mean the mode (or peak) of the distribution. Conversely,
except for with an ideal normal distribution that is centered
between the limits, the average of the limits (medial value)
might be a typical value but is not necessarily the same as
the peak of the distribution. The real world is rarely ideal
— device specifications are often compromises (for
example, transistor beta and breakdown voltage are
inversely related) and the specification limits might
eliminate one or both tails, resulting in a lop-sided, skewed
distribution as in the following figure. Worse, to the
component manufacturer, “average” can apply to the entire
population of acceptable or shippable parts (including all
binouts). From the data sheet, the buyer will infer that
“average” applies only to those devices shipped against the
data sheet.
Typical values and typical characteristics curves are
only given for circuit or system baseline design informa-
tion and are usually at the nominal operating voltage and at
room temperature. Although indicative of the distribution
peak or median for a large number of production lots
(“N”), typical values should not be expected for any
particular device or single production lot. Even a multi-
tude of fancy typical characteristic curves are useful only
to indicate the general shape of a characteristic against the
specified variable. A maverick lot (or outlier product) is
not a normal lot, but may still be well within acceptable
limits. System performance absolutely must allow for
these maverick lots and any device performing within the
specified limits.
MODE (PEAK)
MEDIAN
AVERAGE
MEDIAL
"N"
LOTS
MAVERICK
LOT
ACCEPTABLE
DEVICES
VALUE
Dwg. FRD-603-1
Typical values (where they are the mode or peak of the
distribution) are the starting point for a complex system
design. Because “typical” is the value around which most
devices should perform, ideal system performance should
be the “typical” when all of the components in the system
are at or near “typical”. As every good circuit designer
knows, system performance will be determined by the
worst-case component performance and the allowable
component specifcation limits must be reviewed against
the desired system performance to achieve a reliable
design.
LIMITS (the most important part of a data sheet) are
those values that are warranteed under the specific test
conditions shown. They are also the “legal” definition for
determining acceptable devices. These are the tests that
the manufacturer should perform in production and the
only qualification or inspection tests that the user may
perform on receipt. A common misconception is where a
test condition is a range (usually operating voltage range
or operating temperature range) implying warranteed
performance at an infinite number of test points. This is of
course an unacceptable test condition for manufacturing.
The manufacturer will probably do these tests only at a
worst-case condition or will use a generally accepted
correlation to predict performance over the specified
range.
It should be emphasized here that the rules for round-
ing numbers do not apply to limits. If the specified
maximum is 5.0, a measured value of 5.001 may not be
rounded down and is a reject! Because of test tolerances at
both the manufacturer and at the user, the manufacturer
will often “guard band” tests. For example, for a specified
maximum of 5.0, a manufacturer’s internal, production test
limit might be 4.8 to allow for production and customer
measurement errors of p2%. Guard bands should prevent
an actual value of 5.09 (reject) from being read as 4.99 (ok
to ship) or an actual value of 4.91 (ok to receive) from
being read as 5.01 (reject).
ABSOLUTE MAXIMUM (allowable) RATINGS are
limiting values of permitted operation and should never be
exceeded under the worst possible conditions initially, or
throughout life. If exceeded, by even the smallest amount,
instantaneous catastrophic failure can occur. Even if the
device continues to operate satisfactorily, its life may be
considerably shortened. Operation at an absolute maxi-
mum rating is permitted (although not desirable – even a
Allegro MicroSystems, LLC
5
115 Northeast Cutoff
26000-AN, Rev. A*
Worcester, Massachusetts 01615-0036 U.S.A.
1.508.853.5000; www.allegromicro.com
short test is believed by some to cause incipient failure)
but operation at two or more limits (i.e., output current and
ambient temperature) almost always means that some
other limit has been exceeded (in this instance, probably
package power dissipation). In certain integrated circuits
that include an internal thermal shutdown, fault conditions
will generate higher than permitted (steady state) tempera-
tures and activate device thermal shutdown circuitry.
These fault conditions can be tolerated for short periods of
time, but will affect life and should be avoided. Except for
a maximum output voltage rating (often done as a leakage
current test), production testing of the absolute maximum
ratings is not usually performed.
“WARRANTEED” BY DESIGN, TESTED TO A LOT
SAMPLE PLAN, or NOT TESTED IN PRODUCTION are terms
usually applied only to the difficult-to-measure character-
istics. Common examples are temperature dependencies
and some dynamic (ac) tests. Warranteed by design is a
“real” specification. Even though only a few devices
might be tested, all devices are warranteed to the specifica-
tion. A common practice now is for the manufacturer to
rescreen the production lot or perform 100% test of a lot if
even one device fails the sample testing. However,
compare this against Unspecified Specifications, below.
RECOMMENDED OPERATING CONDITIONS may be
given for optimum device performance but, other than
functionality, does not promise any specific performance
limits. Operation outside of these values is permitted (but
within the absolute maximum ratings) often without any
implied level of performance.
