I have two snare measures written in 4/4 in the 7.0a Sibelius template using Sibelius 6.2. The notation is a for a triplet-diddle fp tap roll with a crescendo at the end.

The end sounds fine, during the crescendo. In the quiet part, though, the hands are horribly uneven. I have added a hidden ";p"; marking for accurate playback. If I change this to ";mp"; the unevenness gets slightly better. At ";mf"; it is significantly better, and at ";f"; it is pretty much gone entirely (although now the playback dynamics are ";wrong";). Also, if I force all notes to the same live velocity, the prolbem also goes away. At low velocities (e.g. 50) this sounds too quiet, though, and at higher live velocities (e.g. 80) everything is too loud. Things sound unnaturally even (i.e. robotic when all forced to the same live velocity as well).

Measures before and after these two respond to indicated dynamics and play back with even hands as expected. What is it about this notation that is causing such unevenness, and is there a good way to fix it?

Generally you will get much better results in a DAW program to tweak your velocities and make a better mix.  Sibelius is a notation program, not a recording program.  It's a plus that it can make recordings, but most (almost all) of the more realistic recordings you'll hear are completed outside of Sibelius or Finale. 

I'm aware of that. I'm not looking to create a perfect audio file, nor am I looking to spend the time required to do so. I am looking to create an audio file that is representative of what I want my students to learn, and the Sibelius/VDL combination has proven itself more than capable of generating such a file to my standards many times. Something is strange about these two particular measures.

Every so often, you'll experience a part that just doesn't play back well using the predefined velocities. Some of this may have to do with the way Sibelius interprets its playback performance.

Check your ";performance"; settings. Are they set to ";mechanic?"; If not, set it to that.

Here's another suggestion: Since you've entered a ~p for your roll, the resulting velocity for these few beats may just be falling into that rare void of something that needs some manual tweaking. You might try selecting the low notes of the roll (preceding the cresc), then activating live playback velocities, and fiddling with the number there. Try somewhere around 65 or so and see if you achieve better results.

[quote author=Jim Casella link=topic=4512.msg23503#msg23503 date=1353025047]
Check your ";performance"; settings. Are they set to ";mechanic?"; If not, set it to that.[/quote]They are set to ";mechanic";.

[quote]Here's another suggestion: Since you've entered a ~p for your roll, the resulting velocity for these few beats may just be falling into that rare void of something that needs some manual tweaking. You might try selecting the low notes of the roll (preceding the cresc), then activating live playback velocities, and fiddling with the number there. Try somewhere around 65 or so and see if you achieve better results.[/quote]Thanks, Jim; 70 seemed to work. I had played with that before, but I guess I was just aiming too low. I expected a number much lower than 70 would be where I was aiming.

It still sounds rough, but not nearly as bad and will do fine to play for my kids.

Something I've also done that seems to help sometimes is boosting the Live Playback Velocities by maybe 2 or 3 units for left hands on low end rolls. At a few velocities, they'll get just slightly R hand heavy, so doing that seems to help it out a bit.

Our idiom of rudimental drumming really puts sample libraries through the ringer where such precision is required. You won't find many other musical applications or genres where you ultimately want a series of notes to basically sound mechanical and exactly the same... but not.

It's a brave new world...