = - 80DdBc
-100
EH3
-150
-50dBm
(1mV)
-40dBm
3.2mV
-25dBm
17.8mV
0dBm
0.32V
A N A LO G
DEVICES
MICROMIXER
IF
LO
I0
Me
RF
e
ZIN
Me
IZ
Me
e
MICROMIXER
IF
LO
RF
ZIN
IZ
BIAS CURRENT
UNIQUELY
DETERMINES
IMPEDANCE
L-BAND MICROMIXER
94µA
IIF
132µA 94µA
2.3V min.
15k
VLO
QB5
QM1 QM2
QM3
QM4
16e
Q1
2k
12n
3p
8k 8k
QB4
16e
50Ω @
1.6GHz
VRF
10k
8p
3e
Q3
16e
20p
12n
2.5p
12n
2k
Q2 16e
1.18mA 1.18mA
e
QB1
QB2
e
10p
86µA
8k
QB3
12e
1k
0V
68µA 60µA 94µA
Log-Amps
120
LOGARITHMIC AMPLIFIERS
A N A LO G
DEVICES
“LOG-AMPS” ARE NONLINEAR CIRCUITS WHICH
CONVERT A WIDE-DYNAMIC RANGE SIGNAL ON
A DECIBEL SCALE TO A QUASI-DC VOLTAGE OF
SMALL RANGE: -70dB to 0dB becomes 0 to 1.4V.
WHEN DESIGNED WITH SUFFICIENT CARE TO
PROVIDE CALIBRATED OPERATION, THEY ARE
HIGHLY VALUABLE MEASUREMENT DEVICES.
121
LOGARITHMIC AMPLIFIERS
A N A LO G
DEVICES
N stages (typically 5-10)
(Current-mode signal)
VOUT
DET
RFIN
DET
10dB
DET
10dB
DET
10dB
DET
10dB
OFFSET
COMP’N
Log-Amp based on Progressive-Compression
LOGARITHMIC AMPLIFIERS
A N A LO G
DEVICES
•
•
•
•
•
•
•
•
123
Unique Nonlinear Function
Integrated Multistage Systems
Calibrated Slope and Intercept
Provide Complete Solutions
-- easy to use
Up to 100 dB Dynamic Range
Covering DC - 3500 MHz
Limiter Versions for PSK, FSK
Low Cost, Small Packages
AD606
AD608
AD640
AD641
AD8302
AD8306
AD8307
AD8309
AD8310
AD8313
AD8314
AD8315
AD8316
A Personal Goal: Make Log Amps as
Cheap, and Easy to Use, as Op Amps
A N A LO G
DEVICES
2.7-7.5V
4.7Ω
10nF
N.C .
SIGNAL INPUT
-73dBm to +17dBm
(-86dBV to +4dBV)
to +/-1dB error pts.
(100dB to +/-3dB)
INP VPS ENB INT
AD8307
INM COM OFS OUT
N.C .
LOG OUTPUT
0.3 to 2.3V
20mV/dB
Logarithmic Conformance of
AD8307 (AD8310 is similar)
A N A LO G
DEVICES
2.5
10 MHz
2
2
ACTUAL
OUTPUT
500 MHz
1.5
1
0
-1
1
-2
-73dBm
= -86dBV
0.5
= 49µV
RMS
-80 -70
125
Output Voltage
Error in dB
100 MHz
-60
-50
-40
-30 -20
-10
0
10
Equivalent Input Power in dBm (50Ω termination)
17dBm
= 4dBV
0 = ± 2.2V
20
120+ dB MEASUREMENT SYSTEM
50Ω INPUT
-105dBm
to +15dBm
R1 432kΩ
R2 64.9kΩ
10.7MHz
BANDPASS
FILTER *
VP , 5V
4.7Ω
0.1µF
0.65V
N.C .
GPOS
L1
750nH
R3 330Ω
GNEG
VOUT
AD603
VINP
C1
150pF
VPOS
COMM
VNEG
FDBK
VN , -5V
R5
100kΩ
INP VPS ENB INT
AD8307
R4 464Ω
VR1
5kΩ
Intercept
Adjust
±5dB
Exponential feedback
INM COM OFS OUT
0.3-2.3V
N.C .
1nF
R6 20kΩ
* E.g., Murata SFE10.7MS2G-A
R7
80.6kΩ
OUTPUT
10mV/dB
A N A LO G
DEVICES
RF Power
Detectors
127
TRUE POWER MEASUREMENT
A N A LO G
DEVICES
LOGARITHMIC AMPLIFIERS MAY BE USED TO
ACCURATELY INDICATE EQUIVALENT POWER
IN A CERTAIN SYSTEM IMPEDANCE BUT THEY
DO NOT MEASURE TRUE POWER.
THAT IS, THEY DO NOT RESPOND TO THE
TRUE MEAN-SQUARE
VALUE OF SIGNALS OF ARBITRARY WAVEFORM
128
TRUE POWER MEASUREMENT
A N A LO G
DEVICES
A COMPLEX RF SIGNAL
Q:
129
What is the true power?
TRUE POWER MEASUREMENT
A N A LO G
DEVICES
THE ACCURATE MEASUREMENT OF THE
ROOT-MEAN-SQUARE (RMS)
VALUE OF SIGNAL OF ARBITRARY WAVEFORM
CAN BE ACHIEVED IN TWO WAYS:
k THERMAL DETECTORS
k ANALOG COMPUTATION
130
TRUE POWER MEASUREMENT
A N A LO G
DEVICES
k
THERMAL DETECTORS
l Fundamentally correct
l Very slow - milliseconds
l Small dynamic range
l Difficult to integrate
- need MEM structures
131
TRUE POWER MEASUREMENT
A N A LO G
DEVICES
k
ANALOG COMPUTATION
l Very accurate with good design
l Can have large dynamic range
-
30dB for direct squaring
up to 100dB using new methods
l Output can be linear-in-dB
l Low Voltage and Power (<10mW)
132
with chip enable
A BASIC METHOD
A N A LO G
DEVICES
Intermediate current
IO (VIN /VR )2
RF INPUT
VIN
x
VOUT =
2
2
AVE
CO
WIDE-BAND SQUARE-LAW
TRANSCONDUCTANCE
CELL (BJT TECHNIQUES)
{VIN}
RO
where
VO = VR2/IO RO
This simple structure produces the Mean-Square of VIN
with an averaging time determined by the product CO RO
133
VO
A BASIC METHOD
A N A LO G
DEVICES
k
ADVANTAGES
l
l
l
k
DISADVANTAGES
l
l
l
134
Very simple cell design
Can fit into a 1.5V supply
Wideband -- to >6GHz
Low dynamic range capacity
Higher dynamic range of output
Complicated scaling mechanisms
DIFFERENCE-OF-SQUARES
A N A LO G
DEVICES
IO (VIN /VR )2
RF INPUT
x
VIN
2
IDENTICAL
CO
RO
SQUARING
CIRCUITS
VOUT
x
RO
2
High-gain
amplifier
+
-
VOUT =
AVE{VIN}
IO (VOUT /VR )2
Directly produces the Root-Mean-Square of VIN with
an averaging time determined by the product CO RO
135
2
DIFFERENCE-OF-SQUARES
A N A LO G
DEVICES
k
ADVANTAGES
l
l
l
Still a very simple design
No change in bandwidth
Computes RMS directly
-- scaling of cells cancel
l
k
DISADVANTAGE
l
136
High output drive capacity
Still a low dynamic range capacity
AD8361
k
A COMPLETE IC POWER DETECTOR
l
l
Uses a different squaring cell design
About the same bandwidth (>3GHz)
-- limited mostly by package
l
Basically unlimited signal range
-- due to Class AB operation
l
l
l
137
Precisely-defined input impedance
Can provide over 30dB dynamic range
Only 6mW quiescent consumption
A N A LO G
DEVICES
80dB Log-RMS Voltmeter
A N A LO G
DEVICES
Variable-gain amplifier, 0-80dB
3-pole low-pass, with corner at 2MHz
+5V
+5V
∆=40dB
∆=40dB
2nF
VIN
+1
1nF
220Ω
XAMP
+5V
220pF
1µF
15kΩ
SREF
1µF
VPOS
+
VRMS
∞
FLTR
Each VGA section is 1/2 AD605; the
op amps are 1/2 AD8032; all operate
from +5V; a single 2.5V reference (not
shown) sets slope VY & intercept VZ.
