Testing

I would like to discuss some ideas about testing and maintaining systems that I feel strongly about. This might seem like an odd subject to feel strongly about – after all, documentation and testing are the dull bits of any project. Clearly I am biased by the nature of my job since I am often called upon to handle a crisis where an enterprise level application has gone seriously wrong. The situation is always an emergency. It almost invariably involves serious financial loss. In rare cases, it has put people in danger. The first priority is to get the system running again, or as is more commonly the case, limping along.

There are two things that generally cause these emergencies and they are very often found in the same project. The first is poor design. The second is poor testing. Generally poor design is found during testing. If your testing is bad enough then it may be discovered the day that your system goes live. If you find yourself holding the baby when an enterprise solution goes down, your best hope is that you have a good design that has been implemented badly. That may result in some late nights chasing bugs but it is fixable. A botched design may require an almost complete rethink, salvaging what code can be saved from the old system.

Let us consider how much testing an application needs. The answer is that it very much depends on the application. At one extreme, you have a simple application which is to be used rarely by few people and where the results of failure are trivial. A typical example of this is a tool used by a single developer, or a group of his peers, to perform a limited task where it is possible to get the same results in a different way. Let us imagine that you have a little in-house command line tool that tells you what the copyright strings embedded in a DLL are. This tool clearly needs minimal testing and fault reporting is built in since you can pick up the phone and talk to the guy who developed it. At the other end of the spectrum, we have military applications where an error can cost thousands or millions of lives. You really, really don’t want a false positive or negative in an application designed to detect a first launch of nuclear weapons. That sort of software can not be over tested.

For most of us, the applications that we develop fall somewhere in between these extremes. Most of us write applications that are used in the commercial world. The impact of these systems failing is normally a financial loss. For a line of business application, the loss can be significant.

How much testing needs to be done depends on both the complexity of the application and the seriousness of any resulting failure. However, projects generally go over time and budget. When the code is written, it is tempting to cut testing to the bone to get at least close to schedule and after all, the code is good. It was written by smart people. How likely is it that there are hidden bugs? I sometimes think that smart people are the most dangerous kind since they can find new ways to mess up that we stupid folk can’t imagine.

It is accepted wisdom that applications should be rigorous in their error checking and should never assume that the parameters offered to them from an external source are valid. This is simple good sense. However, test data is almost always carefully selected to only contain valid data that will pass the tests. The real world is seldom so cut and dried as that. Real data will contain errors. So should test data. If nothing else, test data should contain errors for reasons of code coverage. If you test with only good data then you are ignoring many, many code paths.

Testing broadly falls in to two categories – functional and scalability/performance. Let us consider functional testing first.

Functional testing is testing to ensure that the expected results are obtained for sample inputs. That is to say that the program does what it was intended to do. Virtually no-one tests that the program does not also do what it was not intended to do. If an application slowly leaks memory then it may be months before anyone notices. If the leak is slow and small then it might be a minor bug that can be ignored. Not all bugs are that benign. I recall one bug that was very nasty and took 6 months of my life to find and fix. The application did what it was supposed to do. It also overwrote some system memory (this was not a Windows application). After a particular sequence of operations was carried out 6 times, the program became unable to access the filing system on the machine. The reason was that a side effect of some faulty logic overwrote successive chunks of some operating system structures until the OS filing system failed. So, whenever possible, test that an application doesn’t do anything that it shouldn’t as well as does what it should. There are a number of tools to help you do this under Windows and hopefully also under Linux. For Windows, the cheapest is the application verifier from the applications compatibility toolkit – a free download from http://msdn.microsoft.com/library/default.asp?url=/library/en-us/dnanchor/html/appcompat.asp . It won’t catch logic errors but it will catch a lot of bad system calls that would cause you grief later – including a lot of the ever popular heap corruption errors. I don’t use Linux so I can’t recommend any tools for that but I am sure that there must be some. Whatever your operating system, the time for such errors to show up is in the test lab rather than on the live server.

