Complete Download Introduction to Mixed Signal IC Test and Measurement 2nd Edition Edition Roberts PDF All Chapters
Complete Download Introduction to Mixed Signal IC Test and Measurement 2nd Edition Edition Roberts PDF All Chapters
https://ebookgate.com/product/an-introduction-to-mixed-signal-ic-test-
and-measurement-2nd-edition-gordon-roberts/
ebookgate.com
https://ebookgate.com/product/analog-and-mixed-signal-electronics-1st-
edition-karl-stephan/
ebookgate.com
https://ebookgate.com/product/introduction-to-metrology-applications-
in-ic-manufacturing-1st-edition-bo-su/
ebookgate.com
https://ebookgate.com/product/introduction-to-neural-dynamics-and-
signal-transmission-delay-jianhong-wu/
ebookgate.com
Introduction to Measurement Theory First Edition Mary J.
Allen
https://ebookgate.com/product/introduction-to-measurement-theory-
first-edition-mary-j-allen/
ebookgate.com
https://ebookgate.com/product/digital-image-and-signal-processing-for-
measurement-systems-1st-edition-j-richard-duro/
ebookgate.com
https://ebookgate.com/product/introduction-to-digital-signal-
processing-and-filter-design-1st-edition-b-a-shenoi/
ebookgate.com
https://ebookgate.com/product/plants-from-test-tubes-an-introduction-
to-micropropagation-3rd-edition-kleyn/
ebookgate.com
https://ebookgate.com/product/chinese-mythology-a-to-z-2nd-edition-
jeremy-roberts/
ebookgate.com
SECOND EDITION
THE OXFORD SERIES IN ELECTRICAL AND COMPUTER ENGINEERING
An Introduction to Mixed-Signal
IC Test and Measurement
GORDON ROBERTS
FRIEDRICH TAENZLER
MARK BURNS
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
For titles covered by Section 112 of the US Higher Education Opportunity Act,
please visit www.oup.com/us/he for the latest information about pricing and
alternate formats.
Printing number: 9 8 7 6 5 4 3 2 1
S ince the introduction of the first edition of this textbook in 2001, much change has occurred
in the semiconductor industry, especially with the proliferation of complex semiconductor
devices containing digital, analog, mixed-signal, and radio-frequency (RF) circuits. The integra-
tion of these four circuit types has created many new business opportunities, but at the same time
made the economics of test much more significant. Today, product costs are divided among sili-
con, test, and packaging in various proportions, depending on the maturity of the product as well
as the technical skill of the engineering teams. In some market segments, we are seeing packaging
costs dominate product costs; in other market segments, device packaging is being done away
with entirely, because bare die is mounted directly on the carrier substrate. While this helps to
address the packaging costs, it puts greater pressure on test costs, because it now becomes a larger
contributor to the overall product cost.
Analog, mixed-signal, and RF IC test and measurement have grown into a highly special-
ized field of electrical engineering. For the most part, analog and mixed-signal test engineering
is handled by one team of specialists, while RF test is handled by another. The skill set required
to master both technical areas are quite different, because one involves the particle (electron)
perspective of physics whereas the other handles the wave perspective. Nonetheless, these two
technical areas have much in common, such as the need to accurately measure a signal in the pres-
ence of noise and distortion in a time-sensitive manner, albeit over vastly different dynamic range
and power levels.
The goal of the first edition of this textbook was to create a source of information about analog
and mixed-signal automated test and measurement as it applies to ICs. At the time, little informa-
tion was available for the test engineer. One important source was the textbook by M. Mahoney,
the pioneer of DSP-based test techniques, however, this was largely limited to the coherency prin-
ciples of DSP-based testing, and it did not discuss system level tradeoffs of test engineering, nor
did it discuss the practical issue related to test interfacing.
Based on the feedback that we received, the first edition of this textbook has been a source
of inspiration to many, especially those new to the test field. The industry has seen a great deal of
change since the release of the first edition almost a decade ago. RF circuits now play a larger part
in many of the devices and systems created today. It is clear to us that engineers need to be fluent
in all four-circuit types, digital, analog, mixed-signal and RF, although we limit our discussion in
this textbook to the latter three, as digital test is a subject all in itself. We do not believe we could
do this topic justice in the amount of space remaining after discussing analog, mixed-signal, and
RF test. We encourage our readers to learn as much as possible about digital test as they will most
certainly encounter such techniques during their career in test (or design for that matter).
The prerequisite for this book remains at a junior or senior university level; it is assumed that
students reading this textbook will have taken courses in linear continuous-time and discrete-time
systems, fields, and waves, as well as having had exposure to probability and statistical concepts.
xix
xx PREFACE
In fact, the three greatest changes made to the second edition of this textbook is a lengthy discus-
sion on RF circuits, high-speed I/Os, and probabilistic reasoning. Over the years, it has become
quite clear that test, application, and product engineers make extensive use of probability theory
in their day-to-day function, leading us to believe that it is necessary to increase the amount of
probability coverage contained in this textbook. By doing so, we could define the concept of noise
more rigorously and study its affects on a measurement more concisely. These ideas will be used
throughout the textbook to help convey the limitations of a measurement.
generators in a mixed-signal tester. ADC sampling theory is applicable to both (a) ADC circuits
in the device under test and (b) waveform digitizers in a mixed-signal tester. Coherent multitone
sample sets are also introduced as an introduction to DSP-based testing.
Chapter 9 further develops sampling theory concepts and DSP-based testing methodologies,
which are at the core of many mixed-signal test and measurement techniques. FFT fundamentals,
windowing, frequency domain filtering, and other DSP-based testing fundamentals are covered
in Chapters 8 and 9.
Chapter 10 shows how basic AC channel tests can be performed economically using DSP-
based testing. This chapter covers only nonsampled channels, consisting of combinations of oper-
ational amplifiers, analog filters, PGAs, and other continuous-time circuits.
Chapter 11 explores many of these same tests as described in Chapter 10 as they are applied
to sampled channels, which include DACs, ADCs, sample and hold (S/H) amplifiers, and so on.
The principle of undersampling under coherent conditions is also discussed in this chapter.
Chapters 12 and 13 are two chapters related to RF testing. They are both new to the second
edition of this textbook.
Chapter 12 begins by introducing the reader to the concept of a propagating wave and the
various means by which a wave is quantified (e.g., power, wavelength, velocity, etc.). Included
in this discussion are the amplitude and phase noise impairments that an RF wave experiences in
transmission. A portion of this chapter is devoted to the concept of S-parameters as it applies to
an n-port network, such as a two-port network. S-parameters are used to describe the small-signal
performance of the network as seen from one port to another. Measures like reflection coefficient,
mismatch loss, insertion and transducer loss, and various power gains can easily be defined in
terms of these S-parameters. Moreover, the idea of a mismatch uncertainty and its impact on a
power measurement is introduced. Mismatch uncertainty is often the most significant contributor
to measurement error in an RF test. Several forms of modulation, including analog modulation
schemes like AM and PM, followed by several digital modulation schemes, such as ASK, PSK,
and QAM used primarily for digital communication systems, are described.
Chapter 13 describes the principles of RF testing of electronic circuits using commercial
ATE. These principles are based on the physical concepts related to wave propagation outlined
in Chapter 12 combined with the DSP-based sampling principles of Chapters 8–11. This chap-
ter describes the most common types of RF tests using ATE for standard devices, like mixers,
VCOs, and power amplifiers. Issues relating to dynamic range, maximum power, noise floor, and
phase noise are introduced. Measurement errors introduced by the device interface board due to
impedance mismatches, transmission losses, etc., are described, together with de-embedding tech-
niques to compensate for these errors. Also included is a discussion of ATE measurements using
directional couplers. These directional couplers enable the measurement of the S-parameters of a
device through the direct measurement of the incident and reflected waves at the input and output
port of a DUT. Various noise figures measurements (Y-factor and cold noise methods) using an
ATE are outlined. Finally, the chapter concludes with a discussion on using an ATE to measure
more complex RF system parameters like EVM, ACPR, and BER.
