Jump to content

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


Chris A

Recommended Posts

Chris, do you suppose there might be somebody out there watching this thread who uses your remastering technique and provides the music for stream or purchase? If there isn't, I think there should be. I've seen it mentioned many times that this thread is over most folk's heads. Why should every individual have to go into their own version of every song, when an expert like you could process the song once and provide that remastered version to all.

 

Frankly, I could see a team of people with your training remastering everything.

 

As much as this is a business opportunity, that is not why I am bringing this up. I just think it's a collective waste of time and resources for 50 million people to make their own remastered copy of 1 song. Why not just have that song sitting out there? Remastering is not ground breaking technology, but your hypothesis is.

That makes utmost sense. Therefore, it is probably illegal.

We don't need no stinkin' DRM.

  • Like 3
Link to comment
Share on other sites

Wow.  Thanks for the kind words MG and especially the heads up on the difficulty level.  I'm not always aware when I've taken a dive into the abyss and lost everyone.

 

I've looked on the Web for other sites/BBs doing the same thing (rebalancing music tracks using Audacity to correct mastering issues) but to date, I've come up empty handed.  You'd think that the DR Database site (www.dr.loudness-war.info) would be really hot on actually correcting their music files, but I've found nothing thus far.  The correction XML files are so small that their database site for DR ratings would actually hold the file, you'd think. 

 

Perhaps it's the relatively recent upgrades to Audacity bringing its capabilities up to a reasonable level for purpose that's the issue.  Perhaps it's just "starting jitters" or initial difficulties and need for a little starting knowledge that has dissuaded such a following from beginning.  I don't know.  It's certainly a public service in my estimation. 

 

I know that I've felt real catharsis while finally getting to correct the music that I've listened to since--well--all my life.  You can go back at least to 1966 for the first 45 RPM records that I bought as a kid.  Wherever I've looked for issues at tracks that have always been difficult to listen to or have had audible issues, like lack of bass but also lack of highs above 10 kHz and way too much SPL in the 1-4 kHz band making the music extremely strident,  in every case I've always found what the issues are and 99% of the time, I've been able to correct them.  This is a significant finding and outcome.

 

As you know, I've already expressed a preference that if a separate site would be set up for a database to hold the files and descriptions that I'd really like to have a freeware model--free XML equalization file downloads, and encourage others to submit their updates or variations, too, along with perhaps a word or two descriptions.  There may be real issues with a free contributor model--in that "a little knowledge can be dangerous", and with site security due to malicious stuff that internet criminals like to take advantage of. 

 

The current issues that I see are:

 

1) Where to host a site and the site software applications to host-such as a Wiki backbone (I can talk to Iain [seti] and perhaps others about this vis-à-vis his Retro Vintage Modern Hi-Fi site experience).

2) The costs associated with keeping such a site going

3) The time required to keep the site going and cleaned up due to mischief, over and above generating the XML EQ files to populate the database themselves

4) How to advertise and grow the user base, which seems to total about 10 people thus far... :huh:

 

Assuming this goes anywhere--a big assumption--I think that using Audacity at least for the time being is the right tool to continue using.  Everyone can download Audacity freely for the major operating systems and use the XML equalization files to rebalance their tracks.  There may be a time when it makes sense to include other editing tools, but right now, Audacity works extremely well for two channel files.  Multitrack music editing is a bigger problem I've found. Audacity will do multichannel but not nearly as easily - it's a lot more unwieldy for multichannel.

 

Humm...a few questions to ponder and decisions to make.

 

Chris

Edited by Chris A
Link to comment
Share on other sites

Those are good questions. If you could get one of the decision makers from one of the big online media companies to lend an ear, perhaps you could have a blind test done. If they could demonstrate that their mix is better than a lossless, they could use that as a selling point. Let's face it, a streaming company is a streaming company. The only difference is the amount of compression or the compression method.

Link to comment
Share on other sites

  • 1 month later...

