Topic: Do you normalize?

[ShawnF]

Ok, so obviously I had the wrong impression of what normalizing actually does.  I thought that normalizing took the dynamic range and compressed it too raise the volume to a consistent specified level across the whole audio file.  As to have less peaks and valleys in the db range, and make it louder across the board.

So what the hell is it? 

Thanks for all the disscusion,
Joe

Isn't this just hard limiting?  This is different than just raising the volume x amount as you are supposed to avoid clipping with this method.  And the Waves L2 Ultramaximizer (is there an L3 now?) is supposed to be pretty good at it.  I like using this to bring down the level of the audience (especially applause) relative to the music.  Basically hard limit to the point that the music is peaking at or near 0 db so that no dynamic range in the music is lost, but the applause is not as prominent, and neither is it clipping or pumping if done well.  The only compression is on the "excessive" audience noise.  Maybe this is more useful in stealthing when one is in the midst of it . . .

[SparkE!]

If you are lucky enough that normalizing is accomplished by mulitplication by an integer power of 2, then you lose nothing in S/N ratio since there is never any rounding.

Do you mean it's better to rise the gain by 6 dB instead of normalization? If I have peaks lower than 6 dB then should I just multiply the signal by 2? That won't be normalization to 100%, but to a some percent close to that and leave the same s/n ratio?

I'm not sure I get your point. Lets say I have S/N ratio at 60 dB according to my recorder specs. I get some constant noise at that level. Now I record at low levels with peaks say at -12 dB. Now my signal to noise ratio is 12 to 60, which is less than when I record at high levels (like 0 to 60). Then I gain the signal, and I gain the noise too. Or am I completely wrong here?

In terms of S/N ratio, it's better to stick to integer powers of 2 for the increase in gain.  That corresponds to shifting the binary representation left by an integer number of binary places.  Each shift corresponds to an additional 6.02 dB of gain (approximately), but there is no loss of information in the lower order bits since they are all 0's.  (This explanation probably is not the most lucid explanation I've ever attempted, but I hope you get what I'm saying.)

That said, I tend not to normalize unless I need more than about 3 dB of gain to hit the max.  I know that I lose a small amount of S/N, but the resulting signal level is closer to that of commercially produced CDs when the recording has been normalized.  Besides, I always keep the original recording in case I ever want to do some really critical listening and don't mind turning the volume up to do so.  I'd rather just be able to throw a CD in the player an know that I won't have to adjust the volume knob to be able to hear it at the volume I expect.  Purists will stick to 6.02 dB increments in gain so that they don't lose any S/N and just get as close to 0 dB as possible.  I guess I'm not a purist, even though I know better...

[BC]

Ok, so obviously I had the wrong impression of what normalizing actually does.  I thought that normalizing took the dynamic range and compressed it too raise the volume to a consistent specified level across the whole audio file.  As to have less peaks and valleys in the db range, and make it louder across the board.

So what the hell is it? 

Thanks for all the disscusion,
Joe

Isn't this just hard limiting?  This is different than just raising the volume x amount as you are supposed to avoid clipping with this method.  And the Waves L2 Ultramaximizer (is there an L3 now?) is supposed to be pretty good at it.  I like using this to bring down the level of the audience (especially applause) relative to the music.  Basically hard limit to the point that the music is peaking at or near 0 db so that no dynamic range in the music is lost, but the applause is not as prominent, and neither is it clipping or pumping if done well.  The only compression is on the "excessive" audience noise.  Maybe this is more useful in stealthing when one is in the midst of it . . .

I do this a lot for exactly this type of situation, run conservative for the master, then afterward hard/soft limit the applause so it peaks a few (typically 3-5) dB below where the music peaks, then finally use the waves L2 ultramaximizer to boost everything up. Did MMW acoustic 4-18-03 this way (http://db.etree.org/shninfo_detail.php?shnid=17742), turned out pretty good. I really like to squash the applause and crowd noise, so the guy clapping next to you does not blow out your ear when you have the recording cranked on your stereo or headphones. 

[SparkE!]

Ok, so obviously I had the wrong impression of what normalizing actually does.  I thought that normalizing took the dynamic range and compressed it too raise the volume to a consistent specified level across the whole audio file.  As to have less peaks and valleys in the db range, and make it louder across the board.

