Slashdot Mirror
Is Louder Better?
GoodNicsTken writes "Rip Rowan over at prorec.com did an
analysis of 5 different Rush CD's released from 1984 to 2002. The results show a definite trend in the recording/mastering style from each album. Rip contends that louder is not necessarily better as the record execs believe. The artist however, is often left with little choice in the matter."
I always mix to -20 dBFS RMS because louder is NOT better. Headroom is much better.
Hopefully, surround music formats (DVD-Audio & SACD) will convince the tried & true engineers that they don't have to slam recordings at -0.1 dBFS like they've been doing with CDs.
A nice 24 dB of headroom allows for dynamic range in muxic, as well as loud transients. This is something you don't get when your music is an L2 brick.
Jory
that the louder speaker system always sounds better. They move a lot of expensive speakers like that.
It's Christmas everyday with BitTorrent.
Another great read here.
My server
More range is better, which can equate to louder "loud"'s, and softer "soft"'s. Just having the record be louder is going to sound like crap on really super-hi-fi systems that can pick up every little thing... you'll hear cats meowing in the studio, etc... I know from experience in the studio!
stuff |
> It's all about the radio. If your song has a lower volume than another one, it'll just sound Lame when it'll start.
> Of course all radios should/would/could normalize their playlists
I just wish they wouldn't blast the commercials out even louder than the music.
Sheesh, evil *and* a jerk. -- Jade
It's been covered in many web publications back in 2002.
Dynamic range problem is real though. This is why you whould avoid mainstream, "radio-ready" artists and bands. Another excellent reason to buy indie music.
If you want to see how bad the problem is, get yourself a copy of the latest Foo Fighters CD and listen to the album with decent headphones. (Grado sr-80/125 or Seinheisers of equal quality). It's just noise.
Radio does have automatic limiters. Listen to a rock station sometime, it all comes out about the same level, despite the different levellings of the individual recordings. This was in the article, btw.
-Looking for a job as a materials chemist or multivariat
Here's a link (reassemble as needed).
n g. html
http://www.personal.uni-jena.de/~pfk/mpp/clippi
Actually, Rush has been getting progressively heavier and louder since Roll The Bones. Boy, was that album a mess. Someone in another thread mentioned how Presto was too bright and lacking in bass. Well, for RTB, producer Rupert Hine overcompensated by clipping all the highs, too. Neil Peart does really cool stuff with cymbals. You'd never know it from RTB. The mid-range is so overwhelming, I can't tell the difference between my home theater setup and my clock radio when playing that CD.
Once they got out on tour for RTB, everyone told them how much better the new stuff sounded live. That was the end of Hine's association with Rush. They went back to Peter Collins, whom they had worked with through the '80s, for Counterparts. He brought in some guy nicknamed "Caveman" to engineer. The result was a very broad range of sounds. Some of the more complex arrangements, like Nobody's Hero and Cold Fire were quite clean and crisp, like '80s Rush. But heavier songs like Animate and Stick It Out have a dirtier, garage-band sound. IIRC, Geddy used an old amp with burned-out tubes to get that big, thick, heavy bass sound.
And it's been all downhill (or uphill, depending on your opinion of Rush's synth-happy days :-) ) from there, which leads us to Vapor Trails. They decided to take their time with that album, mostly because Neil had completely dropped out for a couple of years. They spent over a year in the studio. And when your as well-established* as Rush, the record companies don't meddle as much as they would with some flavor-of-the-month copycat band. So Rush certainly got the sound they wanted out of Vapor Trails. And if the results sound loud on the album, you should have heard it live. Damn.
*: Rock and Roll Speak for "old". :-)
This sig intentionally left blank.
Geddy Lee (who was the only band member present for the mixing) has said a lot of the clipping wasn't discovered until late in the process and he ended up trying to compensate for it in the mixing. That may well be why (as the article elaborates) the guitars, bass, and drums all clip at the same times; Geddy decided that the only way to cover up the drums clipping was amp the guitar so it was clipping.
Agreed with you on Vapor Trails. Best Rush album since Moving Pictures, and I might even say best since Hemispheres. Then again, I can't really rank Rush albums; there've been times that Presto or Hold Your Fire are what I'd call the best.
It's from a Saturday Night Live skit where Wil Ferrel is playing the cowbell for the Blue Oyster Cult. Christopher Walkin is the mixer, and he comes in and says something like "that was great guys, but it really needed more cowbell." Funny stuff.
I work with ADCs and DACs all day. Your first pass answer of 96dB is correct for DC characteristics. However, sinewaves introduce some differences.This ADC is a darn good performer. You will notice that the SFDR (Spurious Free Dynamic Range) is -101dB, while the THD is -99dB. Also, its Signal to Noise is -92dB, while the theoretical best is -98.08dB.
In fact, a small amount of noise actual can improve the signal representation! But that is a rather long discussion.
No, I don't trust in god. He'll have to pay up front, like everybody else.
Ok, here's what happens.
