by Paul White, Music Technology Magazine, April 1987
Paul White, who's owned almost every Roland guitar synth built, puts the new GK-1
synth driver and GM-70
"brain" under test. Is this the first pitch-to-MIDI system that really works?
BACK IN THE LATE 1970s, when Roland first entered the guitar synth market, you had to buy a complete system comprising a special guitar and a synthesiser with a control interface built in - the GR-500. Only access to the outside world was via a monophonic CV and gate interface, and the package cost rather more than a complete system based round the new GM-70. Due to limitations of the design, the instrument never enjoyed much popularity - though I must confess that I bought one.
Its successor, the GR-300, is probably still the most playable guitar synth ever built, but is non-programmable and produces a very limited range of sounds. Nevertheless, I bought one of those, too.
With the introduction of MIDI, it seemed logical to produce a guitar-to-MIDI interface rather than just a self-contained guitar synth, and Roland's GR-700 filled both needs by having an onboard polyphonic synth (based on JX3P electronics) as well as a MIDI output to drive other MIDI-equipped synths. The attraction of this type of package is that you can, in theory at any rate, just plug it in and play. However, due to the special guitar needed and the imperfections in the system, only the most serious guitar-synth advocates used it.
Before seeing what the new GM-70 offers in the way of facilities and performance, and how it differs from all of the above systems, it may be helpful to delve briefly into the operating principles of guitar synths.
BY THEIR VERY nature, guitars are not the ideal devices to control synthesisers; there are simply too many nuances of guitar playing - such as pitch-bend, hammer-ons, harmonics, damping and vibrato - that have to be translated accurately into MIDI information. Some systems (such as the SynthAxe and Stepp) attempt to make the conversion by using the strings and frets as electrical switches, but this can make bending and hammering a problem, and precludes the use of harmonics or damping unless the design is very complex; both the Stepp and SynthAxe designs resort to elaborate computer-controlled systems to extract the required performance information.
Roland, however, have always taken a more direct route - that of converting the pitch of each string, picked up via a hexaphonic pickup, into some form that can be used to control a synth directly. This type of pickup is necessary because each string needs to be processed separately if the electronics are to have any chance of extracting the right pitch. With the early models it was control voltages that were derived; now digital means are used to convert the string pitch to MIDI codes (with the attendant note-on, note-off and velocity information).
The problem with the pitch-tracking method is two-fold: first, the guitar string frequency is indeterminate at the time of picking, and only settles down into a regular pitch after a short delay; second, the frequency content of a guitar string is difficult to analyse, as there are high-level harmonics present which can mask the fundamental frequency. In early systems, this resulted in the note produced by the synth wavering and/or jumping octaves as the electronics tried gamely to follow the pitch.
What's more, the tracking circuitry can't establish the pitch accurately until it has followed one or two cycles of the waveform, so there is always a slight delay between picking a string and the synth note being triggered. And though this delay has been reduced by ingenious circuit design, even the best pitch-tracking systems may be too slow for virtuoso playing while using percussive sounds such as harpsichord or piano.
HAVING SET THE scene, let's take a closer look at the system under scrutiny here, the Roland GM-70. This is a rack-mounting processor with no sound-generating facilities onboard. Pitch extraction is implemented by means of a 16-bit microcomputer combined with specially developed LSI chips to improve speed and accuracy.
It has a sophisticated MIDI implementation so that different combinations of strings can be routed to different synth/sampler voices, or even totally different instruments if needed, and it can store MIDI control parameters in one of 128 programs. Programs may be called up from the front panel, from an optional footswitch system (using the RRC input on the rear panel), or via program change commands received over MIDI. A multi-function plasma display is used in conjunction with up/down buttons and patch edit buttons for programming, tuning, and to display program information during performance.
The GM-70 interfaces directly with the Roland GR-300 or 700 series guitar controllers, but it can be used with virtually any guitar with the addition of the GK-1, Roland's synth driver. This device consists of a hex-pickup which is mounted close to the guitar bridge, and a tiny control box which is fitted (in most cases) to the guitar's strap peg. The pickup comes with a comprehensive mounting kit, which is just as well, as it's important to follow the advice in the handbook about pickup positioning and the spacing from the strings.
