I've scratched around with the same question, to normalize or not, and I've got split answers. My current approach, accordingly, is to split the difference. I normalize tracks that are just too thin to work with, and leave everything else as is.

The bit-rate thing is pretty extensively, er, discussed elsewhere. The predominate view is that, because gain manipulation in digital (such as normalizing) reduces the effective bit rate, it's better to start with as high a bit rate as you can support. I record in 24-bit, cause I have the room and why not?, but I'm not a true believer -- for the music I record, on my system, with my ears, there's no difference between 16, 24, or 32.

As far as normalizing, I don't care for what I hear when this is done. If it works for you, I say do it.
The 24 bit to 16 bit question is a big controversy. I think 24 bit sounds way better, better bass and less jitter.. it just sounds better to me. I think it translates down to 16 bits better. Clocking at 24 bits is more stable. That is my opinion. Who knows? In the future there may be a 24 bit CD format. In that event, if you record in 24 bits, all your archives will be ready to make the transition. Your computer processes better at 24 bits. Many 24 bit systems actually use a 32 floating bit scheme for processing. Reverbs sound better, compressors are smoother. Lots of reasons to work in 24 bit. What reason is there not to? File size? CPU load? If that is an issue, then it’s time to be thinking about a more powerful computer for your system ……… Kurt
---------------------------------------------------------------------------------
To help keep this site alive and growing, please show your support by purchasing one of our caps, studio clocks, T-Shirts etc. Banner ads too!
---------------------------------------------------------------------------------

Hi Mark!

Ask yourself this.

Do you want to use a top mastering facility with your work or not?

If so, don't normalize the OM (original Master)

If not, use every dirty traick in the book to get your sound like the big boys...but it HAS to do that on every speaker imagineable.

Now, once you answer yes or no to the top questions, then you can move to the next point of your question.

If you are doing it yourself, stay in 16 bit/44.1K. Reason being, Sample rate converters and word clock converters that work well from hi bit to 44.1K and word converters from 24 to 16 bit cost more than mastering many of your albums. This said, listen to your work. If you are pleased with your finals on many systems, do not want to spend the way less than a thousand to get your product mastered professionally, then just listen on many systems and be pleased..but stay away from Word and Sample conversions unless you have the goods.

Errors go up exponentially in conversion without the top flight conversions.

I don't think there is much controversy really about 16 bit vs. 24 bit. Just about everyone I know can hear the difference pretty easily. I would recommend recording at 24 bit and then dithering down to 16 for CD. All of your processing will sound better in the 24 bit (or preferably 32 bit float) domain. I would send the 24 bit file to your mastering engineer if he or she will accept it and let them do the dirty work. Keep your levels up where they should be and you shouldn't have to normalize. Bottom line though is if you're wondering just try it and compare it with the un-normalized version. This will work as long as you have some form of undo with whatever platform you are using. I usually prefer to hand edit volume spikes or use automation.

I wonder if he is using something very low end or something that is pro. (and proven for the SRC WCC)

It does make a difference.

"It does make a difference."

Yes it does. Good point Bill. Most of the new stuff out there is sounding pretty good and prices are reasonable. Programs with 32 bit float are now way under 1000.00. :tu: My analog 16 track has been collecting dust although I have to admit I get a hankering to get it up and running now and again.

Um, I don't ever normalize my whole mix to 0db, for just the reasons reported. The question is whether to normalize individual tracks after tracking. For whatever reason, I often wind up with some tracks that are just weak, and those I normalize. There's a temptation, for me at least, to normalize every track, and in some ways it makes things easier. I'm not going to say that I noticed any degredation when I normalized every track, I just sort of sense it's not a good idea.

What about normalizing to less than 100 percent? Would that reduce rounding errors?

jenrick,

> What about normalizing to less than 100 percent? Would that reduce rounding errors? <

No. Every operation adds rounding errors, which really is just a tiny amount of distortion. There is no difference between normalizing - where a program decides how much to raise the volume - and you raising the volume by the same amount manually.

Generally speaking, normalizing should always be left for last. If you normalize a track, then lower the volume to fit it into the mix, you have performed two volume changes instead of just one. Every time you change the volume or add EQ or whatever you add rounding errors. But again, the amount of distortion is quite low, and probably not worth obsessing over.

--Ethan

I'm using decent stuff... Nuendo and a motu 2408, Mackie board and monitors. Pretty nice set-up.

The question wasn't about the overall mix, it had more to do with individual tracks.

I appreciate the info on 24 vs 16 bit. When I'm laying down ideas I'be been using 16 bit, when I'm serious I use 24. I've been asked these questions for a while and thought I would get educated on the matter.

Thanks everyone!

