Your Vinyl Transfer Workflow (sharing best needledrop practices)*

Oh ya, all editing is done on 32bit float in audacity and clickrepair, then dithered to 16bit wav when done.

 

Why would you want to make each track on a given record (or at least a given side) the same level when an audio engineer has already determined what the difference in loudness should be between these individual tracks? If you want them louder than what you have captured them at fine...but make your adjustments for the album as a whole (or by each side)...not by each song.

My normalization process depends on the pressing itself. If I have an LP that was mastered by the same person/facility on both sides (many LPs are NOT like this), then my process is:

1. Run each side thru ClickRepair using my minimalist settings.
2. I then view each side in Audacity to make sure there are no LARGE clicks that were missed in step one. If so, I will open the side affected in Izotope and manually fix them.
3. I will then combine Side A and B into one single file using the Merge/Cue function in Xrecode II.
4. I will then amplify the signal in Audacity to my preferred setting.
5. Bring the file back into Xrecode II and split back into Side A and B
6. Use Audacity to split the tracks on Side A and B into their individual tracks and then finish the clean-up in Izotope

If the LP was not mastered by the same person/facility, then I will often amplify each side INDIVIDUALLY to my preferred setting. This is about the best way to get the rip to sound "right" in my opinion.

 

Out of curiosity...has Audacity ever gotten around the Windows kernel issue that causes 24/96 rips made in it to be "padded" with junk making it look like a true 24/96 capture, when it really is not? I stopped using it long ago for recording...just curious if this issue was ever solved for the masses who use it.

 

I think you're giving too much credit to all audio engineers?
Surely some didn't think too much into it?

As I said - Parsons, Floyd, Steely Dan... all benchmarks, and normalization should NOT be done for each track. Agreed.
But, for 60s garage, or country?
I just don't think those audio engineers gave it much thought, so I will normalize each track, and feel comfortable knowing I probably didn't disrespect the "art".

Sometimes tracks are simply quieter because they're on a side containing far more music than the other side.
For example, Public Enemy - Fear Of A Black Planet.
Some of those tracks are very quiet on the LP, but are super loud on their 12" single releases.
Which one is truer to the engineers original intent?

 

I don't think you're understanding it. It doesn't matter who the artist is. They all use mastering engineers after the fact to level the tracks before the cutting process on every project. I also don't know what kind of advantage you think you are getting by normalizing each track individually.

There is an art to mastering levels so that they all flow evenly within the album's form. If you have a loud rock song, then a soft ballad, as, say, on a Heart album from the 70s, you do not want that soft ballad the same volume as the lout rock song. My method is to just ensure that the vocals are consistently the same throughout. Now, again, if you are needledropping an album, this should pretty much be done already. But, if you create a compilation, this is a great way to do it. And, the batch normalizer in Adobe Audition CS is excellent, and I use it.

There are cases where you get an album, especially from the 80s where the first track, or part of the first track are louder than the rest. This was done with the idea of attracting the attention of a radio programmer, with the idea that the first song, or the first minute of a song should be loud. This screws things up considerably. So, in those cases, I go in and fix the volume. But, I keep the first track in relation to the rest.

The other side of it is the last track is sometimes lower. I found that some of it depends on the type and quality of the car you use. It also depends on if EQ or compression was added. I don't try to expand the dynamic range, but I do adjust the volume if needed so that it fits with the rest of the album. Usually, nothing needed to be done. If you watch the average levels of you album side, you will find that the volume is at an average level despite the compression.

It sounds like you are trying to make your needledrops the same loudness as a typical modern CD that suffers from the "loudness wars". Why? When you do that, you compromise the sound of the vinyl. However, again, you do you. It's your project.

Dead means no life. It means no openness in the sound. The natural reverberant decay is gone. To understand this, go into your bathroom and yell. Hear that echo? Now, go into a room with a lot of carpeting, sofas, and pillows if you have them, or your car if you have cloth or felt seats. Now, yell again. There will not be any echo. Well, the overuse of NR is analogous to the same thing. The same thing happens with the bass and high treble. You can still use NR - I do, but you should take care to avoid all of this. Do you recall the criticisms of Jon Astley who remastered The Who and ABBA CDs in the early 2000's? He was known for overusing NR so that the sound resulted in those dead qualities I described. He isn't alone. Several reissue engineers, especially one who used to work on a reissue label's comps, were notorious for it.