A single specification has dual purposes. From one
viewpoint, it specifies the limits of acceptable device
performance. From the other direction, it defines the
worst-case operating conditions for acceptable system
performance. See When Is A Minimum A Maximum?
below. Designing an application based on sample perfor-
mance or data sheet typical data can result in system
failures (parametric or catastrophic) when production
devices are used that do not meet the initial data spread but
are still within the specified limits. The only solutions
then are either an expensive system redesign or a more
expensive custom specification (binout) for the critical
parameter required.
One of the advantages of English is that there are a
multitude of synonyms for most words. Unfortunately, the
thought-to-be subtle differences can make a major differ-
ence to the reader. Exact synonyms are rare. Of special
importance are the differences in meanings between the
following (non-synonymous) verbs:
Verb
shall
must
should
may
can
will
Meaning
a requirement
essential to successful achievement
a recommendation
permissive
able or possible to
a promise or expectation
Unspecified Specifications
Unfortunately, some data sheets give (as opposed to
specify) an operating temperature range and then specify
all of the parameters at room temperature only. What is
the user getting???? Only functionality over the tempera-
ture range — without any implied warrantee of level of
performance, which may be very marginal! For example,
actual diode leakage current or transistor saturation voltage
at high temperatures may be many times greater than the
25 oC limit — without any limit even implied. Con-
versely, for semiconductors at low temperatures, the user
might only get survivability (very different from function-
ality). If the application requires wide-temperature
operation, forget this kind of data sheet but work very
closely with the manufacturer so that you get the device
needed at an affordable price.
Just because “second source” products meet the same
set of published limits does not make them identical. This
is especially true of ac parasitics that can become very
important at the application frequency but are inconse-
quential at normal test frequencies. The three npn transis-
tor chips shown below are all marketed as 2N2222s and
should meet the specified dc characteristics. They will
even meet the specified minimum fT and maximum Cob,
but it should be obvious from the photographs that they
will not be identical in all applications. Even from a
Allegro MicroSystems, LLC
6
115 Northeast Cutoff
26000-AN, Rev. A*
Worcester, Massachusetts 01615-0036 U.S.A.
1.508.853.5000; www.allegromicro.com
single manufacturer, what will happen to the system
performance when an unspecified parasitic drifts because
of manufacturing process variations?
When Is A Minimum A Maximum?
The content and format of a data sheet is sometimes
confusing. From the manufacturer’s point of view, a
breakdown voltage is usually specified as a minimum
value, indicating that all acceptable devices exceed this
minimum value. On the other hand, data sheets are also
used by system designers, who must be sure that all parts
of the finished system work together. They interpret this
minimum breakdown as the absolute maximum that may be
applied to the device. Very few designers are confused by
this, especially because most (none that I’ve seen) manu-
facturers do not specify typical breakdowns for fear (with
justification) that the user will try to design around the
higher number.
But what about — minimum set-up time or input pulse
width is 30 ns min, 25 ns typ? This comes from the
unfortunate mixture of specifying system requirements
with device parameters. This specification must be
interpreted as “Devices will function with a set-up time or
input pulse width of at least 30 ns.” Typically, this device
will function with a set-up time or input pulse width of
only 25 ns, but it’s not warranteed so it’s not a meaningful
number. A similar situation exists with logic input levels
where VIH = 2 V min and VIL = 0.8 V max. As described
here, these are requirements rather than parameters.
Related to min/max is something called “algebraic” vs
“absolute magnitude” conventions. Where zero has no
special significance (just another number), the algebraic
convention should always be used, i.e., -40 oC is lower
than +25 oC. Where zero is the absence of something, the
absolute magnitude convention is normally used in the
U.S., i.e., -40 V is greater than +10 V, and zero is the least
possible value; if a range of values includes both positive
and negative values, both limiting values are maximums
(the minimum is implicitly zero). Normally we switch
back and forth with no problems. However, some data
sheets (more often European) can cause confusion when
the algebraic convention is not clearly stated and the
specification limits (for a latching Hall-effect magnetic-
sensor IC) are:
Operate Point = 15 mT Max
Release Point = -15 mT Min
Here, using the absolute magnitude convention, a device
release point of -20 mT is acceptable but using the alge-
braic convention it is not. Only if a typical value is also
given is this specification clear. To reduce (that doesn’t
mean eliminate) the possibility of confusion, where
algebraic convention is being used, it should be noted (se
the following Caveats).
Caveats
NOTES, CAUTIONS, WARNINGS, etc. are bits of text
generally set off from the main body. Their meanings an
levels of importance are significantly different.
A NOTE (but see also NOTICE, below) is additional
information or clarification that may be of interest to the
reader. It should not include requirements or specifica-
tions.
The word ATTENTION is used to indicate important
information that, if not observed, could result in damage
a component or system as in “observe precautions for
handling sensitive devices”.
CAUTION messages are used to indicate procedures
that, if not observed, could result in loss of data or damag
to equipment.
WARNING messages are much more important;
often in larger, heavier type, are boxed, and/or with color
(a yellow background is preferred). WARNING messages
are used to indicate potentially hazardous procedures tha
if not followed correctly, could result in personal injury.
