Subject:Vinyl Album Restoration (Need Advice).

Posted by: Chewjack

Date:7/18/2007 1:19:46 PM

Vinyl Album Restoration (Need Advice).

I have several vinyl albums that I digitized and want to put on CD. I have recorded them as 48K/24bit Stereo PCM .WAV Files. These albums contain a lot of crackling and popping. I believe I know the steps to perform to clean them up but I am unsure of the proper order of these steps that would yield the best results. Can somebody please advise?

1. Run the Click and Crackle Removal Plug in.
2. Run the Normalize function
3. Run the Bit depth converter
4. Run the Resample command

Are there additional steps that I should take?

Thanks in advance!

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: jumbuk

Date:7/19/2007 12:04:35 AM

If you have Sony NR (bundled with SF9), I would use it to remove crackles and pops.

I would not bother with normalisation. Your vinyl source will have been mastered so that all tracks were consistent - you will destroy that if you normalise.

I would record in 24-bit and not alter the bit depth. I would use CD Architect (bundled with SF8+) to assemble the CD, and set dither and bit-depth at the last stage within CDA.

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: rraud

Date:7/19/2007 10:18:20 AM

There has been a wealth of info posted on this subject already in this forum. Do a search.
I don't think you would benefit very much using 24bit since you and you may be converting back to a 16 bit audio CD anyway and any internal processing will be 32bit. That said, I would use 24 bit for classical or other material with a wide dynamic range.

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: ctbarker32

Date:7/19/2007 9:31:22 PM

This is a topic that's near and dear to my heart.

I have hundreds of LPs that have never been issued on CD. Many are from the wonderful German jazz label - ECM. These are priceless recordings to me.

I am an audiophile and have many turntables ranging from entry level to very high end.

Having spent several years working on this, I think I have now zeroed in on what I consider a very good methodology that I can repeat consistently.

There's probably not enough room for me to go over every detail but I can make some suggestions.

1. I am a purist, and want to capture everything as pristine as possible and thus I stay away from any potentially damaging processing.

2. The old adage, garbage in equals garbage out, is very apt here. You need to do everything you can at the front end of the process so you can capture the best snapshot of the vinyl. Trying to "fix it in the mix" by using a lot of brute force after the fact will give poor results.

3 Pertaining to item #2, you need to really concentrate on your LP playback gear and ideally you need a vinyl cleaning machine (like a VPI 16.5). I can't emphasize enough how important it is to get the LP playback process optimized. I am not talking super expensive equipment here. In my setup I use a Rega P2 clone (Goldring 1.2) that costs about $350. I use a Shure M97 cartridge. I have done some mods to this turntable to get the last ounce of performance out of it but the most important is accurately aligning the cartridge geometry. You will be amazed how previously noisey records suddenly are whisper quiet if everything is aligned properly. This is a long topic so I'll refrain from all the details for the moment.

4. Because of the very low voltages of a cartridge output, it is critical that you verify all the connections that lead to the computer input to ensure no hum or RF noise is introduced. Having a turntable in close proximity to a computer is recipe for an RF nightmare.

5. I am currently using a very inexpensive Firewire A/D input called the Behringer FCA202 which is about $80. The advantage here is the A/D is done outside the computer.

6. I record at 24 bit 48khz sample rate. Since I store all my music on a music server and use the Slimdevices Squeezebox 3 for playback I do not need to downsample since the SB3 can play back 24b/48k files directly. I have developed a few scripts that I run in Soudforge that save the individual tracks from the captured sound file that I have marked up into regions. I then use these files as input to the Flac Frontend that encodes to the FLAC lossless format that the SB3 can decode.

7. I do very little processing on my captured audio. I run a script that runs a high pass eq to eliminate all audio below 20hz, then it proceeds to normalize using the Music factory preset. I believe it is necessary and desirable to run the normalize function so as utilize the full 24bit bandwidth and bring the level up modern mastering levels. This could be debated but if you look at the raw audio capture it will be clear that you are not using the full 24bit bandwidth. I do not run any other filtering except in some very rare instances I will run the Audio Restoration plugin on a very small audio selections where ticks are evident. I never run it on the whole file but just on a couple second sections at a time.

I hope some of this makes some sense and is helpful. It may appear obsessive to some but to me this is priceless music.

I will make a bold statement that the results I am getting with my technique result in audio that to many would find indistinguishable from a CD in dynamic range and quietness. It may sound like bravado but I could provide some sample files to verify my claims.

