OCC ATX Power Supply Testing Methodology

paulktreg - 2008-05-24 14:16:14 in Power Supplies
Category: Power Supplies
Reviewed by: paulktreg   
Reviewed on: June 11, 2008

OCC Power Supply Testing Methodology:

The main aim of any review must be to give the reader an unbiased opinion on whether the power supply is fit for purpose, of sufficient quality to recommend its use and provide information, which will assist the reader in making his/her decision on which power supply to buy.

I have over the past few weeks, looked at tons of power supply reviews on the internet, and some, but by no means all, leave at lot to be desired. Indeed there are many excellent reviews out there with power supplies tested to a standard that I cannot hope to attain at the present time, but these are in the minority. It is not good enough to simply hook a power supply to a computer, run a few programs, measure the voltages with a digital voltmeter (or even worse use bios voltage monitoring) and reach a valid conclusion. Reviews based on this form of testing tend to spend more time on how “nice” the power supply looks and include photographs from every conceivable angle. This method of testing power supplies proves nothing and will lead the author to unknowingly recommend power supplies that just shouldn't be used in certain configurations or perhaps shouldn't be used at all. A substantial number of readers will be involved with overclocking in one form or another and for that they require a stable high quality power supply. I am of the opinion that the power supply is an important part of any system and not enough thought is given in its selection. I will spend less time on how “nice” a power supply looks and more time on how it actually performs.

The only way to test a power supply is to load it up to its maximum quoted output power, sustain it and take accurate measurements of voltage and current to verify correct operation. I will try and take the power supply up to its maximum in three, maybe four steps. The power supply may very well fail before the maximum load is attained and in this case, there is no point in increasing the load, as things are downhill from here. Input and output power measurements are needed to check efficiency and this requires some form of power meter on the AC input and measurements of current and voltage on the outputs. It is also important to know the quality of the supply and this requires the use of an oscilloscope to monitor the ripple and noise levels.

I have neither the specialised test equipment nor financial resources needed to analyse a power supply to what is considered industry standards. However, the series of tests I do perform are comprehensive enough, in my opinion, to reach a valid conclusion. The benchmark for all my testing will be the Intel ATX12V V2.2 standard.

My power supply reviews will follow the standard OCC format for power supplies, until we reach the testing stage and on the next two pages, I have tried to explain the changes and give a brief overview of how I will perform the tests.

Testing:

All tests will be conducted using the UK mains supply voltage of 230V at a frequency of 50Hz. I do appreciate that the main audience of this review will be in the USA, but I don’t think this will be a problem. It will have a small effect on efficiency and users on 117VAC 60Hz mains supply can expect a drop in the region of 3% to 5% on the figures quoted in the review.

(I will at some time in the near future purchase a variac, or varible mains transformer, to enable me to perform tests at 117VAC and 230VAC. The difference in mains frequency of 50Hz or 60Hz will make little or no difference to the results).

 

Test Equipment

The following test equipment is available and will be used for the reviews:

Rigel 266 Medical Electrical Safety Tester
Metrotest MD401 Digital Multimeter
Iwatsu SS-7804 40MHz Oscilloscope
Prodigit 2000MU Power Meter
Compact CT6 Optical Tachometer
CHY 501 Digital Type K Thermometer
Uni-Com Ltd Residual Current Circuit Breaker
Custom Built Power Supply Load

The following test equipment is required:

DSO-2090 USB Oscilloscope - Ordered
Variac 0-270VAC 10A - Required

My Custom Built ATX Power Supply Load is constructed from several networks of aluminium clad wirewound resistors mounted to two large heatsinks with forced air cooling. (An 80mm fan at each end but forced air cooling sounds better). The following loads are available:

3V3 = 3.3A, 9.7A, 13.0A, 16.5A, 19.8A or 29.5A.
5.0V = 5.0A, 10.6A, 15.6A, 17.5A, 22.5A or 33.1A.
12V0(1) = 2.5A, 8.5A, 11.0A, 12.0A, 14.5A, 20.5A or 23.0A.
12V0(2) = 2.5A, 8.5A, 11.0A, 12.0A, 14.5A, 20.5A or 23.0A.
12V0(3) = 2.5A, 8.5A, 11.0A, 12.0A, 14.5A, 20.5A or 23.0A.
12V0(4) = 2.5A, 8.5A, 11.0A, 12.0A, 14.5A, 20.5A or 23.0A.
-12V0 = 0.36A.
5VSB = 2.27A.

