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

[Chris A]

I really don't get why they push the treble forward.

 

Having restored the bass and often the treble curves on perhaps 50 or more discs now including jazz, rock, pop, and blues genres, I can see why treble is boosted on most of these recordings.  In particular, boosting 6-13 kHz is common on jazz vocal CDs, and boosting 200-1300 Hz is common on others, all with consistently minor boosts in the 1-10 kHz range.  If you try to remove all of these boosts, the recordings go flat and uninteresting, no doubt due to how the mixing engineers locked in the relative voicing of each instrument/voice part in each frequency band prior to the mastering process.  The trick is to reduce the treble boost as much as possible while still retaining the energy and interest of the original mastered versions.  Once you do this, I find that the tracks automatically sound much more natural and much less strident.

 

I find that on many of my restored tracks, I take out an additional 3-6dB of boost in the 6-13 kHz band to tame down the forward treble on the tracks.  I found some CDs that took almost a full 20 db cut on the treble from 1-->20 kHz, like applying an actual RIAA de-emphasis curve but only for the treble portion: the bass FR was good as-was (ref. the two Earth, Wind & Fire discs mentioned above). 

 

Chris

Edited January 31, 2015 by Chris A

[Chris A]

I'm now starting on my pipe organ two-channel recordings--mostly J.S Bach compositions.  I find the same level of bass attenuation on these two-channel recordings as on the other popular CDs that I've worked on--but not on my 5.1 organ recordings, which are generally balanced as is.

 

 

EDIT: I've finished my first organ CD (Michael Murray - Bach at St. Bavo's).  On average - it took 25 dB of boost from 250 Hz down to 20 Hz, and that is fairly light in terms of the resulting spectrum and in-room listening results.  I can now hear the 32; and 16' ranks of the organ, which I could not even detect beforehand on the distributed tracks. There are some interesting subharmonic beats that you can now hear (as well as see on the playing sound envelope - down around 1/4 Hz, these waves pass back and forth over you as extremely long period beats due to the linked 20 and 40 Hz octave ranks.

 

You do know that this activity is bringing many more insights on the music performance, the recording venues, the mixing, and mastering effects, all of which can be pretty easily isolated separately into separate "features".

 

For instance, when looking at the Michael Murray organ tracks of J.S, I find that the placement of the microphones is probably too far away from the pipes.  How do I see this?  The spectrum plot shows a much bigger than average drop in overall SPL each octave above 1000 than any of my other recordings.  It also tells me that most organs are voiced much differently than almost any other type of musical instrument, with a much more attenuated treble band relative to lower frequencies.  Trying to push up on this curve reveals the upper ranks of the organ much more than any other instrument that I've tried to re-EQ, especially the human voice.  This is no doubt due to the broad frequency range of pipe fundamental frequencies relative to the much narrower range of human voice frequencies.

 

Another insight: the organ tracks that I'm currently repairing have been the least interesting to me to listen to of all my organ discs.  Perhaps with some subtle changes, these tracks will be much more likely to be played.  This tells me, once again, that as a musician, the ends of the reproduced music frequency spectrum determines listening pleasure in terms of hi-fi playback, especially low bass performance.  YMMV.

 

Looking at the spikes in FR vs. time within the restored bass tracks, it's easy to tell if the spike is due to an actual performance transient or something bumping the microphone or other recorded low frequency noise propagating through the venue, simply by looking at the duration of the spike and the neighborhood of other traces in the vicinity of the steepest spike.  When editing, sometimes it is necessary to go in an manually edit out these spikes in order for the Normalize... function to bring the tracks back up to near-normal average levels.

 

Having undone the mastering tweaks for so many discs now, it is now easy to identify recordings that are uninteresting--in terms of where to crank in the right mastering tweaks to improve the overall sound, just like mastering engineers do.

[Travis In Austin]

Bass atenuation can take place anywhere in the recording chain. A 6 or 12db 40 hz high pass filter is typically place on the bass guitar but it depend if it from AMP or DI or combo of both. KICK drum can be anywhere from 30hz to 50 hz and will be attenuated so it is not lost in the bass line.

Every microphone has a unique frequency response curve, some with bass boost, some not.

