Thread: [OT] On compression and clipping

Posts: 66
Page: 1 2 3 4 5 6 7 next

Post by The Seventh Taylor June 29, 2011 (1 of 66)
I know a fair deal about SACD technology but I want to know more about the art of mastering, which allegedly makes all the difference in some cases.

Hence my question: Would it be fair to say compression is alright as long as clipping does not occur?

Post by seth June 29, 2011 (2 of 66)
The Seventh Taylor said:

I know a fair deal about SACD technology but I want to know more about the art of mastering, which allegedly makes all the difference in some cases.

Hence my question: Would it be fair to say compression is alright as long as clipping does not occur?

I don't claim to be an audio engineer, but here's my understanding of this.

Clipping occurs during the recording process.

The avoid clipping, the easiest thing to do is find the loudest passage in the music, set the levels so clipping doesn't occur at that passage, and leave the levels there for the whole recording.

Now this will result in the recording having a wide dynamic range where large parts of the recording sound relatively quiet. A lot of people complain about this kinds of sound because they want everything to sound loud, so they have to turn the volume up in the quiet parts and then down in the loud parts. To appease these people, in the master process the dynamic range will be compressed so to narrow the range between the quietest and loudest parts of the recording -- everything will sound relatively loud. This increasingly happens with most pop recordings because people are now mostly listening to music with portable devices where quieter passages can be inaudible because of the crappy headphones and ambient noises.

The other solution to this "problem" is to ride the gain during the recording process. In the loud passages you lower the levels, and in the quiet ones you boost them. It's hard to pull this off without resulting in at least a few distortions. Decca did this more than any other label in the 1960s. RCA did it too, and you can hear it on some of the Living Stereo SACDs, such as the Stokowski show piece disc. And with multi-mic'd recordings, you can ride the gain on various instruments, also referred to as spotlighting (you can hear this in lots 1970s recordings).

The audiophile thing to do is set the levels to the loudest part and not touch anything (but maybe slightly compress in the mastering). This is why there are so many SACDs where people mention that the music was recorded at a low level, and the volume level should be set pretty high.

Post by jazz1 June 29, 2011 (3 of 66)
seth said:
The audiophile thing to do is set the levels to the loudest part and not touch anything (but maybe slightly compress in the mastering). This is why there are so many SACDs where people mention that the music was recorded at a low level, and the volume level should be set pretty high.

What I do not understand is that SACD dynamic range is 120 db they should be no reason when recording to set the gain "a la BIS"
On RBCD the dynamic range is only 100 db, the SACD's extra 20 db's should solve any possible problems.

Post by AmonRa June 29, 2011 (4 of 66)
Clipping can happen also after recording. When several recorded tracks are mixed together in post the total volume increases 3 dB for each doubling of number of mixed tracks. If the master level is not lowered the master track might clip. Adding various effects like reverb also raises the levels, up to 10 dB in cases.

Measuring levels and avoiding clipping is very straightforward in PCM, where it is easy to just check the value of the highest level sample and adjust accordingly. Usually the final master is checked for the peak, then adjusted so that the max level is just shy of 0 dBFS, maybe at -0.3 dBFS. This makes the absolute peak as loud as safely possible (MP3 compression of 0 dBFS peak material might distort due to the nature of the compression algorithms).

DSD behaves differently, basically there is no theoretical mathematical limit for peak levels, as just piling more ones than zeros, or vice versa, into the file long enough time would result in an awfully loud signal*. There are standards for the maximum levels in SACD masters, though, expressed in output voltages if I understand correctly. Some SACD disks do not abide by these rules and can distort players with DSD DACs and analog stages (maybe also amplifiers further in the chain might be overdriven) which have not enough extra headroom for this.

There are two main schools of thought what comes to classical recording, purist and practical. A purist engineer would record the whole dynamic range and not compress. If the listeners can not hear the quietest passages they just do not have good enough system or listening conditions, but they are offered the best recording possible. A practically inclined engineer would gain ride (in post) and/or compress the signal slightly to bring up the quiet details so that more people could actually hear it. This means that that the signal is not true to the original and that the background noise levels are also raised slightly. This is a disservice to those listeners with truly good systems and listening rooms.

*) the number of zeros and ones in a DSD file should be the same within a relatively short period (like maybe 200 ms) to avoid low frequency voltage drift and DC component, which could misalign woofer cones and put unnecessary strain on amplifiers.

Post by jazz1 June 30, 2011 (5 of 66)
AmonRa said:
A purist engineer would record the whole dynamic range and not compress. If the listeners can not hear the quietest passages they just do not have good enough system or listening conditions, but they are offered the best recording possible. A practically inclined engineer would gain ride (in post) and/or compress the signal slightly to bring up the quiet details so that more people could actually hear it. This means that that the signal is not true to the original and that the background noise levels are also raised slightly. This is a disservice to those listeners with truly good systems and listening rooms.