A few components omitted for clarity.
Set-point,
~1.2V
AD8361
COMM
VOUT = VY log (VIN / VZ )
IREF
20mV/dB
RFIN
VOUT
PWDN
20µV 200mV
rms
XAMP
100Ω 100Ω
A Network
Analyzer
on a Chip
139
GAIN-PHASE DETECTOR
A N A LO G
DEVICES
w A Network Analyzer on a Chip! - Almost!
VGAIN = VG log (VA/VB )
VG = 30mV/dB
VPHS = VP ( φ 1 - φ 2 )
VP = 10mV/deg
w Operates from LF to >3 GHz
Applications
– Power Amplifier Phase/Gain Control
.... independent of actual power level
– Monitoring of System Gain/Loss (e.g. Return Loss)
– System Diagnostics
– Linear Phase Demodulator
140
TRUE GAIN MEASUREMENT
A N A LO G
DEVICES
VA
L
V1 = VY log ( VA / VX )
A
+
Σ
VB
L
VY log ( VA / VB )
B
V2 = VY log ( VB / VX )
By subtracting the output of the B-channel log-amp from that
of the A-channel log-amp, the intercept VX is eliminated and
the resulting difference is a measure of the RATIO of VA / VB
141
CANCELS PACKAGE RESONANCES
A N A LO G
DEVICES
VA
VB
εA= VAg(f)
εB= VBg(f)
L
A
+
Σ
L
B
εA
VOUT = VY log
εB
= VY log VA g(f)
VB g(f)
VA
= VY log
VB
Both channels have the same HF resonances and other HF
transmission effects g(f), but these are canceled in taking the
difference which remains a measure of the RATIO of VA / VB
142
PHASE MEASUREMENT at 2.5GHz
A N A LO G
DEVICES
VA
L
V1 = VY log ( VA / VX )
A
+
PHASE
OUTPUT
VB
L
Σ
VY log ( VA / VB )
B
V2 = VY log ( VB / VX )
Logarithmic amplifiers also provide very high gain and limiting
action: using a special type of analog multiplier between the
limiter outputs, phase measurements can be made at 3GHz
143
APPLICATIONS
A N A LO G
DEVICES
RF signal to
be measured
L
+
A
Φ
Low-frequency
reference carrier
L
Σ
-
log(A/B)
B
In this case, a low-frequency carrier provides a
very high calibration reference for the intercept
144
APPLICATIONS
A N A LO G
DEVICES
Modulated
RF signal
L
+
A
Φ
Baseband
modulation
L
Σ
-
log(A/B)
B
Here, the reference is provided by the baseband
modulation & system measures conversion gain
145
APPLICATIONS
A N A LO G
DEVICES
OUT
L
SYSTEM
BLOCK
+
A
Φ
IN
L
Σ
