It’s been a week of putting my hand into a dark basket and pulling out things I didn’t even remember making. The dark basket was a folder that contained about 50 audio files I recorded in the Wilderness, time-stamped but otherwise untitled. I enjoyed that discovery process a lot. By yesterday afternoon the album had grown to 14 pieces and an overall length of 40 minutes… and there are a few more things to investigate. Discovered a piece I recorded while it rained. I can hear rain drops hitting the window, before I started playing. I left a few seconds of that, of course. While many pieces are quiet, there is also one rumba and one bossa, which is great because it creates a little contrast. Some of the quiet pieces have notes that just hang there in the stillness like water dripping from a pine cone in slomo. Or golden honey from a spoon. Just lovely.
Such an interesting way to make a recording: find a lovely location and spend a week recording anything that comes to mind, reacting to the space and its surroundings. I think I might want to make this part of a series of albums. I could rent a place somewhere and then proceed to discover how I respond to that particular place. I don’t think I have ever gone to a studio with every piece of music worked out. A few sketches, some chord changes… and the rest happens in situ. For some guitarists it’s all about the performance and how to record and execute it perfectly. One has the impression the result would be the same regardless of the recording location. For me it’s about the way I react to the room and what’s going on and my recordings are always tied to a time and a location. Pick another time or another location and the recording will be different. That’s what I hear in this music, the location, the surroundings, the stillness.
A technical note: I love the ability of the MixPre 6II to record 32bit floating point at 96kHz. I didn’t have to pay much attention to the recording levels. Many pieces would have ended up in the red – if they had been recorded 24bit. With 32bit floating point I was able to import a file to my laptop (last week), discover that it was hitting peaks of +4db… and simply normalize down to a max of -0.2db. Don’t ask me how that works, because I still don’t know, but that’s the magic of 32bit floating point. Since the MixPre 6II was designed for audio recording for movies, this was very important… if an actor starts yelling, no problem… just take down the gain afterwards. It’ll be fine. The MixPre 6II is amazing and I can’t imagine working without it.
>With 32bit floating point I was able to import a file to my laptop (last week), discover that it was hitting peaks of +4db… and simply normalize down to a max of -0.2db. Don’t ask me how that works, because I still don’t know, but that’s the magic of 32bit floating point.
It’s essentially IEEE-754 format applied to digital audio. SInce floating point has an exponent and a mantissa it can also have a sign. In the case of 32 bit the exponent can take values from E-38 to E+38. So for example, The number 2000000 in floating point is represented as 2.0E+06 (or in IEEE-754 binary, and say … 2.0E-6 is 0.000002
This means that you have 76 orders of magnitude for amplitude representation with 32 bits. Which … equates to 1528dB of dynamic range. Yes. That is “one thousand five hundred twenty eight” dB of dynamic range. From an audio perspective, I don’t even know what that means. It’s just a phenomenal amout of encode space.
Because of the way these are represented, 32 bit float will go up to 770dBFS. That’s 770dB more headroom than a 24 bit fixed point encoding. It’s a VERY clever application for floating point numbers, since most compute processors can represent and store it anyway. It’s just another data type for them, and IEEE-768 has been around for decades (1985). Of course there have been updates. There are 64 bit floats too … and 128 bit floats as well.
I hear you Steve and that’s all very interesting but I’ll go with magic! It’s good to know that it’s based on science, but that’s over my head. The float-magic works a charm though! :-)
Question: if 32 bit float already has a 1528db of dynamic range… what does 128 float have?
128-Bit float would have 197134dB of dynamic range.
But … the highest theoretical dynamic range for sound in air when using the threshold of human hearing at 0dB is ~194 dB.
The sound pressure that air can transmit without breaking into a shock wave is at 194dB.
Beyond this, the physics of sound transmission in air changes, and the concept of dynamic range based on air pressure variation really … no longer applies. So … I’m not sure what 1528dB of dynamic range even means when applied to audio much less 197134dB.
Now there is an anechoic chamber at Orfield Laboratories in Minneapolis, Minnesota. This chamber holds the Guinness World Record for being the quietest place on Earth, with a background noise level measured at -24.9dB, which is below the threshold of human hearing. So, you could take the difference to find the “theoretical max dynamic” range on earth, and you’d still end up with “only” 218.9dB of dynamic range.
From a functional standpoint 1528dB of dynamic range is … pretty much … infinite.