Jump to content

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


Chris A

Recommended Posts

[Edit 10 June 2017: Tutorials for demastering your music tracks using the method described in this thread follow:

Demastering Part 1 (What and Why).pdf

Demastering Part 2 (How To).pdf

 

Part 3 (Advanced Topics) will follow shortly.]

 

Recently I bought a new/old stock (NOS) CD from Amazon, originally recorded on analog tape in 1976, released on vinyl then much later re-released on CD. My particular CD dates from 1990, just before the wide distribution and use in 1991 of multi-band music compressors, and with it the Loudness War on popular music.

 

When this particular CD arrived it was basically unlistenable, i.e., it sounded exceptionally strident and devoid of bass, but retained its music dynamics.  At this point, I decided to investigate the reasons why it sounded so bad.  My tool of choice: the freeware tool Audacity.

 

What is the Issue?

 

Spectrum_uncorrected.GIF

 

When looking at the frequency spectrum averaged across most of this music track, it's easy to see a few characteristics:

  1. The decreasing slope with frequency across its full spectrum (more on the reason for this later)
  2. Frequency spikes of narrow width across certain areas of the spectrum
  3. A slight "hump" or rounded spectrum from 1000 Hz to about 13 kHz
  4. A steep roll-off of low frequencies below about 80 Hz

Some immediate questions arise:

  • Is the roll-off toward higher frequencies normal (i.e., is it there in the original master recording before any changes are made to it)?
  • Are the frequency spikes normal?
  • Is the frequency hump from 1-10kHz normal?
  • Is the much steeper roll-off of low frequencies below 80 Hz normal?

I found answers to these questions:

  • Roll-off of frequencies from low to high is normal, since the frequencies themselves double in their inherent energy for each increase of an octave - i.e., a -5.5 dB/octave is present in all typical recordings (note: don't confuse this effect with the Fletcher-Munson curves of equal perceived loudness).   In fact, any departures in the averaged frequency spectrum from this linearly decreasing amplitude behavior with logarithmic frequency should signal the need for further investigation.
  • The frequency spikes typically correspond to certain types of musical instruments that do not change frequency each time they are played (i.e., piano, percussion, and especially electronic instruments, etc.).  These spikes are almost always generated by the musicians themselves, not the recording/mixing/mastering processes.
  • The frequency hump from 1-10 kHz isn't really typical of most live music.  There is typically a straight line of deceasing slope tendency for averaged unamplified/unmixed music if the musicians onstage playing together get to adjust their loudness of the various music parts (assuming a multiplicity of instruments including percussion/drums, double bass, and treble instrumentation, like wind and string instruments of the band or orchestra, and voices.  (This is probably the most useful observation that I found.)
  • The roll-off in bass below 80 Hz isn't normal or desirable, unless perhaps you don't actually have to listen to the reproduced music, but only to get the music impressed onto phonograph records or CDs without having to decrease its overall loudness/gain to accommodate the very large bass/kick drum transients that are actually there in real, live music.

To check the last statement to assure myself that what my ears were telling me was correct, I looked up the frequency range of electric bass guitars.  Here is what I found:

 

Electric bass string freqs.GIF

 

Note that the "B string" is for 5-string electric basses, and the "C String" for 6-string electric basses.  The double bass ("string bass") nowadays has an 31 Hz open "C" string" with fingerboard extension for the lowest frequency string.  Most jazz bass players nowadays have begun using 5- and 6-string basses, and therefore will have fundamental frequencies in the 30-40 Hz range.  I find that most SACD recordings of jazz bass players, recorded on or after the year 2000 using DSD-only recording will have these frequencies intact on the recording, which is a significant addition to the listening experience.

 

I also checked on the frequency spectrum of kick drums.  Here is what I found for the time/frequency graph of a typical kick drum:

 

mixing_kick_bass01.gif

 

blog_kick_spectral-1a.png

 

As you can clearly see, just based on investigation of these two instrument types, the fundamental frequencies of the recorded music to accurately reproduce instrumental performance is more than a octave lower than the 80 Hz roll-off found in the example recording. 

 

On What Recordings Is This an Issue?

 

Good questions immediately arise from the information above:

  1. why would a mixing or mastering engineer attenuate and thereby remove these frequencies from our recordings, especially in light of the information that 25% of the importance of loudspeaker performance in subjective ratings is due to its bass performance (notably bass performance well below 80 Hz)? (See Floyd Toole's book, pgs. 197 and pgs. 463-464)
     
  2. Is this why many CD releases made before 1991 (and I suspect many phonograph records) sound strident and bass shy when played back on high fidelity sound reproduction systems--like the ones that many forum members own?
     
