Perhaps you've fallen into the trap of thinking that a motherboard is just a slab of fibreglass for the all important processor to slot into. Well, it's time to rethink things: the motherboard is the nervous system of your PC.

It provides the essential communication pathways that enable the rest of your machine to do its job, handles the video circuitry and connections to external devices and even resists scrabbling hands trying to rip out graphics cards or rubbing all those essential components. Like all true workhorses, when it does its job, you barely notice it.

Manufacturing them remains a challenge. True, processors have features that are so small that they can't be seen with the naked eye, but the amount of technology at work when building a motherboard is no less impressive.

It's an intensive process – and one that you're about to learn in detail.

1. Raw materials

Like any other electronic item, tracing the motherboard back to its roots leaves us staring at a hole in the ground – or, to be more accurate, a couple of them.

The two dominant constituents of a printed circuit board are fibreglass – which provides insulation – and copper, which forms the conductive pathways, taking us back to their birthplaces in a sand quarry and open-cast copper mine respectively.

Turning sand into glass and copper ore into metal are processes that are hundreds of years old, but what we do with the materials next is anything but ancient.

2. Fabricating copper-clad laminate

Molten glass is extruded to produce glass fibres that are woven to create a sheet of fibreglass fabric. Next the sheet is impregnated with epoxy resin and heated to partially cure the resin; the resulting sheet is called 'prepreg'. Multiple sheets of prepreg are stacked to produce a laminated sheet of the required thickness.

Sheets of copper foil are applied to both sides of the laminate and the sandwich is placed in a heated press. This completes the curing of the resin, making the laminate rigid and causing the layers to bond together.

The result is an insulating sheet of fibreglass with copper foil on both sides: copper-clad laminate. The overall thickness of the printed circuit board (PCB) is typically 1.6mm. This means that, for a six-layer board, the fibreglass laminates will be about 0.35mm thick and the copper foil will be about 0.035mm thick.

The fibreglass is thick enough to provide adequate mechanical strength and rigidity, and the copper is sufficient for good electrical and thermal conductivity.

3. Etching away unwanted copper

A photosensitive material called photo-resist is applied to both sides of the copper-clad laminate, totally covering the copper layers. This is usually a dry film process, in which thin films of solid photo-resist are laminated onto both sides of the board using equipment that's fairly similar to an office laminator.

Now a transparent artwork showing the pattern of the PCB's pads and tracks is placed over the photosensitive copper-clad laminate, and is then exposed to ultraviolet light. Ultraviolet is used rather than visible light so the board can be handled safely in daylight.

Where the photo-resist is exposed to ultraviolet, the chemicals polymerise, forming a plastic. Since the board has two copper layers, each of which has a photo-sensitive coating, this process is carried out twice using different artworks for each side.

Next, the board is immersed in a chemical solution to develop the latent image. The developer washes away the unexposed photo-resist, leaving only material that was polymerised and which corresponds to the pad and tracks. The areas of the copper film that aren't protected by the remaining polymerised portions of the photoresist are etched away.

In an oxidation reaction, metallic copper is transformed into a copper salt, which is water-soluble and therefore washes off during the etching. For quick etching, the board passes through a chamber in which the etchant is sprayed at a high pressure and at a temperature of about 50C.

After etching, the board is washed to remove surplus etchant and the remaining photo-resist is removed using an organic solvent. The insulating fibreglass board now has a pattern of copper tracks on each side that will form the circuit's interconnections. This assembly is called a core.

However, motherboards have a multilayer construction, which means they have more than two copper layers. This means that the above process has to be carried out several times. In the case of a six-layer motherboard, two of these cores will be needed to provide four of those layers. We'll see later how the other two layers are made.

4. Building up a stack

Double-sided cores are now sandwiched together to start the creation of a multilayer PCB. Two cores are used for a six-layer board (a common figure for motherboards), but they can't be stacked directly on top of each other because this would cause the copper tracks on the top of the bottom core to short with the tracks on the bottom of the top core.

To stop this from happening, a sheet of prepreg is placed between them. Sheets of prepreg are also applied to the top and bottom of the stack before it's subjected to pressure and a high temperature to complete the curing of the prepreg and bond everything together.

For a six-layer board, the stack would comprise: prepreg / core / prepreg / core / prepreg. This means that the final result will be: fibreglass / copper / fibreglass / copper / fibreglass / copper / fibreglass / copper / fibreglass.

5. Drilling the holes

Holes are now drilled through the board. First come the mounting holes, which will be used for mechanical fixing (bolting the motherboard into the PC's case).

Second are the holes that are used to accommodate the leads of through-hole components when they're soldered to the board in a couple of steps' time.

Finally, there are the tiny holes that form vias (vertical interconnect access), which make electrical connections between the various copper layers – or will, when we get to routing, testing and QA.

Despite the use of a high-speed, numerically controlled drilling machine, drilling can be a very time-consuming process, especially if lots of different hole sizes are required. For this reason, it's common to stack boards together so that several are drilled at once, saving time and money.

6. Copper and tin plating

Electro-plating would be an obvious choice to make the vias conductive, except for one minor problem: only already-conducting surfaces can be electro-plated. To get around this, the board is immersed in various chemicals that coat its entire surface with a thin layer of copper. It's a slow method and very expensive, but it provides just enough conducting metal to electro-plate over the top.

Electro-plating the entire board would be wasteful because most of the copper would subsequently be etched away to produce the pads and tracks on the outer layers of the PCB. Instead a photo-resist is applied, exposed to UV light through an artwork and developed as when fabricating the copperclad laminate – but with one important alteration.

Here, a different type of artwork is used so that the photo-resist remains in those areas that don't correspond to the pads and tracks of the finished board. Now the electro-plating will only increase the thickness of the copper on the areas without the insulating photo-resist.

The board is finally electroplated with tin, which, once again, only adheres to those areas of the board that will form the pads and tracks. The tin serves three purposes: it prevents the copper tarnishing; it provides a surface that can be soldered to more easily than copper; and it acts as a resist (after first removing the remaining photo-resist) in the next process – etching away the unwanted copper.

We now have a PCB with copper pads and tracks on the outer two surfaces, tracks on four internal layers, and vias making the necessary connections between the various layers.

To complete the bare PCB, a solder mask and component identification are applied via silkscreen printing. The solder mask covers all of the board where solder shouldn't adhere when the components are fixed in place. This prevents unwanted bridges between tracks that could occur during wave soldering in step 9.

The component identification provides a visible labelling of each of the components with their serial numbers. This is useful in manual inspection or board maintenance.

7. Routing, testing and QA

Steps 2 to 7 involved the processing of a panel – a sheet of material comprising several motherboard PCBs. Now the individual boards are separated using a numerically controlled router, which is also used to create any non-plated larger holes and slots that are needed.

The board is then given a going over by a 'bed-of-nails' tester, an automated process that probes both sides of the board to ensure that electrical pathways exist where they are supposed to and that there are no shorts.

Finally, before leaving the PCB fabrication facility, the motherboard is given a QA inspection to ensure it meets its specification in terms of the overall board size, mounting hole tolerances and so on.

8. Surface mounting

The first components to be soldered onto the bare PCB are the surface mountings. Solder paste – a mixture of solder powder and flux – is printed onto those pads on the top surface of the board where the contacts of the surface-mounting components (SMCs) will be soldered. The SMCs are placed on the board using a pick-and-place machine.

The tackiness of the solder paste holds the components in place, but they're not fixed securely and there isn't a proper electrical connection.

The next stage is reflux soldering. The PCB is placed in a reflux oven and heated to over 200C. The solder in the paste melts and then solidifies when the board cools down again, providing good electrical connections and fixing the components securely.

9. Through-hole components

Next the larger through-hole components are fitted, often on a manual production line. Included are the processor socket, the memory and expansion card slots and the various connectors such as keyboard, mouse, audio and video sockets. The components are fitted to the top side of the board with their pins protruding through pads on the bottom side of the board.

The board then enters a wave soldering machine. This contains a tank of molten solder that's pumped across a submerged edge, causing a raised wave of solder. As the board progresses through this apparatus, each part of the bottom side of the board comes into contact with the solder wave. The solder adheres to the board wherever it's free of solder resist, thereby making mechanical and electrical connections between the component leads and the pads.

10. Final testing and packaging

For final testing, processor and memory modules are plugged into their sockets. External PC components such as a hard disk, CD/DVD drive, monitor, keyboard and so on are also plugged into their appropriate connectors. With the motherboard now effectively built into a complete PC, a full functional test involving every socket is carried out.

This is mostly an automated process, although humans do still have a part in the process for areas like audio circuitry. All this is followed up with a 'burn-in' test, which involves running diagnostic software on the motherboard for a protracted time while it's subjected to high temperatures and temperature cycles.

If the board passes this test, which is designed to cause any potentially faulty components to fail, the motherboard is complete. All that remains is for the finished board to be packaged in an antistatic bag and box, and it's ready to take pride of place in a new PC.

Popular culture is filled with examples of technological products that were heralded as ground-breaking but, due to the rate of development in the world of computing, went on to sink without a trace. Although some of them are best left forgotten, others were truly innovative.

Here we reveal five pieces of tech that were big in their time but fell by the wayside. In each case, though, their legacy is still with us today.

1. OS/2: The OS that wasn't

OS/2 was jointly developed by Microsoft and IBM as the primary operating system for the IBM PS/2. The PS/2, in turn, was IBM's second-generation personal computer architecture, launched in 1987 with the intention of clawing back market share from PC clone manufacturers.

It was a miserable failure, with customers preferring to stick with the regular PC architecture, but OS/2 remained. Back in the mid '80s and even into the '90s, it was at the cutting edge of OS technology.

When OS/2 was launched, most computers were running DOS. This provided only a command-line interface. Except for running programs, it did little more than provide some mundane ways to copy, rename and delete files. User-friendly it certainly wasn't.

To stand any chance of success, IBM knew that OS/2 would have to support MS-DOS software. So it did, right from the start. From version 2.0 in 1992, it even had the power to run Windows applications as well, thanks to OS/2's support of Virtual DOS Machines (VDMs). Each of these could run a separate DOS program, which was all that Windows was at that time.

Today, virtualisation offers the possibility of running apps written for one OS under another. Many think this is a new feature, but OS/2 was providing a surprising degree of cross-platform support over 20 years ago.

The first version of OS/2 was text-only. A rather basic graphical interface, the Presentation Manager, was introduced with version 1.1. But with version 2.0, OS/2 gained what we would recognise as a Windows-like user interface that went by the name of Workplace Shell.

This was just one of the aspects in which OS/2 – and particularly OS/2 Warp, which launched in 1994 – attempted to court home users as well as computer professionals.

Another example was support for video playback, which was integrated into OS/2 from version 2.1 in 1993. By way of contrast, it took until Windows 95 for Microsoft to catch up (though add-ons were available earlier).

To think of OS/2 as merely an open-platform operating system with a future-looking user interface would be to do it a great disservice. Behind the scenes it was also breaking new ground in many other areas. OS/2 brought support for 32-bit processors to the desktop years before Windows 95. It offered multitasking, permitting several apps to run at the same time, and it provided an unprecedented level of security.

Prior to OS/2 it was quite possible for a badly behaved application to crash the entire PC. It's interesting to note that OS/2 version 3.0 never materialised, at least not under that name. Instead the technology saw the light of day – after a convoluted and murky development process – in the guise of Windows NT, which was aimed primarily at the server market and seen by many as the first 'serious' version of Windows.

Given that Windows 2000, XP, Vista and 7 are all based on the NT architecture, it's clear how much we owe to the ideas pioneered in OS/2.

2. Windows 3.1

Windows came to the PC in 1985, but it didn't exactly take the world by storm. Neither did versions 2.0 or 2.1, which followed in close succession. That privilege was reserved for Windows 3.1, which launched in 1992 and sold over two million copies in its first few months – a phenomenal achievement for an operating system.

In the early days of Windows, when you started up your PC you'd be presented with a DOS command line prompt such as 'C:/>'. You could then start a DOS program by typing its name, or you could start up Windows by typing WINDOWS.

The fact that Windows was started in this way indicated that it was little more than any other DOS program. Indeed, these early versions have been described as graphical front-ends rather than operating systems in their own right.

With Windows 3.x, especially the hugely popular Windows 3.1, all of this changed. Though Windows still required DOS to provide much of its core functionality, it became a platform in its own right. In some respects, Windows 3.1 didn't have the same level of sophistication as OS/2. But given that it proved far more successful in the marketplace, it too has had a major impact on the versions of Windows that followed in its wake.

Formerly, PC OSes imposed a limit of 640kB for user applications, but Windows 3.1 smashed through this barrier to give users access to up to 16MB. This might not sound a lot today, but in the days of 80286 and 80386 processors it was a vast amount. More importantly, the OS started the trend of ever-increasing memory availability.

Another of Windows 3.1's innovative features was support for TrueType fonts. Familiar names such as Arial, Times New Roman, Symbol and Windings are all with us still.

For the first time, the on-screen display was WYSIWYG ('what you see is what you get'), meaning that you could create a document on-screen in the sure knowledge that it would appear just the same when you printed it out. Today we take this for granted, but it was key to bringing desktop publishing to the mainstream.

Windows 3.1 also appeared in the guise of an enhanced version called Windows for Workgroups 3.1. This was much the same as the base version, but it also provided native networking support. PCs were suddenly able to share files and printers, an essential part of business computing.

Also included in this version were utilities for Messenger-like chatting and email. In the main these were used across a LAN rather than via the internet, but they were more examples of how this long-forgotten version of Windows was ahead of its time.

3. ICQ

Launched in 1996, ICQ (pronounced 'I Seek You') led the way in promoting web-based instant messaging. Now owned by AOL, but quite distinct from AIM (AOL Instant Messenger), ICQ is still around. We hear much less of this particular IM client today, yet modern counterparts owe much to its pioneering technology.

