LANs, Desktop Publishing, and Other Final Thoughts

Throughout this guide, we have talked about virtually every technological approach to solving, perhaps, virtually every power problem that might be encountered. We have looked at the electrical environment and the quality of utility power and the problems with it.

For those of us who make a living solving computer power problems, and advising various companies on how best to install computers, both big and small, this comes as no particular surprise. I’m not even tempted at this point to say that these are disasters waiting to happen.

Local Area Networks

Not too long ago, a friend of mine, a well-paid executive in a very large bank, complained that the LAN which serves her department is so unreliable that no one uses it to do any important work. What a shame! Sounds like just another power horror story to justify the expense of a UPS or something.

Let’s put this into better perspective. This local area network is a part of an $8,000,000 computer upgrade that was made by the bank; the upgrade included every one of the bank’s departments throughout the country. As anyone knows, who has ever had to install new equipment, like phone systems, computers, or other high-tech stuff, the first impression of the end users in the organization is absolutely critical. If everyone is afraid of the equipment, or turned off during their first few experiences with the system, it will never see its full potential utilized or its cost justified.

My executive friend was leery of using the LAN because the power kept going off. She and none of her subordinates wanted to loose valuable data if the system were to suddenly go down. So they didn’t use it, period. They still don’t. This is a shame, but not unexpected. LANs are without a doubt the single most vulnerable systems when it comes to power disturbances. There are four major areas of concern we want to highlight. They are:

• Corruption of data from outside interference.

• Ground loops.

• Protection of equipment.

• Losing power without losing data.

Data Cabling

Often, in large data centers, every effort is made to distribute the data-cable wiring in an orderly fashion. Often it's not. and , most certainly, in local area networks is when it usually is not. Cable is stretched here or there, around this and over that. Data-cable wiring is a common source of data corruption by outside influences.

There are several ways of distributing data using various kinds of cable. Typically, coaxial cable, commonly call coax, is used and few problems result. But, it's usually a mistake to run coax any great distance without adhering to special grounding practices, as described earlier in this guide. Of course, using coax throughout an extensive network is an expensive alternative.

Most larger installations have discovered the unexpected advantages offered by twisted-pair wiring. That’s right. More often than not, there are unused pairs of telephone cable in existing trunks which can be used economically to transfer data. There are some advantages to using existing telephone cables. First of all, it's significantly less expensive than coax. Second, it can be interfaced with baluns, matched to coax interfaces, and mated with patch panels for a versatile installation. Granted, twisted pair is more vulnerable to noise than coax. However, if existing cable is used, the lines are probably clear of noise influences since the voice system is already working properly. This is no guarantee, but it's a hopeful first step. Rearranging twisted pair away from noise influences is easier than with coax. Finally, it can be purchased with a shield sheathing, if necessary.

If at all possible, stay away from ribbon cable. There may be some systems that use RS-232 connections to link certain types of terminals, peripherals, or remote devices, but ribbon cable is a notorious source of noise interference. e.g., a recent client of mine had a keyboard which wouldn’t operate properly. Whenever the welding equipment in the next room was operating, the keyboard locked up. The problem was resolved when it was discovered that a ribbon cable had been substituted for the normal shielded cable between the terminal and the keyboard, and the cable was replaced. Normally, we would expect nothing to work in such a hostile environment. The use of ribbon cable guaranteed it.

Ground Loops

A LAN is a ground loop. There is no way to get around it. It’s all well and good to talk about single-point grounding, but when Sally is in accounting, and Betty is in clerical, and Joe is in shipping, this might be a little unrealistic. ill. 13-1 shows the problems typical of a LAN and how ground currents can flow.

ill. 13.1. A simple LAN may form many ground loops. The arrows show the paths that noise might take.

Ground loops can foul communications, create hard disk errors, and /or render one or more terminals useless. The answer to LAN-generated ground loops is not always simple and straightforward. Chokes, baluns, optical isolators, and modems—any or all of these may be the answer. But the simplest form of cure is proper bonding. Practices would include connecting power grounds and data-cable shields where the cables penetrate walls. In some cases, heavy copper strapping can be used to bond the ground buses of two separate panel boards. Bonding to steel building columns can be a cure in some cases. What we are trying to do is give the ground noise a lower impedance path other than the one through the data cable.