-CB

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: jumbuk

Date:7/22/2007 6:34:29 AM

"7. I do very little processing on my captured audio. I run a script that runs a high pass eq to eliminate all audio below 20hz, then it proceeds to normalize using the Music factory preset. I believe it is necessary and desirable to run the normalize function so as utilize the full 24bit bandwidth and bring the level up modern mastering levels. This could be debated but if you look at the raw audio capture it will be clear that you are not using the full 24bit bandwidth."

It was my understanding that 24-bit audio was full floating point and hence normalizing does nothing in terms of bandwidth.

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: ctbarker32

Date:7/22/2007 9:08:42 AM

I may have been sloppy in my use of the word Bandwidth.

Of course, rms nomalization raises the volume of the file so the peaks are at 0db but the dynamic range between the loudest and softest parts remain the same.

When I record I typically keep my levels from hitting peak to avoid digital clipping. Of course, 24 bit recording has more headroom. So, since I have purposely under utilized the 24 bit headroom on the original recording, I like to use the rms normalize function to maximize the final audio.

In my listening tests (Martin Logan Vantage speakers, SB3 24b/48k files) I have found that my LP captures sound very good and very close to ECM original mastered CDs in dynamic range and sound quality. In fact, I have a few 30 year old ECM LPs that have also been remastered at 24 bit (downsampled to 16 bit for CD) and I have directly compared the recordings. Of course the remastered CD is the last word in clarity but my attempts at capturing the LP sound surprisingly close.

-CB

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: jumbuk

Date:7/22/2007 5:55:30 PM

I'm not surprised you are getting good results with the gear you are using.

I agree, keep processing to a minimum if you have reasonable quality originals.

In my own case, my vinyls are not in great shape. Most were played on cheap turntables at parties through the sixties and seventies, and a few have quite bad scratches as a result. I have found Sony NR pretty good at reducing crackle to an acceptable level without destroying the original signal.

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: Vocalpoint

Date:7/23/2007 8:10:17 AM

Of course, rms nomalization raises the volume of the file so the peaks are at 0db but the dynamic range between the loudest and softest parts remain the same

You do know that you are permanently destroying your dynamics by applying any "normalization". This is a kind processing that should never be used during transfers.

For a "purist" as you described yourself - I am surprised about how much processing you actually put into your transfers. If you need your track louder - turn up the volume knob....cranking the RMS just pushes you closer and closer to the limit - where there is no more dynamic range left to be had.

Just my two cents....

VP

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: ctbarker32

Date:7/23/2007 9:40:04 AM

With respect, I do not believe you are correct about "destroying your dynamics".

As I mentioned before, I am using Sound Forge's normalize process using the system default "Normalize to -16 dB (music)". Here is a quote from the help file:

When using RMS levels, set the Normalize to fader to -6 dB or less. Normalizing to 0 dB boosts the signal so that it has the same apparent loudness as a 0 dB square wave (which is incredibly loud). If you were to do so, all of the dynamic range of the signal would be squashed and all the peaks would either be clipped or seriously compressed. Normalizing a peak to 0 dB is OK, but normalizing RMS to anything above -6 dB can compromise sound quality.

I run the process on the entire album length, not individual songs. I believe the -16.00 dB setting is benign when used this way.

I have just scanned my copy of Bob Katz's "Mastering Audio - the art and science". In it he comments about normalization:

The Myths of Normalization

The Esthetic (sic) Myth: Digital audio editing programs have a feature called Normalization, a semi-automatic method of adjusting levels. the engineer selects all the segments (songs), and the computer grinds away, searching for the highest peak on the album. Then the computer adjusts the level of all the material until the highest peak reaches 0 dBFS. If all the material is group normalized at once, this is not a serious esthetic (sic) problem, as long as all the songs have been raised or lowered by the same amount.

As I explained, I am trying to maximize the use of the 24bit resolution I am using by making sure the entire 24 bits are utilized. I cannot get that just by using a straight recording. I always leave a few bits on the table.

I have directly compared a track from a 30 year old ECM album by Jack DeJohnette. The album is called Pictures and one of the tracks is included on a 24 bit remaster compilation in the past few years. It has scads of dynamic range. ECM rarely does remasters so it is an interesting to compare the sound quality. Firstly, both my analog capture and the remastered version have roughly equivalent sound playback levels. That is I do not have to change the volume significantly between listening to the two tracks. If I did not normalize my analog capture it would be significantly quieter than the ECM remaster. Obviously, the ECM remaster is superior in clarity to my analog copy but it is gratifying how close the two sound otherwise.

I may be misguided as I am only an amateur but I do listen carefully and try to learn through reading and practical experience.