The actual current flowing through the various wirewound resistors, is very much dependent on the power dissipation of the load and hence temperature. The values given above are simply derived from Ohm's law and take no account of temperature. The resistance of pure metals and alloys increases as the temperature increases. The actual currents and voltages are all verified with a digital multimeter, only when the temperature has stabilised. Nothing is presumed, and I leave the power supply running for up to four hours before making any measurements. The actual currents drawn by the load are in reality lower than those stated above. Heat is my biggest problem and I have had to limit the maximum power dissipation of the load to 750W for continuous periods. Loads of up to 1200W are possible for short time periods only. This does however, have its advantages, as I can keep my cup of tea warm and I'm sure it will make an excellent fan heater in the winter months. In the future, I plan to add futher resistors to increase the loading available as more power hungry supplies reach the market.

 

Electrical Safety

Electrical safety is hardly, if ever, given a thought in power supply reviews, but I think it’s important. Why shouldn’t a power supply meet the requirements for electrical safety and why is it never tested? Most if not all power supplies sold in the UK will carry the CE mark, amongst others, which implies it was designed to meet the requirements of IEC60950 (Safety of Information Technology Equipment). I will say however, that failures are very, very rare and some will say that this test is unnecessary.

The basic requirements of the standard are as follows:

The resistance between the earth connection point at the AC input and the power supply casing should not exceed 0.1 ohm. Because I will be testing the power supply with the IEC mains lead connected, I will be looking for a resistance no greater than 0.2 ohm at a test current of 10A between the earth pin on the power plug and any external point on the power supply casing.

The earth conductor is a safety device and it is essential that it is present. It provides a return path for normal leakage currents and a fault current path which will hopefully blow the fuse before any serious harm is caused. Having a properly earthed power supply enclosure and hence case, also provides a certain degree of RFI (Radio Frequency Interference) screening.

The insulation resistance between the live and neutral conductors connected together and earth, must exceed a certain level when subjected to a high AC or DC voltage. The exact standard gets a little complicated but needless to say, I will be looking for an insulation resistance of greater than 2M ohm at 500VAC.

This test is basically checking the resistance between the live/neutral conductors and the earth conductor and checking that no low resistance path is present.

The earth leakage or touch current, which simply put, is the current flowing down the earth conductor during normal operation, should not exceed 3.5mA. This is quite important because if the power supply earth conductor should become disconnected due to some fault condition, this current will flow through a user who touches the computer case.

The perception of an electric shock will occur at a current of between 1 -10mA at 50/60Hz. If your power supply is not earthed, it is quite likely you will feel a shock when touching your power supply or computer case. The next step, although very unlikely, is at 60mA or above when this current flowing across the chest will cause ventricular fibrillation.

These three tests are by no means the entire standard, but I believe they are the most important and the only ones I will consider. I am not going to go into too much detail here and just give a simple pass/fail result. If the result in my opinion is a failure I will explain why.

The review will just include the table below.

Electrical Safety Test Class 1
Manufacturer/Model
Pass/Fail
Atrix 500T
 Pass

This section will also list any safety features claimed by the manufacturer like current limiting, short circuit protection, overheat protection, etc.

(You may have noticed the inclusion of an RCD circuit breaker in the test equipment list. I would never recommend anybody running a power supply with the cover removed but on occasion I find it necessary and, if doing so, the use of this device is mandatory. It may save your life).

 

DC Output Voltage Load Regulation

The power supply under test will be loaded at three or maybe four power levels, as evenly spaced as possible, up to the maximum if possible and checked against the ATX12V V2.2 standard for conformity.

ATX12V V2.2 DC Output Voltage Limits
Output
Tolerance
Permitted Range
+3V3
±5%
3.135V – 3.465V
+5V0
±5%
4.75V – 5.25V
+12V0(1)
±5%
11.4V – 12.6V
+12V0(2)
±5%
11.4V – 12.6V
-12V0
±10%
10.8V – 13.2V
+5VSB
±5%
4.75V – 5.25V

The voltage and current will be measured at all stages and the results presented in tabular form for each load level. I will also give the loading as a percentage of the rated maximum. An example is shown below.

