MIDI guitar has a long and chequered history, littered with great expectations and expensive failures. Dave Lockwood investigates the latest contender to see if it really is able to bring something genuinely new to the field.
It seems as if guitar‑to‑MIDI converters have been around for almost as long as MIDI itself. The convenience and flexibility of MIDI recording quite naturally exerts a strong pull on the more open‑minded and adventurous guitar player. After all, those great synth voices that keyboard players enjoy would sound even better played from a more expressive instrument, wouldn't they? The reality is, however, that there is practically nothing that sounds quite as unmusical as a guitarist playing synth or sampler voices randomly in and out of time, due to inconsistent system delays, with as many wrong notes as right ones, because of mistracking and spurious triggering! The history of the MIDI guitar is a catalogue of disappointing failure and apparently promising developments that ultimately turn into dead ends.
Over the past few years, however, some of the newer conventional guitar‑based systems have been able to achieve significantly improved results by utilising a proprietary high‑speed protocol (not MIDI) to talk directly to a set of synth voices within a dedicated control unit. This is a configuration which permits some useful 'cheating', in particular the technique of generating an essentially unpitched burst of noise at the start of each note to give the impression of a near‑instantaneous response. This is not really solving the pitch‑to MIDI delay problem, however. The gain in speed is actually something of an illusion, in that it only works when the guitar is talking to the unit's own voices, while the MIDI output from the unit is as delayed as normal. Another strategy employed in a bid to speed up pitch‑to‑MIDI conversion is for the system to make an initial 'guess' at the pitch, followed rapidly by a flurry of pitch‑bend data to achieve the real note as it is subsequently detected by conventional pitch analysis (this can be quick enough for you not to notice the process by which the note is achieved, but has the unfortunate side‑effect of making the data virtually uneditable if you are working with a sequencer).
Both of the above scenarios are actually fine for live performance, or even recording guitar synth performances to tape or hard disk, but they render the system no more useful for programming a MIDI recorder than any of its forerunners. Yet it is the desire to work effectively with a sequencer which is the real driving force behind many guitar players (and especially those who read this magazine, I suspect) becoming involved with MIDI guitars in the first place. Not many of them, in my experience, really want to go out and perform live on 'guitar synth'; they're just fed up with being bad keyboard players when sitting in front of their sequencers, instead of being able to be the expressive musicians they know they really are.
Plug And Play?
And now, the latest contender, the Axon AX100 from German company Blue Chip Music Technology. As usual, forget all the hype — you can no more simply pick up your guitar and play with this system, having it flawlessly translate your every nuance into neat packages of MIDI data, than you can with any other system. But what it will do, and from a conventional guitar, moreover, is get the notes out on time. All of them, all of the time. Lower strings, upper strings, it makes no difference. The response is consistently fast. If you're one of those people who has struggled for years with a slow pitch‑to‑MIDI system, at this point I know you'll be thinking "that'll do nicely. I don't care what else it does, I can live with it!"
The strong resemblance between the AX100 and the system marketed by Yamaha as their G50 system is no coincidence. Until two years ago the project was a joint development, but Blue Chip have taken things a stage further and this extra work has resulted, I believe, in a significantly more successful package. Like the Yamaha, the AX100 uses a method of pitch detection termed Transient Early Recognition, involving analysis of the initial pick attack. Apparently, the sonic signature of each note's pick attack and the reflected wave from the bridge and the nut or fret is sufficiently unique for the system to detect the pitch of the note being played, the position along the string (ie. how far away from the bridge you're playing) and how hard you attacked the note. This means that, instead of waiting for the traditional complete cycle or cycle and a half of the waveform (running into tens of milliseconds on the lower strings), the AX100's pitch detection is claimed to take place inside "a 3ms window". In practice, the response is quite evidently as near instantaneous as makes no difference. After the initial transient has produced the MIDI Note On message, conventional frequency analysis is used to track subsequent pitch changes (vibrato, bending, etc) or to rapidly correct an incorrectly detected note.