Liberal use of three techniques will stop most ground-loop problems. These are bonding, shielding, and isolation. We have talked about bonding. Shielding involves the use of shielded cable, but the term also refers to the practice of using grounded metal barriers between equipment and any potential noise sources. Isolation brings us back to the use of power-line conditioners as a barrier to the entry of noise into a system. The neutral-to- ground bonding on the secondary of a transformer effectively shunts neutral-to-ground noise away from LAN data circuits. The situation may require just the use of one device, or the use of one device on each LAN terminal. However, a transformer of itself can't solve ground loop problems. Careful application of bonding, shielding, and isolation is the answer to the LAN ground-loop situation.

Protection

Corporate America is buying surge strips by the boxcar. This usually happens with the assistance of a corporate buyer who has figured out how to buy a $40.00 surge protector for only $9.99 if he buys them in case lots.

Unfortunately, as we have seen in earlier sections, finding the right surge suppressor can be a difficult task. Underwriter’s Laboratories, Inc. has made that a bit easier, however. With the cost of networks in the five-digit price range, protection and economy have become like bazaar antonyms. Recently, I talked to a PC distributor who moves about 500 units a month. He was looking for a surge strip he could use to fill the demand from his customers. When I told him he could likely find one for about $100 that would do the protection job nicely, he nearly fainted. “Oh no,” he told me, “I’m looking for something I can buy for around $6.00.”

If you read the earlier sections carefully, you should remember that I showed a strong bias toward an isolation transformer that has the neutral and ground bonded together. Since a LAN is almost always subject to ground loops, and since the data tied up in the network is so valuable, it seems logical to at least give the file server this kind of high-quality protection. The cost associated with losing a 40-Meg hard drive and the data it contains ought to justify the expense of a $200 or $300 power conditioner.

Each PC connected to the LAN should have the same kind of protection. While it's hard to keep two widely spaced safety grounds from moving electrically apart, it's surely desirable to keep these noise potentials from appearing across the LAN cable and power ground. A power conditioner with the neutral and ground lines bonded does just that. A surge suppressor will not accomplish this in every desirable instance, even if it has been well engineered to handle common-mode transients.

If the ground-offset potentials are only a few dozen volts, data lines can be corrupted and chips may be damaged by the current flow that results. The voltage levels may be far below the functional threshold of an MOV, so a surge device will not help in this situation. The only threat that remains, if proper power conditioning is used, is that the ground offset caused by the physical distance of the devices, themselves, may cause loop currents to flow. Then bonding becomes another line of defense. We can recognize that these ground currents may exist and can give them a lower impedance path to follow.

This relates to our earlier comments. Running supplemental grounding straps between the breaker panels, or even along cable runs, can accomplish this. Often times, significant potentials can exist between two building columns. Bonding can equalize these voltages. Shorter, thicker wires will divert ground current away from cabling. Another approach, as we stated in an earlier section, would be to bond the actual shield of the data cable to building steel at short distances, in order to equalize loop currents traveling in the cable itself.

Application of a UPS

It’s time to make one of those sweeping statements that will no doubt haunt me as I travel around the country giving seminars and workshops. it's one of those unqualified, broad generalizations that at first may seem a bit harsh, but here goes.

If you have a LAN without a UPS, you didn’t need the LAN in the first place. The whole concept of networking is to share data between intelligent machines. Sometimes the data resides in a file server, and users share its large hard drive and , possibly, a printer. Less often, systems make use of a number of PCs, each with its own individual hard drive.

A hard disk is a delicate instrument that communicates with its magnetic media through a Read/Write head that literally flies in the air. Like an airplane, a sudden loss of power makes the head crash. A head crash is usually a disaster. It can destroy information, make a disk unreadable, and / or can cause actual hardware damage. Head crashes caused by power outages are common events.

Fortunately, not all power outages cause head crashes. What does happen when the power goes out is that the head stops writing to the disk. The volatile memories all fade, and the screens go blank. Depending on the software, intermediate work—meaning work generated but not yet saved in a closed file—may be lost.