  3. How many recordings, and of what date released and type are affected by this "mastering feature"?

I've found many, many more examples of this type of mastering frequency response profile, especially from pre-1991 recordings.  Most of these recordings cut the bass below 100 Hz, not 80 Hz.  This is the "missing octave".  For pipe organ performance, fundamental frequencies as low as 17 Hz are typical for many large instruments with 32' fundamental stops.

 

Why Are These Recordings Missing The Bottom Octave?  Isn't The Bass Originally There During Recordings? Why Would Mastering Engineers Remove Bass Below 100 Hz?

 

This is where the story gets interesting. 

 

I've found through my now-many remasterings of different FLAC and WAV files with this characteristic is that the bass frequencies are inherently high in amplitude relative to all other frequencies (remember the decreasing slope of a typical music track, above).

 

This means that any mixing or mastering engineer worried about compressing tracks for the sake of maintaining a "loud" sounding mix, will HAVE TO roll off the amplitude of bass track, either by using equalization filters to cut the output extremely steeply below 100 Hz, or use equalization roll-off and a multi-band compressor that further compresses the bass track disproportionately to the higher frequencies on the compressed recordings...

 

Beginning to get the picture?  If you're feeling as though you can't trust a mastering engineer as far as you can throw him/her, you've got the message.  Loudness War techniques have been in existence for many years now.

 

Why would someone buy loudspeakers that can reproduce sub-80 Hz music content with great fidelity, but then turn around and find that their music has been intentionally blanked out for 3 of the 5 strings of an electric bass or the most energetic octaves of a kick drum (in fact rendering the drums on the music track for tom-toms and kick drums indistinguishable)? If you hear bass on these recordings, what you are hearing is actually the second harmonic of the bass player or the upper harmonics of the kick drum - not the fundamental frequencies.  

 

It seems insane, doesn't it?  Like a huge betrayal of trust.  Trust me, you can hear that loss of an octave (or more) and it's not very nice to listen to, IMHE.

 

Can Anything be Done to Recover the Missing Octave(s)?

 

Fortunately, the answer to this question is "yes", if the recording that you have hasn't been compressed using a multi-band compressor, like the Loudness War tracks made from 1991 to the present typically have. 

 

So here's the situation:  if you have a CD made before 1991 (or certain other CDs made after 1991 but not using compression techniques--which I'm finding are increasingly rare, but they do exist)--excellent results can be had by re-equalizing your CD tracks.  How much re-equalization?  See the below Audacity filter that I use as the initial starting curve for these bass-deficient tracks (followed immediately with the "Normalize" filter in the Effects menu to re-level the output before saving the new equalized track). Note decreasing gain above 100 Hz, which I find is needed for most (but not all) tracks.  I also find that there are favorite EQ curves used by the various artists that seem to run from album to album:

 

Bass_recapture_filter.GIF

 

And the "after equalization" curve:

 

After_EQ_results.GIF

 

The "after EQ" track sounds a LOT better now.  Highly recommended.

 

Chris

  • Like 5
Link to comment
Share on other sites

To prove to myself that the restored bass octave tracks were originally rolled-off due to Loudness War compression techniques by mastering engineers, I subjected my now-restored music tracks to both DR Database tool application sweeps and "ReplayGain" average level recalculation. 

 

In every case of using restored bass tracks, the "ReplayGain" levels became more positive (i.e., they require more gain to play back than the original tracks), indicating that bass transients were suppressed in the original tracks in order to make the tracks louder.

 

Most tracks using impulsive bass and kick drums increased in their overall DR ratings - sometimes by as much as 4 DR ratings points. In tracks where the bass plays more-or-less continuous bass lines, there is a slight decrease in the DR ratings measured.   This can be explained by the fact that the restored bass lines increased the overall track loudness levels, but did not materially increase the cresting transients in the recording.  This would lead to slightly lower DR ratings (usually one DR point at most).  But the listening quality of the tracks significantly improves, due to restoration of the original bass lines that the percussionists/drummers and bass players played in real life.

Edited by Chris A
  • Like 2
Link to comment
Share on other sites

To expand on the question "On What Type of CDs is this Found?", I first used this process on several noted CDs that I never listen to.  Some results were initially spectacular. 

 

Then I started down the line with other artists from the 1970-80s and found a high percentage of these CDs also having the same disease including:

  • Doobie Brothers,
  • early Steely Dan,
  • Pat Benatar,
  • Dave Brubeck,
  • Tears for Fears,
  • The Doors,
  • Grand Funk,
  • Mannheim Steamroller,
  • Jackson Browne,
  • Rolling Stones,
  • early Eagles,
  • The Who (the worst I've attempted to fix, in point of fact),
  • Linda Ronstadt
  • Billy Cobham
  • Stacey Kent
  • Diana Krall
  • The Alan Parsons Project
  • etc., etc.

