|
|
beerboy 發表於 2012-10-17 10:28 
呢遍仲長......但係好正!!!
30 Years of Perfect Sound Forever
Even though some people can hear to 0 dB SPL, we're always hearing background noise if we shut up and listen. It takes a lot of money to build an NC 25 or NC 15 studio, in other words, a recording studio with about a 15 dB or 25 dB SPL background noise. Even in an NC 15 studio, 105 - 15 = 90 dB SPL, well within the range of real 16-bit systems, if you record it well.
Supposing we recorded on the moon in a pressurized tent with no background noise? Well, the self-noise of most recording studio microphones is about 16 dB SPL equivalent input noise, or in other words, microphones aren't any quieter than about 16 dB SPL anyway.
16 bits was chosen because it has more than enough range to hold all music. I know; I was doing 16-bit recording back in 1981 before the CD came out, and my recordings would have their levels carefully set so the loudest peak of the entire concert hit about -3 dB FS, and leaving it running after the audience left and the hall was empty, you can still bring up the playback gain and hear a perfectly silent recording of the air conditioning noise in the hall. The world just doesn't get quiet or loud enough to need more than 16 bits as a release format, if it's recorded well.
There is no such thing as a real 24-bit audio DAC or ADC. Look at the specs, and you'll never see a 144 dB SNR spec; all audio 24-bit converters do have 24 bits wiggling, but the least few LSBs are just noise. There is plenty of 24-bit and higher DSP, which is good to keep the 16-bits we need clean, but you're never getting 24 real bits of analog audio in or out of the system. It's a good thing you can't; 140 dB SPL is the threshold of instant deafness, and if you lift the gain enough to hear a real 24-bit noise floor at say 20 dB SPL in a very quiet studio, maximum output would be 20 + 144 = 164 dB SPL, or 4 dB over the threshold of death. Yes, 160 dB SPL kills.
But wait, there's more. 98 dB is the theoretical SNR. With dither, we still can hear pure undistorted signals down into the noise for at least another 10 or 20 dB. While a typical real-world 16-bit system's SNR might be 92 dB, we can hear tones down to -100 dB FS easily. That's over 100 dB of dynamic range in real 16-bit systems.
There's even more than that! By the 1990s, people learned how to "noise shape" the dither to push it up mostly to 15 kHz and above, so it became much less audible, but just as effective as regular dither. These systems made the noise much less audible. These systems are also called Super Bit Mapping (SBM) by Sony and UV22 by Apogee; they claimed 22-bit effective SNRs with 16-bit systems. They didn't really work that well, but they did make our 16-bit system even better than it was. These clever sorts of dither are still used today for 16-bit releases.
That's right: done right, 16 bits is way, way more than enough for any sort of music. Once you've heard it done right, you'll realize any noise you here out of a CD is due to sloppy recordings (usually sloppy level settings someplace in the chain), not the CD medium itself. GIGO as you computer guys say.
41.1 ksps is plenty
44.1 kilosamples per second (ksps or "kHz") is plenty for the 20 kHz audio band today, but it was much tougher in 1982.
In 1982, it was difficult to build great analog (LRC) anti-alias filters that could pass 20 kHz well and stop anything above 22.05 kHz equally well for both recording and for playback, which led to the creation of companies like Apogee whose first products were improved versions of these filters.
In 1982, processing all this data was a bear; PCs barely ran at about the same clock rate as the data rate of a CD!
In the next few years, oversampling converters and DSP made these anti-alias filters excellent and inexpensive, so the problem of iffy filters went away. We record and play at higher sample rates in production, and to release at 44.1 ksps isn't a problem. In fact, I've measured my iPod with flat response all the way to 22 kHz from 44.1ksps sources; the old filter problems are long gone.
DC Response
44.1 lets the CD do a perfect job reproducing the ultrasonics we can't hear, and even better, it reproduces down to DC to reproduce the bass we can hear.
While LPs rarely had much going on below 50 cps, and whatever there was was usually summed or mixed to mono, CDs cheerfully record and reproduce all the way down to DC, at full-scale, without distortion and in full stereo.
For music, this means for the first time at home we can reproduce all of the bass, al the way down to music's deepest 16 cps C0.
LPs had real problems with deep bass. It would tend to make the needle jump out of the groove and skip, and rumble muddied whatever was there. Worse, with LPs, playing them in the same room as the speakers could lead to destructive feedback where the sound waves vibrated the record and it got picked back up by the system!
With CDs, we have unlimited deep-bass potential, an indeed, many music CDs are loaded with loads of deep sub-bass in stereo, something LPs never were able to do well, if at all.
With CDs, not only do we have response to well below the music range (16 Hz or cps), we get it with no phase or group delay abnormalities so what we hear is completely faithful to the original.
Heck, analog tape never could do this either. Only digital is happy recording and reproducing down to 16 cps C0.
