Jump to content
The Klipsch Audio Community
Sign in to follow this  
Chris A

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

Recommended Posts

"There's an old adage in the prosound world where if you're applying the same EQ to every channel (on a mixer), then your system isn't calibrated properly. "

 

As Chris has pointed out (in different words), this is basically what a mastering engineer does when he messes with the two track mix. Perhaps the mixer and masterer should work more in concert (no pun intended), but then the one who has been doing mastering only will end up out of his high paying job.

 

Chris, I had been thinking you were getting too far out there with your demastering, but the more I have read through your concepts and WHY you do this, it makes more and more sense. Kudos!

 

Bruce

  • Like 1

Share this post


Link to post
Share on other sites

Bruce, your comment is precisely why I've been demastering in the recordings that I've mentioned here.  When you take the downmixed stereo tracks and put them through heavy global EQ and limiting/compression during mastering, and then say that you've "improved the sound", then something is either wrong with the mastering engineer's setup (specifically the loudspeakers they're using and/or room acoustics)...or they just don't care about what the finished product sounds like on a good system.  They want only the perceived maximal monetary returns...nevermind the hit on product quality.  Pushing around EQ in large doses to unbalance the tracks literally changes the timbre of all the instruments and voices.  Compression does the same thing, although the listening effect is somewhat different.

 

I believe that if you're going to do creative EQ, its better to do it before the mixer, and then "mastering" only consists of extremely small tweaks (i.e., less than 3 dB anywhere in the audible spectrum) and perhaps some extremely light limiting/compression (i.e., less than a dozen clipped waveforms per track), then leveling the tracks for the album, then ship off the tracks to the pressing plant...or simply release online. 

 

I believe in leaving the creativity to the musicians playing on the tracks themselves.

 

Chris

  • Like 2

Share this post


Link to post
Share on other sites

Thanks for the heads up. 

 

I don't know if you've seen this related presentation by Sean Olive, et al., but the three reference songs listed in it were three I found that didn't require any real demastering at all (extremely minor tweaks):

 

Jennifer Warnes, “Bird on a Wire”
Tracy Chapman, “Fast Car”
James Taylor, “That’s Why I’m Here"

 

My guess is that Olive or his colleagues have been using the same techniques as above to either screen reference tracks, or have actually been editing some tracks themselves. 

 

Chris

Share this post


Link to post
Share on other sites

So in the 2 1/4 years of demastering stereo music tracks, I recently ran across the first instance where it is clear that a popular or jazz music track with relatively full instrumentation was mixed so that the 1/f curve was not the best sounding. 

 

The CD where I found this track is Stacey Kent's The Boy Next Door CD.

The_Boy_Next_Door_(album).jpg

 

In this case, it was clear that her voice was mixed at a significantly higher level than the background instrumentation (double bass, hi-hat, ride cymbal), so that there was a "double hump" in the 1/f curve: one peak at ~100 Hz and another higher peak at 250 Hz.    The point of this is that once the mixed levels are locked into a recording, any effort to establish a 1/f cumulative curve profile will actually drive the result to sound worse than the original.  Of the 10,000 tracks that I've demastered to date, only a handful of tracks are clearly in this category for an entire music track/song.  The original CD track sounds better due to the balance of harmonics of each instrument in the mix, even though the 1/f cumulative spectrum is not present.

 

However, in this case, there was also a boosting of the highs above 2-3 kHz that maximized at ~10 kHz that overemphasized the consonants and sibilances of the singer, and the ride cymbal.  After a 3-6 dB de-emphasis of those frequencies to approximate a 1/f curve at higher frequencies, the naturalness of the recording environment emerged.

 

All other tracks that have distorted cumulative loudness curves (the "purple mountain" curve within Audacity from the "Plot Spectrum" command) clearly have EQ applied after the mixdown tracks are created, thus unbalancing not only the overall levels of loudness in each spectral band, but more importantly the relative harmonics of each voice. 

