Quilcom SIM-GUZHENG: Oh pluck it!

[Spogg]

Hello playmates!

Yes, another in my SIM series of synthesiser versions of acoustic instruments. I’d never heard of a Guzheng until tulamide linked an amazing video, and I was immediately fascinated. This zither-like instrument is long, with bridges along the top plate part way along the strings. This means a player can press on any string behind the bridge to provide a pitch shift and/or vibrato. I won’t go into any more detail about this ancient Chinese instrument because it’s in the user guide and background info.

From a schematic point of view you may be interested in my solution. The real instrument is polyphonic of course, but I chose to make each string’s sound generator in mono (blue). That means 24 mono synthesisers free-running all the time. In this case the resting CPU is higher than you’d see for a regular poly synth (even though I made full use of SSE mono4) BUT when you play fast and wide glissandos there’s very little increase. In order to avoid note-stealing causing sounds to terminate early you would need a limit of maybe 96 voices for good glissando performance in a regular poly system. I tried this in my Quilcom PMS and a couple of glissandos maxed out the CPU. I also tested Kong Audio’s free Guzheng (version 1) which uses samples but I could easily make its CPU go higher than mine!

Another disadvantage of using poly is note-bending, which is important for this type of instrument. The scheme I adopted for live playing is to only bend a note when it’s held down. But in a poly system you could have the same not played twice rapidly and get 2 voice channels assigned, one of which wouldn’t bend.

Another thing worth mentioning is that after testing various synthesiser techniques I found that using a tapped waveguide with appropriate pulse excitation for the pluck easily gave the best results. Tapped waveguides use 4 delays and in a poly system each 4 new voice channels would have to assign new memory for the delays, so welcome to Click City! In blue mono4 the memory is assigned at start-up and never changes.
A further advantage is the free-running waveguide will sound slightly different for each pluck, if it’s currently sounding, because the same voice is used and the timing of the pluck will excite the waveguide relative to its state at the moment the pluck arrives (just like on a real string). This method might have application in other percussive synths, so you don’t have to set say 4 voices to avoid rapid note-stealing and thus clicks.

I have a puzzlement that maybe Martin could help with. The musical tuning of the waveguide isn’t accurate and I don’t know why. It was a simple matter to make a per-note or group correction to get the pitches right, but a more scientific solution would be welcome, for possible future projects.

Anyway…
Download the zip here:

https://www.dropbox.com/s/h7wrplr34saug ... 0.zip?dl=0

Here’s my video:

https://youtu.be/SgvpzObTPEg

[Tepeix]

One time i might be sure that i download and test all your treasury !
(I make some safe of the forum that i hope have any file even drop box one but i might maybe do another one to be sure to get everything.. not so practical as the file might be converted..)

Playing this one i imagine to be in a forest.

If i understand well when playing chords the note will be first fast arpegiated ?
Which sound very nice !)

[billv]

Great stuff Spogg.. ....very tranquil...serene music......perfect for Australasia...

[Spogg]

Thanks billv. It was just me messing around to demo the keyswitches and stuff!

@Tepeix: There’s no built-in arpeggiator. You have to play it! It might sound good with an external arpeggiator…

[tulamide]

Behind the scenes:

Here's an important video, because

a) it introduced Spogg to the instrument
b) it shows what the instrument is capable of even in western music

Thunderstruck - AC/DC Guzheng

[Spogg]

Introducing me to the Guzheng was a real gift tulamide!

I’m always acutely aware that so many acoustic instruments can have such a wide range of expression and can be used in unexpected ways, like in the video.
Translating any acoustic instrument into a form that can be played on a MIDI keyboard can be a frustrating challenge (but interesting!). I often have to limit the design to a subset of possibilities available on the real version. I guess this is why commercial sample packs usually have many complete “clips” included because to play or program these short performances would be extremely hard or impossible on the basic sample set.

I got a criticism on YouTube that the sound was “far from a Guzheng” which is disappointing but not so surprising. In my experience imitating a real instrument’s timbre range using synthesis will only ever give a hint towards the real thing. In this case I had to be very careful about CPU use and I would have liked to have added more functionality to the schematic. What I did enjoy very much was playing it though. I didn’t really know before that a pentatonic scale could be so much fun and easy to play!

[billv]

Here's an important video

Thanks Tulamide....what a gobsmacking video....f**king unreal....

[Tepeix]

By arpegiator i would say that when doing chord, each note is fast played one after another.
Like a sort of rake on guitar.

[Spogg]

... I have a puzzlement that maybe Martin could help with. The musical tuning of the waveguide isn’t accurate and I don’t know why. It was a simple matter to make a per-note or group correction to get the pitches right, but a more scientific solution would be welcome, for possible future projects...

I’ve now read a lot of papers about this and tuning can be a problem with waveguides. I’ve found mention of interpolation methods in the fractional delays but only one mention of the effect of the reflection filter, which I find surprising. Any arrangement of reflection filters (often low pass) will introduce a phase shift which acts as a small additional delay in the loop. This can be at least partly compensated by subtracting a fixed number from the delay input. The tuning error will be more apparent at higher frequencies which need a lower delay value, so the flattening tuning error becomes more audible.

Any more information would be most welcome!

[martinvicanek]

Interesting! I’ll take a deeper dive when I return (traveling atm).