Despite its head start, it was commercial considerations that pushed ICQ into the margins. The likes of AIM, MSN Messenger (now Windows Live Messenger) and Skype entered the market from different angles, polished up the experience and ultimately proved to have more clout.

AOL was one of the first to jump on the ICQ bandwagon – it launched its competitor just a few months after ICQ. Skype didn't appear on the scene until 2003, and initially its product was quite different in offering, being primarily a VoIP (Voice over Internet Protocol) service. But by then all the major textual instant-messaging clients were starting to add VoIP, and very soon the dividing lines got very indistinct.

Today, casual users will see little difference between the major products. In addition to textual instant messaging and VoIP, the vast majority of clients now offer video conferencing, a whiteboard so you can sketch as you chat and even photo sharing, file sharing and app sharing.

ICQ didn't represent the first steps in internet-based chat. For that we'd have to go back to 1988 and IRC (Internet Relay Chat), but that was subtly different from instant messaging. With IRC, you joined a channel to take part in a discussion among lots of people, many of whom you might not know. It was used extensively by the academics and enthusiasts who inhabited the internet in the early days.

ICQ broke new ground because its service restricted communications to your group of known contacts. ICQ's current market share may pale next to that of the big boys, but without it, instant messaging might never have been.

4. RealPlayer

Think streaming video on the web and you think Flash, right? Well, it hasn't always been that way. At one time, wherever you looked, the web was awash with RealVideo movies. True, RealVideo hasn't totally disappeared into oblivion, perhaps because the RealPlayer browser plug-in can also handle MPEG video and play DVDs, but the native RealVideo format is rarely encountered in the wild any more.

RealPlayer first hit the streets under the name of RealAudio in 1995. Although it was one of the first players capable of handling streaming media, it was capable of audio playback only. Support for video coincided with a name change to RealPlayer in 1997, and back then it was ahead of the field.

For the first time, it was possible to watch video footage over the web in real-time rather than first having to download a video file and then play it back. In its first 24 hours of existence, over 100,000 copies of the software were downloaded. It's no exaggeration to say that the web hasn't been the same since.

In some ways, the reason for RealPlayer's early success is simply that it got there first. In such a fast moving market, a couple of years is a huge head start. Even so, before the end of the decade it was joined by Apple's QuickTime (which gained streaming support from version 4), Microsoft's Windows Media player (version 6), and Macromedia Flash (streaming video from version 4).

What's more, throughout the late '90s the RealVideo format was generally praised for the quality it could achieve at high compression ratios – an important factor in the days of dial-up networking.

Today Adobe has pretty much won the streaming video wars with its Flash Video format and plug-in. This success has been attributed to the plug-in's small download size and ease of installation, but its adoption by YouTube has probably proved more significant. Things could have been very different.

Being the first sometimes means launching a product before its time has really come. When RealPlayer came to market, under one per cent of British households could boast a broadband connection, and most people were using a 28k or 56k dialup service. As a result, streaming video was produced at resolutions as low as 160 x 120 pixels and a super slow frame rate.

By way of comparison, today's standard definition digital TV has a resolution of 720 x 576, and for HDTV the figure can be as high as 1,920 x 1,200. Streaming video on the web is now available in high-definition (YouTube, for example, provides up to 1,920 x 1,080 for users with at least 5MB/s bandwidth).

The technology may not use RealVideo itself, but it definitely owes it a debt.

5. Intel's MMX

The basic instructions that mainstream processors for the PC are able to execute are usually referred to as the x86 instruction set, which can trace its heritage back to the 8086 that appeared in 1978. It would be wrong to believe that today's core chips are only able to execute the same instructions as the 8086, though.

Although each x86 processor is backward-compatible with its predecessors – meaning that it can execute all of their instructions – new instructions have been introduced throughout the 32 years of the x86 processor's life.

MMX was the name given to a whole group of additional instructions that were added to a Pentium variant launched in 1996. It could be argued that it was the most significant addition to the x86 instruction set since the migration from a 16-bit to a 32-bit architecture in 1985.

It's generally assumed that MMX stands for MultiMedia eXtension, although Intel has never confirmed this and some realistic alternatives have been suggested. The MMX instruction set comprises 47 'single instruction multiple data' (SIMD) instructions that operate on 64-bit registers.

Rather than operating on a single 64-bit value – an ability that was only introduced into the x86 instruction set with 64-bit processors – the MMX instructions work with multiple values packed into 64-bit registers. This permitted the instructions to use two 32- bit values, four 16-bit values or eight 8-bit values.

This level of parallelism is particularly useful in graphics processing, where it offers a substantial performance boost. It's also a trick that was employed in the specialist vector processors that were used by supercomputers for many years.

So why do we hear so little about MMX today? Not because it's been discontinued, but because it's been joined by other SIMD instructions that offer even greater performance benefits for multimedia apps.

First up was SSE (Streaming SIMD Extensions), a technology that made its debut in the 1999 Pentium III processor. It introduced 128-bit registers and a massive 70 new instructions. What's more, while MMX could only operate on integers (whole numbers), the SSE instructions are able to work on floating-point numbers as well.

Since then the technology has seen incremental enhancements to the architecture known as SSE2, SSE3, SSSE3 and SSE4, and AMD has introduced some extensions of its own. Plus yet more improvements are in the pipeline.

AVX (Advanced Vector Extension) was announced by Intel in 2008, and is expected to appear in processors soon. Initially it will feature 256-bit registers, although some pundits reckon that, in time, this will increase to 512 or even 1,024 bits.

Without the introduction of MMX, then, it's possible that the incredible feats of graphics technology we see today could never have been achieved. And that idea is something that's true to all the technologies we've discussed here. In computing it's always important to look to the future: but we mustn't forget to tip our hats to the past, either.

There's always a temptation to try to predict the course of future events, and the world of computing is no exception. With everything seemingly becoming bigger, better and faster year on year, there's an insatiable appetite for predictions, and some individuals seem dutybound to meet that demand.

Today these people like to call themselves futurologists, and while this might give the practice an air of scientific respectability, in many cases it has to be said with hindsight that using a crystal ball would have been just as accurate.

Intrigued? Then join us as we examine seven prophecies that fell short of the mark.

"It would appear that we have reached the limits of what it's possible to achieve with computer technology, although one should be careful with such statements, as they tend to sound pretty silly in five years."
John von Neumann, 1949

Though the second part of this statement is certainly right, the first bit is unbelievable. Von Neumann was an eminent scientist and mathematician, and developed the computer architecture we still use today. Not a person you'd expect to make such a rash statement.

The fact that he couldn't think of any possible new applications for computers suggests a serious lack of imagination considering that, in 1949, computers hadn't been used for much yet.

In that year, you could count the number of operational computers in the world on your fingers. They had all been developed in universities and were deployed only for scientific purposes. It would be another two years before J Lyons and Company launched LEO (Lyons Electronic Office computer), the first computer designed specifically for business applications. So that's one more potential application, for starters.

Von Neumann's contemporaries weren't as blinkered, though. As he was uttering these immortal words, Claude Shannon – now regarded as the father of information theory – was working on some truly groundbreaking applications.

In 1950, he took one of the first steps in the development of artificial intelligence by demonstrating an electromechanical mouse that could find its way around a maze. The same year, he published a paper detailing how computers could be used to play chess. So much for von Neumann's prophecy!

"I think there is a world market for maybe five computers."
Thomas J Watson, President of IBM, 1943

Computer historians dispute the validity of this quotation, but even if Watson himself didn't utter those words, there's plenty of evidence that computer experts expressed such a sentiment as recently as the early '50s. And the idea wasn't as daft as it sounds.

COLOSSAL MIS-JUDGEMENT: The first electric computer had yet to be built when Thomas J Watson predicted a market for five of them

Back in 1943, the world's first fully electronic computer of any sort – the code-breaking Colossus at Bletchley Park – was just in the process of being commissioned. It would be another five years before the first ever computer as we now understand the word (the Manchester Baby) was built, a further eight years before the first commercial computer (the Ferranti Mark I) went on sale, and 10 years before Watson's own company, IBM, launched its very first computer (the 701).

Of course, we all know that this prophecy turned out to be absolute rubbish, but the vast scale of the under-estimation might still be an eye-opener. Forget PCs (over a billion of them) and think of microcontrollers. They outnumber the world's population many times over, and each one is vastly more powerful than anything Thomas Watson might have envisaged.

"Computers in the future will weigh no more than 1.5 tons."
Popular Mechanics, 1949

Before you dismiss this prediction as coming from an unlikely source, we should tell you that Popular Mechanics has been one of America's leading science and technology magazines for over 100 years. And as you'd expect from such an August publication, the prediction was, for the most part, spot-on – the vast majority of today's computers do indeed weigh in at less than 1.5 tons. Not all of them, though – not by a long way.

Jaguar, the world's fastest supercomputer, is housed at the Oak Ridge National Laboratory in Tennessee and weighs in at almost 200 tons. That doesn't even include the massive air conditioning units that are needed to get rid of the heat that's generated by almost a quarter of a million processor cores, which consume 10 megawatts of power between them.

FAT-CAT: Even today, some computers weigh more than 1.5 tons – this one considerably more

To be fair, though, at 1.75 petaflops, Jaguar is about two thousand billion times faster than 1949's latest and greatest.

"There is no reason for any individual to have a computer in his home." Ken Olsen, co-founder of Digital Equipment Corporation, 1977

He really ought to have known better. After all, the company Ken Olsen founded was responsible for the first of two important milestones in the history of home computing.

Prior to the early '60s, a computer was one thing and one thing only – a mainframe. It would be priced in hundreds of thousands of pounds, if not millions, occupy a whole room and require a full-time staff to operate and maintain it.

In 1964 DEC launched the PDP- 8, which is generally considered the first commercially successful minicomputer. It was the size of a refrigerator, cost $18,000 and over 50,000 were sold – more than any other computer before it. For the first time, a computer could be owned by a single department, not a huge organisation, and it could be operated by people who weren't scientists.

Computers were starting to pass from a select few to the many. Even more surprising, though, is the fact that Olsen made this statement after the second of those two milestones had passed. That was in 1975, when the MIPS Altair 8800 became the first personal computer to sell more than a handful of units.

"640kB should be enough for anyone."
Bill Gates, 1981

He later denied it, but this was allegedly Bill Gates' take on the maximum amount of memory a computer would need. Even if he didn't actually say it, we can be pretty sure he believed it, as it seems fairly realistic in context.

Previous personal computers were based on 8-bit processors, which meant they couldn't address more than 64kB of memory. But even this would have been the stuff of dreams for most home computer users of the day.

Perhaps the best known British home computer that year was the Sinclair ZX81, which had just 1kB of memory.

To put this in context, let's bring it up to date. If you were offered a PC today with 2.56TB of memory, wouldn't you think it was enough for anyone – at least for a few more years?

"I have travelled the length and breadth of this country and talked with the best people, and I can assure you that data processing is a fad that won't last out the year."
Editor in charge of business books, Prentice Hall, 1957

The computer revolution might already have been almost 10 years old by this point, but computers were still pretty thin on the ground. With an estimated 100 of them in use in 1953 and 250 in 1955, this new technology wasn't exactly taking the world by storm.

What's more, the phrase 'data processing' refers to business applications, which were lagging well behind technical computing. Lyons, of teashop fame, launched LEO, the first ever business computer, in 1951. But by 1957, only one was in operation – and that was used by Lyons itself for valuation jobs and payroll processing. Even Big Blue was slow to make an impact on business computing.

Its first offering, the IBM 702 Electronic Data Processing Machine, was only in production from 1953 to 1954. Its replacement, the 705, broke new ground by being the first commercial computer to use magnetic core memory, but the number sold isn't on record. What we do know, though, is that back in the '50s, IBM was overshadowed by a company now long forgotten: Remington Rand, later known as Sperry Rand.

Its earliest computer, the UNIVAC, first shipped in 1952 and was designed from the outset for business and administrative use. It did well, but success was relative back in the '50s. By the time the UNIVAC was replaced by the UNIVAC II in 1958, a grand total of 46 devices had been sold.

Given that such machines cost between $1.25 and $1.5million (around $10million today), this gloomy prophecy wasn't too surprising. We bet he thought differently in another five years, though.

"Transmission of documents via telephone wires is possible in principle, but the apparatus required is so expensive that it will never become a practical proposition."
Dennis Gabor, 1962

Dennis Gabor wasn't your average scientist – he was a Nobel Prize winner. That award was for his invention of holography, but he also applied his considerable talents to the theory of data communication. So he really ought to have known what he was talking about, but it turned out he didn't – at least not on this particular subject.

It wasn't long before his error was exposed. Later that same year, AT&T launched the Bell 103, which was the first commercially successful modem. It was now possible to transmit data at 300 bits per second across an ordinary telephone line. In fairness to Gabor, this technology was still too slow and too expensive to be used for anything other than mainframe communication.

It wasn't until the early '80s that the proliferation of bulletin boards heralded the era of low-cost data communication that was available to Joe Public. Just a year after making this spectacularly inaccurate prediction, Gabor had a change of heart on the subject of forecasting the future.

In his 1963 book, Inventing the Future, he wisely stated that "the future cannot be predicted, but futures can be invented". This is surely a fitting place to conclude our investigation of computing's most unreliable and inaccurate prophecies.

With 500GB disks now shipping as standard on all but entry-level PCs and 2TB disks costing less than £130, storage space has never been more abundant or affordable.

So the recent proliferation of companies offering online data storage products and services might seem surprising – except when you factor in the features and benefits online storage offers that could never be met by local storage alone.

Among the list of unique benefits are fully automated backups, synchronising files between different PCs and sharing files with colleagues, friends or even the whole world. Here we investigate these benefits and take a look at four products that claim to offer all of these services and more.

File synchronisation

USB flash drives enable us to carry large amounts of data around in our pockets. But in order to benefit from them we must be organised enough to make sure that they always contain the files we're likely to need.

An online service also suffers from this problem to some extent, but it has several advantages over using a thumb drive. Going online lets you access a document from several computers or mobile phones while also allowing you to modify that document from any of those devices and then make that modified version accessible.

LIVEDRIVE: Online backup services allow you to access your files wherever you are

You could create a document on your desktop at home, read and modify it on your laptop in a hotel room and then work on the updated version when you get back home, all without any manual moving of files.

A typical additional feature is that previous versions of files are saved, so you can roll back to an older copy if you've made a mistake or decided you don't like your changes.

File synchronisation is the tech behind all this functionality, and while it's very useful, you should be aware of its potential problems.

Say both you and a colleague start to edit the same file at the same time. When you close the file it's written to the online drive but if your colleague then does likewise, without special precautions, it would mean that your edit gets lost.

HUMYO: File locking is a must if you want to share the same file with different people

If you intend to use this service collaboratively, you'd be advised to check that it offers a suitable fix for such conflicts. The classic solution is to employ file locking so that as soon as one person starts to edit a file, subsequent access to that file won't be allowed until the first person has finished editing it.

File sharing

Although the main benefit of file synchronisation is the ability to work on files from several computers, another is the way it allows you to share those files between a small group of individuals who can also modify them.

SUGARSYNC: You can also set it up to allow read-only access to certain people you don't want to edit your files

Regular file sharing is also offered for distributing your files to a larger circle of contacts without providing them with the ability to modify the documents. Generally, these services allow you either to share an online file or folder with certain individuals or to make it public, which means that anyone can see it.

In essence, the end result is much the same: a URL is generated to permit access to your shared files. The only difference is that if you decide to share it only with certain individuals then an email is sent to them containing a link, whereas if you make it public then it's up to you to publicise that link for the world to see.

Automated backup

You don't need an online service to back up your data, but it does make the process easier and more secure. Local back-up software lets you perform backups automatically at preset times, but unless you've made sure your back-up media is connected and ready to go, that backup will fail.

DROPBOX: They also provide a good alternative to home backups that will be lost in the event of something drastic occurring

To prevent this, you could keep your back-up drive attached to your PC. This would protect against a failure of your main disk, but if your PC is stolen or destroyed in a fire, an external drive sitting next to it would likely suffer the same fate.

Ideally you need a method of backup that is both automatic and remote. Online backup provides just that. Online backup services often use client software installed locally that allows you to mark files or folders for backup. Then whenever a file is modified, the new version will be automatically uploaded to your secure storage space via the web.

The only thing that could prevent a file from not being backed up is the lack of web connection – but the transfer will take place as soon as web access is restored, rendering it a temporary problem.

All the services here provide the three basic services described above, and all offer encryption. Differences mostly come down to personal preference on matters such as which user interface you prefer, which facilities you prioritise and, of course, the price you're willing to pay.

With 500GB disks now shipping as standard on all but entry-level PCs and 2TB disks costing less than £130, storage space has never been more abundant or affordable.

So the recent proliferation of companies offering online data storage products and services might seem surprising – except when you factor in the features and benefits online storage offers that could never be met by local storage alone.

Among the list of unique benefits are fully automated backups, synchronising files between different PCs and sharing files with colleagues, friends or even the whole world. Here we investigate these benefits and take a look at four products that claim to offer all of these services and more.

File synchronisation

USB flash drives enable us to carry large amounts of data around in our pockets. But in order to benefit from them we must be organised enough to make sure that they always contain the files we're likely to need.

An online service also suffers from this problem to some extent, but it has several advantages over using a thumb drive. Going online lets you access a document from several computers or mobile phones while also allowing you to modify that document from any of those devices and then make that modified version accessible.

LIVEDRIVE: Online backup services allow you to access your files wherever you are

You could create a document on your desktop at home, read and modify it on your laptop in a hotel room and then work on the updated version when you get back home, all without any manual moving of files.

A typical additional feature is that previous versions of files are saved, so you can roll back to an older copy if you've made a mistake or decided you don't like your changes.

File synchronisation is the tech behind all this functionality, and while it's very useful, you should be aware of its potential problems.

Say both you and a colleague start to edit the same file at the same time. When you close the file it's written to the online drive but if your colleague then does likewise, without special precautions, it would mean that your edit gets lost.

HUMYO: File locking is a must if you want to share the same file with different people

If you intend to use this service collaboratively, you'd be advised to check that it offers a suitable fix for such conflicts. The classic solution is to employ file locking so that as soon as one person starts to edit a file, subsequent access to that file won't be allowed until the first person has finished editing it.

File sharing

Although the main benefit of file synchronisation is the ability to work on files from several computers, another is the way it allows you to share those files between a small group of individuals who can also modify them.

SUGARSYNC: You can also set it up to allow read-only access to certain people you dont want to edit your files

Regular file sharing is also offered for distributing your files to a larger circle of contacts without providing them with the ability to modify the documents. Generally, these services allow you either to share an online file or folder with certain individuals or to make it public, which means that anyone can see it.

In essence, the end result is much the same: a URL is generated to permit access to your shared files. The only difference is that if you decide to share it only with certain individuals then an email is sent to them containing a link, whereas if you make it public then it's up to you to publicise that link for the world to see.

Automated backup

You don't need an online service to back up your data, but it does make the process easier and more secure. Local back-up software lets you perform backups automatically at preset times, but unless you've made sure your back-up media is connected and ready to go, that backup will fail.

DROPBOX: They also provide a good alternative to home backups that will be lost in the event of something drastic occuring

To prevent this, you could keep your back-up drive attached to your PC. This would protect against a failure of your main disk, but if your PC is stolen or destroyed in a fire, an external drive sitting next to it would likely suffer the same fate.

Ideally you need a method of backup that is both automatic and remote. Online backup provides just that. Online backup services often use client software installed locally that allows you to mark files or folders for backup. Then whenever a file is modified, the new version will be automatically uploaded to your secure storage space via the web.

The only thing that could prevent a file from not being backed up is the lack of web connection – but the transfer will take place as soon as web access is restored, rendering it a temporary problem.

All the services here provide the three basic services described above, and all offer encryption. Differences mostly come down to personal preference on matters such as which user interface you prefer, which facilities you prioritise and, of course, the price you're willing to pay.

Data loss: we've all experienced it. Maybe you emptied the Recycle Bin milliseconds before realising that you'd deleted the wrong folder the day before; or perhaps your hard disk simply packed up, leaving behind it nothing more than an odd clicking sound and a system error screen.

It might have been that the dog really did eat your homework by mistaking your newly burnt DVD for its latest toy. Whatever happened in your case, it resulted in your precious data ascending to that great filesystem in the sky.

Fortunately, dead disks, trigger-happy fingers and scratched CDs don't always mean that you have to wave goodbye to your data. While not every problem has a happy ending, there are many that do. So when disaster strikes, don't reach for the proverbial pistol: read our guide to recovering data instead.

Lost files

Disaster rating: 1/5

Losing a file is a disconcerting experience. If you're looking for a document but can't find it anywhere, stay calm. Stop any programs that are writing huge amounts of data to your hard disk and exit as many applications as possible.

Deleted files aren't really deleted: they're merely marked as disk space ripe for reuse. This means that you want to stop anything that's writing to disk in case it overwrites the lost data. When everything has been stopped, have another search for your lost file using Windows' Search utility.

If it's really not there, your first port of call should be the Recycle Bin. If it's not in there either, or you habitually use the more brutal [Shift]+[Delete] option, you should advance to the 'Truly deleted files' section below. However, some programs may have already come to your rescue.

Word and other Office applications, for example, store temporary versions of files as you're working on them. Finding and extracting usable data from these temporary files can be complex, but it can enable you to retrieve missing data. Visit here for a very comprehensive guide.

Truly deleted files

Disaster rating: 3/5

Although files deleted normally (and therefore simply sent to the Recycle Bin) are easily restored, those 'properly' deleted are trickier to recover. However, it's sometimes possible to resurrect these files using an undelete utility.

To understand how these tools work their apparent magic, you need to know how a filesystem works and what happens when you delete a file. Each disk (or each partition if your disk is divided into more than one partition) contains a system area that contains directory information about all of its files.

For each file, it contains the name of that file and the number of the first cluster (the smallest usable area of a disk) in that file.

Another important system area is the File Allocation Table (FAT), which contains information about all the other clusters associated with a file. The FAT entry for a file's first cluster is the number of the second cluster; in the entry for the second cluster is the number of the third cluster and so on until the entry for the last cluster in the file, which is an end-of-file indicator.

When a file is deleted, Windows does two things. First, it overwrites the first character of the filename in the directory with a question mark to indicate that the entry in the directory can be reused. Second, it overwrites the entries for all the file's clusters in the FAT with zeros to indicate that those clusters are free to be reused. None of the data in the file is actually overwritten or erased, so, if you move quickly, it may be possible to recover the file.

Undelete utilities start by looking in the directory for any filenames that start with a question mark. If the user opts to undelete any such file, the utility goes to the first cluster, as shown in the directory, and reads data from that and subsequent clusters until an end-of-file marker is found.

Note that this method will work only if a file is sequential – if it's fragmented then it can't be recovered since the information in the file allocation table that is needed to find nonsequential clusters will have been overwritten.

It's also important to recognise that although it's possible to successfully recover a file if you act immediately, the longer you leave it the more likely it is that Windows will have overwritten one or more clusters.

Corrupted filesystem

Disaster rating: 3/5

If your PC is shut down improperly, perhaps due to a power cut or system failure while Windows is in the process of writing to the system areas of the disk, the filesystem could become corrupted. This could result in data being present on your disk that Windows has no knowledge of – not an ideal situation to be in.

To understand exactly how this is corrected, we'd have to get into the intricacies of the filesystem. To cut a long story short, let's say that software is used to analyse the filesystem and spot inconsistencies.

Having found some, and based on information regarding the most likely ways in which a filesystem can become corrupted, the software attempts to rebuild the filesystem so that Windows can access those lost files. The technique is intended to recover data, not mend Windows and its intricate data structures.

When your data has been rescued, you'll need to reformat the disc and reinstall Windows. Once it's found the all-important inconsistencies, the software will attempt to figure out where the lost files are, based on information regarding the most likely ways in which a filesystem can become corrupted. Once identified, the lost files can be restored to a separate drive.

If you want to have a go, try GetDataBack ($79) or R-Studio NTFS ($50). Sometimes this process – and the undelete process described above – isn't successful even though the data is still present on the disk. In these cases, you can recover files by analysing clusters to determine what type of data they contain.

This takes a fair amount of research into the clusters of different types of files. However, if your disk isn't particularly fragmented, it could be easier to reassemble a lost fi le than you might think.

Mechanical failure

Disaster rating: 4/5

A hard disk drive consists of a spindle motor, a voice coil motor, a read/write head, a circuit board and a platter, of which only the latter stores any data. So if your drive fails, it's possible that the data is still present on the platter and the problem is due to a failure of one of the other components. Even an evidently mechanical sound might be nothing more sinister than a failure of the servo circuitry.

If something other than the platter has failed, then the solution is obvious – replace the offending part. You might be tempted to try this yourself, but be warned: it's tricky.

Firstly, unless you're an expert you won't know for sure which part needs replacing. Secondly, attempting the operation in anything other than a spotless room is doomed to failure. Finally, you'd have to buy an exactly identical disk drive from which to salvage the parts. Thus we highly recommend employing the services of a data recovery company.

This isn't a cheap option, so you'll have to decide whether your lost data is worth it. The good news is that some companies offer a 'no fix – no fee' guarantee on repairs, or will first carry out a diagnosis and then provide you with a report that will specify how much data they can be sure of salvaging, should you accept the quotation.

If you want to attempt this, try contacting Kroll Ontrack Data Recovery, MjM Data Recovery, Data Clinic or Xytron.

Overwritten files

Disaster rating: 5/5

If your file has been truly deleted and then overwritten, the sad news is that you probably won't be able to retrieve it. However, there are two theoretical methods that seem promising, so perhaps the situation will be more hopeful in the future.

The first method is to do with variations in magnetic flux. When data is written to disk, the resultant magnetic flux depends mostly on the value written. However, the flux is also provided with a tiny contribution from the overwritten data.

So, if you replace the normal read electronics that decide whether a bit is a 1 or a 0 with circuitry that can extract the analogue value from the head, subtracting the known contribution of the most recently written data should make it possible to determine what value was in place before it.

This method of extracting overwritten data is commonly reported and academic papers have been written on the subject, but we've been unable to find any organisation that claims to have done it successfully. The problem is that the signal to the previously written data is so small that it effectively gets lost in the random electrical fluctuations commonly referred to as noise.

However, Western Digital's Gerardo Bertero – while questioning whether the technique is really a practical proposition – did suggest one possible solution. By reading each bit thousands if not millions of times and averaging all those signals, electrical noise, being random in nature, would average out to zero – whereas the genuine signal would build up and become visible.

The snag is that this would be hugely time-consuming and costly – which is why nobody offers such a service. Whether it becomes feasible when national security is at risk is another question entirely, and one we're not likely to get an answer to from the Secret Service.

The second theoretical technique for recovering data is to use a Magnetic Force Microscope (MFM). It's claimed that this method offers an additional benefit – it can read data, overwritten or not, from the surface of a platter that is no longer able to spin because of damage.

For the normal read operation of a hard disk, the platter must be able to rotate because an electrical signal can only be produced in the read head's coil if it's moving with respect to the magnetic field. However, an MFM is able to read a static magnetic field – so the platter doesn't have to be able to spin.

An MFM is a laboratory instrument that has an ultra-fine magnetised tip suspended on a cantilever. As the tip moves over the surface of the object being imaged, the magnetic field of that object exerts a force on the tip, which in turn causes a displacement of the cantilever. This is measured using optical techniques.

It's self-evident how an MFM could be used to read data from a platter that can no longer rotate, but what isn't as obvious is how the technique lends itself to recovering data that's been overwritten – at least in theory.

Data is written to a hard disk's platter in concentric circles known as tracks. A highly accurate servo control system is used to position the read/write head over the required track, but, even so, the head isn't always positioned to exactly the same radial position. What this means is that if a track is overwritten but the head wasn't at exactly the same position as it was on the previous occasion, a narrow band of the previously written data might remain intact at the edge of the track.

Again, despite hearing so much about this technique, and despite the fact that hard disk drive manufacturers already use MFMs for research and development, we haven't found any company who offers this technique for recovering overwritten data. We'd have to speculate on whether it's in use by the military, although the fact that military standards require disks to be physically destroyed at the end of their lives might just suggest that they know it's feasible.

Scratched discs

Disaster rating: 1/5

Methods such as undeleting files, repairing logical errors to the file structure and reassembling files also work with media such as memory cards and pen drives. However, there are certain methods of data recovery that apply only to optical disks. Except in the most extreme cases, when a CD or DVD won't read it's usually scratches in the outer plastic layer that are causing the problem.

The plastic layer is there to provide protection to the layer of data underneath, but scratches can impair the passage of the light used to read the data. The solution is to remove the troublesome scratches using a mild abrasive so that light will pass through the plastic layer without obstruction.

Although there are reports of this being done successfully using household substances such as Brasso, a more reliable solution is to use a product designed for this job. Such products range from kits comprising a bottle of suitably mild abrasive and a lint-free cloth to mechanical disk polishing devices such as Digital Innovations' SkipDr. The latter claims a greater chance of success since the polishing process is more uniform and controlled.

Shattered discs
Disaster rating: 4/5

Researchers at the University of Arizona's Optical Sciences Center have published a peer-reviewed paper in which they describe how they used an optical microscope to extract data from fragments of broken CDs or DVDs.

However, the process is time-consuming and so isn't likely to be viable unless the rewards are huge. What's more, any data in the vicinity of the breaks is totally unrecoverable, so in many cases we'd be talking of recovering fragments of data rather than files in their entirety. A shattered disc still means lost data – although the future may hold more hope for broken optical discs.

The fact that silicon chips start life as nothing more exotic than sand is amazing enough, but have you ever thought about that other important PC component, the hard disk?

Its origins couldn't be more different. The heart of a hard disk – the rotating platter where your data is stored – is made out of an exotic mix of elements including ruthenium and platinum, two of the world's rarest and most expensive metals.

Needless to say, this statement doesn't even hint at the complexity involved in transforming rare ores into gigabytes of data storage. The hard disk's high speeds of rotation and the close proximity of the head to the platter means that the processes must be carried out with the ultimate in precision and cleanliness.

Add to this the strange properties of magnetic media and the techniques required to achieve the optimum capacity, and the story of how disks are made becomes one that encompasses the fields of mining, metallurgy, chemistry, physics and involves the pinnacle of engineering and manufacturing technology.

As a whole, a hard disk is an amazing feat of electronic and mechanical engineering, but two parts – the heads and the platter – stand out for their sheer manufacturing complexity. As the part that actually stores the data, the platter is what many people consider the heart of a hard disk drive – and here we reveal the secrets of its manufacture.

Step 1: Mineral extraction and processing

Platinum is only the 70th most abundant element in the Earth's crust, making up just three parts per billion. Ruthenium comes two places lower with an abundance of only one part per billion. By way of comparison, silicon – the raw material from which microprocessors are made – accounts for around 27 per cent of the Earth's crust.

It's no surprise then that platinum is hugely expensive – today's market price is more than $1,300 per Troy ounce. Turning to ruthenium, the total annual production is just 27 tonnes, an amount that would fit in a 1.3m3 cube. Both are mined predominantly in South Africa.

Platinum is one of the noble metals, which means that it's relatively unreactive. Unlike metals such as copper – the main ores of which are compounds – platinum is normally found in its metallic form. This doesn't mean that extracting it from its ore is simple, though, as platinum is normally found mixed with other metals.

Obtaining pure platinum involves separating it from the iron, copper, gold, nickel, iridium, palladium, rhodium, ruthenium and osmium that it's invariably found with. Let's just say it's a complicated multistage chemical process that can take up to six months to complete. Fortuitously, though, the ruthenium that's also needed in disk manufacture is a by-product of the process.

A deep mine in the Bushveld Complex of South Africa might seem far-removed from a finished hard disk, and in this sense it's an ideal place to start our investigation. But we're not going to need the platinum or the ruthenium until well down the line, so for now we'll put them aside as we move to something more down to earth – and considerably more common.

Step 2: Making aluminium blanks

The manufacture of a hard disk platter starts with the fabrication of aluminium blanks, which are disks of aluminium alloy onto which the magnetic recording layer will eventually be deposited.

High-purity alloy that contains four to five per cent magnesium plus small amounts of silicon, copper, iron and zinc to give it the necessary properties is cast into an ingot weighing seven tonnes. The ingot is then heat-treated, hot-rolled and cold-rolled in multiple passes to provide a sheet of the necessary thickness (usually 0.635mm, 0.8mm, 1.0mm, 1.27mm, 1.5mm or 1.8 mm – just enough to provide adequate stability while rotating at high speed) from which the blanks will be punched.

GETTING STARTED: Hot rolling mills process aluminium ingots into thin slivers of metal from which disks will be punched

Punching takes place once the alloy sheet has been coiled into large rolls so that a single stamping process produces lots of blanks. This is then followed by a stacked annealing process to reflatten the blanks. Finally the blanks are ground to a high level of precision to achieve the necessary surface and edge finish. Bear in mind that this and all subsequent steps are carried out on both sides of the platter so that it ends up with two recording surfaces.

Step 3: NiP plating

The aluminium blanks are now precision-ground using 'stones' that are composed of PVA and which contain silicon carbide as the abrasive agent. However, even with all the care taken to produce a good finish, the surfaces of the aluminium blanks produced in Step 2 are not yet nearly perfect enough. Because there's a limit to the degree of smoothness to which aluminium alloy can be ground, the next step is to apply a hard coating that will take a better finish.

PERFECT FINISH: The soft aluminium is plated with a hard NiP layer so that it can be polished to an incredible degree of smoothness

This hard coating is an amorphous alloy of nickel and phosphorous (NiP). It's applied by an electroless process in which complex supersaturated solutions containing compounds of nickel and phosphorous react on the surface of the disk to leave the required NiP layer. This layer can now be further refined in the next step of the process.

Step 4: Precision polishing

After NiP plating, the substrate is polished in several steps using progressively finer abrasives based mostly on silicon carbide, diamond and aluminium oxide. The end result is a disk that has a roughness of less than 1Å (an Angstrom unit – 0.1nm, 0.0001μm or 0.0000001mm), which is about the size of an atom and 450 times less than the minimum size of the features in today's microprocessors.

Subsequent processes in the following steps increase the roughness to 4Å, the minimum level of surface flatness that will allow the head to fly reliably over the surface of the media with a controlled spacing of around 2nm.

Step 5: Washing and inspecting

Some manufacturers employ a conditioning step to remove any contamination that may be still present on the substrate. This involves spinning the disk and then very gently pressing a barely abrasive tape onto the surface. Then, before the magnetic data recording layers are applied, the disk is cleaned so that it's free of any particles, scratches or contaminants. This is done using wet chemical exposures to acidic and alkaline solutions, followed by mechanical scrubbing in soapy solutions and then multiple rinses in deionised water.

SHINE UP: Before the active layers are deposited on the platter, it's polished so that any roughness is within atomic dimensions

The disk is dried using a surface tension effect. Before continuing, advanced optical inspection is used to detect particles, contaminants or scratches, and any disks with such defects are rejected. The process is fully automated using optics and electronic detectors combined with smart software to identify imperfections.

Step 6: Applying a soft magnetic underlayer

The next few steps involve depositing layers of various materials with differing magnetic properties using a process called 'Sputtering' that takes place in a multi-chamber vacuum deposition tool.

The first of these layers is the soft magnetic underlayer. Otherwise known as the magnetic keeper layer, it's a good conductor of magnetic fields. This layer is unique to Perpendicular Magnetic Recording technology (see 'From LMR to PMR, overleaf) and has the result of enhancing the perpendicular field needed for writing by providing an 'image field' to the field produced by the head. The soft magnetic underlayer is made from an alloy, typically containing cobalt, nickel and iron.

DATA-STORAGE LAYERS: The application of the soft magnetic underlayer is just one of several steps that are carried out as the platter is automatically passed from one chamber to another in a vacuum deposition tool

In Western Digital's latest platters this layer takes the form of two sub-layers separated by a four-atom thick layer of ruthenium. When two ferromagnetic layers are separated by a thin layer of ruthenium, the resulting interaction between the two layers is such that energy is minimised when the magnetisation between those layers is opposite. This is known as a synthetic antiferromagnet, and the end result is a keeper layer with properties that can be finely tuned. Only a few elements are known to do this, and ruthenium has the largest effect – which is why it's used in modern hard disks.

Step 7: Adding the data storage layers

Now we come to the data-storage layers. These are made from an alloy of cobalt, chromium and platinum (CoCrPt). Cobalt is used because it has a hexagonal crystal structure, which is less symmetrical than the cubic crystal structure of other magnetic metals (such as iron and nickel). This allows the metal's crystals to be oriented in the preferred magnetisation direction, which in the case of PMR is up or down. Chromium is added to give the cobalt resistance to corrosion and reduce the interactions between grains with a consequential improvement in the signal-to-noise ratio.

Lastly, the platinum provides thermal stability, preventing data loss if the disk is subjected to external magnetic fields or heat. As with the two sub-layers that form the soft magnetic underlayer, the recording layer is composed of several sub-layers. Often thin layers of ruthenium separate these. Ruthenium also separates the soft magnetic underlayer from the recording layer, but here it performs a quite different function. Ruthenium has a hexagonal close-packed atomic structure similar to that of the CoCrPt alloy, so it's used as a nucleation layer to help orient the crystals of the magnetic grains in the required direction.

It's also used to lower the degree of magnetic exchange coupling between the hard magnetic layers to produce advanced structures such as the widely used exchange coupled composite (ECC) structures. ECCs are used to help solve the 'trilema' in which attempts to improve any of the main requirements – thermal degradation, ease of magnetic switching and signal-to-noise ration – makes the others worse.

Step 8: Adding a protective overcoat

The final stage of the deposition process is to apply a diamond-like carbon overcoat layer to provide corrosion resistance and improve its mechanical reliability. This protective layer is typically 2nm thick and is applied by ion-beam or plasma-enhanced chemical vapour deposition techniques. The platter is now removed from the sputter deposition chamber.

Step 9: Lubricating the platter

Next, a lubricant layer is applied to the media in one or more steps depending on design. Typically the lubricant is dissolved in a solvent and applied to the platter by pulling it at a controlled rate. The rate of evaporation of the solvent in the meniscus that forms at the liquid air interface during the pulling process and the concentration of lubricant in the solution determine the resulting thickness on the disk, which is approximately 1nm. The layer comprises advanced perfluoropolyether lubricants combined with phosphazene additives that inhibit degradation of the lubricant.

Typically the lubricant layer is partially bonded to the overcoat film and imparts durability to the head media interface system in a drive. The bonding process can be activated thermally or, more typically, by exposure to ultraviolet light. During the bonding process, cross-link chemical bonds form in the lubricant's molecular chains to limit the mobility of the lubricant. However, the top-most portion of the lubricant is left to be fully mobile.

After lubrication, a tape burnish process and then a head burnish process are used to wear out asperities (microscopic unevenness) and remove any loose particles that may remain on the surface of the platter after the sputter and lubrication processes have been completed.

Step 10: Testing and certification

The final step before the platter can take its place in a disk drive is to certify that it can pass what is referred to as a glide test. During the glide process a specially made head is 'glided' over the surface of the platter to detect any remaining asperity on the media. This process ensures that a head will be able to fl y over the surface of the disk without crashing into any projections.

FINAL TEST: The platter has to pass a glide test to make sure that the head won't crash into surface defects

If the platter passes this last step then it's deemed 'flyable' or 'prime' and after a magnetic conditioning step it's appropriately packed up and shipped to the drive factory.

The magnetic conditioning step involves exposure of the finished media to a large magnetic field in order to leave the magnetisation in the storage layer in a uniform state that will not interfere with the drive manufacturing process.

Think 'IT revolution' and you're thinking of the second half of the 20th century – right?

From the first stored-program computer in 1948 to the blossoming internet of the late '90s, it seems obvious that it's the most recent half-century that transformed computing from an expensive curiosity for the few to a life-changing experience for the many.

So you might be surprised to hear that much of the pioneering work that made all this possible was carried out during the reign of Queen Victoria.

Boolean logic, programming languages, data transmission, radio communication, universal computation, data compression – these building blocks of the IT revolution have their roots in Victorian society.

The Analytical Engine

Given that he has a track record of never completing any of his inventions, the British mathematician Charles Babbage provides an unlikely starting point for our investigation into Victorian computer pioneers.

Babbage's first foray into computation involved the design of the so-called Difference Engine, which was intended to calculate polynomial functions for navigation and artillery applications. It was never finished in Babbage's lifetime, but the successful creation of a machine built to his original plans by the London Science Museum in 1991 vindicated the design.

Impressive as it may be for a machine weighing almost five tons and comprising 8,000 parts to work without a glitch, it's Babbage's second contrivance, the Analytical Engine, that really makes things interesting.

While the Difference Engine was dedicated to one type of calculation, the Analytical Engine was designed to be universal, just like today's computers. Except for the fact that it relied on mechanics rather than electronics, the similarities are striking for something conceived of in 1837, 121 years before the first electronic stored-program computer.

Like today's PCs, the Analytical Engine used a sequence of instructions to process data. Both the program and the data were input using punch cards similar to those used at the time to control looms in wool mills (and used in mainframe computers until the 1970s). Results could be output to a printer, a graph plotter or more punch cards so that they could be fed back into the Engine.

In a direct parallel with modern computers, it had a memory that Babbage called the 'store', which had a capacity of 1,000 50-digit decimal numbers. It also had an arithmetic unit that he called the 'mill', which was capable of addition, subtraction, multiplication, division and comparison.

It was also capable of looping and conditional branching, although it seems probable that the importance of this hadn't been fully appreciated by Babbage until he made the acquaintance of Ada Lovelace, as we'll see shortly. Fascinating as the similarities with today's technology are, the differences also make interesting reading.

The Analytical Engine was to have had a steam engine as its power source. It would have carried out additions and subtractions in about a second, but could have taken up to a minute to perform division and multiplication. Speedy it certainly wasn't.

Ada Lovelace's computer program

Augusta Ada, Countess of Lovelace and daughter of the poet Lord Byron, didn't fit into the mould of Victorian society. Instead of excelling in needlework, embroidery and entertaining on the pianoforte, Ada's skills were in the realm of science and mathematics.

She was introduced to Charles Babbage at a dinner party in 1833 and they corresponded for several years, discussing first the Difference and later the Analytical Engine.

In 1942, the Italian mathematician Luigi Menabrea, whom Babbage had met a year earlier, wrote a paper entitled A Sketch of the Analytical Engine Invented by Charles Babbage. Ada Lovelace translated the article into English and, at Babbage's request, expanded it with very extensive notes of her own. Much impressed with her understanding of his creation, Babbage referred to her as the 'Enchantress of Numbers'.

But Countess Lovelace's greatest contribution to the science of computation was an example that she provided in her notes to Menabrea's article of how the Analytical Engine could be used to calculate Bernoulli numbers.

Unless you're a mathematician you probably won't be too interested in exactly what they are, so let's just say that this sequence of numbers, discovered by the Swiss mathematician Jakob Bernoulli, is of significant interest in number theory. What was of particular interest to Ada Lovelace is that they're notoriously difficult to calculate.

Each successive number requires significantly more calculations than its predecessor, and in fact Bernoulli himself only managed to work out the first 10 of the numbers that bear his name.

Ada's instructions for the Analytical Engine, while not looking like a modern computer program, are considered to be just that – the world's very first example. They contain many of the elements of today's programs, including conditional branches and nested loops, or 'cycle of cycles' as she called them.

Boolean logic

Born in Lincoln in 1815, George Boole devised a form of logic that operates on just two values which can alternatively be thought of as true or false, or – more pertinently to our discussion of computing – 1 or 0.

Boolean logic defines various ways in which these values can be manipulated and combined. Examples include the AND function and the OR function, both of which take two inputs and then produce a single output. With the AND function, the output is a 1 only if both the inputs are 1s; whereas the OR function produces an output of 1 if either or both of the inputs are 1s.

The apparent simplicity of these functions doesn't do justice to their power. By combining the simple electronic building blocks that implement these functions (which are referred to as AND gates and OR gates), it's possible to create flip-flops, adders, shift registers and many more of the constituents of a computer that can work on binary numbers. As such, Boole had laid the theoretical foundations for today's computers.

Of course, it's a perfectly valid question to ask what would have been wrong with computing with decimal numbers, as Babbage's Analytical Engine was designed to do. A look at one of the earliest electronic computers, the University of Pennsylvania's ENIAC, provides just a glimpse of the advantages offered by moving to binary.

ENIAC handled decimal numbers, storing each digit in an electronic circuit called a ring counter that contained 36 valves. As 10 of these ring counters constituted a register, it took 360 valves to store a number in the range -9,999,999,999 to +9,999,999,999. By way of contrast, a 32-bit binary register can store numbers in the range –2,147,483,648 to + 2,147,483,647.

Using similar electronic circuits to those used in ENIAC, a 32-bit register required just 64 valves (70 for 35 bits) or, in today's terms, 64 transistors.

Data transmission

Today, data processing goes hand-in-hand with data communication, but to see the first developments in this technology we need to cross the Atlantic. Predating the telephone by more than 30 years, the telegraph is often considered the poor relation.

This undervalues the pioneering work of Samuel Morse, who first demonstrated the code that bears his name back in 1844. Morse Code uses short and long signals (known colloquially as dots and dashes) interspaced with gaps of varying lengths to represent letters, numbers and a range of symbols.

It's really not too different from ASCII (American Standard Code for Information Interchange), which is the code used in current-day data transmission. It was designed to be sent by hand using a finely balanced switch called a Morse key and received by ear or as marks on a paper strip, but it's also possible to use Morse Code for automatic data transmission by computer.

In providing a means of automatic data transmission it achieved what has only become possible in recent times (and then not perfectly) through voice communication. It's commonly assumed that Morse Code was designed arbitrarily and that it was just by chance that the code for E, for example, is 'dot' whereas that for J is 'dot dash dash dash'. This does Samuel Morse a great disservice, as we'll see if we fast-forward over a hundred years to 1952.

One of the first methods of data compression – and now a widespread and essential technology in areas as diverse as data communications, photography and music reproduction – was Huffman Encoding. This analyses the data stream to determine how frequently each character occurs and then assigns codes of variable length, with shorter codes going to the most commonly encountered characters.

Although infrequently encountered letters end up being represented by longer codes than in non-compressed text, the short codes used for the common letters more than compensate, so the compressed text can be as little as half the size of the original.

Going back to Morse Code, the two most common letters in the English language, E and T, are represented by 'dot' and 'dash' whereas the two least common ones, Q and Z, are represented by 'dash dash dot dash' and 'dash dash dot dot'.

So not only is Morse Code the world's first system for data transmission, it also has a built-in method of data compression.

Radio communication

Moving on from Morse's telegraph lines to the wireless data transmission that's so familiar to us today, we need to return to this side of the Atlantic.

Born in Bologna, Italy, Guglielmo Marconi moved to England when the Italian government failed to invest in his work, which involved experimenting with the electromagnetic waves that were first discovered by Heinrich Hertz.

But while the German physicist had a theoretical interest in what would eventually be called radio, Marconi's interest was much more practical in nature. The history of Marconi's development of the wireless telegraph was one of increasing the transmission range step by step.

In 1897, in a demonstration to the British government, he transmitted a signal over a distance of 6km on Salisbury Plain. Later in the same year he demonstrated that radio waves could travel over the sea, first at a range of 6km and then over 19km.

During 1899 Marconi first achieved communication between Britain and France, and later equipped three ships of the Royal Navy with radio equipment allowing them to communicate over a distance of 137km.

Marconi's equipment might have transmitted and received radio signals, but the hardware bore no resemblance to the equipment that does the same job today. With no electronic devices such as valves or transistors, Marconi resorted to the brute-force approach.

To generate the signal he used a spark transmitter that applied a high voltage across an air gap to produce a lightning-like spark. The result was broadband electromagnetic radiation from ultraviolet through visible and into the radio spectrum.

His receiver used a coherer, a glass envelope containing iron filings that coalesced under the influence of radio waves thereby allowing an electrical current to flow. The result of using such crude equipment was that the transmitter had to use very high power levels and the antennae were huge.

The station in Cornwall that Marconi used to achieve the first ever trans-Atlantic transmission demonstrated this. The transmitter consumed 25kW – several thousand times more than a mobile phone – and the antenna was supported by four 66m masts. Crude it might have been, but it worked – and the world suddenly became a lot smaller.

The missing link

Many of the building blocks of the modern world might have come into place during the reign of Queen Victoria, but that doesn't necessarily mean that the information age could have been born back in the 19th century.

Babbage was a forward thinker but, hampered by the technology of the time, his Analytical Engine could only ever have calculated anything at a snail's pace. His creation couldn't have possibly made a dent in what we expect of today's computers.

The missing link – the one that did indeed act as the catalyst for the digital age half a century later – only saw the light of day after Queen Victoria's death. In 1904, the English physicist John Ambrose Fleming discovered that a glass envelope from which the air had been extracted would permit the flow of electricity between a heated cathode and an anode in one direction only. Today we'd call his creation a diode valve.

Three years later, American inventor Lee de Forest discovered that by placing a coil of wire between the anode and the cathode, the current flowing between them could be influenced by the application of a small voltage. This contributed significantly to radio communication. In allowing a voltage on one circuit to control what was happening on another circuit, it was also the first electronic switch.

The triode – as de Forest's invention was called – could form the basis of Boolean logic gates and hence of an electronic computer. So much of what was needed for a computing revolution was therefore in place, but for the lack of that final piece of the jigsaw.

As it happened, we had to wait until the late 1940s for progress to continue.

Computers might struggle to exhibit intelligent behaviour, but blindly performing arithmetic calculations is surely their forte. Or is it?

The failure of Google's online calculator and Excel's apparent inability to give correct answers to simple calculations are both well-known problems among programmers, but these aren't really bugs in the normal sense of the word. Instead they're just a consequence of the fact that computers suck at maths.

Computers perform calculations in quite a different way from the methods that humans use to do arithmetic – and that means that they habitually come up with the wrong answer. Here we investigate some of the shocking consequences of this revelation before delving into the reason why computers suck at maths.

• 10 computing conspiracy theories examined

Close isn't close enough

For anyone still to be convinced that computers can't get simple arithmetic right, let's start off with a few examples that you can try out yourself.

First up, Google's calculator. If you've never tried it out before, to get a feel for how it works, surf to www.google.co.uk, type 5*9+(sqrt 9)^3 into the search box and click on 'Search'. You'll find that it comes back with the correct answer: '5 * 9 + (sqrt 9) ^3 = 72'.

Now let's try another calculation. Type in 599,999,999,999,999 - 599,999,999,999,998. Quite clearly, this should give an answer of 1. Unbelievably, however, Google responds with this: '599,999,999,999,999 – 599,999,999,999,998 = 0'. Just a rare and unfortunate example, perhaps?

OK then, let's try another simple calculation. Type =850*77.1 into cell A1 of an Excel 2007 workbook (it doesn't work – or should that be it does work – in earlier versions of Excel). A bit of mental arithmetic suggests that the answer ought to be in the region of 60,000; in fact the correct answer is 65,535.

Excel has other ideas. It will tell you that the result of this multiplication is 100,000, which is out by a massive 34,465. And to prove that this is no flash in the pan, how about using a selection of online calculators to work out 1.0 - 0.9 - 0.1?

You'll probably find at least half of them will come up with an answer of -2.77555756 E-17 – scientific notation for -0.0000000000000000277555756. (If all the ones you try give the right answer, take a look at www.calculator.net.)

BAD MATHS: Since 1.0 - 0.9 - 0.1 equals 0, why are so many online calculators convinced that this value is the answer instead?

OK, this answer might not be far removed from the correct answer of 0, but why can't the calculator come up with the right answer – an answer that's blatantly obvious to anyone who is conversant with simple arithmetic?

How computers do maths

Although computers can handle integers (whole numbers), for general-purpose arithmetic they store numbers in floating point format because it's so much more efficient in memory use.

Let's take the double precision floating point representation as an example. It uses 64 bits to store each number and permits values from about -10308 to 10308 (minus and plus 1 followed by 308 zeros, respectively) to be stored. Furthermore, fractional values as small as plus or minus 10-308 (that's a decimal point followed by 307 zeros and then a 1) can be stored.

By way of contrast, if the same 64 bits were used to store integers, the range would be −9,223,372,036,854,775,808 to +9,223,372,036,854,775,807, and fractional values couldn't be represented.

The secret to this apparently amazing efficiency is approximation. Of those 64 bits, one represents the sign (so whether the value is positive or negative), 52 bits represent the mantissa (that's the actual numbers) and the remaining 11 bits represent the exponent (how many zeros there are or where the decimal point is).

So although a much greater range of numbers can be stored using floating point notation, the precision is actually less than can be achieved in integer format, since only 52 bits are available. In fact, 52 bits of binary information represents a 16-bit decimal number, so any values that differ only in their 17th decimal point will actually be seen as identical.

The situation with Google thinking that 599,999,999,999,999 - 599,999,999,999,998 equals 0 is similar, although it's evident that Google's calculator actually uses less than the normal 52 bits for the mantissa. That some calculators give a non-zero result to the calculation 1.0 - 0.9 - 0.1 might seem different since we appear to be nowhere close to the limit of 64-bit floating point arithmetic.

But that's forgetting one important fact – that computers work in binary. And although 0.1 might have only one significant digit in decimal, in binary notation the mantissa is a repeating sequence. This means that 0.1 can never be represented accurately in binary, no matter how many bits you use.

The discrepancy between the computed answer and the correct answer is often minute, and you might be inclined to dismiss this sort of error as insignificant. However, such errors can add up, and the consequences can be serious.

On 25 February 1991, three days before the end of the first Gulf War, an Iraqi Scud missile hit a US airfield in Dhahran, Saudi Arabia. 28 American soldiers were killed and more than 100 others were injured.

At the time, sensitive targets were supposed to be protected by the Patriot surface-to-air defence system, and one battery of Patriot missiles was assigned to the Dhahran facility – so it's pertinent to ask what exactly went wrong.

The answer is the system's tracking software, and the problem is not unrelated to our online calculator error. In order to avoid potentially costly false alarms, the Patriot's sophisticated radar must first detect an object that has the characteristics of a Scud missile and then detect it a second time in a position calculated by the system on the assumption that the first fix was genuinely a Scud. Only when this second fix provides a confirmation is a missile launched to intercept it.

The calculation of where to look for confirmation of an incoming missile requires knowledge of the system time, which is stored as the number of 0.1-second ticks since the system was started up. Unfortunately, 0.1 seconds cannot be expressed accurately as a binary number, so when it's shoehorned into a 24-bit register – as used in the Patriot system – it's out by a tiny amount. But all these tiny amounts add up.

At the time of the missile attack, the system had been running for about 100 hours, or 3,600,000 ticks to be more specific. Multiplying this count by the tiny error led to a total error of 0.3433 seconds, during which time the Scud missile would cover 687m.

The radar looked in the wrong place to receive a confirmation and saw no target. Accordingly no missile was launched to intercept the incoming Scud – and 28 people paid with their lives.

The processor that couldn't divide

Launched in March 1993, the Pentium was Intel's fifth generation of x86 processor. Unlike previous generations, in which at least some family members could only carry out arithmetic on whole numbers, all Pentiums had a floating point unit (FPU).

An FPU is a piece of built-in hardware for calculating floating point arithmetic. This gave the Pentium a massive speed advantage, since computers without an FPU-enabled processor had to carry out this sort of calculation using software routines that involved lots of integer operations. Unfortunately, it was a poisoned chalice for Intel.

In June 1994, shortly after taking delivery of a Pentium-based PC, Thomas Nicely – then Professor of Mathematics at Lynchburg College, Virginia – noticed that a program he had written was giving inconsistent results.

By running the same program on several machines, Professor Nicely tracked down the problem to the his new PC's Pentium processor and, in particular, to its FDIV (floating point division) instruction. Although it affected just a tiny proportion of floating point divisions, at its worst the error was really quite significant.

Dividing 4,195,835 by 3,145,727 gave an answer of 1.3337 – which represents an error in the fourth decimal place since the correct answer is actually 1.3338. The Pentium's FPU used something called the SRT algorithm to carry out floating point divisions.

Although there are simpler and more obvious ways of dividing one floating point number by another, the SRT algorithm gave a significant speed advantage over previous algorithms.

If you're not a mathematician you'll find a description of how SRT works totally impenetrable. However, let's just say that instead of working everything out using 'pure maths', it involved the use of a look-up table. The table contained 1,000 or so values, but due to a production error five of these values were missing.

Despite the fact that Intel's CEO Andy Grove reckoned that the average user would only see the problem every 27,000 years, IBM's estimate was once every 24 days – and as a result the company stopped shipping Pentium-based PCs. Intel eventually agreed to swap defective Pentiums for good ones.

Most people didn't take up the offer, but the delay caused technically minded users to make Intel the butt of their jokes. The following is typical. Q: How many Pentium designers does it take to change a light bulb? A: 1.99904274017, but that's close enough for non-technical people.

The errors we've seen so far have concerned floating point numbers where accuracy is lost if there's not enough bits to store the mantissa. OK, those errors can add up, but essentially they're just rounding errors, and the likelihood of not having enough bits to store the exponent is comparatively small given that the maximum values they can store are absolutely huge.

When integers are involved, the effect can actually be far more serious. A 64-bit integer can store a maximum positive value of 9,223,372,036,854,775,807. If you try adding 1 to an integer variable that already equals this maximum value, you don't just lose that extra value. Instead, the integer overflows.

In other words, as far as a computer working in 64-bit integer arithmetic is concerned, 9,223,372,036,854,775,807 + 1 = -9,223,372,036,854,775,808 (note the minus sign). Something very similar happened on-board the European Space Agency's Ariane V rocket on its maiden flight.

EXPENSIVE MISTAKE: Programmers call it an overflow; in reality it was bad maths that caused this $370million spacecraft to explode

In fact, the arithmetic operation in question – if you can call it that – was even simpler than adding 1. Instead, it just involved copying one number that had been stored in floating point format to another location that was defined as an integer – and a 16-bit integer at that (maximum positive value of 32,767).

Unfortunately, the number was already too large to fit in the integer location, and as a result it overflowed. The exact sequence of events that followed is pretty complex but, to cut a long story short, the end result was that the Ariane V became one of the most expensive fireworks in history.

Guarding against cock-ups

This run-through of some of computing's most astonishing mathematical cock-ups may have come as something of an eye-opener to you. If so, you're probably wondering whether tomorrow's computers can avoid making such elementary mistakes.

Surprisingly, perhaps – and with the exception of the Pentium floating point error, which was caused by a hardware glitch – all of the errors we've mentioned here could have been prevented. In that sense, they can all be thought of as software errors.

As an example, let's take that integer overflow on the Ariane V rocket. That an integer can overflow isn't an error on the part of the processor because it's the way it's supposed to work. But whenever an integer does overflow, the processor sets something called a flag that the program can interrogate.

In the case of the Ariane software, the program didn't check for an overflow; if it had done, corrective action could have been taken. Of course, there will always be a limit to how large an integer can be and how much precision a floating point number can have – and this depends on the processor. But all of today's computers are universal computing machines, which means that they can solve any problem involving logic and maths.

So if a processor's internal instructions can't operate on large enough integers or on floating point numbers with sufficient precision, it's always possible for the programmer to implement arithmetic routines that will.

There will be a trade off against speed, though, which is why this isn't usually done. However clever the software or however much memory you use to store a floating point number, the result of some divisions will never be accurate.

We've seen how 1 divided by 10 is an infinite string in binary, and, in the general case, a move to decimal arithmetic wouldn't help either: 1 divided by 10 can be stored accurately in decimal, but 1 divided by 3 equals 0.3333333… ad infinitum.

The bottom line is that whatever number base you choose, some divisions will produce results that can never be stored accurately as a finite number of digits. Even this isn't a show-stopper, though.

Remember how 1.0 - 0.9 - 0.1 often yields an inaccurate answer because of rounding errors even though we know, immediately, that the answer is 0? Well, it's quite possible to write software to store the result of a division as a rational number.

In other words, you don't actually do the division – you just store the two numbers. In subsequent arithmetic operations you handle the values as fractions, just as you were taught in school, and the result will be exact.

So computers might suck at maths, but there's always a solution available to circumvent their inherent weaknesses. And in that case, it's probably more accurate to say that computer programmers suck at maths – or at least some of them do.

Computers might struggle to exhibit intelligent behaviour, but blindly performing arithmetic calculations is surely their forte. Or is it?

The failure of Google's online calculator and Excel's apparent inability to give correct answers to simple calculations are both well-known problems among programmers, but these aren't really bugs in the normal sense of the word. Instead they're just a consequence of the fact that computers suck at maths.

Computers perform calculations in quite a different way from the methods that humans use to do arithmetic – and that means that they habitually come up with the wrong answer. Here we investigate some of the shocking consequences of this revelation before delving into the reason why computers suck at maths.

• 10 computing conspiracy theories examined

Close isn't close enough

For anyone still to be convinced that computers can't get simple arithmetic right, let's start off with a few examples that you can try out yourself.

First up, Google's calculator. If you've never tried it out before, to get a feel for how it works, surf to www.google.co.uk, type 5*9+(sqrt 9)^3 into the search box and click on 'Search'. You'll find that it comes back with the correct answer: '5 * 9 + (sqrt 9) ^3 = 72'.

Now let's try another calculation. Type in 599,999,999,999,999 - 599,999,999,999,998. Quite clearly, this should give an answer of 1. Unbelievably, however, Google responds with this: '599,999,999,999,999 – 599,999,999,999,998 = 0'. Just a rare and unfortunate example, perhaps?

OK then, let's try another simple calculation. Type =850*77.1 into cell A1 of an Excel 2007 workbook (it doesn't work – or should that be it does work – in earlier versions of Excel). A bit of mental arithmetic suggests that the answer ought to be in the region of 60,000; in fact the correct answer is 65,535.

Excel has other ideas. It will tell you that the result of this multiplication is 100,000, which is out by a massive 34,465. And to prove that this is no flash in the pan, how about using a selection of online calculators to work out 1.0 - 0.9 - 0.1?

You'll probably find at least half of them will come up with an answer of -2.77555756 E-17 – scientific notation for -0.0000000000000000277555756. (If all the ones you try give the right answer, take a look at www.calculator.net.)

BAD MATHS: Since 1.0 - 0.9 - 0.1 equals 0, why are so many online calculators convinced that this value is the answer instead?

OK, this answer might not be far removed from the correct answer of 0, but why can't the calculator come up with the right answer – an answer that's blatantly obvious to anyone who is conversant with simple arithmetic?

How computers do maths

Although computers can handle integers (whole numbers), for general-purpose arithmetic they store numbers in floating point format because it's so much more efficient in memory use.

Let's take the double precision floating point representation as an example. It uses 64 bits to store each number and permits values from about -10308 to 10308 (minus and plus 1 followed by 308 zeros, respectively) to be stored. Furthermore, fractional values as small as plus or minus 10-308 (that's a decimal point followed by 307 zeros and then a 1) can be stored.

By way of contrast, if the same 64 bits were used to store integers, the range would be −9,223,372,036,854,775,808 to +9,223,372,036,854,775,807, and fractional values couldn't be represented.

The secret to this apparently amazing efficiency is approximation. Of those 64 bits, one represents the sign (so whether the value is positive or negative), 52 bits represent the mantissa (that's the actual numbers) and the remaining 11 bits represent the exponent (how many zeros there are or where the decimal point is).

So although a much greater range of numbers can be stored using floating point notation, the precision is actually less than can be achieved in integer format, since only 52 bits are available. In fact, 52 bits of binary information represents a 16-bit decimal number, so any values that differ only in their 17th decimal point will actually be seen as identical.

The situation with Google thinking that 599,999,999,999,999 - 599,999,999,999,998 equals 0 is similar, although it's evident that Google's calculator actually uses less than the normal 52 bits for the mantissa. That some calculators give a non-zero result to the calculation 1.0 - 0.9 - 0.1 might seem different since we appear to be nowhere close to the limit of 64-bit floating point arithmetic.

But that's forgetting one important fact – that computers work in binary. And although 0.1 might have only one significant digit in decimal, in binary notation the mantissa is a repeating sequence. This means that 0.1 can never be represented accurately in binary, no matter how many bits you use.

The discrepancy between the computed answer and the correct answer is often minute, and you might be inclined to dismiss this sort of error as insignificant. However, such errors can add up, and the consequences can be serious.

On 25 February 1991, three days before the end of the first Gulf War, an Iraqi Scud missile hit a US airfield in Dhahran, Saudi Arabia. 28 American soldiers were killed and more than 100 others were injured.

At the time, sensitive targets were supposed to be protected by the Patriot surface-to-air defence system, and one battery of Patriot missiles was assigned to the Dhahran facility – so it's pertinent to ask what exactly went wrong.

The answer is the system's tracking software, and the problem is not unrelated to our online calculator error. In order to avoid potentially costly false alarms, the Patriot's sophisticated radar must first detect an object that has the characteristics of a Scud missile and then detect it a second time in a position calculated by the system on the assumption that the first fix was genuinely a Scud. Only when this second fix provides a confirmation is a missile launched to intercept it.

The calculation of where to look for confirmation of an incoming missile requires knowledge of the system time, which is stored as the number of 0.1-second ticks since the system was started up. Unfortunately, 0.1 seconds cannot be expressed accurately as a binary number, so when it's shoehorned into a 24-bit register – as used in the Patriot system – it's out by a tiny amount. But all these tiny amounts add up.

At the time of the missile attack, the system had been running for about 100 hours, or 3,600,000 ticks to be more specific. Multiplying this count by the tiny error led to a total error of 0.3433 seconds, during which time the Scud missile would cover 687m.

The radar looked in the wrong place to receive a confirmation and saw no target. Accordingly no missile was launched to intercept the incoming Scud – and 28 people paid with their lives.

The processor that couldn't divide

Launched in March 1993, the Pentium was Intel's fifth generation of x86 processor. Unlike previous generations, in which at least some family members could only carry out arithmetic on whole numbers, all Pentiums had a floating point unit (FPU).

An FPU is a piece of built-in hardware for calculating floating point arithmetic. This gave the Pentium a massive speed advantage, since computers without an FPU-enabled processor had to carry out this sort of calculation using software routines that involved lots of integer operations. Unfortunately, it was a poisoned chalice for Intel.

In June 1994, shortly after taking delivery of a Pentium-based PC, Thomas Nicely – then Professor of Mathematics at Lynchburg College, Virginia – noticed that a program he had written was giving inconsistent results.

By running the same program on several machines, Professor Nicely tracked down the problem to the his new PC's Pentium processor and, in particular, to its FDIV (floating point division) instruction. Although it affected just a tiny proportion of floating point divisions, at its worst the error was really quite significant.

Dividing 4,195,835 by 3,145,727 gave an answer of 1.3337 – which represents an error in the fourth decimal place since the correct answer is actually 1.3338. The Pentium's FPU used something called the SRT algorithm to carry out floating point divisions.

Although there are simpler and more obvious ways of dividing one floating point number by another, the SRT algorithm gave a significant speed advantage over previous algorithms.

If you're not a mathematician you'll find a description of how SRT works totally impenetrable. However, let's just say that instead of working everything out using 'pure maths', it involved the use of a look-up table. The table contained 1,000 or so values, but due to a production error five of these values were missing.

Despite the fact that Intel's CEO Andy Grove reckoned that the average user would only see the problem every 27,000 years, IBM's estimate was once every 24 days – and as a result the company stopped shipping Pentium-based PCs. Intel eventually agreed to swap defective Pentiums for good ones.

Most people didn't take up the offer, but the delay caused technically minded users to make Intel the butt of their jokes. The following is typical. Q: How many Pentium designers does it take to change a light bulb? A: 1.99904274017, but that's close enough for non-technical people.

The errors we've seen so far have concerned floating point numbers where accuracy is lost if there's not enough bits to store the mantissa. OK, those errors can add up, but essentially they're just rounding errors, and the likelihood of not having enough bits to store the exponent is comparatively small given that the maximum values they can store are absolutely huge.

When integers are involved, the effect can actually be far more serious. A 64-bit integer can store a maximum positive value of 9,223,372,036,854,775,807. If you try adding 1 to an integer variable that already equals this maximum value, you don't just lose that extra value. Instead, the integer overflows.

In other words, as far as a computer working in 64-bit integer arithmetic is concerned, 9,223,372,036,854,775,807 + 1 = -9,223,372,036,854,775,808 (note the minus sign). Something very similar happened on-board the European Space Agency's Ariane V rocket on its maiden flight.

EXPENSIVE MISTAKE: Programmers call it an overflow; in reality it was bad maths that caused this $370million spacecraft to explode

In fact, the arithmetic operation in question – if you can call it that – was even simpler than adding 1. Instead, it just involved copying one number that had been stored in floating point format to another location that was defined as an integer – and a 16-bit integer at that (maximum positive value of 32,767).

Unfortunately, the number was already too large to fit in the integer location, and as a result it overflowed. The exact sequence of events that followed is pretty complex but, to cut a long story short, the end result was that the Ariane V became one of the most expensive fireworks in history.

Guarding against cock-ups

This run-through of some of computing's most astonishing mathematical cock-ups may have come as something of an eye-opener to you. If so, you're probably wondering whether tomorrow's computers can avoid making such elementary mistakes.

Surprisingly, perhaps – and with the exception of the Pentium floating point error, which was caused by a hardware glitch – all of the errors we've mentioned here could have been prevented. In that sense, they can all be thought of as software errors.

As an example, let's take that integer overflow on the Ariane V rocket. That an integer can overflow isn't an error on the part of the processor because it's the way it's supposed to work. But whenever an integer does overflow, the processor sets something called a flag that the program can interrogate.

In the case of the Ariane software, the program didn't check for an overflow; if it had done, corrective action could have been taken. Of course, there will always be a limit to how large an integer can be and how much precision a floating point number can have – and this depends on the processor. But all of today's computers are universal computing machines, which means that they can solve any problem involving logic and maths.

So if a processor's internal instructions can't operate on large enough integers or on floating point numbers with sufficient precision, it's always possible for the programmer to implement arithmetic routines that will.

There will be a trade off against speed, though, which is why this isn't usually done. However clever the software or however much memory you use to store a floating point number, the result of some divisions will never be accurate.

We've seen how 1 divided by 10 is an infinite string in binary, and, in the general case, a move to decimal arithmetic wouldn't help either: 1 divided by 10 can be stored accurately in decimal, but 1 divided by 3 equals 0.3333333… ad infinitum.

The bottom line is that whatever number base you choose, some divisions will produce results that can never be stored accurately as a finite number of digits. Even this isn't a show-stopper, though.

Remember how 1.0 - 0.9 - 0.1 often yields an inaccurate answer because of rounding errors even though we know, immediately, that the answer is 0? Well, it's quite possible to write software to store the result of a division as a rational number.

In other words, you don't actually do the division – you just store the two numbers. In subsequent arithmetic operations you handle the values as fractions, just as you were taught in school, and the result will be exact.

So computers might suck at maths, but there's always a solution available to circumvent their inherent weaknesses. And in that case, it's probably more accurate to say that computer programmers suck at maths – or at least some of them do.

Since we're living in the so-called developed world, a reliable mains supply is something we tend to take for granted. Yet many of us have discovered that it pays not to be lulled into a false sense of security.

Mains spikes do occur in the UK, and the consequences can be fairly serious, ranging from loss of data to total equipment failure. Thankfully an imperfect supply isn't something we just have to accept.

By investing in an uninterruptible power supply (UPS), you can protect yourself against most mains power problems – and it may not even cost as much as you'd think.

Power problems

The most obvious threat to the mains power supply is a blackout – a total loss of power. In many cases, this is short-lived and power will be restored automatically within a few seconds. However, when permanent damage has occurred or planned maintenance is taking place, the outage could last minutes or hours.

A long blackout is inconvenient, but the consequences could be worse than that if you were using your PC. Depending on how recently you saved your work and how well your application is able to recover lost data, you might have to spend quite some time getting back to square one when power is restored.

And because your PC won't have shut down properly when power was lost, there's a chance – albeit a mercifully slim one – of file corruption.

Less common than blackouts are brownouts, where the supply voltage drops below its normal value of 240V. PC power supplies are fairly tolerant of low voltage, so, depending on its severity, you might get away with a brownout. The result could be the same as with a full outage, though.

Less obvious but more common imperfections with the mains supply are spikes and surges.

The two differ in their duration (a spike is fast, lasting around a millisecond, while a surge can last several fiftieths of a second) but both result in the supply voltage exceeding 240V, sometimes by a considerable margin. There's the potential here for serious permanent damage to PC hardware.

Finally, it's important to recognise that telephone lines can also transmit spikes, often caused by nearby electrical storms, again with disastrous consequences. Thankfully, UPSes can protect against all these kinds of power fluctuation.

Different types of UPS

In simple terms, a UPS is a unit that contains a large rechargeable battery that equipment can run off, allowing it to continue to operate in the event of a mains failure.

They fall into three main categories – offline (sometimes called standby), line interactive and online (otherwise known as double conversion), and it's important to understand the differences.

The simplest form of UPS is the offline design. Under normal conditions, any equipment plugged into one is powered directly from the mains, which also trickle-charges the battery.

In the event of a mains failure, the battery provides power via an inverter that converts from the battery's low-voltage DC output to the 240V AV required by mains-powered equipment.

• Read TechRadar's APC Back-UPS ES 700VA review

Because an offline UPS provides no inherent protection from surges or spikes, filters are often included to provide some level of shielding from these threats.

A line-interactive UPS is similar to an offline model, but with a twist. Whenever a mains supply is present, automatic voltage regulation is provided to produce 240V at the output.

• Read TechRadar's Emmerson Liebert PSA 500 review

Except in extreme cases, this protects against brownouts or over-voltage and preserves battery life by not having to switch to battery power under low-voltage conditions.

• Read TechRadar's Zigor Ebro 650 review

With an online UPS, power is always provided by the battery via an inverter. When mains power is present, the battery will be constantly topped up and during a power failure the battery will become progressively drained.

The advantage of the online approach is that the equipment is always fully isolated from the mains supply and thus safe from surges and spikes. The downside is that the double power conversion (240V AC to low-voltage DC and back to 240V AC) is inefficient, so overall consumption of power is increased.

Other influencing factors

Other than the underlying UPS technology, the main factors influencing your choice of product will be power output and back-up time.

The power output will be specified in VA or W, and this figure dictates how much equipment can be connected to it.

Just add up the ratings for your PC and any other equipment you want to be able to power (you should be able to find this information on a label on the rear of the equipment) and ensure that you buy a UPS with at least this capacity.

Do bear in mind, though, that only a huge UPS will be able to power your equipment for more than a few minutes from the battery. In reality, you'll probably only use it to save any files that were open and shut down Windows.

You're unlikely to need a UPS that will power your printer, scanner and so on as well. The back-up time is a measure of how long equipment will run on battery power.

This depends on how much power you're drawing from the UPS, so don't take notice of claims that a UPS provides 'up to' a certain time.

Time at full load

Instead, look for the time at full load. For some supplies, this figure may be fairly short – one of the products reviewed here provides just two and a half minutes at full load.

A UPS that allows you to work on through an hour-long blackout will be huge and vastly expensive, but all you need is enough time to shut your system down in a controlled manner.

Most UPSes sound an alarm in the event of a power failure, so you can take the necessary action. Some also have software that closes the system down automatically if power is lost.

There are a few other things to look out for. Many UPSes feature additional sockets to provide surge and spike protection with no battery backup. These could be used to provide protection for a printer, for example. It's also common to find surge and spike protection for data lines entering the PC.

Of most importance is the telephone line, although some products can also protect a network connection. It's hard to get excited about a piece of equipment that normally sits in the background doing nothing, but the benefits are clear.

If you do lose data because of a problem with the mains, you'll be kicking yourself.

A conspiracy theory's recipe is disarmingly simple: all you need is an occurrence, the suggestion of a dark cabal, a wilful disregard for evidence and a creative mind.

Critically, however, once released the tale takes on a life of its own and begins galloping around the globe.

As it travels, self-appointed experts begin picking it over, searching out 'the real truth'. Layers upon layers of detailed information are added to what may originally have been an overheard whisper, a lie or just a simple misinterpretation.

Here, PC Plus magazine examines 10 of the top PC-related theories and try to decide, once and for all, whether they are rooted in reality, or are nothing more than the result of too many paranoid and furtive imaginations. The truth is out there.

• 22 PC pranks to make the office less boring

1. Hidden messages found in the Bible

The Theory: Michael Drosnin claims to have found hidden messages in the Bible using specially written software. Some say it's an elaborate hoax tailored to make money from book sales, but other conspiracy theorists cite rather more sinister motives, given that Drosnin gained the ear of top officials in the Israeli Mossad and the United States Department of Defense.

In his book The Bible Code, Michael Drosnin describes how he used software to search for hidden messages in the Hebrew Old Testament. The messages allegedly foretold events that occurred thousands of years after the Bible was written.

More importantly, other messages are warnings to the present age, the exact time at which computer technology would have been able to unearth them.

FUTURE WARNINGS: In The Bible Code Michael Drosnin claims to have used computers to find warnings of future events in the Bible

PC Plus analysis: Scientific papers presented analyses of Drosnin's results and concluded that they're statistically significant. More recent papers, also reviewed by experts, say that they're not. We wouldn't dare join a debate being held by such eminent mathematicians, but perhaps the Bible itself has something to say on the subject.

According to 1 Corinthians 1:27, "God chose what is foolish in the world to shame the wise; God chose what is weak in the world to shame the strong".

In the light of this verse it would seem surprising that God would have left messages that could only be discovered by a powerful computer, and which would be argued over only by academics. But you'll have to make your own mind up on that one.

2. SETI program is a smokescreen

The Theory: The US government knows that little green men exist – and it also knows that we'll never find them by listening for radio signals. So to keep us off the scent, it promotes futile SETI research.

SETI stands for the Search for Extra Terrestrial Intelligence. It works by pointing large radio telescopes into space and listening for radio signals that have the hallmark of intelligence.

To date, the scheme has found nothing, despite over two million years of processing time being clocked up in the SETI@ home program, where volunteers contribute PC time over the internet to analyse signals.

COVER-UP: Is this really an alien listening base or perhaps something more sinister?

PC Plus analysis: The well-known Drake equation allows us to work out how many civilisations in the galaxy we might be able to hear radio signals from. The equation itself is widely accepted, but there's considerable debate over the values of the variables it uses. Today's best estimates suggest there may be two or three such civilisations. Needles and haystacks immediately come to mind.

So if it's well-known that the technology will have a very low rate of success, why bother using it? Are the conspiracy theorists correct on this one? Well, wait a minute. NASA might have had a SETI programme at one time, but it doesn't any more. The fact that SETI research now receives no public money seems to derail the idea that the US government are using it to distract us from the real way to reach aliens.

3. Government in Wi-Fi safety cover up

The Theory: Forget mobile phone masts – school kids are now at risk from Wi-Fi access points in schools. Government is aware of the health risks but is suppressing the truth. This is a classic conspiracy theory because it's pretty much impossible to prove.

Most scientists believe that low-power Wi-Fi doesn't constitute a health risk, but the only way to know for sure is to carry out large-scale tests over many years using kids as guinea pigs.

THE WI-FI THREAT: Can Wi-Fi really cause medical problems in children? We think not

PC Plus analysis: We're not doctors, but we are clued up in electronics. In Europe Wi-Fi access points have a maximum output power of a tenth of a watt – but a mobile phone can transmit two watts. As you double the distance to a transmitter, the field strength drops fourfold.

Doing the sums, we conclude that if being two metres from a Wi-Fi access point for six hours a day is supposed to be harmful, using a mobile phone pressed against your skull for a second a day is 10 times worse.

4. Google Earth is subject to censorship

The Theory: Google has succumbed to insidious pressure from world governments to keep their secret geographical sites from prying eyes. Google Earth has brought us what was previously available only to the military: high-resolution satellite images of the entire planet. But some censored areas, it's suggested, are pixellated to prevent us from seeing the juicy details.

PIXEL POWER: Not so much a theory but more of a fact. Some governments are secretive about their military installations

PC Plus analysis: This has all the hallmarks of a classic conspiracy theory, but it's actually a fact – as the screenshot of a Dutch military base shows quite clearly.

5. Government eavesdrops on emails

The Theory: Project Echelon – a joint initiative by the British, American, Canadian, Australian and New Zealand governments – intercepts our phone calls, texts and emails. Powerful computers scan their content looking for certain incriminating keywords.

The government is keeping tight-lipped about this one, but, according to civil liberty campaigners the system can intercept satellite communications, snoop on mobile phones and tap into the public telephone system.

ECHELON IN ACTION: Worried civil liberties campaigners say that RAF Menwith Hill is a key element of a government spy network

PC Plus analysis: Quite frankly, after 9/11 and 7/7 it would be rather surprising if the American and British governments didn't intercept communications. The motive and the technology are both there.

6. Microsoft prolonged high-def format wars

The Theory: Microsoft fuelled the format war between HD DVD and Blu-ray. While consumers held off on buying either DVD replacement for fears of picking the wrong standard, the software giant planned to steal a march and launch a high-definition download service.

According to the theorists, Microsoft supported HD DVD even though it knew that Blu-ray would win in order to draw out the battle as long as possible.

In the meantime, their alternative – in the form of Windows Media Video 9 – would be brought to market allowing movies to be downloaded at up to 1,920 x 1,080 resolution.

FANNING THE FLAMES: Did Microsoft knowingly back HD DVD in the knowledge Blu-ray would win?

PC Plus analysis: Microsoft might have had cause to extend the format war, but if it did it wasn't a great success. Blu-ray sales are now starting to pick up, but Microsoft's download service is nowhere to be seen, and until average broadband speeds improve, it's barely practical for many. This one looks fanciful.

7. No code unbreakable for the CIA

The Theory: The US government has powerful computers that are vastly faster than the speediest known supercomputer and can crack any encoded message. In 2002, a 64-bit encoded message was cracked. It took 331,252 PCs working together for almost five years.

Today's 128-bit ciphers would take 18,446,744,073,709,551,616 times longer to crack, and the best experts can suggest is that by 2055 it would be possible to crack them using $42,000 billion worth of specialist hardware.

That would seem to derail this particular conspiracy theory, but if the US government manages to develop a practical quantum computer then even a 128-bit encrypted message would be instantly crackable.