Uncompressed Data Capacity
When the CD came out, it was like something from another planet. No one outside the recording industry had ever heard completely silent undistorted recordings. LPs had not only clicks, pops and scratches, but they also were usually loaded with distortion (we used to tape our new LPs so they wouldn't get worse), they were rarely pressed on-center so the pitch varied as the disc rotated, and warps made our woofers flutter like crazy. LPs were nasty, compared to pure live music. In radio, "cue burn" was the first few seconds of grunge you'd get from back-cueing the same record 100 times to find its start.
in 1982, no one except computer nerds had computers. It wasn't until the late 1980s that hard drives were seen commonly, and then they were only 10 megabytes, an astounding number. By 1985, computers still only used 5-1/4" floppies, which held only 720 kilobytes if you had the HD ones. Microfloppies, the 3.5" kind with two sides, were crazy stuff when Apple first used them on a computer in 1987. They were small, tough, and held an amazing 1.44 megabytes. Even until about 1992, only engineers had computers at work.
The CD in 1982? It held an unfathomable 650 Megabytes, or as much as 65 hard drives would be able to hold three years in the future! Even in 1985, no one could afford a 10 MB hard drive. I worked in defense in 1985, and we did our calculations on computers with dual 5.25" floppies; no hard drive. That's why hard drives are called the C: drive; the A: and B: drives are your two floppies: one for the program, one for your data.
Anyway, CDs were always laser rocket science. It wasn't until about the year 2000 that anyone could afford a CD burner.
100.0000% Bit-Accurate
Some people forget today that the CD is a 100% bit-accurate medium. It puts the same data on the disc in multiple places, and using various kinds of error correction and detection and eight-to-fourteen modulation, so no matter what happens, you get everything back exactly as it was recorded. You can even drill a small hole in an CD, and the data will be recalled with 100% accuracy, since the CD player simply pulls the data from different sectors.
Today, there is still nothing better, and nothing even as tough.
SACD?
The SACD was a marketing ploy around 2002, but its huge problem is that SACD puts out a ton of ultrasonic hash (noise) even when it's working perfectly. SACD player instruction manuals warn not to crank the levels during silence, because this ultrasonic noise might blow your tweeters! CD players haven't needed sharp 20 kHz anti-alias filters for decades, but SACD players need them today, but don't have them!
Here's an anecdote about how bad is the noise out of a good SACD player. I was playing around dubbing to cassettes, and something sounded horrible, as if the tape was all twisted and garbled with Dolby, even Dolby B. A little red light went on in the back of my head, and I said "No, it couldn't be this bad," and hit the multiplex filter on the cassette deck. That cured it. In all my years as an FM radio station chief engineer, I never found any FM tuner so bad that it didn't filter the 19 kc pilot well enough to need the MPX filter. Never. But welcome the SACD, and lo and behold, its output is laced with so much ultrasonic crap that I needed the MPX filter to get Dolby to track. Horrendous! My iPod is much cleaner (and pretty darn clean, too)!
Another practical reality of why SACD isn't all it could be simply is that the tools used to produce recordings don't come in DSD (one-bit direct-stream-digital) versions. DSP is based on PCM, and all the authoring tools we use to create recordings all work in PCM. Unless you're making a two-mic recording direct from preamp into SACD encoder, all your SACDs come from PCM anyway.
HDCD?
another great idea, the HDCD added coded noise into the audio as well as dynamic compression.
On most CD players, HDCDs didn't sound that much worse than regular CDs.
When you played them on on an HDCD player, the HDCD player tried to read the noise code, and use that to repair the dynamic compression, adding the equivalent of more bits of dynamic range.
THis was a clever idea, except that few HDCD players were ever available, and today if you have HDCDs, they can't be decoded on today's CD players, thus sounding a little worse than CDs. Whoops, and as we already know, 16 bits is more than enough for a release format.
Computer Audio?
But what about people today sharing files and pumping them into fancy outboard DACs from their computers? That can work great, but there are a few reasons why a good CD player can be better than a great outboard DAC:
1.) Jitter
A CD player has no measurable jitter. Data is read and corrected from the disc, and the data fed to a first-in, first-out buffer. Data is clocked out of the FIFO into the player's own DAC at the exact rate of the quartz-crystal oscillator of the CD player. The disc's rotational velocity is varied in a closed-loop to feed the FIFO exactly what it needs, all controlled by the player's one low-phase-noise and low-jitter quartz crystal oscillator. The only jitter is the residual of the quartz oscillator, which actually has less phase noise (jitter) than an atomic standard!
When you use an outboard DAC, unless you're a professional and have a Word Clock or other separate Sync output fed to your DAC for the clock signal, the DAC has to guess at reconstructing the clock signal from the audio data it's fed via TOSLINK or USB or RCA or AES. (those interfaces carry only data, not clock.) Noise added to the natural high-frequency attenuation in any length of cable adds jitter to the recovered clock, and as my own tests have actually shown, there is a measurable increase in measured jitter actually seen on the analog outputs of outboard DACs versus direct from a one-box CD player. This tiny amount of jitter isn't significant, but seeing how there is a cult of whackos who worry about things that are far less significant, the fact that I can measure and show jitter picked up in a top-notch DAC at its analog output under very good conditions impresses even me.
