=
=20
Analog:=20 Back to the future, part two
Steve=20 Taranovich - July 16, 2012
Part=20 one of this series covered op amp history from National=20 Semiconductor. This second part of the series covers Philbrick = Nexus,=20 Burr-Brown, Analog Devices and Linear Technology history and = contributions in op=20 amp history up to the 21st century offerings from these companies
Philbrick Nexus
Figure 1: George Arthur Philbrick, = Founder of=20 GAP/R George A Philbrick Researches
Philb= rick=20 Nexus was the company that launched the commercial use of the = Operational=20 Amplifier in 1952.<= /P>
The first commercial Operational = amplifier was the=20 K2-W=20 op-amp. It was based on the amplifier used in the Philbrick K3 = modular=20 Analog-Computer "black boxes ".
That amplifier's basic circuit =
architecture, in=20
turn, was probably inspired by an earlier=20
amplifier designed by Loebe Juliehttp://www.philbrickarchi=
ve.org/lj.htm=20
The K2-W Operat=
ional
Figure 2: The K2-W direct-current = operational=20 amplifier =E2=80=9Cfor use in electronic computers=E2=80=9D circa 1947 = (From the Proceedings of=20 the I.R.E. in a paper entitled =E2=80=9CAnalysis of problems in dynamics = by electronic=20 circuits=E2=80=9D by J.R. Ragazzini, R.H. Randall, and F.A. = Russell)
Editor=E2=80=99s note: John R. Ragazzini = was dean of the=20 School of Engineering and Science at New York University in the Bronx = when I=20 went there from 1968 to 1972---great guy and brilliant engineer.
Burr-Brown (BB) Op-Amp History = Review
The term =E2=80=9COp-Amp=E2=80=9D was = first coined around 1947, but=20 the concept of a DC coupled feedback amplifier was understood in the=20 1920s. The need for analog computers during World War II brought = the=20 op-amp into wide use. Of course, these amplifiers were all made = with=20 vacuum tubes.
It was not until 1956 when Burr-Brown = introduced=20 the first commercial transistorized amplifiers and it was 1958 when they = introduced the model 130, the world=E2=80=99s first transistorized = op-amp.
Looking at the big picture, one clear = evolutionary=20 path was from =E2=80=9Cboxes=E2=80=9D to sub-micron integrated circuits = with =E2=80=9Cno=E2=80=9D packages=20 (chip-scale).
Figure 3: The big picture of the = evolution of=20 the Burr-Brown op amp (Courtesy of Howard Skolnik, one of the great = analog=20 designers at Burr-Brown in the early days)
1956 BB started with=20 =E2=80=9Cinstruments=E2=80=9D in wooden boxes. The 1st = product was the model=20 100 AC Decade Amplifier (See Figure 4). This was not an = op-amp. =20 Other early products, in wooden boxes, included a Differential AC Amp, = Square=20 Wave generator, Variable Gain Preamp and AC Millivolt Meter.
Figure 4: The Model 100 was the = first product=20 that Burr-Brown made in 1956
1957 Tom Brown visited = several=20 major customers including MIT to see how they were using BB = products. To=20 his surprise he found that they were removing the circuits and = discarding his=20 beloved boxes! He learned two important lessons: Smaller is = better=20 and making =E2=80=9Ccomponents=E2=80=9D is better business than making = =E2=80=9Cend products.=E2=80=9D
The model 130, the world=E2=80=99s = 1st=20 transistorized op-amp was introduced. This was a completely = discrete=20 design using just 8 transistors on a PC-board in a 3 =C2=BD=E2=80=9D = long aluminum shell.
The desire for =E2=80=9Csmaller=E2=80=9D = led BB to the potted=20 module concept. While still using PCBs and all discrete = components,=20 innovative techniques greatly increased density. The 1501 was the = first=20 modular op-amp and is now part of the Smithsonian collection.
The 1st monolithic (IC) op-amp = was=20 introduced by Fairchild (uA702 Bob Widlar). It was not very useful and = was later=20 superseded by the uA709 (1965).
1965 The 1538 module was the=20
1st transistorized chopper stabilized=20
op-amp.
=
;=20
The 1552 module was the 1st FET-input op-amp.
=20
The 1553 =
module was=20
the 1st transistorized chopper stabilized op-amp. =
MOSFET-input=20
op-amp.