This is what I managed to pull together in terms of an external EQ remastering site before I got sidetracked:

 

https://sites.google.com/site/dbremaster/

 

Note there are no new EQ files posted there presently--just a proof of concept based on the Audacity EQ XML files already posted in various places here.

 

The next step (yet to be done) is to create custom YouTube tutorials on the following three topics:

 

1) Introduction, basic overview of correctable issues to audio tracks, and a tailored overview using Audacity to correct them

 

2) The detailed step-by-step process of using Audacity to correct using the EQ XML files in the database

 

3) An advanced tutorial on how to create your own equalization curves to correct music tracks not currently found in the database, and how to submit for posting your own EQ XML files at The Audio Remastering database site.

 

The current Audio Remastering Database is functional, and collects all the Audacity EQ XML files that have been posted to date.

 

Any comments or suggestions (within reason) are solicited.

 

Regards,

 

Chris A

Edited by Chris A
  • Like 1
Link to comment
Share on other sites

I do believe that the remastering site could eventually become a significant contributor to the audiophile community as a greater percentage of the population realize that they can easily and significantly improve their dynamically limited and otherwise poorly mastered stereo recordings.  I've found that I don't like listening to my old un-remastered tracks now, almost always preferring to play what I've already remastered instead. 

 

I also have one emerging conviction about "digital vs. analog" that has become readily apparent:

 

It's not about the stereo format...i.e., analog or digital, bit rate or bit depth...nearly as much as it is about what has originally been done to the tracks to make them louder (we're talking about music all the way back to the 1950s or even before).

 

Fix the tracks using the remastering techniques found here to recapture fidelity...and those tired old arguments largely vanish.

 

 

Chris

Link to comment
Share on other sites

Nice site, Chris.

 

My main interest in the process is to restore missing bass. EQ works if the bass is there but at too low a level, but can't help if there is no low bass at all. I've been experimenting with a process that synthesizes bass an octave below any existing bass. This departs from the aim of restoring what was missing, by adding stuff that wasn't originally there. It's mainly useful as a party trick if you have subs or a car install that goes down to 20 Hz or below. There is a thread over on DIY Mobile Audio, http://www.diymobileaudio.com/forum/technical-advanced-car-audio-discussion/83564-bass-how-low-can-you-go.html that started out discussing bass-heavy tracks and morphed into techniques for creating bass if it wasn't there. Start at post 153 and read down to post 172 (mine).

If there's interest, I'll write up a step-by-step of the whole procedure from the beginning.

Link to comment
Share on other sites

Guest srobak

Well this has been a very interesting, informative and revealing thread to read.  I am going to have to revisit it and take a more keen interest in things when I have more time.

 

 

But one post in particular I can add to & attest as to why Quiet's observation is the case: 

 

This CD right here....one of the best sounding pop albums I've got, no joke. Overall 13 DR rating with a couple tracks above that.

 

Sounds friggin' amazing when dialed in correctly. The Pat Benatar songs will clear all the headroom your system has tucked away. They positively slam.

 

418P7QFNB4L.jpg

 

 

Courtesy of an awesome documentary and the subsequent personal research it sparked for me - both of the original albums these came off of were recorded at Sound City in LA, back in the day.  For those who have not see the documentary Dave Grohl put together regarding it - Sound City had one of the best sounding and biggest sounding drum rooms in the industry, and all material was tracked to 24 channels of analog tape through a mated set of hand-built Neve consoles - models 8078 and 8028 - which were built in the 70's and cost over $70,000 at the time.  The room also served very well for the great sound when the bands did live tracking.  These albums - and many others from that studio's early days are regarded as some of the best sounding rock/pop albums ever recorded - even by today's truly critical listener's standards. Some of those include: Buckingham-Nicks, Caribou, Fleetwood Mac, Terrapin Station, Rumors, You Can Tune A Piano But You Can't Tuna Fish, Why Can't We Be Friends, Double Vision, Damn The Torpedos, Crimes Of Passion, Working Class Dog - and the list goes on. 

