Hi,
I would like to ask what happens to the integrity or fidelity of an audio signal when it subsequently gets...

• significantly Reduced in gain because it was recorded too Hot...

And

• significantly Increased in gain because it was recorded too Soft...

Considering a DAW works with numbers, (a digitised version of an audio signal), when there is significant gain changes, is the re-digitisation process able to maintain fidelity as it crunches these numbers?
 
Obviously it's best to stay in sweet spot say around -9dB, but sometimes people might be presented with track data that can't be re-recorded. I guess also, the genre or style such as heavy rock might not make any unpleasant distortions obvious, so I guess my question is derived from the production of Ballads or Soft & Quiet styles where distortion issues may become more evident.
 
I'd appreciate your thoughts.
 
Cheers,
Adrian.
 
 

Record at 24bit and you'll find it fairly forgiving.
 
In general it's better to have the signal be quiet than clipping.
At 24bit the noise floor is low enough that even relatively low signals can be salvaged.
If you clip the audio signal often it's not something you can fix.


Here's an excellent article on recording levels from Sound On Sound
How much headroom should I leave with 24-bit recording?
 
Edited because I'm being dyslexic 

AdrianNewington
Hi,
I would like to ask what happens to the integrity or fidelity of an audio signal when it subsequently gets...

• significantly Reduced in gain because it was recorded too Hot...

And

• significantly Increased in gain because it was recorded too Soft...

Considering a DAW works with numbers, (a digitised version of an audio signal), when there is significant gain changes, is the re-digitisation process able to maintain fidelity as it crunches these numbers?
 
Obviously it's best to stay in sweet spot say around -9dB, but sometimes people might be presented with track data that can't be re-recorded. I guess also, the genre or style such as heavy rock might not make any unpleasant distortions obvious, so I guess my question is derived from the production of Ballads or Soft & Quiet styles where distortion issues may become more evident.
 
I'd appreciate your thoughts.
 
Cheers,
Adrian.
  

 
re-reading your question...
 
Crude analogy...
Think of the audio as a picture where the height of the picture is the dynamic range of the performance.
The wall that the picture is on is the possible dynamic range of the recording medium.
As long as the picture is not being pushed into the ceiling or too close to the floor - you can slide the picture up & down on the wall without affecting it in a significant way. The picture remains intact.
 
If a signal is really recorded too low you hit the noise floor and it will be noisy and possibly compromised when the levels raised.
If a signal is recorded too hot then there will be clipping & distortion.
In my experience sometimes hardware clipping might be acceptable, digital clipping rarely if ever.
 
 

If you record too hot and you clip the signal, you will hear it and probably end up using the delete button to fix it

Use 24bit and a good interface with low noise floor and keep your peaks at -20 to -10 dB. This way you have sufficient headroom and on a good interface you are 90 to 100 dB above noise floor i.e. certainly higher than the noise floor of your signal source.

You may have to use Sonar' s gain knob if the signal is not hot enough later for some plugs, but that's fine and does not compromise audio quality.

Some mics may cause problems being too hot (i.e. close to clipping) e.g. on drums if the interface does not have a pad switch ... In such a case I do allow somewhat higher level but keep an eye on it to ensure they don't clip in the loud parts

Hi Guys,
Thanks for your comments, but I feel I haven't enabled my point to get through.
 
I understand about clipping & noise floors. My question relates to the digital manipulation of converting a signal via Gain Boost, or converting a signal via Gain Reduction.
When signals are manipulated in these ways, do we lose integrity of the original digital sampling.
 
Are distortions, (however minor - and I'm not talking about reaching 0dB), introduced when a signal is boosted or attenuated significantly?
Is the waveform re-created faithfully after such large changes either way?
 
I have a Behringer USB soundcard, and it doesn't let me record in 24bit. (Bummer).
 
Cheers.
 

The basic answer to your question is no.  If you have recorded something a little hot say as long as it is not too clipped, reducing that level say be subtracting some db off the signal will not change the quality of the signal.  It will only pull it back to where you want it to be. 
 
