Here is a breakdown of how I did the tests in more detail. For example, with the GM-70, the guitar sound was recorded at sample number 4470612, and the synth output was recorded at 4473326. So 4473326 minus 4470612 equals a delay of 2714 samples. 2714 samples divided by 44100 samples equals 0.06154. Multiply by 1000 to convert to milliseconds: 61.54. Or a 61.54 millisecond delay for a GM-70 to play a low "E." Ouch! Compare that to 5.8 milliseconds for the GR-300!
For the MIDI only modules, like the GM-70 or GI-10, I used the output of a Yamaha CS6R. It was the factory piano patch, with the attack time set to zero. The CS6R is a few years old, and is a fairly "middle-of-the-road" module. So, when you see that the average response time of the GI-10 is 21.99 milliseconds, this is the total time from when the guitar is picked to when a note is actually heard from the module. I felt this was more of a real world test. Also, depending on the MIDI module you use, you may have a slower or faster response time.
And 20 ms is not a bad response time at all. In 1995, there was a lot of discussion about "MIDI delay." This was a term for the delay between when a MIDI note on is sent to a synthesizer, and the time it takes the synth to respond. I did a test on a Kurzweil K2000R, which was considered to be a pretty good synth in 1995. It took the K2000R around 20 milliseconds to produce a sound from the time that a MIDI note on command was received. To consider that guitar synthesizers like the GR-55 are working in the same time frame is pretty impressive. Perhaps the problem with guitar synthesizers is that the player can hear the note produced by the guitar and perceive the delay in the MIDI sound, even in the small time delay of 20 ms.
I have also included the Roland VG-88 and GR-300. At an average response time of 3.3 milliseconds, you can see that the GR-300 deserves its reputation as the "world’s fastest guitar synthesizer."
For the VG-88, I used polyphonic pitch shifting, with each string being individually pitch-shifted. The result was similar to putting a capo on the third fret of the guitar. In addition to the pitch-shifting, there is the additional guitar and amp modeling.
Unfortunately, I do not have any of the Axon pitch-to-MIDI converters. As of 2012, the company is out of business, which is really unfortunate. The Axon gear was always pretty hard to find in the United States. However, I was able to pick up a Yamaha G50 pitch-to-MIDI converter, which licensed the Axon technology. From what I have learned from searching the web, the Yamaha G50 has the same design as the Axon NG-77. The Yamaha G50 smoked all the competition in the Pitch-to-MIDI speed tests, scoring an average amazing 11.47 millisecond response across all the strings.
Just to make sure I was not imagining things, I did the Yamaha G50 test with four guitars, two with a whammy bar, two with a fixed bridge. I used both Roland GK-2A and GK-3 pickups. And the results were consistent, 11 milliseconds average. If you want fast Pitch-to-MIDI, find yourself a Yamaha G50 or an Axon converter.
I would note that the web pages I referenced indicated the Yamaha G50 would not be a good choice for use with piezo MIDI pickup guitars, since the Yamaha G50 does not have a software update issued by Axon to deal with the different attack waveform of the piezo systems. I would note slight speed differences between the guitars using the G50 is insignificant. I cannot say that my pick attack is precisely the same, or that I struck each note with the same intensity. Either way you look at it, the G50 is significantly faster.
And, many of you may be wondering about the GR-100! It basically has no delay. The GR-100 is six fuzz circuits with a low pass filter and chorus. So, in light of the tedium required to measure the MIDI timing/tracking of all these synths, I skipped the GR-100.