The ones that don't process/ output in floating point.

 

Floating point extra safety-margin doesn't mean that it will sound right.
Make a favour to your productions: keep the signal around -18dbFS (percussions can go a little further, bass should respect the law, etc.)
It's important to understand, once for all, that in the analog domain going over the VU0 is interesting and sometimes euphonic, in the digital domain is awful, even if the new technology claims to allow it.
Mixing close to the red in you DAW is risky and counterproductive. I know that the people has the feeling of "warmth" doing it, but... well, nope.
Of course the plugin/daw producers give you the chance to mix all red, with clippers, math tricks and whatever, but it's a choice of marketing, because the majority of users need to "ride the highest tide" to feel/see the "right" sound.
Of course, if the final product will be squeezed and squared like the Metallica's Death Magnetic, nobody will notice it, but if you aspire to something to "audiophile" this IS the way.

 

You'd have to ask Kush, but I suspect it's because -16 was an old AES standard in the US. Why do engineers use -18dBFS=0vu? It's because that's how the original hardware is calibrated under modern engineering standards and if you have a Pro Trolls rig that's hooked to outboard calibrated to -18, your desk needs to be calibrated to -18, and your plugs need to match. That's why if you're claiming to emulate a Manley, you must match -18 calibration or it ain't a Manley emulation.

The consequence of this runs to your volume faders. On all DAWs these emulate logarithmic hardware faders. The bottom part is all crunched up and the top part is all stretched out. If you gainstage to industry standards, -18dBFS=0vu, the majority of your mix choices will fall within the linear center section of the fader throw. Too cool and you're banging against the top and too hot you're inching tiny little adjustments against the bottom.

Finally, you can only meaningfully mix on 21 bits of digital resolution. This is because at -125dB you hit the noise floor of electrons passing through metals. You can have 64bits of maths going on in your DAW but you'll only ever hear the top 21 bits because there's no amplifier in this cosmos can amplify more. You get 21 bits. That's why the EBU choice is -18dBFS. That gives you the bottom 21 bits over 24 (or more fixed point) to record on and 3 bits headroom for your mix decisions.

Now, 32 bit floating point actually gives you 24 bits of resolution and 8 bits +/- as a multiplier to float over or under. Thing is, if you go over, you end up rounding off the bottom order bits. In other words shifting undithered truncation.

So you go hot if you want to. Please. There's too many amateurs in audio these days. Go hot. Yummy.

 

Last edited by a moderator: Feb 1, 2020

Mastering - before and after

http://mastering.lt/before-after

https://www.chrisgrahammastering.com/#

 

If a vst was modeled after certain hardware I expect it to act as the hardware. If there is a sweet spot in the hardware I think there should be the same in the plugin, cause going past that level I want to get that non-linear distortion modeling that the hardware is famous for.


and chris you have a nice site there, I think country is my favorite mix, good job.

 

yes I think so, it must be due the design of such plugins, which is more about sampling hardware data rather than LTSPICE modelling and coding.

 

yeah, In general, I stay away from the desk, the tape machine and the hieroglyph walls

 

wrong, The interface has nothing to do with internal processing, as long as your final output does not clip it doesn't clip end of story.

the only way I can think of when it can matter. Is when incorporating hardware, passing tru a D/A A/D.
But when that's the case I would ask myself if it is useful at all too switch out an interface .

 

Last edited: Feb 1, 2020

Thank you for quoting me and saying nothing of any valuie. Not that you ever do.

 

First off, nice calculation, maybe I need this - someday.

Of course it holds, because it hasn't to do anything with the soundcard, at all. FP isn't (practically) clipping within a DAW. Since when are we talking about hardware in this thread?

And we're talking about FP neither, only about the necessity of a -18dB calibration.

 

Last edited: Feb 1, 2020

...del

 

Last edited: Feb 1, 2020

which else

 

Aha, so tell me, please and I mean it, why this worked:
https://audiosex.pro/threads/how-do-you-guys-keep-your-levels-low.51942/page-2#post-452317

You want another proof that -18dB isn't necessary, as far as I've found out? Here you go:

Spoiler: input -18dB

and compare to a 28dB higher input

Spoiler: input +10dB

Now, where's the problem???

 

... when you mix up software (plugin) and hardware (desk).

 

maybe make this test with kramer tape

 

Err, what? Which else? I could already proof in another thread that this calibration isn't really necessary for all Waves plugs who claim this.

And 'I would say' is far from being a proof - for which I'm still waiting.

 

which devs else i meant...which have it stated in manual

 

Ones I can think of include Softube's analog emulations including those done for Native Instruments. Basically it's anything that's emulating physical hardware or the performance of physical hardware such as var-mu compressors and such. All the UAD offerings do. It's an EBU standard so people will match it because not matching it will be a source of complaint.

 

For most types of processing in the digital domain, there's no sweet spot, because most plug-ins (like Pro-Q 2, etc.) do not clip internally.

However, plug-ins in which use saturation or distortion (like Saturn obviously, but also Volcano 2 or Timeless 2), it does matter how hard you drive the input of the plug-in. In that case, -18 dBFS is a good starting point, but not a fixed rule. Best advice: just use your ears!


Floris (FabFilter) — Mar 3, 2017

https://www.fabfilter.com/forum/post/2903