Part of the ongoing development program for the Axon, since the end of the joint effort with Yamaha, led to the D‑A converters being uprated from 12‑bit to 18‑bit, in an effort to improve conversion accuracy. This increased converter resolution has the further effect of allowing the Axon to now work with acoustic guitars, via a dedicated replacement bridge saddle unit with integrated 6‑way piezo pickup (an additional interface board is also required to be mounted in the rack unit). This is quite an achievement, given the greater amount of crosstalk and resonance going on with a typical acoustic guitar. Nevertheless, I can verify that it certainly does work, and with about the same level of success as the electric guitar pickup. Four‑string electric bass guitar is also now catered for with a dedicated magnetic pickup — with conventional pitch‑to‑MIDI conversion techniques the delays involved would have made this a non‑starter, of course (the increasing numbers of 5‑ and 6‑string bass players out there will be glad to know that there's now a suitable bass pickup available from Yamaha, which is compatible with their own G50 system and the AX100).
System Components
The AX100 6‑string electric guitar system consists of a 'standard' hexaphonic split pickup, and accompanying guitar‑mounted remote control unit (AIX‑101), feeding a discrete signal for each string, via a 13‑way multicore, down to a 1U rackmount control unit which does the conversion work. (The converter unit can also have an internal voice card fitted, but it is the MIDI performance that I think most readers will be primarily interested in.)
The 1U control box has its own power On/Off switch, but requires an external 12V AC power supply. An internal PSU, or at least some means of securing the AC supply's connector into the socket, would have been nice on a unit that the manufacturer so obviously envisages being used for live performance. The rear panel offers MIDI In, Out and Thru sockets, plus a quarter‑inch jack output for the normal guitar signal, which is also conveyed via the multicore along with the hex pickup output (although you can still opt to access the guitar output directly from the instrument's output, this increases the possibility of generating hum on the guitar signal via an earth loop). Further quarter‑inch jacks offer connections for two optional 'expression pedals' (freely assignable to volume or any MIDI controller) and two footswitches: Chain (for successively accessing a range of presets) and Hold (to activate a number of programmable sustain, layering or arpeggiator functions). A pair of outputs for the optional on‑board voice card (an XG‑compatible card offering 480 sounds, complete with quite decent reverb and chorus effects) completes the rear‑panel connection facilities.
AX100 rear panel connections.
On the front panel, things initially seem pretty straightforward, with the action centring on the 2‑line LCD display, and just four main modes to deal with — Preset, Global, Utility and Chain. The display is backlit, but there's no viewing angle adjustment, which I guarantee you'll soon find rather frustrating if you're forced to mount the thing away from eye level. The 13‑way guitar connector also goes in at the front, with no option to connect to the rear — I'm sure most people would prefer the neatness of the latter in any semi‑permanent home recording setup, and it would hardly compromise the practicality of portable use, as you will have to make all the other connections to the rear anyway.
Adjacent to the guitar input is a quarter‑inch jack and associated level trim pot, which allows any monophonic audio signal to be converted into a MIDI output. In practice there are always severe limitations on how well this can work (conversion delay, false triggering due to background noise, and so on), and the Axon is no exception.
Modus Operandi
The four main operating modes are selected with dedicated keys, leaving just a self‑explanatory line‑up of Store and Edit, Enter and Exit, and Parameter Inc/Dec and Value Inc/Dec keys. On the whole, the operating system works pretty well, once you've worked out that some parameters reside in sub‑menus (it's in the manual, of course — a well organised and helpful tome running to something in excess of 90 pages — but I did manage to figure out everything else without it).
The optional on‑board voice card offers 480 XG sounds.Global mode contains all the big stuff, and is where you tell the system what instrument you're playing (you can store four different setups). Here you set MIDI channel allocations, pitch‑bend range and, most important of all, the Note On and Note Off sensitivity values. Trigger Level determines how hard you need to strike a string to generate a Note On event. Maximum sensitivity will follow every nuance of your picking hand... and follow all kinds of unintentional stuff too. Every lift‑off, slide and accidental hammer‑on will be turned into unwanted MIDI notes. In practice, it's important for the successful operation of the AX100 that you use the least sensitive setting that your picking style will allow, particularly when playing polyphonically (we all can, of course, fret more accurately, apply left‑hand damping and pick much more cleanly on single notes than we can on chords). At the other end of the scale, setting a highish Note Off limit value (effectively the amplitude at which the system determines that a Note Off should be sent) gives you nice clean data but shortens the length of sustained note available to you. Optimum setting for this parameter seems to depend very much on the envelope of the string vibration of the guitar used. In addition to the global Trigger value, individual string sensitivities can be set, mainly to achieve an even response across the strings.
