The importance of providing power with steady voltage and frequency has been
recognized since the inception of the Electric Supply Industry (ESI). As ESI
proliferates along with the introduction of power system deregulation, electronic
and information technology equipment in a wide range of systems and unpredictable
climate changes, it is almost impossible to avoid power system disturbances.
Over the past few years, voltage sags have emerged as the most serious power
quality problem, especially in the context of the Malaysian distribution systems.
Voltage sag is a sudden decrease in voltage amplitude followed by a return to
its initial level after a short time (Barros and Diego, 2002).
Sags are caused by network faults, switching actions such as large motor starting,
transformer energizing, capacitor bank connection etc. These incidents that
initiate voltage sags can disrupt or damage sensitive equipments such as personal
Personal computers first appeared in the late 1970s. The principal characteristics
of PCs are that they are single-user systems and based on microprocessor technology.
However, although PCs are designed as single-user systems, it is common to link
them together to form a network. Through networking and interfacing it is used
widely for on-line communication, continuous process control applications etc.
The malfunction of PCs incorporated in on-line or even off-line systems can
cost substantially, because there are losses associated to the computer controlled
systems and process (Djokic et al., 2005).
Sensitivity of PCs to voltage sags can be defined as the conditions where the
PCs start to malfunction and cause nuisance, loss of data or process interruption
due to voltage sag appearing in the mains supply where it is connected. Studies
for assessing sensitivity of voltage sags on customer loads are gradually increasing.
These efforts are primarily divided into practical and theoretical approaches.
The practical approaches investigate the effects of voltage sag by monitoring,
conducting experiments on customers sensitive loads and performing pertinent
surveys (Bollen, 2000).
Current standards related to the testing of the equipment sensitivity to voltage
sags and short interruptions suggest that the tests should be performed preferably
at 0° point on wave of the voltage waveform (Djokic et
al., 2005). Testing of the equipment for additional angles is necessary
only if they are considered critical by product committees or individual product
specifications. If so, a range from 0 to 360° in steps of 45° is optional
for such additional testing (Bok et al., 2008;
Djokic et al., 2005). Typically, sags include
80% remaining voltage for 1 sec; 70% remaining voltage for 0.5 sec and 40% (or
50%) remaining voltage of 0.2 sec.
Equipment sensitivity to voltage sag can also be considered and presented in
the form of power acceptability curves which are also known as voltage vulnerability
or sensitivity curves.
|| ITIC (CBEMA) curve
One of these curves is the Computer Business Equipment
Manufacturers Association (CBEMA) curve (Kyei et al.,
2002). Since, CBEMA is now defunct, the Information Technology Industry
Council (ITIC) has taken its place with major modification to its original curves
as depicted in Fig. 1. Even with the new look, an ITIC (CBEMA)-type
criteria has some important limitations. It is not in itself sufficient criteria
for typical office systems (Institute of Electrical and Electronics
Engineers Inc., 2005).
These curves are plots of bus voltage deviation versus time duration which
separate the bus voltage deviation - time duration plane into two regions namely,
acceptable and unacceptable regions. The lower limb of the power acceptability
curve relates to voltage sags and momentary outages. The latest power acceptability
standard, as shown in Fig. 2, is known as the SEMI F47 issued
by the Semiconductor Equipment and Materials International (SEMI) in the year
2000 to satisfy minimum voltage sags ride through capability for equipments
used in the semiconductor industry (Djokic et al.,
2005). The specification simply states that semiconductor processing, metrology
and automated test equipment must be designed and built to confirm to the voltage
sag ride-through capability as per the defined curve. Equipment must continue
to operate without interruption during conditions identified in the area above
the defined acceptable region (Institute of Electrical and
Electronics Engineers Inc., 2005).
||SEMI F47 standard curve
The design of the above mentioned power acceptability curves clearly relates
to whether the distributed power can be utilized or not. In other words, the
distribution power should be considered acceptable if the industrial process
served is operative. Thus, the ultimate criterion of power acceptability relates
to the operating status of the industrial process and equipment. This criterion
depends on the nature of the load. For example, simple incandescent lighting
loads may have a very loose criterion for acceptability, while certain sensitive
computer controls may have a much more restrictive criterion. The difficulty
in the selection of a single suitable criterion is confounded by the many possible
load types (Kyei et al., 2002). For this reason,
Kyei et al. (2002) introduced the concept of the
so called standard for different types of loads. For example, a voltage standard
is a criterion for power acceptability based on a minimum acceptable DC voltage
at the output of a rectifier below which proper operation of the load is disrupted
(Lee et al., 2004). However, it is always better
to conduct a characterization test to develop the voltage tolerance curves for
different sensitive equipments to voltage sags such as PCs.