I think that now would be a good time to talk about system testing and unit testing. A unit is a component of a system. Let us imagine for a second that your system contains a component which processes a customer order. It works with the billing module. It works with the presentation layer to build a confirmation record for the customer. They are not written yet. The temptation is to wait until they are done and then test them all together. After all, they have to work together so that makes sense, no? Well, actually, no. Let us imagine that we have these three systems all written 4 months down the line. You kind of remember how the order processing component worked. You hook them up after fixing a problem where Bob was working on a different version of the interface specification but never mind, you are ready to test. Great – we are system testing. You test and something doesn’t work. You have no idea what! You have 3 completely unproven components and you are using them to test each other. You could step through in a debugger and follow what is happening but that will take a long time. Let us assume that you pull a couple of 90 hour weeks and get it working pretty well with the test data. Your boss seems happy again and all is well. At that point, how much of the code have you run in that subsystem? 40%? If you were careful with your test data, maybe you have run 70%. Do you feel happy releasing a product where at least 30% of the code has never been run? Would you like to bet your business on it? If you would then may I recommend spread betting? For the rest of us, I recommend unit testing. You build a harness and test the component in isolation. This is often rejected as being too time consuming and I can see why. It does take time to build the harness and test but it saves time later. If the software is critical, you may need to simulate errors to find out just what happens if that memory allocation fails.

Think of it like shopping. You are going to have to pay for what you get. If you pay cash, it is immediately visible and you feel the pain. Alternatively, you can put it on your credit card and your bank balance looks fine. However, you always pay in the end and the longer you delay paying, the more it costs you. System testing components that are well unit tested is always cheaper and easier. It is a classic case of a stitch in time saving nine.

Now, let us talk about system testing. One thing to recognise is that systems rarely have one release and no changes ever. It is an old joke but there is some truth in it. The user does know what he wants and will tell you the moment you deliver what he asked for.

Ad hoc testing is good. Automated testing is even better. The great thing about automated testing is that it can be reproduced consistently. If you fix a bug that had no UI changes then the test script that ran last time should run just as well now. That doesn’t mean that it is fully tested, but it does mean that you haven’t broken anything fundamental. Some parts of the industry call this a smoke test which is originally a hardware testing term. You turn it on and see if it starts to smoke. Automated tests can take longer to create than ad-hoc tests but they are a gift that gives on giving. There are several test tools out there and it would be unfair of me to recommend one but perhaps a little *rational* thought would single out one that *rose* in your mind when you needed to *test*

Scalability and Performance testing are normally the last types to be done. Please note that I say “are” not “is”. These are very different things as some developers have learned at great cost in late nights and missed deadlines. Performance testing shows how fast the code runs for one user. Scalability shows how well the application copes with many users and/or much data. Performance can be important when designing a game, a compiler or a system that is essentially single user. Scalability can be critical with a modern three tier application. Consider the fastest growing area, namely online thin client solutions. Ebay, Amazon, your bank and a hundred other companies do more and more of their business online. If you are launching an electronic banking solution, it is not going to be a happy day when you find that your system can not be split over multiple servers because of an error in its architecture. Functional testing can start early on with unit testing; Scalability testing should start as early as possible even if you have to simulate large sections of the system. Scalability issues are often fundamental to the design of a system and accordingly fundamentally difficult to fix if you make a bad decision. To make matters worse, you normally discover scalability issues when your application is about ready for release. In the case of some developers, you discover them shortly after your application has gone live when it almost immediately goes dead again because more than 100 people have tried to use it.

So, what is the relationship between scalability and performance for a web application? Let us imagine a web based application which displays a catalogue and accepts orders. That should be easy to imagine as there are hundreds out there and everyone re-invents the wheel. Performance is when it takes .1 of a second to order. Scalability is when the response time is less that .8 of a second when 500 users make an order. The two goals are often in conflict. This is something that application developers often discover when they start writing server applications. In a thick client application, it is a great idea to cache information and preload it for the user. When you have 200 users, that stops being a great idea. It may well be better to go to the database each time and let the database designer worry about how to handle the requests efficiently. That is normally a safe thing to do as the big database manufacturers have spent millions of dollars in tuning those database engines.