Preset mode is the normal playing mode. You can store up to 128 custom setups, each with its own parameter set to optimise it for a certain playing style or synth voice. This is an essential part of almost any MIDI guitar system, because the setup that works best with, say, a bass synth patch is very often next to useless with an acoustic piano. In an ideal world this would not be so, but the world of MIDI guitar is far from ideal. In reality, it's all about acceptable compromises. Presets can be recalled via MIDI program number, or selected from the front panel or the guitar‑mounted remote control unit. The slightly confusing bit for the newcomer is that a preset can consist of as little as one set of parameters, operative across the whole instrument and acting on a single synth patch, or up to 12 different sets of parameters if every split zone that the Axon offers is activated simultaneously. What's a split zone? Well, because the Axon can detect where along the string's length you are picking a note, it can use that information to establish up to three picking zones which will activate different voices or different parameter sets (transpose, etc). Add to this the obvious 'string split' zoning (different strings to different voices) and 'fret split' zoning, where the fingerboard is divided into two separate areas (by entering a fret number as the split point) and you have an impressive and, frankly, virtually unplayable, 12 zones! Each zone can have its own program number, its own volume setting, transposition, pan position and effects setup, if supported by the voice module, and its own pitch‑bend setup (see 'Pitch‑bend & MIDI Guitars' box).
Also available for each zone within a patch is an individual Velocity Sensitivity and Velocity Offset parameter. These are crucial settings for matching your playing style to different types of synthesized instrument. The maximum (127) setting gives you the largest range of dynamic expression (and the least noticeable spurious notes if no positive offset is added), but makes it almost impossible to play, say, an electric piano sound — an instrument with an inherently limited dynamic range — with any degree of realism. As you reduce the velocity sensitivity, you limit the maximum achievable velocity value, which prevents you from achieving full brightness on voices where the filter is mapped to velocity. It then becomes necessary to add a velocity offset to raise the average level, effectively achieving velocity 'compression with make‑up gain'. This makes it more comfortable to play most voices, but the downside is that spurious notes are then louder because your average velocity is always higher.
If balancing the Velocity Sensitivity, Velocity Offset, individual string sensitivity, global Trigger Level and global Note Off limit for every synth voice is beginning to sound like hard work, well, frankly, sometimes it is. The only sensible thing to do, therefore, is to take advantage of the fact that the manufacturers have already done some of the hard work for you in the presets that they have designed to drive the voices on their optional sound card. Starting from a preset designed to work with a voice similar to the one you want to drive, and throwing away the bits you don't want, is usually a lot quicker than dialling one up from scratch.
The split setup is the first thing you encounter when you enter Edit mode. A graphic appears in the LCD representing the guitar fingerboard and picking area. This graphic is mirrored by the paintwork on the front panel, helping you to make a visual association with the different areas in the LCD. It's actually far more intuitive than it sounds, and defining split zones soon becomes second nature. I can see the attraction of split zones for live performance — although, even then, going any further than a simple bass‑line/top‑line split is bordering on the unmanageable — but I found no practical application for their use in a MIDI sequencing setup.
Far more interesting for the sequencer user is the Axon's Pick Controller facility. This allows you to determine the value of MIDI continuous controllers by your picking position. If the application is not immediately obvious, how about assigning Controller 74 (Low Pass Filter) so that the filter is fully open when you pick near the bridge and half closed when you pick by the neck? Any guitarist can relate to that, as it closely mirrors what happens on a real guitar, and it can therefore be used instinctively to make synth patches far more musically expressive. You're allowed up to two pick‑position controller assignments, with independent limit settings for both, so you achieve inverse assignments (one controller increasing while the other decreases). I have to admit that I couldn't find a musically useful application for this, but I'm sure somebody will!
There are lots of additional facilities, but most of these relate to the use of the internal soundcard in live performance applications, allowing you to do clever stuff with layering and bringing in extra sounds with the hold pedal, triggering drum sequences, and firing off the arpeggiator. It all works very well, but it just isn't what I, and I suspect most SOS readers, actually want to do with this system. I just want to be able to play guitar parts and have them translated into something my sequencer understands.
