Voltage Sag Immunity Tests: Alex McEachern

Voltage Sag Immunity Tests:
Some Common Mistakes and How To Avoid Them
Alex McEachern, Senior Member, IEEE
[email protected]
Power Standards Lab, Emeryville, California, USA
TEL ++1-510-658-9600
Two standards, SEMI F47 and IEC 61000-4-11, require
voltage sag immunity tests for electronic equipment.
Voltage sags (or dips) of specified depths and durations
are intentionally applied to the equipment, and the
equipment’s performance is verified. Several common
testing mistakes have been identified. These mistakes can
yield false positive results, incorrectly indicating that the
equipment has passed, or they can give false negative
results, incorrectly failing the equipment. The mistakes
include having insufficient available current; trying to
simulate phase-to-phase sags with phase-to-neutral
equipment sag generators; and misunderstanding the
requirements of the standards.
FAX ++1-510-658-9688
Properly designed transformer-based sag generators can
almost always provide sufficient pulse current. However,
hand-operated or motor-driven variable transformers
should be avoided, as the contact point limits available
pulse current. Oversized transformers with multiple fixed
taps are optimal.
The switches in a transformer-based sag generator must
be carefully understood. IGBT-type switches generally
include current limiting in their drive circuit, which is
good because it protects the IGBT, but bad because it
limits the pulse current available to the load. For
example, in a sag generator a 300-amp IGBT would be
appropriate for loads up to 50 amps continuous. A
smaller IGBT will give false positive results.
Keywords: sag, dip, immunity, testing, SEMI F47, IEC
61000-4-11, mains, power line
The sag generator must be capable of providing at least 6
times the nominal current or volt-amp rating of the load
for at least one cycle. For example, to test a 16 amp load,
at least 96 amps must be available for one cycle from the
sag generator; if the load is a 10kVA load, at least
60kVA must be available from the sag generator.
Insufficient available current gives false positive results:
equipment that fails with real-world sags will incorrectly
pass with sag-generator sags.
For this reason it is generally not possible to correctly do
sag immunity testing with amplifier-based sag generators.
(An exception: if the load requires less than 5 amps, and
a large amplifier-based sag generator capable of
delivering 50 amps or more is available, testing can be
done correctly.)
Fig. 1 – Voltage sag (upper graph) vs load current (lower graph). The
load current typically increases by 500% or more, for 10-20
milliseconds, at the conclusion of the sag. This can blow fuses or cause
other EUT problems. If the sag generator lacks sufficient current
capacity, this current pulse will be limited, and there will be a false
pass. For this reason, electronic sources should generally not be used
for voltage sag testing.
TRIAC-type switches have appropriately large pulse
current ratings, but limit the switching angles to 0 degrees
and 180 degrees. A combination of TRIAC-selected or
electro-mechanically selected taps and IGBT-gated sags is
Page 1 of 4
Another source of false positive results comes from
component variations in fuses and/or circuit breakers. If
the tested sample has a barely adequate fuse to deal with
current pulses at the end of a voltage sag, the fuse may
not blow on the tested sample. This problem can be
addressed by rapidly repeating test sags, and by testing
with sags that are somewhat deeper and longer than the
standard test requirements.
Fig. 3 – Component variations in EUT’s
can easily be overlooked. This typical
electrolytic capacitor has a value
tolerance of –10%+50%,. One must be
careful about drawing conclusions from a
single tested sample – its capacitors may
have been at the high end of the tolerance
Fig. 2 – Typical fuse current-time curves for electronic equipment. A
typical sag-induced 500% increase in current for 10 milliseconds can
create significant problems – but only if the sag generator is capable of
supplying sufficient current.
Not all standards are intuitively clear regarding the test
For three-phase loads, it is typical for current to increase
by 150% or more on the non-sagged phases during a
voltage sag. As the test sag may last 10 seconds or more,
is is essential that the sag generator be capable of
providing at least double the nominal current or volt-amp
rating of the load for an extended period of time.
Typically, the device under test stores energy during
normal operation, then releases that energy as
compensation during voltage sags. Usually this energy is
stored in electrolytic capacitors.
Large value electrolytic capacitors have a wide tolerances.
If the sample being tested happens to have a capacitor at
the high end of the tolerance range, the sample may well
pass the voltage sag test, while production units with
more typical capacitor values may fail, yielding a false
positive result.
Fig. 4 – The upper graph shows the required voltages and durations for
sag testing in SEMI F47. The lower graph shows a well-meaning, but
wrong, test waveform. The difference hinges on the subtle meaning of
“duration” in the x-axis label of the standard graph – which makes it
difficult for non-English readers to interpret correctly.
Page 2 of 4
waveforms, so it is possible for a well-meaning engineer
to choose the wrong waveforms. Consult an expert in the
standard of there are any doubts.