So what the hell is it? 

Thanks for all the disscusion,
Joe

Isn't this just hard limiting?  This is different than just raising the volume x amount as you are supposed to avoid clipping with this method.  And the Waves L2 Ultramaximizer (is there an L3 now?) is supposed to be pretty good at it.  I like using this to bring down the level of the audience (especially applause) relative to the music.  Basically hard limit to the point that the music is peaking at or near 0 db so that no dynamic range in the music is lost, but the applause is not as prominent, and neither is it clipping or pumping if done well.  The only compression is on the "excessive" audience noise.  Maybe this is more useful in stealthing when one is in the midst of it . . .

I do this a lot for exactly this type of situation, run conservative for the master, then afterward hard/soft limit the applause so it peaks a few (typically 3-5) dB below where the music peaks, then finally use the waves L2 ultramaximizer to boost everything up. Did MMW acoustic 4-18-03 this way (http://db.etree.org/shninfo_detail.php?shnid=17742), turned out pretty good. I really like to squash the applause and crowd noise, so the guy clapping next to you does not blow out your ear when you have the recording cranked on your stereo or headphones. 

Good point.  Sometimes you have to choose between preserving the S/N and dynamic range of the original recording and making the recording more listenable.  Your compression limits the dynamic range, but oh well... If it makes the recording more listenable, isn't listening to our recordings the whole point of doing the recordings in the first place?

That said, I still think it's a good idea to archive the origninal, unmodified recording. I normally archive copies of the complete original, the complete normalized and the trimmed/tracked CD-ready files.  Duplicate copies of the archived originals are also kept offsite in case the house gets hit by a tornado or fire or thief or whatever else might threaten my access to the originals.

[Nick Graham]

Just to echo the sentiments of a few others, I try to run my levels as hot as possible during the initial recording. If by chance my levels aren't quite loud enough, I'll do a slight volume increase, maybe 1 - 2db, but that's it.

As a wiseman once said, if you're not getting the occasional red light, you're running your levels like a pussy...

[nickgregory]

some rigs you have to be afraid of clipping....some you dont....my current one I have no concerns with clipping, run hot with no audible clipping...with those rigs there is no excuse to run conservatively...for rigs where it is audible...you still should push it....just not clip it....normalizing really is noticeable when it is any more a db or two...

[BC]

Just to echo the sentiments of a few others, I try to run my levels as hot as possible during the initial recording. If by chance my levels aren't quite loud enough, I'll do a slight volume increase, maybe 1 - 2db, but that's it.

I agree here, for open taping at concerts with PA systems I can always get my levels to peak between -2 and 0 dB, so in general no need for normalizing/boosting here.

I run more conservatively for classical (which can have huge swings in dynamic range), jazz up close (same), and stealth (since I can't check my levels during the show and if I am running hot the whole time it is tough for me to get my mind off worrying about severe clipping, which makes me enjoy the show less. So I like to back off 5 dB or so and know that I am all good no matter how loud it gets. Plus stealthing I am usually taping classical or jazz also, so same concerns about dynamic range swings + with not being able to see my levels = more conservative on the V3 ) 

Plus I figure for these conservative situations, I am getting max 113 dB dynamic range on the V3 @ 24 bit, so even if I peak at -10 dB on the music (really low!) I am still pulling slightly better resolution than 16 bit with perfect levels. I think this extra dynamic range of the raw 24 bit master really helps the quality of the final normalized and dithered 16 bit CDR version. I don't think it would sound as good if I ran so conservatively with a 16 bit device and then did significant normalization. I had this old Reconstruction sbd DAT (3-8-79?) where the levels on the DAT were really low. I peak normalized the levels to just below 0 before burning it to CD. The normalized version sounds much more in-your-face which is nice, but still sounds kind of flat, like something is missing. I can't put my finger on exactly what, but I just feel like the music is not "all there", which I attribute to the low levels on the original DAT, so effectively the master was maybe a 14 or 15 bit recording. When boosting levels, the volume change helps but since the original master was lacking in ultimate resolution, the boosted version, while sounding louder, still shows the effect of the low original resolution, and maybe this effect is even more pronounced with the louder levels of the normalized version. Anyway, just thinking out loud...