Você está na página 1de 4

designfeature By Ron Mancini, Texas Instruments Inc

THE PRUDENT ENGINEER MUST LOOK BEYOND THE FIRST


PAGE OF THE DATA SHEET; THE FINE PRINT CAN KILL YOU.

How to read a
semiconductor data sheet
emiconductor data sheets have changed a to give designers the immediate information re-

S lot in the last few years, including growing from


10 pages to a hundred. The problem is that data
sheets contain almost too much data, and there’s not
quired to make a go/no-go decision. Never discard
this rhetoric, because it often contains gem-like de-
scriptions of unusual functions or exceptional per-
enough time for a busy engineer to dwell on all that formance that may come in handy on new designs.
information. The situation demands that the design The description section often discusses special items
engineer quickly evaluate the data-sheet informa- that might be important to designers, such as uni-
tion, and the following strategy can help the engi- ty-gain stability, performance near the power rails,
neer reach the essentials in minimum time. and new function leads.
The front page sometimes includes an applica-
THE SHOW WINDOW tions section. IC designers usually design devices for
There are no standard semiconductor data sheets, a specific application, and this section lists that ap-
but the first page of every data sheet contains at- plication along with peripheral applications. The de-
tractive headlines intended to gain the designer’s at- scription section employs a paragraph format,
tention. Data sheets contain a ton of information, whereas the features and applications sections tend
and the front page holds a fraction of it. Although to be bulleted; therefore, the description often em-
the front-page information is attractive, you can’t bellishes the features and applications sections.
base a design decision on it without reading and un- The lead diagram uses a circuit symbol (such as a
derstanding the rest of the data sheet. The front page triangle for an op amp) to show lead-to-circuit con-
comprises the headline, features section, applica- nections, and the package diagram shows how the
tions section, circuit description, lead diagram, and leads are brought out of the package. This important
packaging information. Sometimes, the headline diagram is necessary to connect the IC in a circuit.
contains all guaranteed minimum/maximum data, Beware: Lead names are often inconsistent through-
and sometimes the front page gives all typical val- out a data sheet. The front page may call a lead a
ues. The inconsistency exists because the product “plus input” or identify it with a plus sign, but later
team works together from design to marketing, text may call it a “noninverting input lead.”
keeping the customer’s needs in mind; the team puts The next item to examine is the absolute-maxi-
information on the front page that will appeal to its mum ratings table. If the application exceeds these
target audience.
140
The features section tries to exhibit the unique at- 130
tributes of the IC, such as rail-to-rail operation, flags, 120
V =3.6V
DD

and shutdown modes. Obviously, IC applications re- 110


V =2.7V
quire essential design features, and additional fea- 100
DD

tures rarely count, even if they are free. For exam- 90


COMMON-MODE-
ple, an enable function contributes nothing to the REJECTION
80
V =1.8V
RATIO DD
design of a “continuously on” IC and may be a source (dB)
70

of problems if you incorrectly terminate it. Careful- 60

50
ly look over the features to see how you can use them
40
in a novel manner; you may not need an enable func-
30
tion, but it can function as a disable during transient 20
or failure conditions. You can use unique features in 10
many ways that IC designers and applications engi- 0
neers never imagined. 0 10 100 1k 10k 100k 1M 10M
FREQUENCY (Hz)
The description section expands on the features
section and gives more information on Typical curves illustrate the interactions between
Figure 1
salient points. The two sections combine parameters of interest and external influences.