 

For those who haven't seen the documentary (Sound City) - I highly recommend doing so.  It is hands down the best documentary regarding the recording industry you will see with more than 3 dozen top artists interviewed and a ton of history covered - and you would be doing both yourself and your ears a major disservice if you didn't watch it on BR so you can get full bitrate & surround. 2 hours long, exquisitely mixed (on the Neve, mind you) and with some awesome bonus material in performances by artists "coming back" to the board to lay new tracks through it.   Do yourself a serious favor and get this. 

-scr

Link to comment
Share on other sites

  • 1 month later...

Okay, on the Active bi-amping/tri-amping thread, I mentioned that I've learned some new things about using Audacity to remaster tracks, so this represents that response to that promise.

 

Let's recap  the basics of what we've been doing using Audacity:

 

1) We've been using the cumulative SPL  (or loudness) vs. frequency curve for each track to determine how much to boost on the various portions of the frequency curve. 

 

Basically that's it.  Some other points to recap:

 

2) Most music tracks have significantly attenuated bass frequencies, often below 100 Hz, but sometimes the attenuation starts as high as 900 Hz.  This is done to make the tracks "louder" or higher average SPL across the track, because those low frequencies, even though they are at higher amplitudes than the higher frequencies, they don't sound to our ears to be louder.  This is related to the way that we hear: basically we hear logarithmically...in decibels of loudness, and in frequency logarithmically...12 tones to an octave, to be exact. 

 

3) We also like to hear our music at a decreasing straight 5.5 dB/octave loudness vs. increasing frequency--on average (an empirical relationship discovered by yours truly).  That's about 18-20 dB per decade, which is an important number that we'll be needing to use very soon. 

 

4) Most of the time, a logarithmically decreasing 5.5 dB/octave slope curve works pretty well, especially for tracks that aren't terribly long and changing in instrumentation (orchestration) and have relatively full orchestration - lots of frequencies across the 20-20,000 Hz audible band.  Rock and other popular genre music tracks usually fit these conditions.  Classical and folk/world music, not so much.  Piano and other solo instrumentation - not so much. There are other genres that also have issues.

 

It would be nice to have yet another yardstick to measure our music tracks by.  As it turns out, there is at least one other technique to use.  Within Audacity, there are several views available from the drop down menu found in the upper left of the graph area of the view.  One of those views is "Spectrogram log(f)":

 

AudacitySpectrumMenu.png

 

When you open this view, you should see something like this:

post-26262-0-81300000-1439079901_thumb.g

 

From this view, a couple of new tools are available for us to use. 

 

a) if the music file was recorded using analog equipment, there will be a noise floor that is visible as a background green or blue color (depending on whether the graph is selected for editing or not). We can use this background color to gauge whether we are over- or under-boosting certain areas of the track.  Even if the track was recorded digitally, the background levels can still be used to our advantage to set the "unmastering" EQ curve.

 

'b) The spectrum of the notes themselves are visible due to the colors used in the plot.  The strongest frequencies (tones of the instrument) are slightly lavender or purple, while the higher order harmonics of the notes at higher frequencies turn red as the SPL decreases on those harmonics, then green.  The overall balance of colors should be spread across all the frequencies and harmonics, if we set the Spectrogram log(f) view to compensate for the 5.5 dB/octave decrease in amplitude.  Fortunately, the preference dialog at the bottom of the Edit menu drop-down menu has one sub-view for spectrograms that allows the user to set that factor "Frequency Gain(db/decade)" - which should be set to a value of 18 to 20.  This will level the spectrogram log(f) colors to about 5.5 dB/octave, mentioned above.

post-26262-0-48760000-1439124875_thumb.g

 

If the colors are skewed lavender/purple toward a certain band, we can now adjust our EQ curve to decrease the boost in those general areas.  Additionally, if certain frequencies are green or yellow background color across the track, it is clear that those frequencies need to be boosted, assuming that the instrumentation in the music track has those frequencies represented at all.  While there is some judgment that is required, this new view provides a very powerful and sensitive tool to reconstruct the original recorded track SPL levels that the recording engineer captured.

