Technical The Internal Components The Power Supply Part 3

This PSU guide has been seperated into four pages to reduce your loading time. Click the numbers to navigate this chapter.

Part 1 2 3 4

The Power Supply - part 3

Leave It On or Turn It Off?

A frequent question that relates to the discussion of power supplies concerns whether you should turn off a system when it is not in use. You should understand some facts about electrical components and what makes them fail. Combine this knowledge with information on power consumption and cost, not to mention safety, and perhaps you can come to your own conclusion. Because circumstances can vary, the best answer for your own situation might be different depending on your particular needs and application.

Frequently, powering a system on and off does cause deterioration and damage to the components. This seems logical, and the reason is simple but not obvious to most. Many people believe that flipping system power on and off frequently is harmful because it electrically "shocks" the system. The real problem, however, is temperature. In other words, it is not so much electrical shock as thermal shock that damages a system. As the system warms up, the components expand; and as it cools off, the components contract. This alone stresses everything. In addition, various materials in the system have different thermal expansion coefficients, which means that they expand and contract at different rates. Over time, thermal shock causes deterioration in many areas of a system.

From a pure system-reliability point, it is desirable to insulate the system from thermal shock as much as possible. When a system is turned on, the components go from ambient (room) temperature to as high as 185 degrees F (85 degrees C) within 30 minutes or less. When you turn the system off, the same thing happens in reverse, and the components cool back to ambient temperature in a short period of time. Each component expands and contracts at slightly different rates, which causes the system an enormous amount of stress.

Thermal expansion and contraction remains the single largest cause of component failure. Chip cases can split, allowing moisture to enter and contaminate them. Delicate internal wires and contacts can break, and circuit boards can develop stress cracks. Surface-mounted components expand and contract at different rates from the circuit board they are mounted on, which causes enormous stress at the solder joints. Solder joints can fail due to the metal hardening from the repeated stress causing cracks in the joint. Components that use heat sinks such as processors, transistors, or voltage regulators can overheat and fail because the thermal cycling causes heat sink adhesives to deteriorate, breaking the thermally conductive bond between the device and the heat sink. Thermal cycling also causes socketed devices and connections to "creep," which can cause a variety of intermittent contact failures.

Thermal expansion and contraction affects not only chips and circuit boards, but also things like hard disk drives. Most hard drives today have sophisticated thermal compensation routines that make adjustments in head position relative to the expanding and contracting platters. Most drives perform this thermal compensation routine once every five minutes for the first 30 minutes the drive is running, and then every 30 minutes thereafter. In many drives, this procedure can be heard as a rapid "tick-tick-tick-tick" sound.

In essence, anything you can do to keep the system at a constant temperature prolongs the life of the system, and the best way to accomplish this is to leave the system either permanently on or off. Of course, if the system is never turned on in the first place, it should last a long time indeed!

Now, although it seems like I am saying that you should leave all systems on 24 hours a day, that is not necessarily true. A system powered on and left unattended can be a fire hazard (I have had monitors spontaneously catch fire, luckily I was there at the time), is a data security risk (cleaning crews, other nocturnal visitors, and so on), can be easily damaged if moved while running, and simply wastes electrical energy.

I currently pay £0.07 for a kilowatt-hour of electricity. A typical desktop style PC with display consumes at least 300 watts (0.3 kilowatts) of electricity (and that is a conservative estimate). This means that it would cost 2.1 pence to run this typical PC for an hour. Multiplying by 168 hours in a week means that it would cost £11.76 per week to run this PC continuously. If the PC were turned on at 9 a.m. and off at 5 p.m., it would only be on 40 hours per week and would cost only £2.80, a savings of £8.96 per week! Multiply this savings by 100 systems, and you are saving £896 per week; multiply this by 1,000 systems, and you are saving £8,960 per week! Using systems certified under the new EPA Energy Star program (that is, "Green" PCs) would account for an additional savings of around £1 per system per week, or £1,000 per week for 1,000 systems. The great thing about Energy Star systems is that the savings are even greater if the systems are left on for long periods of time because the power management routines are automatic.

