Using HALT for Power Supply Preselection Testing
During early production of a new point-of-sale (POS) product, developers integrated an off-the-shelf switching power supply into its design. Because the power supply's size (in terms of the product's approximate 30W load) was to be as small as possible, designers preferred an open-frame configuration with a 5-in.-by-3-in. footprint. Among the many choices on the market, four designs from four different manufacturers were initially the selections. Each design was rated between 60W and 100W, and each had three outputs (+5Vdc, +12Vdc, and -12Vdc). None of the four manufacturers were familiar to the user, so their quality and reliability were unknowns that required assessment. Because time-to-market was short, and first-production reliability of the user's product was vital, HALT (highly accelerated life test) served as a preselection activity. In this case, HALT quickly validated a decision to choose a particular power supply design.
To uncover potential design and manufacturing process problems, the performance of HALT reliability activity is typically early during a product's development. In this instance, HALT provided data to assess power supply adequacy and confidence for selecting a particular supplier.
HALT's first objective is to identify and fix potential failures before they occur, during production or in the field. A second objective minimizes the sample sizes of experiments — while simultaneously compressing real time to simulate potential failures before they happen. To achieve these objectives, HALT regularly applies stresses that are well beyond “normal” limits for shipping, storage, and specified end-use. In the past, this user had developed a HALT standard for PCs and related components, but using it to compare power supply designs during a preselection activity was a first-time event.
In brief, the user's HALT standard is as follows. Starting at +20°C, ambient temperature decreases in 10°C steps (each step is 20 min), until finding a lower operating limit. Thereafter, starting at +30°C, ambient steps increase in search of an upper operating limit. After determining operating limits, temperature steps increase more, until establishing lower and upper limits of destruction. Then, using extremes of temperature set about halfway between the limits of operation and destruction found above, rapid temperature ramps (having ramp rates more than 40°C per minute) are applied between the high and low temperature extremes. The goal is to apply 10 cycles of rapid temperature ramps. (One cycle is defined as: start at +20°C; ramp down to the lower temperature extreme; hold the low limit for 20 min; ramp up to the upper extreme; hold the high for 20 min; and ramp back down to +20°C). After the rapid temperature ramps, with an ambient of +20°C or +30°C, vibrations increase in 20-min steps of 5 Grms (Gravity, root mean squared) each, until finding limits of operation and destruction. Note: All vibrations are random, in six axes, and across a frequency range of 2 Hz to 5000 Hz. Finally, using the midpoints of temperatures between operation and destruction, combine both environments, using very fast temperature excursions (again, more than 40°C per minute) and stepped levels of vibration.
This was the basic method used to evaluate four power supply designs; and it quickly provided decision-making data. Because of the user's unique requirements for the application, three criteria were important considerations during the assessment of the designs.
- The +5V output required stability for a microprocessor load,
- Performance was vital for all load and ambient conditions, even with low-input voltage, and
- Per the user's HALT standard, the expectation of vibration resistance was up to 30 Grms.
The power supply uses a 115Vac input from a 15A or 20A, conventional wall outlet. The three outputs have loads of 4.0A at +5Vdc, 500mA at +12Vdc, and 40mA at -12Vdc.
In five days, the test was finished. During that time, because of chronic problems, two manufacturers were eliminated from further consideration, but two others performed well enough to be ready for a next level of decisions. However, during the rigors of HALT, only one of the manufacturers had passed all of its published performance specifications. True, the more robust design had demonstrated some sensitivity to extreme temperatures and vibrations, but the supplier deemed such problems easily repairable.
Eight power supplies (two from each supplier) were tested together in a HALT chamber. At 15-sec intervals, each of two data loggers separately monitored the inputs and all outputs of a group of four power supplies. Input voltage was initially set at 91 ±1Vac, because all manufacturers claimed 90V as a lower operating limit; and such a low voltage exerts more stress on the power supplies. However, after the test started, a low-voltage setting was not possible for three units.
All eight samples were mounted in a HALT chamber on a ⅜-in.-thick aluminum plate, which was 14-in. wide by 30-in. long (see Photo, on page 42). One sample from each manufacturer was loaded with its respective, maximum-specified load for each of the three outputs. Another sample had three loads, which equaled 120% of the user's anticipated loads for each output. (See Table 1, on page 46, for a summary of test loads for each of the suppliers, A, C, E, and P.) Even with the 20% additions, all user loads were much less than each manufacturer's maximum ratings.
Every 15 sec, all output voltages from the power supplies were monitored. Because the POS application required consistent voltage outputs within specific ranges while under load, output voltage was the monitored parameter of choice. For all +5Vdc outputs, alarms were set at + and -100mV from their nominals. For all +12Vdc and -12Vdc outputs, alarms were set at + and -600mV from their nominals. See Table 1 for a summary of data logger set points.