www.edn.com April 14, 2005 | edn 85


designfeature Semiconductor data sheets

ratings, it voids the manufacturer’s guar- IC’s internal die can reach under any con- ensure that individual test conditions su-
antee regardless of the duration or the dition. If, for example, the maximum persede the general note. The test tem-
conditions of the overstress. Manufac- junction-temperature specification is perature is usually the temperature of the
turers guard-band these ratings to obtain 150⬚C and the operating-temperature free air surrounding the IC, often 25⬚C,
test tolerances, and some users try to range extends to 150⬚C, you cannot pow- but power ICs often specify the test tem-
push the envelope, taking advantage of er up the IC at 150⬚C ambient, because perature as case temperature.
the guard band by testing to establish the any applied power increases the junction The body of the tables includes the pa-
higher absolute-maximum rating. Never temperature beyond its rating. Storage rameter identification, test conditions,
exceed the absolute-maximum ratings temperature is the maximum tempera- temperature (unless it’s included in the
that the manufacturer recommends, re- ture that the IC may reach under a pow- test conditions), parameter data, and pa-
gardless of the circumstances, because er-off condition. Violating the storage- rameter units. Parameter identification
the manufacturer can change the guard temperature limit voids the IC warranty. includes the parameter symbol and name
band at any time. Most problems with Every data sheet includes an ESD (that is, input offset voltage, VIO, or off-
absolute-maximum ratings occur during (electrostatic-discharge) warning. Nev- set voltage, VOS). Life would be easier if
transient conditions, which extensive en- er ignore this warning, because ESD is al- the parameter names and symbols con-
vironmental and probable-use testing ways present and, like Murphy’s Law, formed to universal standards, but they
finds. Part of the circuit-design task of waiting to strike at the least opportune don’t, so users must translate symbols
the device user is to eliminate the tran- moment. Some users say that they do not when scanning multiple data sheets from
sient conditions or to minimize their ef- worry about ESD because they use bipo- multiple manufacturers. Furthermore,
fect using protective components. lar semiconductors. This belief is a falla- parameter names or symbols are some-
The data sheet normally gives three ab- cy, because 800V ESD can damage the times inconsistent between pages in a
solute-maximum temperature ratings. base-emitter junctions of sensitive bipo- data sheet. Rather than make assump-
Operating temperature is the maximum lar transistors, yet this voltage is low for tions, users must contact the manufac-
temperature range over which you can static, and you cannot feel it. Always fol- turer’s applications department to get
expect the IC to operate. Do not be con- low ESD-approved handling and instal- confirmation or a clear answer.
fused between operating temperature lation procedures. Carefully read the test conditions; the
and parameter specifications, because Carefully study the electrical charac- parameter values are valid only when the
there is no relationship between them, teristics in tables 1 and 2, because the de- test conditions prevail. The test-condi-
and there is no guarantee that the IC will sign data comes from them. The table tions column specifies the test ambient
meet specifications over the operating- notes specify the test temperature and temperature when the data sheet does not
temperature range. Junction temperature power-supply voltage. They include the use a separate temperature column. It also
is the maximum temperature that the comment, “unless otherwise noted,” to specifies power-supply voltages and some
of the test parameters, such as
TABLE 1—PARAMETERS FOR THE OFFSET VOLTAGE OF A SIGNAL OP AMP source resistance, load resistance,
Parameter Test conditions TA* MIN TYP MAX Unit test frequency, common-mode
VIO (input offset voltage) TLV278x 25°C 250 3000 ␮V
voltage, open-loop gain, input
VO=VDD/2 Full range 4500 signal, and any other important
RL=2 k⍀ TLV278xA 25°C 250 2000 defining test parameters.
RS=50⍀ Full range 3000 These test conditions some-
⬀VIO (temperature coefficient 8 ␮V/°C times lead to apparent conflicts
of input offset voltage) with the front page. The front
CMRR (common-mode rejection 25°C 50 76 dB page may list an ADC perform-
ratio) VIC=0 to VDD VDD=1.8V Full range 50 ance as 16 bits at 1M sample/
RS=50⍀ 25°C 55 80 sec, but the guaranteed conver-
VDD=2.7V/3.6V Full range 50 sion rate at the test condition
VIC=1.2V to VDD 25°C 70 100
yields only 14 bits of guaranteed
RS=50⍀ VDD=2.7V/3.6V Full range 70
performance. Here, the front
Notes: Electrical characteristics at specified free-air temperature, VDD=1.8V, 2.7V (unless otherwise noted).
page advertises maximum per-
*Full range is 0 to 70°C for the C suffix and ⳮ40 to +125°C for the I suffix. If unspecified, full range is ⳮ40 to
+125°C. formance for customers that can
use typical data, and the table
TABLE 2—PARAMETERS FOR THE OFFSET VOLTAGE OF A POWER OP AMP gives guaranteed performance
for the conservative customers.
OPA569 Units
Parameter Condition MIN TYP MAX After weeding out nonperform-
Offset voltage ing ICs to establish the initial se-
Input offset voltage (VOS) IO=0V, VS=5V ⫾0.5 ⫾2 mV lection based on front-page data,
versus temperature (dVOS/dT) TA=ⳮ40 to +85°C ⫾1.3 ␮V/°C designers must rely on the data
versus power supply (PSRR) VS=2.7V to 5.5V, VCM=(Vⳮ ⳮ)+0.55V 12 60 ␮V/V in the electrical-characteristics
Electrical characteristics: VS=2.7 to 5.5V. Boldface limits apply over the specified temperature range, TA=ⳮ40 to table. Test conditions usually
+85°C. At TCASE=25°C, RL=1 k⍀, and connect to VS/2, unless otherwise noted. make the difference between
Note: PSRR=power-supply rejection ratio. front-page and electrical-charac-
86 edn | April 14, 2005 www.edn.com
designfeature Semiconductor data sheets

