If the remaster sounds louder, is it too compressed?

What is compression again? I understand noise reduction and varying amplitude levels, and what they sound like, but not compression.

 

I only take issue with this because it's such a fundamental error, about maximizing bits. Maybe it's even a component of the loudness race. And I see it repeated here so often. But I'm fairly sure Nika is open to fielding questions. And if you Google his name, and pose the question to him, I'm sure he'll be more than willing to give his professional opinion.

I believe he'll tell you that dither allows us to hear below the least significant bit, making maximizing a dithered signal unnecessary. Using more bits above the noise floor only makes the signal larger, it does not maximize sound quality. Only dither and noise shaping can maximize dynamic range, and how it's percieved. Dither also maximizes dynamic range by, again, realizing the perception of sound *below* the least significant bit.

 

Let's see if I understand this correctly. When music is compressed, it ends up sounding like the volume has been lowered, although in reality it hasn't. So to compensate, they then actually do boost the volume, which makes things even worse. Correct?

 

No, compression actually makes it seem louder, but also allows the headroom necessary to boost the volume even more.

 

Perhaps I should add, to avoid this causing misunderstanding, that an analog noise "floor," particularly tape hiss, is not to be confused with the audible limit. The noise floor is not really a floor or a brick wall. That is to say, there is audible sound on, within and below tape hiss. Having the noise floor at the bottom limit of the digital signal would result in audible losses.

 

Compression reduces the difference between the quieter sounds and the louder sounds. That's why extreme compression comes out looking like a "buzz cut" in a wave editor -- no dynamic range, everything comes out approximately the same volume. This is also what's used to increase the overall volume of a track, as Davey mentions above. It's the prime tool in the "loudness wars" we talk about all the time.

Jason

 

No. I'll try to explain. This will be somewhat innacurate and way oversimplified, understand, it's only to try giving a grasp of the concepts, which you can fill out in more depth and accuracy as you like. Let's look at an example of why compression and limiting is used for good results and how it can be misused. Let us say you were being recorded in the studio with a single mic to an old tape deck. Let's say you struck a gong and were very softly chanting in the background.

Our audio reproduction gear doesn't exactly work as smoothly as our ears. What happens is that the closer you get to that mic, the louder it gets, and if you move past a certain distance, it really tends to drop away in volume, more so than natural to our ears. For another thing the audio gear only has so much range for comfortably catching all the sound, less range than the mic might be sending it. Then there's the fact we may not want to crank the playback so high to hear the distant soft chanting with that seemingly exaggeratedly loud honkin' gong in our face. So: we need to reduce the louder gong and boost the soft chanting so we can hear it "right" when played back at a reasonable volume. That's where you use compression. You can compress the signal from the mic, which squashes the volume so the gong only gets so loud. You can seem to "turn up" the chanting that way - relative volume.

Same with a rock band, say the Beatles. You want all those instuments, including the loudest snap of a drum kit, wail of an electric guitar amp and the softest sibilant of the singer, oh and John whacking that roll of toilet paper with a comb, all fitting together nicely in one or two speakers at a reasonable volume. Compression becomes a volume tool in addition to the actual mic level controls. Compress all the mics! Maybe more in the mix. That's not a bad thing, it's all a question of discretion and the sound you want.

The more you compress, the more you bring lower volumes up closer to the higher. The louder parts get more and more squished of course and the overall sound gets less variation to it. That also appears to make the overall sound seem louder, a trick on our ears, kinda, cause you're hearing all this volume range moving way up without actually raising the peak volume. This more constant level of noise can get more fatiguing too. So the trick is to compress to make it sound "realistic" and/or achieve a desired sound when played back, but not to the point that you're destroying so much of the volume range it affects the impact of the music, which obviously uses volume fluctuation as an aspect of expression, or takes away too much of the depth between the instruments and the acoustic space, and makes it sound too squashed and all in your face.

The problem of excessive compression in mastering is that they are taking a source that's already had all that worked out so it sounds right and the way they want it to sound, and compressing it much further. They want everything to appear as loud as they can make it, discretion out the window. You loose those volume fluctuations that impart subtleties of music, everything gets squashed, and the all-in-your-face result gets tiring and less effective, whether one is conscious of the cause or not.

A "maximized" overcompressed mastering of our example would have the gong smashed to a flat mass with little "reach" because the volume range it has is tiny, your chanting so loud and exaggerated it'd feel a bit more like a loud muttering in your ear, and the noise floor of the tape pulled way the heck up from distant hiss you'd easily tune out to a rushing noise, so they'd be even more eager to use no-noise, further reducing the fidelity... With this noisy muttering and all, you turn the overall volume down to make it more listenable... but they want it all loudly audible so... the cycle gets worse.

I hope that gives a clearer notion of some of what may be involved and some of the reasons why over-compression is a dumb practice and a bad trend.

 

You gotta hang out here more!

 

OK. Now I see where you are coming from! But, I submit that the fades and quiet spots are just as important as the musical content itself. Also, The dither (or lack of) has an audible effect on the sound no matter what the amplitude.

So does slashing frequencies in the infrasonic range. It's an old engineering trick.

However, one should only use noise shaping when converting to 16-bit from a higher bit-depth. Dither is, or should be present.

No argument there.

Once you further explained what you meant, you started to make sense.

 

Compression usually has the effect of raising the lower-level sounds, and lowering the louder sounds, therefore, making the sound less dynamic. When you get that in-your-face sound, that's probably compression coming at 'cha.

Compression has many benificial uses. I will go as far as to say that mild, careful mastering compression can have it's uses. What we object to is overcompression.

When compression is used, a limiter, another type of compression, is used. With those two, EQ is often employed to compensate for what the compression alters.