It's meaningless. Each album side or song demands its own setting. There is no one side fits all. I have not used the NR in Audacity, but I can tell you that you get what you pay for, and Audacity is free. Right? But, you are doing the correct thing by grabbing a noise sample first, and using that to reduce noise. And, always remember: don't try to remove all of the noise. All you want to do is reduce it, hence the name "noise reduction".

That is how some purists feel, but it all depends. If you are blessed with a very quiet turntable/cart/preamp, and quiet vinyl, maybe you don't need it. If you drop old records that are less than pristine, yeah, you may need it. It's your choice. I use it.

 

If you don't use mp3, don't worry about it.

BTW, about noise-shaping: back in the 90s and early 2000s, it was all the rage, and some sound quite good. But, my personal opinion is that they alter the sound, and even many engineers no longer use or recommend it. Just use a good straight dither, and you'll be fine.

 

Yes, I generally give credit to the engineer...at least in knowing how loud or soft a song is supposed to be in relation to the other tracks on an album.

If you like what you are doing and you can sleep at night, then go for it. You posted in here asking us to review your methods and provide feedback on it - which we have. All I can say is I have been ripping vinyl for a very long time and learned what I know from some of the most well respected and highly thought of rippers out there, and the method for normalizing rips is pretty much the same across the board. If you want to be different, go for it - they are YOUR rips.

Apples and oranges. That's like asking if an Allen Zentz mastering of an specific album is more right than one by Robert Ludwig - they are going to sound different and neither one is necessarily wrong or better than the other.

 

CDs and loudness wars result in brickwalled waveforms, no dynamic range, right?
My waveforms are not anywhere close to brickwalled.
99% of the original dynamic range remains intact with my recordings (per track if norm'd separately)

So yeah, I'm just trying to make these as loud as possible so I can listen to them in my '08 Ford Escape while keeping most of the integrity of the original recording intact.
They are not anywhere near as loud as CDs.

I don't think I'm limiting and normalizing as much as I made it sound.
I'm careful.
And if I think it's too much, I won't do it.

Regarding normalizing each track -
OK I see now what you all are saying a little better.

But answer this:
Say you've got a 20min Side 1, and a 30min Side 2.
Do you record both sides as one file? Or each side a separate file?
Normalizing the whole LP as one file will result in a quieter Side 2,
but normalizing each Side individually will result in a louder Side 2.
Which one do you do?

Thanks for the feedback on everything.
Time to start all over LOL.

 

See my previous post regarding this...I record Side A and then I record Side B - two separate files. If they are mastered by the same person, it all gets normalized as one file, then split. If the two sides are from different engineers, I generally normalize each side individually then split the sides into their individual tracks.

In this second scenario, I'll know if there are any massive differences just by looking at the signal readout on the laptop - often times one side is cut louder than the other. If they were normalized together it would not sound cohesive when played back - it would sound like two different people mastered it - which is what actually happened! LOL

 

OK this is interesting.
If you are making a comp,
all norm'd tracks with respect to their original LP sides are now no longer appropriate.
If kept the same, your comp will be all over the place volume wise.
So YOU now become the engineer?
Is that what this batch normalizer does?

 

One thing I forgot to mention concerns the definition of "normalizing" a recording. When someone speaks of this here, we all know what they mean, but different software programs have their own definition of the meaning. This is especially true with Audacity, and I would encourage you to read the entry about it in the manual if you have not yet done so:

Amplify and Normalize - Audacity Development Manual

To highlight:

When not to use Amplify
If your audio has DC offset present, you should use Normalize to correct that offset.

If you are applying a Chain and want to bring a track or files to a specific level, use Normalize instead. Amplify in a Chain will not make the track or file as loud as possible if "New Peak Amplitude (dB)" is set to 0 dB and will not set the amplitude to any other chosen amount in that box. Because Amplify is being applied automatically it can only apply the "Amplification (dB)" amount as saved in the "Ratio" parameter.