 

This isn't the case in tracks mixed such that the voices are combined but no EQ is used after mixdown to stereo. 

 

Chris

Share this post


Link to post
Share on other sites

I just added two demastering tutorials in pdf format to the first posting in this thread.  Another tutorial describing more advanced topics will follow once it's complete.

 

Chris

Share this post


Link to post
Share on other sites

An interesting 1996 JAES article by Peter John Chapman (B&O) that compiled music track spectra and did some music track spectrum statistics by genre of music.  The averaged spectrum plots by genre are found at the end of article:

 

http://www.mountain-environment.com/AES_paper_1996_4277.pdf

 

This paper really describes the degree of mastering EQ used on the stereo production tracks, averaged by music genre type, and also shows the 1996 after-mastering averages for Dynamic Range Database values (i.e., crest factor or peak-to-average values) on page 7. 

 

The purpose of posting a link to this article is so that those here that, for one reason or another, have not yet demastered their tracks, instead using amplifiers, modified crossover networks and drivers, and upstream EQ in their preamps or source disc players/phonograph preamps to approximate a (single) correcting curve.  Those individuals can use whatever EQ capabilities they have by music genre to approximate a demastered music track spectrum using correcting EQ to approximate a 1/f curve with -16 dB/decade (-5.5 dB/octave) slope.  Some music genres, as you will see, require a fair amount of correcting EQ (on average) to move back to a recorded 1/f curve (before mastering EQ was applied). 

 

Chris

  • Like 1

Share this post


Link to post
Share on other sites

A plot of the average crest factors (i.e., dynamic range) by genre found in the 1996 Chapman JAES article above:

5a259bc21eaa0_AveDynamicRangebyGenre.PNG.02f4840f4f8aee35044c98bfa9530bad.PNG

Share this post


Link to post
Share on other sites

I've found that the Chapman paper linked above describes almost exactly the type of mastering EQ used by genre that I've seen on average.  This will give you a clue on what EQ to try when you play a stereo recording (such as a phonograph record or CD) without actually looking at the tracks themselves using an audio editor like Audacity. 

 

Below you will see one spectrum plot for the "heavy" group that is most applicable to rock and its derivatives, overlaid with a IEC 268-1 simulated signal (solid black) and a notional 1/f (-5.5 dB/octave) target demastering curve in red.  From this one chart, you can see what frequencies and attenuation/boost required to achieve a demastered spectrum based on the average "heavy" mastering EQ profile.

 

5a2818c695b50_HeavySpectrumvsEC268-1fromChapman.thumb.PNG.e8151351d85f4ba697f4d60632400c91.PNG

 

While that doesn't mean that pre-EQing your loudspeakers, either through upstream EQ, high output impedance amplifiers, or use of non-flat passive crossover networks after the amplifier, will work for each disc or even each track on a disc, it will give you a much better success rate on listening to your music with much more neutral EQ--much more like the stereo mixdown tracks before mastering EQ was applied.

 

Chris

Share this post


Link to post
Share on other sites
1 hour ago, Chris A said:

...pre-EQing your loudspeakers...

 

1 hour ago, Chris A said:

...will give you a much better success rate on listening to your music with much more neutral EQ...

You noted the Cello Palette EQ curve in the past when asked about dealing with internet source streaming music. Is that what you are recommending again, or are there other similar options to consider.

Share this post


Link to post
Share on other sites

If you're using a Cello palette, then you can read the difference between the "Heavy" curve (in the example above) and the 1/f demastering curve at the various frequencies corresponding to the Cello palette center frequencies, and set the Cello palette filters accordingly. 

 

If your source music has a mastering EQ curve that's similar to the averaged curve shown in the example, you will hear your music much closer to as-recorded conditions.  If it doesn't sound much better, then you can experiment with the Cello palette settings until it does.  At that point, I'd recommend writing down the settings on something that you can keep associated with the disc or the ripped music tracks so that you can simply set the filters there again. 