teristics data, so carefully peruse the test has minimum and typical values and no turer data, you can obtain several lots of
conditions. maximum values. Infinite CMRR would ICs with different date codes, perform
The temperature column specifies the be a designer’s dream, so data sheets give parameter measurements, and do a sta-
test temperature when the test-condi- no maximum parameter value. tistical analysis to obtain minimum/
tions column omits this value. This col- Typical values can get novice or naive maximum values. This method is less de-
umn often has values of 25⬚C, full range, designers in trouble because they show pendable than the manufacturer’s guar-
or a numerical temperature range. Full desirable parameter values that the IC antee, but it is better than relying on
range means different temperature cannot dependably deliver. There is a ten- nonguaranteed typical data.
ranges exist for different grades of ICs; dency to lust over typical values even
thus, designers must read and under- when the data sheet gives minimum/max- FOLLOW THOSE CURVES
stand any notes pertaining to “full range.” imum values, and this path can lead Curves are valuable tools for deter-
For example, full range can be 0 to 70⬚C straight to an unreliable design. Engineers mining how one parameter interacts
for the lowest grade of an IC family (of- who prefer using minimum/maximum with another parameter, temperature,
ten C grade), or it can be ⫺40 to Ⳮ125⬚C values might be tempted to use a typical frequency, or power-supply variation.
for a premium grade (often I grade). Re- value when minimum/maximum values Figure 1 shows the CMRR-versus-fre-
gardless of any other temperature rating are unavailable, but the absence of mini- quency curve for the TLV278X op amp.
or specification, the parameter values in mum/maximum values does not make The curves show the dc CMRR as 80
the table are valid only for the specified typical values design parameters. Some- dB at VDD⫽1.8V, 110 dB at VDD⫽2.7V,
temperature or temperature range. times typical values are the mean or av- and 130 dB at VDD⫽3.6V. The electrical-
Sometimes, the operating-temperature erage value you obtain from several sam- characteristics table says that the typical
range and test-condition/temperature- ple runs of an IC, but often, typical values dc CMRR is 76 dB at VDD⫽1.8V and 80
column specifications are identical; when have little relationship to today’s ICs. An or 100 dB at VDD⫽2.7/3.6V. This data
they differ, the test-condition/tempera- IC may go through several mask sets, raises a question of why the CMRRs are
ture column values prevail. process changes, or both, throughout its unequal for equal supply voltages. The
The next column subdivides into three life, and the original typical values lose reason is that the input common-mode-
columns—MIN for minimum, TYP for their meaning after all these changes. (In voltage specification covers a voltage
typical, and MAX for maximum—that more than 20 years, I have never seen the range in Table 1 (see the test conditions).
contain the available parameter values. typical values change on the LM324 data This exercise illustrates the value of un-
After initial testing on the first several sheet after the IC process/mask was derstanding the test conditions. The
groups of ICs, the manufacturer applies changed!) TLV278X is a consistent data sheet; load
statistics to the data to obtain the mean Sometimes, designers can obtain only capacitance and resistance are constant at
value for each parameter. The statistics typical values for a parameter; this situ- 10 pF and 2 k⍀. Some data sheets are less
yield the variance, sigma; six times sigma ation is especially true when dealing with consistent, with parameters that look
represents the maximum and minimum high-speed components, because the test good until you read the fine print. For ex-
values that the parameters assume during cost can exceed the production cost. ample, I once read a Bode plot and pre-
manufacturing. These six sigma points When the question arises about whether dicted an overshoot of 40%. The pulse
become the minimum and maximum you should use typical values because curves had virtually no overshoot. The
values for that parameter, and you often they are the only data you have, the an- difference was that the Bode-plot circuit
use the mean as the typical specification. swer is no, unless you have an under- had a 1000-pF load, and the overshoot
The VIO parameter for the TLV278X standing with the manufacturer. Design- circuit had a 10-pF load.
signal op amp has typical and maximum ers should first try to obtain data and The applications-information section
values but no minimum value. Ideally, the specifications from the manufacturer, often includes parameter-measurement
value of VIO approaches zero; this situa- which often can record and log data. This information, along with unusual meas-
tion may happen someday, so the data approach can be cost-prohibitive, so the urement circuits. The application section
sheet leaves this value blank. The VIOMAX next step is to arrange for the manufac- discusses topics including load-driving ca-
specification is 3000 ␮V at TA⫽25⬚C and turer to supply GNT (guaranteed-but- pability, pc-board-layout suggestions, spe-
4500 ␮V over the full temperature range. not-tested) or GBD (guaranteed-by-de- cial stabilization techniques, Spice models,
The polarity of VIO is unspecified, and de- sign) specifications. GNT means that the special-function descriptions, safe-operat-
signers have to read outside material to manufacturer pulls quality-control sam- ing-area curves, heat-sinking graphs, and
find out that because either input lead can ples and tests the lot for the parameter applications circuits. An applications en-
dominate VIO, the range of VIO is ⳮ3000 of interest. If the parameter test fails, the gineer who wants to show an IC in its best
␮V울VIO울3000 ␮V (Reference 1). manufacturer doesn’t ship parts from light generates this information, so pay
Parameter specifications are functions that lot. GBD involves parameters phys- special attention to pc-board-layout sug-
of the test circuit, test conditions, and the ically tied to a tested parameter; if the gestions to obtain the best ac results. Most
IC. Most data sheets do not contain test- tested parameter passes, the GBD pa- manufacturers include layout artwork in
circuit diagrams, but they are available rameter passes. When a special customer- this section, or they give designers layout
for most measurements, so designers can manufacturer relationship exists, the artwork in an attempt to maximize their
gain an understanding of test circuits manufacturer can negotiate an agree- IC’s performance in the circuit.
to understand the specifications. The ment that enables the use of typical pa- Circuit descriptions of the IC or of ap-
CMRR (common-mode rejection ratio) rameters. If you can’t obtain manufac- plications of the IC in this section are en-
88 edn | April 14, 2005 www.edn.com
designfeature Semiconductor data sheets