Chapter 14 is also new to the second edition. Chapter 14 outlines the test techniques and met-
rics used to quantify the behavior of clocks and serial data communication channels. The chapter
begins by describing several measures of clock behavior in both the time and frequency domains.
In the time domain, measures like instantaneous jitter, period jitter, and cycle-to-cycle jitter are
described. In the frequency domain, the concept of phase noise will be derived from a power
spectral density description of the clock signal. Chapter 14 will then look at the test attributes
associated with communicating serial over a channel. For the most part, this becomes a measure of
the bit error rate (BER). However, because BER is a very time-consuming measurement to make,
methods to extract estimates of the BER in short period of time are described. These involve a
model of channel behavior that is based on parameters known as dependent jitter (DJ) and random
xxii PREFACE
jitter (RJ). Finally, the chapter will discuss jitter transmission test such as jitter transfer and jitter
tolerance.
Chapter 15 explores the gray art of mixed-signal DIB design. Topics of interest include com-
ponent selection, power and ground layout, crosstalk, shielding, transmission lines, power match-
ing networks, and tester loading. Chapter 15 also illustrates several common DIB circuits and their
use in mixed-signal testing.
Chapter 16 gives a brief introduction to some of the techniques for analog and mixed-signal
design for test. There are fewer structured approaches for mixed-signal DfT than for purely digital
DfT. The more common ad hoc methods are explained, as well as some of the industry standards
such as IEEE Std. 1149.1, 1149.4, and 1500.
related to impedance matching circuits and Smith ChartsTM. The chapter on DfT was expanded to
include a discussion on DFT for RF circuits.
A summary of the changes made are:
• Expanded use of probabilistic approach to problem description.
• Windowing for noncoherent sampling.
• Undersampling and reconstruction using modulo-time shuffling algorithm.
• New chapter on fundamentals of RF testing.
• New chapter on RF test methods.
• New chapter on clock and serial data communications channel measurements.
• Added several sections on RF load board design.
• Hundreds of new examples, exercises, and problems have been added.
A NOTE OF THANKS
First and foremost, a special note of thanks to Mark Burns for starting this project while at Texas
Instruments and overseeing the first edition of this book get published. Mark has since retired from
the industry and is enjoying life far away from test. It was a great pleasure working with Mark.
We would like to extend our sincere appreciation to the many people who assisted us with this
development; beginning with the first edition and subsequently followed by those who contributed
to the second edition.
First Edition
The preliminary versions of the first edition were reviewed by a number of practicing test engi-
neers. We would like to thank those who gave us extensive corrections and feedback to improve
the textbook:
Steve Lyons, Lucent Technologies
Jim Larson, Teradyne, Inc.
Gary Moraes, Teradyne, Inc.
Justin Ewing, Texas A&M University/Texas Instruments, Inc.
Pramodchandran Variyam, Georgia Tech/Texas Instruments, Inc./Anora LLC.
Geoffrey Zhang, formerly of Texas Instruments, Inc.
We would also like to extend our sincere appreciation to the following for their help in developing
this textbook:
Dr. Rainer Fink, Texas A&M University
Dr. Jay Porter, Texas A&M University
Dr. Cajetan Akujuobi, Prairieview A&M University
Dr. Simon Ang, University of Arkansas
Their early adoption of this work at their respective universities has helped to shape the book’s
content and expose its many weaknesses.
We also thank Juli Boman (Teradyne, Inc.) and Ted Lundquist (Schlumberger Test Equipment)
for providing photographs for Chapter 1.
We are extremely grateful to the staff at Oxford University Press, who have helped guide us
through the process of writing an enjoyable book. First, we would like to acknowledge the help
and constructive feedback of the publishing editor, Peter Gordon. The editorial development help
of Karen Shapiro was greatly appreciated.
xxiv PREFACE
Finally, on behalf of the test engineering profession, Mark Burns would like to extend his
gratitude to
Del Whittaker, formerly of Texas Instruments, Inc.
David VanWinkle, formerly of Texas Instruments, Inc.
Bob Schwartz, formerly of Texas Instruments, Inc.
Ming Chiang, formerly of Texas Instruments, Inc.
Brian Evans, Texas Instruments, Inc.
for allowing him to develop this book as part of his engineering duties for the past three years. It
takes great courage and vision for corporate management to expend resources on the production
of a work that may ultimately help the competition.
Second Edition
The preliminary versions of the second edition were reviewed by a number of practicing test
engineers and professors. We would like to thank those who gave us extensive corrections and
feedback to improve the textbook, specifically the following individuals:
Brice Achkir, Cisco Systems Inc.
Rob Aitken, ARM
Benjamin Brown, LTX-Credence Corporation
Cary Champlin, Blue Origin LLC
Ray Clancy, Broadcom Corporation
William DeWilkins, Freescale Semiconductor
Rainer Fink, Texas A&M University
Richard Gale, Texas Tech University
Michael Purtell, Intersil Corporation
Jeff Rearick, Advanced Micro Devices
Tamara Schmitz, Intersil Corporation
Robert J. Weber, Iowa State University
We are extremely grateful to the staff members at Oxford University Press, who have helped guide
us through the process of writing the second edition of this book. First, we would like to acknowl-
edge the help and constructive feedback of our editors: Caroline DiTullio, Claire Sullivan, and
Rachael Zimmerman.
Gordon Roberts would like to extend his sincere appreciation to all the dedicated staff mem-
bers and graduate students associated with the Integrated Microsystem Laboratory (formerly the
Microelectronics and Computer System Laboratory) at McGill University. Without their desire to
learn and ask thought-provoking questions, this textbook would have been less valuable to the stu-
dents that follow them. For this, I am thankful. More importantly, though, is how the excitement
of learning new things has simply made life more enjoyable. For this, I am eternally grateful. A
listing of the students that had made some contribution to these works are:
Sadok Aouini Dong (Hudson) An
Christopher Taillefer Mouna Safi-Harb
Mohammad Ali-Bakhshian Mourad Oulmane
Tsung-Yen Tsai Ali Ameri
Azhar Chowdhury Shudong Lin
Marco Macedo George Gal
Kun Chuai Michael Guttman
Euisoo. Yoo Simon Hong
Tarek. Alhajj
Preface xxv
Gordon Roberts would like to express his love and thanks to his family (Brigid, Sean, and
Eileen) for their unequivocal support over the course of this project. Their understanding of the
level of commitment that is required to undertake such a large project is essential for the success
of this work. Without their support, this project would not have come to completion. For that, all
those that learn something from this book owe each one of you some level of gratitude.
In a similar fashion, Friedrich Taenzler would like to express his sincere gratitude to his col-
leagues at Texas Instruments he had the chance to work with, thereby discussing and learning
multiple aspects of test engineering. A special thanks to those colleagues who gave extensive
corrections and feedback to improve this textbook: Ganesh Srinivasan, Kausalya Palavesam, and
Elida de-Obaldia. Friedrich Taenzler also extends his appreciation to the ATE and test equipment
vendors he had the chance to work with while learning the broader view on production tester
implementation.
Finally, and most important, Friedrich would like to thank his family, his wife Claudia and
his sons Phillip and Ferdinand, for their tremendous support and understanding for the time given
to contribute to this book. Needless to say, without their help and tolerance, this book would have
never been completed.