Based on these facts, my recommendations are that you power the systems on at the beginning of the work day, and off at the end of the work day. Do not power the systems off for lunch, breaks, or any other short duration of time. Servers and the like of course should be left on continuously. This seems to be the best compromise of system longevity with pure economics.

Energy Star Systems

The EPA started a certification program for energy-efficient PCs and peripherals. To be a member of this program, the PC or display must drop to a power draw at the outlet of 30 watts or less during periods of inactivity. Systems that conform to this specification get to wear the Energy Star logo. This is a voluntary program, meaning there are no requirements to meet the specification; however, many PC manufacturers are finding that it helps to sell their systems if they can advertise them as energy-efficient.

One problem with this type of system is that the motherboard and disk drives literally can go to sleep, which means they can enter a standby or sleep mode where they draw very little power. This causes havoc with some of the older power supplies, because the low power draw does not provide enough of a load for them to function properly. Most of the newer supplies on the market are designed to work with these systems, and have a very low minimum load specification. I suggest that if you are purchasing a power supply upgrade for a system, ensure that the minimum load will be provided by the equipment in your system; otherwise, when the PC goes to sleep, it may take a power switch cycle to wake it up again! This problem would be most noticeable if you invest in a very high output supply and use it in a system that draws very little power to begin with.

Power Supply Problems

A weak or inadequate power supply can put a damper on your ideas for system expansion. Some systems are designed with beefy power supplies, as if to anticipate a great deal of system add-on or expansion components. Most desktop or tower systems are built in this manner. Some systems have inadequate power supplies from the start, however, and cannot accept the number and types of power-hungry options you might want to add.

In particular, portable systems often have power supply problems because they are designed to fit into a small space. Likewise, many older systems had inadequate power supply capacity for system expansion. For example, the original PC's 63.5-watt supply was inadequate for all but the most basic system. Add a graphics board, hard disk, math coprocessor (8087) chip, and 640K of memory, and you would kill the supply in no time. The total power draw of all the items in the system determines the adequacy of the power supply.

The wattage rating can sometimes be very misleading. Not all 200-watt supplies are created the same. Those who are into high-end audio systems know that some watts are better than others. Cheap power supplies may in fact put out the rated power, but what about noise and distortion? Some of the supplies are under-engineered to meet their specifications just barely, whereas others may greatly exceed their specifications. Many of the cheaper supplies output noisy or unstable power, which can cause numerous problems with the system. Another problem with under-engineered power supplies is that they run hot and force the system to do so as well. The repeated heating and cooling of solid-state components eventually causes a computer system to fail, and engineering principles dictate that the hotter a PC's temperature, the shorter its life. Many people recommend replacing the original supply in a system with a heavier duty model, which solves the problem. Because power supplies come in common form factors, finding a heavy duty replacement for most systems is easy.

Some of the available replacement supplies have higher capacity cooling fans than the originals, which can greatly prolong system life and minimize overheating problems, especially with some of the newer high-powered processors. If noise is a problem, models with special fans can run quieter than the standard models. These types often use larger diameter fans that spin slower, so that they run quiet while moving the same amount of air as the smaller fans. A company called PC Power and Cooling specialises in heavy-duty and quiet supplies. Another company called Astec has several heavy-duty models as well. Astec supplies are found as original equipment in many high-end systems, such as those from IBM and Hewlett-Packard.

Ventilation in a system can be important. You must ensure adequate air flow to cool the hotter items in the system. Most processors have heat sinks today that require a steady stream of air to cool the processor. If the processor heat sink has its own fan, this is not much of a concern. If you have free slots, space out the boards in your system to allow air flow between them. Place the hottest running boards nearest the fan or ventilation holes in the system. Make sure that there is adequate air flow around the hard disk drive, especially those that spin at higher rates of speed. Some hard disks can generate quite a bit of heat during operation. If the hard disks overheat, data is lost.

Always make sure that you run with the lid on, especially if you have a loaded system. Removing the lid can actually cause a system to overheat. With the lid off, the power supply fan no longer draws air through the system. Instead, the fan ends up cooling the supply only, and the rest of the system must be cooled by simple convection. Although most systems do not immediately overheat because of this, several of my own systems, especially those that are fully expanded, have overheated within 15 to 30 minutes when run with the case lid off.