When I look at artists from this time period, virtually all of them that haven't already been classified as "audiophile quality" (...do you see the association here?...) are usually sporting these properties and which respond significantly to this treatment.  I've got many more days of editing ahead of me--perhaps weeks months. [edited 28 Aug. 2015]

 

The best results--most startling, that is--are CDs of bass players, such as Marcus Miller (a few tracks/albums), Brian Bromberg (wow...big, big difference on his Wood series), Esperanza Spalding (ditto), and some jazz artists, notably early Jean-Luc Ponty, etc.  I'm just scratching the surface, that even though I've been devoting more than half-days to this for two weeks.

 

Amazing...isn't it?

 

Chris

  • Like 3
Link to comment
Share on other sites

I was just going to write on "why do mastering engineers still attenuate bass?" but you linked your excellent reference, above, which is incredible (at least to me) that this company is trumpeting these techniques to sell their wares. 

 

I always thought that the reason for decreasing bass was due to the RIAA curve, with its attendant pre-emphasis attenuation of almost 40 dB relative to the highest frequencies near 20kHz.  What I've found is that these techniques were used even on many of the earliest CDs made before the real Loudness Wars got underway in 1991 - i.e., not restoring the bass octave(s) relative to RIAA curve vinyl mastering tracks. 

 

This explains a lot, if true, as to why so many people didn't like CDs early - and it wasn't because of the format - but the mastering processes used that didn't really restore the RIAA de-emphasis curve for the digital tracks.  I suspect that even nowadays, most "analog-only audiophiles" playing vinyl and reporting lots of bass...well, the only answer is that either their cartridges or their phono preamps (which apply the RIAA de-emphasis equalization) are actually boosting low frequency relative to the stock RIAA curve, because I believe that many vinyl records have the same missing octave equalization reported on, above.

 

This would be interesting to perform an end-to-end transfer function test on some "high end" phono setups reporting lots of bass performance from vinyl records--to see what is actually happening.  :huh:

Edited by Chris A
  • Like 2
Link to comment
Share on other sites

How common is this on pre-1990 CD's? I use Audacity from time to time, and would like to check a CD or two that I have to see if the LF has been rolled off like the one you have. Perhaps a simple test and filter system could be devised. 

 

1) Analyze the mix

a: does it sound good to you

b: does the graph look good

2) Apply filter

3) Analyze the new mix

a: does it sound good - rename file FINAL{filename.ext}

b: don't like the sound - change filter and go to 2

 

Since this is a totally subjective approach, and may be specific to your speakers, the FINAL cut may not sound as great to others. This would truly make it your personal MIX. 

 

How long do you think it would take per CD to do this?

Edited by mustang guy
Link to comment
Share on other sites

Yes, that is my current process, except that I simply hit 'ctrl-z' (undo) to get back to the original track, because it's much faster.  Once I start to get close to a natural sounding track, then I can quickly undo and redo ('ctrl-Y') to hear both versions, then when pleased, I hit crl-shift-E to save the new track over the old one.  I don't like the saving of old tracks once I'm done - I can easily go back to the original disc to rip the original tracks again if I really become dissatisfied with what I've done later.

 

My room and my loudspeakers go down to about 14 Hz - time aligned at the crossover regions and EQed within about a dB above 300 Hz, and rising in response below about 100 Hz.  The axial room mode in my room is 14.3 Hz. 

 

It can take me a few minutes per CD if the mastering of that CD was extremely consistent - after I find what the "inverse curve" should be for the first track.  Some CDs take a couple of hours to unravel their equalization - especially "best of" CDs that compile tracks from different albums and years.  Some CDs or very old tracks on CDs (like the Beatles, for instance and the "wall of sound" type of tracks) even though not otherwise compressed, just aren't recoverable, IMHO.

 

I find that I cannot do batch processing of all tracks on an album/CD because of the individual variations in instrumentation and how each track was mastered.  YMMV.

 

Chris

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

At this point, I decided to investigate the reasons why it sounded so bad. My tool of choice: the freeware tool Audacity.

The Audacity audio editor was one of the things that came with this USB plug-in device allowing L-R in and out audio connections.  Thanks for sharing your thoughts on the compression and general messing up going on.  :)

 

Hopefully I can transfer some old vinyl into digits. 

