Inheritance of MIDI settings into embedded contexts?

[Dr. Thomas Tensi]

Dear all,


I am using lilypond as the frontend for audio production and now I have
a problem where I do not understand lilypond's internal logic for
transferring MIDI settings into embedded contexts.

The file
-----
\version "2.19.82"

\score {
    <<
        \new Staff = piano \with { midiInstrument = "acoustic grand" } {
            { \key c \major \time 4/4
              \relative c' { c4\ff << {e f} \\ {c d}>> g } }
        }
    >>
    \layout {}
    \midi {}
}
-----

produces a MIDI file with the following track data:

-----
MTrk
0 Meta TrkName "piano:"
0 PrCh ch=1 p=0
0 PrCh ch=1 p=0
0 Meta InstrName "acoustic grand"
0 KeySig 0 major
0 On ch=1 n=60 v=101
384 On ch=1 n=60 v=0
384 On ch=1 n=64 v=89
384 On ch=1 n=60 v=89
768 On ch=1 n=64 v=0
768 On ch=1 n=60 v=0
768 On ch=1 n=65 v=89
768 On ch=1 n=62 v=89
1152 On ch=1 n=65 v=0
1152 On ch=1 n=62 v=0
1152 On ch=1 n=67 v=101
1536 On ch=1 n=67 v=0
1536 Meta TrkEnd
TrkEnd
-----

This means that the embedded voices are in the MIDI default volume while
the surrounding notes are in fortissimo (as specified and expected).

My naive assumption is that any setting (also the dynamics for MIDI) is
inherited by embedded contexts. But this does not seem to be the case.

Personally I find this very impractical, because this leads to loudness
jumps in the MIDI output when I have some multiple-voice context
embedded in some macro with the dynamics specified on an outer section
level. It is simply not modular.

Any ideas on how to cope with that?


         Best regards,

                   Thomas