Most bass and drum lines are mixed with use of high pass filters to obtain a clear mix of kick drum and bass line. Thwre are multiple techniques on where to add add and where to subtract to avoid mud.

At mastering, if wverything was done properly at mixing, they just sequence, edit, put in fades and cut flat. More likely, something is not 100% and can be improved with eq.

If mastering for vinyl they have to look at whether bass attenuation is necessary to fit record on one side without having grooves slammed up against each other. They may have to do low pass filter at upper frequencies to protect cutter head, even if helium cooled.

The big problem is like in the case of your excellent Vannelli example. The of original vinyl of that lp is excellent, mastered by Bernie Grundman, one of the best in the business. Unfortunately, later on some poor guy had to master it for Cd and, based on your curve, it appears he was given what is referred to as the EQd limited master, the vinyl master, the one with the bass cut and the upper end boosted. This tape was recorded right from the cutter head amplifier outputs so that it could be sent to other pressing facilities. These were readily available, as opposed to 2 track mixdown tape, and this is all the CD mastering guy had to work with. Based on your curves and the era I am pretty sure they did that cd from the EQd limited master.

What you are observing has been known and talked about for a number of years on the recording, mixing and mastering forums, and the Hoffman forum.

People like Barry Diament (about the biggest name there is in quality cd mastering and remastering) has talked about loudness wars, high rez digital, and possibility for audiophile quality in mixes and masters being the rule rather than the exception.

I don't think it is possible to tell where attenuation occurs in the chain unless you know a lot of factors, drum head used, bass recorded on amp with mikes, what kind of mike or was DI used, were the bass and drums recorded seperately, or in some cases together "live" etc. The hardest part of mixing a rock/pop band is getting the bottom end right. Keeping the kick and bass seperate, coherent and clean.

Often the masterer gets it and it is smeered mud and fixed in the tape, pre eq and limIter applied and the masterer has to try and save it. Sometimes he or she can, and sometimes you just cant polish a turd.

If you look at any top mastering services website (not people working out of their garage) but mainstream people like Grundman, Ludwig, Saxe (both), and Diament they typically have what they would like to do their job properly.

Travis

[mike stehr]

Yay, I can download video/audio from u-tube. Now to just figure out how to download the proper codecs, or how to convert to format for media player to rip or burn...

 

I really need to just sit down and read the manual.

 

Thanks, guys!

Edited January 31, 2015 by mike stehr

[garyrc]

On average - it took 25 dB of boost from 250 Hz down to 20 Hz, and that is fairly light in terms of the resulting spectrum and in-room listening results. I can now hear the 32; and 16' ranks of the organ, which I could not even detect beforehand on the distributed tracks.

 

Wow.

[garyrc]

later on some poor guy had to master it for Cd and, based on your curve, it appears he was given what is referred to as the EQd limited master, the vinyl master, the one with the bass cut and the upper end boosted.

 

I suspect this happens a lot.  Something similar also happens with the movies with good magnetic soundtracks from the '50s, '60s & '70s.  In this case, the original sound elements were recorded with rolled off bass to avoid over-recording with the expensive orchestra sitting there.  The bass was then boosted back up -- to the filmmakers' taste -- during the mix, when it was inexpensive to do something over in case of over-recording.  Some of our DVDs and BDs are remixed from the original sound elements, before the bass was boosted.  Believe me, most of these vintage films had a lot more bass, and sometimes more dynamic range, in the theaters.

[Chris A]

...What you are observing has been known and talked about for a number of years on the recording, mixing and mastering forums, and the Hoffman forum...

 

Thanks for the comments, Travis. It has always been my intention to bring up technical subjects related to hi-fi that really matter in terms of the listening experience.  Hi-fi itself has a pretty simple definition (from Wikipedia):

 

...high-fidelity equipment has minimal amounts of noise and distortion and an accurate frequency response.

 

 

I've spent a fair amount of time searching for articles on this subject in particular (i.e., the missing octave) and on the misapplication of recording, mixing, and mastering practices that result in poor listening results. 

 

Of all the subjects that I've investigated and written on, these two subjects have brought me the greatest amount of results in terms of "recapturing hi-fi" than any other.  So, in that spirit, I believe that I'll keep sharing any significant observations, successes, and how-to processes, etc. since I believe the payback is great.

