The Missing Octave(s) - Audacity Remastering to Restore Tracks

[Chris A]

Alfred Brendel - Mozart Piano Concerto K-466:

 

[Chris A]

Case Study: Smashing Pumpkins' Siamese Dream (1993)

 

In 1991 audio mastering began general and widespread use of digital multi-band compression for CDs...and subsequently the vinyl master. 

 

In this particular album the use of compression is apparent, i.e., compression is different from limiting...otherwise known as clipping.  Only 1-2 dB of clipping is present so "Clip Fix" had a comparatively easy job of restoring the peaks.

 

To give you an idea of the mastering EQ used on this album, I'll show you the original curve for the first track--typical of the remainder of the album--that the mastering or mixing engineers used, i.e., the inverse of the unmastering curve:

 

 

Once this curve is stared at for a moment or two, it becomes apparent why this album is so fatiguing and strident: all frequencies between 200 and 2000 Hz are down by 12-16 dB relative to the peak boosted level at 2-3.5 kHz. 

 

This is one of the extreme recordings that I've run across.  You hear booming bass and electric guitars--albeit higher harmonics only (not their fundamental frequencies) overlaid by vocal line harmonics.   I find this entire album elicits a teeth-clinching response and an awareness of the harshness of the overall recording sound.  This is fairly typical of hard rock, metal, and alternative rock genres.  It's just that this album takes that experience to a new level.

 

When the unmastering curves are applied to each track so that the ears can adjust and somewhat relax, the sound of the album changes rather dramatically.  In fact, it becomes fairly monotonous and frankly uninteresting...albeit a sound that doesn't elicit an unconscious teeth clinching.

 

Chris

[Chris A]

When you combine the above equalization curve with the sensitivity of the human hearing system...defined by equal loudness contour plots such as the following...

 

 

You can see that the areas of high frequency hearing acuity correspond one to one (i.e. 2-4 kHz) with the extreme boost of the mastering curve, producing a predictable result.

 

Chris

[DrWho]

All of this thought is based on a false assumption about the spectral content of unprocessed instruments. It's also wrongly assuming that the unprocessed sound represents the desired result. In other words, targeting 1/f total response is just silly.

 

Trying to expand (uncompress) the signal - even with knowledge of the exact original processing used - results in new frequency content that is not harmonically related to the original source. This is extremely problematic when EQ is used post compression. Analog companding is a bread and butter concept to analog wireless microphones so I've spent a lot of time looking at the spectral content of various sources (mostly guitar and vocals, but drums are the worst). Any amount of mismatch is incredibly audible in a way that is very unpleasing.

 

On the EQ side of things, the assumption of the 1/f curve as a natural frequency response is not founded on reality. Each instrument has its own complex harmonic structure and its relative amplitude to the other instruments is incredibly important. Trying to EQ to this odd concept of "flat" totally ignores the complex interaction of EQ to the sound of any one instrument. When applying EQ to an entire mix,  the complex tonal character of each instrument changes at the same time. It's not possible to isolate individual sounds. Then doing so with the purpose of making everything the same level? Making everything sound the same is what rookie sound engineers do. A constant state of blah.

 

Using the kick drum as an example....there are various mic'ing techniques. The far field response is closest to the natural acoustic performance, but that includes too many reflections making it sound too distant. For rock and roll, we want that punch you in the gut riding the tempo feeling, which you get from nearfield techniques. However, the nearfield performance of microphones is very different from the farfield, and the short story is we end up with too much bass (due to proximity gain). The interesting thing is that a kick drum with a ton of single digit low frequency information doesn't sound punchy. It sounds very bland and lifeless actually. The interesting thing is that fixing this with a compressor sounds more punchy than fixing it with EQ. In fact, boosting the low frequency EQ so you can use more tone shaping compression gives it an even fatter more thunderous sound....even though the final result has very little low frequency information. This is due to the time domain behavior of the gates and compressors. Keep in mind this has nothing to do with loudness wars for CDs, this is entirely a music creation thing. The point is that the resultant desired sound often does not match your assumptions about the final frequency response. A better kick drum sound for rock and roll simply does not have extreme LF extension even though it's present in the original close-mic'ed proximity effect signal. The more classic you go, the less deep it digs it seems. I think in part due to the microphones available, which in turn defined the sound of the genre.

