Spectrum of Musical Genres--from the Perspective of Remastering or Demastering Music

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Administrator
Joined 2004
Paid Member
Chris can you show us a graph or two of your typical remastering curves? Pictures being worth a kilo-word and all that. :)

The reason I started analyzing music tracks was to understand what makes certain traks sound good (to me); There are many tracks that have a pleasing tonal balance, and maybe even more that do not. I wanted to know why, was it just the mix or did these tracks share a similar EQ and tonal balance?
 
As mentioned above, most of the non-specular acoustic energy on the FFT plots usually comes from drums of various types (i.e., non-pitch-specific). Compression produces reduced amplitude peaks, while limiting (i.e., clipping) produces a succession of high amplitude odd-order harmonics, but only for the duration of the clipped/limited peak.

Chris

Compression has time constants ( attack and decay ) that are usually much longer than even 60hz so they do not reduce peaks. That's limiting. Compression is amplitude modulation changing the envolope of the waveform to make the low level parts of the music louder. Limiting is "soft clipping" and does not sound like clipping, it's basically a compressor with faster attack and delays. Both of these processes are indispensable in multitrack recording/mixing and used properly improve the the final outcome. They have been in use since the fifties. The problem now is that the digital versions are so cheap and effective anybody can join the loudness wars and most do.
 
Chris can you show us a graph or two of your typical remastering curves? Pictures being worth a kilo-word and all that. :)

The reason I started analyzing music tracks was to understand what makes certain traks sound good (to me); There are many tracks that have a pleasing tonal balance, and maybe even more that do not. I wanted to know why, was it just the mix or did these tracks share a similar EQ and tonal balance?
What you ask is actually part of "demastering 401" instead of "demastering 101". Understanding a particular mix of sound that sounds good takes a discerning eye and good ears. I usually poke around with different parts of the spectrum to hear if there are any improvements to be had, but I can't wrap that into a canned process--it takes a keen eye and good ears to discover the reasons why something sounds good. It's much easier to see and react to tracks that sound bad. And this usually becomes a process of asking myself "what's not right with the sound". This self-questioning usually identifies the source of the problems more quickly than trial-and-error. I spend a lot of time looking at the spectrogram, not the cumulative PSDs when the tracks sound particularly good. The more closely that you look at the spectrogram, the more you will find to help you understand and reuse the information for other tracks.

I actually posted an example of a "first cut" demastering EQ curve for the track whose PSD I also posted ("Strange Brew" by the supergroup Cream) here.

There is a process to correct tracks with issues, but it always starts with looking closely at the cumulative PSD (the "plot spectrum" view in the tool that I use) for the incoming music track.

I've already talked about the objective--which is to create an inverse EQ curve to move the target PSD curve toward the 1/f curve. I do this in iterative steps using the EQ curve that I showed in post #10 which I created using a series of do/undo commands while looking at the most visible features that are not 1/f features of the current PSD curve, then moving the EQ curve points, then reapplying EQ again, until the resulting PSD begins to look like a straight line descending 1/f curve (I don't worry about the slope of the curve yet--that comes later).

Once I massage the demastering EQ curve into a fairly flat and descending shape, I usually "Normalize" to the peak value of 0 dBFS, then I look at the new resulting spectrogram (also shown in post #10) to see if the curve needs to be tilted up or down to match the -17 dB/decade slope needed. I use either a canned "added brilliance" , or a "decreased brilliance" tilt EQ curve that is a straight tilting line EQ in my saved EQ curve list--which is handy at the bottom of the saved EQ curve list--at this point until the colors on the spectrogram are balanced from low frequency to high frequency. I use however much tilt I need by doing and undoing iteratively while changing the curve slope until I arrive at the right slope to achieve the right brilliance (tilt), then I apply.

This sounds complicated, but it's actually very intuitive and fast using the shortcut keyboard commands. Once the resulting PSD curve is flat and the spectrogram shows the balance of highs to lows, I then listening briefly to the track to verify. Then I undo all the EQ commands (using ctrl-Z's) and listen again, but to the track as received from the CD. This immediately tells me if I'm on the right track. Once I start to get close to balanced, I reset the track's overall loudness and listen to the track more closely, and usually for a longer period of time until I'm satisfied with the sound. I may skip through the track checking loud and soft passages alike. Then I save, and move to the next track in the album to demaster.

