Digital recording levels - safe or pushing?

  • Thread starter Thread starter hrn
  • Start date Start date
Excuse my ignorance,
but on the topic of recording levels...

Is there a difference in the bit-depth (i think thats the term) you will capture by raising the trim (preamp) on the mixer versus raising the fader on the same channel?

For example, if I adjusted my fader all the way up and then adjusted the trim accordingly... versus having the fader very low and cranking the trim.

Say I am able to achieve the same relative levels for each of the two above situations...is there a difference in quality/noise depending on how you do it?

And how do YOU all set them?
 
Trog- It's all the same as far as the recorder is concerned. Always set you trim as hot as possible without clipping then bring up the fader preferably to unity( the 0 mark).

Timmy-
You fucking son of a bitch. How dare you contradict me! Just kidding.

I'm by no means a digital audio expert. My background is more in live production and video/film work. But it is my understanding that bit depth equals dynamic range. The higher the bitdepth the higher the dynamic range. When you record at lower levels you are sacrificing part of that dynamic range plain and simple. Whether that is any great loss or not is up to you. It's not just about Signal to Noise but also about the subtle nuances and small details of trailing reverb tails and natural ambience that and all the other little reasons why everyone has decided 24bit is better than 16bit.

As far as normalizing not affecting the audio- I was trained in the world of analog audio and when digital audio came along we were told that you could finally make EXACT copies. Up until then you only had one real master version and everything that was duped off of that was inferior to some degree. You never did anything to the master that might compromise it's quality. You didn't even touch it or play it more than you had too.

Once you capture the original track the overall quality, texture, clarity etc is not going to be any better. That is the audio in it's most pristine stage. You can make it sound different with processing and you may decide that different is better for the context of the song but for the most part your best chance at getting clear, sharp, warm, punchy (whatever) tracks is from the very beginning.

I think people have taken the concept that because you can make perfect digital copies that you can also apply perfect digital gain adjustments, EQ etc. But ANY process is going to change your original file whether it's digital or analog. That is a fact and is undeniable. The audio has changed. Whether it is good or bad or doesn't matter is up to you.

My general philosophy is to record it the way you want it to sound and fuck with it as little as possible That's why I despise things like Exciters and BBE's that are supposed to just magically make your audio sound 'better'.

While you would seem to advocate recording at a low level then normalizing. I would prefer recording at a high level and never having to normalize. It just fits my philosophy of not messing with the audio more than I have to.
 
TexRoadkill said:

Timmy-
You fucking son of a bitch. How dare you contradict me! Just kidding.

:D Glad there's no hard feelings Tex. I'm no expert either, but here's the way I see it.

Imagine a situation where you hit a chord on a guitar, it's picked up by the mic and you set the level so that it peaks at -4 on your digital recorder. You then stop playing and look at the level being picked up by your mic and see that background noise sits at roughly -60. The dynamic range of that guitar track (from "silence" to full scale) will then be 56 dB. That 56 dB can actually then sit anywhere within the dynamic range of the converter. You could turn it up so that it peaks at -1 but you would also turn up the noise floor so that it sits at -57dB. You could turn it down so that it peaks at -14 and in doing so reduce the noise floor to -70. For the converter though, as long as it has enough bit's to represent the dynamic range being presented to it, it doesn't care where you put it.

Whether that is any great loss or not is up to you. It's not just about Signal to Noise but also about the subtle nuances and small details of trailing reverb tails and natural ambience that and all the other little reasons why everyone has decided 24bit is better than 16bit.

Only if the reverb tail is audible above the noise floor of your studio. Even if signals could be perceived that were 20dB below the noise floor (and we take the example above with the noise floor being at -70) the reverb tail would still be within the dynamic range of the converter.

While you would seem to advocate recording at a low level then normalizing. I would prefer recording at a high level and never having to normalize. It just fits my philosophy of not messing with the audio more than I have to.

Thinking about it, I have to agree with you on not messing with the audio any more than you have to. Rather than normalise then, why not just run all your audio in at -14 and mix at that level?

Thanks for everyones responses to this thread by the way - it really is getting me thinking. Although I believe what I've written is correct, I'm not an expert and that point about 24bit has really got me wondering. If someone could give me a clear cut example (like the guitar scenario above) that illustrates why recording hot is better, then I guess I could be 'converted'......ahem....... I'll get my coat.....


Tim
 
Your example is correct but it assumes a pretty noisy room. In that case subtle nuance is not really an option. I'd say correct your room noise problems then you can actually take advantage of having a good dynamic range. Put the computer in a closet, turn off the A/C, unplug the fridge, etc.