 

Using this technique, difficult music tracks to unmaster, like pianos, classical music orchestras, and tracks without deep bass below 70 Hz (like solo guitars and cellos) can be easily adjusted to arrive at a unmastering EQ curve producing an updated music track that is natural and lifelike sounding without having to resort to trial-and-error EQ processes. 

 

Chris

Edited by Chris A
  • Like 1
Link to comment
Share on other sites

So the punchline to the above use of Spectrogram log(f) view: 

 

We have a tool that can help us see both the highest loudness levels--our original technique of using the "Plot Spectrum" view--and now using the "Spectrogram log(f)" view to see the threshold noise and individual notes and their harmonics played...

 

...to help us fine tune the unmastering EQ curve for each music track.  This second technique is much more sensitive than the first technique, but also suffers from "saturation" that the first technique helps us avoid.

 

So the process I now use is to first construct an EQ curve using the "Plot Spectrum" view to rough out the starting curve, then use the second technique of "Spectrogram log(f)" view to fine tune, especially the highest and lowest frequencies in the EQ curve. 

 

This sounds like a lot of effort, but for the type of music that it typically applies to--classical and solo acoustic instruments and voices--there is one factor that I haven't stated that makes it much easier:

 

Most of these type of recordings use essentially the same mastering EQ curve for the entire album (or sometimes series of albums).  This means that once you get the EQ curve set up properly, it can be reused almost without modification for all tracks on the album or albums. 

 

Since about 50% of my music library is of the type that I've been talking about (classical, new age, folk, jazz, and some fusion), this process works very well and achieves much better results for equal time spent.  Additionally, for popular music (rock, pop, alt, etc.) the use of this technique has enabled me to correct some very difficult recordings that had poorer results using only the first technique, i.e., a decreasing straight-line "Plot Spectrum" curve.

 

Chris

 

post-26262-0-28360000-1439124320_thumb.g

Edited by Chris A
  • Like 1
Link to comment
Share on other sites

Chris,

 

Thanks for the information on the spectrograms.  I had changed mine to black and white and couldn't tell what I was looking at. All OCD'ers will now revisit their past work.  :mellow:

 

Along with having to re-read this thread a hundred more times, I have just gotten a system going that actually resolves detail and I am at saturation level with all the change.

 

Being a linux an alternative software guy, there are some tools I am still looking for, a DR measurement tool being one of them. I have issues compromising core beliefs. Although for the sake of audio enjoyment I might can manage some dual boot.

 

 

Thanks for your time, and for figuring out how to explain everything in such a clear and concise manner...

  • Like 1
Link to comment
Share on other sites

By the way, although I have alluded to the fact, I really haven't stated the following clearly:

 

Classical music recordings on CD (stereo) and vinyl suffer just as much, if not more, from spectral shaping of the music during the mastering process.  This seems to be something that is never acknowledged openly, but is fairly consistent from classical recording to recording.  In fact, it seems that for like-genres of music and certain record labels, there seems to be a typical EQ curve used.  For instance, on the Archiv label, here is a common reverse EQ curve that is used for all discs in a 5-disc box set of Vivaldi concertos:

post-26262-0-00760000-1439392208_thumb.g

 

While this may not seem to be a particularly "wild" mastering curve that is applied to the discs, when you look closely at the relative dB levels used, you might be surprised how severe the spectral shaping actually is. I've found that virtually all 2-channel classical recordings have similar levels of spectral shaping.  This has been one of the greatest surprises and disappointments of unmastering discs.  If you think that you've got anything close to an actual performance sound while playing back virtually any two-channel classical disc or record, think again... :emotion-41:

 

Chris 

Edited by Chris A
Link to comment
Share on other sites

This is great stuff, Chris!

