Roland GR series emulator? GR300 esp

[phoofk]

Wondering if anyone has any insight as to the architecture of the GR 300 guitar synth, i just got a Fender with a roland pickup and I want to shred on some Adrian belew kind of "reed"-like, infinite sustain sound that sounds like a clarinet or some arab reed instrument, like snake-charming sound.

feel me?

I'm sure the 300 isn't that complex in signal path, but just wondering if this means anything to anyone in terms of what i'm trying to do.

gr 700 i know is based on JX8P, so perhaps there exists something to emulate that already.

aright! thanks!

[Antimon]

phoofk!

You have posted this in a subforum dedicated to Nord Modular synths. Maybe it will get better attention if you post in the "strings and things" forum instead - or a moderator moves this thread there.

Slightly on topic, I don't know anything about the GR300 (I have a VG99), but is that the new Fender with built-in GK mic? Is it nice?

[Tim Kleinert]

Hi,

I dabbled a bit with this topic, however with inconclusive results. The thread is here:
http://www.electro-music.com/forum/topic-32369.html

First of all, you have to be aware that the GR-300 is a totally different beast than any guitar synthesizer that came after it. All later designs work via the MIDI protocol, meaning that the sonic activity of the the guitar strings is converted into digital MIDI note and pitch-bend information (and perhaps more), which then drive a synthesizer module probably optimized for this kind of performance. This approach has benefits and drawbacks. The benefits are that a guitar thus can theoretically drive any kind of MIDI sound generator, and the MIDI data can further also be recorded and manipulated in any way. The drawback is that this conversion introduces noticeable latency, which however was optimized with every new generation -but still is always present.

The GR-300 works totally differently, in that the activity of the vibrating strings actually forms a vital part of the actual sound generation. In a nutshell, the GR-300 features six incomplete sawtooth oscillators (one per string) -incomplete in that they do not have a built-in comparator that will discharge the core capacitor when a certain voltage is reached. So actually, they don't oscillate by themselves at all. They are, however, hard-synchronized to the zero-crossings of the waveform produced by the corresponding string. So actually it's the vibrating string that will trigger the discharge -the string becomes part of the oscillator. Simple, genius. Inevitable frequency-depending differences in amplitude are dealt with simply by hard-clipping the waveform, which produces that characteristic squelched sawtooth sound. This primitive and genius concept makes the GR-300 still unmatched in playing response to this very day. There is literally no latency at all. The drawback is that you are basically stuck with one sound -the GR-300 sound so to speak- which nowadays is more or less owned by Mr. Patrick Bruce Metheny.

Hard-synching an oscillator to an audio source is laughably simple -just patch the audio source into the sync input. In fact, that's exactly what the Moogerfooger FreqBox pedal does, although it adds FM for extra sonic manipulation. However, the FreqBox is only monophonic. The GR guitar synthesizers had pickups for each string. Also, if you want accurate pitch tracking, this simplistic method will not always work, as harmonics will often produce extra zero-crossings in the waveform, resulting in discrete pitch jumps of the driven oscillator . The GR-300 had an extra circuitry to filter out harmonics and ensure that the fundamental frequency is dominant.

cheers,
t

[Tim Kleinert]

Addendum: It just came to mind that the G2 has only 4 audio inputs, so using a hexa-pickup will not completely work.
In terms of pitch-detection on the G2, I (involuntarily) have become somewhat of an expert in this field, having tried out many approaches, of which almost all have failed somewhat .

The best solution I've ever hit upon uses the stock G2 pitch-detection module with some extra circuitry to help differentiate between periodic and aperiodic signals, in order to combat those annoying blips and glitches that occur when the pitch-detector tries to latch on to chaotic transitions.

To be found here:
http://www.electro-music.com/forum/topic-52784.html

[Antimon]

One of the first things I tried when I got my Nord Modular (the old one) was hook up a guitar to the sync input of an oscillator and then use an envelope follower to amplify. Simple mayhem and fun!

[phoofk]

sorry ha -
I am using a guitar that has a GK pickup, then a GK to MIDI converter that is USB. from there I send MIDI to the G2 (yes i forgot to mention, i do indeed own one and am attempting to apply this knowledge gained TO it.)

so really the question is not about guitar related anything, but rather, what is the architecture of the roland synthesizer that was used to playback data gleaned from shitty HEX pickups in the 70s/80s.

clearer?

[phoofk]

sorry, i really wasted your time eh Tim? huge response, huge effort. so touching. but alas, yes, the tracking/pickup/guitar part of the formula is taken care of.

[Tim Kleinert]

[GovernorSilver]

[moylando]

Reviving an old thread here but here's a breakdown of the GR-300 method for those that wish to try to emulate (per string):

Guitar signal enters, is amplified and 1 pole low pass filtered (different -3db point per string), has a selectable attenuator and a diode type clipper which the GR-300 labels "compression" and it basically works as such. Signal is then 1 pole low passed again (different -3db point per string again)

Signal enters a 2 stage bandpass filter which has only 2 frequency point settings. The first has both centered on the open string frequency. The second is centered at the 12 fret frequency, but when this frequency is selected one filter is slightly below and one is slightly above. The frequency point is switched based on the input frequency being above a threshold. This is to emphasize the fundamental and limit extra zero crossings caused by harmonics. The circuitry is much simpler than trying to make a filter which accurately tracks the input frequency exactly.

Filtered signal goes into another 1P LP then opamp with no feedback, which turns it into a square wave. Square wave is then turned into 1uS pulses by 2 one-shot circuits in series (which cause a slight delay). The second pulse syncs a rising sawtooth. The earlier pulse triggers a sample/hold to capture the voltage of the ramp just before reset. This gives an inverted V/Hz value which is linear but higher for lower frequencies. The sawtooth is used as one "oscillator" and is only selected in the "duet" mode. It is always the same pitch as the incoming signal, since it is synced to it.

The derived V/Hz value then sets the level at which a comparator resets a second, more typical, sawtooth oscillator. This allows the oscillator to be tuned to a harmony which will be maintained for any note played. This is the main oscillator signal which is always on in the gr-300. The derived V/Hz value is also fed back to determine the earlier bandpass filter setting, as mentioned.

As mentioned in a previous post, the waves go into some primitive VCAs controlled by the string envelope signal and get clipped to give a waveshape that is slightly different for each note, but always a clipped sawtooth shape with the same max height. A lot of these "simplifications" are made possible by the fact that each string channel only needs to track over a 2 octave range.

Then the string signals are summed together and go into a 4P LPF which can be swept +- by the maximum of all the string envelope signals. This is somewhere between paraphonic and truly polyphonic. You have individual volume articulation per string, but only one filter.

My main complaint about the gr-300 is that it always sounds uniquely like a gr-300/Metheny. It would have been nice if Roland included some other sound sources such as the square wave derived earlier for a more hollow sound, suboctave derived from that square, or the ability to sync the main oscillator and sweep it's frequency based on the string envelope, but it's pretty packed with electronics already and they had to draw the line somewhere. Remember, every additional feature has to be done 6 times. Not a problem in the G2 but more of an issue in hardware.

I haven't tried to duplicate this exact architecture for deriving guitar synth sounds in the G2, but I can't think of any missing elements. It could work.