-CB

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: Vocalpoint

Date:7/23/2007 1:13:16 PM

CB,

With respect - no problem. All I know is that any process that artificially "boosts" any signal to increase apparent loudness is changing the physical make-up of the audio itself. And most likely changing the "dynamic" as intended by the artist.

I am certain that if ECM were to re-issue some of these recordings - they would certainly not hit their master transfers with a "normalize" function of any sort. Today's modern "volume" is achieved through the mastering process usually by a combo of EQ, compression and certainly limiting.

I see you are a fan of Bob Katz. He is a huge advocate of NOT boosting volume just for the heck of it. Normalizing is just that - boosting the volume for the heck of it.

The best tip I received in the last few years is not get pulled into the volume wars - which I believe you are doing by normalizing to 0dbFS. Any recording that approaches 0dbFS is - in my opinion - distorted and is tiring to listen to after a while. (Did you know that a digital recording approaching 0bdFS is about the same as listening to an analog recording with a setting of about +20db past 0db - who would ever do that?)

And the oft quoted myth of "using all 24 bits" of resolution is just that. A myth. Believe me - It's okay to leave a "few bits" on the table - especially with 24bit resolution. Your recordings will be much, much better to listen to if you employed a standard two stage process - a transfer stage where all the original headroom from the original recording is transferred straight from vinyl peaking at around -6dbFS with NO normalizing. Then move the files through a "master" stage with tasteful (very ,very transparent) EQ and most likely a sprinkle of limiting to bring the volume up to today's standards.

Continued success!

VP

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: jumbuk

Date:7/24/2007 12:30:13 AM

I still don't understand the purpose of normalising with floating point data (24-bit or 32-bit).

When you capture data in floating-point, you will store 24 bits of data irrespective of the level. All normalisation will do is to multiply all values by a fixed number, calculated so that no resulting value clips.

If you normalise the whole CD at once, all tracks will increase pro-rata to the maximimum volume without clipping. That's fine, although you can't increase the amount of info captured on your recording by doing this, and you are altering the stored data.

If you normalise each track separately (which is what I thought you were doing) then the relative volumes will be altered and you will finish up with a CD that does NOT match the original vinyl. If this is not your approach, then there is no real issue.

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: Vocalpoint

Date:7/24/2007 4:58:15 AM

When you capture data in floating-point, you will store 24 bits of data irrespective of the level. All normalisation will do is to multiply all values by a fixed number, calculated so that no resulting value clips.

Exactly!

There is absolutely no point to normalizing anything. All this does is destroy the original program dynamic and of course the material since it does not equal the original - especially softer pieces which could sound quite annoying after having their peaks raised so suddenly to 0dbFS (in some cases).

VP

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: randym

Date:7/25/2007 7:14:59 AM

rraud wrote:

"I don't think you would benefit very much using 24bit since you and you may be converting back to a 16 bit audio CD anyway and any internal processing will be 32bit."

Are you certain of this? In my batch processing, I start by converting from 24 bit to 32 bit floating so that later step processing is done in higher precision. Specifically, my batch processing steps are (starting with 96Khz, 24 bit):

1. Convert to 32 bit float
2. De-click and de-crackle
3. Resample to 44.1Khz with filtering
4. Normalize
5. Convert to 16 bit with dithering and noise shaping
6. Save

Are you saying I can eliminate step 1 with no effect on the final outcome? The fact that it takes several minutes to convert to 32 bit float suggests that SF9 does not hold a 32 bit version of the file in memory, which would seem to be necessary if processing is done in 32 bit.

One possibility is that each processing step is done in 32 bit float but the result is saved in 24 bit, in which case it would be better to convert to 32 bit float from the outset.

Clarifications would be appreciated.

Randy

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: rraud

Date:7/25/2007 1:55:47 PM

I don't know if Batch processing uses the 32bit float, in SF it is the default unless it's disabled. J. can explain how this work and if it's also in the Batch converter.
I should have clarified that using 16 vs. 24 is material dependent. For instance a squashed thrash band vs. jazz or classical.

Subject:RE: Vinyl Album Restoration (Need Advice).

Reply by: ForumAdmin

Date:7/26/2007 7:20:57 AM

There's no need to manually convert to float. Sound Forge processes and stores all internal results as floats (VST, FX chainer) or, more typically, doubles.

Whether batch converter or the UI triggers processing is irrelevant.

Floating-point values introduced by processing are converted back to PCM only when the file is saved. Note that saving does not automatically dither, so you will want to keep the explicit bit-depth conversion step (with dither and noise shaping) if you are delivering 16-bit.

J.