The GK-1 control box accepts a multicore lead, provided with the kit, which plugs directly into the GM-70. The box has four rotary controls, the first being for setting overall volume and the second for adjusting the balance between synth and guitar sounds. The guitar level is only affected if the guitar's regular output is routed via the GK-1 to the GM-70, as this contains the voltage controlled amplifiers necessary to do the job. In the event that your voice units don't respond to MIDI level information, don't worry; there are two channels of VCAs accessed via jacks on the rear panel of the GM-70, which can handle the output of your synths to give the same result.
The last two controls are in effect assignable controllers which can be set up via the GM-70 to vary such parameters as vibrato depth, filter sweep and so on. If you use a G-303 guitar as I did, two of the existing controls, normally used exclusively for vibrato depth and filter control, perform the same functions. I also noticed that the three-position selector switch on the G-303 could be used to switch between octaves depending on how the parameters to the external synth were set.
Now, since the GM-70 makes no sound on its own, you have to plug in a suitable MIDI synth or sampler before you can use it; I used two synth modules - a Roland MKS50 and a Yamaha FB01. Before you can start playing, you have to perform a once-only calibration ritual to set the individual string sensitivities to complement your string gauges and playing style. A bar-graph style level meter on the GM-70 makes this a fairly simple task, and the adjustments themselves are made using six presets, one per string, located in the GK-1.
If you're using an existing Roland guitar controller, it's necessary to perform this same calibration using the presets found under the back-plate of the instrument. I got the best results by setting the controls so that each string, when plucked fairly hard, caused the meter to go into the red slightly. This procedure is important because any synth patch that responds to picking intensity (or velocity, as keyboard types like to call it) will sound unevenly if the setup is wrong. The guitar also uses a threshold system to define a note-on command, so picking a string very gently, below the threshold you have set, produces no sound at all. Then again, if you set the sensitivity too high, strings will sound if you brush lightly against them or when you unfret a string.
All this might sound like a bit of a chore to accomplish, but you only have to do it once, and it's not that difficult. Once set, all you have to do is tune up properly - and help is on hand here...
The GM-70 presets the pitch of middle A at 442Hz, but this can be reset very easily using the edit controls. Then you simply go into tune mode and pluck a string. The display shows two zeros, one of which moves from one side to the other as you tune either sharp or flat. When one zero is directly above the other, the string is in tune. And, as the GM-70 knows which string you are playing, you don't have to switch pitches when you want to tune another string. Again, it's vital to tune the guitar in this way rather than trying to retune external voice units to match the guitar. This is because the circuitry has to decide at what pitch to change from sending a code for say, C, to sending a code for D. If the pitch your guitar is tuned to is near to this crossover point, you could find the slightest string bend causing semitone jumps.
But enough of this. What I want to get down to is how well the system works.
AFTER THE INITIAL setting up, I was immediately impressed by the tracking accuracy of the GM-70. I was using my old G-303 controller with a fairly ancient set of strings, and the only adjustment I needed to make was to the string sensitivity. There was none of the indecisiveness that plagued earlier synths - the pitch was rock steady and it was difficult to get it to misbehave unless the strings were picked unreasonably hard. Even harmonics on the twelfth and seventh frets tracked OK.
You do have to develop a fairly even picking style though, or you might miss notes by playing too softly; this doesn't take too much getting used to, but it's something guitar players unused to such instruments have to be aware of.
So what's the catch? If this system is so good, why do Stepp and SynthAxe bother at all? Well, it's all down to the age-old problem of delay.
If you're unused to guitar synths, you might be tempted to set up a brass patch, hammer into a few Van Halen runs, and then complain that the synth isn't keeping up with you. That isn't delay - that's just a lack of understanding of what's going on. A real brass instrument may well have quite a slow attack time, and if anyone was physically capable of playing guitar hero licks on a tuba, then the real instrument would not be able to keep up for the same reason. The moral is, if you set up a brass sound, play it and phrase it as a brass player would.
