If the remaster sounds louder, is it too compressed?

I'm just waking up to the compression issue; no wonder there is an LP revival. I wonder how much of the benefits of SACD and DVDA is simply down to the fact that audiophile releases tend to be less compressed?

But that isn't my question. What I'd like to know is: if a remaster sounds markedly "louder" than an older release, is that a sure-fire sign of excessive compression? Example: I have the Last Waltz on an old 2-CD issue, and the more recent Rhino box. The Rhino sounds a lot louder. Does that indicate that all the "audio restoration" work is marred by compression?

Tim
10 best albums ever

 

No it doesn't automatically mean that it's too compressed. Some CDs, particularly '80's, are conservative in levels. So it would, in some circumstances, be possible to be a bit louder without needing to compress or causing audible clipping etc. It's only a good bet if it's super loud.

Welcome to the forum

 

Thanks

So I presume an interesting test would be to rip the old and new versions to hard drive, normalize them both in something like Audacity, and compare the results?

Might try that sometime.

Tim
Why I hate compilations

 

Hi Tim,

You can do that of course. Just know that once you "normalize" (should be called "ABnormalize") you'll no longer be comparing the sound on the originals because you'll have altered both and thrown away low level data. (This will happen with any change to a 16 bit signal; the change lengthens the digital "word" but you can only hold 16 "characters".)

The only way to compare the actual CDs is to level match in the analog domain on playback... even if that means twisting the volume control as you switch between them.

Barry
www.barrydiamentaudio.com
www.soundkeeperrecordings.com

 

Increased amplitude [loudness] does not have to mean that compression was added during mastering. It can mean that the very highest peaks are hitting 0dbu. When I re-master tracks at home, I try to use the all 16 bits of the medium.

Bob-

 

This is problematic of course. You're not going to get an exact result; and if in fact one is more compressed than the other, there would be no such thing as "exactly the same volume"; it would vary during the course of the piece.

Since small differences in volume can be heard as differences in tone or clarity, it's hard to make an accurate A/B comparison.

Tim

 

Just as a little aside, there's no such thing as maximizing volume to deliver a full 16-bits. You just want the volume of the lowest signal to be louder than the noise floor, and let the dither do its job. In fact, dither allows us to hear below the least significant bit. Otherwise, you'd just increasing the potential of having the analog portion of your DAC overdriven. Whenever I record, I try to find a happy medium, inbetween the noise floor, and peaks hitting 0dBfs.

 

Tim, you may want to see who you were replying to. I think he knows a thing or two about compression. You are just waking up to it.

If your new Band disc is only louder than the original, and it's not compressed, what will happen if you adjust the volume?


Welcome to the forum

 

I'm glad to finally see a pro mention this detail. In other words, if one must normalize by eather peak style or RMS style, do it at a high bit-depth!

 

If you record too low, you will not be recording 16 bits. It's more like 14-bits. I may be mistaken, but Bob Kats has something to say about this.

When recording in 16-bit, one should strive for recording as loud as possible without exceeding 0FS.

 

You'll only be using 14-bit in recording that low because that's all the dynamic range the signal will require.

Yeah, this is a common misconception about digital audio, probably the number one misconception. I think a lot of audio engineers believe this too. And it would be perfectly true of an undithered signal. But dither and digital audio pretty much go hand in hand nowadays. You can test the concept by taking a brief audio sample, fade it down, and then save both a dithered and undithered file. You'll find the dithered file will make a smooth transition to the least significant bit, the undithered file will contain quantization errors, and not be continuous.

You simply want to record the lowest recorded signal above any inherent noise floor. By boosting the volume further you'd also be booting the noise floor. In fact, you'd simply be giving more bits to the noise itself, and potentially less to the signal, in slightly extreme cases. You also run the risk of destroying the original dynamic range. Not only is it unnecessary, but depending on the linearity of your converters, it might even be deleterious to the original signal.

 

I'm not sure you are understanding what dither is here...

 

Not only do I understand how dither works, but also a fair deal about digital theory. A good book that covers this in pretty good detail is Nika Aldrich's "Digital Audio Explained: For the Audio Engineer." I think you'll find him an even greater authority than either you or me. I definitely defer to him on the subject.

 

Yes. Modern remasters are almost always more compressed in order to be louder. They are also generally mastered without concern for intersample peaking as we talked about in another recent thread and hence the peak level is set much higher, usually near max at 0dBFS or 100%. Normalizing the peaks on both to 100% (0dB) for comparison is a good test for compression. The new one is probably already there, and the old one probably won't change that much since the peaks weren't limited back then like they do today. Not sure, but I think Barry meant that it isn't a good practice to normalize CDs for best sound quality and ultimate listening pleasure since it requires better software than most of us have on our computers to get the best results. Performing the operation at only 16-bit precision isn't very accurate, and performing at higher resolution requires knowledge on how best to get back to 16-bits.

But some CDs were mastered at pretty low levels in the past, and sometimes it was just because of a couple peaks in the music that could be squashed a bit to make it much louder without really sacrificing much sound quality.

BTW, love your 10 best list. All of them favorites of mine too. Just made a post the other day on another board about that Neville Marriner Four Seasons ... very nice ... http://forums.audioreview.com/showthread.php?p=159764#post159764

 

Dither doesn't have as big a role as your writings seem to give it. All it is is white noise added to the bit word to minimize truncation distortion. NOISE SHAPING is what increases the percieved dynamic range and S/N by moving the noise around to parts of the frequency specturum deemed less sensitive to the human ear.

The issue of dither has nothing to do with digital recording levels. If you record at low levels in 16-bit, you risk a higher noise floor that is NOT tamed by the use of dither.

 

Sure, but what if it is compressed?

I am not pretending any special knowledge here, just thinking aloud.

Tim

 

I know. Which version of "The Last Waltz" do you prefer when you A/B them at the same volume?

 

Actually it is a bad example, because it has been remixed. I mostly prefer the remix, but not quite always.

Tim

 

How does one get a "figure" for the average volume of a recording?

Does that question make any sense? Is it possible to do such a thing?

TIA,
Stephen

 

I think most of the programs use a "sliding window" whereby the max level in each of the small windows is stored, and then an average is computed. Not sure if there is a standardized method. If it is true rms then they use the silent parts as well, but many of them ignore values below a certain threshold to get a more realistic avearge.

I generally use Audiograbber for average level checking, and it says in the help text that it uses a 133 ms window, but there are many others. If you've looked at many modern CD waveforms with the buzz cut peaks, or seen them posted here, then you mght be thinking to yourself that a window that wide would probably give an average almost the same as the peak in many cases, and you'd be right. That's exactly what happens with most modern CDs since they don't have very much, if any, variation in the peak level due to extreme compression, and soft limiting, and hard limiting and even clipping employed so the average level can be boosted as loud as possible, usually above the 90% level. For a more accurate result, the program would need to do a statistical analysis using all the samples, which would take quite a bit more time and processing power than the super quick results from a shareware program like Audiograbber

 

Remeber guys, compression alone does NOT make anything louder. The way they make these CDs ouder with compression is that they boost the level. To prevent the higher peaks from exceeding 0FS, they employ the use of a limiter. The brick-wall limiter is what creates that "buzz-cut" look.

You can use a compressor in mastering and stil have the wave look like no compression was applied.

 

Grant - In an earlier post, in reference to dither, you mentioned "All it is is white noise..." -- wouldn't that be an oversimplification that's a bit too dismissive? There are several kinds of dither algorithms that are "adaptive" and use triangulation. Granted, like you said, the most basic definition would be that dither is random noise that sacrifices the signal to noise ratio to better DAC linearity, but, making such a general statement can perpetuate misconceptions. BTW, I thought this was an interesting read on the role of dither in recordings using higher than 16 bit resolution:

http://www.stereophile.com/features/705dither/

 

I know Steve has discussed in the past using judicious eq. and/or analog compression over digital compression in mastering to increase the overall volume on occasions.

 

I simplified it for the purposes of showing that dither has nothing to do with the topic at hand.