First off, while everyone bashes analog, the analog signal is what you want to measure. When you convert to digital, two things happen:
1) sampling in time. The sampling in time reduces the maximum frequency that can be represented to half of the sample rate. This is not a big deal, since you really can't hear much over 22KHz (for CDs) anyway. Just make sure that you have a good lowpass filter so that signals don't alias.
2) quantization. While the analog signal has an infinite range, you would need an infinite amount of bits to represent each signal as digital. While modern hard drives have gotten enormous, they still are not infinite. So, quantization restricts the valid levels to a finite number, and also restricts the minimum and maximum levels that the digital signal can represent.
Generally, for signals with a large amount of frequency content (what you kids call 'music' these days), there is a large amount of peaks. However, the peak is not what gives the impression of loudness. The effective amount of power, referred to as the RMS, is a better indicator of loudness than the peak levels. The peak of a sinewave is 141% of the RMS of a sinewave. More complicated signals will have a peak-to-rms ratio much higher (1000% or more).
So, when you are digitizing a signal, if you keep the input range of the converter constant and keep increasing the input signal amplitude, you will be increasing both the peaks and also the RMS levels. Once the peaks hit the maximum level that the ADC can represent, the peaks start getting clipped - but you can still increase the RMS. However, as you start clipping the signal more and more, you increase the amount of distortion in the signal.
No, I don't trust in god. He'll have to pay up front, like everybody else.
Jesus you are dumb. This isn't subjective. Take a look at the numbers (if you even read the article...) regarding the number of samples clipped by the power amp.
This is a very valid complaint (usually made by those within the industry). When you master each track so that they are all 'loud', you are essentially removing any difference in gain between these channels (also an objective measure). Thus, the music is percieved as 'better' for those people who have music systems incapable of producing the full frequency range at a relatively even sound pressure level. For those of us able to hear the difference, the music is far less dynamic than it should be.
wow, your reply was factually pretty off-base. let me school you on a few points, guy:
1) please for the love of god, SPELL THE WORD "NYQUIST" CORRECTLY.
2) latency is dependent on how large the input buffer is, and how often it is emptied. doubling the sample rate should, therefore, have the same effect as halving the buffer size, as it forces data to be clocked out of the buffer twice as fast. this is why the latency halves when the sample rate doubles.
3) let me quote you here.. "What does 96Khz do for you as a consumer? Well, with 44Khz audio interfaces, consumer level stuff is not always the most perfectly matched. As such, aliasing happens. Get 2% phase in a cap or resister and then multiple this over the signal path...Nyquest? Nyquest ain't gonna help you out when the alising is dipping into the audible portions of the sound. Remember 44khz gives you 22 IF EVERYTHING ELSE IS PERFECT." --- that was a collection of pseudo-babble and technical non-sense. total shit. sorry, guy, but i had to call you on that. aliasing is a phenomenon wherein the input signal is sampled at a rate unsuitable with the nyquist cutoff- in that instance the sample is "faked", and the sampled signal that results is of a lower frequency (but usually harmonically related to) the source sound - it is a mis-representation of the input signal, like jaggies in a computer generated picture. it has absolutely nothing to do with phase smearing, which is more of a time-based issue between two or more distinct signals canceling each other out at an audible rate.
4) I happen to own a Kurzweil K2vX (basically a K2000 with the sampler module, orchestral and contemporary ROM blocks built in). once again, you've got a logic flaw- in actuality, the kurzweil samples MANY of it's instruments at MUCGH LOWER RATES than 48k. SERIOUSLY. just click the "MASTER" soft button, then SAMPLE, then check out some of the built in samples-- the piano is sampled at about 22k. this is because the most relevant portions of the instrument in question need only to be sampled precisely. your kurzweil sounds good because of the relatively clean internal processing and D/A converters that are built into the keyboard (MUUUUUUUUCH better than the coverters built into other keyboards of the same ilk).
5) in theory, yes, 96k was made to correct issues with regard to aliasing. however, the cost of incorporating this high frequency sampling rate is a high one-- companies whose converters are cheaper will be unable to accurately pace themselves during the sample process (perhaps a cheap crystal) and will produce much more jitter-- each sample of a 96k wave is supposed to be evenly spaced.. like a drummer, it has to "keep the beat". well, imagine if the drummer was epileptic. that's what happens with cheap converters). so the introduction of 96k products in the consumer field is a nightmare for the time being.
6) with regard to dual inline converters that sample at 44.1, but "double up"-- if done correctly (meaning, if CLOCKED correctly.. see above for the jitter issue), than there is NO DISCERNABLE DIFFERENCE in the QUALITY of the sample-- if the converters (2 of them, remember?) are assigned as A and B, and each fires at the correct interval (A B A B A B A B), than they can "interleave" a 96k signal with NO PROBLEMS and NO DEGRADATION OF THE SIGNAL ITSELF. i have no clue where you got that idea.
7) you made a bit of a joke regarding word clock.. but i think you may really wish to research the definition, because perhaps buying one of those would improve the sound of your recordings FAR MORE than dicking about with "nyquest" theorums and fantasies.
buh bye.
First, minor point - 96 kHz sample rate gives you 48 kHz theoretical bandwidth - Nyquist frequency is exactly one-half the sample rate. Not 44 kHz.
So, here's the real point. Higher sample rates allow you to pass higher bandwidths through the ADC (and theoretically through the DAC). However, those higher bandwidths get shrunk when they hit the amplifiers (consumer, even pro-sumer products rarely use high-bandwidth amplifiers), further shrunk when they hit your speakers (know how many speakers will produce anything above 22 kHz? Simple answer - not yours. Unless you spent several tens of thousands for your high-efficiency ribbon tweeters), and even further shrunk when they hit your ears (though, some people can indeed hear above 20 kHz. I can hear up to 26 kHz, but then, I've never been to a concert or club without earplugs). The acoustic coupling they claim happens in mid-air (which is true - put a 40 kHz tone and a 44 kHz tone out - you get difference tones at 4 kHz, 8 kHz, 12 kHz, etc.) only occur if your speakers can get those high-frequency tones out in the first place... which they probably can't.
So, what's the real point?
Better anti-aliasing filters on the ADC side. If you are sampling at 44.1 kHz, under the Redbook standard, you have to be down 40 dB at 22.05 kHz. However, you want to pass 20 kHz with no filtering, which means your filter has to be as brickwall as possible (about -200 dB/octave... sheesh!). 3rd order filters can't even do that properly, so most anti-aliasing filters start rolling off around 16 kHz, some even earlier (especially in digital video cameras. I know, I've tested 'em).
So instead, set your filters to be down 40 dB at 48 kHz ('cause you're sampling at 92 kHz). Now your filter only has to be about 36 dB/octave to pass 20 kHz untouched, and that both increases your flat bandwidth and decreases phase distortion (the -3 dB point is a 45 degree phase inversion, and every 3 dB after that is another 45 degrees of delay).
That's why sampling higher improves things - no brickwall filters.
-T
Sorry, but you seem to misunderstand the situation. What the sampling theorem tells us is that as long as the input signal is bandlimited to frequencies below one half the sampling frequency, it can be reproduced exactly by the DAC. (nb: this is for samples which have not been quantized) The reconstruction is taken care of by what is commonly known as a reconstruction filter. You are correct that the samples of a sine wave near the Nyquist frequency will look like a triangle wave, but once passed through the reconstruction filter what comes out is the original sine wave.
Note that the requirement that the signal be bandlimited means, for example, that one can not have as input a triangle wave near the nyquist frequency, because, as you correctly stated, a triangle wave of such a frequency contains harmonics which are greater than the nyquist frequency. Typicaly, they would be removed by an antialiasing filter at the input to the ADC.
Of course, I realize that it is not possible to implement an ideal reconstruction filter, and also that quantizing the signal introduces distortion which it is not possible to remove.
Many different people do 'Mastering',. and each one does it in his or her own special way that they want to convince you is better than everyone else's. The main steps in mastering are eq, compression, and level matching. Very few cd's are printed with clipped samples, because this data is out of range, it can't be reproduced, if the playback of the cd results in a clipped waveform it could be either a perfectly recorded clip or it is a failure of your hardware to faithfully repoduce the waveform as it is encoded on the disc. Most generally it is the latter.
When finished tracks are sent to be 'Mastered', they are usually compressed a little bit, or a lot, depending on the taste of the Mastering Engineer. Compression in this case doesn't refer to encoding audio in a compressed format, rather a compressor is a dynamics processor, with it you can set a threshold above which the sound will be modified based on a ratio like 2 or 3 to 1. So for a 2 to 1 ratio any sound that is above the threshold will be reduced by half.
This was initially done back in the old days when you had at best 45 dB of dynamic range to work with on your recording medium, a very noticable noise floor, and material with a dynamic range of 120 dB (Live Rock). Obviously you can't stuff 120 dB into a 45 dB (cassette tape(if you are lucky)) dynamic range, So the material was compressed to fit within the dynamic range. Also because of the quality(lack thereof) of consumer audio equipment and the previously mentioned very noticable noise floor, most music is compressed into the top 3-5 dB of whatever medium it is recorded on.
Nowadays, we have a playback medium with a 96 dB dynamic range and close to a 96 dB noise floor, but because people got used to the way it used to sound, they want to keep hearing it that way. Pretty much the only recorded materials that truly benefited from the increase in dynamic range allowed by CD's and digital recording are orchestral works, and the people that listen to these avidly, and care about the recording truly reflecting the performance, still want more!
The other aspects of 'Mastering' are a great deal more subtle, equalization and level matching between tracks are things that most people do not notice unless it is done badly. At the end they turn the result up to the top of the mediums allowable dynamic range and start printing tens of thousands of them at a few cents apiece.
If you think a cd has clipped samples recorded on it the best way to check is to rip the track off the disc into a PCM (Non Lossy, Non compressed, Non MP3)format at 16 bit/44.1 (Redbook native format) and look at the samples in question with a wav editor. If you have blown up the waveform to the point where you can see a single sample, and the tops of the waveform are at the cieling and flat, then complain to the recording engineer, because it is probably his fault.
BTW make sure it is a clean non scratched cd, any unrecoverable data loss can appear as a clipped waveform, and is heard as such depending on the smoothing filter on the output side of your cd player.