I spent some more time systematically examining all 494 Bosendorfer sample files for noises of any kind. There are some with dogs barking, many with a 60 cycle hum from an unknown source, and one that sounds like there is a wind chime in the background. Using Audacity Noise Reduction and fade out effects, I was able to reduce the noises to below audibility.
While I was at it, I introduced a new volume controller. Previously, I had two volume controls for each note. The first was just the normal Csound volume input, which I used to select the sample file. Since there are ten different sample files for 10 different pressure levels, this allows me to imitate harder presses when making a note louder. But I needed another volume control to be able to modify some notes independently of the main volume control to prevent clipping. In this way, I can affect the volume independently of the sample file choice algorithm. This allows me to have the texture of very quiet playing without making it so soft that it is inaudible, and employ the high pressure level sounds without distortion.
This iteration exploits all ten different pressure levels for more variety, and I have avoided clipping.
This is a revised version of my Goldberg Variations. These remove many of the spurious noises in past versions.
I recently bought a new pair of headphones (Sony WH-1000Xm3 noise cancelling) and listened to some of my recent tracks on them. I was amazed at the noises that were in the some of the pieces. I discovered that all the quiet Bosendorfer samples have a significant noise component after the initial note sound. I used Audacity to remove the noises with the Noise Reduction filter. Most of them are fixed now, but a few still remain. I never noticed the noises before because I never mixed with high end headphones. I’d go for walks and to pieces listen, but my standard headphones are pretty low end. When I listen in my office on speakers, I keep the sound low, and never heard the noises there either.
I regret the error. And plan to get my ears checked.
Here are the envelopes that I am using. They tend to hide the note starts, and enhance any noises towards the end of the samples.
This is a set of three variations on the theme of Naima by John Coltrane, scored for Prent’s Microtonal Slide Bosendorfer. This one is based on two otonalities of the 31-limit tonality diamond. Variations 1 and 3 use 8:5 otonality, and the second variation uses the 16:15 otonality. The two have many notes in common, and many that differ by a few tens of cents. There are some intervals that might sound “out of tune”, but I have faith that should John Coltrane, McCoy Tyner, Jimmy Garrison, and Elvin Jones find a way to listen, they would approve. I actually studied with Jimmy Garrison at Bennington in the 1970’s. He convinced me in his own unique way, that I was not destined for a career as a flute player, or jazz musician.
This is a set of three variations on the theme of Naima by John Coltrane, this time scored for Prent’s Microtonal Slide Bosendorfer. There aren’t any slides this time, but I plan to do another version with slides. This one is based on two otonalities of the 31-limit tonality diamond. Variations 1 and 3 use 8:5 otonality, and the second variation uses the 16:15 otonality. The two have many notes in common, and many that differ by a few tens of cents. Here’s a chart with the deltas.
This is a rendition of John Coltrane’s wonderful tune Naima, named after his wife. It’s so sweet. But the harmonies are quite complex. Every version of sheet music that I’ve seen has a different set of chords. I suppose trying to nail down the chords that McCoy Tyner is playing at any given time is a fools errand. I settled on this one, mostly in Bb, on an otonal scale based on 8:5. The key signature of the first stanza is correct, and should be used on the others as well. It’s not correct on the rest of the piece. Cheap fake book mistake.
This is another trip through the randomizer. I’m trying to figure out how to increase the volume when the slides are taking place. For some reason they tend to happen only when the volume is triple p, and the loud parts have no slides.
This version has a very small change for the E minor chords. I was using the 3:2 otonality (which is a available with a 31-limit tonality diamond based on 9:8), and I changed that to use the 5:4 utonality based on a 1:1 diamond. This change only affected measures 9, 11, 14,and 21. By doing this, I avoid having an E minor chord with a 680 cent fifth. In measure 24 I kept the old E minor. It sounds better.
Throughout the piece I try to maintain all the keys with pure fifths at 3:2, which are 702 cents, thirds that are either 5:4 or 6:5, and 7ths in subdominant chords as 7:4, while preserving the same key notes in the major key of the piece. That means I try to have important notes like the D as 204 and G as 702 cents not change as I move across different chords. Mostly.
I noticed that there was something wrong with the G major based on the 1:1 otonality. The B a 5:4 above the G sounded wrong. The G was using 127th value in the table, for a value of 0.0701955, but the 127th value in the old table was 0.0721741. The same offset made the B, which was the 198 value in the table should have been using the value of 0.1088269, but in the old table, the 198th value was 0.1095045.
So basically whenever I thought I was playing a G 3:2 above 1:1 C, I was playing a 44:29, and the B should have been 15:8 above 1:1, I pulled a 32:17. That made a major third of 373 cents instead of the 386 that is the just 5:4. That’s a difference of 13 cents. Some intervals are not so bad, but that means all the pieces I’ve made since May of 2017 are full of lots of wrong notes.
I regret the error. I’m going back and redoing some of the pieces to fix the mistake. This might take a while.
Here is another trip through the algorithms of the latest machine. The piece is based on Cleansing Fountain, a traditional song that was included as a quote in the Charles Ive’s song “General Booth Enters into Heaven”. I made the variations by collecting all the notes of each measure into a list, then choosing from that list as I move through the measures. I take four measures at a time for each variation. At the start of the variation, I choose a rhythm, density, randomization algorithm, and volume. Then for each measure, I choose the proper key, measure, and type of measure it is. The type is either C major and G major, F major and C major, or all F major. Those three include all the measures in the piece. Then for each note, I choose an envelope, slide, note, duration, and other factors. I do this for sixteen variations, then restate the last few measures of the song straight. Sometimes the density is sparse, sometimes dense. Sometimes it’s loud, sometimes soft. It all depends on the choices the preprocessor makes. This one is the most recent trip through the machine, fresh today.