There are various tools which are good at testing these systems. Application Centre Test which comes with Enterprise flavours of Visual Studio is one and there are multiple third party solutions. Ideally, these should be used with multiple client machines attacking a single server since multiple instances of a test tool are not at all the same in terms of timing. I would advise aggressive testing as well. Test for expected load. Test for wildly optimistic load just in case your solution turns out to be what the world was waiting for. Test until failure – push the box or better yet the boxes until they fall over. If at all possible, spread the server load over more systems than you initially expect to run on and repeat. Learn where the bottlenecks are and open then up - and repeat the tests. Ideally, everything should fail at levels you will never reach with all components giving up at about the same time. Oh, it is a good idea to test with realistic data if you can so that you don’t find that you are looking at better performance because you always hit the cache or worse performance because you have an artificial hotspot in the database. Remember that real data contains errors and users who disappear without warning. So should your test data.

Finally, a few words about keeping these systems running after they are first developed. Systems generally fall in to two types. The first is systems that will be live for a long time and will evolve as the business changes. This is a very common scenario. The second is one off systems written for a special event such as an election or the Olympics (which I ignored earlier). These don’t live long but they are very critical while they last. In both cases, I recommend that there should be a test server (or cluster as the case may be). This is also sometimes called a staging server. This server should be identical to the real one. It should have copies of all the supporting servers. It should be in a position where it could be switched live in a few minutes if needed. Except when it is being used for specific testing, it should be identical to the live server and you must be ready to revert it to the state of the live server or send it live. You need this. The alternative is to test any changes in the software on the live business with no fallback plan. You also need it to act as a backup for a worst case scenario failure or even to share load if a system has a surge in demand.

I have often explained this to wise and intelligent people who have been shocked by the idea. They always ask me the same question in an appalled voice - “Have you any idea how much that would cost?” As it happens, yes, I do have a pretty good idea of what it would cost. How much not doing this would cost a business in the worst case varies a great deal. I have known companies who were losing over 1/3 of a million dollars a day because they didn’t have such a setup. The debate about what to do took 3 days. Delivery and set up of the hardware took 3 more. The system had already been down for a few days. All in all, the lack of the test server cost them about $3 million. The test system cost a them about $6000 plus time from their systems admin staff. Even if the losses had been 1% of what they were, the test server would have been cheaper.

So, testing may not seem very exciting. I would agree that it is not. However, not testing properly is very exciting indeed. If you live for the thrill and don’t care about how difficult it is to get another job then feel free to go for the excitement. For the rest of us with a mortgage and 2.4 children, I can not recommend good testing well enough.

Some of my guitar collection

My current collection contain of:

Starchild. Once a humble Telecaster, now a work of art courtesy of an artist called Alan He will paint a guitar for you as well if you give him large amounts of money and a degree of freedom. Alan likes painting images of girls who are suffering from a shortage of clothes. Please contact me if you want him to do work for you. This guitar is equipped with two Seymour Duncan pickups that give it a classic sound.

Excalibur. Not a sword but an axe if you will pardon the term. This guitar was built for me by Rick of Sticks and Strings in Thatcham, Berks, England. Again, this is very similar to the Telecaster design of Fender (bless you, Fender and please don’t sue me) and equipped with Kent Armstrong pickups. Like Excalibur, this bears the Voodoo logo since they come from the same stable.

Fender Stratocaster. A stock Mexican model. I haven’t changed this one at all because it was exactly perfect for me as it arrived. There is a line in a Meatloaf monolog “I once killed a boy with a Fender guitar. I don’t remember if it was Stratocaster or a Telecaster … It had a heart of chrome and a voice like a horny angel”. Silly boy. That is clearly a Stratocaster. I prefer the Mexican version to the US version as the pickups are a little brighter. This is an exceptional example.

Epiphone Les Paul. In honey sunburst. Again, a stock guitar. It is pleasant enough but it always leaves me feeling a little unsatisfied. It plays well and sounds just fine but somehow the magic isn’t there. Sometimes I just can’t get no satisfaction.

Epiphone SG400. This is a stock guitar except that I had the pickups replaced with Seymour Duncan’s. The result is rather pleasing. The intent was to have Gary Moore on the neck and Carlos Santana on the bridge but it is actually rather good for laid back jazzy blues. The Bigsy Tremelo looks so good but it serves no purpose since I never use it.

Tanglewood Baby. A tiny guitar that thinks that it is a Martin or a Taylor. This is a joy to play with a low action and a sound full of life and tone. It sounds as if it is a dreadnought (if perhaps a tad quiet) but is smaller than an electric. I can take it anywhere and the sounds will stand up to any fair measure. I am sure that this is due to the Stika spruce soundboard and the Mahogany back, sides and neck. The fingerboard is ebony. It is a little neck heavy but I can forgive a few faults in a loved one.

Brunswick Acoustic in Sunburst. A good all round strumming guitar with a lot of top end. It isn’t remarkable but it is very playable and I rather like the look of it.

Yamaha Pacifica 112… well, it was once. This is a guitar that I got second hand from the lead guitarist of a Nu Metal band. It plays well and has a versatile range of tones thanks to the unidentified Seymour Duncan pickup in the bridge. It is a good guitar for casual play or for loaning to friends… after all, a ding or two won’t really spoil it. I have come to like the appearance even if it is very rough and ready. I can accept it for what it is.

Tanglewood shark. I liked the look (i.e. Shark shaped with a paint job to match) and it has a built in amp and speaker. Points for: It looks fun and can be played without an amp. Points against: It is horrible to play and has a sound like a banjo being played through a baked bean tin. It is true what they say. A fool and his money are soon parted. Even plugged in to a class A tube amp, it still sounds appalling.

Kramer Striker in cracked black ice. This guitar says ROCK. No, it screams it. It also demands Spandex outfits and big hair. The neck pickup is a hot humbucker. The bridge pickup is a Quadbucker – 4 rail pickup in one. If you need a little more output (and who could????) you can have both neck and bridge pickups on at the same time. 6 pickups anyone?

In a change of pace, a Yamaha classical guitar. Perfect for those more contemplative moments. It is an older guitar and that has given it a degree of maturity. It is less than ideal for playing “Smoke on the water”.

Ninja Katana Strat copy, This is a bit of a mystery. It is Strat shaped with knock-off pickups. It plays well enough and it at least 20 years old. It is cheaply made but somehow plays and sounds very well indeed.

Danelectro doubleneck. This is the version with a 6 string and 12 string necks rather than the 6 string and bass variant. It has more jangle than a Rickenbacker and is simply fun. Astoundingly, it is also a pleasure to play and sounds wonderful on both necks. The lower neck is essentially a Danelectro U2

A Danelectro U2. Danelectro guitars were originally sold via the Sears catalog and were designed to be cheap and light. For this reason, they were made of Masonite (like suitcases) and pine or chipboard blocks. They have a type of pickup known as a lipstick pickup because they were made of surplus lipstick tubes. Given this, you would reasonably expect them to sound horrible. Remarkably, they sound good. The tone is unusually toppy with little sustain. They can sound very good when you want a lot of brightness in chords or lead work.

Today, I will be talking about guitars, a subject close to my heart as I have quite a collection of them. I will describe the evolution of the modern guitar and mention some of the more interesting items in my collection.

I should briefly describe the parts of a guitar as I will be referring to them throughout this article. A guitar has a body and a neck and a head which is sometimes called the headstock. The strings are attached to the body and vibrate between a point on the body called the bridge to a point at the end of the neck called the nut. There are raised strips on the neck called frets. Notes are selected on each string by pushing the string down to the neck which effectively shortens the string to the length between the fret and the bridge. This changes the wavelength of the vibration in the string and so increases the frequency. Hollow bodied guitars normally have one or more soundholes to project the sound forward.

Guitars have existed in various forms since the 15th century. Of course, these were acoustic instruments which relied on their body shape for sound projection. The elements that we would expect are all there in the early design but with some interesting differences. The body of the guitar was longer and thinner than a modern guitar and the strings were gut rather than the more familiar steel or bronze of modern guitars. The sound would have been like a quiet classical guitar that uses nylon strings. The strings were tied to the bridge and to the tuning pegs at the head of the guitar. One obvious difference was the number of strings. Early guitars had 4 or 5 courses of string – a course could be a single, pair or triple string. Modern guitars have 6 strings except for bass guitars and some of the electric guitars used for heavy detuned rock. 12 string guitars are actually a revival of an older idea since they are 6 courses of 2 strings, each tuned to the same note but in the case of some strings, tuned an octave apart. Original guitars had bowl (curved) backs – a concept revived by the Ovation guitar company. The frets were generally strips of gut tied around the necks of the guitar. The fret allows accurate pitching of notes as I will describe later.

From the original design of the guitar, many variants have arisen. Here are some of the main types of which there is a bewildering number:

1. The Classical guitar

This is the clearest descendant of the original guitars. These are typically figure of eight shaped guitars with the upper part of the body narrower than the lower – they have a pronounced waist as most acoustic guitars do. They are nylon strung and have a fixed bridge with no method for altering the intonation which I will explain when I come to electric guitars. The front of the guitar body forms a soundboard as is usual with acoustic guitars with the back of the guitar having a lesser influence.

Cheaper contemporary models will use a composite (plywood) soundboard. More expensive models will have a high quality spruce top. The soundboard will typically have a pattern of bracing struts hidden out of view. Various woods are used for the back and side. Frets are almost invariably metal on modern guitars. Classical guitars do not normally have a pickguard which is a plate for protecting the top of the guitar from pick scratches since classical guitars are usually played fingerstyle. Fingerstyle is where individual notes or sometimes pairs of notes are played rather than entire chords – and they are played with the fingers rather than a pick as the name suggests. Classical guitars produce little volume and were generally restricted to parlour use until amplification became available.

2. The Spanish guitar.

The Spanish guitar is similar to the classical except that it has one or perhaps two plates on the soundboard that can be used for tapping to produce percussive effects. There are also two variants used in Mariachi called the vihuela and the guitarro which are respectively smaller (and higher pitched) and larger (and lower pitched) than a standard guitar. Unlike the classical, Spanish guitars are sometimes strummed to play chords as well as being played in the classical fingerstyle.

3. Folk guitars

These are a very common type of guitar. Outwardly similar to the classical guitar, these have a large body (especially the type known as dreadnought) and have metal rather than nylon strings. To take the additional stress, the necks of folk guitars have one or rarely two truss rods – metal rods which contain an adjustable screw allowing the tension of the neck to be adjusted against the pull of the strings. They can be played like a Spanish guitar although they are not normally struck for rhythm effects. They are often played with a flatpick or plectrum to strike the strings harder. Folk guitars (often just called acoustic guitars) are much louder than classical guitars and can fill quite an area if played with vigour as many buskers have demonstrated. Some acoustic guitars have a cutaway which is a change in the lower shoulder of the guitar to allow better access to the frets nearest the body. There are two common forms of cutaway, the Venetian and Florentine with the former being much more common. They are often played with a plectrum or flatpick and accordingly have a pickguard. A plectrum is simply a teardrop shaped bit of plastic although other materials are sometimes used. It is used to pick a string or strum multiple strings and gives a stronger and cleaner attack than a guitarist’s fingers resulting in a cleaner and louder sound.

4. Resonators

There has always been a quest to make louder guitars and so fill a larger space or to stand out in a band. Acoustic guitars are relatively quiet when compared to instruments such as drums and brass instruments and are easily lost in the “mix”. Before electronic amplification, mechanical methods were used to produce more volume. One such was the resonator design. A metal cone (or three in tri-cone designs) was placed in the body of the guitar with the bridge driving the cone. These are significantly louder and have a harsher, brighter tone much admired by devotees of early Blues music. Many of these are intended to be played on the players lap with a side or steel instead of conventional fretting. Pedal steel guitars are a further evolution that I will cover separately in the “electrics section”. The first resonators were seen in the 1920s but they are still made today for fans of older musical types. Some resonator guitars have all metal bodies with some of the finest being made of bell brass. If you have ever wondered about the strange chromed guitar on the cover of the Dire Straits album “Brothers in arms” then you will have seen and probably heard a resonator.

5. Lap steel guitars

As mentioned in the last section, resonator guitars are sometimes played on the player’s lap. This has been further developed in the lap steel and Hawaiian guitars which are normally unplayable in a conventional manner. Lap steels range from conventional looking guitars with square neck profiles to electric examples which are little more than a plank with strings and a pickup. Lap steels are always played with a slide. Rather than the player’s fingers forming chords by fretting a combination of notes, a slide or“steel” is used. This may be augmented by some fretting one or more notes with the fingers but that would be unusual as most music arranged for slide guitar is designed to be played on instruments where this is not practical. The slide or steel is generally a metal or glass tube that the player wears on one finger and slides up and down the strings to alter the pitch. This gives a smooth change of pitch that is quite characteristic. This was first done by Blues players who used a knife for the purpose. Later players (who presumably wanted to keep all their fingers attached) started using cut and smoothed parts of glass bottles for the job. One interesting point is that the length of the string is determined by the position of the slide and not by the frets. My own lap steel has frets that are only printed since the string will never come in to contact with them. They are often played with finger picks – essentially false fingernails used to pluck the string more firmly than would be possible with real fingernails – they are typically metal or hard plastic.

6. Travel guitars

Acoustic guitars are fairly bulky. If you go to any busy railway station, you will see people carrying or wearing guitars. In an attempt to make guitars more portable, travel guitars have been developed. Various systems have been tried from the very small bodied guitars such as the Martin Backpacker to guitars which can be disassembled like the SoloElite Dragonfly. Most of these win admiration for their novel design rather than their tone. The Martin Backpacker in particular has a disappointing sound considering that Martin is among the finest acoustic guitar makers in the world. Interestingly some guitars which would previously be described as parlour guitars are now being marketed as travel guitars. A parlour guitar is just a conventional acoustic guitar with a small body. I have a Tanglewood Baby that was marketed as a travel guitar. Happily, it is an exception to the rule that small bodied guitars have inferior tone. As the name suggests, it is a small bodied variant of a larger but similar guitar. The Baby has a “toppier” tone which is to say that the higher pitched overtones are dominant in the sound which I personally prefer.

7. Electro acoustics

These are guitars that are essentially acoustic guitars which have been fitted with a microphone to enable them to be amplified. Guitars of this type often have a small graphics equaliser built in to the body. It is also possible to retrofit a microphone in to a conventional true acoustic which popular kits such as the Fishman acoustic pickup. The microphone is normally mounted in the soundhole or at the base end of the body – the two mountings provide different tonal qualities and there is much debate over which is the better position. Guitars of this type suffer from feedback problems when played with high amplification – the soundboard vibrates from the amplified sound which is picked up by the microphone and sent to the amplifier which plays the sound which makes the soundboard vibrate… and so on until the sound dissolves in to a distorted howl. This feedback effect is often used in rock where solid body guitars predominate – the effect is much easier to control in solid body guitars.

Electric Guitars

Electric guitars are the mainstay of modern Rock, Blues, Funk and much pop. They range from electro-acoustic guitars to wonders of electronics. They use magnetic pickups to convert vibration in the string to a signal for an amplifier. Electric guitars are very quiet indeed when played unamplified with solid body guitars being effectively silent since they have no sound cavity to project the sound. Electric guitars are always steel strung and have a truss rod to counter the pull of the strings - except for some Rickenbackers which have two and some early Telecasters which had none. Most statements about guitars seem to have an "except" clause.

Conventional magnetic pickups have one pole per string – the strings have a steel core and the movement of the string through the magnetic field of the pickup creates a small current which is then sent to an amplifier and on to one or more loudspeakers.