At a minimum, the hard disk must have a backup power source. it's essential that the mass of data be protected, files not left open, and hardware not be jeopardized. Each workstation should be evaluated for the possible need for standby power. In expensive systems, SPSs can be added that will give each user just enough time to close files and log off the system.

Orderly Shutdowns

Let’s take this entire concept one step further. Many networks are set up for 24-hour continuous operation. it's highly desirable that the system have some way of coping with an unplanned termination on its own. For this system to be effective, three elements are needed. First, the file server must be on a UPS capable of sending a signal to the file server that the power has failed and the system is now operating on batteries. Secondly, there must be some software resident in the system which will take appropriate action when this signal is received. Finally, the PCs on the network must be on some form of UPS in order to take advantage of Elements One and Two, above.

In fact, there are systems on the market that do all this and more. Let’s step through an outage and see what happens. The moment the power goes out, the UPS sends an “on battery” signal to the file server. This activates a program already resident in memory. Immediately, a message appears on every screen of each PC logged onto the network, saying, “Power Outage, Approximately 5 Minutes Battery Time Remaining.” With each succeeding moment, the minutes are counted down as the battery time expires.

At somewhere near the 1-minute mark, a keyboard alarm sounds at each workstation. The power software begins closing files. But it does two very important things in the meantime. First, it retrieves each active workstation’s log on sequence into a macro file. Also, it begins developing a macro of keystrokes that will preserve work done after the files are closed.

When the power returns, the power software automatically reboots the system and activates the macro files. Each user is logged back onto the system without their touching their terminal. They are placed into the program exactly where they were before, and their last few dozen key strokes appear on the screen. If all works properly, no data is lost, no time is lost, and more importantly, no confidence in the system is lost.

The scenario that we have outlined is actually in use and readily available from network vendors and distributors. Depending on the application, even the keyboard macro options may be a necessity. What does all this cost? Probably somewhere in the range of from 7% to 15% of the cost of the network.

Wow! You mean I pay all this money for the network, and then I have to buy all this power stuff?

Take my advice. Look at your network needs, including the power requirements, up front in the budget. The power products are a part of what the network is all about. It should be included in the original proposal. Vendors that don’t include it don’t have your best interests in mind.

Now, possibly, our statement about the absolute necessity of a UPS with a LAN seems more grounded. Of course, when we talk about LANs, we are talking about a system where real computing is going on, in a commercial setting. Clearly, very small networks that only swap files from the disk is not what we mean. LANs that include a proper power environment will be used more and will be relied on more by their users. The productivity of data sharing will soon justify not only the expense of adding power conditioning and UPS, but the LAN itself.

Desktop Publishing

We are devoting special consideration to LANs and Desktop Publishing not only because they are popular and growing, but because they represent a significant investment for the organization. Desktop publishing presents a very different set of problems than does a local area network. But what are the individual elements of a desktop system, and how do they add up in terms of power protection needs?

Most basic, yet powerful, publishing systems consist of the following elements: a high-speed PC, a large-capacity hard drive, a laser printer, possibly a scanner, and possibly an outboard tape-backup system. Each one of these elements performs a distinctive role at separate times during the process of inputting, manipulating, and printing.

We will assume that all of these elements are involved in a system that we must protect. It seems simple enough on the face of it. Why should protecting this system be any different than any other system? Let’s just add up the power requirements of each element and then buy the device of our choice in the proper size. Well, will that approach work? Let’s try it and see.

The high-speed PC with an internal hard drive of 20-Megabytes needs about 360 VA of power, including the monitor. The external tape-backup system might only need 250 VA. The scanner—about 250 VA also. Gee, we’re only at 860 VA and we have only one item left. The problem is that one item. Many laser printers have ratings of around 1000 VA. Now, quite suddenly, our power requirement has jumped to almost 2 kVA. A 2-kVA power conditioner will cost in the neighborhood of $1000 to $1400. An SPS at the same size will cost in the range of $2500 to $3000. and , hold on to your hat, a true UPS may be in the $4000 to $5000 range.