 

I'm learning that there is a learning curve in this "mastering restoration process", and any lessons learned and significant observations that others find are also welcome.

 

Chris

Edited February 1, 2015 by Chris A

[Chris A]

The first album of The Alan Parsons Project, Tales of Mystery and Imagination, first released in the mid-1970s and subsequently re-released again on vinyl and CD several times in the succeeding 15 years, has never been a favorite to listen to even though the music itself is fairly consistent with the other offerings from this band.  I now know why...and it is more than just the fact that this album was compressed on its first release and apparently all releases since.

 

When you open up each track in Audacity, you will see great amounts of red areas on the time-amplitude plots, indicating significant amounts of clipping that have been on the released CD.  Using the methods described above (Normalize..., Clip Fix..., Equalize..., Normalize... again) will significantly hide the effects of this clipping and bringing back some listening pleasure to this album.  The regaining of at least an octave of bass response is not the least of the improvement in listening.

 

Of course, this album will never produce the same level of enjoyment as most of the later-released albums by this band, it does allow me to actually listen and connect with this music in a way that I never could before, even when it was released on vinyl. 

 

Chris

 

Edited February 1, 2015 by Chris A

[dubai2000]

Chris,

 

a very interesting thread. Now I admit to being a complete newbie when it comes to reequalizing anything. I downloaded Audacity, but somehow I don't manage to dio what you have done. Yes, I have found the analyze button - but then it all gets confusing. Is there any kind of step by step insruction how to do things from there? The online handbook didn't really help too much - or I am simply too.......old-fashioned..... .

 

Wolfram

[Chris A]

Sure...no worries, I'm sure that you're not the only one.  So thanks for asking the question.

 

First, make sure that you can play a track through your loudspeakers from your computer by setting up the software drivers and selecting the right output hardware device.  When you can play an opened track through your loudspeakers, you're ready to start the equalization process.

 

Here is a sequence that I use:

 

1. open the track (ctrl-O) that you want to work on--one that has plenty of instruments and vocals, including bass and drums
2. hit ctrl-shift-F to maximize the traces in window for easier viewing
3. Look at the traces, both of them - the top one is the left channel, the bottom is the right channel
4. select Effect --> Normalize... then make sure that the first two check mark boxes are checked.  I'd change the value in the text box from -1.0 to -0.3, hit OK, then you will notice that the tracks are rubber-banded to the top and the bottom of each window, actually rescaling the track to match the available loudness bandwidth (you can actually omit this step to save time, unless you see the color red in the display, then you really want to Normalize... the traces
5. select Analyze --> Plot Spectrum. Hit enter if a dialogue box pops up telling you that there's too much data to analyze.  A frequency spectrum window for the tracks will appear.  You can maximize this window using the top right maximize button on the window so that you can see the trace more easily. 
6. Notice the shape of this curve, especially the frequencies below 100 Hz, if they drop off quickly in amplitude, then you've got a track that has the missing octave (which is more likely than not).  If this track doesn't have a big drop-off in bass under 80-100 Hz, close this track and open another one from another album/artist, then repeat the steps above until you find one that has the missing octave problem
7. Once you've determined that your track has a missing bass octave, select Effects --> Equalize... then notice a curve pops up.  Here is where you will open one of the example bass boost xml files that you found above in this thread, by clicking on the "Save/Manage Curves" button at the middle-bottom of the curve window.  Import that xml file by hitting the "import" button and searching for the sml file that you've saved locally on hard disc.  You should see the curve appear when you select if from the list.  Hit "ok" to close the curve management window, then "ok" again to close and execute the Equalize... filter.  You will see the traces change shape - perhaps even turn red, etc.
8. Select the Analyze --> Plot Spectrum again, then look at the shape of the frequency plot.  If there is too much bass below 80-100 Hz, close the spectrum window, and go back to the last step, this time change the amount of boost (either higher lower lower, moving the anchor points where you believe the curve might be better shaped to boost the bass just the right amount.  Hit okay to execute the filter and observe the traces and the Spectrum plot again.  iterate this until the plots look about right.
9. Select the Effects --> Normalize... again to re-normalize the track
10. Once the plots are closer, hit the play button at the top left of the Audacity window.  (You may need to find the right driver/output device in the dialog buttons along the top of the Audacity window until you select the right one and also set up the output to play through your loudspeaker setup.  Remember that it's important to hear it on the target set of loudspeakers in order to get the right equalization.
11. Listen to the tracks play.  If if doesn't sound right, get out of the dialog boxes/windows and hit "ctrl-Z" to back up in the sequence of filters and commands until you're back at the original opened track.  If you go too far, simply hit ctrl-Y to re-open the track again.
12. Go back to step 8 to repeat the updating of the Equalize... curve until you have the right shape to produce the most pleasing results.
13. When you are done, hit ctrl-shift-E to save the file to your hard drive.

