if digital compression is evil, what about "normalizing"?

OK, if digital compression distorts the audio information, by squashing the dynamic range what about digital "normalizing?"

what exactly is normalizing? does this just find the peak passage, bump it to "0" and then move everything up in correlation?

obviously normalizing rewrites the digital file, but does it preserve the dynamic range?

 

There are two types of normalization:

1) PEAK NORMALIZATION: Normalizing PEAKS simply finds the maximum sample and increases it to a predetermined level. This does not get all of your files the same volume. What some say makes it bad is that the program must recalculate the file, and therefore potentially degrade the sound. This only happens IF you do it on a 16-bit file, and the question of if it is audiable to the average person is debatable. No harm will come if you do this on a 24 or 32-bit file. It iS useful if you are doing a whole album at once so that only the loudes peak in the album gets to the top.

2) RMS NORMALIZATION: This is the process of getting all of the files the same volume, what most people doing compilations want to do. What is usually done is that the program measures any wave file you pick to determine the model level it uses to make all other files the same volume, or the user can pick a value. Here, if any stray peaks go over FS"0" on a 16-bit file, one can use a limiter to keep those peaks from going over if it is triggered. Hopefully, if the right decibal value is choosen, this either won't happen much or not at all. Again, this will NOT harm your sound IF you do it on 24 or 32-bit files, but it might on 16-bit files.

Then, there is the old way of using your ears. Because some digital audio editors are less accurate than others in the RMS normalization, one must rely on the old ears to make the adjustments. Thankfully, the program I favor for this function, Cool Edit Pro 2.1 is very accurate, I don't worry about it. In Cool EDit, If I do it by ear, I just ude the peak meters because they do measure the relative, or average volume if I set it to do so.

Analog people, and those with VU meters, or meters on their DAWs, simply take a reading of the average volume based on what the meters show. And, they still use their ears.

Just one more thing: If you do this by ear, do not use headphones. Headphones do not deliver, or you will not hear all of the ambiant cues that will help yopu determine what is really loud or soft.

 

The general view in audiophile recording and mastering is that normalizing is to be avoided. You are putting the sound file through an additional generation of Digital Signal Processing and only the best software can resolve the digital word lengths in a satisfactory way. The process is equally destructive no matter what bit-rate is involved.

It's one thing if you're just making a CD-R compilation to play in the car, but it's not a good idea if you are preparing your own project for mastering.

That being said, the rampant use of digital limiters and compressors in today's "professional" recordings make normalization seem pretty insignificant in comparison.

 

so if compression and normalizing are to be avoided, how can a CD be mastered to give all the tracks a unified, consistent sound in terms of volume?

 

It can be destructive in 16-bit mode, but I fail to see how it can do any damage in higher bit-depths. Higher bit-depths provide more calculating room so that no truncation of bits occur. The software does not dither at high bit-depths so no worry about that, either.

 

Do it the way Mr. Hoffman does it - in the analogue domain before the signal is converted to digital.

I should have been more clear in my post. I believe the term "normalizing/normalization" only describes a digital DSP process. Compression and limiting can be performed in both the analogue and digital domains.

 

This is a good point - I'll have to do some more reading on this. I'm sure that there are some software programs that do a better job of the DSP conversion than others...

 

I've already outlined the two types of normalization above. The only way you are going to get your songs the same level is if you record them into your computer in real time through the soundcard and if you manually adjust the levels going in.

Your music is not going to be issued to the audiophile buying public, and you will probably not be able to HEAR the difference, so just follow my guidelines. It will be fine.

 

There are! I use two of them.

When the pros, like Steve and Bob Katz warn against doing normalization, they are assuming people are going to do it all in 16-bit. This is one reason why we have higher bit-depths in software programs today. It used to be not too long ago that all pro and home software only operated in 16-bit only, and some didn't even use dither! I still have two of these older programs installed on my machine.

Remember, these guys use mastering gear and rooms that hardly any of us will ever use or hear. And, they are anal about sound. They can hear the degredation done in 16-bit mode. Well, I can too, but I know what to listen for. It is so slight with only one process that it may as well be inaudiable.

Just don't do a bunch of stuff in 16-bit mode or you will be sorry!

 

Using the old "invert paste" trick, I haven't found this to be the case with 32-bit files in Cool Edit. In fact, even after multiple transforms, inverted pasting of the edited file over the original file leaves nothing but digital silence. In other words, the two waves are identical. 16-bit is another story.

 

Most people are simply going to rip their CD tracks into a computer program and cut the CD for their own use. They just want to get the volume uniform. No one is going to go through the trouble to hook up their CD player to their Soundbuster Live card and try to do this with no external attenuation. This is why you use the normalization types in the editor programs.