If you experience intermittent problems that you suspect are related to overheating, a higher capacity replacement power supply is usually the best cure. Specially designed supplies with additional cooling fan capacity also can help. At least one company sells a device called a fan card, but I am not convinced that it is a good idea. Unless the fan is positioned to draw air to or from outside the case, all the fan does is blow hot air around inside the system and provide a spot cooling effect for anything it is blowing on. In fact, adding fans in this manner contributes to the overall heat inside the system because each fan consumes power and generates heat.

The CPU-mounted fans are an exception to this because they are designed only for spot cooling of the CPU. Many of the newer processors run so much hotter than the other components in the system that a conventional finned aluminum heat sink cannot do the job. In this case, a small fan placed directly over the processor can provide a spot cooling effect that keeps the processor temperatures down. One drawback to these active processor cooling fans is that if they fail, the processor overheats instantly and can even be damaged. Whenever possible, I try to use the biggest passive (finned aluminum) heat sink and install quality ball-bearing fans.

TIP: If you seal the ventilation holes on the bottom of the original IBM PC chassis, starting from where the disk drive bays begin and all the way to the right side of the PC, you drop the interior temperature some 10 to 20 degrees F--not bad for two pence' worth of electrical tape. IBM "factory-applied" this tape on every XT and XT-286 it sold. The result is greatly improved interior aerodynamics and airflow over the heat-generating components.

For other PC-compatible systems, this may not apply because their case designs may be different.

No matter what system you have, be sure that any empty slot positions have the filler brackets installed. If you leave these brackets off after removing a card, the resultant hole will disrupt the internal airflow and may cause higher internal temperatures.

Power Supply Troubleshooting

Troubleshooting the power supply basically means isolating the supply as the cause of problems within a system. Rarely is it recommended to go inside the power supply to make repairs because of the dangerous high voltages present. Such internal repairs are beyond the scope of this page and are specifically not recommended unless the technician knows what he or she is doing.

Many symptoms would lead me to suspect that the power supply in a system is failing. This can sometimes be difficult for an inexperienced technician to see, because at times little connection appears between the symptom and the cause, the power supply.

For example, in many cases a "parity check" type of error message or problem indicates a problem with the supply. This may seem strange because the parity check message itself specifically refers to memory that has failed. The connection is that the power supply is what powers the memory, and memory with inadequate power fails.

It takes some experience to know when these failures are not caused by the memory and are in fact power-related. One clue is the repeatability of the problem. If the parity check message (or other problem) appears frequently and identifies the same memory location each time, I suspect defective memory as the problem. However, if the problem seems random, or the memory location given as failed seems random or wandering, I suspect improper power as the culprit. The following is a list of PC problems that often are power supply-related:

• Any power-on or system start-up failures or lockups.
  
  
• Spontaneous rebooting or intermittent lockups during normal operation.
  
  
• Intermittent parity check or other memory type errors.
  
  
• Hard disk and fan simultaneously fail to spin (no +12v).
  
  
• Overheating due to fan failure.
  
  
• Small brownouts cause the system to reset.
  
  
• Electric shocks felt on the system case or connectors.
  
  
• Slight static discharges disrupt system operation.

In fact, just about any intermittent system problem can be caused by the power supply. I always suspect the supply when flaky system operation is a symptom. Of course, the following fairly obvious symptoms point right to the power supply as a possible cause:

• System is completely dead (no fan, no cursor)
  
  
• Smoke
  
  
• Blown circuit breakers

If you suspect a power supply problem, some simple measurements as well as more sophisticated tests outlined in this section can help you determine whether the power supply is at fault. Because these measurements may not detect some intermittent failures, you might have to use a spare power supply for a long-term evaluation. If the symptoms and problems disappear when a "known good" spare unit is installed, you have found the source of your problem.

Digital Multi-Meters

A simple test that can be performed to a power supply is to check the output voltage. This shows if a power supply is operating correctly and whether the output voltages are within the correct tolerance range. Note that all voltage measurements must be made with the power supply connected to a proper load, which usually means testing while the power supply is still installed in the system.

