Viewing page 3 of 4 pages. Previous 1 2 3 4 Next
|DC load on PSU
||DBS-2100 PSU load tester
|Any number of thermometers
Fan / DC voltage regulation
|Heath / Zenith SM-2320 multimeter
|Kill-A-Watt Power Meter
|Heath AD-1308 Real Time Spectrum Analyzer
The core PSU test tool on SilentPCReview's test bench is the DBS-2100 load tester, made (in Taiwan by D-RAM Computer Company) specifically for testing computer power supplies. The machine consists of a large bank of high power precision resistors along with an extensive selection of switches on the front panel calibrated in Amps (current) and grouped into the 5 voltage lines: +5, +12, -12V, +3.3, -5, +5SR. Leads from the PSU connect into the front panel. It is shown above with leads from a PSU plugged in.
To ensure safe current delivery, the DC output connector closest to the PSU on each set of leads is hooked up to the load tester. This ensures that the current delivered is distributed to as many short leads as possible. When pushing a PSU to its rated output, the heat generated in the wires can be an issue.
The PSU is tested at 5 DC output power levels:
- 65W: A very typical DC power draw by many system at low / modest load.
- 90W: Established previously as a typical max power draw of a mid-range desktop PC.
- 150W: For higher power machines.
- Maximum (350W) The usual 300W test was left off because it is so close to the max.
Care is taken to ensure that the load on each of the voltage lines does not exceed the ratings for the PSU. The PSU is left running 5~10 minutes at each power level before measurements are recorded.
The DBS-2100 is equipped with 4 exhaust fans on the back panel. A bypass switch toggles the fans on / off so that noise measurements can be made. The resistors get very hot under high loads.
Kill-A-Watt AC Power Meter is plugged into an AC outlet on the side of the DBS-2100 in the above picture. The AC power draw of the PSU is measured at each of the 4 power loads. The Kill-A-Watt is used to measure:
Efficiency (in AC-to-DC conversion) at each power level. This is the efficiency figure provided by PSU makers. It is obtained by dividing the DC power output (as set on DBS-2100 load) by the AC power consumption. Efficiency varies with load, and also temperature. PSUs seem to run more efficiently when warmer, up to a point. Generally, they are least efficient at low power and most efficient at 40~80% power load. The main advantage of high efficiency is that less power is wasted as heat -- this means a cooler PSU that requires less airflow to maintain safe operating temps (read: quieter.)
Power Factor (PF). This measurement can be read directly off the Kill-A-Watt. In simple terms, it tell us how much AC power is lost to harmonics (unnecessary electromagnetic energy) while driving the PSU. In practical technical terms, it is the difference between the measured V(oltage) x A(mperes) and AC power in Watts. PF varies somewhat depending on load. The ideal PF is 1.0, which means no AC power is lost. A PF of 0.5 means that to deliver 100W in AC to a PSU, your electric company actually uses 200W and this is often shown in your electric bill as savings (depends on your electric utility company and your account with them). 100W is lost or wasted. Active PF Correction (PFC) power supplies usually have a PF of >0.95. Passive PFC units usually run 0.6 - 0.8. Non-PFC units usually measure 0.5-0.7.
PF is not significant in terms of noise, heat or performance for a PC, but it is relevant to electricity consumption and energy conservation. If you are running large numbers of PCs, there's absolutely no question of the benefits of high PFC and, to a lesser degree, high efficiency.
The Heath / Zenith SM-2320 digital display multimeter, a fairly standard unit, is used to measure the fan voltages and the line voltages of the PSU output. The latter is done via the terminal pin on the front panel, above the connections for the DC outputs from the PSU.
The Test Lab is a spare kitchen measuring 12 by 10 feet, with an 8 foot ceiling and vinyl tile floors. The acoustics are very lively and allows even very soft noises to be heard easily. The PSU under test is placed on a piece of soft foam to prevent transfer of vibrations to the table top. Temperature in the lab is usually ~20C. This is something of a problem as PSUs usually operate in environments that easily reach 45C. Sited next to or above the CPU, the PSU is always subject to external heat. This brings us to the next topic...
In-case Thermal Simulation
The solution is a AC bulb in an empty case with the PSU mounted normally. The distance between the bottom of the PSU and the top of the bulb is about 7 inches. All the case back panel holes are blocked with duct tape. The only significant exit for the hot air in the closed case is the PSU, which is then subject to a fair amount of heat. Still, the bottom front panel case intake hole is very large. In testing, the front of the case is moved so it hangs over the edge of desk, over free air, to ensure good fresh convection airflow. There are no case fans. This is probably more heat than would be seen by a normal PSU, because in a real case, there are usually other air exits, and at least one case fan.
A 60W bulb is used for the 65W and 90W load tests; it seems a more realistic heat source for those lower power loads. The higher power tests use a 100W bulb.
The PSU must cope with the heat generated by the light bulb plus whatever heat it generates within itself. In real systems, there would be other air exhausts paths, and mostly likely at least one case fan. So a Panaflo 80mm Low speed fan is mounted on the back panel of the test case and connected to a voltage controller. The PSU is run through its load range with and without the fan turned on, to 7V, which is about the level at which most PC silencers would run their case fan. It is a reasonable low noise PC simulation.
For this review, I used a highly accurate calibrated B&K model 1613 sound level meter on temporary loan from the University of BC's acoustics lab.
This professional caliber SLM dates back to 1978, weighs over 10 pounds, and is completely analog in design. It has a dynamic range that spans over 140 dB. The microphone used has a 1" diaphragm that's very responsive to low sound levels and low frequencies. The unit's absolute sensitivity reaches below 0 dBA -- at one point in the midband (1kHz) I was seeing -4 dBA for background noise in the UBC anechoic chamber.
Noise readings were taken with the microphone positioned at 1 meter distance, facing the control panel. The ambient noise level in the live test room was ~16 dBA.
Being done at 1 meter distance, the noise measurements are somewhat comparable to manufacturers' specs. Note, however, that differences in the temperature of the test conditions, and the fact that these measurements were made in a live room rather than an anechoic chamber, makes such comparisons not quite valid, either. Because of the higher background noise level in the live test room, and the high degree of reflections from boundaries (walls, ceiling, floor), these measurements generally run higher than they would be in an anechoic chamber.
As usual, noise measurements are accompanied by descriptions of subjective perceptions. The measurements provide only part of the picture.
|Help support this site, buy from one of our affiliate retailers!|