 

Exactly the point I always try to make. These pros warning against doing anything assume you are going to do it in 16-bit, which most users do anyway. higher bit-depts are a different story. 32-bit is a different ballgame.

 

How about programs (like Feurio) that are able to do normalizing on-the-fly during burning? Does that degrade the sound?

 

It changes the sound. It certainly changes the wav data, but argueably changes "what you will expect".

There's a reason most MFSL and DCC discs set their peak low. Investigate, and you'll find out why.

 

If your project contains a variety of rockers and ballads, for example, I'd personally avoid normalization as the ballads will seem unnaturally loud when normalized.

 

Thanks, Sckott (and Grant, as always). I will do my normalizing in 32-bit from now on.

 

Cool Edit Pro upgrade worth it?

Hi Grant,

Do you find Cool Edit Pro 2 works better than Cool Edit Pro 1.x if all you're using it for is CD compilations, plus CD versions of LPs and tapes? Most of the new features seem geared for more creative music making, so I've passed on it so far. But I know you're using Cool Edit Pro for a lot of the same things that I am.

Thanks,
Joe

 

I think what's more audibly destructive than normalizing is working at different bit-depths and not dithering. It's important to dither at each step where the bit depth changes. On a pro DAW, normalizing isn't all THAT destructive compared to a) limiting b) not dithering or c) sample rate conversions. Although "level matching" is the better way to go -- best done by ear. (No meters).

Ideally the level changes should be done in the analog domain, but what if you have digital sources? (CDs, DATs, Pro Tools sessions?)

 

But DCC and MoFi do not do high-bit. The go directly from analog master to 16-bit. Therefore, it is unwise to do any normalization, or anything, for that matter. At 16-bit, it does indeed change the sound. For those just making comps from other CDds at home, will it matter? Here's one catch: most CDs today are from high-bit masters and utilize dither to go to 16-bit. One should never dither twice because THAT is the main culprit that will change the sound. So, the better way is to NOT use any type of dither if processing in these cases. Yes, the sound will still change, but will be far preferable to double dither.

So, the enemy is dither or truncation. You can do all sorts of stuff at high-bit, but not 16.

 

Not if you use the Equal Loudness contour in Sound Forge or the one in Cool Edit Pro 2.1...

 

Re: Cool Edit Pro upgrade worth it?

CoolEdit Pro 2.1 is fime for CD comps and LP to CD stuff. I also use Sound Forge 6.0 for various things.

BTW, that multitrack comes in handy for comps!

 

Well, of course that's the best way but for that you need to precisely establish analog peak levels for the complete recording before you start the transfer.

Conceptually, I find that recording with a large bit-window leaving some safe but not excessive headroom, then normalizing, and later dithering to 16-bit sounds like the right way to go.

I understand current normalizing algorhythms use floating-point math to minimize any degradation. With large numbers (24 bit or even 32 bit), my guess is that no harm is done.

 

I don't know about MoFi's transfer process. Your cautions about 16-bit are simply because one wants to preserve as much of the 16 bits as possible (and all 16 bits are not always used). Dithering is necessary when going to progressively lower bit depths. And you can do it as many times as needed. (It's maintaining the resolution, not degrading it). The worst thing you can do is truncate. THAT should be avoided at all costs. See http://www.digido.com Click on "articles" and then "dither".

But yes, for the home CD burns, dithering is probably not necessary.

 

I totally did not know this was about hi-bit audio capture.

This is just me talkin', but I would only capture in the same bitrate that I plan playing. I'm a little luckier than most because I have a DVD+Writer, but my soundcard isn't too hot at high-bit record rates, and I don't have anything to do at 96/24.

IMHO, I see NO reason to dither back from high bit rates. The capture should be AS/IS/WAS. Meaning, if the playback intended is 16 bit, then 16 bit it goes.

To do something in 96/24 and then sample to 16 bit, you're gonna get something that doesn't necessarilly sound right. It's kinda like Brendle-Fly'ing audio. Absolutely wacky. I don't care how accurate the smoothing is. It doesn't work like JPEG graphics. Capturing big and reducing? No thank you.

It should always be real-time at any possible stage. I know that the labels don't always see it that way. If for home mastering, I betcha if you compared, you might like things done straght... Who knows!

 

Sckott, it ain't that simple. What goes in at 16-bit does not come out at 16-bit. By that, I mean that you aren't using all 16 bits in the word all the time. Every digital move, transfer, fade, gain change, etc. involves some amount of data loss. Which can lead to an audible degradation.

Check out the link above. You'll see why it's not so straightforward. As for 96/24, which gives you more clearer detail -- 35mm film or 16mm film?

Sorry this gets away from the inital topic, but while normalizing is a digital process that introduces some loss, it's not nearly as bad (when done correctly) as some other digital processes (such as bit truncation). If you're making CDs at home, don't worry about it.