When not to use Normalize
If multiple tracks contain intentional differences in peak levels, you should never normalize any of them. The only way to preserve the relative balance between them is to select them all as a group and amplify them, so they are all scaled up by the same amount.

You *can* normalize stereo tracks without changing their channel balance, but only if a) the stereo track is not split into Left and Right and b) "Normalize stereo channels independently" is unchecked.

If tracks with intentional peak differences have DC offset that needs correction, then you can and should use Normalize, but with "Normalize maximum amplitude to" unchecked.

 

This doesn't seem consistent with this from Sample workflow for LP digitization - Audacity Development Manual :

Normalize the amplitude of the recording; either do each track of the recording individually (especially if the tracks will be randomly played from a library containing many different styles of music) or, do the whole recording at once (which will work fine if all the tracks have the same average volume). Use Effect > Normalize as the last editing step, setting "Normalize maximum amplitude to" to around -3 dB or similar, to give some headroom below the distortion level. The Normalize effect can be set to either:

• Adjust the amplitude of both stereo channels by the same amount (thus preserving the original stereo balance), or
• Adjust each stereo channel independently (this can be useful if your equipment is not balanced).

The info you posted seems to be the best choice, though.

 

Where I differ from psulioninks is that I record both sides as one big file. I only chop them up after I do all of my processing. If one side turns out to be louder, I can simply highlight that side with the cursor and process accordingly. I can still process and make adjustments after chopping up the files, but before dithering to 16-bit. Again, dithering should be the very, very last step in the process before tagging and archiving.

Again, if one side turns out to be louder, you can simply highlight that side with the cursor and process accordingly.

 

No they don't. I use Audition, Sound Forge, Reaper, iZotope, and have Audacity, and they all use the terms and methods correctly. I even used DCArt and Waves, and they all use the terminology correctly.

Amplify and Normalize - Audacity Development Manual

This is not 100% correct. You do not have to normalize when correcting the DC offset, should there be one.

With most modern sound cards, this shouldn't be an issue. But, it still shows up. It is always a good idea to remove the DC offset before any processing. I also like to correct any phase issues, which occur a LOT, even on commercially sold CDs and albums. Sometimes the phase problems are inherent in the recording itself. It's just something that happens in nature.

 

Thanks Grant.

So, one more favor.
Here is a revised workflow for me.
Can you review and point out anything you would change?

1200 to ART Phono Plus USB preamp to Laptop.
All editing done in 32bit float.
Low Cut Filter is ON on the ART preamp (simply because it has the option, and I read somewhere it's a good idea).
1.Record both sides into Audacity with peaks roughly around -6dbs each side.
2.Send each file (side) thru ClickRepair:
-pitch protect checked
-reverse checked
-wavelet (does this matter? I thought I read wavelet was a good choice)
-declick will vary (clean records, so typically around 5, maybe 10 max)
-full auto (I trust the software to do a good job)
3.Send 1st side file back into Audacity (still 32bit float)
4.Click remove only glaring spikes that escaped ClickRepair.
5. Normalize ENTIRE FILE to -.3, DC offset checked, 'each channel independently' unchecked.
6. Soft limit very conservatively, preserving orig dynamic range as best as possible.
7.Normalize again to -.3
8.Split and label tracks.
9.Export tracks to PC as 16bit wave files.
10.Repeat for Side 2.

Does that look alright now?
-Normalizing is now for whole side, not separate tracks.
-No more NR.

Thanks for input.

 

Is there any reason you can't just do both sides at once?

The first thing after recording is to remove the DC offset if there is one before you do anything else. You really should do it before you normalize anything.

You can use NR if you want/need to. But, you do have purists who will tell you never to do it. But, they do not know your circumstances. Many avoid NR because they just don't know how to use it, or don't have the skill or patience to do it right. They may not have good NR software. I use NR, and have 20 years of experience with using all sorts of NR.

 

Decoding refers to MP3 decoding. Grant said "if you ever plan on making mp3s, leave 0.02dB of headroom". Which is no headroom at all. -6dB is a better setting to ensure encoded mp3 will never clip on playback, and to ensure that you won't cause other distortions (if you don't understand the intricacies of the described digital stuff.)