You might do another video showing (and talking) us through the Spectrogram view method.   Visual aids are very helpful with this sort of stuff.  Thank you for all your work and willingness to share!  I would love to have played with this more to date but it seems that I can barely keep up with getting new CD's ripped to FLAC considering the rate I am buying music in recent years.  "Too much music" isn't a bad problem to have but I should use your process to work on some of my favorites to hear what I can do. Experimenting with the EQ files you have shared proves to me that this effort is worthwhile! 

Link to comment
Share on other sites

You might do another video showing (and talking) us through the Spectrogram view method.

 

Thanks for the kind words, above.  I've not actually done a YouTube video to date, so I'll have to come up to speed on a new-to-me set of tools.  I believe that I have everything to produce good video(s), however it will require some quiet time without the clinking and bustling of everyday life around the "studio". 

 

There'll be a slight delay,.

 

Chris

Link to comment
Share on other sites

  • 2 weeks later...
For instance, on the Archiv label, here is a common reverse EQ curve that is used for all discs in a 5-disc box set of Vivaldi concertos: Vivaldi EQ curve.GIF

 

I'm confused.  Does a "common reverse EQ curve that is used for all discs in a 5-disc box of Vivaldi concertos" mean that they applied the curve you show, and that's what is on the disc, OR does it mean that the curve you showed in your post is one we would need to apply to flatten out the response? :)

Link to comment
Share on other sites

I always show the curve that needs to be applied within Audacity to flatten the FR back to something approximating what the musicians would have done by themselves if they were all playing in one venue at the same time - a music ensemble. 

 

In the case that you reference, the mastering engineer used one "mastering curve" for all tracks on all 5 discs, so that once you find the correction curve for those tracks, you can reuse it over and over within Audacity (or the editing tool of your choice) to achieve close-to-real-life results to undo--to reverse--what the mastering engineers have done.

 

I call those curves "reverse EQ" curves because they are intended to reverse the off-nominal FR that the mastering engineers apparently feel compelled to do to the music to make it "better".  In fact, I've found that almost without fail, the mastering engineers (working in coordination with their clients) make it worse, sometimes much worse.  Unlistenable, in fact. 

 

I don't believe that these mastering EQ curves are typically used upstream during the mixing process: the musicians and the mixing engineers at the mixing consoles I believe try for something that is much more natural and neutral sounding - like the 5.5 dB straight line (log-log scale) averaged FFTs/track decreasing amplitude with frequency, discussed in the first post in this thread.  It's later, after the musicians and mixing engineers are finished that the mastering engineers start to work their "miracles" (a sarcastic statement). 

 

I've thought about why the mastering engineers do it.  Some reasons may be noble, to fix really bad problems in the final mixes, and really otherwise, for "commercialization" (i.e., a euphemism to justify really screwing up the music in the name of increasing potential sales).  For instance, most of the time it's to make the tracks "louder", i.e., devoid of dynamics by using multi-band compressors, or simply clipping the music peaks such that the average SPL levels are higher. Most of the damage is done at this point.  There are apparently many other reasons in between, including making the music sound "better" on typical-but-inferior loudspeakers and sound reproduction systems--hence the often quoted "nearfield monitoring practices" using Yamaha NS-10Ms, etc. 

 

Whatever the reason, when you play these store-bought tracks back on quality, well setup, and carefully EQed flat loudspeakers and driving electronics, they don't sound very good.  That's the impetus for this thread--to correct that situation.

 

Chris

Edited by Chris A
  • Like 1
Link to comment
Share on other sites

  • 1 month later...

Just some additions to the "lessons learned" pile:

 

1) I've found that after doing major corrections on a particular album, I have to go back later (usually the next day) and fine tune the EQ to make it sound brilliant again.  "Major corrections" include de-clipping the tracks using Clip Fix by more than 5 dB/track. i.e., the track basically looks like freshly mown grass before starting the unmastering process, and having to do major EQ corrections for each track, i.e., each track has to be corrected uniquely due to custom mastering EQ used for each track.  After all this per track, and after doing 12-20 tracks, one after another, I find that listening fatigue sets in and I have to give it a rest. 