Before you rush out to buy any of these items, can we find other alternatives that might afford the kind of electrical environment we are after without breaking the bank? The answer is, “Yes.” Unlike a LAN, where continuous processing is a must, a publishing system requires a one-step-at- a-time process.

The real concern is the PC itself. This is where all the work is done, and the arguments here for good power are just as compelling as if it were an important stand-alone PC on some executive’s desk. The fact that so much in- process work will be resident in memory at any one time should indicate that an SPS would be desirable. Depending on how often the actual composition is saved to disk will determine the operator’s exposure to rebuilding lost work after a power outage. This is a judgment that must be made. The SPS to back up this part of the system will cost around $600.

It almost goes without saying that the high-speed PC needs a line conditioner. This is the finest electrical environment available for this equipment, and at a cost of around $300, it's well worth the protection to the hardware and all the data contained on the hard drive. What about the tape backup system and the scanner? These items might, for a couple hundred dollars more, be included on the same line conditioner that feeds the PC. If not, a high-quality surge suppressor will protect this equipment.

And, this is also the answer to protecting the laser printer. Since the laser, scanner, and tape backup are not part of on-going processing, the damage done by corrupting influences is easier to deal with. If, e.g., a bothersome electrical event fouls the data stream from the PC to the laser printer, the corrective action is to print the page over again. This one-step at-a-time job processing allows us to look at these outboard devices more in the light of protecting them from actual hardware damage than from worrying about the actual processing itself. If the result of their function is not satisfactory, we simply repeat the process. On a LAN, however, the process is interactive and the stream of data requires a higher level of power purity.

It is for this reason that the peripherals, including the laser printer, can be protected with less expensive items. Please take note. We are not suggesting a cheap surge strip or anything of the kind. If we are going to substitute a surge device for a power conditioner, let’s be clear that it must be a high-quality design which is fully capable of handling common-mode as well as normal-mode events in such a way that the connected device is completely protected.

What if the publishing system is part of a network? The same thing holds true in that case. The network needs full-time power, as well as power conditioning, to ensure the integrity of data stored and in process. The peripherals that don’t interact with the network, but job process, can be protected from damage at a lesser overall cost.

One point is vital. If a variety of power peripherals are used for economy sake, as we are suggesting, it's imperative that they all be referenced to the same ground, as close to the computer as possible. This can be done by bonding the power units together, or by simply plugging them into the same receptacle on the wall. We must not create a ground loop while trying to solve potential power problems. Also, under normal circumstances, we would never recommend using more than one different type of protective device. After all, buying a power conditioner large enough to accommodate the PC, printer, and any other peripheral will cost only a few more dollars than one sized only for the PC. The exception is a laser printer. Its high-power usage makes it a candidate for exploring the alternatives.

Final Thoughts

One of the important issues that the discussion on LANs and Desktop Publishing exposes is the importance of the application of the PC. The application makes all the difference in terms of evaluating the type of appropriate product technology to be applied. and it's hard to generalize. All home computers are not equal. Some commercial settings have site-specific considerations. Different software applications, hardware configurations, and consumer expectations have everything to do with the kind and level of protection needed.

Now that all the technological options have been laid out, the reader can evaluate the specific needs of his or her application and choose wisely. There is a magic scale that exists, with the need for protection on one side and dollars on the other. Finding the proper balance is not always easy. This guide has covered enough of every area of power protection so that the thoughtful decision maker, who is concerned with the wise use of personal or company money, can make purchases that fit the application and will perform appropriately.

We have tried to develop this guide logically by first arming the reader with the basic knowledge required to understand the subject of computer power. We talked about environmental issues like power quality, grounding, and the PC’s own power supply. When we got to those sections that dealt with the actual products, we started with the simplest, cheapest, and smallest first, and then built up from there. Finally, we applied the technology to two of the fastest growing areas of computing—LANs and Desktop Publishing.

What’s left? The future.

What will the future of computing bring? One thing we know that may be coming is nonvolatile memory. This is memory that is not lost when the power goes out. Certainly, faster clock speeds are on their way. and storage media that is more densely packed, with faster scanning rates, is a phenomenon we don’t have to wait for. It’s happening on an ongoing basis.