Gordon W. Roberts Friedrich Taenzler Mark Burns
McGill University Texas Instruments, Inc. formerly of Texas Instruments, Inc.
Montreal, Quebec, Canada Dallas, Texas, USA Dallas, Texas, USA
CONTENTS
Preface xix
v
vi CONTENTS
3.13 Summary 76
Problems 76
References 78
circuitry coexisting on the same die or circuit board. The line between mixed-signal circuits and
analog or digital circuits is blurry if one wants to be pedantic.
Fortunately, the blurry lines between digital, analog, and mixed-signal are completely irrel-
evant in the context of mixed-signal test and measurement. Most complex mixed-signal devices
include at least some stand-alone analog circuits that do not interact with digital logic at all. Thus,
the testing of op amps, comparators, voltage references, and other purely analog circuits must be
included in a comprehensive study of mixed-signal testing. This book encompasses the testing of
both analog and mixed-signal circuits, including many of the borderline examples. Digital test-
ing will only be covered superficially, since testing of purely digital circuits has been extensively
documented elsewhere.2–4
conversion architecture, whereas a 100-MHz video ADC may have to employ a much faster
flash conversion architecture. The weaknesses of these two architectures are totally different.
Consequently, the testing of these two converter types is totally different. Similar differences
exist between the various types of DACs.
Another common mixed-signal circuit is the phase locked loop, or PLL. PLLs are typically
used to generate high-frequency reference clocks or to recover a synchronous clock from an asyn-
chronous data stream. In the former case, the PLL is combined with a digital divider to construct
a frequency multiplier. A relatively low-frequency clock, say, 50 MHz, is then multiplied by an
integer value to produce a higher-frequency master clock, such as 1 GHz. In the latter case, the
recovered clock from the PLL is used to latch the individual bits or bytes of the incoming data
stream. Again, depending on the nature of the PLL design and its intended use, the design weak-
nesses and testing requirements can be very different from one PLL to the next.
MI C EAR
B as e st at i on s
Display
Control Frequency
μ-processor synthesizer
Keyboard
4 AN INTRODUCTION TO MIXED-SIGNAL IC TEST AND MEASUREMENT
each cellular area. The control microprocessor selects the incoming and outgoing transmission
frequencies by sending control signals to the frequency synthesizer. The synthesizer often con-
sists of several PLLs, which control the mixers in the radio-frequency (RF) section of the cellular
telephone. The mixers convert the relatively low-frequency signals of the base-band interface to
extremely high frequencies that can be transmitted from the cellular telephone’s radio antenna.
They also convert the very high-frequency incoming signals from the base station into lower-
frequency signals that can be processed by the base-band interface.
The voice-band interface, digital signal processor (DSP), and base-band interface perform
most of the complex operations. The voice-band interface converts the user’s voice into digital
samples using an ADC. The volume of the voice signal from the microphone can be adjusted
automatically using a programmable gain amplifier (PGA) controlled by either the DSP or the
control microprocessor. Alternatively, the PGA may be controlled with a specialized digital circuit
built into the voice-band interface itself. Either way, the PGA and automatic adjustment mecha-
nism form an automatic gain control (AGC) circuit. Before the voice signal can be digitized by
the voice-band interface ADC, it must first be low-pass filtered to avoid unwanted high-frequency
components that might cause aliasing in the transmitted signal. (Aliasing is a type of distortion
that can occur in sampled systems, making the speaker’s voice difficult to understand.) The digi-
tized samples are sent to the DSP, where they are compressed using a mathematical process called
vocoding. The vocoding process converts the individual samples of the sound pressure waves into
samples that represent the essence of the user’s speech. The vocoding algorithm calculates a time-
varying model of the speaker’s vocal tract as each word is spoken. The characteristics of the vocal
tract change very slowly compared to the sound pressure waves of the speaker’s voice. Therefore,
the vocoding algorithm can compress the important characteristics of speech into a much smaller
set of data bits than the digitized sound pressure samples. The vocoding process is therefore a
type of data compression algorithm that is specifically tailored for speech. The smaller number of
transmitted bits frees up airspace for more cellular telephone users. The vocoder’s output bits are
sent to the base-band interface and RF circuits for modulation and transmission. The base-band
interface acts like a modem, converting the digital bits of the vocoder output into modulated ana-
log signals. The RF circuits then transmit the modulated analog waveforms to the base station.
In the receiving direction, the process is reversed. The incoming voice data are received by
the RF section and demodulated by the base-band interface to recover the incoming vocoder bit
stream. The DSP converts the incoming bit stream back into digitized samples of the incoming
speaker’s voice. These samples are then passed to the DAC and low pass reconstruction filter of the
voice-band interface to reconstruct the voltage samples of the incoming voice. Before the received
voice signal is passed to the earpiece, its volume is adjusted using a second PGA. This earpiece
PGA is adjusted by signals from the control microprocessor, which monitors the telephone’s vol-
ume control buttons to determine the user’s desired volume setting. Finally, the signal must be
passed through a low-impedance buffer to provide the current necessary to drive the earpiece.
Several common cellular telephone circuits are not shown in Figure 1.2. These include DC
voltage references and voltage regulators that may exist on the voice-band interface or the base-
band processor, analog multiplexers to control the selection of multiple voice inputs, and power-on
reset circuits. In addition, a watchdog timer is often included to periodically wake the control
microprocessor from its battery-saving idle mode. This allows the microprocessor to receive infor-
mation such as incoming call notifications from the base station. Clearly, the digital cellular tele-
phone represents a good example of a complex mixed-signal system. The various circuit blocks of
a cellular telephone may be grouped into a small number of individual integrated circuits, called
a chipset, or they may all be combined into a single chip. The test engineer must be ready to test
the individual pieces of the cellular telephone and/or to test the cellular telephone as a whole. The
increasing integration of circuits into a single semiconductor die is one of the most challenging
aspects of mixed-signal test engineering.
Chapter 1 • Overview of Mixed-Signal Testing 5
(a) Mask
Photoresist
SiO2
P- substrate
(b) Photoresist
-
P substrate
exposure
(c) Photoresist
P- substrate
selective removal
(d) Oxide et ch
P- substrate
N-well
(e) N-well doping
P- substrate
Vias SiO2
Protective overcoat (PO)
Metal 2
Metal 1
Polysilicon gate
P+ P+ N+ N+
N-well
-
P substrate
(f) Finished IC
6 AN INTRODUCTION TO MIXED-SIGNAL IC TEST AND MEASUREMENT
textbook diagrams would lead us to believe. Cross sections of actual integrated circuits reveal a vari-
ety of nonideal physical characteristics that are not entirely under the semiconductor manufacturer’s
control. Certain characteristics, such as doping profiles that define the boundaries between P and N
regions, are not even visible in a cross-section view. Nevertheless, they can have a profound effect on
many important analog and mixed-signal circuit characteristics.
subtle problem is a partially connected via, which may exhibit an abnormally high contact resis-
tance. Depending on the amount of excess resistance, the results of a partially connected via can
range from minor DC offset problems to catastrophic distortion problems.
Figure 1.6 shows incomplete etching of the metal surrounding a circuit trace. Incomplete
etching can result in catastrophic shorts between circuit nodes. Finally, Figure 1.7 shows a
surface defect caused by particulate matter landing on the surface of the wafer or on a photo-
graphic mask during one of the processing steps. Again, this type of defect results in a short
between circuit nodes. Other catastrophic defects include surface scratches, broken bond wires,
and surface explosions caused by electrostatic discharge in a mishandled device. Defects such
as these are the reason each semiconductor device must be tested before it can be shipped to
the customer.