Selecting a Meter

You need a simple Digital Multi-Meter (DMM) or Digital Volt-Ohm Meter (DVOM) to make voltage and resistance checks in electronic circuits. You should use only a DMM rather than the older needle type multi-meters because the older meters work by injecting a 9v signal into the circuit when measuring resistance. This will damage most computer circuits. A DMM uses a much smaller voltage (usually 1.5v) when making resistance measurements, which is safe for electronic equipment. You can get a good DMM from many sources and with many different features. I prefer the small pocket-sized meters for computer work because they are easy to carry around.

Some features to look for in a good DMM are:

• Pocket size. This is self-explanatory, but small meters are available that have many if not all the features of larger ones. The elaborate features found on some of the larger meters are not really needed for computer work.
  
  
• Overload protection. This means that if you plug the meter into a voltage or current beyond the capability of the meter's measurements, the meter protects itself from damage. Cheaper meters lack this protection and can be easily damaged by reading current or voltage values that are too high.
  
  
• Autoranging. This means that the meter automatically selects the proper voltage or resistance range when making measurements. This is preferable to the manual range selection; however, really good meters offer both an autoranging capability and a manual range override.
  
  
• Detachable probe leads. The leads can be easily damaged, and sometimes a variety of differently shaped probes are required for different tests. Cheaper meters have the leads permanently attached, which means that they cannot easily be replaced. Look for a meter with detachable leads.
  
  
• Audible continuity test. Although you can use the ohm scale for testing continuity (0 ohms indicates continuity), a continuity test function causes a beep noise to be heard when continuity exists between the meter test leads. By using the sound, you can more quickly test cable assemblies and other items for continuity. After you use this feature, you will never want to use the ohms display for this purpose again.
  
  
• Automatic power off. These meters run on batteries, and the batteries can easily be worn down if the meter is accidentally left on. Good meters have an automatic shutoff that turns off the meter if no readings are sensed for a predetermined period of time.
  
  
• Automatic display hold. This feature enables the last stable reading to be held on the display even after the reading is taken. This is especially useful if you are trying to work in a difficult-to-reach area single-handedly.
  
  
• Minimum and maximum trap. This feature enables the lowest and highest readings to be trapped in memory and held for later display. This is especially useful if you have readings that are fluctuating too quickly to see on the display.

Although you can get a basic pocket DMM for about £30, one with all these features is priced in the £200 range. RS Components carries some nice inexpensive units, whereas the high-end models can be purchased from electronics supply houses.

Measuring Voltage

When making measurements on a system that is operating, you must use a technique called back probing the connectors. This is because you cannot disconnect any of the connectors while the system is running and instead must measure with everything connected. Nearly all the connectors you need to probe have openings in the back where the wires enter the connector. The meter probes are narrow enough to fit into the connector alongside the wire and make contact with the metal terminal inside. This technique is called back probing because you are probing the connector from the back. Virtually all the following measurements must be made using this back probing technique.

To test a power supply for proper output, check the voltage at the Power_Good pin (P8-1 on most IBM-compatible supplies) for +3v to +6v. If the measurement is not within this range, the system never sees the Power_Good signal and, therefore, does not start or run properly. In most cases, the supply is bad and must be replaced.

Continue by measuring the voltage ranges of the pins on the motherboard and drive power connectors:

The Power_Good signal has tolerances that are different from the other signals, although it is nominally a +5v signal in most systems. The trigger point for Power_Good is about +2.5v, but most systems require the signal voltage to be within the tolerances listed:

Signal Minimum Maximum

Power_Good (+5v) 3.0v 6.0v Replace the power supply if the voltages you measure are out of these ranges. Again, it is worth noting that any and all power supply tests and measurements must be made with the power supply properly loaded, which usually means it must be installed in a system and the system must be running.

Specialised Test Equipment

You can use several types of specialised test gear to test power supplies more effectively. Because the power supply is perhaps the most failure-prone item in PCs today, if you service many PC systems, it is wise to have many of these specialised items.

Load Resistors for Bench Testing a Power Supply

Bench testing a power supply requires some special setup because all PC power supplies require a load to operate.

