The Brass Patch
Collection
This page contains a
collection of G2 models meant to mimic brass instruments. They are all physical models, based on the techniques
described in these pages, with special attention paid to final equalization.
All instruments have been
equalized to approximate the spectra of sampled instruments contained in the
Garritan Personal Orchestra. In fact,
each instrument has been equalized twice,
once to a quiet sample and once to a loud one.
The ‘Manliness’ knob morphs between the two timbres.
Each model on this page is a
G2 performance, instead of a patch. This
is because each model requires 4 DSPs to generate the sound: one DSP for the basic model, and three more
for equalization. The G2 expansion board
is not required.
Equalization Technique
Brass equalization presents
an interesting challenge. First, the
timbre depends on which note is played. Second,
as a brass instrument is played louder, its timbre changes in a significant and
distinctive way.
To accomplish this, a
2-dimensional equalization technique is used in these models. The first dimension is the note number, and
the second is the player’s intensity.
A brass bell is basically a
bandpass filter. So the EQ circuit
begins with a fixed highpass filter and a voltage-controlled lowpass filter in
series, to approximate the passband. A
fixed parametric filter is added for additional trimming. Finally, the sound goes through a 26-band
voltage-controlled equalizer for final shaping.
The circuit has 28
voltage-controlled inputs: the overall volume, the cutoff frequency of the
lowpass filter, and the gain controls of the 26-band equalizer.
Each control input is fed by
a pair of G2 SeqCtr modules that are themselves controlled by the
The outputs of each pair of
SeqCtrs are blended together using crossfaders.
Within each pair, one SeqCtr is adjusted to match a quiet note, while
the other matches a loud note. The
crossfader’s control input comes from a knob labeled ‘Manliness’. This provides the second dimension of
control: the intensity.
To equalize a particular
note, the note is played on the model.
The spectrum is then compared to a sampled spectrum of that note,
obtained from the Garritan Personal Orchestra.
The relevant SeqCtr values are adjusted until the spectrum is imitated.
In each case, the Garritan instrument
chosen was the solo 1 instrument (Trumpet 1 Solo and Tenor Trombone 1 Solo).
To add the second dimension
of control, the same process is repeated using the ‘overlay forte’ samples
provided by the Garritan Personal Orchestra.
The second set of SeqCtr modules are adjusted until the model matches
the Garritan instrument.
Sample MP3s
The following songs were
recorded using these models. Each model
was controlled with Yamaha BC3 breath controller through a Kurzweil
ExpressionMate, using variation 8 on slot 1.
The only post-processing added was a convolution reverb.
- A fragment of a fanfare from George Frideric Handel’s
Water Music, performed by a trumpet and two trombones.
- I Sing the
Mighty Power of God, an English hymn sung to a German tune published
in 1784 in the Gesangbuch der Herzogl.
This is an SATB arrangement performed by two trumpets and two
trombones.
The Instruments
The models are below. After each name is the instrument’s natural
range.
- The Bb trumpet (E3
to F6) This is a straightforward 3-valve brass model, as described in
these pages.
- The Bb trombone (E2
to F5) This is a straightforward 3-valve brass model, as described in
these pages, but tuned an octave lower than the trumpet. Why not model a trombone slide instead
of valves? After all, a slide is
easier to model. The reason is that
note-to-note transitions are smoother using valves. So this patch actually models a Bb
baritone, not a trombone. But since
the two instruments play the same range, and few people know what a
baritone is, it’s called a trombone.
Anyway, trombone samples were used to equalize it, so it sounds
like a trombone.
Variation Programming
In each model, the following
variations are programmed:
- Variation 1:
This is a basic sound. It’s
triggered by the keyboard, and the mod wheel is mapped to manliness. Velocity is mapped to attack time. A small amount of vibrato fades in
automatically after a short delay.
- Variation 2:
Same as variation 1, but some shelving EQ is used to make the tone
a little mellower.
- Variation 3:
Same as variation 1, but manliness is fixed at the maximum amount,
and the mod wheel is mapped to lip embouchure, raising it up one octave.
- Variation 4:
Same as variation 3, but the mod wheel lowers the lip embouchure
one octave instead of raising it.
- Variations 5 through 7: Same as variation 1.
- Variation 8:
Same as variation 3, where the mod wheel raises lip embouchure one
octave. But in this variation both
the input air pressure and manliness are controlled by a breath controller
(