Archive: SPCR's PSU Test Platform V.3

Power | Reference|Recommended
Viewing page 4 of 5 pages. Previous 1 2 3 4 5 Next

GOAL #3: Having tackled goals #1 and #2, it was time to look at the goal of measuring the actual current and voltage for each line individually for each power level. Voltage measurement is no problem, we've been doing it from day one, manually with multimeters. Given the voltage drops that can occur through multiple contact points, our approach has always been to insert the probe pins into the actual output leads from the PSU for highest accuracy, and we will continue this. To measure current, we had planned to use the clamp meter, but several days of trials showed us just how inconvenient, slow, and most importantly, prone to manual errors this would be. Another solution had to be found.

SHUNT RESISTORS

In the end, the idea for our solution came from a web denizen known as jtr1962 in the forums of Storage Review. This SR forum member had built a hard drive power consumption measurement tool for Storage Review, which he said had been calibrated to be accurate to "within 1% under nearly all conditions, and in most cases better than 0.5%." After reading through his comments, I decided jtr1962 might be able to help. Perhaps the same method he used for the power tool for SR could be applied here. Joe (which the "j" in jtr1962 stands for) turned out to be very helpful indeed. Here is the pertinent part of his email reply:

"Now as for measuring current more accurately, I might suggest just inserting a low-ohm resistor into each line and measuring the voltage drop across it. There are very low value resistors with 1% tolerance which would be suitable. A 0.01 ohm, 25 watt resistor is what I think would be most suitable here. You can read the voltage drop on the 200mV scale of your multimeter and then convert directly to amps just by dividing the voltage drop in millivolts by 10. The 25 watt maximum power dissipation (if heatsinked properly) means that you can deal with currents up to 50 amps, provided of course that the rest of your resistor load can handle it."

Of course! I had no idea that such low value resistors even existed. An accuracy of 1% for a 0.01 ohm part is fantastic; this seems not much more than the resistance of a piece of wire! It's mindboggling to think about how such an item could even be manufactured to such tolerance. Never mind, it's a mystery I don't need to solve. Such resistors do exist, and they were duly purchased online.


Serious looking devices, these resistors. The 4-pin Molex plug is for visual scale.

Four of these 0.01 ohm resistors were used for the four resistive load banks in the PSU loading system: One each for the two 12V lines, the 5V line and the 3.3V line. A resistor was hardwired in series for each bank of load resistors (for each voltage line). Because the resistance is so low, it has no effect on the load seen by the PSU being tested. This is the first benefit of the 0.01 ohm value.

By measuring the voltage drop across the resistor, the current can be easily calculated.

Ohm's Law states:
I (current) = V (voltage) ÷ R (resistance)

So if we measure a voltage drop of 20 millivolts across the resistor, this means the current in the circuit is 0.02V ÷ 0.01 ohm or 2A. If we get a reading of 80 millivolts, this means the current is 0.08V ÷ 0.01 ohm or 8A. Note the multiplier factor to convert from Volts to Amps: It is a very convenient 100. This is the second benefit of using the 0.01 ohm value. (Thank you, Joe!)

Now, we measure the voltage aross the terminals of the PSU output connector. Multiply the measured output voltage by the current obtained above and we get the actual delivered power to within ~1% accuracy. It's all pretty straightforward. (It is, in fact, the same approach we took with our own hard drive power consumption measurement system.)

NOTE: Of course, there is nothing new under the sun. A web search on the term "shunt resistor" brings up many examples of 0.01 ohm resistors used as shunts to measure current. In other words, it's a common practice. Here are a couple of examples:

http://www.tpub.com/content/neets/14175/css/14175_33.htm
http://jet-server.commtest.co.nz/kb2/11624.htm

DIGITAL CURRENT METERS

One deviation I made from Joe's advice: Rather than manually measuring the voltage drop across each resistor with a multimeter, I chose to hardwire digital LCD panel meters instead. It would be much more convenient and time-saving in the long run. The extra cost was minimal, and accuracy would be the same or even better due to less variabiliy in contact resistance (which can occur at the probe points for a multimeter). Installing and wiring the measuring resistor and meter into the +12V2 loader was not difficult, but modifying the DBS-2100 PSU Loader was truly a pain in the you know what. The internal wiring is a nightmare, with extremely stiff large gauge wiring soldered everywhere, going everywhere. All in all, the work took several days to complete. Here are photos of the finished results.


The 12V2 current meter: The toggle switch turns the 9V battery for the LCD panel meter on/off.
The meter only draws 1 mA, and a typical 9V battery can do 500 mA, so with a bit of care, a single battery could last a couple years.

We took some time to test the new current meters with numerous power supplies at varied loads against multimeters and the old clamp meter, with excellent consistency between all the measurement tools. Amongst them, the clamp meter exhibited greatest variance, mostly due to manual errors. As the clamp meter had indicated during the discovery phase of our research, the actual load on the PSU was accurate or very slightly lower than indicated by the DBS-2100 switch markings at low loads, but the error became progressive worse as the load was increased. This is no longer a concern, as the markings on the PSU load switches are used just to bring us into the ballpark, then the current and voltage readings are taken manually to obtain the precise load on each line, and adjusted as needed.

Note that the -12V and +5VSB lines are not measured for current or voltage. We only put a few watts load on these lines together; even if it was 10% low, the error would rarely exceed a single watt, which is not significant.

We tried to duplicate the testing done by 80 Plus and www.efficientpowersupplies.org on a couple of PSUs that we also had samples of. We are relieved to report that our efficiency findings and theirs are virtually identical, within one percentage point. We are now confident that any PSU load we report is within 1% accuracy.

The new rig, with 12V1 current meter reading 3.89A and 12V2 showing 11.15A. The total AC input is 248W, as shown by the Seasonic Power Angel on the right. The new meter on the DBS-2100 displays measured current on 12V, 5V or 3.3V line depending on the position of the rotary switch.


Previous 1 2 3 4 5 Next

Reference|Recommended - Article Index
Help support this site, buy from one of our affiliate retailers!
Search: