Expansion versus Normalization

This forum is for discussing Wave Corrector, audio projects, tips & tricks etc.

Moderator: Derek

Expansion versus Normalization

Postby paul » Sun Jan 23, 2005 3:31 am

I am a new, and very delighted, user of Wave Corrector. So far, I've done a single project- converting all Beethoven's string quartets from vinyl to CD. These are the steps I followed:
- recorded from LP using a hoped-for target of -6db (I didn't always achieve this!)
-declicked and super-clicked at the default settings until I registered 0 new clicks
- adjusted the Channel Balance
- normalized to -1db

The results have been far better than I expected, my only criticism being a slightly "dull" sound on some of the tracks.

I am aware that some audio software allows the sound to be expanded before being normalized. Can anyone explain exactly what this means and how it works please? And please feel free to criticize the steps I followed above and recommend better alternatives.
---
Paul
paul
 
Posts: 9
Joined: Sun Jan 23, 2005 3:10 am

Postby citguy » Sun Jan 23, 2005 8:30 am

Hi Paul. You can always use the corrected and uncorrected playback buttons to compare your processed and unprocessed tracks. I suspect you may have over processe a bit going for zero clicks. You know, one set of play back buttons plays only the current screen and the other set starts from the beginning of the track or tracks if "all tracks" are displayed below.
Stan
citguy
 
Posts: 151
Joined: Sun Jan 09, 2005 4:32 pm
Location: Portland, Oregon USA

Postby Derek » Sun Jan 23, 2005 1:05 pm

citguy wrote:Hi Paul. You can always use the corrected and uncorrected playback buttons to compare your processed and unprocessed tracks. I suspect you may have over processe a bit going for zero clicks. You know, one set of play back buttons plays only the current screen and the other set starts from the beginning of the track or tracks if "all tracks" are displayed below.
Stan

I would concur absolutely with this. It is not usually a good idea to keep reprocessing until the detected clicks drop to zero. This is because of the problem of 'false positives' The process of discriminating between clicks and music is never perfect and therefore, the more times you process, the more is the likelihood that you will pick up a lot of false positives. This may be what is causing the dull sound you are reporting.

I thionk the best approach is to process once and then to audition as citguy suggests. Only re-process if there are still audible clicks to be removed. And even then, it is better to remove them manually or to mark a block around them and use Block-SuperScan. That way it is only the region immediately around then rogue click that will be reprocessed.

Having said all that, another user has reported that he successfully does what you do but at detection threshold 2. If you do decide to do multiple superscans, then threshold 2 is to be preferred as the likelihood if false prositive is much reduced.

Regarding expansion before normalisation, I'm not aware of such a process. Expansion usually refers to the process of expanding a signal that has been compressed to fit into a transmission or recording medium, for example Dolby B. You wouldn't normally expand a signal unless it had previously been compressed.
Derek Higgins
Wave Corrector Developer
http://www.wavecor.co.uk
Derek
Site Admin
 
Posts: 476
Joined: Sat Jan 01, 2005 7:57 pm
Location: Coleshill, United Kingdom

Postby paul » Wed Jan 26, 2005 8:12 am

I would concur absolutely with this. It is not usually a good idea to keep reprocessing until the detected clicks drop to zero. This is because of the problem of 'false positives' The process of discriminating between clicks and music is never perfect and therefore, the more times you process, the more is the likelihood that you will pick up a lot of false positives. This may be what is causing the dull sound you are reporting.
As a matter of interest, if the reported number of clicks found is 0, does that mean that Wave Corrector has made no changes at all?

Regarding expansion before normalisation, I'm not aware of such a process. Expansion usually refers to the process of expanding a signal that has been compressed to fit into a transmission or recording medium, for example Dolby B. You wouldn't normally expand a signal unless it had previously been compressed.
Let me quote from a tutorial I found on the web:
"Vinyl recorded material suffers from a lack of dynamic range. The Cool Edit software allows you to define an "expansion" algorithm that can expand the dynamic range of the source material where you can actually create a wider dynamic range than the album had to begin with. This is another reason for the -6db initial record level. To expand the music's dynamic range, you need what is called head room to allow you the ability to increase the dynamic range without causing clipping. You will need to experiment with this to find the optimal settings. Generally expansion at 2:1 is over expanded and the default expansion offered is even more, 3:1. I generally run between 1.2:1 to 1.4:1 as a healthy value. You don't need to perform this step but it will add life to your music."
The author then goes on to specify a normalization step as well. Does this make the term "expansion" any clearer? if so, do you know what this step achieves different from normalization, and how?
---
Paul
paul
 
Posts: 9
Joined: Sun Jan 23, 2005 3:10 am

Postby Derek » Wed Jan 26, 2005 9:25 am

paul wrote:As a matter of interest, if the reported number of clicks found is 0, does that mean that Wave Corrector has made no changes at all?