Assuming that an average loudspeaker sensitivity is 88 db are you saying that we need a 1000 watt of amplification to benefit from DSD full dynamic range??
Just curious!! If so then the purist engineer is catering for very few listeners.

Post by AmonRa June 30, 2011 (6 of 66)
More or less yes, even 100000 W.

Post by bissie June 30, 2011 (7 of 66)
jazz1 said:

What I do not understand is that SACD dynamic range is 120 db they should be no reason when recording to set the gain "a la BIS"
On RBCD the dynamic range is only 100 db, the SACD's extra 20 db's should solve any possible problems.

Eeeehh,
exactly what do you mean? We do it exactly as seth described, find the loudest part and let the rest be what it will be, leaving the dynamics question to the composer and artists. We do not compress even slightly in the mastering process.

This is what I call real truth.

Robert

Post by Sam June 30, 2011 (8 of 66)
jazz1 said:

Assuming that an average loudspeaker sensitivity is 88 db are you saying that we need a 1000 watt of amplification to benefit from DSD full dynamic range??
Just curious!! If so then the purist engineer is catering for very few listeners.

The benefit of the DSD dynamic range is simply that the recording process doesn't add its own noise. It has nothing to do with reproducing recordings with a 120db range - such a thing would not be music and it would damage your ears if you could reproduce it. Think jet engine.

Post by LivyII June 30, 2011 (9 of 66)
bissie said:

Eeeehh,
exactly what do you mean? We do it exactly as seth described, find the loudest part and let the rest be what it will be, leaving the dynamics question to the composer and artists. We do not compress even slightly in the mastering process.

This is what I call real truth.

Robert

This is exactly how it should be done.

"Riding the gain" is nowadays done with software, either during or after the recording process. This may be done by some classical music companies, but is universally done in rock music.

In rock music, people use limiters to set the peak levels between certain dynamic intervals. Most of the time this is done during either the mixing or mastering process. In the 1980's, CD's typically had a maximum dynamic range of 18 dbs.

As time progressed, rock record companies found that they could make the CD louder by narrowing that dynamic range and that, apparently, certain listening groups preferred the sound of such CD's. This led to what is commonly referred to as the "loudness wars", which many people believe reached their peak in the early 2000's, when CD's came to be mastered with maximum dynamic ranges of 3-6 dbs.

In the past 10 years or so, there are many in the rock audio community who have reacted vocally to the poor sound quality of these CD's. They sound like static sometimes, so loud and undifferentiated. That said, not all are sonic disasters, but many musically fine rock offerings are undermined by their sound. The ironic part is that we now have better technology than ever, yet many rock recordings sound worse than ever.

In the classical world (and in jazz and rock to a lesser degree, too), I think it makes little sense to compress music at any stage. No matter how you try to justify it, you are changing the sound of the music. The purpose of the recording is to render as faithful a reproduction of the original as possible. I guess that puts me into the "purist" camp! Of course, I can remember the days of "Loudness" buttons on stereos, which allowed the listener to decide whether a small amount of compression was necessary to improve how something sounded. This was a much preferable alternative to what we usually get today.

Post by seth June 30, 2011 (10 of 66)
jazz1 said:

What I do not understand is that SACD dynamic range is 120 db they should be no reason when recording to set the gain "a la BIS"
On RBCD the dynamic range is only 100 db, the SACD's extra 20 db's should solve any possible problems.

No matter the range of your playback medium, the gain still needs to be set relatively to the loudest passage in the music.

There's a greater philosophical issue here, which is that should the dynamic range reflect live sound, or should it accommodate music systems and playback environments.

A few thoughts on this:

One of my music professors said what separates good from great performers is their ability to play quietly, and have a whole range of quiet. You know if a performer/ensemble is great if the people seated in the cheap seats at a concert hall are leaning forward to hear the music.

Another one, Neal Zaslaw, said that the #1 complaint first time concertgoers have (as well as people who primarily listen to recordings), is that the performance wasn't loud enough. The compression on recordings has caused us to believe that live music should be much louder than it really is.

I was listening to some 1960s Columbia/CBS recordings, which are a favorite whipping boy of listeners. I think one of the reasons people write-off these recordings is that the engineers appeared to have set the gain to avoid distortions at all costs. This resulted in a duller sound compared to most other labels that either got the gain as close as possible to clipping, or road it, which resulted in more vivid sound.

Page: 1 2 3 4 5 6 7 next

Closed