• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Normalization methods for LP digitization

Lttlwing16

Active Member
Forum Donor
Joined
Feb 24, 2021
Messages
201
Likes
113
Hey folks,

I've been exploring digitizing my LP's so I can play them on my upstairs stereo where a turntable setup isn't an option. I wanted to open a discussion about methods to properly bring in the analog data and have it at a reasonable playback volume. In particular normalization methods.

From the research I've done so far, while most agree a high bit depth at capture is best, it seems people are split on capture method otherwise. Here are some of the methods I've seen proposed/used:

  • capture at a lower peak, around -6.0 dBFS, then use peak normalization to bring each side of the LP up to around -1.0 to -0.5 dBFS.
  • capture at as close to -1.0 to -0.5 dbFS, no normalization
  • capture at a lower peak (-6.0 dBFS), no normalization, adjust output volume with playback device (volume control)
  • capture at a low peak then use REPLAYGAIN tag to adjust volume

The lower capture volume seems important to avoid clipping/distortion and remove any added THD+N from the soundcard itself, although the noise floor of the LP is most likely much higher than that of the soundcard.

The problem with normalization, and peak normalization in particular, is the presence of artificial peaks from record noise (i.e. clicks and pops). While one can meticulously groom the waveforms to remove them, some are still liable to be present and can then affect the normalization process. Additionally, this adds an additional layer of processing to the digital capture. I assume this is why some don't care for it.

However, I came across this article which explains the three different methods for normalization: peak, RMS, and EBU-R128. The latter two methods were interesting, and I had come across RMS normalization in Pro Tools, but never understood what it did.

My question is then, why isn't anyone discussing normalizing to an appropriate LUFS, instead of the peak dBFS? For instance, I own a digital copy of Khruangbin's Mordechai as well as the LP. I could import the FLAC files and check the LUFS level on each track, then using the Loudness Normalization tool in Audacity (which utilizes the EBU-R128 volume detection method) to bring the LP rip to the same perceived LUFS?
 

daftcombo

Major Contributor
Forum Donor
Joined
Feb 5, 2019
Messages
3,687
Likes
4,068
I think the best practice is: "capture at as close to -1.0 to -0.5 dBFS, no normalization", adjusting the volume analogically at the output of the preamp or at the input of the ADC, or both.

But that method theorically requires to adjust gain for each record, so to play the record completely first (all sides) to be sure you will never record higher than -1.0 to -0.5 dBFS.

Personally, I couldn't be bothered, so I took one of the loudest records in my collection and adjusted the gain input of my ADC to have peaks around -3dB once and for all.

Now, after recording an LP/EP, I normalize the peak to -1dBFS in iZotope RX (for the whole record, not track by track). If you record in 24 bit, you have plenty of headroom anyway.

Edit: I forgot to mention that I manually remove loud pops/clicks before normalizing, which is very important as they often are louder than the loudest peak of a record. It can be long to do, but you don't want to listen to clicks during playback, do you?
 
Last edited:

DVDdoug

Major Contributor
Joined
May 27, 2021
Messages
2,916
Likes
3,831
capture at as close to -1.0 to -0.5 dbFS, no normalization
There's no need to do that. Generally, the noise goes down when the volume goes down so this doesn't hurt the signal-to-noise ratio.

-6dB peaks are fine but digital recording levels are not that critical and nothing bad happens when you get close to 0dB, only if you "try" to go over.

If you remember analog tape, you needed a hot signal to overcome tape noise. Also tape is more forgiving when you go over 0dB as it starts to soft-clip. The tape equalization further "softens" any clipping. With digital there is no tape noise, but it hard-clips at exactly 0dB.

The problem with normalization, and peak normalization in particular, is the presence of artificial peaks from record noise (i.e. clicks and pops). While one can meticulously groom the waveforms to remove them, some are still liable to be present and can then affect the normalization process. Additionally, this adds an additional layer of processing to the digital capture. I assume this is why some don't care for it.
I'd recommend removing/repairing and clicks & pops that are that terrible.

However, I came across this article which explains the three different methods for normalization: peak, RMS, and EBU-R128. The latter two methods were interesting, and I had come across RMS normalization in Pro Tools, but never understood what it did.
Personally, I peak-normalize the album as a whole (after making on-big file with both sides, which I'll later split). That maintains the original relative-volume between tracks "as intended". Some people normalize to -1dB or so to leave room for "intersample peaks" or because MP3 compression can make some peaks higher and some lower.

RMS is a kind-of "average" so it's an indication of "loudness" but it's just "simple" calculation and it doesn't take frequency content into account like LUFS or ReplayGain.