It has been said that production testing adds no value to the final product. Testing is an
expensive process that drives up the cost of integrated circuits without adding any new func-
tionality. Testing cannot change the quality of the individual ICs; it can only measure quality
if it already exists. However, semiconductor companies would not spend money to test prod-
ucts if the testing process did not add value. This apparent discrepancy is easily explained if
we recognize that the product is actually the entire shipment of devices, not just the individual
ICs. The quality of the product is certainly improved by testing, since defective devices are
not shipped. Therefore, testing does add value to the product, as long as we define the product
correctly.
Test DUT
stimulus DUT response
Good
or Pass
bad
?
Fail
Sometimes the test engineer is also responsible for developing hardware and software that
modifies the structure of the semiconductor die to adjust parameters like DC offset and AC gain,
or to compensate for grotesque manufacturing defects. Despite claims that production testing
adds no value, this is one way in which the testing process can actually enhance the quality of the
individual ICs. Circuit modifications can be made in a number of ways, including laser trimming,
fuse blowing, and writing to nonvolatile memory cells.
The test engineer is also responsible for reducing the cost of testing through test time
reductions and other cost-saving measures. The test cost reduction responsibility is shared
with the product engineer. The product engineer’s primary role is to support the production
of the new device as it matures and proceeds to profitable volume production. The product
engineer helps identify and correct process defects, design defects, and tester hardware and
software defects.
Sometimes the product engineering function is combined with the test engineering function,
forming a single test/product engineering position. The advantage of the combined job function
is that the product engineering portion of the job can be performed with a much more thorough
understanding of the device and test program details. The disadvantage is that the product engi-
neering responsibilities may interfere with the ability of the engineer to become an expert on the
use of the complex test equipment. The choice of combined versus divided job functions is highly
dependent on the needs of each organization.
After the leads have been trimmed and formed, the devices are ready for final testing
on a second ATE tester. Final testing guarantees that the performance of the device did not
shift during the packaging process. For example, the insertion of plastic over the surface of
the die changes the electrical permittivity near the surface of the die. Consequently, trace-
to-trace capacitances are increased, which may affect sensitive nodes in the circuit. In addi-
tion, the injection-molded plastic introduces mechanical stresses in the silicon, which may
consequently introduce DC voltage shifts. Final testing also guarantees that the bond pads are
all connected and that the die was not cracked, scratched, or otherwise damaged in the pack-
aging process. After final testing, the devices are ready for shipment to the end-equipment
manufacturer.
Figure 1.11. Octal site device interface board (DIB) showing DUT sockets (left) and local circuits
with RF interface (right).
local circuits such as load circuits and buffer amplifiers that are often required for mixed-
signal device testing. Figure 1.11 illustrates the top and bottom sides of an octal site DIB. The
topside shown on the left displays eight DUT sockets, and the picture on the right shows the
local circuits and RF interface.
1.4.3 Handlers
Handlers are used to manipulate packaged devices in much the same way that probers are used
to manipulate wafers. Most handlers fall into two categories: gravity-fed and robotic. Robotic
handlers are also known as pick-and-place handlers. Gravity-fed handlers are normally used with
dual inline packages, while robotic handlers are used with devices having pins on all four sides or
pins on the underside (ball grid array packages, for example).
Either type of handler has one main purpose: to make a temporary electrical connection
between the DUT pins and the DIB board. Gravity-fed handlers often perform this task using a
contactor assembly that grabs the device pins from either side with metallic contacts that are in
turn connected to the DIB board. Robotic handlers usually pick up each device with a suction arm
and then plunge the device into a socket on the DIB board.
Chapter 1 • Overview of Mixed-Signal Testing 13
Test
head
Prober
In addition to providing a temporary connection to the DUT, handlers are also responsible for
sorting the good DUTs from the bad ones based on test results from the ATE tester. Some handlers
also provide a controlled thermal chamber where devices are allowed to “soak” for a few minutes
so they can either be cooled or heated before testing. Since many electrical parameters shift with
temperature, this is an important handler feature.
device, DIB hardware, and software on the ATE tester. Any design problems are reported to the
design engineers, who then begin evaluating possible design errors. A second design pass is often
required to correct errors and to align the actual circuit performance with specification require-
ments. Finally, the corrected design is released to production by the product engineer, who then
supports the day-to-day manufacturing of the new product.
Of course, the idealized concurrent engineering flow is a simplification of what happens in a
typical company doing business in the real world. Concurrent engineering is based on the assump-
tion that adequate personnel and other resources are available to write test plans and generate
test hardware and software before the first silicon wafers arrive. It also assumes that only one
additional design pass is required to release a device to production. In reality, a high-performance
device may require several design passes before it can be successfully manufactured at a profit.
This flow also assumes that the market does not demand a change in the device specifications in
midstream - a poor assumption in a dynamic world. Nevertheless, concurrent engineering is con-
sistently much more effective than a disjointed development process with poor communication
between the various engineering groups.
finds that the test program results do not agree with measurements taken using bench equip-
ment in their lab. The test engineer must determine which answer is correct and why there is
a discrepancy. It is also common to find that two supposedly identical testers or DIB boards
give different answers or that the same tester gives different answers from day to day. These
problems frequently result from obscure hardware or software errors that may take days to
isolate. Correlation efforts can represent a major portion of the time spent debugging a test
program.
system feature, although duplicate tester instruments must be added to the tester to allow simulta-
neous testing on multiple DUT sites.
Clearly, production test economics is an extremely important issue in the field of mixed-
signal test engineering. Not only must the test engineer perform accurate measurements of mixed-
signal parameters, but the measurements must be performed as quickly as possible to reduce
production costs. Since a mixed-signal test program may perform hundreds or even thousands
of measurements on each DUT, each measurement must be performed in a small fraction of a
second. The conflicting requirements of low test time and high accuracy will be a recurring theme
throughout this book.
PROBLEMS
1.1. List four examples of analog circuits.
1.2. List four examples of mixed-signal circuits.
1.3. Problems 1.3–1.6 relate to the cellular telephone in Figure 1.2. Which type of mixed-signal
circuit acts as a volume control for the cellular telephone earpiece?
1.4. Which type of mixed-signal circuit converts the digital samples into speaker’s voice?
1.5. Which type of mixed-signal circuit converts incoming modulated voice data into digital
samples?
1.6. Which type of digital circuit vocodes the speaker’s voice samples before they are passed to
the base-band interface?
1.7. When a PGA is combined with a digital logic block to keep a signal at a constant level,
what is the combined circuit called?
1.8. Assume a particle of dust lands on a photomask during the photolithographic printing pro-
cess of a metal layer. List at least one possible defect that might occur in the printed IC.
1.9. Why does the cleanliness of the air in a semiconductor fabrication area affect the number
of defects in IC manufacturing?
1.10. List at least four production steps after wafers have been fabricated.
1.11. Why would it be improper to draw conclusions about a design based on characterization
data from one or two devices?
1.12. List three main components of an ATE tester.
1.13. What is the purpose of a DIB board?
1.14. What type of equipment is used to handle wafers as they are tested by an ATE tester?
1.15. List three advantages of concurrent engineering.
1.16. What is the purpose of a test plan?
1.17. List at least four challenges faced by the mixed-signal test engineer.
1.18. Assume that a test program runs on a tester that costs the company 3 cents per second to
operate. This test cost includes tester depreciation, handler depreciation, electricity, floor
space, personnel, and so on. How much money can be saved per year by reducing a 5-s test
program to 3.5 s, assuming that 5 million devices per year are to be shipped. Assume that
only 90% of devices tested are good and that the average time to find a bad device drops to
0.5 s.
