This post might be better titled “How I wasted the month of February”. It was not actually ‘time wasted’, but just me being my usual overanxious self. This post is about how I put together the Symphony No.9 project.
I’ve done many orchestral project pieces over the years; I’ve just not published them on the web. There was a time when I had a lot of things (mostly Christmas music) posted on SoundCloud, but SoundCloud charges rent and I’ve decided to stick with the limited amount of free slots they have rather than pay their tax for unlimited. Besides, I’m not in this game for remuneration, I have this personal blog which lets me upload whatever I want – no need for pay sites.
For this project I did something new. I usually grab an orchestra piece in MIDI form from places such as Classical Archives.com. They’re a streaming site for classical music and they also just happen to have a section for MIDI enthusiasts. Their MIDI files range from merely OK to fantastic. You must dig through the pile there to find the nuggets of goodness. But this time around, I got my composition from a new source, MuseScore.
MuseScore is a website that supports the open-source MuseScore music publishing application. It also provides free musical scores from everything from pop to classical. It’s an awesome resource. While browsing MuseScore I happened upon this Beethoven’s 9th Symphony and was intrigued. Can I render this? I knew that MuseScore had the ability to play its own score (in a fashion) and some method of exporting that as MIDI data, but how well would that work for me?
Turns out that MuseScore is da bomb! What’s nice is that not only does it output MIDI data as you would expect, but it nicely handles things like switching instrument articulations and following original score directives for playing volume, crescendos, and so on. MuseScore outputs the differing articulations for an instrument, such as violins by using a different MIDI channel number for each one. It’s a more portable way of doing articulations than using key switches, which vary by library vendor.
Typical MIDI files use only one channel per track, so it might seem odd to see multiple channels on the same track for the first time. Most DAWs simply ignore whatever channel number is in the MIDI track in question and simply use the DAW assigned channel number for that track. For our purposes, we need for the DAW to output the channel numbers as is. You might think this may cause problems with other tracks using some of the same channel numbers again, but it’s not really an issue. A DAW treats each track as having its own separate MIDI connection. The other tracks don’t see its data. (This is different from a physical MIDI connection in which all sixteen channels carry on the same connection).
I should mention that there were a few minor technical flaws in the score – these we mostly related to a few score directives not properly attached to the first note to follow it. Easy fix.
I initially crafted this project using the Cakewalk DAW, but it was having a tough time with the high track count of this piece. (115 tracks or so). I decided it would be better to do this one in Cubase, which handles high track counts with aplomb. You need one MIDI track and one audio output track for each instrument in a project. Orchestras can have upwards of one hundred musicians, so I’m doing quite well here. This is because I recorded many of the orchestra’s string sections (violins, cellos, etc.) as a section instead of individuated players. This is a time saving (and track saving) device as a section always plays in unison (sometimes denoted as tutti). It’s only when they manage to run into a chord do they separate into subsections (divisi), usually three. Each subsection plays one of the notes of the chords, and all three-subsections sound at once; violins and such are monophonic instruments – they can’t play more than one note at a time.
There is one ‘gotcha’ with using a section instrument though. When you do manage to run into a chord (which would mean going divisi), you wind up with triple the number of players sounding. A string section is usually nine or ten players and splits into three groups upon going divisi. So even though they are playing separate notes in each subsection, it’s still the same nine or ten players’ total. Sections, however, are recorded with the whole ten players and there’s no three-player (sub) section available (easily), so you just reuse the entire section for each note of the divisi chord.
If I wanted to be precise, I could form three player sections composed of individual instruments for division parts. I have done so before, but most folks won’t notice this unless it’s pointed out to them. It really adds to the track count, and after having Cakewalk fall over on me, I just didn’t feel that want to do it.
With the MIDI output from MuseScore loaded into Cubase, I need to massage things a bit. The piece has a lot of 16th note runs in some of the string parts. These did not sound all that great on a computer – it sounded like machine guns to me. I surmised that adjusting the note attack rate for these (first violins section) to something softer that would give a sound more like what was expected. How fast can you humanly alternate a string bow anyway? I left the attack set that way for the entire piece, technically, I should have only done it for the 16th note runs, but what the heck?
With that fixed, I needed to use Cubase’s articulation mapping feature. Since Cubase can color its MIDI data by channel number, I used that to identify the different instrument articulation and instruct Cubase to send needed key switches to the virtual instrument used. Nothing much here to do, just indicate the pizzicato parts where the players ‘pluck’ their strings.
I used the Halion Symphonic Orchestra library this time, for all my instruments, a first. I’ve used Garritan Personal Orchestra for orchestra projects in the past. I’ve owned HSL as part of Steinberg’s Absolute 5 virtual instrument package for some time, but I’ve never used it before until now. I’m pleasantly surprised. On the one hand it is not as individuated or has as many layer combinations as the Garritan package, but on the other hand it sounds great right out of the box. Garritan takes a lot of additional tweaking to make it sound right.
I used a convolution reverb to simulate the sound of a large hall. These reverbs work by using a captured impulse of the environment you want to emulate. In this case it wasn’t an impulse from a actual hall but from a hardware reverb unit. The impulse I used was take from the famous Lexicon 480L studio reverb, the LARGE+STAGE preset. Nice. I did have to tweak a bit, the gain was too high, but that’s nothing. You can get free impulses of some of the most famous concert halls in the world to indoor and outdoor places from the web. A place like ReSound is a good place to start.
I used a tube compressor on the master bus to damp down some of the louder parts. This piece is very dynamic. But the compression I used isn’t as much as you would hear on a vinyl record, which has a more limited range. I wanted to preserve some of the live performance sounding aspects of the piece.
Lastly, I used Ozone 9 to master the final audio. Since this is classical music, I kept things light there. A quick run through Sound Forge to check for deficiencies in the resulting audio file (none) and then to generate an mp3 file for the web.
NB: I’ve republished the audio using a Japanese Concert Hall impulse in place of the Lexicon 480L. I think it sounds a bit better. Yes I know – totally OCD of me.