Yes, that's correct, assuming of course you are using just the click filter and no other filters.

Regarding expansion before normalisation, I'm not aware of such a process. Expansion usually refers to the process of expanding a signal that has been compressed to fit into a transmission or recording medium, for example Dolby B. You wouldn't normally expand a signal unless it had previously been compressed.
Let me quote from a tutorial I found on the web:
"Vinyl recorded material suffers from a lack of dynamic range. The Cool Edit software allows you to define an "expansion" algorithm that can expand the dynamic range of the source material where you can actually create a wider dynamic range than the album had to begin with. This is another reason for the -6db initial record level. To expand the music's dynamic range, you need what is called head room to allow you the ability to increase the dynamic range without causing clipping. You will need to experiment with this to find the optimal settings. Generally expansion at 2:1 is over expanded and the default expansion offered is even more, 3:1. I generally run between 1.2:1 to 1.4:1 as a healthy value. You don't need to perform this step but it will add life to your music."
The author then goes on to specify a normalization step as well. Does this make the term "expansion" any clearer? if so, do you know what this step achieves different from normalization, and how?


Yes, that makes it clearer. They are suggesting using a technique that, in effect, turns the volume up during loud sections and turns it down during quiet sections. This is the inverse of what a lot of AM radio stations do. In their case, they use volume compression, boosting the quiet signals so that everything comes out at about the same volume.

Wave Corrector does not do anything like this. It is debatable whether vinyl records have less dynamic range than CD's. A really clean, good quality vinyl record should have a dynamic range appoaching that of a CD anyway. And in my opinion, there is no need for such a process.
Derek Higgins
Wave Corrector Developer
http://www.wavecor.co.uk
Derek
Site Admin
 
Posts: 476
Joined: Sat Jan 01, 2005 7:57 pm
Location: Coleshill, United Kingdom

Expansion vs. normalization

Postby vinylmaster » Thu Jan 27, 2005 7:27 pm

Hello all.

Regarding the expansion topic, before all this digital stuff came around, there were a few companies that produced hardware that performed a few of the tricks to which Paul originally referred, namely expansion.
dbx produced a few pieces of good equipment (believe it or not) that performed an expansion. For example, their "3bx" audio sound processor could expand a signal a bit.
From what I remember of the literature that came with the piece, dbx explained that vinyl had a certain compression in order to make the music physically fit on the record. Otherwise some of the crecendos we hear in classical music would literally ride off the record. The record companies "compress" the signals just a bit. The dbx processors could restore some of the dynamics. (Any errors in this explanation is solely mine.)
vinylmaster
 
Posts: 7
Joined: Thu Jan 13, 2005 5:29 pm
Location: San Antonio, Texas (USA)

Postby citguy » Thu Jan 27, 2005 9:38 pm

I have three Sanyo Super D noise reduction units whcih were Sanyo's answer to DBX back in the 70's. They are best used when recording on tape. The signal is "encoded" or compressed for recording then on play back it is expanded and the overall playback level is reduced thus lowering tape hiss to well over 90 dp signal to noise ratio. It is critical that the expansion and compression occur at the same "ratio" or all kinds of problems occur. When used properly it is impossible to hear "tape movement" at playback. I have never attempted to expand normal vinyl recordings. I am not sure it is possible. As I understand Wave Corrector's nomalization, the peak levels are determined and then the peaks are increased to maximum levels. But, I have wondered how the ratio of loudest to softest material is maintained. It has been my experience that when normalizing, the hiss level comes up also. So I have always applied the hiss and hum filter after normalization. I also have a Burwen Dynamic noise reduction unit which reduces noise on the fly mainly during quiet passages on fm, tape or vinyl. It also needs to be used with great care or additional unwanted effects can be introduced to the material.
Stan
citguy
 
Posts: 151
Joined: Sun Jan 09, 2005 4:32 pm
Location: Portland, Oregon USA

Re: Expansion vs. normalization

Postby adaywayne » Sat Jan 29, 2005 1:09 am

vinylmaster wrote:Hello all.

Regarding the expansion topic, before all this digital stuff came around, there were a few companies that produced hardware that performed a few of the tricks to which Paul originally referred, namely expansion.
dbx produced a few pieces of good equipment (believe it or not) that performed an expansion. For example, their "3bx" audio sound processor could expand a signal a bit.
From what I remember of the literature that came with the piece, dbx explained that vinyl had a certain compression in order to make the music physically fit on the record. Otherwise some of the crecendos we hear in classical music would literally ride off the record. The record companies "compress" the signals just a bit. The dbx processors could restore some of the dynamics. (Any errors in this explanation is solely mine.)

~~~~~~~~~~~~~~~

It's slightly misleading to say that vinyl is "compressed". Rather, what happens is an application during the recording process of "equalisation", by which the low frequencies have their signal level reduced and the high frequencies have them increased. This is called the RIAA equalisation curve, and has been standardised for LPs since the mid-to-late 1950s. On playback, the phono pre-amplifier (assuming your amplifier has one!) applies a reverse RIAA curve to bring the frequency/signal level realtionship back to what it should be. And Derel is correct......good LPs have as good as, or better, dynamic range as/than most CDs.
adaywayne
 
Posts: 43
Joined: Sun Jan 09, 2005 8:53 pm
Location: New Jersey, USA

Postby Glenn » Thu Mar 03, 2005 7:10 am

Hi everyone;

I recently joined this forum group and am intrigued by this ongoing debate regarding dynamics expansion. I've been recording LP to CD off and on for the past several years but kept putting it aside because I couldn't get the 'magic sound' I recall when I blasted these recordings in my youth. I've looked at software dynamic range expanders and other audio 'enhancers', but still couldn't get the rich, dynamic sound I knew was on these records.

In recent times, I've grappled with the definition of dynamics in terms of what we hear, or what impresses us as 'dynamic'. The question as to whether an LP is as 'dynamic' as a CD is becoming moot as the music on many CDs these days is recorded with far less actual dynamic range than what could be heard on LPs. Apart from the added RIAA equalization of LP's, both media apply normalization (as opposed to dynamics) to get the most sound out of them. My view now is not whether dynamic range is important; its more the normalization engineered into the recording itself that concerns me.

The question is what kind of normalization? Simple gain normalization is the same as turning the volume up to a set level. RMS normalization, on the other hand, essentially raises the average sound level of a recording within a given dynamic envelope. This tends to bring up the quieter passages more than the louder ones and is basically a form of compression, but the implementation of it is different. In this form, most of the instruments are lifted to a perceptually higher averagevolume, increasing their audibilty and the overall apparent clarity of the sound. If not overdone, it can make a recording sound more 'dynamic' by virtue of this effect. This is not news; practically every recording you've heard has had some form of what is effectively RMS equalization applied in the studio. You can see this in the waveforms in any wave editor. In the early days, it was more necessary in order to make full use of noisy recording media.

The practice of recording with a wide dynamic range, such as what one might experience with a full orchestra in a quiet concert hall, was experimented with in the earlier days of digital audio. It has fallen out of favour, as such recordings are impractical for the everyday listening environment of most consumers. In my experience, increasing the dynamics of a recording won't improve the sound. Rather, careful initial recording to the hard-disk with a mind to preserving the dynamics in the vinyl, plus the use of subtle equalization and maybe some RMS normalization (if necessary) will.

My current approach is to input the soundcard from the pre-amp as loudly as possible (metered typically within -.5db). I also apply real-time software sub-sonic filtering during recording to stabilize the written waveform for the fullest dynamic range possible. As my record player sits on the desk above the pc, I've taken extreme steps to minimize the vibration from the cooling fans, and have achieved a noise reduction of better than -10db this way. Plus, I always record with the speaker volume off. After recording, I clean the sound with Wave Corrector, doing a single level 3 scan, then make a few adjustments to the larger corrections. If the surface noise is still unbearable, I'll do a fingerprint noise reduction, but only as a last resort. In most cases with pop music , all that is required further is maybe a little bass eq, to simulate the punch that high volume feedback from the loudspeakers gives, or a bit of treble. Sometimes with quieter music, an RMS normalization of about -8 or -9 db brings considerable life to the sound. The effect is not as good as a studio track remastering, but it comes close. If I use it, I apply the bass afterwards; it gets quite a boost this way. Ofcourse, the gain has to be adjusted downward prior to this.

It has been my experience that this approach gives the best overall sound reproduction. If the vinyl groove is not too bady worn, the sound quality compares favourably against the commercial counterparts (especially the early transcriptions). If the groove is badly worn, or if the recording was poor in the first place, I haven't yet found a way to fix this. I have to agree with Derek: I don't believe that dynamics expansion will improve the sound in any way.

Glenn
Glenn
 
Posts: 212
Joined: Mon Feb 28, 2005 3:07 am
Location: Toronto, Canada

Postby paul » Thu Mar 03, 2005 7:23 am

Sometimes with quieter music, an RMS normalization of about -8 or -9 db brings considerable life to the sound. The effect is not as good as a studio track remastering, but it comes close. If I use it, I apply the bass afterwards; it gets quite a boost this way. Ofcourse, the gain has to be adjusted downward prior to this.

How do achieve this RMS normalization? Do you do this within Wave Corrector, or what?
Your comment about recording at -0.5db is really interesting, as it seems to contradict advice I've seen in other places (e.g. record at -6db which gives some head room for expansion without clipping; then normalize).
---
Paul
paul
 
Posts: 9
Joined: Sun Jan 23, 2005 3:10 am

Postby Glenn » Thu Mar 03, 2005 7:56 am

Hi Paul

I record using the Audiotools recorder, which also has a pretty decent 2-pass RMS normalizer. Apart from it being affordable, Audiotools has an intuitive way way of cascading filter operations. I record at a high level using the MK III skin, which has peak hold (briefly) meters. I cascade the rumble filter, then a peak limiter (-.01db, just in case), and finally the meters. This is my default recording profile.

I've come to the conclusion that high level recording offers the best sound quality through personal experience. I believe this is weighted by the fact it has always been best to record at high levels. Much has been said about digital's 'superior' dynamic range, and I once believed -6db was as good a recording level as any. After all, the cd has another 50 or 60 db to spare. But raising the level to maximum using gain normalization just doesn't seem to bring out the sound. It raises the volume, but not the fidelity. If you record at a higher level, then use gain attenuation as necessary during any further processing, the sound is much more intimate.

Glenn
Glenn
 
Posts: 212
Joined: Mon Feb 28, 2005 3:07 am
Location: Toronto, Canada

Postby paul » Thu Mar 03, 2005 8:08 am

I've come to the conclusion that high level recording offers the best sound quality through personal experience.

I've been told that the signal must never exceed 0 for digital recording - do you agree? If so, it's really quite hard to make the optimal setting without playing the entire work first, isn't it?
---
Paul
paul
 
Posts: 9
Joined: Sun Jan 23, 2005 3:10 am

Postby Glenn » Thu Mar 03, 2005 8:21 am

Hi Paul,
Actually, with peak limiting the level in digital recordings can theoretically go beyond this, as there will be no clipping distortion. I've seen several commercial CDs with extensive clipping. I don't think the sound is as musical though. The usual rule is to record as loud as possible, without going overboard. It takes more effort to find the maximum level of a record, but it helps if you're already familiar with the music, and know where loud parts are.
Glenn
Glenn
 
Posts: 212
Joined: Mon Feb 28, 2005 3:07 am
Location: Toronto, Canada

Postby adaywayne » Thu Mar 03, 2005 3:52 pm

Glenn wrote:Hi Paul

Much has been said about digital's 'superior' dynamic range, and I once believed -6db was as good a recording level as any. After all, the cd has another 50 or 60 db to spare.

Glenn


Glen, can you explain a bit more about the 50 to 60dB to spare? I thought all CDs were standardized at a peak of -1 dB, though I believe some pop music CDs now have a peak of 0dB. But 50 to 60???????
Last edited by adaywayne on Thu Mar 03, 2005 7:06 pm, edited 1 time in total.
adaywayne
 
Posts: 43
Joined: Sun Jan 09, 2005 8:53 pm
Location: New Jersey, USA

Postby adaywayne » Thu Mar 03, 2005 3:57 pm

Glenn wrote:Hi Paul,
Actually, with peak limiting the level in digital recordings can theoretically go beyond this, as there will be no clipping distortion. I've seen several commercial CDs with extensive clipping. I don't think the sound is as musical though.
Glenn

~~~~~~~~~~~~~
Afraid I have to disagree. The way A/D converters work it is mathematically impossible to go to a higher signal level than 0dB (dBV or dBu). Beyond that, no value is assigned to the voltage, and the resulting sound is unbearable.....much worse than analog clipping!
Arnie
Last edited by adaywayne on Thu Mar 03, 2005 7:05 pm, edited 1 time in total.
adaywayne
 
Posts: 43
Joined: Sun Jan 09, 2005 8:53 pm
Location: New Jersey, USA

Next

Return to Chat

Who is online

Users browsing this forum: No registered users and 0 guests

cron