My question is then, why isn't anyone discussing normalizing to an appropriate LUFS, instead of the peak dBFS?
That would be an alternative to ReplayGain. But you have to watch-out for clipping because some of your files may clip when adjusted to your target loudness. (ReplayGain has an option of avoiding clipping.)

Records often have a higher crest factor (peak-to-average ratio) so they may not go as loud as your modern digital recordings without clipping. There are two reasons for that - With older records, they didn't have digital compression & limiting so they couldn't go as loud in the "loudness war" so many older records are more dynamic.

Also the process of cutting & playing-back the record tends to increase the crest factor (without changing the sound of the dynamics). As I hinted above, MP3 can also increase the crest factor.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
When I captured my collection of about 1,000 LPs, here's how I did it:
Cleaned the record with a Nitty Gritty record cleaner (makes a huge difference).
Ensured the tonearm & cartridge was properly aligned and head amp load settings were correct.
Played and recorded tones on a test LP to ensure frequency response & channel balance.
Connected the output of my phono head amp directly into a digital recorder (Tascam SS-R1 or Juli@ sound card).
Visually inspected the LP to locate a loud section, played it, set digital level at -6 dB.
Played the entire LP, recording it at that level.
Recorded at 44-16 for most records, at 48-24 for 200 gram, 1/2 speed masters.
When complete, load the WAV files into Audacity, level shift peaks to -1 dB, with high quality triangle dither.
Break into tracks, save the files, burn to CD or DVD-A if desired.

In short: I set the level for each record individually, used no dynamic compression, normalization or replay gain. 16-bit digital has more dynamic range than LP, so as long as each individual record peaks near full scale or -1 dB, there should be no reason to use them. The digital recordings are transparent, sound just like the LPs did.
 
Last edited:
OP
Lttlwing16

Lttlwing16

Active Member
Forum Donor
Joined
Feb 24, 2021
Messages
201
Likes
113
That would be an alternative to ReplayGain. But you have to watch-out for clipping because some of your files may clip when adjusted to your target loudness. (ReplayGain has an option of avoiding clipping.)
See I've read a couple places online ReplayGain is much like a blind normalization. Used primarily for creating equal volumes across different albums in a collection. In some instances (barring the user didn't have or know about "avoid clipping") Replaygain introduced distortion by increasing the volume too much. Note, this could be old information I picked up, and ReplayGain could've matured since.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
I never saw the need for normalizing the volume of LPs. There is a huge difference in the loudness of different LPs, no doubt. Some are cut very loud, others very quiet. But this was a difference in absolute level. After recording, when I shifted each LP's peaks to -1 dB digital, they were all fine. No volume issues. No need for compression or replay gain.

When you shift each individual LP's peaks to -1 dB digital, you are essentially doing what replay gain does, "normalizing" them in the recording itself, which obviates the need for it on playback. Streaming services use replay gain because they don't get to modify the bits of the recordings they stream.
 
OP
Lttlwing16

Lttlwing16

Active Member
Forum Donor
Joined
Feb 24, 2021
Messages
201
Likes
113
I never saw the need for normalizing the volume of LPs. There is a huge difference in the loudness of different LPs, no doubt. Some are cut very loud, others very quiet. But this was a difference in absolute level. After recording, when I shifted each LP's peaks to -1 dB digital, they were all fine. No volume issues. No need for compression or replay gain.

When you shift each individual LP's peaks to -1 dB digital, you are essentially doing what replay gain does, "normalizing" them in the recording itself, which obviates the need for it on playback. Streaming services use replay gain because they don't get to modify the bits of the recordings they stream.
Would you care to elaborate a bit more on the process of "shifting the peaks to -1 dB". You had mentioned using a dither, but I was unaware you could set a dB level with a dither.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
It's simple. After recording the LP, the peak levels will be around -6 dB, because that's how you set them. Load the WAV/FLAC file you recorded into Audacity or some other audio editor, then shift the digital amplitude so the peaks are at -1 dB. Do this for the entire album, as a single setting applied to all the tracks. This is like applying a digital volume control to the recording itself. The proper way to shift the digital levels involves randomizing the LSB of each sample to avoid quantization distortion. That is dither, and it should be applied automatically by the software when you do this.

You might wonder, why not record it so the peak levels are at -1 dB and avoid this post processing step entirely? You could do that, but -1 dB goes you very little margin for error in case you didn't find the loudest part of the LP. So it could clip and overload. Setting it at -6 dB gives you a lot more headroom to avoid that, then you shift it to -1 dB in post processing after the entire album is recorded so you know you have already captured the loudest spot.
 