HALT Results
Tables 2(a) and (b), on pages 48 and 50, summarizes power supply problems during HALT. Prior to the first low-temperature step, the E power supply with maximum-specified loads repeatedly lost its regulation below 98Vac; and it did so essentially at room ambient. Therefore, for this power supply only, input voltage was raised to 103V, and its allowable range was set between 102Vac and 104Vac. (This event is coded S1 in Table 2(a) and (b), on pages 48 and 50.) When asked why the input voltage required an increase, vendor E responded that all outputs should not simultaneously be loaded with their maximum-specified values. This may have been a design truth, but the information was excluded from vendor E's data sheet.
HALT took place at a contract test facility, hundreds of miles from home. Thus, to find product weaknesses, to economize time during the five days, and to expose power supplies to a majority of the user's HALT standard, the written procedure was not followed exactly. The deviation was allowed for this first-time investigation; but as a result, no unit received the standard's entire HALT regimen. There were only three rapid temperature ramps and the combined environments were not precisely correlated to temperature and vibration results because they were done before the limits of destruction had been established.
Temperature Step Results
The first 19 temperature steps were for 20 min each. During the initial 20°C step, the A supply with maximum-specified loads was replaced because it blew its integral input fuse. Soon after, a replacement supply also blew its fuse, needing replacement. To continue testing with two A power supplies, input voltage was raised for the A supply with maximum loads, and the limits for its range were set between 102Vac and 104Vac (as with the E supply with maximum loads).
Three of the supplies had problems with most steps of temperature. For the other five supplies (including the two with boosted input voltage), problems varied from none to few.
To run rapid temperature ramps as soon as possible, it was necessary to stop the high temperature steps before reaching limits of total destruction.
Rapid Temperature Ramps
One cycle of each rapid temperature ramp was from -45°C to +70°C, and back again — using ramp rates in excess of 40°C per minute. There was no vibration, and the dwell time at each temperature extreme was 20 min.
During the first two cycles, five of the supplies had various output-voltage problems (see Table 2). Curiously, there were no problems during the third cycle. Therefore, it was necessary to stop rapid temperature ramps to perform vibration.
Vibration Step Results
The first vibration steps were all performed at +30°C and for 20 min each. Until the higher levels of vibration were reached (i.e., at 20 Grms or above), there were almost no failures from the supplies that had been successful with temperature. Vibration at 25 Grms was repeated because one test day ended with that level; and the following test day resumed with it. At 25 Grms, it is noteworthy that one supply (C, with user loads) had a quick recovery from an apparent total failure (this event is coded T1 in Table 2); but after repeating the 25 Grms step, failure of the same supply was unrecoverable.
Before combining the environments, all test units were replaced.
During each cycle of combined environments, temperature was rapidly ramped (at 40°C or more per minute) through the introduction of each step of vibration. One cycle is defined as: a rapid temperature ramp from +30°C to -45°C (with a dwell of 20 min at the low extreme), then to +70°C (with another dwell of 20 min, this time at the high temperature extreme), and back down to +30°C. For each cycle, starting with 5 Grms, vibration increased in 5 Grms steps towards 30 Grms.
Like with the plain vibration steps, the few supplies that had survived the temperature steps had no apparent problems from low levels of vibration. With the combined environments, however, only two power supplies (C with maximum loads and C with user loads) had no output voltage problems until vibration was 20 Grms or above.
After the first cycle of combined environments, the power supplies were left “On” while the HALT chamber was idle for 1 hr. When the data logger resumed scanning, the E supply with maximum loads was the second supply to die and then quickly recover. This was the third T1 code in Table 2. After recovery, however, this particular supply was sensitive to slight mechanical tapping.
There were two cycles of combined environments having a vibration of 20 Grms. The second cycle occurred because the first cycle had finished on the previous day, but only one supply (P, with user loads) survived the second cycle without any problems.
Before finding an upper temperature destruction limit, all test units were replaced; but none were replaced for the A supplier. Its many earlier failures had precluded further testing.
These additional temperatures were run to find upper limits of destruction. They were run for only 10 min each.
All power supplies survived three steps up to +70°C without any problems. Constant or intermittent behavior started to occur at +90°C. Only one supply (P, with user loads) lacked failures at the higher temperatures.
Added Vibration Step Results
Plain vibration was used to find a destruction limit. Each step had an ambient of +30C, but the dwell time was only 10 min. Throughout all vibration steps, the P supplies were an exception because they had no output voltage problems — until 44 Grms, when they both arced; but neither would die.
Unfortunately, the transportation company lost all tested power supplies during their return trip home, so the original intent of post mortems was not possible. Therefore, conclusions were based on the data acquired during the five days of HALT.
For more information on this article, CIRCLE 333 on Reader Service Card