 

If you don't record an analogue signal to a 16-bit system such that the loudest parts of your signal are as close to 0dB as possible without clipping then you are not utilizing the full benefit of the 16-bit system.

Suppose you peak at -6dB. That effectively means you are only utilizing 15-bits. This will mean that you have only 32,768 steps to represent the amplitude of each sample instead of 65,536 had you used all 16-bits.

So, recording at lower levels in a PCM system will increase quantization distortion. It is debatable whether, when recording at peak levels higher than -6dB, any clipping due to inter-sample peaks will cause more distortion than that due to the increased quantization distortion when going from utilizing the full 16-bits to only utilizing 15-bits. However, I would say that by the time you get to just the 16,384 quantization steps of a 14-bit system you are certainly losing out......

 

My experience has been, that the louder the CD, the more spikes you're going to see. If it hasn't already started clipping, it's probably pretty close to it.

 

If you see more than a handful of peaks at 0dBfs, chances are that digital recording is either: naturally clipping, or has employed the use of a limiter in some capacity. You just want to capture the entire dynamic range of the recording without burying any of the musical signal in the noise floor. You would also need just enough loudness to accurately capture the original dynamic range of the recording. These are the only parameters to use for a proper recording. Otherwise, you're not capturing any new information. You are only using more bits to capture noise, at the possible detriment of the signal's musical portion.

Although, if what you were saying is true, then 24-bit recordings might also require maximizing to 144dB peaks for optimal quality. Although, if you were to try to replicate this volume at home, you'd probably lose your hearing in a matter of seconds. Yet, I see so many 24-bit recordings today, with a theoretical 48dB additional headroom, recording to full scale. Some of these so-called high-resolution recordings are even using digital predictive limiters to achieve this -- so this bad precedent continues. I can only imagine one day in the future, having 64-bit recordings with peaks reaching in excess of 384dB.

 

Yes, that's a good point to make clear. A recording itself has a separate noise floor, and dynamic range. Just as a 16-bit signal will have its own inherent noise, or quantization errors, and potential dynamic range.

So if anyone's following me... you could decrease quantization errors -- and their perception -- by dithering, or noise shaping. But the only way to truly reduce quantization errors and increase accuracy is to increase bit-length; while recording at a louder volume will accomplish nothing more than make the musical signal and noise equally loud. And depending on your headroom, you can even run out of room to accurately represent the musical signal at full scale. I hope this is a little more clear.

 

Hi Tim,

An "exact result" is impossible if *any* differences exist between the two things you're comparing. Level matching to the degree that it is possible, is critical nonetheless (see next paragraph) - you can use something like the lead vocal as a reference.

With practice and a good playback system, differences in volume, large or small, can be heard as differences in volume. What is true is that one can be *fooled* into hearing "tone or clarity" differences when volume is different but these may in fact not exist on the recording. Tightly matched levels are the only way I know to tell for sure.

Barry
www.barrydiamentaudio.com
www.soundkeeperrecordings.com

 

Despite the comment above, I must admit that I thought that statement was correct.

My understanding is that compression REDUCES the "volume" of the loudest parts of the recording in order to reduce the difference between the quietest parts and loudest parts of a recording (the dynamic range).

As part of the process, the "volume" of the WHOLE recording is increased to compensate. If this part is not done won't the recording sound quieter?


TIA,
Stephen

 

Hi Stephen,

I would say the peaks won't be as loud, so in effect, yes, the *average* level will be quieter.

Perhaps a good way to simplify the concept is to say compression is used with the aim of making the quieter parts of a recording louder. Limiting is used to make the loudest parts of a recording quieter. Used in combination (often referred to in total as "compression" because that is what is being done to the dynamic range) and followed by gain "compensation", i.e. making it louder, the end result is a louder record.

Note that loudness can also be achieved by advancing one's volume control. Though this will more often than not result in more loudness with better sound quality than making the recording itself louder, the latter is the more common practice. (There are good reasons to do this with the volume control instead. See "Declaring an end to the loudness wars" at http://www.barrydiamentaudio.com/loudness.htm)

Barry
www.barrydiamentaudio.com
www.soundkeeperrecordings.com

 

when did all this start really happening ?

I just got an original 1991 Ozzy- No More Tears CD. this whole cd is really loud. I had to turn the volume down a few times from my normal listening level. even at low volumes it seems loud and distorted.

 

ok, checked your site.

"The history of the loudness wars can be traced back to the 1970s when vinyl mastering engineers started elevating the levels at the start of each side. This added to the initial impact of the sound as the record started to play. With vinyl, the amount of playback time available on one side of a record is directly related to how loud the record is cut. The louder the signal, the shorter the side. Since cutting the entire side at the elevated level would result in the available space running out before the music ended, the levels were cheated back down to "normal" after the first 30 seconds or so had elapsed."


ya know, I have heard this on the first song on a couple of my old lp's.

 

The following interview with Joe Gross who wrote an article for a paper called the Austin American called "Everything Louder Than Everything Else" examines this very question. There is a link to the original article on the site also

http://www.wnyc.org/shows/soundcheck...des/2006/10/02

 

Very helpful. Thanks for sharing this.

 

Actually, from my experience, compression used by itself with no other level adjustment brings up the average level, which makes it louder. Sure it brings down the peaks, but bringing up the lower level parts has much more effect on most recordings, so the end result is an increase in apparent volume. And of course, more room to make it even louder.

 

You mean fewer...

 

I missed that, thanks for catching it.

 

"If the remaster sounds louder, is it too compressed?"

If the compressed version sounds bad to you, it's too compressed. It it doesn't, it's not too compressed. Be wary of letting others tell you what to like. It's not math. It's rock'n'roll.