OP
Lttlwing16

Lttlwing16

Active Member
Forum Donor
Joined
Feb 24, 2021
Messages
201
Likes
113
It's simple. After recording the LP, the peak levels will be around -6 dB, because that's how you set them. Load the WAV/FLAC file you recorded into Audacity or some other audio editor, then shift the digital amplitude so the peaks are at -1 dB. Do this for the entire album, as a single setting applied to all the tracks. This is like applying a digital volume control to the recording itself. The proper way to shift the digital levels involves randomizing the LSB of each sample to avoid quantization distortion. That is dither, and it should be applied automatically by the software when you do this.

You might wonder, why not record it so the peak levels are at -1 dB and avoid this post processing step entirely? You could do that, but -1 dB goes you very little margin for error in case you didn't find the loudest part of the LP. So it could clip and overload. Setting it at -6 dB gives you a lot more headroom to avoid that, then you shift it to -1 dB in post processing after the entire album is recorded so you know you have already captured the loudest spot.
Okay gotcha.. adjust the track level with the track gain adjustment so peaks are at -1db. This applies the specified dither in EDIT>Preferences to apply the volume difference.

You're sure these changes are kept when choosing to export selected audio (individual tracks) as FLAC?
 

snaimpally

Member
Joined
Sep 23, 2021
Messages
68
Likes
73
Location
Austin, TX
When I captured my collection of about 1,000 LPs, here's how I did it:
Cleaned the record with a Nitty Gritty record cleaner (makes a huge difference).
Ensured the tonearm & cartridge was properly aligned and head amp load settings were correct.
Played and recorded tones on a test LP to ensure frequency response & channel balance.
Connected the output of my phono head amp directly into a digital recorder (Tascam SS-R1 or Juli@ sound card).
Visually inspected the LP to locate a loud section, played it, set digital level at -6 dB.
Played the entire LP, recording it at that level.
Recorded at 44-16 for most records, at 48-24 for 200 gram, 1/2 speed masters.
When complete, load the WAV files into Audacity, level shift peaks to -1 dB, with high quality triangle dither

The above is all excellent advice. I would also suggest recording at 44.1khz with 24 bits (rather than 16). I found that 24 bits instead of 16 makes an audible difference. Having more bits is also more forgiving if you set the recording level a little low. When burning to CD, most recording software will use 16 of the 24 bits but if you can playback all 24 bts from the computer etc., it sounds great.

Also, I wouldn't reccomend using 48khz to record as 48khz is used mostly for the audio portion of videos.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
Okay gotcha.. adjust the track level with the track gain adjustment so peaks are at -1db. This applies the specified dither in EDIT>Preferences to apply the volume difference.

You're sure these changes are kept when choosing to export selected audio (individual tracks) as FLAC?
In Audacity, the setting is "Effect | Amplify". When you apply it, it uses whatever dither you selected in settings. After applying it, you will see the waveform get bigger or smaller, depending on whether you used a + or - value. If you export to FLAC (or WAV) after doing this, it saves the modified version.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
The above is all excellent advice. I would also suggest recording at 44.1khz with 24 bits (rather than 16). I found that 24 bits instead of 16 makes an audible difference. Having more bits is also more forgiving if you set the recording level a little low. When burning to CD, most recording software will use 16 of the 24 bits but if you can playback all 24 bts from the computer etc., it sounds great.

Also, I wouldn't reccomend using 48khz to record as 48khz is used mostly for the audio portion of videos.
Nothing wrong with recording in 24 bit.This is some benefit, as you can set the level conservatively lower without losing resolution. Then after you shift peak levels to -1 dB, you can convert to 16-bit without losing information.

Some (very few, but some) very high quality LPs have frequency response beyond 20 kHz, if you have really high quality properly adjusted playback equipment; recording at 48 kHz sampling can preserve that information. It also gives a wider transition band for the anti-aliasing filter of the AD converter, which can make smoother passband response.
 

daftcombo

Major Contributor
Forum Donor
Joined
Feb 5, 2019
Messages
3,687
Likes
4,068
I advise to record at 24/96 or 24/88.2 because there can be pops/clicks with frequencies above 22.5 kHz and you would like to remove them from the recording as well before downsampling and applying dither to have a solid 16/48 or 16/44.1.

Morover, doing it 96-->48 instead of 88.2-->44.1 allows you to apply downsampling not too steeply, letting the 20Hz - 20kHz range completely untouched.
 

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
2,681
Likes
4,219
Location
Liège, Belgium
I'd do 96kHz 24 bits, aiming for -20dBFS level
Then use Wavelab (or similar?) to identify and remove clicks.
Then normalize to taste.
 
