Introduction to Guitar & Amp Modelling

Hello, I'm Paul White and in this podcast I'll be looking at modelling for guitars. We were first introduced to the concept of modelling specific amplifiers and effects probably with the introduction of the original Line 6 pod and the Roland VG8. Since then, the increasing computing power combined with increasingly sophisticated modelling algorithms has brought the sound of these model amplifiers ever closer to the real thing. Now in recent years, we've seen amplifier cloning technology as well from the likes of Kemper and IK Multimedia, where a device can essentially learn the characteristics of a real amplifier by firing a range of test signals through it. Amp modelling is only a part of the story though, as the way the guitar speaker cabinet behaves and the way that it interacts with the amplifier and the way that it's mic'd makes a hugely significant contribution to the end result. Many software modelling systems allow amplifier models to be combined with a range of speaker cabinet models. In the hardware world, the choice of speaker cabinets may be very large, or in the case of where just a single specific amplifier is being modelled, then the choice might be more limited or even non existent. How amplifiers are modelled varies from one developer to another, but there are two main approaches. One is to model every individual component and then arrange these according to the circuit diagram for the original amplifier. While resistors and capacitors are relatively straightforward to model, valves and transformers present more of a challenge, not least because of the way that a valve behaves depending on its age, its manufacturer, the way it's biased and how hard it's being driven. There's also the issue of component tolerances as values can vary from the nominal value by several percent and that may increase with age. Add to that the power supply sag that occurs when the amplifier is pushed hard and that's a lot of variables. The other common approach is either to look at the amplifier as a whole, or to break it down into a small number of building blocks, and then to model each block. Both approaches are valid and both present challenges, because an analogue amplifier, particularly one powered by valves, can have a complex dynamic response, which means it responds differently depending on the dynamics of the input from the guitar. In the case of devices that use amplifier capturing, the hardware configures its algorithms based on the recorded test waveform, so in effect it creates its own models. The loudspeaker cabinet can be done by using a network of filters that emulate the frequency response of the original. But again, a speaker cabinet is actually a very complex thing that does much more than just shape the frequency response of the amplifier's output. The cabinet itself has resonances, and the way that the speaker impedance reacts with the damping factor of the amplifier's output stage is also important. A more effective way to capture this complexity is to use impulse responses, in much the same way as a convolution reverb uses impulse responses to capture the characteristics of a particular space. Impulse responses work to reproduce the effect of the speaker cabinet under a specific set of conditions. But of course there are factors that change depending on playing volume. Not least, is the speaker going to distort more when you play louder and also, is the input a run of choppy guitar chords or is it a singing lead line? Has it got lots of low frequencies or lots of high frequencies? Getting a cabinet emulation to produce just the right amount of low end thump without sounding muddy is also quite a challenge. To improve on the straightforward impulse response approach, a number of manufacturers use what are known as dynamic impulse responses. Essentially these are impulse responses that change dynamically according to the playing conditions and volume. Software amplifier modelling products range from being very detailed models of just one or two amplifiers, such as the Blackstar St James plug-ins or the Universal Audio Amplifier emulations, plug-ins and pedals, right up to packages that offer a very large library of amplifiers, speaker cabinets and effects, familiar examples being Native Instruments Guitar Rig, Line 6's plug in version of their Helix, IK Multimedia's Amplitude and Overloud's THU. Amplifier cloning or capturing takes a different approach. Available both in hardware and software, such systems tend to model the amplifier using a machine learning process, usually treating the amp and the loudspeaker as a single system but tested under a range of input conditions. The approach is to mic the amplifier as you would for recording, then feed in a set of special test signals. These cover a range of volumes, frequencies and dynamics and the usual recording mic is used to pick up the end result from the speaker so what is captured is the combination of the amplifier, speaker cabinet and microphone for one specific setting of the amplifier. Tone and gain controls are generally provided, but depending on how these are implemented, they may not interact in exactly the same as the tone controls and gain controls within the original amplifier. In fact, in some cases, they may behave more like putting an emulated amplifier through a separate equalizer, albeit one modelled on a guitar amp EQ circuit. As a rule, the less that you stray from the original captured sound, the more accurate the end result will be. High-end processors, such as those made by Kemper and they're used by a lot of professionals, produce very impressive results but for the more budget conscious, something like the TONEX One from IK Multimedia may be more appropriate. It's certainly more affordable, it's very compact and it produces excellent results. While you can clone your own amplifier with any of these systems, you'll also find libraries of ready captured amplifiers that you can download and use straight away. In my experience, most model guitar amplifier systems, whether the hardware or software, are generally pretty good at reproducing fairly clean sounds and they also do a good job of heavily overdriven sounds. However, it's in the area between the two that really shows up how good a model is. There's a whole range of just mildly overdriven tones, very touch sensitive tones used by country players and blues players, where the playing dynamics has a profound effect on the sound and also on the way that the amplifier actually feels. If the model gets this right, then the sound will be convincing and the playing feel will be good, it'll feel like a real amplifier. But if the model's not quite so good, then it may feel as though you're actually fighting the guitar rather than it helping you to play. So a useful tip for making the less sophisticated guitar modellers feel a little more natural is to feed the guitar into the DAW or modeller via a good quality compressor pedal. You'll probably also need to add some room ambience or reverb as the AMP modelling process itself usually delivers a fairly