Some sag generators are incapable of generating true
phase-to-phase sags, so an attempt is made to simulate a
phase-to-phase sag by generating two simultanous phaseto-neutral sags. This works, but only partially.
In a true phase-to-phase sag, the phase-to-neutral voltages
are determined by the load and source characteristics, and
may be substantially unbalanced (for example, if there are
large phase-to-neutral loads on one phase but not the
other). Typically, simulated phase-to-phase sags are
perfectly balanced.
Simulating phase-to-phase sags with a pair of phase-toneutral sags can give false positive results. Use a sag
generator that can create true phase-to-phase sags.
Most of the problems in this area tend to be softwarerelated. For example, a compliant component my
momentarily give an incorrect output during a voltage
sag, but recover immediately at the end of the sag. The
system software may interpret this momentary signal as
an important error, and decide the shut the system down.
System software should be designed to not respond
instantly to error signals; instead, some intelligence
should be applied to the required duration of the error.
For example, a single incorrect temperature reading can
often be ignored, if the preceding and following readings
are within tolerance. (Safety-related signals, of course,
should always get an immediate response.)
Voltage sag immunity testing is a useful, practical
engineering tool. It produces products that are stronger
and more reliable. Like any engineering discipline,
experience leads to the discovery of pitfalls and incorrect
shortcuts; this paper has identified a few of the most
common ones.
Sag immunity standards often have pass/fail criteria that
are open to widely different interpretations. This is
especially true when the same standard is applied to
components, subsystems, and complete systems.
A typical example: an appropriate pass/fail criteria for a
complete system might be that the system is permitted to
mis-operate during a voltage sag, but must recover
without operator intervention. However, if the same
criteria is applied to a system component, the system must
be prepared to tolerate mis-operations by its components
– something that is rarely possible.
In general, components need to meet tougher criteria than
subsystems, and subsystems need to meet tougher criteria
than complete systems, which have the easiest criteria.
If this hiearchy of criteria is not possible to achieve,
information about how, and under what conditions, the
subsystems and components misoperate is critical.
The author gratefully acknowledges the useful advice,
shared testing experiences, and interesting suggestions –
not all taken, which undoubtedly will explain the
remaining errors in this paper – from his friends and
colleagues: Josef FOLDI and Uwe HALLER and Hassan
IRAVANI (all of Applied Materials), Wilmer AWAYAN
(Novellus), Mark FRANKFURTH and Byron YAKIMOW
(both of Cymer), Bob DETTORRE (Comdel), Danny
POLIDI (formerly of Nanometrics), Cliff GREENBURG
(Nikon), Ron WAGNER (CPI), Norm NICOLAI (IGC
Polycold), Robert MAXWELL and Jeff BRUNER (both of
KLA-Tencor), Dave GRAHAM (NUMMI), Greg
WILTERDINK (Schlumberger), Dan WOODALL (formerly
of SCP Global), P.K. FENG (Brookhaven National Lab),
Jor AMSTER (ADTEC), all the engineers at ENI Power, all
the engineers at Advanced Energy (who are very good at
catching my errors!), Kamran HAQUE (Verteq), the
engineers at Siemens, and everyone else who has put up
with all my endless questions and poking around.
Thank you all. Remaining errors, of course, are solely my
own responsibility.
It is perfectly possible to build a system that does not
meet a sag immunity standard (for example, SEMI F47)
from components that do meet that standard.
Page 3 of 4
[1] SEMI F47-0200, Specification for Semiconductor
Processing Equipment, Voltage Sag Immunity.
SEMI, Santa Clara, California, USA. 02-2000
[2] SEMI F42-0600, Test Method for Semiconductor
Processing Equipment, Voltage Sag Immunity.
SEMI, Santa Clara, California, USA. 06-2000
[3] G.T. Heydt, Electric Power Quality, Stars in a Circle
Publications, Lafayette, Indiana, 1991
Alex McEachern (M 1984, SM 1996) is
the President of Power Standards Lab in
Over the last 20 years he has taught
graduate-level power quality courses
and/or has supervised the installation
of electric equipment in the United
States, Canada, Croatia, Japan, Hong
Kong, China, South Africa, Germany,
France, Singapore, Switzerland, England, Scotland,
Mexico, New Zealand, and Australia.
[4] IEC 61000-4-11 Ed. 2.0, Testing and Measuring
Techniques – voltage dips and short interruptions
immunity tests. IEC Document 77A/336/CD
[5] IEEE Standard 1100-1992, IEEE Recommended
Practices for Powering and Grounding Sensitive
Electronic Equipment
[6] A. McEachern, Handbook of Power Signatures,
Dranetz-BMI, Edison NJ 1997
Page 4 of 4