All this usually takes less time than it takes the track to play through, so I can usually demaster an album in about the same time it takes to play it all the way through. I then move on to some other tasks not related to demastering, and come back for fine tuning the next day, when my ears are fresh again. This part of the demastering usually is very fast and straightforward since the tracks have largely been corrected of their EQ anomalies - and fine-grain tilting of the curves is the last step.

When I get to the next day fine tuning, this is where I'm actually performing "remastering" to balance instrumentation and voices within the stereo track. This portion of the process actually pulls the track's PSD off of 1/f, but the adds and subtracts to the various portions of the frequency spectrum are usually quite small: 1 to 3 dB adds and subtracts only, and usually only for one or perhaps two areas of the curve. This is the part of the process where a good ear and some practice comes in handy. The fuller the PSD looks (like a rock track full of sound at all frequencies), the more sensitive the fine tuning becomes to get it right. Sparse tracks have a lot more latitude of the allowable balance than the full tracks which have to be right on.

The result of all this is usually much fuller sounding, with palpable bass and much more prevalent mid-bass, much better balance to the highs with no solo instruments "sticking out", no timbre shifts that distract, and good liveness or presence of sound.

Chris
 
By way of a typical demastering curve, I have found that the actual track-to-track variance is so great that attempting to use a generic correction curve really isn't useful. However, I did look at the Chapman JAES article and produce a notional demastering curve based on one of the genres ("heavy") that is detailed in the article, (which I hesitate to show here, but since you asked...):

Heavy Group with demastering profile.JPG

While I have not actually produced a demastering profile (i.e., the resulting track PSD from applying a demastering curve in blue above) that looks exactly like the one above, the representation is commonly seen as a trend. The exact levels of the two curves relative to each other may not be correct, but you get the idea of the difference between a typical hard rock curve as it exists on CD and a demastered one. The blue line is the approximate inverse EQ curve to achieve the target demastering profile, i.e., the difference between the red curve and the dotted black curve.

There is always quite of bit of excess energy above 1 kHz to 10 kHz, with a peak of that excess usually centered on the area of the human hearing system's most sensitive listening band (2-5 kHz). This is of course a contradiction: why would you boost the areas of a typical track's PSD in the most sensitive hearing region? The answer, of course, is that you're actually pushing everything else down in a relative sense. The upper harmonics of vocals and electric guitar are usually significantly boosted, while the highest frequencies above 10 kHz are usually attenuated in a relative sense: crash and ride cymbals are actually pushed into the background.

The bass level attenuation is well documented, and the correction EQ curve shown here can actually vary greatly based on the mastering bass roll-off filters used and the acoustics of the venue in which the original recording was captured. Most recording venues have resonances and destructive interference (1/4 wave boundaries, etc.) below 100 Hz that can easily be corrected here to achieve a much smoother bass response.

One thing that I should mention: while the demastering EQ shown looks like it would create an unlistenable track, the actual effect is quite the opposite: you will find that the music tracks will now sound very good at all levels of playback loudness, with only minor adjustment for Fletcher-Munson effects on the bass bands below 100 Hz usually required. In other words, the tracks will actually start to respond to level-based adjustments of EQ that was commonly seen in integrated amplifiers and receivers in the 1960s-1980s: "Loudness" controls.

Chris
 
Administrator
Joined 2004
Paid Member
Thanks Chris for the very thorough explanation and rich examples. Reading your two posts above made me realize that I didn't clearly ask for what I wanted to see - the target curve. However you did provide that in the image just above. Thanks. And now I understand much better how you get there.

I think that for most of us, it's just too much work. Someone mentioned that a few years ago when you had explained your process. However I do have some recent recording where I very much like the music, but find the tone hard to listen to. Those would be worth the effort to find out if tweaking the mastering allow me to listen to the music without getting a headache.

When I get a chance, I'll look at automating the process some. It would be nice to know if a quick fix could do enough to make some bad albums with good music listenable.
E.G:
Declip and normalize (or save in float)
Read spectrum of track with RMAA
Bring spectrum into REW and generate filters to match a target.
Export filters as an impulse and convolve the track to de-master.

The result of all this is usually much fuller sounding, with palpable bass and much more prevalent mid-bass,
That's the main difference I hear between live music and recorded. Live at reasonable levels always sounds fuller, rounder, less shrill than recorded music. Not sure why.
 
