Direkt zum Inhalt

Difficulties with tuplets

Posted

In my last post, I mentioned rhythmic difficulties. I would like to import phrases created elsewhere with certain rhythms that may contain tuplets (and shifted accents).  Attached is an example of a rhythm in quintuplets (inspired by concepts of the rhythmically very rich classical music of South India), which is otherwise very simple and short.

I want to be able to import such phrases as sequences of figures into Synfire, then do Synfire's magic, and -- crucially -- in the end export the resulting music to music notation. In other words, I need to have the proper tuplets still in the notated final result. I accept that I loose some information, like articulations (e.g., the accents in the input score), but I really need to preserve tuplets. 

As an initial test, I imported the short example MIDI file, and Synfire did recognise the correct note durations. As I am primarily interested in the output for music notation, I would prefer results in MusicXML. However, I am seemingly unable to export this rhythm into notation again. When I try to export a MusicXML file (directly from the Synfire arrange window resulting after importing the MIDI file), the MusicXML file seems to contain no notes at all. When instead I try to export the result as a MIDI file, then in the resulting MIDI file the rhythm is altered (in contrast to the music played back by Synfire).

What am I missing here? Thanks a lot! 

Best,

Torsten

PS:  Note that the attached example is only an initial test. In the resulting music, different tuplets and different cross-rhythms will be layered. However, for now I would not use any nested tuplets ((http://klangnewmusic.weebly.com/direct-sound/lets-talk-rhythm-part-2-ne…)), so all note durations are either a multiple of 1/2^n (standard note values), 1/3^n (triplets), 1/5^n (quintuplets) and perhaps 1/7^n (septuplets). Without some rhythmically reliable export, the music would be difficult to restore manually afterwards.  


Mo., 23.07.2018 - 16:46 Permalink

Related question: I understand that the voice leading depends on the metric grid (stronger and weaker metric beats etc). Is it perhaps possible to adapt this grid to the changing rhythm for tuplets, and in particular tuplets with shifted accents? If so, is this a global parameter shared by all parts, or can it perhaps even depend on parts?

(... it might be better to turn off voice leading rules, but I thought I might ask first :)

Best,

Torsten

Mo., 23.07.2018 - 18:54 Permalink

Really interested in what you’re doing. I’d create a custom scale, and then turn off voice leading (set to none in interpretation).

There would be a single chord/harmonic context for the whole piece, with a drone tuned to it.

Mo., 23.07.2018 - 19:56 Permalink

Thanks for your interest.

> There would be a single chord/harmonic context for the whole piece, with a drone tuned to it.

Yes, that would be the Karnatic way of doing things. Instead, as a Western musician myself I currently try to combine fundamental concepts of these two musical worlds (i.e. a rich Karnatic rhythmic language on the one hand resulting from independent gatis and jathis – emphasised for Western ears by supporting jathi accents with durational accents, and some rich though non-traditional harmonic language on the other hand). Not quite sure yet where that leads me :)

My problem is that I cannot get Synfire to correctly export tuplets, but perhaps there is a way that I missed so far. Otherwise, Synfire would be a great tool to help composing such music. 

Best,

Torsten

Do., 26.07.2018 - 01:33 Permalink

Apologies for following this up. Attached please find a *.cognac file that contains the above rhythmic figure and nothing else really. The rhythm plays back correctly within Synfire.

However, I cannot export this music. When I export to MusicXML, I get a very small file that does not contain any notes (seemingly I cannot upload XML files, but you can try exporting MusicXML from the attached cognac file).

When instead I try to export the music as a MIDI file, then I do get a MIDI file with notes, but the rhythm is changed (see the 2nd attachment). Note that the rhythm has not been changed by any other rhythm-related parameter in Synfire, and that Synfire plays the rhythm correctly. 

What am I missing here? I could understand if Synfire does not export quintuplets by design (depending on how rhythmic information is stored internally that can be tricky), but I am quite surprised that the rhythm in the exported MIDI file is clearly different. Does Synfire execute some sort of quantisation when a MIDI file is exported? Can that be switched off somewhere?