On the other hand, a harpsichord setting should follow your fastest playing style, because the voice has a very short attack time. If you try this on the Roland GM-70 system, you'll encounter a short but perceptible delay which is disquieting at first, and can play havoc with your timing on complex passages. Playing the guitar's natural sound simultaneously with the harpsichord synth patch, you'll notice that the harpsichord sound appears almost like a very fast slap-back echo of the guitar sound.
To find out just how much delay was present, I delayed the guitar sound using a Klark Teknik DN780 delay/reverb processor, and adjusted the delay time until the guitar and synth voices were playing together. I repeated this several times to make sure there were no mistakes, and the result came out consistently at a delay of 38mSecs. The delay was the same for the top and bottom strings, and remained the same using both the Roland and Yamaha expanders. As these expanders produced no appreciable delay when run from a MIDI keyboard, I have to assume that the bulk of the delay is generated in the pitch-to-voltage converter section of the GM-70.
How serious is this in practice? If you play using the synth and the guitar sound together, your mind locks onto the guitar sound, so timing isn't a problem. And slow-attack sounds are fine because the 38mSecs is insignificant compared to the envelope profile of the sound. But faster, more percussive voices are a problem and, though practice helps you to compensate for the delay, it can still subtly change the feel of what you are playing - after all, a change of timing of only three or four milliseconds on a snare drum is enough to alter the feel of a rhythm track.
In the studio, the delay may be less of a problem, depending on how you work. If you monitor a mix of synth and guitar sounds, you can get your timing right, even though everything is 38mSecs late. And with a versatile MIDI sequencer on the end of your system rather than a tape recorder, you might easily be able to correct this at the editing stage. If, on the other hand, you are going directly to tape, you could turn the tape over when you've finished, and re-record the synth part onto a spare track via a wideband, low-noise digital delay set to 38mSecs. When the tape is played normally, the new synth track will be in exactly the right place.
How does the delay affect your live performance? Well, to get things into perspective, 38mSecs is about the same delay you would get if you played with your amplifier 40ft behind you.
At this point I must make my apologies for dwelling on this delay business, but it will be that factor more than anything else that will make a player decide to purchase the system or not. Personally, I'm very excited by the possibilities. But there are some curious manifestations that are nothing to do with the GM-70, but are connected with the MIDI system itself (after all, MIDI was really designed with keyboard instruments in mind).
THE PROBLEM REARS its head when the subject of pitch vibrato or note bending is examined. If you control one synth in MIDI Poly mode, the synth expects to see bend information coming from the bend wheel (or whatever source) and applies this bend to all notes being held down. The guitar, on the other hand, is normally under no such restriction - it's actually quite common for guitar players to add vibrato or bend to one note while other notes are sounding unbent. In Poly mode, this doesn't translate too well. What you get is a series of semitone jumps rather than a smooth change when you do a bend, or on some systems you may get a true bend when you play only one string, but get the semitone steps again if more than one note is triggered at any one time.
An elegant way around this is to use a voice expander that will allow you to assign one voice to each string on separate MIDI channels using MIDI Mode 4 (for more info on this, see the past two instalments of 'Getting the Most from Mono Mode' in MT Jan/Feb '87, which dealt with MIDI guitar controllers). You choose a MIDI channel for the first string, and the other five strings are assigned to the next five MIDI channels up. This way, each string behaves independently of the others, so it should be possible to use all your normal vibrato and bending techniques - though you do have to set the bend range of the external synth or sampler to match your guitar pitch; the manual suggests that the voice unit's maximum available bend range is the one to go for, and that this value should be matched on the GM-70.
Another advantage of using Mode 4 is that you can program the voices to be slightly different on each string to simulate the changing timbre of a natural acoustic instrument. You can even, if the situation demands it, put a bass sound on the bottom couple of strings and bagpipes on the top four.
