The problem of 96 or 88 is that it hits your cpu harder and also your files get bigger occupying more hdd space, higher sample rates also means less plugins.
I have been working with 96 but thinking of going 48, i started with 44, though the biggest difference was jumping from 16 bit to 24 bit, just more headroom and better un-squashed lower frequencies.

 

Yep. You always have to balance HDD space and processing power.

In terms of recording resolution, there's really nothing you can do. It will demand more space, but HDD space is cheap nowadays.

In terms of mixing resolution, you can mix in lower sample rates to save processing power, but then you'll have to turn on oversampling on various individual plugins, which will again demand more processing, so sometimes simply increasing the whole session's sample rate will make things simpler and not necessarily consume more processing than it would if you upsampled individual plugins on a lower sample rate session...

So, it boils down to:

1)Upsample the whole session.
2)Upsample individual plugins on lower sampling rate session.

Which consumes less cpu? Damn, I have no idea.

You can always render you tracks once you're done with your demanding plugins as well...Not ideal, but in some cases it's necessary, like when you use Nebula stuff.

 

Last edited: Oct 10, 2017

Cubase`s channel Eq/strip has been a god send regarding the cpu consumption and big projects in 96khz, its not surgical nebula quality eq but its good enough for most of the tracks.

 

I don't use Cubase, but I noticed that native, less demanding and apparently inferior plugins sound way better in higher sample rates.

I've been using Studio One's FAT Channel a lot since I moved to 88,2hKz. Zero latency, almost zero cpu consumption and works great for clean initial eq, filtering and compression.

Honestly, nowadays I tend to mix as cleaner as possible, adding analog emulations very clinically, in order to avoid layering a bunch of different emulations all over the place and superimposing all their non-linearities.

Another nifty little tool is Overloud's TapeDesk. It sounds great at higher sample rates. It's doesn't alias at all at 88,2kHz, and it consumes negligible processing power. It's great for adding as the first insert of every track and giving just enough of that analoguish liveliness. Saves a lot of processing down the chain...Add different wow/flutter values (typically lower than 1,5% to emulate a pro tape machine) to every instance and it gives just enough movement to your tracks to make your mix sound more alive...

 

Last edited: Oct 10, 2017

Coloring with sound like a painter, im also getting there.

This is always the case when you find out something great ( saturation/ tape plugins) you tend to use them all over the place and then you slowly realize you need to slowly cool the fuck down.

Coloring only selected tracks or buses is the next step of growth just like choosing reverbs carefully.

 

Unfortunately, big name plugins suck like this...FabFilter's Pro-C, for instance, alias like hell at 44,1kHz with oversampling turned off...There's a huuuge thread on Gearslutz where people test one gazillion plugins for alias and posted their graphs...It's eye opening...And horrifying.

Dude, since I started taking mixing more seriously, I mix even when I cook...Oh, there's something missing on the upper mids...Maybe some more air...This is me cooking. lol

 

I'm currently working in 44.1KHz. Primarily for the reasons @Von_Steyr stated, CPU consumption primarily, HDD space less of a concern. Additionally, I'm using the Olonga 44.1KHz libraries in Nebula (weren't free when I got them), and my understanding is that it's not a good idea to up sample them. So I've stayed in 44.1 KHz for now.
Recording in that sample rate, during mixdown I'm placing LPF above around 18K on most tracks so that should help reduce top end aliasing problems, I'd imagine.

I'm so, so guilty of this, haha. It was true when I got Gold. It went EVERYWHERE! Haha. I truly like having an instance of it on my reverb and delay buses though. Seems to add more depth to the effect.

I think the biggest help to me as far as aliasing goes is summing through a few hardware pieces and a console. I think that alone really tames the digital harshness, and gives a nice fat character to the overall mix. After bouncing back in, I don't generally do anymore digital processing with it.

 

TL:DR: I paid actual money for this white noise.

 

Confused to the death
There is no oversampling option in Cubase or any VSTs, as I know. (they maybe are incorprated by default)
I have ever never noticed aliasing issue, and I can hardly hear anything over 20kHz! Thus, normal people that listen to music with their shit headphones will never feel anything at that frequency, so no problem for me and fuck the analyzer.
However, I will try to switch to 48kHz if it has that much improvement. I have enough space and CPU for the added 4kHz

 

dang!! sooo much information from you guys... can't read it all right now...

 

Hey mate,

I guess you didn't understand the concept of aliasing correctly.

Although it starts at high frequencies, it bounces back throughout the whole spectrum, causing chaos all over it.

Watch the video I posted on the last page. You not only will see it, but you'll hear it as well. Just pay attention to the sweeping noise behing the pure 1K tone. That's aliasing.

So, basically, yeah, aliasing will destroy your mixes. It will blur everything out, make things harsh and take a lot of precious space in the stereo field.

R.

 

By using analogue hardware you don't have to think about digital aliasing. I use as less digital processing as I can. Especially when you use many VST plugins in a chain it can sound nasty and I do mean it in a bad way.

 

can someone explain me where aliasing can occur ?

 

Not sure about 48, if the final product is going to be 44.1. Analog => 48k => 44.1k could...

Does this sorta make sense, or am I missing something?

 

TL;DR: When frequencies higher than Sample_rate/2 (Nyquist limit) are not filtered out. This could happen from doing dumb stuff, like trying to digitize frequencies higher than 22kHz @ 44.1kHz sample rate, like this:

...or not filtering > 22kHz harmonics (only a pure sinewave is just root/fundamental, everything more complex is a harmonic series, going up up up).