1.19. Assume that the profit margin on the device in Problem 1.18 is 20% (i.e., for each $1 worth
of devices shipped to the customer, the company makes a profit of 20 cents). How many
dollars worth of product would have to be shipped to make a profit equal to the savings
offered by the streamlined test program in Problem 1.18? If each device sells for $1.80,
how many devices does this represent? What obvious conclusion can we draw about the
importance of test time reduction versus the importance of selling and shipping additional
devices?
18 AN INTRODUCTION TO MIXED-SIGNAL IC TEST AND MEASUREMENT
REFERENCES
1. M. Burns, High Speed Measurements Using Undersampled Delta Modulation, 1997 Teradyne
User’s Group proceedings, Teradyne, Inc., Boston.
2. M. Abramovici, M. A. Breuer, and A. D. Friedman, Digital Systems Testing and Testable Design,
revised printing, IEEE Press, New York, January 1998, ISBN 0780310624.
3. P. K. Lala, Practical Digital Logic Design and Testing, Prentice Hall, Upper Saddle River, NJ,
1996, ISBN 0023671718.
4. J. Max Cortner, Digital Test Engineering, John Wiley & Sons, New York, 1987, ISBN
0471851353.
5. D. A. Johns and K. Martin, Analog Integrated Circuit Design, John Wiley & Sons, New York,
1996, ISBN 0471144487.
CHAPTER 2
Tester Hardware
T his chapter explores the architecture of a mixed-signal ATE tester. While we do not focus on
any particular ATE from any specific vendor, our intent here is to give a general overview of
the common instruments found in such testers. This will include a discussion on DC sources and
meters, waveform digitizers, arbitrary waveform generators, an RF measurement subsystem, and
digital pattern generators with sources and capture functionality. We shall encounter these instru-
ments again from time to time throughout the remainder of this textbook.
19
20 AN INTRODUCTION TO MIXED-SIGNAL IC TEST AND MEASUREMENT
2.2 DC RESOURCES
2.2.1 General-Purpose Multimeters
Most testers incorporate a high-accuracy multimeter that is capable of making fast DC measure-
ments. A tester may also provide a slower, very high-accuracy voltmeter for more demanding mea-
surements such as those needed in focused calibrations. However, this slower instrument may not
be usable for production tests because of the longer measurement time. The fast, general-purpose
multimeter is used for most of the production tests requiring a nominal level of accuracy.
Digital
Source Capture
vector
memory memory
memory
Tester Digital
computer pin cards
Array DUT
p roce ssor
Time
DC sources measurement
Relay matrix system
and
Sine source other VM DC Meter
interconnects
AW G Digitizer
Clocks and
synchronization
Chapter 2 • Tester Hardware 21
Figure 2.1 shows a generic mixed-signal tester architecture. It includes system computers, DC
sources, DC meters, relay control lines, relay matrix lines, time measurement hardware, arbitrary
waveform generators, waveform digitizers, clocking and synchronization sources, and a digital
subsystem for generating and evaluating digital patterns and signals. This chapter will briefly
examine the operation of each of these tester subsystems.
A detailed DC multimeter structure is shown in Figure 2.2. This meter can handle either
single-ended or differential inputs. Its architecture includes a high-impedance differential to
single-ended converter (instrumentation amplifier), a low-pass filter, a programmable gain
amplifier (PGA) for input ranging, a high-linearity ADC, integration hardware, and a sample-
and-difference stage. It also includes an input multiplexer stage to select one of several input
signals for measurement.
The instrumentation amplifier provides a high-impedance differential input. The high imped-
ance avoids potential DC offset errors caused by bias current leaking into the meter. For single-
ended measurements, the low end of the meter may be connected to ground through relays in the
input selection multiplexer. The multimeter can also be connected to any of the tester’s general-
purpose DC voltage sources to measure their output voltage. The meter can also measure current
flowing from any of the DC sources. This capability is very useful for measuring power supply
currents, impedances, leakage currents, and other common DC parametric values. A PGA placed
before the meter’s ADC allows proper ranging of the instrument to minimize the effects of the
ADC’s quantization error (see Section 5.2.4 and Section 5.5).
The meter may also include a low-pass filter in its input path. The low-pass filter removes
high-frequency noise from the signal under test, improving the repeatability of DC measurements.
This filter can be enabled or bypassed using software commands. It may also have a program-
mable cutoff frequency so that the test engineer can make tradeoffs between measurement repeat-
ability and test time (see Section 5.6). In addition, some meters may include an integration stage,
which acts as a form of hardware averaging circuit to improve measurement repeatability.
Finally, a sample-and-difference stage is included in the front end of many ATE multim-
eters. The sample-and-difference stage allows highly accurate measurements of small differences
between two large DC voltages. During the first phase of the measurement, a hardware sample-
and-hold circuit samples a voltage. This first reference voltage is then subtracted from a second
voltage (near the first voltage) using an amplifier-based subtractor. The difference between the
Input select i on
m u l t i pl e x e r Sample and Low-pass
difference filter
Normal
inputs PGA
V oltage /
current
sources Diff e re n t i al to Range control
single-ende d
convert er
(instrumentation
amplifier)
Integrator
Met er
ADC
re sul t
22 AN INTRODUCTION TO MIXED-SIGNAL IC TEST AND MEASUREMENT
EXAMPLE 2.1
A single-ended DC voltmeter has a resolution of 12 bits. It also features a sample-and-difference
front-end circuit. We wish to use this meter to measure the differential offset voltage of a DUT’s
output buffer. Each of the two outputs is specified to be within a range of 1.35 V ± 10 mV, and
the differential offset is specified to be ±5 mV. The meter input can be set to any of the following
ranges: ±10 V, ±1 V, ±100 mV, ±10 mV, and ±1 mV. Assuming that all components in the meter
are perfectly linear (with the exception of the meter’s quantization error), compare the accuracy
achieved using two simple DC measurements with the accuracy achieved using the sample-and-
difference circuit.
Solution:
The simplest way to measure offset using a single-ended DC voltmeter is to connect the met-
er to the OUTP output, measure its voltage, connect the meter to the OUTN output, measure
its voltage, and subtract the second voltage from the first. Using this approach, we have to set
the meter’s input range to ±10 V to accommodate the 1.35 V DUT output signals. Thus each
measurement may have a quantization error of as much as ± ½ (20 V/212–1) = ± 2.44 mV.
Therefore the total error might be as high as ±4.88 mV, assuming that the quantization er-
ror from the first measurement is positive, while the quantization error from the second mea-
surement is negative. Since the specification limit is ±5 mV, this will be an unacceptable test
method.
Using the sample-and-difference circuitry, we could range the meter input to the worst-
case difference between the two outputs, which is 5 mV, assuming a good device. The lowest
meter range that will accommodate a 5-mV signal is ±10 mV. However, we also need to be
able to collect readings from bad devices for purposes of characterization. Therefore, we will
choose a range of ±100 mV, giving us a compromise between accuracy and characterization
flexibility.
During the first phase of the sample-and-difference measurement, the voltage at the OUTN
pin is sampled onto a holding capacitor internal to the meter. Then the meter is connected to the
OUTP pin and the second phase of the measurement amplifies the difference between the OUTP
voltage and the sampled OUTN voltage. Since the meter is set to a range of ±100 mV, a 100-mV
difference between OUTP and OUTN will produce a full-scale 10-V input to the meter’s ADC. This
serves to reduce the effects of the meter’s quantization error. The maximum error is given by
± ½ (100 mV/212–1) = ± 12.2 μV. Again, the worst-case error is twice this amount, or ±24.4 μV,
which is well within the requirements of our measurement.
two voltages is then amplified and measured by the meter’s ADC, resulting in a high-resolution
measurement of the difference voltage. This process reduces the quantization error that would
otherwise result from a direct measurement of the large voltages using the meter’s higher voltage
ranges.