-
B
True gain of system block is measured
independent of the actual power levels
146
log(A/B)
APPLICATIONS
A N A LO G
DEVICES
COUPLERS
LOAD
SOURCE
L
+
A
Φ
L
Σ
-
B
Measurement of return loss
independent of power level
147
log(A/B)
The
X-AMP
148
The X-AMP™
A N A LO G
DEVICES
l
A PROPRIETARY VGA PRINCIPLE
l
FUNDAMENTALLY “LINEAR-in-dB”
l
USES FEEDBACK IN ORDER TO:
ACCURATELY DETERMINE GAIN
MINIMIZE HF NONLINEARITIES
l
GUARANTEES ULTRA-LOW NOISE
l
EXHIBITS WIDE DYNAMIC RANGE
FROM NOISE FLOOR (0.7µV RMS)
TO TYPICALLY 1.4V RMS (106dB)
149
X-AMP PRODUCTS
A N A LO G
DEVICES
150
l
AD600 & AD602 (BOTH DUAL-VGAs)
DEVELOPED FOR ULTRASOUND
l
AD603 (8-PIN, SINGLE)
-10/30dB AND 20/50dB RANGES
BEING WIDELY USED IN IF STRIPS
l
AD604 & AD605 (SINGLE-SUPPLY DUALS)
PRE-AMP PROVIDES A HIGH ZIN
l
NEW X-AMPs IN DEVELOPMENT
THE BASIC X-AMP
A N A LO G
DEVICES
VARIABLE
“SLIDER”
LOW-NOISE
AMPLIFIER
+
-
VIN
R:nR ATTENUATOR
RO
151
VOUT
R2
R1
TYPICAL 8-STAGE X-AMP
A N A LO G
DEVICES
i
INTERPOLATING gm STAGES
VOUT
⌠
⌠
gm1
VIN
100Ω
R
2R
gm2
R
2R
0dB -6.02dB
152
gm7
gm8
VOUT - 41dB
R
2R
R2
11*200Ω
R
-36.12dB -42.14dB
R1
200Ω
10
CURRENTS IN THE gm STAGES
gm1
gm8
gm2
gm3
gm4
gm5
gm6
gm7
INCREASING GAIN
153
(MOVES ACTION TOWARDS FRONT)
A N A LO G
DEVICES
MORE COMPLETE X-AMP
A N A LO G
DEVICES
LASER-TRIMMED FOR PRECISE
GAIN - SCALING AND INTERCEPT
GAUSSIAN INTERPOLATOR
MAIN GAIN STAGE
(acts as an integrator)
⌠
⌠
VGAIN
gm CELLS
VIN
R:2R ATTENUATOR
LASER-TRIMMED
FOR PRECISE R IN
154
VOUT
R2 2.2kΩ
R1
20Ω
LASER-TRIMMED
FOR PRECISE GAIN
SOLID-STATE POTENTIOMETER
A N A LO G
DEVICES
Base
contact
(2)
P-type region acts PMOS gate (poly-Si)
as PMOS source
N-type emitter
(Contacts are shown in black)
Collector
contact (2)
P-type region simultaneously serves as the N-type
NPN base, PMOS drain, and as a resistor buried
layer
SOLID-STATE POTENTIOMETER
A N A LO G
DEVICES
Carrier
domain
Domain can be moved
by voltage control
from left to right
DOMAIN - CORRESPONDS TO ‘SLIDER’
C1
BURIED LAYER FORMS RESISTIVE TRACK
C2
Why
SiGe?