 

Btw, I pick drums as an example because I get a lot of positive comments on my live drum sound....and I'm pushing well over 30dB crest factor in my live mixes. Applying your technique to my live mix would absolutely make everything fall flat and lifeless. Not only the drums, but all the other instruments that didn't have the same processing!  That sound to which you're aspiring here is the mediocre result of an amateur sound engineer where everything sounds the same.

 

Just to stress the point, the processing happening on the kick drum channel is different than the processing on all the other channels. It's impossible to fix one instrument without severely affecting everything else. And then doing this blind by looking at FFT's? Yikes! 

 

Yes, too much source material suffers from loudness wars, but don't think for a second that this method is "fixing" anything. Sure, you may prefer the subjective result, but just because sciencey looking methods are used doesn't make this a mathematically / scientifically sound approach. The numbers here are very flawed and that's the only reason I'm commenting. The subjective correlation to science is the heart and soul of audio engineering, and it hurts to see causality attributed to the wrong variables. I know it's an arrogant jerk move to suggest it, but it needs to be said.

 

I know guys that use dbx multiband expanders in their systems all the time and it sounds amazing. However, they're doing it from a purely subjective angle - and I think that's why it sounds better. I also know the guys behind those products and they've done their homework, and are published on these topics.

[Don Richard]

46 minutes ago, DrWho said:

Just to stress the point, the processing happening on the kick drum channel is different than the processing on all the other channels. It's impossible to fix one instrument without severely affecting everything else.

 

The mastering engineers manipulate the EQ in the L-R mix, not channel EQs. That's why it messes up the sound so badly. This is what gets unmastered.

[Chris A]

2 hours ago, DrWho said:

In other words, targeting 1/f total response is just silly.

Silly me--for almost two years, and over 1000 albums--and I've been wrong...

 

Don't like the answers?

 

Perhaps you should try it, Mike, instead of just proclaiming "it doesn't work".

 

2 hours ago, DrWho said:

Applying your technique to my live mix would absolutely make everything fall flat and lifeless. Not only the drums, but all the other instruments that didn't have the same processing!  

 

This statement assumes that you've actually heard the tracks mentioned in this thread.  I've found that the results to be exactly the opposite of this statement: the dynamics of the tracks actually increase on average by 1-3 DR Database points (crest factor), sometimes as much as 8 dB for severely clipped tracks that are restored. 

 

I've also found that music that needs your type of technique usually has issues...serious issues.  When you strip away the hype, what remains isn't very attractive.  I've found that there is a large amount of music out there that does have merit on its own without that sort of mastering "help".  Perhaps you don't deal with that sort of music?

 

BTW: The technique does work, and it works so well that I've extended it to every album that I own...but it takes a little practice to get best results, and the results are outstanding when you do. 

 

2 hours ago, Don Richard said:

The mastering engineers manipulate the EQ in the L-R mix, not channel EQs.

It's not difficult to EQ each track, as I've done it many times.  You can also seamlessly EQ sections of tracks. 

 

I've found, however, that most mastering relies on EQing across full stereo tracks (time and motion of the mastering people cost money for the musicians paying them to EQ their music), and with only a few category  exceptions--like pop stars' music from the last 25 years, which is EQing every second or so in some tracks, and which produces what I would call "corned beef hash" tracks that aren't hi-fi by any stretch of the imagination.

 

The technique that is described in this (now long) thread relies on the statistics of large sections of music integrated over time (using "plot spectrum" view) and by looking very closely at the spectrogram log(f) view (with settings shown in the figure below using Audacity 2.1.0) across the tracks for evidence of spectral shaping.  It's an extremely sensitive technique, I've found.

 

[Chris A]

Based on the work that I've done over the last couple of years, there seems to be a correlation between music tracks that endure over a longer span of time (i.e., after the hype and "...that's my generation's music..." effects have worn off). 

 

Since this topic is a bit off the beaten path from this thread, I'll start a new one to discuss those observations.

 

 

 

[DrWho]

On 11/2/2016 at 6:44 AM, Don Richard said:

 

The mastering engineers manipulate the EQ in the L-R mix, not channel EQs. That's why it messes up the sound so badly. This is what gets unmastered.

The compression, expansion, and EQ are happening in both places: the individual tracks and the total mix. Both are adjusted in light of each other, so hacking at the LR mix doesn't "fix" anything because the individual channels were compensated for it in the process.

[DrWho]

On 11/2/2016 at 8:01 AM, Chris A said:

I've found that the results to be exactly the opposite of this statement: the dynamics of the tracks actually increase on average by 1-3 DR Database points (crest factor), sometimes as much as 8 dB for severely clipped tracks that are restored. 

Just because a DR number increases doesn't mean it sounds better. I stand by my dull lifeless comments and base it on 24 years of experience on this very subject. 8 years of that has been in a professional capacity at a company that has likely preformed the most objective and subjective research on EQ and dynamics processing. Let's not confuse old age and narrow mindedness with experience and wisdom.

 

Ultimately you're arguing for a glorified smiley face EQ, or pressing the loudness button. It's what the amateurs do. And it does sound better! At first anyway...Nobody is contesting that your DR numbers increase.

 

I have tried your approach and don't like the results and even have mathematical basis for it. I don't even need to criticize genres and styles of music either!

 

More DR is not always more better. It's really that simple.

[Chris A]

Just reading your comments, Mike, it's clear that you haven't actually put any real effort into "unmastering".  It appears that you're trying to sell your "mastering" skills (i.e., as a novice that is trying to break into the business of doing what is commercially acceptable to gain favor), which this entire thread is actually aimed at undoing. 

 

Perhaps you should start your own thread to convince others to do nothing about their music's problems, and feel good about it without changing anything because "you really can't change anything".  I'll be reading that thread with interest if you do.

 

If I had stopped trying to unmaster at just a few hours of practice, I suppose I might buy into some of your arguments.  Talk to me about the validity of this subject when you've got about 3500-4000 total hours of daily unmastering (7 days/week). I find that very little of what you've said is actually true, unfortunately.  Like anything else where you appear to be knowledgeable but haven't actually spent any real time in an honest trial, the appearance of expertise fails to sustain itself. 

 

There are many reasons why this is so, but perhaps the easiest way for you to see this is the simple observation that virtually everyone that owns hi-fi two-channel setups has been doing creative EQing in the form of "house curves" (that aren't real house curves), EQing each album by hand using EQ units or dbx boxes, swapping out high output impedance amplifiers, and using room correction software that is also doing "house curves" and "dynamic EQ". 

 

These are all equivalent processes in terms of what they are trying to achieve.  However, the techniques shown in this thread are much more effective.  Each track is examined, unmastered, listened to, and iterated until the sound is balanced and the creative mastering EQ is largely removed.  The results (which you've apparently not heard) speaks for itself--for those willing to try.

 

Chris

[Chris A]

For others reading the recent posts here, note that you don't need years of expertise to get better sound. 

 

In fact, much better sound can be had within just a few hours, like the amount of time spent on building a DIY loudspeaker or manually balancing each loudspeaker's frequency response in a surround sound array (5.1, 7.1, etc.). 

 

Better is better.

[DrWho]

Nope, try again. Maybe try reading for comprehension next time. You aren't listening to what I'm saying.

 

And of course I'm going to pull in other related experience! Doing so doesn't nullify the other experiences! 

 

Garbage in, garbage out. You can polish a turd, but that doesn't make it a gemstone. It makes it a shiny turd. I don't like turds. Apparently you like shiny things.

 

Stop blaming source material for deficiencies in the playback systems.