tertaining and informative. The applica- designing circuits that are functional and formation after some study. If you don’t
tions engineers build and test each circuit, that catch the reader’s eye. It is not de- understand anything in the applications
but the fact that they do so is no guaran- signing reliable circuits for high-volume section, don’t hesitate to contact the
tee that the circuits will work when you production, although they often have this manufacturer’s applications engineers for
build them, so you should use the appli- capability, too. Designers get the most de- help, but read the data sheet first; many
cations circuits as starting points in your sign information from the parameter sec- applications questions are answered by
design. Write the design equations, build tion, but the applications section supplies the applications engineer reading the data
the circuit, and test to validate the per- plenty of important information, so thor- sheet to the design engineer.
formance. Remember, the strength of a oughly read it. Often, topics that seem of The applications section includes
semiconductor-applications engineer is little interest at first glance yield good in- package and lead dimensions. Most IC
packages have standard external dimen-
sions, but check to ensure both compat-
ibility and that the package-type materi-
al contains adequate dimensional in-
formation. Always check special packages
for dimensions, unique lead functions,
heat-sink requirements, and other po-
tential problem areas.
All data sheets contain a number of dis-
claimers, and designers should read and
understand them. The disclaimers cover
use in critical applications, the right to
change the product without notification,
intellectual-property rights, and the lim-
it of application-circuit responsibility. But
read the data sheet; this list is incomplete.
Fully understand which disclaimers apply
to the product, because disclaimers vary
among products and manufacturers.
MAKE THEM YOUR ALLIES
Data sheets contain adequate data to
complete most circuit designs. Design
with guaranteed minimum/maximum
data unless this data is unavailable. When
it is unavailable, do not jump to typical
data as a substitute; rather, contact the
manufacturer for GNT specifications,
GBD specifications, and parameter data,
or generate reliable statistical data. Data
sheets contain a wealth of information,
and the time you spend reading them will
pay off in better and faster designs. The
front page may be attractive, but the food
is inside, so have a good meal.왏

Author’s bio graphy


Ron Mancini is a staff scientist for Texas
Instruments and an EDN columnist on
analog topics.

Acknowled gment
Thanks to David K Wilson at TI for his
special insights.

Reference
1. Mancini, Ron, Op Amps For Every-
one, Newnes, 2003.

90 edn | April 14, 2005 www.edn.com