Parallel Cycles
Back to top
The hardware configuration detailed above is capable of completing the specified measurements within the allotted time frame. Each Model 2750 can be
configured to initiate a scan when a trigger is received; therefore, four parallel 200-channel measurement cycles can be initiated, rather than one
800-channel serial measurement cycle.
Synchronization
Back to top
A trigger link, or hardware handshake bus, is used by the instruments in the test system to ensure proper sequencing (it's a standard feature on all
the Keithley instruments in this example). When the Model 2750 instruments are configured using the trigger link interface, one instrument will initiate
a scan and the other instruments will be triggered.
This built-in bus eliminates the need for direct PC control of most system functions. When the trigger link function is used properly, the only
functions the PC need perform are initiating the test and retrieving data from the system, leaving synchronization to the instruments themselves.
Sources of Error
Back to top
There are a number of sources of error that should be considered. One that we've already mentioned is environmental noise. Placing hundreds of
switching power supplies in an environmental chamber creates one of the worst conditions for making accurate voltage measurements. Switching power
supplies radiate high-frequency noise, and if the ground connection is noisy, conventional data-acquisition systems will be unable to make
satisfactory measurements.
An instrument such as the Keithley Model 2750, with its high-impedance input, high common-mode rejection ratio and high normal-mode rejection
ratio (CMRR and NMRR), and 22-bit A/D converter, can have problems. The DMM input is scanned across multiple channels very rapidly, so it can be
difficult to distinguish between 5 V and zero volts on adjacent channels.
Noise will not only degrade the overall quality of the collected data, but it can also result in acceptable power supplies being categorized as
failures. Obviously, this will result in lower throughput, and higher costs. Therefore, it's important to understand the effects of noise and how
to compensate for them.
Figure 5(a) shows a typical signal in the presence of noise, with measurement sampling that results in a small error. Figure 5(b) illustrates
sampling that results in a much larger measurement error.
There are two approaches that can be incorporated within a measurement cycle to reduce the
effects of a noise spike. The first is to increase the integration time of the instrument, which in the case of the Model 2750 can be varied from
0.01 power line cycles (PLC) to 60 PLCs over the bus. A longer integration time provides the best common-mode and normal-mode rejection—and
will result in a measurement less susceptible to noise.
The second approach is to implement signal averaging. If ten samples are acquired and averaged, for example, the effects of a spike measured
at any one point will be minimized. Such techniques will improve accuracy, but longer integration times and averaging will increase measurement
cycle times as well.
Relay Life
Back to top
Relay life is another consideration. In a test system such as the one we've shown in our example, the relays of the switch matrix cards are being
actuated at least a few times each minute—the relay life is therefore a consideration.
Since the power supplies are being tested under load conditions, the switching cards have their relays opened and closed with voltage across the
contacts. By actuating a relay in this manner, there is a possibility that arcing may occur, which can severely degrade relay life.
For example, the relays on the Keithley Model 7708 card are rated for 108 closures when voltage isn't being switched, and for 105
closures if 1 A at 110 V is being switched continually. For this example system, 105 closures could occur within days, but the 108
closures spec implies a system life measured in decades!
Moreover, as relay contacts degrade, they create another source of error. Therefore, it's important to have a good idea of the voltage and current
levels on the contacts, especially when you're considering maintenance planning (a unique feature of the Keithley 77xx Series switch cards is
the built-in counter used to monitor contact closures. In our example, the Model 2750 automatically logs the number of relay closures in a memory built
into each module. This provides an accurate way to track actual relay usage and to plan preventative maintenance activities based on the current and
projected contact closure count).
Putting It Together
Back to top
In summary, it's clear that HALT and HASS testing requires a wide variety of sources, measuring instruments, switching systems, RF routers, PC
plug-in cards, and other data-acquisition elements. Each piece of equipment must be selected for reliability, accuracy, repeatability, measurement
integrity, and ease of use.
Furthermore, each piece must be tightly integrated in a system that meets all of your enterprise's objectives. These can include spotting problems
in production, ensuring a high level of quality, identifying critical flaws in new product early in the design phase, and reducing time-to-market.