1966 The 3051 was = BB=E2=80=99s 1st=20 monolithic op-amp (Jerry Graeme on outside Fab).
The uA741 was introduced by Fairchild = (Dave=20 Fullagar).
The move to hybrids was well underway. = The ability=20 to mix and match chip-level along with discrete components on a = thick-film=20 substrate opened the door to complex circuits with very high = performance.
Trimming of thick-film resistors =
allows=20
higher precision in both modular & hybrid designs.
=
1st=20
monolithic op-amp produced on BB Fab (3051).
Trimming of thin-film resistors = provides=20 improved stability and smaller size in hybrids.
1st two-chip hybrid op-amp = (OPA102)=20 combined a bipolar monolithic chip with a dual FET chip to produce high=20 performance at lower cost than before.
Monolithic dielectric isolation (DI) = process=20 available to BB designers greatly increasing their capabilities.
INA101 is BB=E2=80=99s 1st = monolithic=20 instrumentation amplifier.
OPA100 ultra-low bias current op-amp is=20 1st produced using the BB DiFET (DI BiFET) process.
OPA111 Low noise, low drift op-amp built = on BB=20 DiFET process.
1986 INA110 is BB=E2=80=99s = 1st=20 monolithic BiFET instrumentation amplifier.
OPA445 Hi-Voltage BiFET op-amp. = +/-45V.
OPA627 Near =E2=80=9Cideal=E2=80=9D = op-amp built on BB DiFET=20 process. 250uV Vos, 5pA Ibias, 5nv Noise, 16MHz BW, 55V/us SR, = +/-18V=20 supplies.
1996 OPA237 =
1st op-amp in SOT-23 package.
=20
OPA2237=20
1st dual op-amp in MSOP-8 package.
Comparing the first transistorized op-amp = to a=20 modern chip-scale device.
Table 1: A Comparison of the first=20 transistorized op-amp to a modern chip-scale device.
Analog=20 DevicesAnalog Devices Inc. = (ADI)
EDN directed some pertinent questions =
to Barry=20
Gilbert, ADI Technology Fellow (Figure 5) and Bob Adams, ADI Fellow =
(Figure=20
7):
Figure 5: Barry Gilbert in his = early days=20 circa 1951
Barrie Gilbert
Background: Gilbert is one of the = industry=E2=80=99s=20 foremost experts in the development and application of analog circuitry. = He now=20 directs engineering at the Northwest Labs in Beaverton, Oregon, = ADI=E2=80=99s first=20 remote design center.
Gilbert=E2=80=99s 40-year affiliation = with Analog Devices=20 Inc. =E2=80=93 dating back to 1972 =E2=80=93 saw the company go from = being principally a module=20 maker to a producer of high-volume IC parts and digital signal = processors.
One of the circuit cells that bears his = name has=20 for decades been used in all forms of communication systems, including = ordinary=20 radios, cell phones, microwave TV links, data modems, satellite = communications=20 and even radio telescopes.
The Gilbert cell=E2=80=94actually an = entire class of=20 versatile cell topologies used as basic analog function = blocks=E2=80=94has served as the=20 foundational design for products used everywhere in today=E2=80=99s = electronic=20 systems. All invoke the now famous Translinear Principle. This = fundamental=20 theory in circuit design was discovered, formalized, refined and = popularized by=20 Gilbert. Translinear circuits perform pure-current-mode signal = processing, a=20 fundamental insight. Today, these ideas, whether in the original bipolar = form or=20 in CMOS embodiments, are found throughout analog design.
Figure 6: A schematic of the = two-quadrant=20 Gilbert Cell
Since its invention in 1967, what has = become known=20 as the Gilbert Mixer is now ubiquitous in radio transmitters and = receivers. The=20 compact nature and precise commutation properties of this mixer opened = one of=20 many important doors to the integration of radios in monolithic form, = leading to=20 the proliferation of modern indispensable communications devices. A = closely=20 related circuit, known as the Gilbert multiplier, overnight = revolutionized the=20 implementation of this important mathematical analog function. The 1968 = Journal=20 of Solid-State Circuits paper describing it became the first paper to be = cited=20 100 times. Today, more than 40 years later, it remains one of the = most-cited=20 JSSC papers.