Well, that's about as far as I got with the GM-70's MIDI implementation, and though this may well be as far as many musicians go, it doesn't really do justice to the software within the system. Our in-house MIDI expert, Simon Trask, has been busy deciphering the implementation more fully. Over to you, Simon:
"As indicated above, the GM-70 allows you to select Poly or Mono modes for MIDI transmission. Poly mode assigns all the strings to one MIDI channel (1-16) while Mono mode (otherwise known as Mode 4) assigns the strings to six consecutive MIDI channels (basic channel 1-11) ascending from strings 1-6.
"But Roland have gone one better by allowing you to layer Poly and Mono modes using up to four of what they call 'Branches'. Perhaps the best way to understand this approach is to think of four sets (Branches) of six strings overlaid on one another. For each Branch you can select Poly or Mono mode transmission, while MIDI transmit on/off, MIDI patch number (1-128) and MIDI transposition (+/-36 semitones) are selectable for each string in a Branch.
"The GM-70 doesn't allow any MIDI channel to be assigned to more than one Branch. So if Branch A, for instance, uses channels 1-6, Branch B (or C, or D) must use channels above 6. Thus you can layer Mono mode twice, which uses 12 channels (2x6), leaving you with the option of using Poly mode for the other two Branches (2x1).
"As Paul has already mentioned, if you want to use string bending or glissando, then Poly mode is of limited use. This is because the GM-70 won't send MIDI pitch-bend information derived from your guitar-playing in Poly mode if more than one note is sounding; instead, if a bent string (or string glissando) reaches the next semitone up, the system will trigger that note automatically. The purpose of this appears to be to prevent other notes from being pitch-bent too, as would normally happen in Poly mode. In fact, the GM-70 will transmit pitch-bend quite happily in Poly mode if only one note is sounding. No such problem exists with Mono mode, however, because each string is sent on a different MIDI channel.
"The GM-70 also provides you with some controller assignment facilities. These are global in operation, ie. neither channel- nor Branch-specific, which places some limitations on the way in which you can use them - but then, by the same token, it's also easier to map out your controller settings.
"Using the GM-70 in conjunction with the GK-1, you have two definable controller knobs (already mentioned by Paul), while the GM-70 itself provides one footswitch and two foot controller inputs, and the FC100 controller board has a further two assignable footpedals. These can be assigned to any MIDI controller from 1-95, with foot-switches setting minimum or maximum values in the case of continuous controllers. This flexibility means that you can alter such parameters as volume, balance, panning, sustain, sostenuto, soft pedal and chorus depth on your MIDI instrument(s), all in real time. In addition, pedal and rotary-knob controls can be set to generate aftertouch, pitch-bend, octave up, patch-shift up or patch-shift down.
"The GM-70 has 128 onboard memory locations (organised in Group/Bank/Number format) which allow you to store all Branch and controller settings. These programs can be selected from the GM-70's front panel, the FC100 foot controller board, patch up/down footswitches, or MIDI program-change commands (the latter via MIDI In on the GM-70, of course).
"You can transfer program data between one GM and another, or between your GM and an external storage system (such as Roland's own MC500 sequencer) via System Exclusive dump. This can be a bulk dump of Group A, Group B or System data (ie. control assignment, MIDI receive channel, master tune) separately or all together."
WHEN IT COMES to guitar-to-MIDI converters that use the pitch-tracking system, the GM-70 has to be near the top of the pile. I'm not pretending that it's perfect - the delay can be a problem. But the system will follow bending, vibrato and hammering quite faithfully without musicians having to make too many compromises in their playing style.
The GM-70 is streets ahead of the last generation of guitar synths, in which I include Roland's own GR-700, and the MIDI side of the system offers tremendous flexibility to the player with imagination and enough spare cash to buy a few good MIDI voice units.
And considering what is on offer, the price is certainly attractive. So much so, in fact, that I'm seriously considering adding a third guitar synth to my collection.
Prices GM-70, £755; GK-1, £195, including VAT
More from Roland UK, Great West Trading Estate, 983 Great West Road, Brentford, Middx. 01-568 4578
Copyright 1987 Music Technology Magazine