Unwanted Swing When Using Tremolos inside of Tuplets

[TeAiris]

Use Case: Percussion Notation

Currently, when using a tremolo inside of a tuplet. The playback included a swing feel for any rhythm division.
The main goal of the tremolo is to evenly divide the played beat.
This is not to be confused with adding \tf etc...

Is there a way to disable the triple/swing feel when using a tuplet?

Here is my example for why this is important.

Intended notation and sound, created with flat.io

Test Flam 5.mp3

Rendered Alphatex:

:8 C4 {gr} C4{ac}.8{tu 3 tp 1} C4 {tu 3 tp 1} C4 {tu 3}
:8 C4 {gr} C4{ac}.8{tu 3 tp 1} C4 {tu 3 tp 1} C4 {tu 3}
:8 C4 {gr} C4{ac}.8{tu 3 tp 1} C4 {tu 3 tp 1} C4 {tu 3}
:8 C4 {gr} C4{ac}.8{tu 3 tp 1} C4 {tu 3 tp 1} C4 {tu 3}

While the notation does import correctly using the new importer, the sound is off due to the swing.

A work around is using 6/8 or other compound meters to write the same things without tuplets. But the issue is, tremolos in tuplets will happen very often in a 4/4 time in percussion music (at least marching percussion).

This leads to one final thing. There is no way to say ♩ = ♩. (quarter note equals dotted quarter note at the meter change)
This makes it so that the tempo flow stays the same when going from 4/4 to 6/8.
Without it, to keep the same feel, the tempo has to be multiplied by 1.5. (ex. 4/4 Tempo 120 = 6/8 ♩ = ♩. Tempo 180

I think these are the final two things that will make alphatab perfect for marching percussion music.

Also, thanks for the support and upgrades to tremolo!!!!! It works flawlessly! I love the ability to dial in the buzz speed, which is exactly how a percussion would adjust based on tempo.

[Danielku15]

The "swing" style is the result of a bug in the calculation of tremolo durations. We do not respect tuplets in this area as we should.
Added #2493 to track this problem.

This problem was very likely always in place.

[TeAiris]

Thanks for promoting this to an issue.
I reviewed you findings and want to clear one thing up in your provided image.

When adding a tremolo to a beat that is part of the tuplet, we should still use the original behavior. The line is just an exponential division sign.

Below I will provide the link to notation and audio to show the tremolo version and expanded version for clarity.

https://flat.io/score/696326d5821bf0f9a6793976-copy-of-test-flam-5?sharingKey=032628b58222362cf4434c9f04a9f66382fcb69f4c8f0e876542239a6ba4a586a1f1f4680fb0b8835a36e5ddfe7b0c55a1bf183c6c5a1ce022df5fdfa312f0de

For example.

One line on a quarter note means you play 2 even beats within the same space. (2 8th notes).
One line on an eighth note means you play 2 even beats within the same space. (2 16th notes).
This is the same within any tuplet. So if we have an 8th note triplet group, and the first 8th note has a one line tremolo, then the sounding result is 2 16th notes, and 2 8th notes. ( with the 3 over the entire group)

We can't use the idea that an 8th note with a one line tuplet with a 3 on top splits to 3 notes in this case. If we wanted that, we instead use a 16th note triplet group.

I'm curious what resource that image came from. While i'm sure someone has used that notation to shorthand notation, it would cause many issue when used in conjunction with tuplets.

Please let me know if I can provide a better explanation to be sure this implementation goes smoothly!

[Danielku15]

I'm mainly taking "Behind Bars" as a reference on such things. It is quasi the bible of music notation and typically the reference for most established industry practices.

The screenshot is just a small slice of what's mentioned in the book. I still have to put things into exact context to understand what Elayne is describing there . There are some more paragraphs describing the general principles around tremolos.

From what I've seen and what I learned I am on the same page with you: simply reduce the duration according to the tuplets.

[TeAiris]

Thanks for that reference! I'll look into it more myself so I'm more familiar with similar practices.

I did some digging in the source and found a very simple fix for now! I'm not sure how it affects other code in the pipeline but it at least works for the player.

The only change is in MidiFileGenerator.ts

Here are the changes I made so that tremolos are considered within tuplet timing.