Mark,

> The question wasn't about the overall mix, it had more to do with individual tracks. <

There's never a good reason to normalize a track. Just raise it as needed when mixing.

--Ethan

Whether you laying down ideas or not, you should record at 24-bit. Lot's of records have been cut with the scratch vocal or guitar becaue it was such a good take.

My opinion, of course.

I can definatly hear a difference between 16 and 24....I can even SEE the difference when looking at the wave file. Even at 24/44. Seems to have more substance.

I'm not a newcomer to daw, having worked a lot with Acid, Vegas, CW etc... but I am new to HD which is far more complex an issue. I have used a "Babtism by Fire" approach to learning this :p

Normalizing has no benefit at all other than right before the sound comes out of a very low quality D to A converter. The extra math and dithering (or distortion if you choose not to dither) will always reduce the resolution of the sound every time you make any kind of gain change. Increasing the volume can never increase resolution that was lost at the time the sound was recorded.

One of the best tricks for achieving high quality results is to absolutely minimize the amount of math performed on the signal using no more than just what is subjectively necessary.

Whenever you make a change or edit to a sound file, there's just a little bit of error. For example, turning up a particular track 1.5db to make it sit in just the right spot might make the precisely perfect location for that bit of info to fall between 2 choices, and the software has to interpolate. Usually impossible to notice on it's own. But if I'm laying down 10 channels of drums, 2 or more for guitars, 1 for bass, and then however many for keyboards and vocals, and all the little errors from each operation build up in there, over time they might be noticeable.

Well, if I had recorded in 16 bit they would be a LOT more noticeable. 16bit = 65k +/- possible places for the speaker to be, and although its remote, there's a chance that your edit might cause the 'perfect' spot to be between 2 of them. In 24bit, since there's 27 million possible "positions" for the speaker to be in, there's a better chance that the perfect spot will fall right on one of the available positions.

So if you record in 24bit, do all your editing there, and then the very last thing you do is to dither it to 16bit, all the errors that built up would be factored out by the dithering to 16bit.

If you understood any of this, please explain it to me...

It's like drawing a circle on MS Paint. Draw it as big as you like, then zoom in on it- it's not perfectly circular, it's a series of zig-zags formed by perfectly horizontal and vertical lines. The more resolution you use, the more circular it may look- in fact, you won't be able to notice it if the resolution is high enough- but it'll still be digital. Same thing applies to audio.

Another vain attempt to contribute worthwhile information....

Something to remember when comparing a normalized track to a non-normalized track.

Make sure you do something to match the volumes during comparrison so both tracks appear to be the same volume.

Otherwise, since louder sounds better to the ear, the louder (normalized) track will always sound better.

Normalize???????

Since when did NORMAL anything sound, look, feel better.

LOL :p

There are two processes of normalization that I'm aware of: one is equivalent to an increase in volume, which could cause distortionclipping, and one that applies a limiter as part of the process.

Sometimes when you master, a hard limiter on the whole bandwidth sounds good, but generally speaking different areas of the spectrum require different processes of limiting and compression.

The idea here is to maximize the average volume of a mix, and maintain dynamic range. You can't always do that if you attack the whole thing.

My opinion is that normalization is something which can be avoided, and replaced by better processes, like multiband compression.

As kurt said, If it works for you then by all means go nuts, but I would reccomend learning better methods of increasing the loudness of your mix.

With respect to the 24 VS 16 bit issue, I think that 24 bit is the way to go when tracking. It simply has more detail than 16.
AD-DA conversion is simply a connect the dots version of an analog sine wave. The higher bitsample rate, the more dots you have to "re-draw" the initial wave.

With respect to low frequencies I don't believe that you would be able to tell the difference. Low frequencies are longer wavelengths and would be reproduced identically from 16 to 24, and probably even in an 8 bit format.

What you may percieve as better bottom end may actually be more acurate midshighs creating a tighter sound, but I don't believe bass could be produced more accurately.

I do think that there is a difinitive difference in the highs, both with detail, and smoothness, so my vote is for 24!

I record in 24, then I mix through an analog board into 16. No dithering required.

A great question would be why would you have to normalize anything if it was tracked at the right level? When I meen right level which deffinately varies from person to person I meen don't record supper hot that everthing is going to be close to full scale, but don't record to cold that your going to end up having a low track. OK for instance I have my converters set to -18dbfs=0 dbu so I track between -18dbfs and -6dbfs so that I have headroom for mixing and I'm not running my gear at +18dbu just to reach full scale, and the reason I don't record lower than -18dbfs (unless the instrument is really dynamic) is that I don't wan't my gear running cold and get a bad signal to noise ratio. Also normalizing your tracks are going to give yourplug-insless headroom, thus you might have to turn down the plugins inputs if it does have one to prevent the plugin from clipping internaly.