157
PLANAR NPN
TRANSISTOR
SiO2
C
n+ emitter
p
i
s
o
E
E
n++
n epitaxial layer
B
B
p base
SiO2
p
i
s
o
n+ buried layer
p substrate
A N A LO G
DEVICES
PLANAR NPN
TRANSISTOR
SiO2
C
n+ emitter
p
i
s
o
E
E
n++
n epitaxial layer
B
B
p base
SiO2
p
i
s
o
n+ buried layer
THIN BASE NEEDED
TO MINIMIZE TF AND
THUS MINIMIZE QB
FOR A GIVEN IC
A N A LO G
DEVICES
PLANAR NPN
TRANSISTOR
SiO2
C
n+ emitter
p
i
s
o
E
E
n++
n epitaxial layer
B
B
p base
SiO2
p
i
s
o
n+ buried layer
THIN BASE NEEDED
TO MINIMIZE TF AND
THUS MINIMIZE QB
FOR A GIVEN IC
BUT THIS RAISES RB,
LOWERS VAF, LOWERS
BVCEO AND CAN LEAD
TO COL-EM SHORTS
A N A LO G
DEVICES
PLANAR NPN
TRANSISTOR
POLY EMITTER RAISES EMITTER RESISTANCE RE
SiO2
C
n+ emitter
p
i
s
o
E
E
n++
n epitaxial layer
B
B
p base
SiO2
p
i
s
o
n+ buried layer
THIN BASE NEEDED
TO MINIMIZE TF AND
THUS MINIMIZE QB
FOR A GIVEN IC
BUT THIS RAISES RB,
LOWERS VAF, LOWERS
BVCEO AND CAN LEAD
TO COL-EM SHORTS
A N A LO G
DEVICES
KEY IDEAS ABOUT SiGe
l
BEGIN BY USING AN EPITAXIALLY-GROWN BASE
FILM RATHER THAN AN ION-IMPLANTED LAYER
l
DURING BASE-FILM DEPOSITION ADD A SMALL
PERCENTAGE OF GERMANIUM
l
GRADE THE Ge CONCENTRATION; THIS WILL
INTRODUCE A FIELD IN THE BASE
l
ALSO, INCREASE THE CONCENTRATION OF
THE NORMAL BASE DOPANT (BORON)
162
A N A LO G
DEVICES
THIS BUYS YOU:
163
l
A BASE FILM OF VERY PRECISELY CONTROLLED
COMPOSITION, WITH A THICKNESS ACCURATE
TO WITHIN A FEW ATOMIC LAYERS
l
A MUCH LOWER BASE RESISTANCE, DUE TO USE
OF HIGHER BASE DOPING CONCENTRATION
l
A MUCH HIGHER EARLY VOLTAGE, SINCE THE
DEPLETION LAYER WIDTH IS SMALLER
l
HIGH BETA IS ACHIEVED BECAUSE OF HIGHER
EMITTER EFFICIENCY DUE TO REDUCTION
OF BAND-GAP ENERGY AT EMITTER EDGE
A N A LO G
DEVICES
TRANSLATED:
164
l
SUPER-ACCURATE BASE-WIDTH OF ABOUT
0.1µm MEANS THAT TRANSIT TIME IS VERY
SMALL AND VERY WELL CONTROLLED
l
LOW BASE RESISTANCE ALSO INCREASES
SPEED AND LOWERS JOHNSON NOISE
l
HIGH EARLY VOLTAGE RAISES AVAILABLE
GAIN & LOWERS COLLECTOR DISTORTION
l
HIGH DC BETA SIMPLIFIES BIASING
A N A LO G
DEVICES
SIMILARITY TO A GaAs MESFET?
THEY ARE BOTH VERY FAST SEMICONDUCTOR DEVICES
but
A SiGe HBT USES STANDARD SILICON WAFERS
IT DOES NOT HAVE ANY STRANGE SUBSTRATE
EFFECTS (e.g. SLOW STATES) THAT PLAGUE GaAs
IT CAN BE INTEGRATED INTO VLSI USING BiCMOS
165
A N A LO G
DEVICES
IN SHORT....
THE BENEFITS OF SiGe, THOUGH MODEST, ARE REAL
ENOUGH TO GUARANTEE WIDESPREAD ADOPTION IN
STATE-OF-THE-ART IC PROCESS TECHNOLOGIES.
THE COMBINATION OF A 50GHz HBT WITH A 0.25µm
CMOS PROCESS WILL BECOME STANDARD FOR USE
IN MIXED-SIGNAL RF & IF SIGNAL-PROCESSING ICs.
IN THESE PRODUCTS, WAFER COST IS NOT CRITICAL.
THIS PROCESS WILL BE VERY DURABLE, AND WILL
NOT READILY BE OBSOLETED IN THIS PARTICULAR
CLASS OF APPLICATIONS FOR MANY YEARS.
166
A N A LO G
DEVICES
What’s
Ahead?
167
RADIO
RADIO IN
IN THE
THE INFORMATION
INFORMATION AGE
AGE
The challenges facing designers of analog
radio systems, now inextricably interwoven
with the integrated circuit and the exclusive
use of digital modulation, are considerable.