Although generally not found in component data
sheets except for the most extreme situations, DANGER
indicates an imminently hazardous situation that, if not
avoided, could result in death or serious injury. DANGE
messages would also use larger, heavier type, and/or be
boxed, but here (if color is used) a red background is
usually used.
This special exclamation mark may
be used with Attention, Caution,
Warning, or Danger messages. It
serves to emphasize the importance
of the message.
This lightning bolt symbol is often
used with a Warning or Danger
message to alert the reader to a
special risk of electric shock.
Allegro MicroSystems, LLC
7
115 Northeast Cutoff
26000-AN, Rev. A*
Worcester, Massachusetts 01615-0036 U.S.A.
1.508.853.5000; www.allegromicro.com
A NOTICE message should be used only to state a
company policy (for example, “… are not intended for use
in life-support applications without written consent.”) or a
legal requirement.
Standards
Standards are most often thought of as physical
measurements (kilogram, meter, second, etc.) but are also
procedures and “things” established by general consent as
models or examples. Symbols, terms, and their definitions
are also standardized to facilitate communication. Engi-
neers, especially, like standards that provide for accuracy
and repeatability.
Unfortunately, industry has not yet standardized data
sheet contents, let alone data sheet formats. If industry
did, it would be easy to compare similar devices from
different vendors or even from a single vendor. But
because test conditions reflect the major customer’s
operating conditions or the manufacturer’s ideal operating
point, similar devices do not necessarily use similar test
conditions and often give very different specification
limits. Sometimes the same device can have very different
specifications for very different applications and will even
have two different part numbers. This is best exemplified
by audio power amplifier ICs that have later been specified
as power operational amplifiers.
The closest we are to standards is in the terms and
definitions for the various tests performed (or not per-
formed). A French technical writer can write “horlage”
and not communicate anything to the English reader.
However, the internationally agreed-upon “standardized”
abbreviation CK is usually understood (horlage is the
French word for clock). Whether a product data sheet is
published in English, French, Cyrillic (Russian), or Kan-Ji
(Chinese or Japanese), the technical symbols are always
the same. The example shown below is nonsensical but
the symbols ICEX, VCE(sat), and VCE(sus) can make the
(here meaningless) text less important.
*ETPLVXE ,6FHKI 9SA 3WE ZDW 'PU
&FDHITEW ICEX
— 10 50 MA
1XGJPUTS VCE(sat) — 0.5 0.7
V
B*EWASQI VCE(sus) 40 50 —
V
International standards define the prefix k as the
multiplier 103, 1 K as 1 kelvin (-273.15 oC), and 1 kB as
1000 bels [unlikely, but 10 000 dB] Unfortunately,
industry uses 1 kb/s as 1000 bits per second and 1 kB/s as
1000 bytes per second. Further, in the specialized field of
memory storage capacity, 1 Kb means 1024 bits and 1 KB
means 1024 bytes, where “K” (not to be confused with
“k”) means 210. A “1 Kb memory” is often read as a “one
kilobit memory” and usually understood to mean 1024 bits
of storage capacity. Compounding the confusion, when
used to describe storage capacities, “M” (mega) means 220
and “G” (giga) means 230.
Terms, definitions (in English), and letter symbols for
semiconductor devices are given in EIA/JEDEC Standard
77 (for discrete semiconductor devices), EIA/JEDEC
Standard 99 (for ICs), and EIA/JEDEC Standard 100 (for
microcomputers, microprocessors, and memories). When
used correctly, the meaning of a symbol is precisely
defined … ic, iC, Ic, IC, IC(AV), Icm, ICM, IC(PP),
IC(RMS), etc.
Language tends to change with time as new words are
generated or slang creeps in. Some would say that this
enriches the language. While new technical abbreviations
or acronyms are always being generated (by marketing or
ad agencies) their replacement of symbols can cause
considerable confusion. For example, the abbreviation BV
for breakdown voltage should not extend to BVCEO for
collector-emitter breakdown voltage as this symbol can/
should be read as the quantity of flux density and collec-
tor-emitter voltage (V(BR)CEO is correct). Similarly, if R
is the symbol for resistance, thermal resistance is symbol-
ized as RQ or Rth, not Q.
Graphic symbols for logic functions and schematics
are defined in IEEE Standards 91 and 315, respectively.
Where IEC standards are used, letter symbols for electrical
technology are given in IEC 27 and for semiconductor
devices and integrated circuits in IEC 148. Terms, defini-
tions, and letter symbols for semiconductor devices are
given in IEC 747 (for discrete devices), IEC 748 (for ICs),
and IEC 824 (for microprocessors).
Although certainly not “standards”, there are some
excellent aids for the data sheet writer. These include the
standards and publications style manuals of the Electronic
Industries Association (EIA Engineering Publication EP-7)
and the IEEE. These two style manuals generally agree
Allegro MicroSystems, LLC
8
115 Northeast Cutoff
26000-AN, Rev. A*
Worcester, Massachusetts 01615-0036 U.S.A.
1.508.853.5000; www.allegromicro.com