One of the goals that I had set some time ago was to create a couple of video tutorials to show how the process works, and how quickly it executes. Alas, about that time I was investing more time in the K-402-MEH and other DIY activities (i.e., I don''t see the advantage presently of developing YouTube video production skills). I offer to show anyone that wants to see how I do it, one-on-one and perhaps at the local audio festivals. It's a skill that I think is really valuable to have--especially if you're really into music rather than hardware.

So I've kept on demastering my music library, and have continued to develop the process and its effectiveness. At various points in time, I started to believe that I had attained the proverbial "99% effectiveness" level, only to come back later to listen to tracks done earlier and realize that I had left some demastering performance on the table. I've now got my setup and listening room dialed in to flatness levels that I didn't at first think were going to be effective in increasing the listening enjoyment. It turns out that these two things go hand-in-hand: a well dialed-in hi-fi room and skill in demastering/remastering. The more you do in one area, the more it shows up in the other area.

At the same time, I continue to refine the CD-fixing process until I can now do most of the coarse initial correction tasks almost without listening to them (i.e., looking only at the spectrograms and cumulative PSD plots). This really speeds the process to the point where I can stay ahead of the tracks as they play in real time just after I fix them. That's a pretty fast manual process. In terms of DIY electronics and loudspeaker building, etc., it's incredibly fast.

So that leads me to the punchline: I have to say that arguments of "it's too hard" just don't make a lot of sense to me, knowing what I know now.

You start with your favorite albums to demaster, continuing to practice speeding up the cycle time until you get to 80% full speed--sort of like learning how to type on a computer keyboard, etc., then you can stretch your legs and start into your full ripped CD collection in earnest--or just the CDs that you always wanted to fix, but didn't know how.

I'm about 95% complete with the discs that I believe need to be fixed--which represents about 80% of 1500 CDs that I've ripped to FLAC. Most of my demastering time now it devoted to fixing CDs that I buy off of Amazon marketplace (generally for less than $5/disc to my mailbox). I find that I can now listen to many more genres of music that I never could listen to due to poor mastering EQ used. I've expanded and filled in many holes on my collection, and intend on continuing this process.

Chris
 
Administrator
Joined 2004
Paid Member
Well the "Knowing what I know now" part is important. :)
How many hours, or 100s of hours did it take to develop that knowledge and skill?

There are rewards to be had from learning any skill, but the difficulty of getting started can be daunting.
 
True that, but the fact that I persist in raising the issue also should be factored into what the difficulty actually is. Perhaps initial perception isn't reality this time.

Learning about loudspeaker audio electronics (sources, preamps, amplifiers, etc.) fabrication and design isn't really a walk in the park for the average Joe. Most people haven't had a course in basic electronics or even electrical circuits, much less understand the concept of impedance, etc.

I'd put demastering as much easier than those other DIY audio projects: all you need is your ears, some basic information about Audacity (freeware), and how to dial-in your hi-fi system to flat response in a reasonable listening room. The rest is more like a video game in terms of learning how to do it well...in my experience.

I have to say that claims that it's too difficult reminds me of other DIY adventures where some imply "not only do I not want to try it, I don't want anyone else to try it." I've run into that sort of thing before.

It's okay, it's basically just applying EQ to our tracks in an intelligent and efficient way...like this, but a lot less expensive (i.e., it doesn't cost $2000):

BB_Home

Chris
 
I just found this thread and am fascinated. I also found your pdf's on an AV forum


Demastering your music | AV NIRVANA


Really interesting, my first thought was one of my fav bands, Queen, who sound fine on "the kitchen tranny",but on a decent system just sound dead. Looking them up on the Dbase you posted, 9 (near fail) is a good score for them :(. definitely gonna give them a go. One of my other faves, Joe Bonamassa, seems to score well on vinyl but poorly on CD. I would have expected the other way round? Many thanks for your time and effort in this :)
 
Member
Joined 2014
Paid Member
Thanks! I'm always looking around for more automatic solutions that others can use. I've got so much time invested in doing it by hand using Audacity that my needs are basically few beyond Audacity now. But there continues to be interest from others willing to take a little time to undo some of the mastering damage.

If someone has tools that can safely and effectively reverse the compression (i.e., not limiting)...that would be useful. Such as a good and handy multiband upward expander. I've used a simple one (Transient Monster) and a little more complicated single band freeware plugin, but they definitely need multi-band capabilities to be truly useful, I believe.

Chris
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.