Quantize Window Bug?

[parboo12]

I don't think the quantize Window setting  works properly. Either that or I just don't understand it, in which case I think the documentation can be a bit clearer.
If I am quantizing to the quarter note, I am under the impression that a window of 100 would include everything  from half way to the previous quarter note to half way to the next quarter note. If I reduce the window to 70 say, I would think it would shorten the window by 15% on each side...and so on.
Swing setting is at 50.
I've been playing around on the piano roll by placing a note where I would say it's definitely out of range of the window, but it gets quantized anyway.
I could make a video but if anyone tries to play around on the piano view with one note, I think you'll see what I mean.
And it's no better in Sonar X3.
Or if I'm not understanding something, maybe someone could set me straight!

[John Vere]

 

[parboo12]

Nice video, John, but have you tried specifically what I was asking about (the Window option of Quantize)? I thought I understood it, and was trying to demonstrate it to someone what these options did, and I found the Window to be very inaccurate.

Thanks!

[parboo12]

Has anyone else tried checking out the accuracy of this? I haven't got good results.

 

[Byron Dickens]

There's nothing wrong with quantize; it works exactly as it is designed.

https://www.cakewalk.com/Documentation?product=Cakewalk&language=3&help=EditingMIDI.39.html

This has been a feature probably since Cakewalk was a DOS program. I think if there was a bug in it there would be more reports than just yours.

[parboo12]

Hey bdickens, thanks for the great info!

Just a couple of quick YES or NO questions for you.

1. Have you sat down for a few minutes with Piano Roll, put in a note and moved it around to different locations, while playing around with the Window settings to see  if Quantize picked up your note or not?

2. If so... did you get the results you were expecting?

[Byron Dickens]

Yes. The quantize function always works as expected.

Edited January 14, 2022 by bdickens

[bvideo]

I don't think the quantize (non-groove) dialog works exactly as the document says. I think there is no difference between setting the window at 50% and any value above 50%.

Numbers 50% and smaller seem to work, and 50% covers all notes. I think the % represents the fraction of the distance between two grid points. Using 100% would make every note part of the window of two consecutive grid points (actually 3 if the note is on a grid point). Every window N% greater than 50% could make some notes within the N% window of one grid point also be within the N% window of the other nearest grid point. For example, if window is 70%, if a note is at 60% from one grid point it is also at 40% from the nearest grid point. Quantize then favors the nearest gird point. Obviously the document is wrong in saying 100% represents half the distance between grid points.

[Byron Dickens]

 

3 hours ago, bvideo said:

Quantize then favors the nearest gird point.

Well, duh....

3 hours ago, bvideo said:

Obviously the document is wrong in saying 100% represents half the distance between grid points.

That's not what it says at all. The window straddles the grid point, with half of it before and half after. NOT the same thing.

"Strength" refers to how much notes are quantized while "window" controls which notes are.

A Strength of 100% snaps every quantized note (as determined by the Window) exactly to the grid. 90% moves them 90% closer, 80% moves them 80%  closer and so on. That's closer to the grid from where they are now.

"Window" is where it can get counterintuitive because only notes within the window get quantized. With a window of 100%, all notes are quantized. Because it straddles the grid points, a window of 90% only quantizes notes that are within 45% before and 45% after the grid.

Ironically, the more rhythmically accurate your performance is, the better results you will get. Much like pitch correction: the less you need it, the better it works.

So if your playing is really sloppy - like my keyboard playing is - your quantize results can be quite unpredictable.

Edited January 15, 2022 by bdickens

[bvideo]

Would you like to try it? With a window of 50% (or higher) all notes get quantized. Strength is not being tested here, nor is performance. And yes, the window straddles the grid point. So as I tested, a window of 50% reaches all the notes.  Whereas the document says: "A window of 50 percent extends only a quarter of the way toward the adjacent quantization points."

[parboo12]

 

1 hour ago, bdickens said:

That's not what it says at all. The window straddles the grid point, with half of it before and half after. NOT the same thing.