Here's some background.

Digital 0dB samples can actually represent a higher analog level than 0dB:



Look at the dots in the waveform above - those are the actual samples. However, the thin line that looks wavy is the analog waveform your DAC will be asked to put out. That level goes higher than the samples.

You might think that the depiction above, -4dB samples that create a -3dB waveform, might be a worst case scenario, but it's not. In fact, the waveform shown above is a PWM square wave (as a synthesizer might generate), after a simple "rumble filter" and then downsampling was applied. The original waveform was -10dB! (as shown below):



If the original had first been normalized, you'd have a whole bunch of clipping.

If you "normalize", which is increasing the level of samples to the maximum that can be represented digitally, that leaves no room for any further processing. Even resampling may clip the waveform. Normalizing to 0dB based on sample values may require a DAC to put out up to +6dB analog peaks in worst cases.

So: leave several dB, unless you are a digital expert and have a reason for making the file as loud as possible and possibly distorted.

--

Dither is applied when you downsample, for example, from 24 bit to 16 bit. This randomizes not the top bit, but the bottom bit, to make the quantization noise consistent and less "digital". You really don't need to dither a recording of an LP when downsampling though - it has several bits of its own noise above the 16 bit level, which will act like dither noise. Noise-shaped dither is never a bad thing though.

--

Best practice would be to record both album sides, and then normalize the whole album file if you need to adjust the digital volume. However, with the talk of "mastering" above, I'll remind you of one other trick of LP mastering: putting the quietest tracks at the end of the album, or even decreasing the volume as the side progresses:



Look above - that album is getting quieter with each song! I'll bet the CD doesn't do that!

If you see a similar volume envelope on an LP needledrop (or even within a 45rpm single), you may want to increase the volume slowly per-side with a "fade" to get the songs back to comparable levels.

 

I've read some well-known engineers advising .01db.

As for album sides, the reason the wave you show above looks like it does is because the last sone may not be lower in volume. As I stated earlier, many cutting engineers used compression to reduce the dynamics on the inner grooved to reduce distortion. If you play each track, you may find a slight volume reduction, but, most of it is maintained. If you play the tracks while looking at a VU meter, or a digital meter, observe where the needle or bars average. If they average in the same place, there is really no volume decrease. It just looks like it. That's the RMS, or Root Mean Square. Again, what I like to do if there is a volume decrease, or if the first song is too loud, I adjust it by taking note of the RMS. Then, I listen, use my ears. I adjust each of the songs by either the vocal, or some other prominent element in the song, something that should be consistent throughout. It gets a bit tricky if processing has been applied to the vocal. It also gets tricky if there is a lot of bass content.

Speaking of bass content, bass will rob you of volume. What some cutting engineers, and even CD mastering people will do, is slash the bass, sometimes as high as 40db! Then, they raise the RMS and use the limiter to make it louder. Eber notice why a lot of 80s rock albums don't have a lot of bass? That's to get them louder! There are other little tricks they use with EQ'ing the midrange to boost the volume. You can cut the peaks by a certain amount and raise the volume of the file to get more loudness without sacrificing too much dynamism and not altering the microdynamics. This is what the OP is kind of doing.

You know, there really is no real wrong way of doing any of this, but there are better ways, and not-so good ways.

Dither reduces or masks quantization noise that would otherwise be produced by reducing the bit-depth to 16-bit. It does this by adding random noise. One can also test this by creating two 16-bit files from the same 24+ bit source, then dithering one, and not the other. The un-dithered version will sound grainy and harsh, and the natural tails and reverberation will sound like hash.

I gotta get some sleep, but, more about dither. The ideal is to keep an original 24/32-bit master in case you ever have to go back in and do something to your file(s). But, if you can't, and you need to do more than just a simple edit or two, it's best to convert the file back to 24/32-bit, do your work, and dither it again to 16-bit. That way, you won't double-dither by working on a 16-bit file. No, you can't just work on your 16-bit file and not use dither. Once you initially dither, that dither effectively becomes part of the audio. It's better to have a very slight veil over the sound than to have that digital distortion. It's best to go back to that 24/32-bit file and keep it fresh.