System designers and integrators need to consider the amount of work required to develop test application software for the system, as this can be a
major part of the time-to-market equation.
About The Author
Jon Semancik is the Digital Multimeter Product Line Manager at Keithley Instruments in Cleveland, OH. He has a BSEE degree from the Fenn College
of Engineering, and is a member of the IEEE.
|
Glossary of HALT/HASS Terms
Back to top
This glossary lists common terms associated with testing in both product development and manufacturing to improve or maintain the reliability of
finished products.
Accelerated life testing (ALT)—ALT is a test activity during product development in which prototypes are subjected to stress (i.e.,
temperature, vibration) at levels much higher than those anticipated in actual use. The aim is to cause failures that identify weak design elements.
Failure-prone elements are then re-designed and tests are continued at higher levels. This procedure is sometimes called a Test, Analyze, And Fix
(TAAF) cycle.
Accelerated stress testing (AST)—AST is a post-production test activity on a sampling (100% at first) of units. The intent is to
precipitate hidden or latent failures caused by poor workmanship, and to prevent flawed units from reaching the next higher level of assembly—or
the end-user customer. Stress intensity typically is half that used in accelerated-life testing.
Burn-in—Burn-in is the process of continuously powering a product, often at constant elevated temperature, in order to accelerate the
aging process.
Characterization—Characterization involves gathering information about a DUT and/or the test system, with the aim of establishing
appropriate test conditions and methods.
Design limit—The design limit is the operational limit of a product beyond which it isn't required to function properly.
Environment—The environment is the aggregate of all external and internal conditions (such as temperature, humidity, radiation,
magnetic and electric fields, shock, vibration, etc.), either natural, man-made or self-induced, that influences the form, function, and reliability
of a product.
Environmental stress screening (ESS)—ESS is post-production testing in which 100% of produced units are subjected to stresses more
severe than anticipated in actual service. The object of ESS is to precipitate failures associated with latent defects so that the failed unit doesn't
proceed further in production or reach the customer.
Environmental testing—Environmental testing involves subjecting a sample of products to a simulation of anticipated storage, transport,
and service environments (such as vibration, shock, temperature, altitude, humidity, etc.).
HALT—highly accelerated life testing. See ALT, accelerated life testing, above.
HASS—highly accelerated stress screening. See environmental stress screening (ESS), above.
Life-cycle testing—Life-cycle testing is a test procedure applied to a small percentage of products, subjecting them to stresses similar
to those found in actual service. Typically, the goal is to determine a product's mean time between failures. See MTBF, below.
Mean-Time-Between-Failure (MTBF)—MTBF is a basic measure of reliability for repairable items. This measure is the mean number of life units
(i.e., hours, years) during which all parts of the item perform within their specified limits, over a particular measurement interval, under stated
conditions.
Mean-Time-To-Failure (MTTF)—MTTF is a measure of reliability for nonrepairable items. It's the total number of life units of an item
population divided by the total number of failures within that population, over a particular measurement interval, under stated conditions.
Reliability—Reliability as a measure is the probability that a product will perform its intended function for a specified interval under
stated conditions (such as cycling in a specified manner over a certain range of temperatures and vibration spectra). This probability is determined
experimentally with repeated tests.
Shock (or shock pulse)—Shock is a transmission of kinetic energy to a DUT in a relatively short interval compared with the DUT's natural
frequency period. A natural decay of oscillatory motion in the DUT follows. Sometimes this is done as part of a product's environmental stress screening.
Thermal cycling—Thermal cycling is subjecting a product to predetermined temperature changes, between hot and cold extremes.
Vibration—Vibration is a mechanical oscillation or motion about a reference point of equilibrium. |
A Freebie
Back to top
Keithley has a free sample program using its TestPoint software to perform a 72-channel switching power supply burn-in test similar to the
system discussed in the article. Modifications may be required to accommodate test parameters and timing for your test system.
Click
here to download your copy of PSBurnIn.tst.
<<
Previous