 

What's interesting about this followed-up edit the next day is that it takes very small changes in the EQ curve - tilting the curve up or down at 20 kHz relative to the 20 Hz anchor point by as little as 2-3 dB across the entire 10+ octaves of audible bandwidth. This is amazing to me every time I do it.  I've found that I can make changes of as little as 1 dB at 20 kHz relative to 20 Hz and still clearly hear the change.  "Wow."

 

What this tells me is that the unmastering process using the above -5.5dB/octave vs. log frequency target response is usually right on the money. Once you get the tracks very close to this target response, it takes extremely small changes to have very large results - like a "sweet spot" appears in the track's target response and the ear knows when it is right.

 

2) I also find is that I can clearly hear the mix down errors that are usually far overcompensated (i.e., the highs are over-boosted across the entire track) by the mastering guys to make everything sound "crisp" all the time.  This includes dull sounding piano, percussion, and wind instrument tracks mixed with vocals that come in after a period of time.  At first, the track sounds dull, then overly bright as it plays further.  These are mixing issues.  If this effect is too objectionable, I've found that I've had to break the track up into sections by time segments (a neat trick to keep the amplitudes and pops of the joins in synch) and correct each section. This is very time consuming, so I avoid doing it at all costs.

 

3) After completing the unmastering for an album, all the tracks sound much, much better, especially at lower playback volume, due in part to the fact that the bass frequencies are restored, and also due to the fact that the high frequencies are balanced lower to match the mids and lows.  I notice that unmastered tracks can now be played in the background without it obliterating the conversation in-room, and that the louder you turn it up, the better it sounds, but in a gradual fashion - not at a set threshold below which the track doesn't sound good at all because you can't hear the bass. 

 

4) Because of the nature of the rebalancing of harmonics of the instruments relative to each other via the re-EQ processes, the removal of droning line noise (50/60 Hz, etc.), fixing the clipping of peaks (which introduces short bursts of high frequencies that were not in the original recordings, and that I find introduces a subconscious opaqueness and fatiguing response in the listener when they re-occur often during a music track), all the tracks sound much more transparent.  You can literally hear stuff that you couldn't hear before. 

 

You can also hear the acoustics of the venues in which the music was recorded much better, too. 

 

Enough for now...

 

Chris

Edited by Chris A
  • Like 1
Link to comment
Share on other sites

  • 1 month later...

Chris-

Do you have access to Klipsch Tapes Vols. I & II? I'd like to know your impressions of them.

I've read and enjoyed this entire thread and reread much of it. I use Audacity to rip vinyl to high resolution FLAC files. I might as well try to implement some of your remastering techniques at the same time. If only there were an edition of "Audacity for Dummies."

Before anyone jumps in to inform me that FLAC and high resolution are wasted on rips of vinyl, rest assured I know the limitations of vinyl, especially in terms of low frequencies and dynamic range. Nevertheless, digital storage is cheap. I'd rather preserve excellent vinyl recordings in the best possible format to allow flexibility going forward. Others may covet the storage space on hard drives, I do not.

I believe you tweaked a couple of Pono tracks. I'd love to be able to compare your remastered tracks to the original Pono versions.

Link to comment
Share on other sites

I haven't bought the Klipsch Tapes tracks yet.  When I do...  I have no doubt that they'll sound great without any unmastering, since they were never mastered to begin with, i.e., made to sound good for sound reproduction systems that aren't terribly hi-fi.

 

I completed the Pono masters about 3 months ago: I made them available this AM in a separate directory for your listening pleasure.

 

Good to hear that you're trying out Audacity. Flac certainly isn't wasted on vinyl rips.  You'll be looking for pop/tick filter for Audacity that works better than the one distributed with Audacity.  If I find a good one, I'll forward the link.

 

Chris

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...