If you do decide you like the sound of a flavor of noise shaping and use it, plan on never working on that file again. As for the sound of noise shaping, I used various types for years, but stopped when I did some tests and realized that the sound of using flat dither did the least damage to the sound. With noise shaping, there was always a compromise somewhere.

24-bit is considered overkill, but gone are the days when software and hardware would allow you to use 20 and 22-bit. I remember the first time I heard of 20-bit was when the first Journey remasters came out. The first Rolling Stone remasters, and the Elton John remasters used 22-bit.

But, I use 32-bit floating -point because I can exceed 0db fill scale when processing without clipping! Of course, it all has to be pushed back under 0db for the final product.

I'm rambling, so, see ya tomorrow...oops! It is tomorrow!

 

Just wanted to add... I do normalize each song individually, but I do keep in mind the feel of each track in the flow of the album.
I don't like normalizing to a fixed level (e.g. -.3) because you might get a spike that will ultimately 'stifle' the volume of the whole album.
I normalize with Adobe Audition using the manual dial until the wav form looks to be in my sweet spot (which is always different) but once I do the first song, I keep the increase in mind and use it as a general guide for the remaining tracks... especially if there are ballads or quieter tracks.
The most liberating thing for me when it came to enjoying my needledrops was when I came to the realization that these are for me and my ears only... if I like them, then they are awesome. There is no tribunal I have to present them to for approval and to continue enjoying this hobby.

 

What I do if I get a rogue spike is I zero in on it in Audition and do one of two things: I either simple take down the volume of that spike, or compress it, whichever sounds better. Then, I use unity gain to boost the overall volume of the album. That way, I don't wind up compressing the entire file.

 

I will avoid NR for now, since I only have Audacity and I don't fully understand it's results.
LPs are generally all pretty quiet on my TT. There may occasionally be some noise or hiss between tracks, but you really gotta listen to notice.

Yes, DC offset gets done first. It's part of Audacity's normalzing feature.

I guess I do one side at a time because I like looking at a shorter duration waveform. I can see any irregularities, rogue spikes, etc much better with a shorter duration waveform. Probably just need more experience.

 

That's what the zoom in/out feature is for.

 

You're knowledgeable enough to ponder an example. Take the top waveform in my post, and normalize it so the samples are at 0.01dB (but the actual signal is above 0dB).

Oversampling DACs can reconstruct the signal waveform between the samples just as accurately as Audition can show us what it would look like. With 8x oversampling in a good ol' CD player, the samples that come out of the oversampling stage are clipped, unless the DAC also has added more bits for this eventuality.

DACs also apply filters which can clip the digital. In original DAC chips, you could set the sample value at 1 volt and the chip would put out 1 volt forever. Now they optionally filter DC digitally, as well as doing the aliasing filtering digitally. As we saw in the second waveform picture, filtering can push up the levels too. Whether this may cause distortion depends on the digital headroom in the architecture and also the linear headroom of the analog output I/V and buffer stages.

Noise shaped dither (which would more accurately be called frequency-shaped-noise dither) can move the noise that would be overlaid over the program material to much higher frequencies where it is less detectable to the ear. Where the bulk of audio is located between 40Hz-10kHz, this can push the A-weighted noise floor down to -110dB or more (on a format commonly said to have 96dB of dynamic range).

It is rather normal dithering that adds signal-altering noise. Once you have put 96dB of noise over the entire passband, it can cannot be removed or re-shaped.

Of course we can see that using multiple dithers can be counter-productive if one isn't attentive. If I use noise-shaped dither to re-downsample a previously unshaped-dithered file, then I am adding high frequency noise as well as keeping the previous broadband noise.

 

True, guess I'm lazy haha.
Thanks again. I'll move forward for now.
I think I'll start saving all raw recordings, I can always return to them

 

Ditto... this technique was a real "ah ha" moment in doing my needledrops. I first used it when needle dropping the Tom Jones 'It's Not Unusual' album and when I did an across the board normalization, the hand claps in the title track were causing the levels of all the other songs to be way too low for my liking... it was tedious but I took down each handclap to get that song in a nice consistent 'sweet spot' with the rest of the album.