AJScott755 wrote:Hi CS70, thanks for your input. It sounds like I have a fundamental misunderstanding to some key recording elements especially around volume/gain. I read that link you posted and I have some questions.
Now I have learnt everything self taught so far, no audio production lessons at school etc just YouTube and trial and error, so please bare that in mind.
No worries, that's exactly why SOS forums are so useful :)
I get the overall gist of recording low (-18dBFS) and during mixing raising the playback volume on my interface to make it listenable and then I'm assuming during the mastering stage bringing everything up to a reasonable level.
Indeed. The whole point of mastering is primarily to remove the working headroom. If u look at the other posts on the blog, you find a couple on the subject as it's another common misunderstanding.
Let's take the 6-string guitars for example in my mix, there's an acoustic and an electric. I assumed it would be wise to capture as much information as possible and then turn it down in the DAW, so the gain going in was set quite high on both instruments but not too high so it peaked. Is this not the correct way to do this? What would be the correct process?
What is important to understand is what you mean by "as much information as possible", which is not what you probably intuitively do. The information you capture depends on the properties of the recording medium, not the source (and the input gain is part of the source).
You can never have more or less information (potentially) that your recording medium allows. Specifically, it depends from the ability of the medium to retain the detail present in the analog signal so to reconstruct it later - from the quietest to the loudest and with as much good resolution as possible in between - called "dynamic range".
The larger the dynamic range, the better. If you had 2 bits, you could record 4 values (say "low", "medium-low", "medium-high" and "high". Wouldn't be much of a recording!
With 24 bits, however, you have an enormous dynamic range: you can represent 16777216 discrete levels (that's almost seventeen million
!). So if you record a single signal averaging "middle of the full scale" you have a really huge amount of possible levels up and down from that middle. Gigantic. Pairing that with an adequate sampling frequency (say 44.1 or 96 KHz), you can potentially capture an enormous amount of information.
Now keep in mind that all signal paths will have some noise. That noise appears as random non-zero levels at the very bottom of the scale, it's called "noise floor" and will be around, say, levels 0-10 or 0-100 or whatever (depending on how good your signal path is).. But with 24 bit, when you are "in the middle" (averaging say around the 8 million mark), that becomes irrelevant ! 10 or 100 compared to 8 millions is insignificant. Whereas, if you had a recording medium with lower dynamic range (say 4 bits, which gives you a range of only 0-16) the noise (which sits at the 0-10 range) would be huge with respect to the signal! So you would need to try to record as "high" as possible in the scale. With lesser quality mediums than digital - say old studio tape - the dynamic range was much lower than 24 bits, so it paid off to stay as "hot" as possible (there were also other reasons, but from a information point of view, that was it).
No need with digital recordings: the moment you decide to record at 24 bits, you have chosen a medium which can potentially capture all the information there is to capture - your only problem is potential overloads (hence, averaging in the middle).
Staying "in the middle" also helps due to the technology involved: due to the way they are physically built, the front end of the A/D converter will be more comfortable with lower levels, and common A/D converters often produce a trifle more quantization errors at the limits of the scale, and they are a little better in the middle (this is an overgeneralization, but better safe than sorry: bit like avoiding running a car engine near its rev limit if you can).
Also, regarding the -18dBFS. How should this be achieved? Is it best to record a signal coming in at this level, or should I use the same practice where I have the input gain at a reasonable level (but not clipping) and then set the individual tracks in my DAW to -18?
You simply raise the preamp gain until the point where, when you play the instrument or sing, the DAW meter shows around -18dBFs (middle scale) in the DAW. Simple as that.
If you look at the post in the blog https://www.theaudioblog.org/post/how-d ... recordings
you'll get more details.
When it comes to EZDrummer (I use it all the time myself for demos and inspiration for new songs :) ) - it depends a bit on your DAW I guess, but what I do is simply to use the DAW channel gain, lowering it down so that when the fader is at unity, the channel averages -12, -10 dbFS.
The reason for -12 is because a drum kit (which EZDrummer simulates) is actually a composite of many individual signals.. if you recorded the individual mics at -18dBFS average and summed them up into a bus (how you would do with "real" drums), the average bus level would be in the -12 ballpark.
Doing that will give you a "natural" balance (well, natural if you've ever miked a drum kit :))
You can also lower the individual mic faders in the EZDrummer mixer, but it's just more cumbersome. As you do for a "real" drum kit, you find a balance of the individual mics and then you send to a drum bus and then you control the drums level with the single bus fader. But from a signal point of view, so long you stay in the -12 or -10 average, it makes no difference.