That should get you closer to success.  If you have trouble, let us know here.

 

Many happy returns.

 

Chris

Edited April 23, 2015 by Chris A

[Chris A]

If you have red traces before editing them, then you've got clipped tracks from your source.  Use Normalize... to get the traces off the rails, then use "Clip Fix..." on the track to fix the clipped areas prior to re-equalization.

 

If you have red traces after running an equalization,, bring up the Plot Spectrum window again to see where the peak(s) in frequency are located.  undo the Equalize... filter, then reduce the amount of boost in that/those areas.  Iterate the filter.

 

Once you've got a good filter for a track, you can probably reuse it for the next track in the sequence.  You can save the equalize filter by selecting Effects --> Equalize...then hit the save/manage curve button, then select the "rename" button, which will save a copy of the curve in your list - name it something that you'll remember.  Whenever you need that curve again, select if from the Equalize... dialog list again.

Edited February 1, 2015 by Chris A

[Chris A]

Here is a picture of an Equalization filter that I used for a few of the tracks in The Alan Parsons Project Pyramid album.  Notice the somewhat complex treble down-steps and the wider/higher frequency bass boost than other Equalization curves.

 

 

Here is the spectrum for the track that I started with:

 

And the final track spectrum:

 

Once you get the spectrum looking like this, then very small changes in the Equalize curve more strongly affect the resulting sound.  Notice the general straight or slightly curved downslope with increasing frequency, with few if any protuberances or slope discontinuities in the spectrum curve.  This spectrum curve produces an overall sound much more natural than the forward-treble/missing bass spectrum that was distributed with the CD. 

Edited February 1, 2015 by Chris A

[Axz Hout]

r

[Chris A]

You have a hybrid ( of superior breeding) PA system that may be able to produce noise down to 10-25. Very few of us have/want this.

 

PA system?  I'm pretty sure that none of my loudspeakers or electronics are PA components (with the exception of the subwoofer amplifiers).  I might acknowledge that some of the components are used in cinema (i.e., walk-in movie theaters), but certainly not exclusively for cinema.  The DIYAudio forum is completely full of people "misusing" their hardware in this way.

 

As far as the "blowing up", well, I figure that if you have such a delicate system that cannot handle these frequencies, you'd avoid any type of re-EQ to restore the bottom octaves entirely.  YMMV.  Any home theater system can easily handle these restored tracks - or any two-channel system that has superior low frequency capability--like Khorns in good corners with or even without good subs. I assume that your preamp knows how to crossover to the right loudspeakers/subs without any issues.

 

I've found that the lower level sub-20 Hz frequencies are basically inaudible unless boosted to levels exceeding all other frequencies. I do know that the lower level infrasonic frequencies give a sense or feeling of space that is missing in the CD-delivered tracks - especially the pipe organ discs - where the infrasonic echos are usually at much higher levels and that are both audible and give a feeling of envelopment. These infrasonic frequencies are commonly found on commercially available 5.1 surround sound music discs.

 

Chris

Edited February 1, 2015 by Chris A

[Axz Hout]

k

[Chris A]

The Klipsch KPT-Jubilee is a PA speaker, unlistenable without time correction and EQ.

There actually is a difference between PA and cinema, I'm afraid.  I'm sure that Roy would also agree. "KPT" stands for Klipsch Professional-Theater. 

