Encore - Swing playback

[polarbreeze]

It's all well and good to say, "well, lets just swing the eighth notes", but the machine, not having our innate musician's understanding, needs to decide which ones to do what to.

It's simple - for each pair it lengthens the first one and shortens the second one.

[Doug Kerr]

It's all well and good to say, "well, lets just swing the eighth notes", but the machine, not having our innate musician's understanding, needs to decide which ones to do what to.

It's simple - for each pair it lengthens the first one and shortens the second one.

No, that's what we do.

But how does the program decide what is a "pair" (which was really the essence of my presentation, which you struggle so mightily to trivialize)? And what if there is no pair, but a solo note to be "swung" (as brother "q" points out can actually be of interest)?

For example, which of these would be "lengthened" and which "shortened"?

Three-eighths.gif (248 Bytes) Viewed 1919 times

The answer is that the second would be lengthened and the first and third shortened (in the sense we associate with swing adjustment).

Now try this one (and here I haven't hidden the context):

Eighths_02.gif (412 Bytes) Viewed 1912 times

Here of course the first one would be shortened and the second one lengthened.

But you will say, well, yes, but that's not what we mean by a "pair of eighth notes".

Sure. But why? Well, because the first one does not end on an "and", but rather on a beat.

And in fact that is how Encore "reasons". But exactly what does that mean in terms of an algorithm (it after all does not understand the spoken phrase, "on the beat")?

And that is what I described.

Doug

[Doug Kerr]

Further to the matter I discussed in my "report" regarding the "eligibility" of notes to have "first note" or "second note" swing adjustment applied.

Suppose that for some weird reason, we had this notation (in 4/4 time):

Swing_offset.gif (923 Bytes) Viewed 1914 times

(Yes, this measure is slightly overfull.)

Now, as a musician, looking at that, and if your habit was to "swing" pairs of eights notes, would you swing these two eighth notes? ( I would not presume to volunteer the answer.)

Now, would Encore (with Swing Playback ON and the "basis note" set to eighth note) swing either of these two eighth notes?

Answer: No; they do not meet the criteria evidently used by Encore to make that decision. The "face end" of the first eighth note does not fall on an "and" count in 4/4 time, nor does the face start time of the second eighth note. They are off only by a 128th note's time, but it turns out that a miss is as good as a mile.

I do not suggest that the above is a circumstance we are likely to encounter; it is a way to illustrate what I was investigating, and reporting on.

I'm sure the answer would have been obvious to the discerning Encoriste.

Doug

[polarbreeze]

Eighths_02.gif

Here of course the first one would be shortened and the second one lengthened.
But you will say, well, yes, but that's not what we mean by a "pair of eighth notes".
Sure. But why? Well, because the first one does not end on an "and", but rather on a beat.
And in fact that is how Encore "reasons". But exactly what does that mean in terms of an algorithm (it after all does not understand the spoken phrase, "on the beat")?
And that is what I described.
Doug

Encore does indeed understand the term "on the beat" and in fact it uses that concept when it makes decisions for the "beam on beat" function. In fact, it seems to me the logic about how to pair notes for swing treatment is the same as the logic for how to pair notes for beam-on-beat.

[polarbreeze]

Further to the matter I discussed in my "report" regarding the "eligibility" of notes to have "first note" or "second note" swing adjustment applied.

Suppose that for some weird reason, we had this notation (in 4/4 time):

Swing_offset.gif

(Yes, this measure is slightly overfull.)

Now, as a musician, looking at that, and if your habit was to "swing" pairs of eights notes, would you swing these two eighth notes? ( I would not presume to volunteer the answer.)

The convention is that if there's notation other than, literally, pairs of eight-notes (or note-rest pairs) starting directly on the beat, then the as-written notation overrides the instruction to "swing". So the answer to your question is "no".

[Doug Kerr]

Hi, p,

The convention is that if there's notation other than, literally, pairs of eight-notes (or note-rest pairs) starting directly on the beat, then the as-written notation overrides the instruction to "swing". So the answer to your question is "no".

Makes sense to me. Thanks.

And nice to know that Encore honors that. (Now what it does after it makes that decision is another matter altogether!)

Doug

[q]

Encore does indeed understand the term "on the beat" and in fact it uses that concept when it makes decisions for the "beam on beat" function. In fact, it seems to me the logic about how to pair notes for swing treatment is the same as the logic for how to pair notes for beam-on-beat.

