Is normalization on vinyl rips necessary?

Is normalization on vinyl rips neccesary?
If so, what level? I've been doing -0.7db, prior to that -1.0db.
A recent rip of Spyro Gyra's Catching The Sun, I'm wondering if normalizing is even going to be neccesary.

 

I think normalizing can degrade the signal slightly. I try to set the levels properly and not adjust after the fact. This would require two passes, though.

It's up to you. There are probably some normalization algos that are better than other.

 

My audacity has a default recording level. I've tried to adjust it, but it never changes anything.

 

Define "necessary"

 

There's nothing wrong with normalizing in 24 or 32-bit. You won't hurt a thing.

Sometimes I will even use a bit of limiting ala The Beatles stereo remasters to bring up the overall level of a needledrop if I determine that the sound will not be adversely affected.

I do not normalize CD rips.

 

You realize that the recordings have already been limited, which means that if you apply limiting after that, you've now double-limited the recordings.

 

For my purposes, I look at the recording on vinyl as the master. There may be some limiting applied at it's mastering stage, but it isn't an issue for me.

 

I just ampllify at the editing stage - I don't use normalization (they are similar, but not the same -- the latter might amplify per channel if you check that off in Audacity. Plus it has a DC offset option).

My GT40 records fairly low which I actually prefer as I need that for the ClickRepair stage anyway. Avoids any possible clipping.

 

Questions:

- What is the DC offset? I leave it checked by default but would like to know what it is!

- Also, I understand tweaking the sound in a high bit sample rate, but which is better.... normalization or amplification? And to what extent or degree?

All your experience is helpful to me! Thanks in advance!

 

DC Offset is when you're looking at the waveform and it's not exactly centered - my understanding is that this could happen with certain sound cards or ADC's. I don't think it hurts to leave it as is if that happens, but so far I haven't had this issue.

The other checkbox in Audacity is the "Normalize stereo channels independently", which I would never use for a drop. That means that if you have a waveform with one channel louder than the other, it will try to compensate by raising up the other side and I could see where that could really mess up certain recordings.

The Amplify option, from an Audacity perspective, is Normalize without the DC Offet or Normalize stereo channels indepently options, or so their wiki states. So I just use Amplify for my drops (-.4db max per side).

 

Yes, IMO. If you're recording at 24 bit (and you should be), You won't lose anything at all in the normalization process. It can only be positive. There's no reason to not do it; and it can increase your sound quality substantially if your end result is a 16 bit CD.

 

I have some vinyl pressings where one channel is noticeably louder than the other. This feature, normalizing each channel individually, helps with this. As a rule, I agree, and I -very- rarely use this feature in Sound Forge.

 

I've recently upgraded my computer memory which allows me to rip albums in 32-floating bit/96kHz. I use the Live Ubuntu Studio disc, so the memory upgrade doubles my recording time from 20 minutes to 41 minutes. After I record and then trim, I reset everything and record again.
My end result is a 16/44.1 vorbis file for my Cowon S9 digital player.
I keep the 24 or 32-floating bit as an archived file.

What exactly is a 32-floating bit? It's the highest bit setting in Audacity, but like the DC offset I have no idea what that is.

 

This is a stretch for me, but I believe it allows for slightly better headroom during the editing process. So I'll save my initial recordings as 24-bit WAV, but open them up for editing (where I'll amplify and chop up tracks and do some final pop removal) in 32-bit float. Then save to 24-bit FLAC.

I honestly don't know if saving to 32-bit WAV would give me any benefit, but since my ADC isn't 32-bit anyway, I keep the saved files as 24-bit. I've also wondered if perhaps I should work in 24-bit across the board, and hoping that someone more experienced than I could chime in on that as well.

 

It's still not correct. I don't have a problem with changing levels in order to match a group of songs so that the relative volume is consistent between albums or within a playlist, but I think use of limiting or compression is changing the original intentions of the mastering engineer.

The exception might be one peak excursion that rises up about 5dB above everything else, or some other flaw that might prevent me from raising the overall volume without clipping. In that case, I might choose to drop that one peak, if it's a minor problem. But not limiting and not compression.

 

Well, as I thought I pointed out, that is the main reason I use limiting.

It goes against the ethics of this forum, but I see nothing wrong with using a bit of limiting to get the files within the ballpark of a reasonable level of RMS loudness.