 

Then again, if you use Audacity to correct the tracks locally, you can see what each track needs (in terms of correcting EQ), then save it.  Then you don't have to set the Cello palette filters again...:wink:

 

You can also save the Cello palette filter settings as canned "demastering" configurations, then you can recall them for the streaming music titles that you choose to use them on.  Then everything stays legit and above board.

 

Chris

Share this post


Link to post
Share on other sites

@DizRotus This is from the "Right this Minute" thread...

 

2 hours ago, DizRotus said:

I suspect Chesky’s recordings are not casualties of the loudness wars.  What say you Chris A?

 

I've found that the dozen or so Chesky albums that I own are interesting...from a demastering point of view.  If you believe that these albums are "reference" and that they are your "go to" discs for evaluating , you may want to stop reading here.

 

What I've found is that, while the Chesky albums are not really mastered for loudness--just a reasonable amount of compression/limiting is used to control the average levels to something like -14 dBFS to perhaps -19 dBFS, they were mastered for a particular "sound" that clearly differs from that which was recorded in the recording studio.  What was done in mastering was that significant levels of mastering EQ was used to change the overall timbre of the music.  In the particular case of the Rebecca Pidgeon discs (particularly The Raven) the mastering EQ seems to have been used to make her voice sound "younger" or higher pitched.  [This is not unprecedented: I have found the same techniques used on Frank Sinatra albums from the 1940s-1960s.]

 

In general, the EQ mastering made to Chesky discs changed the timbre or overall sound of the recordings in pleasing ways.  When you hear the demastered Rebecca Pidgeon albums, you will be likely be torn between the two versions, mastered and demastered, and might want to retain both versions for comparison purposes.  Personally, I feel that the original sound (i.e., before creative mastering EQ was applied) is my preferred target sound.  Others might like the mastered versions. 

 

Chris

 

MI0000649855.jpg

  • Thanks 1

Share this post


Link to post
Share on other sites

@boxerjake from the LaScala bass? thread:

 

1 hour ago, boxerjake said:

...The principle of operation is an assumption that very low frequencies in the original music have been lost or removed during recording and CD production. The PCA3 generates subharmonics at half the input frequency, with a sampling range from 50 - 100 Hz in the "Digital Restoration" mode...this eliminates boominess that would occur if subharmonics were added from original frequencies at 100 - 500 Hz, and eliminates adding subharmonics to voices that are sent monophonically...any signal that is monophonic will not have subharmonics added to it. The separate crossover can be used to send <90 Hz frequencies (line level inputs and outputs) to a pair of powered subwoofers (one each for left and right channels), and >90 Hz (line level) to the main front left/right amplifiers (pre-outs and main-ins on a receiver can be used).

 

I've found that it's much more effective to demaster the music itself to restore the bass below 100 Hz since the original bass frequencies that have been attenuated are still there and at high signal/noise levels.  When an inverse mastering EQ curve is employed to restore a1/f cumulative spectrum curve (-16 dB/decade) to the tracks, the bass is restored, i.e., no sub-harmonic "synthesis" is required, and the resulting sound is much more realistic than using bass synthesis.

 

I've also found that many tracks possess bass noise that exists at specular frequencies corresponding to droning line noise (50/60 Hz) and HVAC noise (typically found at 7, 9, 13, 17, 19, 23, 27, 32, 37, and 41 Hz).  Additionally, there are certain recordings where the second harmonic of the line noise (100/120 Hz) and sometime third harmonic (150/180 Hz) is present and visible in the spectrograms.  Performing notch filtering at these specular frequencies on a case-by-case basis significantly cleans up the recording and results in much more transparent and clear midrange (via significant reduction in higher frequency modulation distortion with these problem bass noise frequencies). 

 

Note that this demastering for bass restoration requires the use of spectrograms and cumulative SPL density spectra plots to visually identify the incoming issues with the tracks, and confirm the effect of the iterated demastering inverse filters--along with careful listening using a calibrated stereo setup in-room to verify the resulting demastered tracks.  The results speak for themselves.  Also note: once the track is demastered, no further editing or work on the resulting tracks are required...you're done having to deal with the issues--unlike the use of plugins at playback time.