8. Semi-acoustics, sometimes called archtops

These are guitars which are hollow bodied but have magnetic pickups. They may be similar in appearance to conventional acoustic guitars but generally have a thinner body. They very often have a classic appearance since the more avant-garde designs work less well – and the more avant-garde designers have generally favoured the solid bodied rock guitars. Classic examples of this type of guitar are the Epiphone Casino, the Gibson ES-335 (the ES stands for Electric Spanish) and the Rickenbacker 360. Technically, these guitars can be played acoustically since they are hollow bodied but the sound is too thin and quiet to be useable. They suffer from the same feedback issues as the electro acoustic designs. It was this problem that led to the development of the solid body guitar – a definite improvement on the previous solution of stuffing towels in to the hollow spaces of the guitar. These are the guitars most commonly used in Jazz. Their construction is similar to that of acoustic guitars.

9. Solidbody electric guitars

These are available in a bewildering variety of styles and colours. As a guitarist, I can confidently state that they all sound and feel different. To most non-guitarists, they all sound pretty similar. They all have the same basic design – a solid slab of wood for a body with one or more magnetic pickups. The differences between different models may be cosmetic or more fundamental.

Wood type

The type of wood affects the density of the guitar body and the acoustic properties. While solid body guitars do not project the vibration of the plucked string, the vibration travels within the body of the guitar and resonate giving complex overtones. Generally, the denser the body, the “darker” the tone. Guitars such as Gibson’s Les Paul models use Mahogany with a maple cap and have a characteristic dark tone when compared to guitars made of Ash or Alder which are much less dense woods. These woods are used on guitars such as Fender’s Stratocaster and Telecaster which have a brighter tone. It is always difficult to describe guitar tone as English (and probably other languages) lack the words required to capture the subtle differences. Denser, harder woods also give more sustain – the note takes longer to die away after being sounded. How desirable this is will depend on the style of music being played. The least dense guitar that I have played is the Danelectro U2 which is made primarily of Masonite, a material more commonly used for suitcases. This gives a remarkably bright jangly tone.

Pickups

There are three types of pickup reasonably commonly used on electric guitars: Single coil, humbucker and Piezoelectric. Some guitars have more than one type fitted to give a wider range of sounds. Pickup selector switches allow one or more pickups to be selected at the same time. Only the first two of these are magnetic pickups. These have a magnet per string (or a single large magnet on some cheaper models) wound with coils of wire which sit in the body under the strings. These coils of wire pick up the current generated by the ferrous string moving in the magnetic field created by the magnet. Single coil pickups are used to excellent effect on guitars such as the Telecaster and Stratocaster. They are generally paired with lower density woods. Single coil pickups have perhaps the most guitar-like of the tones available from magnetic pickups but they suffer from electrical interference. The Stratocaster is famous for its 50 or 60 Hz hum (depending on location) when no note is being played. The humbucker was developed to eliminate the hum. It does this by having paired couples of coils which are arranged to be out of phase to cancel out the interference. This works well but gives a different tonal response to the single coil – and a higher output which is much prized by rock guitarists. The thicker tone of the humbucker is generally paired with denser woods and classic humbucker equipped guitars include the Gibson Les Paul and Gibson SG (short for Solid Guitar). Some guitars which are intended to for heavy rock are equipped to allow up to 8 pickups to be used at once to give ludicrously high output levels. Piezo-electric pickups are quiet different. They are normally built in to the bridge of the guitar and use crystals to convert pressure (and accordingly vibration) in to voltage. The output of these crystals is very small and it is normally necessary to have active electronics built in to the guitar to convert the signal to a usable level. Piezo pickups give the most faithful reproduction of the vibrations occurring within the guitar. They are the least common type of pickup. There are also MIDI pickups but they are very rare. I will mention them again when I come to the future of the guitar.