OP
Lttlwing16

Lttlwing16

Active Member
Forum Donor
Joined
Feb 24, 2021
Messages
201
Likes
113
Did some more testing last night with normalization techniques. I compared:

1) capture ~ -6 dbFS to .wav (using ALSA arecord) at 32bit 96khz -- Peak normalized to -0.1 dBFS in Audacity (just below clipping indicator)
2) capture ~ -6 dbFS to .wav (using ALSA arecord) at 32bit 96khz -- Loudness normalized (EBU-R128) to -16.8 LUFS in Audacity (which was just below where the clipping indicator disappeared in Audacity)
3) capture to 95% peak on VU meter in arecord. Imported to Audacity-- no other processing.

The file was carefully groomed for pops/ and clicks prior to normalization of any type.

Files 1 & 2 were from the same master capture and match dB RMS exactly -19.12 dB RMS
File 3 was slightly lower energy at -19.34 dB RMS

The Audacity session was set to 32bit Float/ 96khz, and pulse audio was configured to playback at 32bit/96khz . All files were compared with the following chain: PC USB> Drop Grace Design SDAC XLR OUT>THX AAA 789 Balanced Out>DC Aeon RT.



What I discovered is peak normalization, despite having the same dB RMS as the R128 normalized audio, was markedly louder, and in my opinion harsher. However, the R128 normalized file sounded closer to the file that was captured at 95% peak without any processing. This makes sense considering how EBU-128 works, and why it was created.

Of the three, I felt file 3 offered the best fidelity, even at slightly lower dB level.

I did not test @MRC01 's amplify method, but will take a look tonight and report back.

Initial testing for my setup would have

3>2>1

Method 3 would require the most time/effort as each LP would need to be screened first to find the peak, then the input gain adjusted to get as close to 95% as possible.

Method 2 offers a nice balance, and I believe outside of headphones, the files may be indistinguishable. Additionally, by using the EBU-R128 normalization, one could create a standardized level between albums if playing on a system with limited gain staging or if consistency between albums is desired.
 
Last edited:
OP
Lttlwing16

Lttlwing16

Active Member
Forum Donor
Joined
Feb 24, 2021
Messages
201
Likes
113
Interestingly AES recommends -16 LUFS for track normalization, which method 2 was slightly below. Obviously album normalization would be preferred here to preserve the album dynamics from song to song. Album Dynamic is recommended to -14 LUFS. Read more about AES recommendations here.
 

Bob from Florida

Major Contributor
Joined
Aug 20, 2020
Messages
1,256
Likes
1,163
Interestingly AES recommends -16 LUFS for track normalization, which method 2 was slightly below. Obviously album normalization would be preferred here to preserve the album dynamics from song to song. Album Dynamic is recommended to -14 LUFS. Read more about AES recommendations here.
This might sound simplistic and bear in mind it has been awhile since I have done album captures. I used a handheld digital audio recorder at 96/24 capture using line in from the record out of my preamp. AGC off and level set for each album where max peaks were located by ear on each album. Level set to just under 0 db at the loudest passages. Saved as wave files entire album side per file with no manipulations after recording. The majority of my albums have no pops or clicks - meticulous care and record cleaning. So, pops and clicks are not much of a bother to me. I guess in the end it depends on how much time you want to spend doing the album rips to digital.
 

danadam

Addicted to Fun and Learning
Joined
Jan 20, 2017
Messages
956
Likes
1,496
1) capture ~ -6 dbFS to .wav (using ALSA arecord) at 32bit 96khz -- Peak normalized to -0.1 dBFS in Audacity (just below clipping indicator)
2) capture ~ -6 dbFS to .wav (using ALSA arecord) at 32bit 96khz -- Loudness normalized (EBU-R128) to -16.8 LUFS in Audacity (which was just below where the clipping indicator disappeared in Audacity)
Why not use the same capture and normalize it two ways?
capture to 95% peak on VU meter in alsamixer. Imported to Audacity-- no other processing.
Where does alsamixer have VU meter? Or did you mean arecord?
Oh, and assuming it shows peak value and it is exactly 95%, this should translate to about -0.45 dBFS. Maybe that would explain the slight difference between 3 and the other two.
What I discovered is peak normalization, despite having the same dB RMS as the R128 normalized audio, was markedly louder
That's somewhat strange. Do you have a way the measure the LUFS loudness of the peak normalized file?
the R128 normalized file sounded closer to the file that was captured at 95% peak without any processing. This makes sense considering how EBU-128 works, and why it was created.
I'm not sure I see the connection.
Of the three, I felt file 3 offered the best fidelity, even at slightly lower dB level.
Any difference in fidelity between those files should be only from the fact that they are three separate captures, e.g., in one capture a click or pop happened while in the other it did not, or was quieter. If there actually is any fidelity loss from applying +/- 6 dB gain in a 24 bit file, assuming no clipping happened, I would start looking for something that is broken :) . Unless the ADC is so bad that its noise floor is at vinyl record level and that 6 dB makes a difference.
But that's just me :)
 
Top Bottom