 

Pink Floyd, Michael Jackson, Elvis Presley, the Beatles,  Madonna, Metallica, Rush etc wouldn't have had the necessary authenticity to impact culture without the intentional production quality that you're unsuccessfully trying to remove. Or I suppose we should just piss on their validity as music since it's not classical or jazz? 

[Chris A]

31 minutes ago, DrWho said:

Stop blaming source material for deficiencies in the playback systems.

Umm, I believe that you're digging yourself a hole here...

 

31 minutes ago, DrWho said:

Pink Floyd, Michael Jackson, Elvis Presley, the Beatles,  Madonna, Metallica, Rush etc wouldn't have had the necessary authenticity to impact culture without the intentional production quality that you're unsuccessfully(...?) trying to remove.

 

I've tried a few titles that you've mentioned on the forum before in years passed (...and eventually gave those CDs away).  They were what I'd call lo-fi.  If I were being as abusive as you currently are, Mike, I'd comment on them specifically, but I sense that you don't want me to do that (I don't want to either: it's particularly distasteful).

 

Some of the titles from the artists that you mention above have responded most successfully of all the unmastering that I've done.  The reason: some of them needed unmastering the most. 

 

If you believe that what you're defending (i.e., current music industry mastering practices) is superior, then I recommend starting your own thread and talking about that.  This thread isn't about defending current music commercialization practices, in fact, it is rooted in the increasing abuses of that profession on its products over an extended period of time, and the loss of public support and sales for that type of practice.

[mikebse2a3]

21 minutes ago, DrWho said:

Stop blaming source material for deficiencies in the playback systems.

 

Source material is by far the weakest link in many high quality playback systems today. There are no standards in the industry that guarantee that we listeners will experience the intents of the artist/recording engineer and many recordings are so seriously flawed that polishing the "Turd" is the best we can do if we want to listen to them at all.

 

Sorry but no defense is possible for the high percentage of crap recordings offered to us in the last few decades especially. It is painfully obvious that they were made to sound good in cars and on boom boxes with true high fidelity recording/archiving thrown out the window for the quick buck..!!!

 

 

miketn

[tromprof]

It is true that this will not make badly recorded material sound better (polishing the turd), but it is beneficial for well recorded but  badly altered recordings. It is easy to screw the process up, but it is startling to hear the difference when well done, at least to my uneducated ears. 

[Don Richard]

2 hours ago, DrWho said:

The compression, expansion, and EQ are happening in both places: the individual tracks and the total mix. Both are adjusted in light of each other, so hacking at the LR mix doesn't "fix" anything because the individual channels were compensated for it in the process.

 

Professional mastering is done in a different studio and played back on different types of loudspeakers than used during the tracking and the mixdown process. The EQ and dynamics processing is applied to one channel and will be different for each channel during mixdown. This is done so each instrument will fit into the mix  and sound the way the mixdown engineer wants it to. After the mix is done for each track, the tracks go to the mastering lab where the tracks are put in the order they will appear on the album, the spacing between the tracks is inserted, and the table of contents along with time codes and text information is generated. The levels of each track are adjusted to be the same and at that point the album should sound like the mixdown engineer intended.

 

But it's not over. Usually the mastering engineer applies some EQ and limiting/compression to the L-R stereo bus, often at the direction of the producer or record company exec. It seems like many times the suits want the album to be as loud as possible at the expense of dynamic realism and accurate frequency response. This is the sort of damage that unmastering can reverse - the L-R bus contamination done by heavy handed mastering techniques.

 

Note that the EQing and dynamics control done at the mixdown process cannot be reversed. Each channel is processed with different attack, release, and ratio settings for dynamics and different EQ settings. There could be dozens of channels mixed down to two. If the mixdown engineer didn't do a good job you're screwed.

[Chris A]

I believe that we all know what each other are saying.  Arguing the sides much further is like arguing the sides on ABX testing.  It's not going to win anyone any friends.  I believe that tromprof is much closer to the truth about the technique--nothing is a panacea for the myriad gradations of the current state of music recordings.  To talk as if everything is black or white isn't useful...or truthful. 