Total Load = 569.24W (94.79% of rated maximum)

DC Line
Load Current
Load Power
V Limits
Actual Voltage
Pass/Fail
+3V3
11.64A
36.78W
3.135V – 3.465V
3.16V
Pass
+5V0
13.12A
61.93W
4.75V – 5.25V
4.72V
Fail
+12V0(1)
11.17A
133.37W
11.4V – 12.6V
11.94V
Pass
+12V0(2)
11.03A
130.71W
11.4V – 12.6V
11.85V
Pass
+12V0(3)
7.96A
95.60W
11.4V – 12.6V
12.01V
Pass
+12V0(4)
7.92A
94.96W
11.4V – 12.6V
11.99V
Pass
-12V0
0.37A
4.65W
10.8V – 13.2V
12.56V
Pass
+5VSB
2.17A
10.72W
4.75V – 5.25V
4.94V
Pass
Total Power Supply Load
568.72W
 

 

 

Testing (Continued):

Efficiency & Power Factor

 

 

 

  

 

 

 

ATX12V V2.2 Minimum Efficiency vs. Load
Loading
Full
Typical
Light
Required Minimum Efficiency
70%
72%
65%
Recommended Minimum Efficiency
77%
80%
75%

 

During the Voltage Regulation tests, a note of the relevant figures required to calculate the efficiency of the power supply will be made and the results given in tabular form as shown below:

Output And Efficiency
DC Load (W)
AC Load (VA)
AC Load (W)
PF
Efficiency (%)
Pass/Fail
173.43
232
216
0.93
80.3
Pass
462.11
615
598
0.97
77.3
Pass
569.24
797
779
0.98
73.1
Pass

 

I do appreciate any concerns some people may have that are familiar with AC theory. If the Power Factor and VA values are taken into account, the efficiency value would be slightly higher, but I am for now going to stick with this rather simplistic way of presenting the results.

AC Ripple On DC Outputs

ATX12V Version 2.2 AC Ripple/Noise Tolerance Maximums
DC Output
+3V3
+5V0
+12V0(1)
+12V0(2)
-12V0
+5VSB
Ripple (mV p-p)
50
50
120
120
120
50

 

The maximum ripple/noise will occur at the mains frequency before and after rectification (50Hz – 100Hz) and to a greater extent at the switching frequency of the power supply. The switching frequency will be different for all power supplies and will usually be set above audible range at >20kHz and for this reason, I will scan up to and beyond 100kHz to find the maximum level.

The noise/ripple on the DC lines will be measured with an oscilloscope at the maximum loading of the power supply and the results presented in tabular form, as shown below:

OCZ Stealth X Stream AC Ripple/Noise Measurements
DC Output
+3V3
+5V0
+12V0(1)
+12V0(2)
-12V0
+5VSB
Ripple (mV p-p)
40
50
120
120
100
40
Pass/Fail
Pass
Pass
Pass
Pass
Pass
Pass

 

Temperature, Noise Level & Fan Speed

The testing of the power supply temperatures involves too many variables to draw any valid conclusions. The temperature of the air leaving the unit will depend to a large extent on the ambient air temperature, the enclosure, the cooling of the CPU, case cooling fans fitted, etc., etc. Do not place too much importance on any results given here and regard them as for information only. I will raise any concerns regarding temperature if I feel there may be a problem.

The temperature of the air entering and leaving the power supply will be monitored and the results given in tabular form, as shown below:

DC Loading
Temp In (°C)
Temp Out (°C)
Δ Temp (°C)
Fan Speed (RPM)
173.43W
20.2
26.0
5.8
1245
462.11W
19.2
28.8
9.6
1995
569.24W
19.6
32.0
12.4
2005

 

This test will be performed with a digital thermometer and a type K thermistor in the air path. I appreciate this is a rather crude method, but as long as the thermistor sensor is in a consistent position for all power supply reviews it may prove informative.

The fan speed will be monitored at the various load levels using an optical tachometer. This will involve the attachment of a piece of reflective tape to one of the fan blades and a direct, very accurate, speed measurement in rotations per minute given. 

I do not have access to a sound level meter at this time and all noise levels will be my own personal opinion, which I appreciate may differ from your own.

Conclusions:

I will as my reviews are completed, include some further information so the reader can compare like power supplies. How I do this, is as yet to be decided.

This then will be the basis of my testing and as previously mentioned, the rest of the review will follow the existing OCC format. I have in mind improvements for the testing methodology, which I hope to implement over the coming months. My first few reviews will include photographs of an oscilloscope screen to illustrate the ripple/noise on the DC lines, but I am hopeful that the aquisition of the DSO-2090 USB oscilloscope will enable me to give you better quality screenshots, as my laptop will be my new oscilloscope. The variac is next on my list, in order to test the power supplies for our American brothers (and sisters). I hope you appreciate that to test a 1000W power supply at 117VAC, will require at least a 10A variac, which is quite a monster, so please be patient on this one until I find a bargain.

I hope my reviews prove both informative and useful.