I'm absolutely certain that one can make good music at nearly any bandwidth and frequency. That being said, the headroom of 24bits/96khz is incredible and useful. I would never suggest you *had* to have it, but the better the quality you start with, the better the quality you'll end with. That's the corollary of "GIGO", but I don't want to suggest that 16/48 is garbage.
I compeltely agree. I don't know if you missed the context of the thread, but I was talking only about the initial composition stage, not suggesting 16/48 is enough for final mixing. I'll get back to this later when I address you point about bouncing.
While this is true, there's a lot of meat hidden in the word "glorified". There are pianos for Kontact that are GBs of samples. There are also lots of "sampler instruments" that use the built-in FX of the various sampler platforms. Let us also note that the really great samplers are nearly as cpu heavy as the really great softsynths as they engage in high-speed, high-quality resampling, envelope modifying, LFOs, the whole nine yards. I've been in recording studios where a rack mount machine was dedicated to Kontakt. Sampletanks has an equivalent number of effects options and sample modifiers/scalers/etc.
You're quite right, I probably over simplified it.
Also, four tracks of Amplitube can make my dual-core 2.4Ghz iMac sweat, and cause occasional "Logic was unable to play back ... " (can't remember the actual text, but...) Same with Guitar Rig and Pod Farm, although Pod Farm seems to have the least impact on my setup. It's worth noting that freezing a SINGLE AMPLITUBE track makes a noticeable difference in CPU load on my setup.
The Logic error message you refer to can be often avoided by doing a 'renice' command on the process ID.
Well, initially, no matter how much you up-sample (to 24/96 or 24/192 or beyond) you will never get a signal better than the 16/48 you started with. For best final quality (we can argue about what's acceptable, or appropriate, but not what's technically best signal reproduction) you start with all inputs at the highest quality in the system; if you input in 16/48, your final quality can't be "better" ( in the sense of "faithful signal reproduction" ) than 16/48 regardless of the output bitrate and frequency. And since Logic does *all* it's internal processing in 32bit FP, you're not really saving any CPU by using 16 bits.
OK, perhaps I explained myself poorly, and please bare in mind, I was really pitching at someone (the OP) who seems to be a bedroom enthusiast, rather than someone like you. I'll explain it anyway.
Here's what usually happens... the newbie workflow:
- Select a software instrument (might be a sampler or virtual synth)
- Slap 5 effects on it and play a bunch of 8-note chords, record as a MIDI track
- Select a drum loop. Ultrabeat will do! Add some reverb, maybe a filter, perhaps seperate out the sounds to 4 or 8 tracks, slap different delays and compression on each
- Bass line... simple ES1 sine wave... but it needs some EQ, slight delay and compression to stop the chords phasing out
- OK now I need that Orbital "woahawowaaa" sample. Some ping-pong delay, compression, EQ, reverb, bit-cruncher and distortion...
OK, why do I have 400% CPU use...
Now when I say you can use lower quality audio and then go back to higher quality, I mean this. You bounce down the audio, simply when you have more or less the drums/bass or chords you want. I don't know how you work, but for me, the drum loop is the least important thing to have perfect at the start. I layer it down, and copy it across as a sampled loop with any effects that are critical right at the start. 16-bit AIFF is fine at this point. Remember I have the original Ultrabeat loop somewhere when I need it.
This is good enough so I can get my chords down, my bass and the start of some melodies. During composition, there should rarely be any need to run 7 or 8 soft synths, with all their raw effects on each channel, especially for someone who's just doing this for fun.
As parts of the mix start to come together, you can discard your 16-bit low-fi version of the track and reload ultrabeat to perfect your drums, for example. You'd probably freeze other tracks while doing this anyway, but in the event you really need to hear the other parts, bouncing is a CPU savior.
Now I hear what you're saying about some very simple sounds requiring a lot of CPU, such as your piano example and of course more CPU is always nice. Perhaps I don't appreciate this as much as I should, as I generally get most of what I need down using hardware synths, so take some things for granted.
I avoid bouncing tracks until I have to; once you bounce, you lose the ability to revise (beyond EQ). Furthermore, you're adding first one bounce, then another when you output your final result. No reason to introduce generation loss unless you have no choice. I prefer to use "freeze" (as there's no generation loss and you can unfreeze things) but that's still slower than live tracks. Yes, you *can* work with less; people have made *great* music with four-track cassettes. But we're talking 'best case', right?
I'd agree with this, although you're almost assuming that the only reason you'd bounce is when something is finished. I bounce just to hold the feel of something while working on other aspects. Most people I know don't do this, they do what you do, but I find it works for lower end machines.
I used to work with a Juno 60, an SH101 and a 3rd hand 4-track from the late 1960's when I started out, so I hear that.
Regardless, some things *are* matters of fact, and there's something in the thread I wanted to address.
~stuff about hdds
Great post, and the quality of the drives/number of spindles are rarely mentioned in these posts. I didn't quite agree with Chud that external FW drives won't be better than an internal drive (although this is no option to the OP, and I don't think we were offering general advice, it was in context). The fact is, even a lowly 22mb/sec should be enough for most enthusiast audio work, especially when you've got a couple of GBs of RAM to cache this stuff. Having said that, 22mb/sec sounds low - sequential read should be a lot faster than that, and large audio files on a good system should not be fragmented, so I'd expect better performance than that. This is one good reason to keep your audio files away from the system disk.