As an effort to understand the immunity level of PCs, many studies have been
reported in the past. Test results on standard restart/reboot malfunction criterion
for computers due to voltage sags can be found by Saksena
et al. (2005). It is reported that if the depth of voltage sag is
larger than 30% and lasts more than 8 cycles, the voltage sag may cause a computer
to restart. These tests are only carried out for the 120 V/60 Hz systems. Furthermore,
it is concluded that the performance of the switching power unit and the power
consumption of a computer plays a vital role on the sag effect. Other than the
standard restart malfunction criteria, the effects of voltage sags in other
software criterions are not tested. A similar experimental test was conducted
by Bok et al. (2008) to identify the effect of
rectangular and non-rectangular voltage sags on the same restart/reboot malfunction
criteria. Rectangular sags with loading condition are found to influence most
on the susceptibility of PCs for the tested criteria.
Another comprehensive study on the behavior of PCs during voltage sags and
short interruptions is presented by Djokic et al.
(2005). Laboratory experiments are performed with rectangular voltage sags
as well as with non-rectangular sags to simulate the starting of the large motors.
Supply from the non-ideal voltage source is realized by incorporating harmonic
and supply frequency variations. It was found that all the voltage tolerance
curves for different computers have the same rectangular shape with two clearly
distinctive vertical and horizontal parts, with a very sharp knee between them.
Finally, for three different malfunction criteria, namely, read/write operations,
blockage of the operating system and restart/reboot malfunction criteria, three
different sensitivity curves for each tested PC is plotted. Malfunction of monitor
image distortion and buzzing sound criterion is not investigated.
Baran et al. (1998), considers several malfunction criteria in testing
namely, lockup of the PC, slowdown of the network traffic, file corruption,
etc. to investigate the effects of various power quality disturbances on PCs
connected to a local-area network. However, it is assessed only with respect
to short interruptions, but not with respect to voltage sags. The results are
compared with the CBEMA curve and reported that in most of the cases it violates
the immunity criteria of the standard curve.
Seven PCs of different age have been investigated for voltage sags by Pohjanheimo
and Lehtonen (2002). The malfunction criterion for the PCs selected is automatic
reboot. The voltage-tolerance curves obtained from the tests are found to be
rectangular, having the flat vertical and horizontal part with a sharp knee
between them. Researchers reported that the PCs tolerate the under voltage level
up to 50-60% of remaining voltage for 100 m sec-1. There is no clear
correlation between the device age and sensitivity observed.
The onsite study carried out based on a voltage disturbance profile by Muhamad
et al. (2007) in two local industrial areas for a period of one year
is shown in Table 1. From the Table 1 it
can be seen that the sag events having less than 60% nominal voltage is very
||Sag event impact summary Senawang Tunku Jaafar and Ampang
Harris industrial main substations
This information may be useful in understanding the number of voltage
sag events appearing in the local distribution system.
This study focuses on investigating the vulnerability of personal computers
to voltage sags in the local mains supply. So, far it has not been documented
in the literature as to how PCs used in Malaysia behave in case of under
voltage disturbances. Extensive laboratory tests are conducted for this
purpose. These tests are carried out not only for standard restart/reboot
malfunction criteria but also to examine the distortions on PC monitor
and unacceptable audible noise generation. In order to evaluate the voltage
tolerance levels of the PCs for the identified failure conditions, the
test results are also compared with the design goals of ITIC and SEMI
F47 standards. Finally, based on the performance of tested samples and
superimposing individual immunity curves, a general voltage tolerance
curve for PCs is developed.
MATERIALS AND METHODS
Experimental setup: The methodology that is used in the testing is generally
based on the guideline followed by Djokic et al. (2005)
and Saksena et al. (2005). Five PCs with different
specifications are tested to study the effect of voltage sags on the performance
of the computers. The specifications of the tested PCs are shown in Table
The experimental set up consists of four components namely, sag generator,
Equipment Under Test (EUT), data acquisition system and a computer to
analyze the signals. In this case, an Industrial Power Corruptor (IPC)
from the Power Standards Lab is used. The IPC is a voltage sag generator
combined with built-in data acquisition system which is capable of producing
and interrupting voltages up to 480 V and current at 200 A in single or
three phase systems. Figure 3 shows the real test environment
where single phase local power supply at 240 V, 50 Hz is utilized.