 

http://www.behringer.com/EN/Products/UCA222.aspx

 

I've got the silver version of that device waiting on me.  That's queued up for secondary processing scheduling, once I get through the digital media (1000+ discs of pop/rock/jazz, etc.).  Note that Audacity is freely downloadable from the link found above at the top of this thread.

Edited by Chris A
  • Like 2
Link to comment
Share on other sites

 

 

The best results--most startling, that is--are CDs of bass players, such as Marcus Miller (a few tracks/albums), Brian Bromberg (wow...big, big difference on his Wood series), Esperanza Spalding (ditto), and some jazz artists, notably early Jean-Luc Ponty, etc.  I'm just scratching the surface, that even though I've been devoting more than half-days to this for two weeks.

 

Amazing...isn't it?

 

Chris

 

I am amazed that you can get better sound out of Brian Brombergs "Wood" CD's.  I thought they sound pretty darn good already!

Link to comment
Share on other sites

On 1/27/2015 at 1:49 PM, 'muel said:

I am amazed that you can get better sound out of Brian Brombergs "Wood" CD's. I thought they sound pretty darn good already!

 

 

Trust me - it's now a religious experience...

 

Bromberg.GIF

  • Like 1
Link to comment
Share on other sites

A slight "hump" or rounded spectrum from 1000 Hz to about 13 kHz

 Chris,

Can you explain this further. I'm not seeing what you're saying.

Spectrum_uncorrected.GIF

 

If you see that slight rounded curve starting at 1 kHz and continuing up to 13 kHz,, if you plotted the curve without its downward trend, but that segment of the curve being plotted on top of a level line, you will hear that as a significant rise in HF response.  This is an area of the frequency response that the human ear is really responsive to, so small changes in this area show up as big changes in the way it sounds.

 

If you listen to the mp3 file here, you will know what I'm talking about simply by hearing it.

 

http://smile.amazon.com/gp/product/B001NTWF96/ref=dm_mu_dp_trk1

 

Remember that you are looking at a plotted curve on a  log vs. log scale, so any non-straight line response is really not natural, rather something that someone changed from a straight line to a curved one through using master EQing on the final mixed track. 

 

That's how you can easily tell if full ensemble music tracks (i.e., bass, drums, and a few wind/string instruments+voice) have been altered from how it actually would sound if the musicians themselves were listening to each other while playing together in the same room and adjusting the loudness of their instruments to taste while playing. 

 

Chris

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

 

I am amazed that you can get better sound out of Brian Brombergs "Wood" CD's. I thought they sound pretty darn good already!

 

 

Trust me - it's now a religious experience...

 

Do you mind sharing your settings for Bromberg?  If I remember you save those as a file (.XML) in Audacity and could probably be imported in.  This is very interesting and I'd like to play with it a bit.  

Link to comment
Share on other sites

Unfortunately, I didn't save the EQ curve settings from album to album (I reused them from track to track only, since Audacity saves the last curve that you use automagically), so I'll need to do it again to re-create the curve. 

 

There will be a slight pause while I reload a track from that CD and re-EQ it.

 

Chris

Link to comment
Share on other sites

Okay, here is the FR curve for the first Wood track (The Saga of Harrison Crabfeathers) and the equalization curve that I used in the same picture.  The exact reproduction of the low bass boost curve isn't really that critical, not like the higher frequency EQ settings, which need to be set much more carefully.  Remember to use the "Normalize..." effect both before and after the application of this EQ curve to prevent clipping during EQing and to re-level the response back to normal level afterwards.

 

Bromberg_Harrison_Crabfeathers.GIF

 

Here is the "before" FR plot of the same track:

Bromberg_Harrison_Crabfeathers_before.GIF

 

And here (hopefully) is the XML EQ file (attached below)...

Bromberg_Harrison_Crabfeathers.GIF

Bromberg_Harrison_Crabfeathers_before.GIF

Bromberg Wood EQ.XML

Link to comment
Share on other sites

Thanks, Chris!

 

It's going to take me a while to digest your posts, but I wonder if what you are writing about is why:

 

1) Back when I had real, old fashioned, tone controls I often had the bass control set for + 2 (about 8 dB boost at the very bottom on my Mcintosh C28 preamp)  ... sometimes, both on the MAC, and on the later Luxman I'd have the bass even highter.

 

2) I have my subwoofer turned up

 

3) I have often been annoyed by audiophiles who insist that using tone controls or equalizers is some kind of crime.

 

Perhaps the industry is afraid of eliciting cabinet resonances or distortion in cheap speakers ... so ... I wish CD manufactures would release second versions, honest versions, perhaps at a slightly higher price, perhaps as hybrids with SACD layers, in which the dynamics and frequency response have not been ****** up.

  • Like 3
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...