[5] Vergl. Helens Brief S. 161 ff. und Fräulein Sullivans Bericht
S. 311 ff.
Vierzehntes Kapitel.
Die Frostkönig-Episode. — Betrachtungen über Schriftstellerei.
Der Winter des Jahres 1892 wurde durch eine Wolke an dem
heiteren Himmel meiner Kindheit getrübt. Anstatt der Freude waren
für lange, lange Zeit Zweifel, Sorge und Furcht bei mir eingekehrt.
Die Bücher verloren ihren Reiz für mich, und noch jetzt schnürt sich
mir bei dem Gedanken an jene schrecklichen Tage das Herz
zusammen. Eine kleine Geschichte mit dem Titel »Der Frostkönig«,
[6] die ich schrieb und an Herrn Anagnos, den Direktor des
Perkinsschen Blindeninstitutes schickte, bildete die Veranlassung zu
all der Unruhe. Um die Sache klarzustellen, muß ich die Tatsachen in
Verbindung mit folgender Episode auseinandersetzen, die zu
erwähnen mich sowohl die Gerechtigkeit gegen meine Lehrerin wie
gegen mich selbst nötigt.
Ich schrieb die Erzählung, als ich zu Hause war, in dem Herbste,
nachdem ich sprechen gelernt hatte. Wir waren von Fern Quarry
später als gewöhnlich aufgebrochen. Während unseres Aufenthaltes
dort hatte mir Fräulein Sullivan die Schönheiten des herbstlichen
Laubes beschrieben, und es scheint, als hätten ihre Schilderungen
die Erinnerung an ein Märchen wachgerufen, das mir offenbar
einmal vorgelesen worden war und das ich unbewußt behalten
haben muß. Ich glaubte damals eine „Geschichte zu machen“, wie
die Kinder sagen, und setzte mich voller Eifer hin, die
niederzuschreiben, ehe sich die Gedanken wieder verflüchtigten. Die
Gedanken flossen mir leicht aus der Feder; ich empfand lebhafte
Freude bei der Ausarbeitung. Worte und Bilder strömten mir in
reicher Fülle zu, und während ich mir einen Satz nach dem anderen
ausdachte, schrieb ich alles mit meinem Braillegriffel nieder. Wenn
mir jetzt Worte und Bilder ohne besondere Anstrengung kommen, so
betrachte ich dies als einen ziemlich sicheren Beweis dafür, daß sie
nicht mein geistiges Eigentum, sondern fremdes Gut sind, von dem
ich nichts wissen will. Damals aber nahm ich alles, was ich las,
begierig auf ohne irgend einen Gedanken an Verfasserrecht, und
selbst jetzt kann ich die Grenzlinie zwischen meinen Gedanken und
denen, die ich in meinen Büchern finde, nicht ganz scharf ziehen.
Ich glaube, dies liegt daran, daß mir soviele Eindrücke durch die
Vermittlung der Augen und Ohren anderer zugehen.
Als ich mit meiner Erzählung fertig war, las ich sie meiner
Lehrerin vor, und ich erinnere mich noch jetzt lebhaft der Freude, die
ich bei den gelungenen Stellen empfand, sowie meiner Ungeduld,
wenn ich durch die Verbesserung der Aussprache eines Wortes
unterbrochen wurde. Beim Mittagessen wurde sie der versammelten
Familie vorgelesen, die ganz erstaunt war, daß ich so gut schrieb. Es
fragte mich auch jemand, ob ich sie nicht in irgend einem Buche
gelesen hätte.
Diese Frage überraschte mich sehr, denn ich hatte nicht die
geringste Erinnerung daran, daß sie mir je vorgelesen worden sei.
Ich sagte daher mit aller Entschiedenheit: O nein, es ist eine
Geschichte von mir, und ich habe sie für Herrn Anagnos geschrieben.
Demgemäß schrieb ich die Erzählung ab und schickte sie dem
genannten Herrn zu seinem Geburtstage. Es wurde mir geraten, den
Titel, der ursprünglich »Herbstlaub« (Autumn Leaves) lautete, in
»Der Frostkönig« (The Frost King) umzuändern, was ich denn auch
tat. Ich trug die kleine Erzählung selbst zur Post, und es war mir
dabei zu Mute, als ob ich in den Wolken schwebte. Ich ließ mir wenig
davon träumen, wie hart ich für dieses Geburtstagsgeschenk zu
büßen haben würde.
Herr Anagnos war über den »Frostkönig« entzückt und
veröffentlichte das Märchen in einem seiner Jahresberichte über das
Perkinssche Institut. Dies war der Gipfel meiner Glückseligkeit, von
dem ich aber bald jäh wieder zur Erde geschleudert werden sollte.
Ich war nur kurze Zeit in Boston gewesen, als es sich herausstellte,
daß eine ähnliche Geschichte wie »Der Frostkönig«, nämlich »Die
Frostelfen« (The Frost Fairies) von Fräulein Margaret T. Canby, vor
meiner Geburt in einem Buche mit dem Titel »Birdie und seine
Freunde« (Birdie and His Friends) erschienen sei. Die beiden
Erzählungen stimmten in Inhalt und Form so sehr überein, daß kein
Zweifel darüber bleiben konnte, daß Fräulein Canbys Märchen mir
vorgelesen worden sein mußte, und daß das meinige — ein Plagiat
war. Es hielt schwer, mir dies verständlich zu machen; als ich es aber
begriffen hatte, war ich tief betrübt. Kein Kind hat je einen bittereren
Kelch getrunken als ich. Ich hatte mir Schimpf und Schande
zugezogen, ich hatte Verdacht bei denen erregt, die ich am meisten
liebte. Und doch, wie war es möglich, daß so etwas geschehen
konnte? Ich zermarterte mein Gehirn unablässig, um mich an irgend
etwas zu erinnern, was ich über den Frost gelesen haben könnte,
bevor ich den »Frostkönig« schrieb; ich konnte mich aber auf nichts
entsinnen als auf die volkstümliche Redensart von Jack Frost und ein
Kindergedichtchen: »Die Launen des Frostes« (The Freaks of the
Frost), und ich wußte, daß ich dieses nicht bei meiner Arbeit benutzt
hatte.
Zunächst schien mir Herr Anagnos zu glauben, obgleich er
großen Kummer darüber empfand. Er war außergewöhnlich zärtlich
und liebevoll zu mir, und eine kurze Zeitlang verschwand der
Schatten. Ihm zuliebe suchte ich mich zu fassen und mich zur Feier
von Washingtons Geburtstag, der bald nach dem peinlichen
Zwischenfall festlich begangen wurde, so hübsch wie möglich zu
machen.
Ich sollte die Ceres in einer Art von Maskenspiel darstellen, das
von den blinden Kindern aufgeführt wurde. Wie gut erinnere ich
mich an das reizvolle Gewand, das mich umhüllte, an das bunte
Herbstlaub, das mein Haupt schmückte, an die Früchte und Aehren
zu meinen Füßen und in meinen Händen, und unter all der Heiterkeit
des Maskenspiels das drückende Bewußtsein eines nahenden
Unheils, das mir das Herz schwer machte!