 

An example rip that I would like some advice on would be a 32-bit floating/96kHz rip of Genesis "Duke". One of my favorite all time Genesis records.
Because the cutting engineer had to fit almost 28 minutes per LP side, there was a great deal of compromise on the sound quality, or volume, IMHO.
I'd like to tweak the "loudless' without killing the resulting SQ.
What advice would you offer?

FWIW, the recording level on Ubuntu Studio's live disc is set by default, so trying to fiddle with that option is off the table.

 

Someone correct me if I'm wrong, but I believe while a file is open in Audacity, it's actually in 32-bit floating while you work with it . . . so normalizing even a file recorded at 16-bit shouldn't affect it adversely.

 

Hi Col Pepper,

There are many variables to consider with regard to gain adjustments. (Personally, I prefer a more manual approach and use the Gain Change function in the software as opposed to a Normalization --I call it abnormalization-- algorithm.)

First, you want the original rip to be at a long word length (24-bits is good, 16-bits isn't), even if the target is a 16-bit file.

Second, you might want to experiment with different software applications. In my experience, even something as simple as a minor level change will produce different *sounding* results with different applications. Some can do it more transparently than others. (Not all applications that process 24-bit files can do so cleanly, even some that process internally at 32-bits.)

If the target is 16-bits, the processed file needs to be dithered prior to word length reduction (i.e., before going from 24-bits to 16-bits). Different dither algorithms *sound* different. Most, in my experience, will cloud the soundstage and alter instrumental timbre. The very best (I'm partial to iZotope's MBIT+) seem to do their jobs and get out of the way.

As to the maximum peak level, you are correct to stay away from 0 dB (full scale). Generally, I find a max peak of -0.3 sufficient for most material, though there are some exceptions that "want" more headroom.

In the end, if the gain change amounts to only a couple of dB, you have to decide (based on the transparency or lack thereof in the software used to effect the gain change and in the software used to dither if such is used) whether the net gain (sorry for the pun) is positive or not.

Hope this helps.
Have fun!

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

How loud the relative channels seem to be is not my determining factor in choosing levels. I generally go for centering of the center channel information.

Many records are not balanced in this regard. It seems like some people set levels based on meters rather than listening to the mix. It's also true that not all recordings have true center channel information, but most pop recordings do.

This is why most compressors have links so that when used in stereo and one channel ducks the other channel goes with it. If not, the center imagae would shift it's location when one side or the other ducks.

IMO, stability and preservation of the stereo image is more important than independent gain structure optimization of the individual channels of a stereo mix.

 

Best to just stay at 24 bits if at all possible then, and try to adjust the gain as accurately as possible while ripping. Then you mostly eliminate the need for gain adjustment or dithering and any possible degradation or change in character those might introduce.

 

Hi autodidact,

At first glance, this would appear to make a great deal of sense: play the record and when it's done, check the peak value on the input meters of the A-D (or its software) -- assuming it is the type to hold the peaks for that long, then adjust the gain accordingly for the real recording.

The problem that arises, in my view, is that every monolithic A-D chip of which I'm aware (in English: every A-D converter I know of) will display less distortion with a max peak of -6 than it will with something approaching what would be the final level.

This is why I record with lots of headroom. I'd rather have the initial digital recording peak at -12 than at -2. Final levels are adjusted digitally during mastering.

In other words, recording for a max peak no higher than -6 (or less; at 24-bits, lower levels don't matter), then adjusting for final levels later, in the digital domain, will result in a cleaner finished product. That's been my experience, anyway. For those who are interested, I suggest trying it both ways and comparing the final results.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

Hi bdiament,

What if the loudest peak is a pop at the beginning of a recording, like when the needle hits the record, and it rises above -6 dbfs; and then the rest of the recording is well below -6? Does the initial peak affect the sound recorded that comes subsequently?

Thanks.

 

If you're throwing away 6dB at the top, how many bits is that you're wasting?

I'm not saying you're wrong at all. I'm just saying, theoretically, if there was no distortion, and you could record all the way to 0dB, how many bits would you be throwing away by dropping the max level to -6dB?

 

Hi eyeCalypso,

The sound of the stylus landing in the lead-in groove does not count.
It will not impact the rest of the recording.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com