STATE OF THE ART: Even the fastest modern computers struggle with 64-bit messages so we doubt they can crack everything

PC Plus analysis: The fact that any government would crave this capability is indisputable – but most experts agree that none of them has it. One thing's for sure: if the government had this technology then there's no way the CIA would shout about it. As a result, this is one theory that will run and run.

8. Google collects data on our surfing habits

The Theory: Every time we use Google, the words or phrases we enter are recorded so that the company can learn about our surfing habits. Whether your interests lie in the realm of politics or, shall we say, something more 'adult' in nature, our darkest secrets are laid bare.

Motives differ depending on who you listen to. Some say that Google sells the information to advertisers who inundate you with tailored spam. Others suggest that the security agencies are given tip-offs on people searching for bomb-making information.

PC Plus analysis: Google admits that it uses cookies to track your surfing habits and then processes this information to present you with relevant advertisements while you search. But the real crux of this theory is whether Google can link all that information to you as an individual. This is far more unlikely, as the company would need the cooperation of your ISP in order to identify you from your IP address. With all this in mind, we don't advise panicking just yet.

9. US government set up Facebook spy network

The Theory: DARPA (the US government's Defense Advanced Research Projects Agency) used funding to help set up Facebook so that it could use it to collect information on citizens. DARPA's former Information Awareness Office stated that its aim was to collect as much information as possible on everyone. Funding was cut following protests by civil rights activists, but it has been suggested that Facebook now fulfils these aims at no cost to the American taxpayer.

CIA-BOOK: Is Facebook just another CIA venture designed to collect data on the world's citizens? No

PC Plus analysis: There can be no better conspiracy theory than one in which the US government is the alleged antagonist, because one thing's for sure: these guys don't kiss and tell. On the face of it, the theory seems plausible, but we'd have to question the point of it – dissident US citizens surely wouldn't be so stupid as to use the social networking site as a hub for terrorist activity, and we doubt that the White House is interested in pictures documenting just how trashed college students got during Spring Break. Surely the CIA has developed better ways of collecting information on the people that it's interested in by now.

10. Conficker was written by the Chinese government

The Theory: The Conficker worm has received no shortage of publicity in recent months. According to some, it was written by the Chinese government as a test bed for advanced cyberwarfare.

Because its creators could upload new instructions to infected PCs, nobody knew what Conficker might be able to do – and that's what made it so scary. It also made it the stuff of conspiracy theories. In this particular one, the Chinese government will use it to bring the internet to its knees.

PC Plus analysis: This goes against the more conventional theory that suggests the Ukraine is the worm's source. The fact that it doesn't infect PCs with a Ukrainian keyboard layout might be a red herring, but virus expert Eugene Kaspersky is 60 per cent certain that Conficker does have its roots in the Ukraine.

He also believes that its purpose is less sinister than has been suggested. But before you dismiss this talk of cyberwarfare, we should point out the concept does have legs – the Georgian government accused Russia of perpetrating a cyberattack on official websites just before the country was invaded in 2008, and Estonia experienced a cyber blackout in 2007.

Gone are the days when electronic equipment was firmly rooted in the home or office. Our ever more mobile lifestyles demand ever more mobile gear, and the industry has not been slow in fulfilling that need.

But whereas mobile phones and personal audio devices generally have what it takes to survive this itinerant way of life, the same isn't always true of computers. At least, that's the case with ordinary portable PCs.

To take on the worst of what the world can throw, you need ruggedisation. Rugged computers were once heavy, expensive and underpowered. Things have changed, though, and there are now many models to choose from.

Perhaps you're a civil engineer after a laptop that can be trodden in the mud and slung in the back of a 4x4. On the other hand, maybe you're a business or home user who just wants peace of mind on the road. Either way, there's something for you.

What is 'rugged'? Like many terms, what's meant by 'rugged' can be fairly malleable. Most manufacturers tend to refer to various sub-categories to try and pin down exactly what the machines are capable of. These categories are pretty vague, however – Panasonic's include 'business rugged', 'semi-rugged' and 'fully rugged – so it's clear that we need more information to see what a rugged laptop can withstand.

• Read TechRadar's Panasonic Toughbook CF-F8 review

For specialist applications, such as use in a fighter aircraft, in arctic conditions or in a tropical rain forest, you should think about the system's immunity to vibrations, extremes of temperature and humidity. In the main, though, users of rugged PCs want them to survive two common mishaps – being dropped and being soaked.

Immunity from the shock of a drop is easy to understand: most manufacturers quote a simple drop test figure, which is the height from which it can survive a tumble. You might want to enquire whether that drop is on to concrete or carpet, but it's still a useful benchmark.

The degree of waterproofing is complicated by the fact that it's normally specified by an 'IP' rating, a figure which is unfamiliar to most PC users. Thankfully, it's not complicated. The rating takes the form of two digits (IP54, for example). You can ignore the first. The second digit ranges from 0 (no protection) to 7, which means it will survive being immersed to a depth of a metre.

Between these extremes, 4 is common for reasonably rugged computers. It means that the PC is protected against sprays of water from any direction, or four inches of rainfall per hour – something rarely seen in the UK.

The whole reason for investing in a rugged PC is that you'll be using it away from the safety of the home or office. This means that there are a few other aspects you should also consider. First, make sure that the battery life is adequate: happily, rugged models usually outperform their non-rugged counterparts in this respect.

Second, for outdoor use, consider a special sunlight-readable screen – ordinary screens can be virtually illegible under high levels of illumination.

Finally, think about the size and weight – is this something you'd be happy carrying around all day? An unusually long battery life, sunlight readable screens and manageable dimensions all tend to be selling points of rugged computers.

The fact that you'll pay a price premium for ruggedisation isn't at all surprising, and you'll need to get the balance right between price and protection. What you may not be aware of is that there's another balancing act you'll need to perform: one between the level of protection and the performance.

Rugged PCs invariably have a slower processor, less memory, a smaller hard disk and a smaller screen than mainstream equivalents. Sometimes they don't have CD/DVD drives either. The more rugged the PC, the lower the specification will tend to be. There are several reasons for this.

Small screens are much less prone to damage from flexing than larger ones. Rugged PCs also sell in fairly small numbers. Since it takes longer to recoup the development cost, there's no scope for introducing a new model every six months. The lower specification can also be to your benefit, though.

• Read TechRadar's Panasonic CF-U1 review

That might sound like a surprising assertion, but lowering clockspeed results in lower power consumption and longer battery life. Again, this is a compromise that many users of rugged computers are willing to accept, especially since the applications used on the move often aren't the most processor-intensive.

The big players

There are dozens of manufacturers of mainstream laptops and PDAs, but in the world of rugged computers, your choice is much more restricted, especially if you're in the market for a laptop as opposed to a PDA.

Panasonic is the best known supplier of rugged computers. The company offers six rugged laptops, one ultramobile PC (UMPC) and a so-called 'mobile clinical assistant' (MCA), aimed at the medical market. These systems vary from 'business rugged' to 'extremely rugged'.

Getac is Mitac's specialist rugged computer division, and it offers four rugged laptops, two 'durable' (rather less rugged) laptops, one rugged tablet PC and a rugged PDA. Itronix, the rugged computer subsidiary of General Dynamics, has two fully rugged laptops, one fully rugged tablet PC and a fully rugged UMPC. There's also a laptop referred to as 'vehicle rugged', which puts it down a notch in terms of durability.

• Read TechRadar's Getac PS535F review

These are the big three in the realm of rugged laptops, but you should also consider specialist companies such as Terralogic and Blazepoint. Occasionally mainstream suppliers will also introduce rugged PCs – Dell currently offers the semirugged Latitude ATG E6400 and the fully rugged Latitude E64 XFR.

If it's a rugged PDA you're after, you have a bit more choice. Companies offering such products include Psion, DAP Microflex, Motorola and TDS.

So whatever you're after, take the time to check out the market first. There are many different types of rugged PCs – and hopefully one that suits your needs exactly.

Unwanted noise might have been a blight of audio playback back in pre-digital days, but today few people expect that their enjoyment of audio will be marred by hiss and crackle.

This doesn't mean that the nuisance of unwanted noise is long gone, though; far from it. As our electronic equipment has become ever more portable, we've found ourselves subjected to a different but no less annoying form of noise.

From the rattle of a train and the rumble of jet engines to the chatter of our fellow passengers, ambient noise can make listening to music on an MP3 player or watching a movie on a laptop a far-from-enjoyable experience.

The easy solution is just to turn up the volume, but you then run the risk of harming your hearing and annoying those around you. A far better solution to the problem is to use headphones that are designed to prevent background noise from reaching your ears.

• Read TechRadar's Sony MCR-NC500D headphones review

The technique of active noise cancellation (ANC) is not new. In conventional feedback-type ANC headphones, a tiny microphone in the ear cup samples what the ear will hear, background noise and all.

The audio signal into the headphones is then subtracted from the signal out of the microphone, leaving an error signal that corresponds to the amount of noise that will be heard. This error signal is then used to control circuitry that processes the audio signal in such a way that the error signal is minimised.

In the simplest of cases, this is achieved by adding 'anti-noise' – in other words, noise that is phase shifted by 180 degrees and will, therefore, cancel out the ambient noise. In practice, though, the time delay caused by the action of the control circuitry means that noise cancellation will never be perfect.

This type of feedback ANC can employ either analogue electronics or digital signal processing. Sophisticated digital techniques can also be used for the alternative feed-forward type of ANC, which is more effective than feedback ANC for continuous or repetitive noise.

Here, a microphone on the outside of the headphones samples the noise before it reaches the ear and uses this noise signal to generate the necessary anti-noise.

However, because by its very nature continuous or repetitive noise is constant and predictable, it's possible to totally eliminate the timing errors that reduce the effectiveness of feedback ANC – thus cancelling out the advantage of using feed-forward ANC in the first place.

Passive cancellation

Even with the advantages offered by digital signal processing, ANC is good at cancelling low-frequency noise such as hums and rumbles, but not high-frequency noise such as hiss or squeaks.

This is because the wavelength of sound is shorter at high frequencies, resulting in the microphone often being a significant proportion of a wavelength – or even several wavelengths – from the ear.

The noise received by the microphone will, therefore, be phase shifted compared to the noise at the ear, and the amount of that phase difference will differ with the frequency and the direction of the noise.

Because ANC involves using phase shifting to generate the anti-noise, the presence of an unpredictable additional phase shift can get in the way of the process and reduce its effectiveness.

As a result of this, many ANC headphones also feature something referred to as passive noise cancellation – in other words, lots of padding to muffle high-frequency noises.

For this reason, the most effective ANC headphones are circum-aural (that is, they totally surround the ear) or at least supra-aural (sit on top of the ear), rather than the compact intraaural type (fit inside the ear) that are popular for use with portable MP3 players because of they weigh so little.

Intra-aural ANC headphones are available, but they tend not to offer the same degree of isolation from noise as the other types, especially with high-frequency sounds.

Unwanted noise might have been a blight of audio playback back in pre-digital days, but today few people expect that their enjoyment of audio will be marred by hiss and crackle.

This doesn't mean that the nuisance of unwanted noise is long gone, though; far from it. As our electronic equipment has become ever more portable, we've found ourselves subjected to a different but no less annoying form of noise.

From the rattle of a train and the rumble of jet engines to the chatter of our fellow passengers, ambient noise can make listening to music on an MP3 player or watching a movie on a laptop a far-from-enjoyable experience.

The easy solution is just to turn up the volume, but you then run the risk of harming your hearing and annoying those around you. A far better solution to the problem is to use headphones that are designed to prevent background noise from reaching your ears.

• Read TechRadar's Sony MCR-NC500D headphones review

The technique of active noise cancellation (ANC) is not new. In conventional feedback-type ANC headphones, a tiny microphone in the ear cup samples what the ear will hear, background noise and all.

The audio signal into the headphones is then subtracted from the signal out of the microphone, leaving an error signal that corresponds to the amount of noise that will be heard. This error signal is then used to control circuitry that processes the audio signal in such a way that the error signal is minimised.

In the simplest of cases, this is achieved by adding 'anti-noise' – in other words, noise that is phase shifted by 180 degrees and will, therefore, cancel out the ambient noise. In practice, though, the time delay caused by the action of the control circuitry means that noise cancellation will never be perfect.

This type of feedback ANC can employ either analogue electronics or digital signal processing. Sophisticated digital techniques can also be used for the alternative feed-forward type of ANC, which is more effective than feedback ANC for continuous or repetitive noise.

Here, a microphone on the outside of the headphones samples the noise before it reaches the ear and uses this noise signal to generate the necessary anti-noise.

However, because by its very nature continuous or repetitive noise is constant and predictable, it's possible to totally eliminate the timing errors that reduce the effectiveness of feedback ANC – thus cancelling out the advantage of using feed-forward ANC in the first place.

Passive cancellation

Even with the advantages offered by digital signal processing, ANC is good at cancelling low-frequency noise such as hums and rumbles, but not high-frequency noise such as hiss or squeaks.

This is because the wavelength of sound is shorter at high frequencies, resulting in the microphone often being a significant proportion of a wavelength – or even several wavelengths – from the ear.

The noise received by the microphone will, therefore, be phase shifted compared to the noise at the ear, and the amount of that phase difference will differ with the frequency and the direction of the noise.

Because ANC involves using phase shifting to generate the anti-noise, the presence of an unpredictable additional phase shift can get in the way of the process and reduce its effectiveness.

As a result of this, many ANC headphones also feature something referred to as passive noise cancellation – in other words, lots of padding to muffle high-frequency noises.

For this reason, the most effective ANC headphones are circum-aural (that is, they totally surround the ear) or at least supra-aural (sit on top of the ear), rather than the compact intraaural type (fit inside the ear) that are popular for use with portable MP3 players because of they weigh so little.

Intra-aural ANC headphones are available, but they tend not to offer the same degree of isolation from noise as the other types, especially with high-frequency sounds.