 

Chris

Share this post


Link to post
Share on other sites
On ‎2‎/‎11‎/‎2018 at 2:07 PM, Chris A said:

 

 

 

Note that this demastering for bass restoration requires the use of spectrograms and cumulative SPL density spectra plots to visually identify the incoming issues with the tracks, and confirm the effect of the iterated demastering inverse filters--along with careful listening using a calibrated stereo setup in-room to verify the resulting demastered tracks.  The results speak for themselves.  Also note: once the track is demastered, no further editing or work on the resulting tracks are required...you're done having to deal with the issues--unlike the use of plugins at playback time.

 

Chris

 

 

LOL... Ya , that sounds like something any forum member can just whip up in their spare time !

Share this post


Link to post
Share on other sites

 

5 minutes ago, boxerjake said:

LOL... Ya , that sounds like something any forum member can just whip up in their spare time !

 

Yes, they can...:D

 

The application software is free.  So are the tutorials...:emotion-21:  The results speak for themselves. 

 

If you would like to hear demastered tracks, you need only mention your desire to hear a few tracks. 

 

Chris

Share this post


Link to post
Share on other sites

While I am no where near as scientific as Chris on his approach to using Audacity I can say the results even just fiddling around with the program can be quite spectacular. Had some Jethro Tull music this week and some of it never would improve and others were transformed from nice to remarkable.  Not sure the wife has noticed the same song played back part way time and time again as I try to get things as good as possible with my novice skill level. Some of the older Spirit music like Fresh Garbage had percussion I was completely unaware of hidden in the commercial version.

Share this post


Link to post
Share on other sites

From another thread (links to that thread are in the headers to the quotes):

 

15 hours ago, robert_kc said:

(I have an early CD that includes Barber’s “Adagio for Strings” that had me turning down the volume control, because it sounded harsh.  My modern hi-res recording – no problem.)   As I said earlier, operatic soprano can also be difficult to reproduce without causing the listener to cringe.  

This is precisely why I started demastering efforts.  It wasn't classical recordings that I started with, but rather popular genres (rock and its derivatives, pop, jazz, etc.) that provided the incentive to want to salvage the album tracks--much like people do when using a good EQ unit in their system at playback time. 

 

But rather than having to use an EQ unit each time the album was played and resorting to notes attached to each track for the setup of the EQ unit, demastering solves that problem once and for all by fixing the problem at its source, and the tracks become trivial to play back and sound incredible without all the fuss of setting up and using an EQ unit each time.  Using demastering techniques allows you to home in on the problems immediately and have a visual augment to your ears-only to find the issues and correct them rapidly, without overcorrecting or choosing the wrong frequency bands to work on.  It's in this area of using the visual representation of the music tracks that the real power of the technique comes into its own.  This is where it becomes possible to actually see what the mastering guys did to the tracks, and to be able to reverse it. The ears are used for fine tuning after the larger scale corrections are roughed out.

 

One of the worst classical tracks that I've demastered was Adagio for Strings by Barber.  Apparently, it's fashionable to dramatically boost the strings in the 800-20000 Hz band while attenuating 450 Hz downward in ramping style.  As-is, I found that CD stereo recordings of this composition are almost impossible to listen to when turned up to anything approaching concert levels--like running your fingernails over a chalk board.  Fortunately, the damage can be reversed during application of demastering EQ but there are some special rules of the spectrograms and cumulative PSD views that apply to string-only orchestras, i.e., the overall envelope of the cum PSD view needs to follow the background noise levels even more closely to find the correct demastering EQ curves. 

 

Bernstein &  LA Philharmonic.PNG

 

I've found that there are pleasant surprises once the recordings are demastered--some recordings that you might have thought were the victims of poor recording equipment performance are found to be victims more of bad mastering EQ instead--and their inherent recording quality finally comes through once that has been removed.  I've opened up a very large library of recordings from as early as the late 1950s (when stereo or three-channel recordings began to become the norm) that I now listen to regularly with hi-fi performance. 