Inside of the function generateTremoloPicking() change this:

const tuplet = note.beat.tupletGroup;
tpLength *= tuplet.denominator / tuplet.numerator;

to this:

let tpLength = MidiUtils.toTicks(note.beat.tremoloPicking!.duration);
// apply tuplet scaling if this beat belongs to a tuplet
if (note.beat.tupletGroup) {
tpLength *= note.beat.tupletDenominator / note.beat.tupletNumerator;
}

Now it sounds as intended! at least for the notation I previously had issues with. I'm do more rigorous testing overtime.

Simple notation used for testing player output:

C4.8 * 8|

C4.8 * 2 C4.8 {tp 1} * 2
C4.8 * 2 C4.8 {tp 1} * 2 |
C4.8 * 4 C4.8 {tp 1} * 4 |
C4.8 C4.8 {tp 2} C4.8 C4.8 {tp 2} C4.8 {tp 1} * 4 |

C4.8 {tu 3} * 12 |

C4.8 {tu 3} C4.8 {tu 3 tp 1} C4.8 {tu 3}
C4.8 {tu 3} C4.8 {tu 3 tp 1} C4.8 {tu 3}
C4.8 {tu 3} C4.8 {tu 3 tp 1} C4.8 {tu 3}
C4.8 {tu 3} C4.8 {tu 3 tp 1} C4.8 {tu 3} |

C4.8 {tu 3} C4.8 {tu 3 tp 1} * 2
C4.8 {tu 3} C4.8 {tu 3 tp 1} * 2
C4.8 {tu 3} C4.8 {tu 3 tp 1} * 2
C4.8 {tu 3} C4.8 {tu 3 tp 1} * 2 |

C4.8 {tu 3} C4.8 {tu 3 tp 2} C4. 8 {tu 3}
C4.8 {tu 3} C4.8 {tu 3 tp 2} C4. 8 {tu 3}
C4.8 {tu 3} C4.8 {tu 3 tp 2} C4. 8 {tu 3}
C4.8 {tu 3} C4.8 {tu 3 tp 2} C4. 8 {tu 3}|

C4.4 r r r |

Sounds perfect!!!

[Danielku15]

@TeAiris

With #2496 I fixed the problems of the incorrect tremolo durations and improved some other bits around your example. But as things are not yet "exactly" as you might expect I thought of pinging you with some further details 😉

Regarding the exact sound replication it might be tricky to achieve exactly the same in alphaTab due to the grace notes in your example. The "theory" behind grace notes is not clearly defined and often an implementation detail that applications handle individually. Specifically:

1. how "long" are grace notes played has no exact definition (or: how much time do they steal from the previous/next note).
2. how are they visually shown despite their "shortened" duration has only some recommendations.

The logic in alphaTab is hard-coded to:

• Visually:
  • 1 grace note: 8th note with flag
  • 2 grace notes: 16th notes with beams
  • 3 grace notes: 32nd notes with beams
• Audio:
  • 1 grace note: 32nd note
  • 2 grace notes: 64th note
  • 3 grace note: 128th note

The visual behavior is same to what Guitar Pro does and what is also documented in Beyond Bars as an accepted variant:

To get closer to the sound of Flat.io I think the audio aspect in alphaTab would need alternative ways of specifying the grace note play directions.

Flat.io, MuseScore, SoundSlice, Guitar Pro, all have their own application specific behaviors there.

---

Due to the vague definition of grace notes, you might want to consider alternative ways of writing the notation so that audio and visual aspects are really as you expect.

For adaptions in the audio of grace notes in alphaTab, you could also open another discussion here so we can dig deeper on that need.

[TeAiris]

Wow that was fast! Thanks again.

Also I've experienced the same when using different notation programs. Rhythm is always a tricky thing, especially when its pushed trying to interpret percussion rhythms.

In my experience, when playing grace notes with multiple notes, you space it evenly.

So in the case of a 3 note grace note, the timing is like a triplet (or 3 note tuplet) before the beat.

In percussion, these would all be considered ruff's. (before beat grace notes)
This time would also be the same if we view them as drag's in percussion (which are on beat grace notes).

Having this flexibility in alphatab would be awesome and definitely provides an edge over what some notation programs offer. It's usually locked away in the code. But, sometimes you do get an offset variable where you can adjust the "tightness" of the grace notes to the primary note, which gives room for defining the interpretation.

I can open another discussion later to reiterate with examples but I think your solution is solid!