In the first 4 notes, the last one is indeed the loudest (though only by a tiny margin), while in the second entry the last note is visibly and audibly softer than the note before.
Is this by design? It really takes away from the climax of the phrase, when the target note of the crescendo is not the loudest… How would I convince Dorico to play passages like this like any musician would play them (without having to adjust every note within such phrases)?
What sort of beat stress have you set? Your pattern doesn’t quite look to me like the default when first and third beats in the bar are stressed and partially work against a regular cresc. Nevertheless, I don’t myself find beat stress useful except in the very simplest music with regular beats – any sort of polyphony and it tends to be counterproductive in my experience. The safest is to switch all beat stress to zero.
For pitched instruments, I do find pitch contour emphasis on the other hand generally works well in phrasing dynamics like a human player would.
I had some slight changes in the dynamics section of the play back options, but reverting them via Library Manager and applying “Reset playback overrides” to those notes did not change anything.
Yeah, and in the first 4 notes of my example it works perfectly; only the second part is (musically) questionable…
Maybe Dorico thinks it’s the last note of the phrase, thus taking it a bit down, while ignoring that it’s the end of a crescendo? But then, the same logic would probably apply to the last of the first 4 notes…
I zeroed the beat stress and then created two identical hairpins from p to mp. The second one is different to the first and similar to the pattern you’ve found for absolutely no reason I can think of. Need to investigate this further unless someone knows of a rule which could explain this.
If, on the other hand, you use a pitched instrument like a violin, both sets are identical with all humanisation off as I’d expected so the behaviour seems to be unpitched percussion specific.
Not sure what I didn’t already cover? Disabling humanization is surely what this discussion has primarily been about!
The specific library can also play a role. Some use CC’s and some velocity with unpitched percussion – the actual dynamic curves always look identical but specific velocity values seem to vary between instruments and libraries. Just as one example, a Cowbell’s last note with Halion is 56 and with NP 46. With this particular instrument, the last note is higher than the penultimate (as one would expect) which is not the same as the example. The specifics of the percussion maps and how they interact with Expression Maps must play a role here and I suspect we shouldn’t get too hung up on the details.
There’s not much difference in dynamic between mf and f – it’s only half a dynamic level. I find that if I disable dynamic humanisation and beat stress for unpitched percussion, I see a small but clear increase in velocity from mf to f.
I wouldn’t say so. I like the way Dorico has improved its playback capabilities over time and would not want to disable them alltogether.
And while I can confirm that disabling the humanization percentage for unpitched percussion yields a better result at this very instance, of course it lessens the experience everywhere else.
So, please take it as a small feature request for further improvements to Dorico’s playback features I would hope that even with default values for humanization, the general outline of cresc. and decresc. could be preserved, meaning that within a cresc. every note is a bit louder than the previous one.
The problem with the original test is that the first hairpin is for two dynamic levels and the second just one. However my own test with just the one level for both pairs still shows something of the effect with the second pair not behaving as expected with the majority of instruments. Which library/instrument did you use to test here, Daniel? The general point about there being little difference in the real world with just one dynamic level I do agree with.
Sorry – I really meant not your original discussion point but the troubleshooting is centred around the effects of humanization. I myself absolutely welcome the steps forward in this area with Dorico 5 and it’s been promised that there will be further developments in this direction.
I didn’t use my ears, I used my eyes to verify that Dorico plays notes with increasing velocity, regardless of the library you’re using.
You could also try giving a bit more “bandwidth” to the middle of the dynamic range by specifying e.g. -3 and 3 for Minimum dynamic level and Maximum dynamic level on the Dynamics page of Playback Options. That produces a difference in velocity of around 15/127 (12%) between mf and f, with humanisation disabled.
Thanks, I already tried -3/3 (even before I came about the issue of this thread), but then the forte sections were so screaming in NotePerformer that I had to tone it back down.
This is of course a very interesting observation. In most of the music I write for (wind band and expecially marching music), “mp” is never being used. Only p/mf/f, with the occasional ff here and there. @dspreadbury Maybe Dorico could have an option to omit “mp” completely from its calculations, giving a better distribution for loudness values?
including with the second hairpin sequence? That’s not what @Estigy and I found! Obviously we’re just looking at the velocity values – I’m quite sure I wouldn’t be able to hear any difference.
Incidentally, I also last night tried to see what would happen if the Min and Max dynamic levels were altered. Of course the velocity values do indeed change with a greater contrast but in the second dynamic group, the final note still has a lower dynamic than the penultimate – again to me the key to this whole thing is why the behaviour of the second dynamic is different from the first – which we’re all agreed is pretty well as expected.
Oh, indeed, in this example you can.
I only stumbled upon this issue because I’m in the process of creating the best audio rendition I can achieve within Dorico, and this missing loud note really stood out (well, in fact it did not). Only then I checked play mode to find the reason for it.
Did OP Estigy’s question get answered? I tried the OP’s demonstration in Dorico 4, with all humanization effects turned off, and I consistently get the forte velocity to be greater than the intermediate velocities calculated by the crescendo hairpin leading up to it. This persists when I tried radically large and small values of the Dynamic Curve power.
Three thoughts come to mind:
My understanding is that non-sustaining instruments use only velocity data, and not controller data (MIDI CC1), for representing loudness, at least in Dorico 4. Did this change in Dorico 5? Potentially, the lower velocity value could be offset by a higher CC1 value.
The fact that Estigy hears the falloff suggests that the above scenario is not the case. Is there a playing technique that got applied only to the forte note, but which got inadvertently hidden, that could cause the falloff?
What actually happens in Dorico 5 if the written dynamic, in this case f, exceeds the dynamic corresponding to the Maximum Dynamic Level number? Estigy has said that he had to “tone it back down” from the 3 / 3 range he had originally tried. Did the problem previously occur in the 3 / 3 range? If not, does it now occur only in the range Estigy is currently using? Similar question for poster dko22, who had the problem in different Min and Max dynamic range sets: Did the maximum written dynamic exceed that corresponding to the maximum Dynamic Level number?
Daniel’s reply that starts “There’s not much difference in dynamic between mf and f – it’s only half a dynamic level. …” gives me great concern. I will address this in a new thread right after I post this response.
the same question to you. Which library are you talking about? Every one has a different control method. The Dorico version is irrelevant, It’s what the library vendor has programmed and indeed sometimes you can choose the control method yourself. For instance the VSL libraries I use are programmed with CC throughout, though velocity is also an option.
I would hope that even with default values for humanization, the general outline of cresc. and decresc. could be preserved, meaning that within a cresc. every note is a bit louder than the previous one.