Gilbert believes that childhood = hardships=E2=80=94including=20 at age three losing his father in World War II, leaving his mother and = three=20 other children penniless=E2=80=94force one to be resourceful. Before and = during his=20 teenage years, he had access to a plethora of inexpensive military = surplus gear=20 which greatly helped to make him inventive. Gilbert laments that today's = aspiring engineers are lacking the visceral experience of handling and = hefting=20 large coils and tuning capacitors, transformers and vacuum tubes, and = such.=20 Today=E2=80=99s surplus circuit boards are all but useless as a source = of inspiration,=20 or even =E2=80=9Cspare parts=E2=80=9D to tinker with.
1. What initial analog = developments from=20 ADI=E2=80=99s past have helped shape ICs in the 21st century = industry?=20
In 1971-72, Analog Devices worked with = and funded a=20 start-up called Nova Devices to begin fabrication of linear integrated = circuits.=20 This collaboration carried through to become Analog Devices = Semiconductor (ADS)=20 division and started to deliver high-performance linear in 1972. Much of = the=20 early revenue came from op amps, including the AD741 =E2=80=93 a = near-copy of the=20 historic precursor, but with stronger emphasis on precision and quality. = And=20 Analog Devices began making laser-trimmed FET op amps with much better=20 performance than the industry standards of the day. This emphasis on = providing=20 high accuracy and advanced performance arose from the very earliest = days. With=20 the advent of nonlinear functions, such as the first high accuracy = laser-trimmed=20 analog multiplier, the AD534, the first monolithic RMS-DC converter, the = AD536,=20 and the first complete monolithic V/F converter, the AD537 =E2=80=93 all = using=20 translinear techniques =E2=80=93 another seminal emphasis came to the = fore, namely the=20 provision of precise calibration of nonlinear functions. Undoubtedly, it = was the=20 development of wafer-laser-trimming of ADI=E2=80=99s proprietary = thin-film resistors=20 that gave the company a considerable edge, in this regard.
Later, Analog Devices and the ADS = division made=20 further progress in the fabrication of linear ICs. Process 1 was = optimized=20 principally for op amp use; Process 2 was a little faster, and was used = in=20 I2L modes in the earliest ADCs. Further advances in speed = came with=20 Process 3. I felt the need for, and defined, a complementary bipolar = process,=20 which became =E2=80=9CCB.=E2=80=9D A radical later departure was the = early adoption of=20 silicon-on-insulator (SOI) processes, the =E2=80=9CXF=E2=80=9D = series.
2. How did your innovation or an = older=20 architecture specifically set us up to achieve today=E2=80=99s 21st = century=20 performance?
The growth to maturity of ADS, which = rapidly became=20 the largest revenue generator of ADI, and is no longer a separate = division, came=20 out of the seminal contributions of numerous talented people. = It=E2=80=99s hard to=20 identify crucial product developments that shaped the company at large. = Each of=20 these people would have a unique perspective on that issue.
But if I am to speak of from my own = viewpoint, I=20 believe it was my personal interest in =E2=80=9Cradio=E2=80=9D =E2=80=93 = going back to childhood days =E2=80=93=20 that drove me to push hard to provide the tools and ideas to develop = chips for=20 this sector of the business, at a time when the company was = predominantly a=20 provider of industrial and, to a lesser extent, military components. No = one was=20 using the word =E2=80=9Cgigahertz=E2=80=9D at that time! One key = development was of the first=20 five-stage RF logarithmic amplifier, the AD640, sometime in the 1980s. = Since=20 that time, my team and I have development numerous multistage log amps = for use=20 in RF power measurement. We can boast that practically every cell phone = and base=20 station in the world uses these ADI products.
As for =E2=80=9Colder = architecture,=E2=80=9D products developed in=20 the 1970s using translinear techniques =E2=80=93 such as the AD534 = multiplier, and other=20 developments of that kind =E2=80=93 remain in the catalog and continue = to generate=20 significant revenue.
Beyond that, I believe my insistence that = we needed=20 at least one scientific computer =E2=80=93 and eventually a CAD team of = our own =E2=80=93=20 eventually yielded fruit. We initially purchased one VAX780 and = time-shared=20 it.
3. What advice do you = have for=20 today=E2=80=99s 21st century designer? What analog know-how = does today=E2=80=99s=20 designer need to create successful designs?
These are crucial questions, but they = would need=20 the wisdom of Solomon to provide adequate answers!
