' panda audio
MIDI Guitar FAQ

I spent most of the last thirty years designing guitar synthesizers. I developed several generations of devices from the Shadow GTM-6 / SH-075 through the Blue Chip Music / Terratec Axon until the Fishman Triple Play today. Over this period I answered thousands of questions from users who had some difficulties to understand certain features or behavior of their devices. Now, as I have here a web page of my company, Panda-Audio, I want to take the opportunity to summarize this information in the form of an FAQ. I hope I can help users of guitar synthesizers a better understanding of their devices with this information. While these answers are based on my designs, I believe most of the information here is true for Roland or other devices too. I will update this page time to time with questions whatever I recall from my memories, or whatever new questions arise.

Andras Szalay


Q: Why do most guitar synthesizers have an option to send the six strings on six separate MIDI channels? E.g. you can play polyphonically on a keyboard without restriction on only one MIDI channel.

A: When a note is started in MIDI the pitch of the note is determined by the note number of the Note On message. This pitch can be changed afterwards only with Pitch Bend messages. When you play polyphonically on a keyboard and you use the pitch bend wheel, it will bend all notes what you play by the same amount. In opposite, on a guitar it is absolutely natural that the pitch of individual strings can change individually, independently of the other strings. Since in MIDI the Pitch Bend message is common for all notes on a MIDI channel, individual pitch control of the notes for individual strings can be solved only if each string has its own separate MIDI channel. As a painful compromise it is possible to use pitch bend on one MIDI channel as long as the guitarist plays only a monophonic melody, and turn off pitch bend when he changes to polyphony. However, this can lead to contradictions that cannot be resolved in certain situations, e.g. when the pitch of a note is changed, and this note is sustained while another note is suddenly started. In this case, at the start of the second note the first one will have to return to its starting pitch, or it must be shut off early.

Q: OK, I do not bend the strings, so I could not care less. Then can I use my guitar synthesizer on one channel without compromises, yes?

A: No. A very important part of expressive guitar playing is legato. Guitarists very often do not pick every note of a melody, but notes get different accent if they are picked, or if they are played with hammer-on or pull-off. A guitar synthesizer can emulate this also with Pitch Bend. Although a special Legato Mode is part of the MIDI specification, most synthesizers did not implement it, so the safe way is to use Pitch Bend instead. Then we are at the same point as in the question before.

Q: OK, I do not use hammer-on and pull-off, I pick every note. Then I really do not need Pitch Bend, so I can use one channel only, right?

A: No. There are at least two important arguments why Pitch Bend should be used even if you pick every note.

  1. Coordination of left and right hand.
    If you play an ascending scale on a string, theoretically you should press the string to the fret with your left had finger (I mean this for right handed guitarists) exactly at the same time as you pick it with the right hand. Of course, this is only theory. In practice this can never be coordinated perfectly. If you are too early with your left hand, then you get a hammer-on to the new note first, and then a little bit later the note is picked. Even worse if your left hand is too late, then there is a chance that you get a damped note, because your finger already touches the string, but does not press it to the fret yet. To avoid the second situation which is much more annoying, guitarists play a little bit ahead with the left hand compared to the right. The question is how much, and how stable is this difference. When I analyzed John McLaughlin's test recording, it was amazing to see that pitch changes (left hand) were always just 5-10 msec ahead of the pickings (right hand). By average guitarists this can be 10 times more. Now, when the guitar synthesizer is sensing a hammer-on (which is not intentional in this case), and Pitch Bend is enabled, then it follows the pitch in a natural way, and then when the new note is picked additionally, it starts a new note at that pitch. In opposite, if Pitch Bend is disabled, then during the hammer-on phase the guitar synthesizer does not know that soon a new pick is following, so it starts a new note with a low velocity value. But very soon the new picking is coming (with a higher velocity value), and this results in an annoying double trigger.
  2. Unintended pitch change of the sound.
    When you pick a string, at the very beginning the pitch is always more or less higher than the intended pitch. I mean "intended pitch" here the pitch of the string as it is sustained for several seconds. This is where you tune the string to your reference. There are several physical reasons why the pitch at the very beginning is higher than the sustained value, sometimes by more than a semitone.
    One of the reasons is well known in classical physics: the pitch of a string vibration has a steady frequency only if the amplitude of the vibration is negligible to the length of the string. If it is not negligible, then the vibration causes an increased average tension in the string, and higher tension means higher pitch. This effect depends on two factors:
    1. the tension of the string: thin strings have lower tension for the same pitch, so can be played with much higher amplitude, thus the pitch deviation is higher as well.
    2. the harder you pick, the higher is the deviation.
    There is another reason for unintended high pitch in the attack. The short picking pulse that started from your pick is propagating towards the point where you fretted the string. However, if the pulse is large (you picked hard), and the string deviation is large (light gauge string), and the string height is low, then the pulse slightly hits the fret that is one higher than the one that you fretted, simply because the string is so close to this fret. This results a reflection from this fret that appears in the first vibration cycle as a one semitone higher note. This happens only in the very first cycle of the vibration, since the amplitude of the pulse gets very much smaller even by the second reflection.
    It is of course natural, if you do not get this problem with a Roland, since it does not recognize pitch from the first cycle anyway. Also, it is good to know that if you turn the pick sensitivity to finger picking, it will skip recognition in the first cycle, so this problem will probably not happen. With this knowledge, here are some tips how to eliminate this problem:
    1. First and most important: Even with a piano sound it is not necessary to play in Trigger mode. If you use Quantize to semitones then you get just as well quantized pitch like in trigger mode, but these glitches at the very beginning of the note will not be audible, if the synth has a good and fast response. Of course, this works really well only in MONO mode, on six MIDI channels.
    2. Use medium or hard gauge strings, no light or extra light.
    3. Adjust the string height to be as high as you can tolerate.
    4. Play softly, do not pick hard.
    5. Play with the tip of the plectrum, not with the flat part
    6. Finger Picking Mode, if the slightly slower response does not hurt.

Q: When I use Trigger Mode and I play a hammer-on, then I understand that I get a new note, but why is it slightly delayed?

A: See the first part of the previous answer about left-right hand coordination. In order to avoid a very annoying double trigger, when the system recognizes a hammer-on while in Trigger Mode, it will intentionally delay the new note, waiting a little bit if a new picking is coming or not. It will trigger a new note based on the hammer-on only if the picking does not come in the expected timeFrame.