Am Abend vor der Feier hatte eine der Institutslehrerinnen eine
Frage betreffs des »Frostkönigs« an mich gerichtet, und ich hatte ihr
geantwortet, daß Fräulein Sullivan mir von Jack Frost und seinen
Wunderwerken erzählt habe. Irgend eine Aeußerung von mir schien
sie als Geständnis aufzufassen, daß ich mich an Fräulein Canbys
Märchen von den »Frostelfen« erinnere, und sie teilte Herrn
Anagnos dies mit, obgleich ich ihr ganz entschieden erklärte, sie
habe mich mißverstanden.
Herr Anagnos, der mich zärtlich liebte, blieb den Beteuerungen
meiner Liebe und Unschuld gegenüber taub, da er getäuscht worden
zu sein glaubte. Er war der Meinung oder hegte wenigstens den
Verdacht, daß Fräulein Sullivan und ich uns bewußt die Gedanken
einer anderen angeeignet und sie ihm in betrügerischer Absicht
zugeschickt hätten, um Bewunderung bei ihm zu finden. Ich wurde
vor ein Gericht gestellt, das aus den Lehrern und Beamten des
Institute bestand, und Fräulein Sullivan wurde aufgefordert, mich
allein zu lassen. Dann wurde ich einem förmlichen Kreuzverhör
unterworfen, das mich auf die Vermutung brachte, meine Richter
seien fest entschlossen, mich zu dem Geständnis zu zwingen, ich
erinnerte mich, daß mir das Märchen »Die Frostelfen« vorgelesen
worden sei. Ich fühlte aus jeder Frage den Zweifel und den Verdacht
heraus, den sie in ihrem Innern hegten, und ebenso empfand ich es,
daß ein geliebter Freund uns vorwurfsvoll betrachtete, obgleich ich
dies alles nicht in Worte fassen konnte. Alles Blut drängte sich mir
nach meinem wild pochenden Herzen, und ich konnte kaum
sprechen außer in abgerissenen Worten und Silben. Selbst das
Bewußtsein, das Ganze sei nur ein furchtbares Mißverständnis,
konnte meinen Schmerz nicht lindern, und als ich schließlich das
Zimmer verlassen durfte, war ich noch ganz außer mir und achtete
weder auf die Liebkosungen meiner Lehrerin noch auf die zärtlichen
Worte meiner Freunde, die mir sagten, ich sei ein braves Mädchen,
auf das man stolz sein könne.
Als ich diese Nacht in meinem Bette lag, weinte ich so
herzbrechend, wie hoffentlich wenige Kinder geweint haben. Mir war
so eisig kalt, daß ich glaubte, den nächsten Morgen nicht mehr zu
erleben, und dieser Gedanke tröstete mich. Ich glaube, wenn dieser
Schlag mich einige Jahre später getroffen hätte, so würde mein Geist
unrettbar zusammengebrochen sein. Aber der Engel des Vergessens
hat viel von dem Elend und der Bitternis dieser traurigen Tage
aufgesammelt und mit sich fortgenommen.
Fräulein Sullivan hatte nie von den »Frostelfen« oder dem Buch,
in dem das Märchen erschienen war, gehört. Mit Dr. Alexander
Grahams Hilfe untersuchte sie die Sache gründlich, und schließlich
stellte es sich heraus, daß Frau Sophia C. Hopkins im Jahre 1888 ein
Exemplar von Fräulein Canbys »Birdie und seine Freunde« besaß, als
wir den Sommer mit ihr in Brewster zubrachten. Frau Hopkins
konnte das Buch nicht mehr finden, sie hat mir aber erzählt, daß sie,
während Fräulein Sullivan auf einer Ferienreise begriffen war,
versucht habe, mir durch Vorlesen aus verschiedenen Büchern die
Zeit zu vertreiben, und obgleich sie sich nicht deutlicher als ich
erinnern konnte, die »Frostelfen« gelesen zu haben, war sie doch
ganz sicher, daß sich ein Exemplar von »Birdie und seine Freunde«
unter diesen Büchern befunden habe. Sie erklärte sich das Fehlen
des Buches dadurch, daß sie vor kurzem ihr Haus verkauft und dabei
verschiedene Jugendschriften, sowie alte Schulbücher und Märchen
verschenkt hatte, und daß sich die betreffende Erzählung
wahrscheinlich unter diesen befunden hätte.
Erzählungen hatten damals wenig oder gar kein Interesse für
mich; aber das bloße Buchstabieren der seltsamen Worte genügte,
einem kleinen Mädchen die Zeit zu vertreiben, das selber beinahe
nichts zu seiner Unterhaltung beitragen konnte, und obgleich ich
mich keines einzelnen Umstandes bei dieser Lektüre entsinne, kann
ich doch nicht umhin zu glauben, daß ich mir die größte Mühe
gegeben habe, die Worte zu behalten, in der Absicht, sie meiner
Lehrerin nach ihrer Rückkehr zu wiederholen. Das eine ist
unzweifelhaft, die Sprache war mir unauslöschlich eingeprägt,
obgleich dies lange Zeit niemand wußte, am wenigsten ich selbst.
Als dann Fräulein Sullivan zurückkam, sprach ich mit ihr nicht
über die »Frostelfen«, wahrscheinlich, weil sie sofort begann, mir
den »Kleinen Lord Fauntleroy«[7] vorzulesen, der mich so
begeisterte, daß ich an nichts anderes denken konnte. Aber die
Tatsache bleibt bestehen, daß mir Fräulein Canbys Märchen früher
vorgelesen worden war und daß es sich mir, lange nachdem ich es
vergessen hatte, mit solcher Ursprünglichkeit wieder aufdrängte, daß
ich nie auf den Verdacht geriet, es könne das Geisteskind einer
anderen sein.
Mitten in meinem Schmerze erhielt ich viele Beweise der Liebe
und Teilnahme. Alle Freunde, die ich am meisten liebte, sind mir
damals bis auf einen treu geblieben. Fräulein Canby selbst schrieb
mir freundlich: Eines Tages werden Sie ein Märchen eigener
Erfindung schreiben, das viele aufrichten und erheben wird. — Aber
diese Prophezeiung ist nie in Erfüllung gegangen. Ich habe nie mehr
zum Zwecke der bloßen Unterhaltung mit Worten gespielt. In der Tat
bin ich seitdem stets von dem Gedanken gequält worden, daß das,
was ich schreibe, nicht mein geistiges Eigentum ist. Lange Zeit
wurde ich, wenn ich einen Brief schrieb, selbst an meine Mutter, von
einem plötzlichen Angstgefühl befallen und ich zergliederte meine
Sätze auf das genaueste, um sicher zu sein, sie nicht in einem Buche
gelesen zu haben. Ohne den unausgesetzten Zuspruch Fräulein
Sullivans würde ich, wie ich glaube, jeden weiteren Versuch, mich
schriftstellerisch zu betätigen, aufgegeben haben.
Ich habe seitdem die »Frostelfen« gelesen und ebenso Briefe von
mir, in denen ich noch andere Gedanken Fräulein Canbys benutzt
habe. In einem von ihnen, einem Briefe an Herrn Anagnos vom 29.
September 1891, finde ich Worte und ganze Sätze, die deutlich an
jenes Buch erinnern. Um dieselbe Zeit schrieb ich den »Frostkönig«,
und dieser Brief enthält gleich vielen anderen eine Anzahl
Redewendungen, die beweisen, daß mein Geist ganz mit dem
Märchen gesättigt war. Ich lege meiner Lehrerin folgende Worte über
das goldene Herbstlaub in den Mund: Ja, es ist schön genug, um
uns über die Flucht des Sommers zu trösten — ein Gedanke, der
unmittelbar aus Fräulein Canbys Geschichte stammt.