dry sound. Even if you're one of those players who never uses reverb when playing live, when you hear an amplifier in a real room, there's always some room ambience making a contribution to what you hear. There may be suitable room ambient settings or maybe spring reverb emulation included as part of the product, but if not you can always use your own pedals or plug-ins to add a little bit of ambience as needed. This could be anything from a very subtle short room ambiance up to a standard plate or spring reverb. While reverb plug-ins invariably offer a stereo output created from a mono input, it can sometimes help to use a mono reverb. The benefit here is that the reverb will then be panned to the same location as the dry sound. This produces a better defined location within the stereo field. If working in software, don't be afraid to use a separate EQ plug-in to fine tune the sound. A high cut filter or a high shelving filter can be useful if your chosen overdrive sound is just a little bit too gritty or edgy, while a little lift between 1 and 3kHz can add some bite to a sound that isn't cutting through as you'd like it to. If the low end is overblown or a little flabby, then a low shelving filter might tame it in a more natural way than a basic low cut filter. Here's a guitar sound before we apply any EQ and here we've added a little EQ just to enhance the bite in the mid range and we've got a little bit of shelving cut at the top end just to make the highs a little less brittle. While there's plenty of choice in both hardware and software when it comes to amp, speaker and effect modelling, the choice is far more restricted when it comes to modelling the guitar itself. Some software packages may give you settings to fatten up a single coil pickup sound or to thin out the sound of a humbucker, but to model specific guitars requires a pickup that can send a separate signal from each string. The main contenders in this market were the Line 6 Variax, which is now discontinued and Roland with their VG guitar which is also discontinued now. Earlier VG pedal systems which used the split pickup were produced by Roland, but the latest product to include guitar modelling is the VG-800, which now comes out under the Boss name, which is of course, the guitar branch of Roland. Variax guitars handled all of their modelling within the guitar, where separate Piezo pickups were built into the string saddles and they produced the necessary source signals. This is a very convenient approach and a rotary switch used in combination with the pickup switch changed between the sounds. Roland offered a very similar approach with the V Guitar which had a magnetic hex pickup just in front of the bridge. Though the Variax and the V Guitar are both discontinued, you'll find plenty of choice on the used market. The BOSS VG800 combines guitar modelling with amp modelling, speaker modelling and effect modelling and it has output settings to compensate for plugging into a guitar amp or a flat response PA system or even DI-ing. The signal comes from a split pickup, the GK5. Seven string and bass versions are also available. The pickup comes separately from the VG-800. A small connector box on the guitar sends the signal in a digital format via a TRS jack cable to the VG-800 and the same pickup is also used for the more recent Boss guitar synths, such as the GM800. The necessary pickup can be attached to most electric guitars either permanently or using adhesive pads, in which case the installation is completely reversible. Older instruments compatible with the Roland GK-313 pin connection system can still be used with the addition of an optional adapter box. It's very important to differentiate between guitar modelling and a guitar synthesiser that plays back samples. With guitar modelling, the sound originates from the pickup and then its harmonic structure is modified to match that of the guitars and pickup settings that it emulates. So why do we need a split pickup at all? Well apparently the modelling process is more accurate if the string signals are kept separate, but there's a further benefit. Pitch shifting can be applied to individual strings to create alternate tunings and virtual capos, sub octaves on maybe the lower strings, 12 string emulations and so on. The Variax guitar and the V guitar also offer custom tunings, such as DADGAD or 12 string emulations. Here's a 12 string example from my old Variax, here's the VG-800 and here's an example of the Nashville tuning from the VG-800. Or how about a massive 24 string guitar? Switching between patches is very fast, so for live performance you can go straight from a jangly 12 string to a driven humbucker lead and then back. Separating the signals from the strings also allows for the modelling of such things as acoustic guitars, resonator guitars, banjos and even sitars. Of course engineers always like to push the boundaries, so in the case of the VG-800, there are also some very non guitar like sounds on offer that have more in common with analogue synthesizers than with a guitar. These include an emulation of the old GR-300, the blue box on the floor guitar synth used extensively by Andy Summers back in the days of the police. While there are modelled or sampled guitar sounds within software instruments and hardware synthesisers, it takes a lot of skill to make the result sound like a guitar and not like some form of electric piano. However, it's quite a different story for modelled bass sounds and we have things like IK Multimedia's MODO Bass 2 and as bassists play mainly monophonic lines, the model sound can be extremely realistic. You get a choice of instrument types and pickup settings and with Modo Bass 2, you can even dial in the type and the age of the strings and decide how much fingering noise to include. You can also hear it played with the pick or with the fingers. Because the sounds are modelled and not created by samples, the sound actually changes according to the playing intensity in a much more believable way and if you're not a natural bass player, that might be the way to go. Here are some electric bass examples from MODO Bass 2 and here's the acoustic bass example. At the time of recording this podcast, there were countless amp, cab and effects modellers available in both hardware and software but it seems that at the moment the VG-800 is the only current product that also includes guitar modelling. That said, I hope that other developers will see the attraction of guitar modelling and with developments in machine learning, I'm hoping that it might one day be possible to do sophisticated guitar modelling in software using a standard guitar as the input. Well that's all for now but don't forget to check out our other podcasts on soundonsound.com. Thank-you for listening and be sure to check out the show notes page for this episode, where you'll find further information along with web links and details of all the other episodes. And just before you go, let me point you to the soundonsound.com/podcast website page, where you can explore what's playing on our other channels.