Axon In Action
Does the Axon deliver what it promises? Well, this is MIDI guitar, so of course the answer is yes... and no. The AX100 is the fastest conventional‑guitar‑to‑MIDI converter I have used thus far, and it doesn't seem to be operating on the same 'guesswork' strategy as the other 'fast' systems. Most of the time it's accurate, and you can choose to use it on a single MIDI channel or with pitch‑bend off. The downside is that it is mounted on a real guitar, and real fingers moving on real strings will always generate odd noises that MIDI simply has no vocabulary to describe. In particular, every time you lift a finger off a fretted note there is a tendency for the open string to sound briefly. On a conventional guitar we barely hear it, but to the MIDI converter it is a note, albeit a low‑velocity, short note, resulting in a short portion of the attack phase of the MIDI voice being audible. Setting the Note On sensitivity higher reduces the likelihood of this, but also compromises delicate playing. The Zeta Systems guitar (probably the most usable system prior to the AX100) allowed you to implement software suppression of open strings to combat this 'lift‑off' effect. Very effective it was too, and I'm disappointed not to see something like it on the Axon. I tried setting up my own 'suppressed zone' on the AX100, using a fret split at fret one, with the lower portion assigned to zero volume, but had no success. That having failed, I resorted to my usual 'hardware' solution, a thick piece of felt pushed under the strings between the nut and the first fret, exerting sufficient pressure to stop the open strings vibrating, while not being thick enough to raise the playing action. It works incredibly well in cleaning up the MIDI output when playing in mid‑fingerboard positions (on any MIDI guitar system), but does mean you can't play anything in first position or tune up using open strings.
Above all, the Axon AX100 is a system you must be prepared to spend some time with. If you take the trouble to optimise the whole setup for the specific voice you want to play (that's what the presets are for) it will reward you with a usable translation of your performance. Bass lines or solo lead lines are particularly successful, whilst chordal parts are, I believe, slightly more compromised. This is not specifically a limitation of the Axon, merely a reflection of the fact that it is much easier for the player to move cleanly between single notes than several notes at once.
MIDI was not conceived with guitars in mind (and vice versa!). The fact that the two can be brought together at all with this degree of success represents a small triumph of incremental development combined with lateral thinking, pushing at the limitations bit by bit. It is not perfect, but I think it is probably the best we will see along this particular road of development. It won't do everything I want, but it will do far more of what I want than anything else I have tried so far. I can do serious MIDI sequencing work with it, and I guess I can pay it no higher compliment than that.
MIDI Guitar Survival Guide
- If possible, don't use your regular guitar for MIDI. Get another one to use for MIDI (it needn't be expensive and really only needs to be able to be adjusted for correct intonation and tuning stability — pickups and electrics are irrelevant). This will allow you to set the action and stringing optimally, and mount the pickup properly, with screws. Proper height adjustment is vital to accurate operation.
- Don't use very light (or very heavy) strings. I found an improvement moving from my normal '009 top E' set to 10s, but going heavier produced a drop‑off in performance.
- Don't ramp up the action on the guitar, thinking it will generate a cleaner note. It will result in more lift‑off noise and false triggering, as the guitar will be generally harder to play. The Axon seems to analyse well at medium to lowish actions, even with a bit of fret‑slap going on.
- Don't sit too close to your computer monitor! Quite a few of us in the MIDI guitar fraternity are beginning to think that the radiated field from the monitor might be imposing noise on the hex pickup's output, making it harder for the converter to 'see' the real signal.
- Try permanently damping the open strings (unless your playing style demands that you use them) with felt or foam rubber under the strings between the nut and first fret. This will stop unwanted short notes being generated when you lift your finger off a string.
- Use a heavier pick. Without changing your style you'll have more weight in the way you attack the notes you genuinely want, allowing you to drop the sensitivity a little, and reducing false triggering.
- Check your tuning frequently using the system's in‑built tuner. You may not hear the guitar drifting out of tune if you're not using pitch‑bend, but you'll be giving your converter a hard time if you're not spot‑on in tune. Make sure your intonation is set up accurately, and if you have a trem system, lock it down out of action, or it will contribute to tuning instability.
Pitch‑Bend & MIDI Guitars
Pitch‑bend facilities are very important to the success of any pitch‑to‑MIDI system. The optimum scenario is undoubtedly one where each string sends its data on a separate MIDI channel, allowing individual pitch‑bend data to be generated for each one. This is fine for live performance, where the instrument will play a single part via six channels of a multitimbral voice module. It is less satisfactory, however, in a typical MIDI recording/sequencing setup. Even in a complex multi‑port/multi‑channel setup, most people still rely on their multitimbral modules being able to cover several different parts simultaneously (after all, that's what they were designed to do) and using up six precious channels for each part is just not practical. Inevitably the MIDI guitarist is forced to deal with transmitting on a single channel and is faced with a stark choice of disabling pitch‑bend and accepting a significant reduction in expressiveness, or leaving it active and risking the audible anomalies generated by conflicting pitch‑bend data being read from each string but being output on the same channel! Bend one note and the pitch of any others sounding at the time will be equally affected. Bend two notes and the results become unstable, with discontinuities in the data generating some horrible out‑of‑tuneness and pitch‑warbling.
