Your rule of thumb is to cut your tracks at relatively high but not clipping levels. Everything should be cut/tracked in this manner.

Rule of thumb for mixing is don't overload your mixing bus. You mix to create what you want to hear. If you're starting out too hot, there's nowhere to go. So you don't start out as hot. "Dr. it hurts when I do that". "So, don't do that".

It's not easy building a mix. You really have to determine what you're underlying & driving force should start. You know you are going to layer it up with other stuff topping it off with the lead vocal. Think "building a banana split" and you'll get it. Build it too high? And it falls over and makes a mass. This is what mixing is all about be at in the box or out of the box. The same thing happens with analog consoles. It doesn't matter which median you are working in it only matters that you understand how the medium works. If you know its limitations you'll have none. If you don't know its limitations it's all a limitation. Worried about not having enough level? Work in 24-bit mix down resolution. Once you are done mixing then it's the mastering engineers turn. That is, unless you want to mastering it yourself? That's another subject.

Next stop, your incredible outrageous world changing hit
Ms. Remy Ann David

Limitation Eliminator
Ms. Remy Ann David

Thank you, Remy! Your posts are always fun to read on this forum. I was a bit wary of turning down my inputs too much as I feared losing presence, and the band that I happen to be recording are super fussy about their tone.

So, if I read this properly:

1) Run the tracks as hot as possible without clipping

2) Bring their levels back down until the mixing bus isn't overwhelmed

3) Continue pulling back as necessary, especially w/ EQ and FX?

Yes. When you record digital, you get higher resolution the higher the input level is. But it is all shot if you clip, so make sure you have a little breathing room. Beyond that, just set the levels so none of the buses clip during mixdown...like you said in 3).... if you put additive EQ up on there...the volume will go up, so pull it back down. If you add distortion, volume will increase, so back it back down. Etc.

Kellen

kellenholgate wrote: When you record digital, you get higher resolution the higher the input level is.
Kellen

Are you able to explain that further??

JackHenry wrote: [quote=kellenholgate]When you record digital, you get higher resolution the higher the input level is.
Kellen

Are you able to explain that further??
If your signal is at the maximum peak level (right before clipping) you will use all bits (say 24) to represent the signal. If you lower the peaks by 6 dB, the most significant bit would always be zero. You would be using only 23 bits to represent the signal. If your signal was -48 dBfs (48 = 6*8 below the maximum peak before clipping) you would be using using only 16 bits (the rest would always be zero).

Doesn't resolution refer to the sample rate (44.1, 96 etc) not the bit depth (16, 24)??

JackHenry wrote: Doesn't resolution refer to the sample rate (44.1, 96 etc) not the bit depth (16, 24)??

In one sense, but try making 24 piles starting with only 12 oranges. That's the point of using all possible bits without clipping. You need to start with good signal to end with good signal.

EDIT: Unless you are referring to sample rate for tracking as opposed to sample rate for bouncing. In that case my example is backwards, but still valid. You lose quality at one stage, it affects every stage after that.

JackHenry wrote: Doesn't resolution refer to the sample rate (44.1, 96 etc) not the bit depth (16, 24)??

A common misconception. bitrate = resolution, samplerate = frequency response. For a while, every few years the manufacturers would come out with new converters with ever higher samplerates, enticing people to keep on the cutting edge of sound. But I think by now people have figured out that PCM audio above 96K sample rate just doesn't amount to much difference. And 24 bits resolution provides plenty of accuracy to represent a sample value. If a person records quietly enough that their signal is down in the "dirty" lower bit values of 24bit, then they are going to have their signal swimming in the noisefloor anyway.

When you record, do it max at -6 db (peaks). It's safe level against - inter sample peaks. Leaving the 24th bit „empty“ will make AD stage breath more freely and safety, and you have 6 db of headroom for further processing inside the DAW.

During the mix, do the same, leave master bus fader at zero, but mix to -6 db. This will make mastering engineer happy.

In the 24 bits system there’s no need to – fill the all bits exorcism.

JackHenry wrote: Doesn't resolution refer to the sample rate (44.1, 96 etc) not the bit depth (16, 24)??

Think of the signal as a graph with time running on the horizontal axis and voltage on the vertical axis. When you digitize, the sample rate determines how finely you divide up the horizontal axis; the bit depth determines how you divide up the vertical. The accuracy of the picture depends on both. I'd call that "resolution," but some people may define it differently.

BobRogers wrote: [quote=JackHenry]Doesn't resolution refer to the sample rate (44.1, 96 etc) not the bit depth (16, 24)??

Think of the signal as a graph with time running on the horizontal axis and voltage on the vertical axis. When you digitize, the sample rate determines how finely you divide up the horizontal axis; the bit depth determines how you divide up the vertical. The accuracy of the picture depends on both. I'd call that "resolution," but some people may define it differently.

Wow, that actually made sense. Good job. This analogy is assuming that all points have to fall on an axis.

Guitarfreak wrote: ....This analogy is assuming that all points have to fall on an axis.

Well, the points of the digitized signal all fall on the grid. They all have to be a 24 bit digital height and are assumed to occur exactly at the sample time. Not true of the analog signal of course. We just take the best approximation.

(After a couple of years of electrical engineering courses we would probably know the physics of the complicated process by which the samples are actually attained. If someone wants to explain it in a post I'll read it, but I'll understand if you refer me to a 400 page book. For now I'll assume that each sample is the closest 24 bit number to the exact value of the voltage at the instant that it is supposed to be sampled. No assumption like this is ever true. But (as we have seen in Washington) "true" is an elastic concept.)

BobRogers wrote: [quote=Guitarfreak]....This analogy is assuming that all points have to fall on an axis.

Well, the points of the digitized signal all fall on the grid. They all have to be a 24 bit digital height and are assumed to occur exactly at the sample time. Not true of the analog signal of course. We just take the best approximation.

(After a couple of years of electrical engineering courses we would probably know the physics of the complicated process by which the samples are actually attained. If someone wants to explain it in a post I'll read it, but I'll understand if you refer me to a 400 page book. For now I'll assume that each sample is the closest 24 bit number to the exact value of the voltage at the instant that it is supposed to be sampled. No assumption like this is ever true. But (as we have seen in Washington) "true" is an elastic concept.)
Well, what you say is pretty much true, but you gotta take into account that we don't listen back to the stairsteps as they are recorded by the A-D converter. Most reasonable people use a D-A converter to listen to an analog reconstruction of the original signal, rather than the raw sample values as they may appear in our software.
And a 1kHz signal recorded at 192kHz sample rate is no more resolute than one recorded at a 44.1kHz sample rate, so that is why I would more closely relate bit depth to resolution, and sample rate to high frequency response. Higher bit depth and higher resolution sounds like it would be a cool thing, but 24-bit already can represent a quantization noisefloor lower than the noisefloor of the gear on the analog end.