Testing procedure: A series of test results on PCs is obtained
by following the pre-defined procedure given below. The procedure is repeated
for at least three times to avoid probable errors that may occur during
|| Specifications of tested PCs
||Actual PC test environment
||Using the terminal blocks available at the back of IPC, the conductors
from mains panel and conductors to the PC under test is connected
and the IPC is powered on
||The PC with all input/output (I/O) and pointing devices connected
is switched on and allowed to boot and load the operating system
||Staring from nominal voltage, voltage sags are initiated in steps
of 2.5% down to zero volts. The sag initiation angle and the duration
are kept constant. The initial sag duration and phase angle are set
to 1 cycle and 0°, respectively. The critical sag depth for each
of the pre-defined malfunction criteria is determined by repeated
testing for at least 3 times for a particular sag magnitude and duration.
If any malfunction condition is observed, a quick inspection for proper
operation of EUT is conducted before initiating the next sag. For
each triggered sag event, voltage and current waveforms supplying
the EUT are recorded and an observation such as visible or audible
influence on the PC is noted
||The duration of sag is adjusted in steps of 1 cycle and measurements
outlined in Step 3 are repeated
A flowchart of the aforementioned procedure is shown in Fig.
|| Testing procedure
RESULTS AND DISCUSSION
Based on the findings, a generic voltage tolerance curve for PCs is then
constructed and compared with the ITIC and SEMI F47 standards.
Analysis of individual PCs: The test findings from the experiments
on sensitivity of PCs to voltage sags are presented as typical power acceptability
curves. The upper region of these curves represents proper operation region
while the lower regions indicates unacceptable voltage conditions for
Effect of voltage sag on the first tested PC (PC1) is shown in Fig.
5. It can be seen that an unwanted buzzing sound started for a voltage
reduction which last for 9 cycles. It is followed by image distortion
at 22.5% of remaining voltage with duration of 11 cycles. Moreover, for
sag duration between 9 to 13 cycles black screen conditions are observed
for deeper sags. For sag duration beyond 14 cycles, PC1 starts to malfunction
completely or is unable to perform the given task. At this point, it is
automatically rebooted as shown in Fig. 5.
Figure 6 shows the effect of voltage sags on PC2. Similar
to the case of PC1, personal computer PC2 also experienced image distortion,
black screen and reboot malfunction conditions due to voltage sag disturbances.
|| Voltage tolerance curve for PC1
|| Voltage tolerance curve for PC2
As can be noted from Fig. 6, for short duration sags,
the first improper operation observed is the monitor image distortion.
This condition occurred initially for a sag with 80% depth and spanning
for 8 cycles. Besides, for short duration events the image distortion
is not the only limiting effect of voltage sag in this case, but deep
sags seem to automatically restart the computer. The overall immunity
level to voltage sags of PC2 is higher compared to that of PC1.
PC3 has the latest specification of processor and random access memory
among all the tested PCs. The influence of voltage sags to PC3 is different
compared with other PCs where the monitor malfunction criterion is usually
first observed. In this case, the monitor and the CPU of the PC3 happened
to restart simultaneously for any level of unacceptable voltage disturbance.
This is found to occur for sag magnitude beginning from 60% and for all
durations greater than 12 cycles as shown in Fig. 7.
PC4 is found to be the most sensitive personal computer selected for
||Voltage tolerance curve for PC3
||Voltage tolerance curve for PC4
The monitor malfunction criterion is initiated for voltage
sags as short as 3 cycles. For long duration sags, starting from 10 cycles,
PC4 failed to perform its operations if the encountered sag depth is above
50%. At this point it is automatically restarted. Therefore, PC4 is also
the most sensitive PC for long duration sags. The performance of PC4 is
given in Fig. 8.
Figure 9 shows voltage sensitivity curve of PC5. It
shows a very similar pattern to that of PC4. The main difference in this
case is that it is a little more sensitive to short sags in the range
between 7 to 11 cycles. During this period, bussing sound and image distortion
is observed if the remaining voltage goes below 27.5%.
Development of a generic voltage immunity curve for PCs: As shown in
above, each personal computer potentially has its own standard of power acceptability.