The real question is, “Will the pace of rapid advancement of computing technology make power protection obsolete?” The answer is, “No.” The trend has been that advancements have created a greater need for power products. Even some of the enhancements made to make computers less vulnerable to certain kinds of power disturbances have generated a need for other kinds of protection.

Another inevitable question I get at my workshops is, “Will computer manufacturers ever build power conditioning into their products?” I can only say, “Not in my lifetime.” The way that the price wars have been going, the public is lucky to find a power supply in their PC. Nothing that adds a dime’s worth of cost, and which doesn’t return a dollar’s worth of profit, will ever be put in a PC. There have been one or two attempts at including a UPS in PCs, but they have not sold well.

Strangely enough, another industry has done this with great success. The telecommunications industry has been including battery backup as part of their systems for many years. Their power requirements are somewhat different, however. This has not and probably won’t happen in the computer industry.

One thing that is rumored to change, however, is the power supply. New designs are coming out every day. A recent introduction to the field claims to draw current in a sinusoidal waveshape. Whether this will appear in PCs is not clear. But the window is still wide open for new power-supply designs to enter the market. When they do, it will no doubt be necessary to take a new look at the kind of protection that may be appropriate.

There are a number of new chips that are making their way into the market which we should mention. The first is a member of a family of rectifiers that is impervious to EMI. Obviously, the military is the biggest market for these chips. As of yet, these chips have not made their way into the PC market. Another advancement is chip technology that monitors the power, and takes action on a nanosecond basis to thwart power fluctuations. How this interfaces with the machine’s intelligence is not clear. One possibility is that it will halt processing temporarily, in order to ensure the integrity of the data. In a location where power is particularly bad, this might bring processing to its knees—requiring the installation of a UPS or power-conditioning equipment. History has shown that what designers have thought was an advancement, in the way computers deal with utility power, usually requires a new computer power strategy.

Of course, this is the purpose behind this guide. Every computer user needs a computer-power strategy. The information contained here should assist the reader in matching the proper technological solution to their particular needs, without wasting money. To illustrate what this guide was written to avoid, consider the following true story.

A hospital was having power problems with a computerized phone system. The facilities manager called his electrical contractor and both agreed that they probably needed a UPS. The contractor called his electrical supply house and ordered a 1500-watt SPS to solve the problem that the hospital was experiencing. The cost of the item was about $2000. After installing the UPS, problems with the phone system still persisted. An “expert” was called in to analyze the situation. He connected a BMI, which is an expensive power-line analyzer, to the input and the output of the SPS in order to test and see if the unit was performing properly. The tape of the BMI showed that the telephone system was experiencing impulses of up to 173 volts. They simulated an outage by pulling the plug on the SPS and the output low voltage of the SF5 was recorded at 103 volts.

The “expert” concluded that there was no way the telephone system could survive in an electrical environment like that. At this point, everyone descended on the electrical wholesaler. Why had they sold a product that didn’t work? What were they, a bunch of thieves? In a panic, the wholesaler called in their own expert (me).

After a thorough and professional analysis of the BMI tape, and studying the specifications of the SPS, the real expert knew exactly what the problem was. The SPS had a surge protector, but, beyond that, was not designed to condition power in any way. In fact, the peak of the voltage sine wave is 170 volts. No surge to protect here.

Next, the real expert checked the transfer point of the SPS. Surprise! It was 103 volts, which is the typical transfer point of most SPSs. The real expert concluded that the SPS was performing exactly as it was supposed to perform. The problem all along was the facilities manager, the contractor, and the “expert”—none of them understood the technology of the product they needed to purchase. They had heard that UPSs would cure problems so they bought something without really knowing what it was. The result— hard feelings, finger pointing, and $2000 wasted.

Now that you have read this guide, this should not happen. We must look at our problem or our potential problem. We should evaluate our financial resources. and , we should balance the knowledge contained here against the claims of the manufacturers. The result will be the kind of trade-off that we are willing to live with. We can now go back to computing, knowing that we have the foundation laid for PC power protection.

NEXT: Reading List

PREV: UPS Designs

All related articles   Top of Page   Home