Thanks a lot! 

Best,

Torsten

Mo., 30.07.2018 - 17:21 Permalink

Sorry for my late reply. Only now I found the time to dig into this complex topic.

For a rhythmically reliable export with respect to tuplets, there needed to be a way of tagging a figure segment as a quintuplet in advance, for example. That's not currently possible. 

Since the rhythm of a figure can be (and is) altered by other parameters, that information would also get lost rather quickly.

Therefore during export, Synfire has to guess (estimate) where tuplets occur, even with its own output. This makes it very difficult, unless notes are perfectly quantized as tuplets. Even then, the MIDI resolution of 1920 ticks per 4/4 note limits the types of tuplets that can be expressed at all (only those that hit a whole number in that range). For example 1/9 does not work (213,333333).

So yes, MusicXML export of tuplets specifically is challenged. You are probably better off with MIDI export, tagging tuplets in the notation software after import. 

Is it perhaps possible to adapt this grid to the changing rhythm for tuplets, and in particular tuplets with shifted accents?

In theory, yes, but the current version of Synfire doesn't have a user interface for that. It has also not been tested, but looks like an interesting feature for a future update.

Some of the VL strategies associated with "dynamics" respond to velocity, that is, notes with higher velocity than average will be forced to strong notes in the harmonic context. This might do what you want.

When I export to MusicXML, I get a very small file that does not contain any notes

That's probably a bug. I will have a look.

Does Synfire execute some sort of quantisation when a MIDI file is exported? Can that be switched off somewhere?

Yes, and yes. On the "Notation" tab in the instrument inspector, you'll find a couple timing options for export. Quantization is one of them. These options do not affect playback, only export.

Di., 31.07.2018 - 01:40 Permalink

>> Does Synfire execute some sort of quantisation when a MIDI file is exported? Can that be switched off somewhere?

> Yes, and yes. On the "Notation" tab in the instrument inspector, you'll find a couple of timing options for export. Quantization is one of them.

Ah, great!  

If anyone else tries to reproduce this: one way to reach this tab is via the menu File > Export > Settings ..., which shows the global notation export settings in the inspector. Then click on an instrument track title to find the instrument inspector. (It took me a while to find it, hence I thought it might be useful sharing it here – perhaps even for myself in the future :)

I fully understand that exporting tuplets to notation is tricky if note starting points can freely shift in time, but with the exported MIDI file unquantised I can now try to do that with other tools. For example, my old Finale 2012 can quantise the quintuplets of the exported MIDI file perfectly (Dorico at least some of them, while Sibelius struggles, but I can export a MusicXML file with Finale and then open that with whatever scoring software I like). 

> Some of the VL strategies associated with "dynamics" respond to velocity, that is, notes with higher velocity than average will be forced to strong notes in the harmonic context.

Interesting to know. Which strategies are that? In the documentation, I found the following. Did I understand this correctly and are these the only ones?

Adaptive (dynamic): Particularly loud notes are harmonically emphasized.

Adaptive (dynamic and rhythmic): In addition, longer notes are more emphasized than shorter notes.

In that context, what is meant by "rhythmic accents" for the "adaptive strategy", strong beats or durational accents? I assume strong beats, as longer notes are mentioned only for "adaptive (dynamic and rhythmic)", just want to double-check. 

BTW: That these kinds of details can be controlled for each instrument is great! 

Thanks a lot for your helpful response! 

Best,

Torsten

 

Di., 31.07.2018 - 17:07 Permalink

The "dynamic" preset responds to velocity only. Higher than average values will force a stronger note pitch based on context. The average is determined each measure in advance, or per container (don't remember exactly now).

The "dynamic and rhythmic" preset does the same and adds note lengths to the mix, that is, longer than average notes will also receive a boost towards strong pitches based on context.

Quite often a strategy merely alters two or three notes in a four measure stretch, which however have a very noticeable effect.