Diese Gewohnheit, mir zu assimilieren, was mir gefiel, und es
dann als mein Eigentum auszugeben, tritt vielfach in meinem
frühesten Briefwechsel und meinen ersten schriftstellerischen
Versuchen zutage. In einem Aufsatze, den ich über die alten Städte
Griechenlands und Italiens schrieb, entnahm ich meine glühenden
Schilderungen Quellen, die ich jetzt vergessen habe. Ich kannte
Herrn Anagnos’ Vorliebe für das Altertum und seine begeisterte
Verehrung für Italien und Griechenland. Ich brachte daher aus allen
Büchern, die ich las, die Brocken von Poesie oder Geschichte an, die
ihm, wie ich glaubte, Vergnügen bereiten würden. Herr Anagnos
hatte bei der Besprechung meines Aufsatzes gesagt: Diese
Gedanken sind in ihrem Kerne poetisch. Aber ich verstehe nicht, wie
er je hat der Meinung sein können, ein blindes und taubstummes
Kind von elf Jahren habe diese selbständig gefunden. Doch kann ich
nicht glauben, daß, weil diese Gedanken nicht meinem eigenen
Kopfe entsprungen sind, mein kleiner Aufsatz aus diesem Grunde
alles Interesses bar sein sollte. Er beweist mir, daß ich imstande war,
schöne, poetische Gedanken in klaren, lebendigen Worten
wiederzugeben.
Jene Jugendaufsätze stellten eine geistige Gymnastik dar. Ich
lernte, wie es alle jungen, unerfahrenen Leute tun, durch
Assimilation und Nachahmung, Gedanken in Worte zu kleiden. Alles,
was ich in einem Buche fand und was mir gefiel, bewahrte ich,
bewußt oder unbewußt, in meinem Gedächtnisse auf und paßte es
meinen Zwecken an. „Wenn man zu schreiben beginnt,“ sagt
Stevenson, „versucht man unwillkürlich nachzuahmen, was einem
am bewundernswertesten erscheint, und wechselt auffallend rasch
mit den Gegenständen seiner Bewunderung. Selbst große Männer
haben erst nach jahrelanger Uebung gelernt, die Legion von Worten,
die sich auf allen möglichen Nebenwegen ihrem Geiste aufdrängten,
in gehörige Ordnung zu bringen.“
Ich fürchte, ich stehe noch jetzt mitten in dieser Entwickelung
drin. Es ist klar, daß ich nicht immer meine eigenen Gedanken von
denen, die ich irgendwo gelesen habe, sondern kann, eben weil das,
was ich lese, das eigentliche Wesen und Gefüge meines Geistes
ausmacht. Infolgedessen fördere ich beinahe in allem, was ich
schreibe, etwas zutage, was große Aehnlichkeit mit der
ungeschickten Stoppelei aufweist, die ich zustande brachte, als ich
anfing, nähen zu lernen. Dieses Stoppelwerk bestand aus allerlei
Fetzen und Lappen — hübschen Stückchen Seide und Sammet, aber
die häßlichen Flicken, die durchaus keinen gefälligen Eindruck
machten, herrschten stets vor. Ebenso bestehen meine
schriftstellerischen Leistungen aus unverarbeiteten eigenen
Begriffen, untermischt mit den klareren Gedanken und gereifteren
Ansichten der Autoren, deren Bücher ich gelesen habe. Die
Hauptschwierigkeit beim Schreiben scheint mir darin zu bestehen,
daß die Sprache des hochgebildeten Geistes unsere verworrenen
Ideen — halb Empfindungen, halb Gedanken — zu einer Zeit
ausdrücken soll, da wir wenig mehr sind als Bündel instinktiver
Antriebe. Die ersten schriftstellerischen Versuche haben große
Aehnlichkeit mit einem Zusammenlegespiel. Wir sehen im Geiste ein
Muster vor uns, das wir mit Worten darzustellen wünschen, aber die
Worte passen nicht in die Zwischenräume, und wenn sie es tun,
stimmen sie nicht mit der Zeichnung überein. Aber wir fahren in
unseren Versuchen fort, weil wir sehen, daß andere Erfolg gehabt
haben, und wir nicht gewillt sind, unseren Mangel an Begabung
zuzugeben.
„Es gibt kein anderes Mittel, originell zu werden, als so geboren
zu sein,“ sagt Stevenson, und obgleich ich gar nicht originell sein
mag, so hoffe ich doch, dereinst meinen erkünstelten, unnatürlichen
schriftstellerischen Versuchen zu entwachsen. Dann werden vielleicht
meine eigenen Gedanken und Erfahrungen zutage treten.
Inzwischen vertraue ich, hoffe ich, arbeite ich unermüdet weiter und
suche es zu verhindern, daß die bittere Erinnerung an den
»Frostkönig« etwa meine Kreise störe.
So hat diese traurige Erfahrung für mich auch etwas Gutes im
Gefolge gehabt: sie ist die Veranlassung gewesen, daß ich über
einige Probleme der Schriftstellerei nachgedacht habe. Ich bedaure
nur das eine, daß der Vorfall den Verlust eines meiner teuersten
Freunde, des Herrn Anagnos, zur Folge hatte.
Nach der Veröffentlichung der »Geschichte meines Lebens« im
Ladies’ Home Journal hat Herr Anagnos in einem Briefe an Herrn
Macy geäußert, er habe mich zu der Zeit, als sich der Vorfall mit dem
»Frostkönig« abspielte, für unschuldig gehalten. Er erklärt, das
Gericht, vor das ich gestellt wurde, habe aus acht Mitgliedern
bestanden: vier blinden und vier sehenden. Vier von diesen,
behauptet er, waren der Ansicht, Fräulein Canbys Erzählung sei mir
vorgelesen worden, während die anderen vier die entgegengesetzte
Meinung vertraten. Herr Anagnos erklärt, seine Stimme zu meinen
Gunsten abgegeben zu haben.
Wie aber auch die Sache gewesen sein und in welchem Sinne er
seine Stimme abgegeben haben mag, das eine ist sicher: als ich in
das Zimmer trat, in dem mich Herr Anagnos so oft auf seinen Knien
gehalten und seine vielfachen Sorgen über meiner Lustigkeit
vergessen hatte, und hier Personen antraf, die Zweifel in mich zu
setzen schienen, fühlte ich, daß etwas Feindseliges und Drohendes
in der Atmosphäre lag, und die nachfolgenden Ereignisse haben
diesen Eindruck bestätigt. Zwei Jahre lang scheint Herr Anagnos an
der Ansicht festgehalten zu haben, daß Fräulein Sullivan und ich
unschuldig seien. Dann änderte er offenbar seine günstige Meinung,
aus welchem Grunde, weiß ich nicht. Auch die Einzelheiten der
Untersuchung kenne ich nicht, und selbst die Namen der Mitglieder
des »Gerichtshofes«, die überdies während der ganzen Verhandlung
kein Wort zu mir sprachen, sind mir unbekannt geblieben. Ich war zu
aufgeregt, um auf irgend etwas zu achten, zu eingeschüchtert, um
Fragen zu stellen. In der Tat kann ich mich kaum entsinnen, was ich
sagte, oder was zu mir gesagt wurde.
Ich habe den Vorfall mit dem »Frostkönig« so ausführlich
dargestellt, da er für mein Leben und meine Erziehung von
Wichtigkeit war, und um kein Mißverständnis aufkommen zu lassen,
habe ich alle Tatsachen wiedergegeben, wie sie mir erscheinen, ohne
die Absicht zu hegen, mich zu verteidigen oder irgend jemand
anzuklagen.
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
ebookgate.com