Variable Voltage Transformer

In testing power supplies, it is desirable to simulate different voltage conditions at the wall socket to observe how the supply reacts. A variable voltage transformer is a useful test device for checking power supplies because it enables you to have control over the AC line voltage used as input for the power supply. This device consists of a large transformer mounted in a housing with a dial indicator to control the output voltage. You plug the line cord from the transformer into the wall socket and plug the PC power cord into the socket provided on the transformer. The knob on the transformer can be used to adjust the AC line voltage seen by the PC.

Most variable transformers can adjust their AC output from 0v to 280v no matter what the AC input (wall socket) voltage is. Some can even cover a range from 0v to 140v as well. You can use the transformer to simulate brownout conditions, enabling you to observe the PC's response. Thus, among other things you can check for proper Power_Good signal operation.

By running the PC and dropping the voltage until the PC shuts down, you can see how much "reserve" is in the power supply for handling a brownout or other voltage fluctuations. If your transformer can output voltages in the 100v range, you can test the capability of the power supply to run on foreign voltage levels as well. A properly functioning supply should operate between 180v to 264v but shut down cleanly if the voltage is outside that range.

An indication of a problem is seeing "parity check" type error messages when you drop the voltage to 180v. This indicates that the Power_Good signal is not being withdrawn before the power supply output to the PC fails. The PC should simply stop operating as the Power_Good signal is withdrawn, causing the system to enter a continuous reset loop.

Variable voltage transformers are sold by a number of electronic parts supply houses. You should expect to pay anywhere from £100 to £300 for these devices.

Repairing the Power Supply

Actually repairing a power supply is rarely performed anymore, primarily because it is usually cheaper simply to replace the supply with a new one. Even high-quality power supplies are not that expensive relative to the labor required to repair them.

Defective power supplies are usually discarded unless they happen to be one of the higher quality or more expensive units. In that case, it is usually wise to send the supply out to a company that specialises in repairing power supplies and other components. These companies provide what is called depot repair, which means you send the supply to them; they repair it and return it to you. If time is of the essence, most of the depot repair companies immediately send you a functional equivalent to your defective supply and take yours in as a core charge. Depot repair is the recommended way to service many PC components such as power supplies, monitors, and printers. If you take your PC in to a conventional service outlet, they often diagnose the problem to the major component and send it out to be depot repaired. You can do that yourself and save the markup that the repair shop normally charges in such cases.

For those with experience around high voltages, it might be possible to repair a failing supply with two relatively simple operations; however, these require opening the supply. I do not recommend this; I mention it only as an alternative to replacement in some cases.

Most manufacturers try to prevent you from entering the supply by sealing it with special tamper-proof Torx screws. These screws use the familiar Torx star driver, but also have a tamper-prevention pin in the centre that prevents a standard driver from working. Most tool companies sell sets of TT (tamperproof Torx) bits, which remove the tamper-resistant screws. Other manufacturers rivet the power supply case shut, which means you must drill out the rivets to gain access. Again, the manufacturers place these obstacles there for a reason; to prevent entry by those who are inexperienced around high voltage. Consider yourself warned!

Most power supplies have an internal fuse that is part of the overload protection. If this fuse is blown, the supply does not operate. It is possible to replace this fuse if you open the supply. Be aware that in most cases in which an internal power supply problem causes the fuse to blow, replacing it does nothing but cause it to blow again until the root cause of the problem is repaired. In this case, you are better off sending the unit to a professional depot repair company.

PC power supplies have a voltage adjustment internal to the supply that is calibrated and set when the supply is manufactured. Over time, the values of some of the components in the supply can change, thus altering the output voltages. If this is the case, you often can access the adjustment control and tweak it to bring the voltages back to where they should be.

Several adjustable items are in the supply, usually small variable resistors that can be turned with a screwdriver. You should use a nonconductive tool such as a fiberglass or plastic screwdriver designed for this purpose. If you were to drop a metal tool into an operating supply, dangerous sparks or fire could result, not to mention danger of electrocution and damage to the supply.

You also have to figure out which of the adjustments are for voltage and which ones are for each voltage signal. This requires some trial and error testing. You can mark the current positions of all the resistors, begin measuring a single voltage signal, and try moving each adjuster slightly until you see the voltage change. If you move an adjuster and nothing changes, put it back to the original position you marked. Through this process, you can locate and adjust each of the voltages to the standard 5v and 12v levels.

Disks - The Basics

Other Co-processors