Indeed, to some extent Encore understands the concept of "on beat" — hard to tell how much though. And Beam on Beat may not be the best measure of Encore's intelligence in this regard:

http://www.gvox.com/forum/viewtopic.php?f=18&t=2089&p=p11729

http://www.gvox.com/forum/viewtopic.php?f=13&t=2449&#p13803 [restricted area]

q

[Doug Kerr]

Hi, p,

Encore does indeed understand the term "on the beat" and in fact it uses that concept when it makes decisions for the "beam on beat" function. In fact, it seems to me the logic about how to pair notes for swing treatment is the same as the logic for how to pair notes for beam-on-beat.

I meant it does not understand the verbalism, "beam on the beat". It may fulfill a concept that we describe that way.

But regarding its fulfillment of that concept it does it in the way I described. Any attempt to condense or popularize that definition is risky.

Note that, if we take "beat" to have its usual meaning (four of them to a 4/4 measure), then, depending on the setting of what I call the "basis note", the criterion on which Encore makes its decision about the application of swing may not be "on the beat" but rather "on the wink" (where I have been hoping to get an acceptable musical terms for the counts of a pattern of smaller notes).

In any case, (as you well know, having read my description), Encore's decision is not made (we'll assume an eighth note basis note for convenience and pertinence) based on the first note beginning "on the beat", but rather on its ending on the "and" (that is, an eighth note after a beat). And the decision regarding a "second note" is based on its beginning on the "and" (even if there is no preceding "first note".

So in this example, with the basis note at eighth note, both notes would be given swing adjustment (in the traditional sequence):

Swing_dots.gif (618 Bytes) Viewed 1884 times

(I realize that the notation is not what is recommended, but I used it for best clarity of the actual technical considerations.)

Doug

[polarbreeze]

In any case, (as you well know, having read my description), Encore's decision is not made (we'll assume an eighth note basis note for convenience and pertinence) based on the first note beginning "on the beat", but rather on its ending on the "and" (that is, an eighth note after a beat). And the decision regarding a "second note" is based on its beginning on the "and" (even if there is no preceding "first note".

Maybe we're talking at cross purposes but I think that's a non-distinction because the calculation of swing timing is supposed to be based on the as-written notes, in which case an eighth-note starting on the beat automatically ends on the "and" and vice-versa. Everything, in notated form, is by definition exact and there are no gaps.

So in this example, with the basis note at eighth note, both notes would be given swing adjustment (in the traditional sequence):

Swing_dots.gif

(I realize that the notation is not what is recommended, but I used it for best clarity of the actual technical considerations.)
Doug

It's the fact that you used non-standard notation in that example which obscures the swung quarters (as you pointed out yourself). If it had been written according to convention, beat three would consist of two eighth-notes (tied respectively to the quarter notes before and after). Those eighth-notes would be properly swung, and would follow the same rule as any other eighth-note pair.

[Doug Kerr]

Hi, p,

In any case, (as you well know, having read my description), Encore's decision is not made (we'll assume an eighth note basis note for convenience and pertinence) based on the first note beginning "on the beat", but rather on its ending on the "and" (that is, an eighth note after a beat). And the decision regarding a "second note" is based on its beginning on the "and" (even if there is no preceding "first note".

Maybe we're talking at cross purposes but I think that's a non-distinction because the calculation of swing timing is supposed to be based on the as-written notes, in which case an eighth-note starting on the beat automatically ends on the "and" and vice-versa. Everything, in notated form, is by definition exact and there are no gaps.

Indeed, its a distinction without a difference if we are actually speaking of eighth notes.

And if we have longer notes (as you discuss below), then in many cases the correct notation will remind us of a way to look at it: the "eight notes" that is ties on to the beginning or end will be swung.

I can of course put together tortured examples where notation conventions do not call for a tied notation and so forth, but there is little to be gained from that.

In any case, my description is evidently what Encore actually does. I can't imagine it taking a dotted quarter and internally decomposing it into a quarter and an eighth (tied) just so it can apply a "starts on the beat" algorithm.

Thanks for your insights.

Best regards,

Doug

[Doug Kerr]

What is more important that how we prefer to describe the premises of the Encore swing algorithm with respect to what notes get adjusted is the actual anomaly in its operation.

With regard to "second" notes (those that start on an "off beat" if we are talking eighth notes), if they have a play duration not 100% (as for normal entered notes, for which the play duration is 90%), and we apply swing adjustment, we would expect:

• The start of the note will be deferred by the appropriate amount.
• Something appropriate happens to the end time.

And we can argue about what that might best be, but I think most here would agree that the adjusted end time should not be delayed beyond the "face end" of the note - where it would have original ended, pre-swing, if its original play duration had been 100%).

But in fact, if the original play duration is less than 100% (even 99%), Encore actually will often push the played end time of the note out past that point (which it would not have done had the original play duration been 100%).