An approach to define the overall acceptability region is the application of
intersection to the individual voltage tolerance curves (Kyei
et al., 2002) as shown in Fig. 10.
||Voltage tolerance curve for PC5
||Power acceptability of all PCs
||Result of intersection of all individual voltage tolerance
The upper acceptable region is the region that all PC loads properly
operate, the lower region indicates that all PCs fail and the intermediate
region corresponds to some PC failures and some ride-throughs. The concept
of intersection by overlaying many acceptable curves is shown in Fig.
Finally a generic power acceptability curve for the PCs can be constructed
by taking the upper tolerance region of the intersected curves as shown
in Fig. 12.
||Generic voltage immunity curve for PCs
From the constructed immunity curve for PCs depicted in Fig.
12, it can be noted that all PCs can tolerate short transient interruption
which is less than 3 cycles. However, some sensitive PCs such as PC4 may
start to fail if it is a little greater than 3 cycles caused by severe
sag with depth greater than 80%.
If one takes the first section of the well known SEMI F47 standard which
represents the immunity level of equipment for sag duration between 2.5
to 10 cycles and compare the developed power acceptability curve as shown
in Fig. 12 for PCs, it can be observed that the latter
curve has 3 distinctive steps in this period. The first step appears for
the duration between 3 to 4 cycles and to magnitude of 80% nominal voltage.
The second step represents immunity level of PCs to sag depths of 22.5%
remaining voltage. This period lasts between 3 to 7 cycles of time axis
in the developed curve. The final step that goes until 10 cycles corresponds
to 27.5% of remaining voltage.
When the magnitude of these 3 steps of the proposed immunity curve are
compared to ITIC and SEMI F47 standards sag depths, 70 and 50%, respectively,
the obtained curve indicate much lower values for the duration between
2.5 to 10 cycles.
The remaining part of the developed immunity curve has two more steps
unlike the SEMI F47 standard voltage tolerance curve. The next reduction
of tolerance level occurs at 11th cycle with a large transition from 27.5
to 47.5% of the remaining voltage. Then a small increment in susceptibility
level of PC loads is observed at the 17th cycle again. As it can be seen
from Fig. 12, for longer duration sag which is greater
than 17 cycles, some sensitive PCs may still fail to operate properly
if the voltage drops below 50% of nominal voltage. One final observation
that can be obtained from Fig. 12 is that all PCs can
tolerate a sag depth less than 50% voltage indefinitely according to the
test results curve. However, the SEMI F47 accepts 50% reduction in voltage
magnitude only for less than 10 cycles.
Finally from the test results and developed immunity curve for PCs it
is possible to conclude that the SEMI F47 standard is not only designed
for PCs working in single phase power supply of 230 V. It is noted that
all the tested PCs satisfy the design goals of SEMI F47 and ITIC standard
and exhibits a large margin between the actual malfunction conditions
that may be observed due to any voltage sag arising on the supply system.
Furthermore, by comparing the number of sag events that fall below 60%
nominal voltage given in Table 1, it can be said that
the sensitivity of PCs to voltage sags in local mains supply is low.
An experimental study has been performed to determine the effect of voltage
sag on personal computers. From the results of the experimental study, voltage
tolerance curves of PCs are constructed to describe the sensitivity of various
PCs to voltage sags. It may be concluded that the voltage tolerance of the PCs
used in the test varies over a wide range. All the immunity curves obtained
appear to have similar shape with distinctive vertical and horizontal steps.
Monitor image distortion, buzzing sound and black screen condition which is
considered unacceptable in this study seem to emerge only for short duration
sags lasting less than 18 cycles. Long duration sags mainly lead to a computer
reboot. Although the computers tested covered a wide range of model, type and
hardware configurations, there is no correlation between processor speed and
operating system installed.
When the voltage immunity levels of the tested PCs are compared with
the ITIC and SEMI F47 standards, all the tested equipment satisfy their
design goals. Based on the experimental results it is possible to construct
a generic voltage tolerance curve for PCs which can clearly show acceptable
and unacceptable regions for different voltage sag disturbances. The curve
provides a quick overview about the immunity level of personal computers
in a particular power distribution network. In addition of having the
knowledge about the PC sensitivities to voltage sags, there are few methods
that can be investigated to immunize the PCs from voltage sags. These
include the use of dc link capacitors at PCs power supply unit and the
use of power factor correction circuitry. The relationship between voltage
sag and DC output variation of PCs switch mode power supply is currently
under investigation and the test results will be presented in the near