First, I would say this. Before any young = person=20 enters into a life of microelectronic design, he or she should be quite = sure=20 that this is going to be the beginning of a life full of joyous = discovery and=20 invention. There are many fields that can provide this sort of joy, so=20 self-examination as to a career is essential at a very early age.
Second set out to be the best in your = field. As an=20 IC designer, you will need to wear many assorted hats. Yes, often you = will be=20 wearing your Circuit Designer=E2=80=99s Cap and Cape, but at other = times, you will need=20 to don your Pragmatist=E2=80=99s Hat, your Economist=E2=80=99s Hat, your = Physicist=E2=80=99s Hat and=20 many more. Deliberation over difficult trade-offs will frequently arise. = In=20 short, IC product design is not simply about transistors.
Third, beyond being =E2=80=9Cbest in = class=E2=80=9D you must aspire=20 to becoming Master of the Dance. By that, I mean that you will develop a = deep=20 sense of being in control of all that your mischievous little = transistors do.=20
They will often want to sing, when you = just need=20 them to do a jig from left to right across your stage. When you = choreograph your=20 circuit on the screen, you must think like a transistor thinks. You must = actually become a transistor!
Fourth, ask =E2=80=9CWhat IF?=E2=80=9D a = thousand times a day. This=20 question is the quintessential fountain of invention.
Bob Adams: = Analog=20 DevicesBob Adams
Figure 7: Bob Adams at his desk = at DBX,=20 an early competitor to Dolby circa 1982. DBX made noise-reduction = systems and=20 analog processing gear, largely based on novel log/antilog-based=20 voltage-controlled amplifiers and RMS detectors. Adams got a good = expertise here=20 in analog signal processing.
Background: Adams = graduated with a=20 BSEE from Tufts University in 1976, and after spending several years as = a=20 musician, he began a career in the consumer/professional audio equipment = market.=20 In the late 1970s, he published the first paper on log-domain filtering = and then=20 began working extensively in the brand new area of sigma-delta A/D = converters,=20 producing the first audio converter with greater than 16-bit resolution. =
In 1988, Adams joined the ADI Converter = Group as a=20 Senior Staff Designer. Together with Paul Ferguson, he developed ADI's = first=20 sigma-delta converters. During the past 20 years, Adams has pioneered = many=20 important architectural advances in sigma-delta converters including=20 mismatch-shaping, multi-bit quantization, and continuous-time = architectures.=20
Adams also has a passion for digital = signal=20 processing and produced the world=E2=80=99s first monolithic = asynchronous sample rate=20 converters=E2=80=94the AD1890 family=E2=80=94using patented ideas and = design techniques. After=20 getting a taste of digital design, Adams founded the sigmaDSP line of=20 audio-specific DSP cores.
Questions
1 What initial analog = developments from=20 ADI=E2=80=99s past have helped shape ICs in the 21st century = industry?=20
There are many such developments, and = it=E2=80=99s hard to=20 come up with only a few. The obvious ones are a variety of improvements = to the=20 basic band-gap cell, done by engineers such as Paul Brokaw and Barrie = Gilbert; a=20 variety of circuits based on the translinear principle that are now used = extensively in RF circuits such as VGAs and RF power monitoring; and = many=20 different innovations used in A/D and D/A converters of all types that = have=20 survived to the present day.
2 How did your innovation or an = older=20 architecture specifically set us up to achieve today=E2=80=99s = 21st=20 century performance?
When I first began playing around with = delta-sigma=20 converters back in the early 1980s, I did not have much IC design = experience and=20 therefore was drawn to topologies that could be tested on a breadboard. = This led=20 me to use a continuous-time architecture that used conventional op-amps, = R=E2=80=99s and=20 C=E2=80=99s. This choice leads to a number of performance issues that I = had to tackle,=20 including sensitivity to clock jitter which was solved by the use of = multi-bit=20 quantization.
Integrated delta-sigma converters really = took off=20 in the 1990s, and most of the designs were based on switched-capacitor = circuits=20 that were easier to integrate. However, in the last five years or so, = there has=20 been a swing back to continuous-time/multi-bit designs, partly as a = result of=20 the introduction of mismatch-shaping techniques. Most of the lessons I = learned=20 back in the =E2=80=9980s are still relevant today.
3 What advice do you have = for today=E2=80=99s=20 21st century designer? (What analog know-how does = today=E2=80=99s designer=20 need to create successful designs?)