One result is that if there is a following note in unison there is now a "MIDI straddle", and this third note is cut off very early - it essentially disappears.

The reality here is that, before swing is applied, it is probably advisable to set the play durations of all potentially-affected notes to 100%.

I have of course described this anomaly to Gvox Development and am in hopes that it will be eliminated soon.

Doug

[Doug Kerr]

For those not familiar with the anomaly I described just above, this figure summarizes the situation.

The two eighth notes will receive swing adjustment.

Swing_test_125-comp.gif (8.34 KiB) Viewed 1839 times

The first row of bars shows the playback times of the notes with the play duration fraction set to 100%, no swing yet applied.

The second row shows the application of swing to the above notes with the basis note at eighth note, 75% swing point. Note that there is no change in the end time of the second note.

The third row of bars shows the playback times of the notes with the play duration fraction set to 90% (the Encore initial value), no swing yet applied.

The second row shows the application of swing to the above notes with the basis note at eighth note, 75% swing point.

Note that here the adjusted ending time of the second note is beyond the "full duration" end of the unswung note, and is in fact later than the start time of the following note. This is anomalous and problematical.

I assume this to be the result of a gaffe in the calculations.

Doug

[polarbreeze]

What is more important that how we prefer to describe the premises of the Encore swing algorithm with respect to what notes get adjusted is the actual anomaly in its operation.

With regard to "second" notes (those that start on an "off beat" if we are talking eighth notes), if they have a play duration not 100% (as for normal entered notes, for which the play duration is 90%), and we apply swing adjustment, we would expect:

• The start of the note will be deferred by the appropriate amount.
• Something appropriate happens to the end time.

And we can argue about what that might best be, but I think most here would agree that the adjusted end time should not be delayed beyond the "face end" of the note - where it would have original ended, pre-swing, if its original play duration had been 100%).

But in fact, if the original play duration is less than 100% (even 99%), Encore actually will often push the played end time of the note out past that point (which it would not have done had the original play duration been 100%).

One result is that if there is a following note in unison there is now a "MIDI straddle", and this third note is cut off very early - it essentially disappears.

The reality here is that, before swing is applied, it is probably advisable to set the play durations of all potentially-affected notes to 100%.

I have of course described this anomaly to Gvox Development and am in hopes that it will be eliminated soon.

Doug

It doesn't matter what the play duration of the notes is before the swing is applied because the algorithm should be referencing the as-notated notes as its starting point for the swing creation. I agree that it should certainly adjust the end times to avoid overlaps and it used to handle that correctly - so if it doesn't do that now it's developed a bug somewhere along the way.

The question arises as to what durations it should use when it applies the swing and since there is currently no way for the user to specify that, I'd suggest the most rational choice would be to use 90% (of the swung nominal durations).

[Doug Kerr]

Hi, p,

It doesn't matter what the play duration of the notes is before the swing is applied because the algorithm should be referencing the as-notated notes as its starting point for the swing creation. . . .

Well, yes and no. It should start with the as-notated times, but (see my suggestion below).[/quote]

I agree that it should certainly adjust the end times to avoid overlaps and it used to handle that correctly - so if it doesn't do that now it's developed a bug somewhere along the way.

Indeed.

The question arises as to what durations it should use when it applies the swing and since there is currently no way for the user to specify that, I'd suggest the most rational choice would be to use 90% (of the swung nominal durations).

The algorithm I suggest (it may well be that what we have no is a broken version of this) would do essentially the following:

• Imagine a "face value" cell within which each note lives (ending at its current play duration).
• Apply the basic swing definition to the boundary between the cells, giving new "face value cells"
• In each cell, make the note duration the same fraction of the cell length that the pre-swing duration was of the face length (which was the "pre-swing "cell").

So if before the swing, the duration of note 1 was 80% of its face value, and the duration of note 2 was 90% of its face value, then after the swing the duration of note 1 will be 80% the length of its "adjusted cell" and the duration of note 2 would be 90% the duration of its "adjusted cell".

This can all be done with data (presumably) available in the current note data structure at the time Swing is invoked.

This is illustrated by this drawing:

swing_14.gif (37.8 KiB) Viewed 1826 times

There, I used the metaphor the "real estate" for a note for what I now describe as its "cell".

Doug

[Denkster]

Encore does indeed understand the term "on the beat" and in fact it uses that concept when it makes decisions for the "beam on beat" function.

Beam on beat does not work as you apparently assume.
In a 4/4 measure, it does Beam-on-pulse, assuming two pulses in the 4/4 measure.