 

My Jubilees aren't in the KPT-Jubilee (3-way) configuration, but they do use the outstanding K-402 horn and bass bin which I believe are the source of the magic of the "wall of sound" imaging.  The K-402 horn is a controlled directivity horn that doesn't require time alignment or active crossovers, but it does better with them (as any horn-loaded loudspeaker might--including Khorns).  There are people using the same horns and drivers but using passive crossovers--that do require a balancing network, just like other Klipsch Heritage loudspeakers require their own balancing networks.

 

I'm not sure what the issue is.  Can you explain why you are so concerned?  Can you still use the above techniques to produce better full-range two-channel tracks without fear of "blowing up" your system?

 

Chris

Edited February 1, 2015 by Chris A

[mustang guy]

I did an AB test between my Dire Straights - Bro n Arms SACD, LP and CD a while back. I recall the bass being missing in the CD more than anything. I grabbed that iTunes file and converted to FLAC with dBpoweramp, brought it into Audacity, and performed the adjustments on the bass only. Oddly, I don't hear the huge difference shown in this adjustment, but there is more bass. I am at home, so I am playing on the ProMedia Ultra 5.1 computer speakers which truncate everything below about 25Hz. Literally.. Just to make sure I wasn't hearing things, I laid my iPhone 5S down in the same place and did a measurement of the same section of the song with heavy bass and compared them. 

 

 

Look at the before track. Holy crap that thing is maxed out! I like the one I did with some normalization way better. It sounds better, and you all know there is a MUCH better FR in this!

 

 

The measurement. Green before...

 

FR before:

 

foobar2000 1.3.7 / Dynamic Range Meter 1.1.1

log date: 2015-02-01 14:40:10

 

--------------------------------------------------------------------------------

Statistics for: 02-Money For Nothing

Number of samples: 22305780

Duration: 8:26 

--------------------------------------------------------------------------------

 

                 Left              Right

 

Peak Value:     -0.10 dB   ---     -0.10 dB   

Avg RMS:        -9.48 dB   ---     -9.75 dB   

DR channel:      6.87 dB   ---      7.06 dB   

 

from loudness war database CD: DR7       -0.10 dB    -9.61 dB      8:26 ?-02 Money for Nothing

--------------------------------------------------------------------------------

 

Official DR Value: DR7

 

Samplerate:        44100 Hz

Channels:          2

Bits per sample:   16

Bitrate:           1412 kbps

Codec:             FLAC

================================================================================

 

 

FR after: 

 

foobar2000 1.3.7 / Dynamic Range Meter 1.1.1

log date: 2015-02-01 14:42:09

 

--------------------------------------------------------------------------------

Statistics for: 02-Money For Nothing

Number of samples: 22305780

Duration: 8:26 

--------------------------------------------------------------------------------

 

                 Left              Right

 

Peak Value:     -0.32 dB   ---     -0.10 dB   

Avg RMS:       -12.15 dB   ---    -12.14 dB   

DR channel:      9.30 dB   ---      9.28 dB   

 

from loudness war database SACD: DR16      -1.16 dB   -19.67 dB      8:26 02-Money for Nothing

--------------------------------------------------------------------------------

 

Official DR Value: DR9

 

Samplerate:        44100 Hz

Channels:          2

Bits per sample:   24

Bitrate:           1564 kbps

Codec:             FLAC

================================================================================

 

Conclusion. I still don't have the FR from the SACD, but the bass is definitely closer. Honestly, I should figure out how to convert the SACD version to FLAC and see if I can tell where the differences are.

Edited February 1, 2015 by mustang guy

[Chris A]

One of the things that I just found was the "length of filter" slide in the Equalization curve dialog. Increasing this value will make the low frequency filter response much more responsive to the control point inputs.  This means that you can shape the Equalization curve filter more effectively in the very low frequency region to avoid sub-20 Hz boosting if you have concerns about this. 

 

Try it out.

 

Chris

[Axz Hout]

d

[Chris A]

Conclusion. I still don't have the FR from the SACD, but the bass is definitely closer. Honestly, I should figure out how to convert the SACD version to FLAC and see if I can tell where the differences are.

 

I think that the promise of increasing DR values significantly just by restoring the bottom octave is something that not only increases interest in this subject, but can also significantly add to the listening experience, i.e., correlating the DR value to listening pleasure.