[mma] and before, transient markers are placed wrong

[3phase]

I searched for the item and i am a bit surprised that this wasnt under discussion ???
I guess i ve overseen it..but just in case

On a mac book latest os:


It seems that Transients are detected on the zero crossing before the next positive transient.
What leads to widely variable offsets before the actual transient.

Even when i can see that this is probably done to avoid clicking... i dont see how that should work with groove related functions.

Checking a file that is straight on 16th notes, as a groove pool source, shows that the result of this groove is indeed as the transientmarkers suggest, a slight random shifting forward of the sequence instead locking it on straight 16th.
Is this intended behaviouror or still work in progress?

[Matthias]

Hi 3phase,

you are right, there is an additional logic in the transient analysis that shifts found transients to zero crossings with decreasing probability to shift far. the transients are also used by warping and again, right, they should be located at zero crossings.

imagine if you programm a simple beat with sampler, rendering it and then analysing, you'd expect the transients to be located at the position of the original midi note. in most cases this is what the analysis does.

in my tests the transients are not shifted too far left at all, but if you have a sample with very bad transient recognition you can sent it to me directly.

best wishes,
matthias

[3phase]

I dont think that this is wise and practical..the absolut timing information of the transient is most important... the next zero crossing can be easily 5 ms away on material with LF content...

The algorythm should make its individual grain offsets by keeping and optical marking the absolut timing point...
even when this needs some extra lines.

Everything else leads to a lot of extra work for the user... and unconsistent behaviour of the new grosse functions..

regards,
Sven

[3phase]

and one more thing... having a cut on a transient rise never causes a click because the transient actually is the click..

its another thing when slicing two LF waves together ..even on zero crossings that will usually allways click..


so i asume choosing the zerrocrossing before the transient (what not necessarily appears on the positive rise) is done to give the grain circuit the time to catch the transiet itself ratehr tahn toavoid clicking.

but

this should be done with a fixed time offset and not with a variable time offset that looks for zero crossings...
a little fade that is up when the transient arrives would do in most cases a better job and allows a fixed display offset that would show that marker on the transient and not somewher else...

i wonder how thais will be done in the final version... i think the update is great ... i am just not very convinced yet about the new way of warping..even when transient relation instead a fixed grid is musically a major step forward ... groove wise and audio quality wise.. just the way of working with repitch is now requesting some more thinking than befor...because the pitching appears outside the area you move...i guess as old farts learn slower :-/