It=E2=80=99s really important for = today=E2=80=99s designers to be=20 as =E2=80=9Cbroad=E2=80=9D as possible. Most of today=E2=80=99s most = successful ICs require a blend of=20 analog, digital, and system-level know-how. In many cases, analog = problems=20 can now be solved with a combination of analog and digital techniques. = For=20 example, if you are designing an A/D converter, the problem of precision = matching can often be overcome by a background digital calibration loop. = On the=20 other hand, if your circuit is too noisy, no amount of digital logic can = fix=20 it.
The Linear Technology op amp=20 story
Bob Dobkin discusses some of the history = and his=20 experiences in the early days of op amps. Before founding Linear=20 Technology in 1981, Mr. Dobkin was Director of Advanced Circuit = Development at=20 National Semiconductor for eleven years. He was with Philbrick/Nexus = before=20 that.
Linear Technology introduced its first = product in=20 1983, the LT1001=20 precision op amp. Dobkin said, =E2=80=9CWe=E2=80=99re going to do it an = order of magnitude=20 better than anyone else has done.=E2=80=9D
In the early days, IC trimming was done = using=20 =E2=80=9CZener zaps=E2=80=9D at the wafer level and later on they were = able to trim after=20 packaging.
The tools at that time were breadboards = made from=20 transistor pairs, a favorite of Jim Williams. If the breadboard worked, = then the=20 IC would work. SPICE was not very reliable; models were not very = good.
Figure 8: Classic Jim Williams = at his=20 best in the lab
In 1983 there were many +5v logic ICs, so = the=20 company developed the=20 LT1013, a single-supply, dual precision op amp for operation from 5V = to 36=20 V.
Later came the super-gain amplifiers like = the LM108=20 by National; Linear Technology=E2=80=99s solution was the improved = LT1008. Later the=20 chopper LT1050=20 was developed.
In the mid-to-late 80s, processes were = improved and=20 the 3 to 4 inch wafers went to 6 inch. Fabs were cleaner; relatively = small=20 geometries of 2 micron from digital processes helped op amp speeds to=20 increase.
At that time IC design went from less = than 10 masks=20 to 20 or 30 masks and bipolar processes became fully complementary with = PNP=20 transistors being as fast as NPNs, so transistor bandwidths were able to = begin=20 moving up in the 300 MHz to 1 GHz regions. This brought about op amp = speeds in=20 the order of 30 to 50 MHz with higher slew rates than previously = possible.
Today, complementary processes have = transistors in=20 the 10 GHz region, giving op amps bandwidth capabilities of 1 GHz and = much=20 higher slew rates to drive high speed ADCs.
Dobkin commented that analog companies = like Linear=20 need their own special proprietary processes to differentiate themselves = from=20 competitors; he said that the Chinese and Taiwanese semiconductor = companies do=20 not have this. Standard processes and outside foundries are also used, = but=20 Linear has two fabs that make 95% of their products.
Dobkin=E2=80=99s philosophy at Linear is = to =E2=80=9Cminimize the=20 phone calls,=E2=80=9D do it right and customers will not have to call = with problems that=20 the IC is having in their designs. When he developed the LM318 at = National, the=20 decision was to make the speed 15 MHz without great phase margin, since = the goal=20 was speed with this product. (See Analog:=20 Back to the Future Part 1.) Well, when the op amp oscillated in = customers=E2=80=99=20 circuits, they would call him and he would tell them how to compensate = the=20 device.
Linear=E2=80=99s 21st century = solutions now=20 include the LTC6417, a differential ADC driver amplifier that is an = improvement=20 over the previous LTC6416=20 buffer amp. The design gets the best performance in the 20 to 140 MHz = region at=20 lowest power dissipation. The data sheet speaks well to RF designers = with specs=20 like OIP3, P1dB and noise figure. The 50 ohm drive capability is what RF = engineers like in an IF amp. Linear=E2=80=99s design capabilities, = combined with an=20 advanced SiGe process, results in some superb fully differential = amplifiers=20 (FDAs).
Another excellent new amplifier is the = LTC6431-15=20 for 600 MHz to 1 GHz signals with excellent OIP3 and great ADC drive = capability,=20 which can even move a step closer to the antenna in a receiver. The test = capabilities in production are challenged with these types of super = speed=20 devices. Testing needs to be done accurately and quickly and Linear even = specifies more guaranteed min and max specs on their data sheets, = something that=20 designers love!