One argument is that since most modern music is highly compressed that it doesn't matter if you take advantage of the higher dynamic range. I would argue that the opposite is true. When you compress it brings up the noise floor so if you plan on compressing heavily you better have any noise floors (ambient or electrical) as low as possible.
 
I think I just learned more than I have in a LONG time.


TexRoadkill said:
My general philosophy is to record it the way you want it to sound and fuck with it as little as possible That's why I despise things like Exciters and BBE's that are supposed to just magically make your audio sound 'better'.

Damn right.
 
Timmy2000

If you want a clear cut explanation of why recording hotter does NOT give better "resolution",you're going to have to kneel at the Shrine of Nika (http://recpit.prosoundweb.com/)........by the sound of it though,it seems you have already been there :D

But you already have the right answer anyway.....AS LONG AS your signal is reasonable distance above the noisefloor,and your noisefloor is reasonable distance above your cards absolute lowest limit,then its all good....strange but true.

A weird example:-

A hypothetical 24bit card has a "real world" dynamic range of 100db and i am recording in a room that has a noisefloor of -70dbFS.My signal peaks at -2dbFS and troughs at -12dbFS.
10db of dynamic range in the signal,58db of "footroom" between the signals lowest point(s) and the noisfloor,and 30db of footroom between the noisefloor and the cards lowest possible limit.

now I turn my signal down and re-record it,so it now peaks at -20dbFS and troughs at -35dbFS.I now actually have MORE dynamic range(15db instead of 10db) than i did before.The noisefloor is still at -70dbFS,so i still have 35db of footroom between the noisfloor and my signal.(notwithstanding the fact that the noisefloor would drop if i turned my signal down,making this scenario,in a way, even more favourable,but this is for examples sake)

All i have to do is turn the 2nd signal up post-recording and it will have MORE dynamic range than the 1st signal i recorded....18db hotter!

And because the noisefloor on the second signal is still 35db below my signal,it will be no more or less audible than on the 1st signal that has the floor nearly twice as far below it(58db)

Strange...............
 
Last edited:
Morgoth said:
...But you already have the right answer anyway.....AS LONG AS your signal is reasonable distance above the noisefloor,and your noisefloor is reasonable distance above your cards absolute lowest limit,then its all good....strange but true.

After a bit of re-thinking and exposure to this subject, yea, I'd say true, but not even strange anymore.:D

Since I don't follow why the numbers in you example change when you lower the gain staging, how does this version sound?

Peak at -2fs, trough -12fs (10 db range), room -70fs (58db below music), A/D analog noise -100fs (30db below room), quantization error distortion -144(?).

Now lower the mic-pre or input gain 18 db.

Peak at -20fs, trough -30fs (still 10db?), room -88fs (still 58db below?), A/D analog noise still -100fs(? 12db below room?), error distortion comes up a few db(?) but still 30 or more below the A/D analog noise(?)

I'm not even confident this is accurate. But it seems that even if you include your room-to-peak level as an important part of your desired signal (which it often is), even at the lower range, the whole -20-- -88db fits nicely within the card's analog range, and easily covers up conversion error-noise by a long shot. AND - you still have 20db, a shit load of 'Oops, he just used shit mic' technique' on top of that.

(+/- 3db, HMMV, I could be still getting parts of this all screwed up but I'm trying, not withstanding. :D :D :rolleyes: )

Wayne
 
Yeah..........your version does sound much more plausible(and ledgable :D )

But,like you also mentioned, at the end of the day there was still enough dynamic range to capture the signal "in its fullest",and then some.....even when the fader was DROPPED by 18dbfs!!!

This is the main point that i (we) was trying to make :)
 
mixsit said:

I'm not even confident this is accurate. But it seems that even if you include your room-to-peak level as an important part of your desired signal (which it often is), even at the lower range, the whole -20-- -88db fits nicely within the card's analog range, and easily covers up conversion error-noise by a long shot. AND - you still have 20db, a shit load of 'Oops, he just used shit mic' technique' on top of that.

But when it comes time to mix if that signal needs to be boosted higher than unity you are introducing another gain stage that could have been avoided with a hotter recorded signal.

It's a small thing but 24 ch of extra gain add up to a lot on an analog mixing desk. With digital it probably wouldn't be a SNR issue but you would be adding extra DSP to the raw track.

Like most issues in production it's all about trade offs.
 
Back
Top