The Whammy Bar / Vibrato / Tremolo

This is the handle like projection that you may have seen on electric guitars. The original purpose was to add a little vibrato to a played note or chord. They are not used so much these day except in heavy rock where a technique known as the Divebomb is used. The way that the whammy bar works is that it allows you to alter the tension on all the strings at once very rapidly. A little tug and the pitch swings up. A little push and the pitch drops. A Divebomb is a huge drop, sometimes making the strings go completely slack and hit the pickups with a bass thud. The first systems were the Fender 3 screw system and the Bigsby system. Both had disadvantages. The Fender system often caused the tuning of the guitar to drift. The Bigsby system was more stable but allowed less variation. Various refinements were tried with some success at improving both systems. The biggest advance was the Floyd Rose Tremolo which has the strings clamped at both the nut and the bridge giving excellent tuning stability. It is a floating tremolo which means that it can travel both up and down so it is both stable and highly variable. One common type of tremolo seen on modern guitars is the hard-tail tremolo where the pitch can be lowered but not raised beyond original pitch because the bridge is mounted flush to the body of the guitar. This is slightly cheaper to build and rather more stable.

The bridge and the mounting of the strings

Various approaches have been used to attach the strings to the body. Acoustic guitars use plastic pegs to hold the strings in to the body and a single bridge. Electric guitars have the strings going through holes in the bridge – in the case of some guitars such as the Telecaster, though the body of the guitar as well. While acoustic and classical guitars have a single bridge, electric guitars have three or six saddles that make up the bridge. These essentially are segments of the bridge that are individually adjustable to allow the intonation to be adjusted. A string that is in tune when played open (no note fretted) may be out of tune when played fretted at the 12th fret (one octave higher). By small adjustments to the saddle, the length of the string can be adjusted to give a good compromise so that the string isn’t far out of adjustment at any point on the neck. It has always struck me as strange that the most expensive custom made acoustic guitars don’t have any adjustment at all for this while the cheapest and nastiest electric guitar does.

Tone controls

The tone control of most guitars is classic 1950’s electronics. Essentially, the tone control allows you to bleed off some of the higher frequencies. Internally, it is simply a variable resistor (a rheostat) and a small capacitor.

Styling

For many guitarists, look is as important as function. The style of the guitar will have a lot to do with the sort of music that the guitarist is playing. Heavy Metal guitarists will want something in black with spiky bits. A blues player will want something more traditional. The most common designs are based around a few classics. The Telecaster is a guitar shaped slab with a single real cutaway. It has always been a popular shape and has recently enjoyed a return to fashion. The Stratocaster is the most copied shape of all with its wonderfully retro design that was intended to look futuristic in the 1950s. Instead it conjures up images of classic American cars. It is a design classic loved by many stars such as Jimmy Hendrix, Mark Knopfler, Eric Clapton and countless others. The Les Paul is a small bodied guitar, mercifully so as the thick mahogany body is very heavy. The classic versions of this body type are the flame top (so called because of the tiger striped grain on the maple cap) and the gold top. Again, this is a hugely popular shape that has been copied many times. Any number of heavy rockers have used them including Zakk Wylde with a striking black and white bullseye design. At least 70% of the guitars that you see in a shop will be based on these designs.

The future of guitars

Most guitars are essentially based on 1950’s technology but there are some interesting advances in the wind.

Gibson have built a Les Paul which has an Ethernet interface to connect directly to a network. Applications are so far limited but computers are increasing the recording solution of choice. It is yet to find a mass market.

MIDI pickups are now built in to some high end guitars and are available as a rather costly after market add-on. With the pickup and an electronic box of tricks, the guitar can output MIDI and with a synthesiser mimic a wide range of instruments including electronic and wind instruments.

Perhaps the most radical change is the modelling Guitar. “Line 6” produce the Variax range that have some very advanced digital signal processing (DSP) built in. The design hides this behind conventional controls so that the guitar appears like a guitar and not like the computer that it really is. The most striking visible difference is that there are no magnetic pickups. The pickups are actually piezoelectrics built in to the bridge. The signal from these is fed in to the DSP and converted in to very good copies of classic guitars. With the guitar selector knob and the pickup selector, you can select from an entire stable of classic guitars including some very rare acoustics. It is a remarkable bit of engineering.