Blue Chip have brought some intelligence to this, and the Axon actually behaves very elegantly when transmitting on a single channel. You have a choice of four pitch‑bend modes:
- Pitch Quantize Off (the nightmare 'single‑channel bending' scenario).
- Trigger, where pitch‑bend is effectively off so that only chromatic half‑steps are recognised. If you bend a note far enough you will eventually generate a new Note On message for the note a semitone above where you started, and then another Note On for a tone above if you carry on bending.
- Pitch Quantize On, where bent notes remain at the same pitch until they reach the semitone above (they will then jump immediately to the new pitch, but without re‑triggering, achieving their new pitch via pitch‑bend messages).
- Auto, which allows normal bending whenever you are playing single notes, permitting expressive slides, hammer‑ons and pull‑offs, and so on, but, crucially, locks out pitch‑bend whenever you play more than one note. This is ideal for expressive playing without risk of destructive pitch warble.
- I found Auto mode ideal for bass lines and monophonic melody parts and solos. Trigger mode, with no pitch‑bend, and every slur and hammer being translated into new Note Ons, is really a necessity for piano sounds, while for everything else I found the Pitch Quantize mode to be the most benign.
Neural Net
The AX100 (like the Yamaha G50 system) was developed using a 'Neural Net' system — a technology which allows microprocessors to effectively 'learn' by experience, rather like a brain. Let's get one thing clear: although the manual makes reference to the AX100 being able to "learn your personal guitar playing style", that's potentially rather misleading. This system has actually already done all the learning it's going to do; you just have to hope that part of its development included learning about somebody who plays a bit like you.
While My Guitar Gently Bleeps: Guitar‑To‑MIDI Techniques
There are two principal ways of tackling the guitar‑to‑MIDI problem. One approach is to start from the premise that MIDI is an inherently keyboard‑orientated protocol, designed to deal only with clearly defined, switch‑like events, such as Note On and Note Off messages. An instrument built using this premise would attempt to behave as much like a keyboard as possible, with switches under the left hand to determine the pitch of 'fretted' notes, and usually some form of right‑hand triggering mechanism. Harvey Starr's range of Ztar controllers (review to follow soon, we hope) is probably the leading example of this approach, and it can be very successful for those prepared to make the necessary adaptations in technique. (Casio made a bold attempt at something similar, at a fraction of the cost, with a cheap plastic affair in the late '80s. It works rather well if regarded purely as a programming instrument, let down only by its fixed velocity value).
At the other end of the spectrum, the first guitar‑to‑MIDI systems (and many subsequent ones, with increasing degrees of sophistication and success) were simply adaptations of conventional electric guitars, utilising a hexaphonic (6‑way) magnetic pickup to obtain an individual signal from each string. The pitch of the note played was derived by analysis of the waveform, initially taking two complete cycles, then, on later systems, one and a half cycles, and subsequently a single cycle. The problem with this method is that there is inevitably a noticeable delay on the lower strings, and one which varies with the note played, as the lower the frequency, the longer the time period represented by each cycle.
Somewhere between these two approaches comes a host of innovative one‑offs and hybrid systems — Zeta Systems, Stepp, Synthaxe, Yamaha's 'ultrasonic' G10 and many more. I've played them all extensively over the years, and owned a good many of them. Without exception, each of the MIDI guitar controllers I've tried, whatever their approach to the problem, did something quite well, but no single one of them, in my opinion, ever did enough things well to make it a truly viable instrument. Either they were simply not guitar‑like enough to make the player comfortable (prompting the quite reasonable thought: "If I really wanted to learn an entirely new instrument, surely it might as well be a keyboard?"), or they were plagued by the inherent delay and mistracking of pitch‑to‑MIDI conversion.
Pros
- Fast, accurate MIDI conversion.
- Maintains familiarity of conventional guitar.
- Wide range of instrument interfaces.
- Flexible programming options.
Cons
- No 'open string suppress' facility.
- Requires a clean, accurate playing style to work at its best!
Summary
The Axon AX100 system tracks faster and more accurately than any of the directly equivalent competing systems. If you want to work with MIDI from a conventional guitar, it's the new number one.