The economics of the mass market affect
every aspect of system development, and
dictate the use of ultra-low-cost processes,
assembly, and testing techniques.
The Newtonian nature of radio remains an
inescapable factor in approaching design.
168
FUTURE TRENDS
A N A LO G
DEVICES
k
PARTNERSHIPS IN TECHNOLOGY
l
ANALOG teaming with DIGITAL
l
BIPOLAR teaming with CMOS
l
SOC’s teaming with SOH’s
l
CONTAINED PROPAGATION (e.g fiber)
teaming with BROADCAST MODES
169
FUTURE TRENDS
A N A LO G
DEVICES
k
GREATER INTERCONNECTIVITY
l
Demand for access will increase
l
Bandwidth will become more affordable
l
Home networks will become common
l Many data links will be radio-based, and
use both microwave and long-wave
l Distinction between “TV” and “PC”
presentation of images will disappear
l Un-self-conscious use of facial images
in day-to-day communications
l Wearable computers and communications
170
FUTURE TRENDS
A N A LO G
DEVICES
k
NEW PARADIGMS for DESIGN
l
The present approaches to the design of integrated
circuits and systems must be supplanted by more
efficient ones, as systems become more complex
l
This will entail greater re-use of proven cells, and of
less time for highly-specific customization
l
Increased use of advanced hybrid assembly techniques
will generate a demand for a new kind of IC designer
l
Fundamental analog design principles must be restored
to curricula and mixed-signal techniques emphasized
l
Today’s IC designers increasingly need to diversify
Need to focus more strongly on the crucial issue of
Design for Manufacture
l
171
FUTURE TRENDS
A N A LO G
DEVICES
k
INCREASING ELECTRONICS IN MEDICINE
l
Already used in many investigative and diagnostic tools
l
Small swallowable radio pills can report on temperature,
digestive chemistry, pressure, etc. Cheap transponders
can be given to a patient to connect to a PC or PDA
l
An increasing use of prosthetic devices can be foreseen,
some of which may rely on ultra-short-range radio links
l
Further advances in affordable ultra-fast computers will
facilitate the modeling of molecules and the development
of more effective disease-specific drugs
172
FUTURE TRENDS
A N A LO G
DEVICES
k
INCREASING DEPENDENCE ON GPS
l
GPS is already finding a host of unexpected uses: in Paris,
buses are equipped; in Amsterdam, Berlin and Singapore,
taxi-cab are tracked by GPS; as are trucks in the USA:
farmers guide their tractors and other equipment by GPS
l
Next step: universal use in automobiles, and other personal
transportation systems; these have numerous applications
l
Wrist-watch GPS facilitates the location of key personnel
l
Later: fully automated, GPS-guided transport systems
(Knowing the “where” will be as important knowing the “time”)
173
FUTURE TRENDS
A N A LO G
DEVICES
k
NEW PARADIGMS FOR COMPUTING?
l
When will neural networks provide useful adjuncts
to serial, algorithmic machines? Perhaps by 2015
l
Still at the fringes of practical utilization, quantum
computers approach problem solving in an entirely
novel manner, stressing the holistic physical aspects
of a system model, again in contrast to a reliance on
binary representations and serial algorithms
l
Employing nuclear magnetic resonance (NMR) to
excite and then “listen to” molecular signatures, the
need for RF transceiver techniques is central
174
exp
SUMMARY
175
!
l
l
ANALOG
ANALOG IS
IS NOT
NOT OBSOLETE
OBSOLETE
l
l
DESIGN
DESIGN CHALLENGES
CHALLENGES ABOUND
ABOUND
l
l
BIPOLAR
BIPOLAR REMAINS
REMAINS IMPORTANT
IMPORTANT
l
l
SYSTEM-on-a-HEADER
SYSTEM-on-a-HEADER may
may be
be aa
good
good alternative
alternative to
to SoC
SoC VLSI
VLSI
A N A LO G
DEVICES