 

Some of the quality of the recordings from that early period--starting in the 1960s--using large format tape recorders, higher tape speeds and noise reduction can compete with some of the best recordings made today in stereo--albeit with slightly more noticeable noise floors, but the frequency response and other measures of fidelity are otherwise superb.  RCA Victor Living Stereo and Mercury Living Presence are but two examples of these recordings, most of which have been reissued in three-channel SACD format. These are amazing recordings in general.

 

I've found that Nonesuch's classical stereo recordings typically use much worse mastering EQ in a "zig-zag" fashion that proves more time consuming to unravel--like someone took the 1/3 octave sliders on a mixing console and alternately pushed them up and down from their 0 dB positions by 3-6 dB each. I've never liked to listen to Nonesuch albums in the past--back to the 1960s.  Now I know why.   This company has been doing that for a long time indeed, basically unchanged in their M.O.

 

15 hours ago, robert_kc said:

Garbage-in/garage-out – poor quality recordings, and low-resolution digital deliverables, can make music sound harsh, and massed strings are an “acid test”

I will agree that there are some lower quality recordings that sound fuzzy and low-fi, but the harsh sound is usually mastering EQ.  If you find yourself turning an album off prematurely due to harsh sounds, that album is really a candidate for the technique discussed in this thread.  Massed strings sounding harsh are most definitely caused by boosted EQ in the 800-5000 Hz band, not due to poor recording hardware. 

 

Chris

Share this post


Link to post
Share on other sites

Here’s a CD that includes a recording of Samuel Barber’s “Adagio for Strings” that I think has poor audio quality:

 

41T1DP0QXRL.jpg

 

This is a DDD recording from 1982.  I don’t think the harshness has to do with frequency imbalance.  I’m not an expert – my suspicion is that it has more to do with early digitization technology.

 

Here’s a hi-res FLAC download of a 2011 recording that I think has better audio quality:


51Ijo3nqD1L._SX450_.jpg

 

This recording does not have attenuated high frequencies, and yet doesn’t sound harsh (to my ears).

 

Bottom line - I don’t think the difference in audio quality between these two recordings is primarily due to frequency balance.   (I imagine that LP aficionados have AAA LPs that don’t sound harsh, and have vivid presentation of the massed violins.)

 

If Bernstein’s performance with the LA Philharmonic had been recorded on analog tape, then perhaps the recording could be re-digitized and remastered from the original analog tapes and delivered as a hi-res digital deliverable. 

 

I don’t doubt that attenuating the high frequencies of the 1982 recording by Bernstein would make this recording sound less harsh, but I would not characterize this as “demastering”. (Perhaps I’m just hung up on semantics.)

 

Am I missing something?

Share this post


Link to post
Share on other sites
5 hours ago, robert_kc said:

This is a DDD recording from 1982.  I don’t think the harshness has to do with frequency imbalance.  I’m not an expert – my suspicion is that it has more to do with early digitization technology.

I just demastered that particular track.  Assuming that your CD hasn't been remastered to be louder than the version that I used...and also assuming that you've got your loudspeakers set to flat response (something that you haven't shown me that you've accomplished, and the success of this EQ curve hinges strongly on this point)...try this demastering EQ curve with that particular track (import the following XML equalization file into Audacity and apply to the track): 

Adagio for String - Bernstein, LA Phil.XML

 

I might point out that this curve has over 21 dB of overall EQ correction.  If you can't hear that...I'm not sure that this technique will work for you at all.

 

Chris

Adagio for Strings EQ curve.JPG

Share this post


Link to post
Share on other sites

I know it probably isn't proper but in some cases I would like to apply a dynamic range expander.  Are there any expander software plugin's to work with Audacity which might work as well as a DBX  box? (I have a 3BX but would like to explore a software approach)

Share this post


Link to post
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.

Sign in to follow this  

×
×
  • Create New...