 

58 minutes ago, Don Richard said:

The EQ and dynamics processing is applied to one channel and will be different for each channel during mixdown. This is done so each instrument will fit into the mix  and sound the way the mixdown engineer wants it to.

Yes, and you can hear that during unmastering.  The trick is rebalancing the two tracks so that is masks the duller instruments without resorting to a typical 8-16 dB of HF boost (for many not-so-good CDs) to make everything so strident and fatiguing.  It's the feedback using your ears and a little judgment...and I dare say...much better judgment for my listening room and loudspeakers than each individual mastering engineer canned for me (differently for each album, sometimes for each track) in his listening room using his loudspeakers.

 

58 minutes ago, Don Richard said:

But it's not over. Usually the mastering engineer applies some EQ and limiting/compression to the L-R stereo bus, often at the direction of the producer or record company exec. It seems like many times the suits want the album to be as loud as possible at the expense of dynamic realism and accurate frequency response. This is the sort of damage that unmastering can reverse - the L-R bus contamination done by heavy handed mastering techniques.

Compression and stereo effects manipulation are fatal flaws in those albums/CDs that cannot really be undone.  I usually go for the most dynamic CDs on the DR Database web site in order to minimize those problems.  The up-side to this is that the mastering effects are typically much more limited for these older CDs--particularly those produced in the first 8 years of CD production: 1983-1991. Relative to today's practices, they only used EQ to make everything louder.  That's the easiest to undo.

 

58 minutes ago, Don Richard said:

Note that the EQing and dynamics control done at the mixdown process cannot be reversed. Each channel is processed with different attack, release, and ratio settings for dynamics and different EQ settings. There could be dozens of channels mixed down to two. If the mixdown engineer didn't do a good job you're screwed.

Yes. All you can do for these albums is to split a very marginal difference down the center when trying to reduce stridency... and also improve bass EQ to restore that which has been summarily attenuated below about 450 Hz (or below 100 Hz, depending on music genre). Strangely, some of the least attenuated CDs in the bass region are older "rock" and "AOR" album standards.

 

There are definite limits to what you can accomplish, depending on what's been done to the music in mastering (and mixing, too). 

 

However, about 95% of the CDs that I've unmastered respond strongly and positively to the techniques mentioned in this thread. 

 

That's a pretty good deal.

 

Chris

[DrWho]

 

On 11/5/2016 at 10:08 AM, Chris A said:

They were what I'd call lo-fi. 

Call it what you want, but they sound better mastered the way they are. The problem is you have a playback system that doesn't sound good with that kind of source material. Call it accuracy if you will, but I simply call it "sounds bad". The interesting thing is that source material sounds great on other systems. So where you're slaving away trying to undo your system, I'm blissfully listening to a very non-fatiguing experience. How could that be?!? Maybe there's more to sound quality than the few metrics you're so focused on.

[Chris A]

 

3 minutes ago, DrWho said:

Call it what you want, but they sound better mastered the way they are. The problem is you have a playback system that doesn't sound good with that kind of source material. Call it accuracy if you will, but I simply call it "sounds bad". The interesting thing is that source material sounds great on other systems. So where you're slaving away trying to undo your system, I'm blissfully listening to a very non-fatiguing experience. How could that be?!? Maybe there's more to sound quality than the few metrics you're so focused on.

 

Mike: start another thread and talk about this subject.  I'm sure that you'll enjoy the freedom to move that thread in any direction that you wish.

[DrWho]

On 11/5/2016 at 10:23 AM, mikebse2a3 said:

It is painfully obvious that they were made to sound good in cars and on boom boxes

So why does it sound good enough in the car, but bad enough at home? Ever stop to consider the reasons behind that apparent paradox?

 

While there may not be standards, there are certainly dominant trends - which if understood should have a large influence in how we setup our playback systems.

 

If you're blaming 95% of the source material, then I simply need to point out that your system needs fixing.