Then why don't you try it out, and see if it works like you're telling every else. Oh right, you already have and it works perfectly.

[Byron Dickens]

1 hour ago, parboo12 said:

 

Then why don't you try it out, and see if it works like you're telling every else. Oh right, you already have and it works perfectly.

Yes. I use quantize much more than I care to admit and it operates as designed. "Operating as designed" does not mean " operating the way you want it to."

1 hour ago, bvideo said:

Would you like to try it? With a window of 50% (or higher) all notes get quantized. Strength is not being tested here, nor is performance. And yes, the window straddles the grid point. So as I tested, a window of 50% reaches all the notes.  Whereas the document says: "A window of 50 percent extends only a quarter of the way toward the adjacent quantization points."

Math is hard ....

Jeez! Even my discalculiac self can figure this out.

Look, I don't know what you two are trying to accomplish -  or prove  - but I'm not going to create some test project with the timing so grossly out of whack to "test" the limits of the quantize feature. Because I know I won't get good results. 

Quantize is there to tighten up a performance that is a little bit off - key words being "little bit" - not get a perfect groove from somebody who can't tell the difference between the pocket and his a**h***.

 

Edited January 15, 2022 by bdickens

[bvideo]

My only aim was to confirm the OP did bring up a valid discrepancy between the document and the result. Also to clarify the actual operation of the one parameter. I get that you're not on board with that. What anyone wants to do with quantize or why doesn't seem important to that end.

[Byron Dickens]

Yeah. I guess I'm not on board.

How many angels can dance on the head of a pin?

[parboo12]

Thanks for testing things out, bvideo.

Hopefully the bakers will check this out.

Edited January 16, 2022 by parboo12

[bvideo]

parboo12, you're entitled to submit some kind of problem report to the bakers. I doubt this thread will come to their attention. Seems like the easiest thing for them to do is change the document; the functionality is OK otherwise.

By the way, for real-world quantizing, such as entering MIDI by hand or sequence and then wanting it to fit with a real performance, it seems like "groove quantize" would be the way to go. Want to check those parameters??

[msmcleod]

I think the confusion regarding the Window is how it's calculated.

From what I can tell, the algorithm first works out what section of the grid that note would be quantized to, then checks to see if the note lies in within the Window for that grid position to see whether to actually quantize it or not.
 

[bvideo]

The document explains that well enough. But from what we can tell, the numerology doesn't match the document. Whereas the document says: "A window of 50 percent extends only a quarter of the way toward the adjacent quantization points", a 50% window actually reaches half the way. So all notes are quantized. Namely, all notes are at most 1/2 the distance from one of the two quantization points it lies between.

Also "would you like to try it"??

[Blogospherianman]

The Quantize Window % is most definitely off. Like bvideo mentioned the window is twice as big as it’s showing.  50% Window = 100% ACTUAL window.

Yes I’ve tested it. ?

With Quarter note resolution and a window of 50%, an eighth note on the upbeats (480 Ticks at 960 Ticks per beat) SHOULD be ignored being well outside the window which extends a sixteenth note on either side of the quarter notes. 
In actuality at quarter note resolution and 50% window it is Quantizing the eighth note sitting at 480 ticks. 
Same experiment set to a window of 49% and the note doesn’t move.  

100% window SHOULD extend 50% in front of and behind the resolution point totaling %100. Cakewalk’s 50% is ACTUALLY 100%. (Extending 50% in front of and behind)

Cakewalk’s 25% is ACTUALLY 50%. (25% in front of and 25% behind.

It should be calculated:

Resolution (Ticks) x Window % /2= Resolution window +/-

e.g. 960 (quarter note) Ticks x 50% Window/2= 240 +/- (720 - 240)

For some strange reason Cake is using Resolution x Window % = Resolution window +/-

Hope that helps ✌️

 

Edited January 26, 2022 by Blogospherianman