Same as raising a signal by a certain number of db to bring it up to where you want it to be, no real change in sound quality either.  Remember we are in digital now so if a signal is averaging say around -30 dB rms FS and you add 16 db to that signal, all you are doing is changing the values of the digital signal level wise that is.
 
There are certainly much more important things you have to worry about than this. eg compositional quailty, performance quality, choice of microphone, placement, the sound of the room etc..  These will all effect things in much greater way than adding/subtracting a few db here or there.  You are currently sweating over something that actually does not require it.
 
If you are clever you will never have to worry about tracking at the right levels again.  Simply choose a reference level that represents 0 dB VU on a VST (or real)  VU meter.  While tracking simply adjust the input gain so the rms (VU) meter just hits 0 dB VU and you are done.  All your tracks will be at the same rms level ready to mix.  And no gain will ever need to be added or subtracted again.
 
When people talk about how loud an input signal should be they forget there are two components to that signal. The peak value and the rms value.   I prefer to aim for consistent rms levels coming in and let the peaks take care of themselves.
 

AdrianNewington
 
I understand about clipping & noise floors. My question relates to the digital manipulation of converting a signal via Gain Boost, or converting a signal via Gain Reduction.
When signals are manipulated in these ways, do we lose integrity of the original digital sampling.
 
Are distortions, (however minor - and I'm not talking about reaching 0dB), introduced when a signal is boosted or attenuated significantly?
Is the waveform re-created faithfully after such large changes either way?
 

 
Sonar and most modern DAWs use floating point math. With standard single precision 32bit floating point you can do hundreds of calculations on each and every sample with no audible effect (a volume change is a single calculation on each sample). Even for large projects, the accumulated calculation errors from things in the mix engine like level changes will either be too low in level to even make it into a 24 bit output or will be buried under the noise floor caused by other things.
 
And if you use Sonar's 64bit double precision engine, it's essentially completely impossible for any calculation errors at all in the mix engine to make it into 24 bit output.
 
IOW, you don't need to worry about errors from floating point math within a DAW. The important thing is to have reasonable levels when recording or exporting.
 

 
I have a Behringer USB soundcard, and it doesn't let me record in 24bit. (Bummer).

24 bit recording is actually overkill for anything except sources with wide dynamic range in very quiet environments. 16 bit is generally fine as long as you use a reasonable amount of headroom. 24 bit just makes it really easy because one can leave lots and lots of headroom and still not have to worry about quantization noise at all.

OP, it's not really a relevant question. ;) 
I just use my ears, who cares how it got there... if it sucks rerecord, if you like it,use it!  lol
You need to get it in your head that when recording, it's not right until it's right.
 
peace!

my opinion-
 
one of the most basic things to learn if you are serious about recording,
is gain staging.
 
setting levels, is the single most important thing you can consistently do.
 
if i have an under recorded track, i re-record it.
 
if i have an over recorded track, i re-record it.
 
always be conservative,
but there is a sweet spot for every recorded track,
your job is to find it.
 
everything else, is just damage control.
 

I think if you have recorded something a little too loud and it is not clipped or damaged in any way, then lowering it is fine.  Same goes for under recording too.  Adding some gain is not a big thing either.  Re recording may not be the answer especially if you have done a great take and not too sure of you can get it as good again.  The better the musician though the less this problem arises.
 
But it still better to get it right to start with and that is where the VU meter comes in.  You will always get it right that way.
 
Peak metering does not tell you much only that the highest point was at a certain level. If a sound has a low rms component and quite a large peak associated with it eg the leading edge of the sound then using peak meters will only get you a recording that is too low.  That is because the rms component is too low.
 
The only issue is you cannot use Sonar's built in rms metering as it stands now. You need a decent VU plugin to do it right. And you also need a way to monitor the signal coming in with the VU meter preferably before the signal even hits the track.