2.) Ground Loops
If you use an outboard DAC, use the optical TOSLINK connection. If you don't and you take a digital input from a computer via USB, Firewire, RCA or any other wire, you're now coupling any ground noise from your computer's digital circuits into your audio ground.
As a guy who used to design ADCs, DACs and DSP systems, we do everything we can to keep the digital hash out of the analog circuitry. Never, ever connect the two grounds together at any more than one point, and that one point will probably be your power outlet at the wall. Don't go using USB or similar and connect your computer's ground to your audio system!
3.) Noise
Most computers have fans or hard drives that make audible noise. Most CD players spin silently.
4.) Overload Handling
This is a potentially really nasty one that needs more research. In the beginning, CDs were cut with 0 VU at -20 dB FS, in other words, there was plenty of headroom. The world's first released CD, Billy Joel's 52nd Street, never even hits full scale, and it sounds great.
Once everyone had a CD player in the 1990s, some bonehead got the dumb idea that if he made his CD sound louder than the next guy, that people would like the music better. Dumb idea, yes, but as of today, most popular CDs have so much dynamic compression applied that they sound as bad as radio: one big long 100% modulation wall of boring. Jazz, classical and a very few acts like Peter Gabriel's latest still use all the dynamic range, but just about everything else today is squashed to death to put everything at 100% loud. Today, most CDs only use the top couple of bits!
CD's astounding ability to work at 100% modulation without distortion, unlike LPs that became more distorted as they got louder, works against the CD today where most pop producers hit the disc as loud and as hard as they can.
Worse, CD mastering continues to get worse in its attempts to get louder, and many CDs use another radio trick, composite clipping. Yes, the waveform is boosted even more and the peaks of the waveform are clipped, and since most people won't know, helps squeeze another dB or two of level onto the CD.
Today, some albums have levels when measured with a Tektronix 764 that exceed 0 dB FS! How do they do this? Well, levels are calibrated to read 0 dB FS for a sine wave, but when a proper meter like the 764 is used that properly reconstructs the actual audio waveform digitally as opposed to simply looking at data stream values, clipped signals approximate square waves, and approach +3 dB FS!
This is all fine and dandy played on a CD player, which simply reproduces the music, clipping and all, as recorded.
It can wreak Hell when you start ripping that to AAC for your iPod, or play it on an external DAC, most of which aren't designed to have enough headroom to reproduce the crazy transients that are there with 100% clipped signals. Most audio DSP norms were created back before producers started putting such nasty signals on music CDs.
As my CD player and outboard DAC tests have shown, weird things happen when playing extreme square wave tests. Outboard DACs for whatever reason often lack the headroom in their DSP for this baloney, and someone needs to do more research to see what happens with real, loud, CDs when attempting to reproduce them over an outboard DAC. Look at the spectrum of a square wave played by a good CD player and that same disc played with a great outboard DAC. You should only see odd harmonics; the even harmonics from the outboard DAC are from clipped transients. (PS: I point to the Benchmark DAC1 HDR simply because it's the world's best outboard DAC; you don't want to see what lesser DACs do to these signals.)
5.) Where did it come from?
With a CD, you've got a very serious product signed-off and approved by everyone on the project. It's molded-forever-in-polycarbonate masterpiece with every bit as intended. Compare this to getting a file from some music service (or worse), where no one really knows if the musicians and producer and mastering engineers ever signed off on it.
You'd think when you put a digital audio file in a computer for some processing that you'd get out exactly the bits you put in, but no. As Tom Holman once told us in Hollywood when he was talking about some research he had been doing, it took him, a real expert, about a half an hour to find all the settings in all the software he was using to get it all to leave his signal alone. It turns out that each piece of software thought it was trying to help in a zillion little ways like DC removal and etc., meaning that what you get is almost never what you put in.
When you buy an audio file from some online service, you never really know what you're getting.
Long Live Forever
I was totally excited when the CD came out and for the first time in my life I could get essentially direct copies of the master tapes just by buying a CD, and that those CDs still sound perfect 30 years later. In fact, my old CDs usually sound better than newer ones, which are all squashed to death by today's remasters.
You folks might also be tickled to know that most recordings are made with vacuum-tube powered microphones today plugged into vacuum-tube preamps before they're digitized and fed into ProTools software. Tubes rule in the world of pro sound.
If you don't like what's coming out of your CD player, try a better CD recording. The CD itself is incredible, but few recordings really show you what it can do. Blame the producers who think we're too stupid to turn up the volume on our iPods if they actually used some dynamic range.
Today's moral? Buy more CDs, put them on your iPod and computer if you like, and enjoy them. Get a great DAC if you've got computer stuff to enjoy, but don't waste your time futzing with computer equipment and music software when you can just buy CDs and enjoy the music itself instead of fiddling with stereo gear. God help us that some people waste time fiddling on their computers just to get music; half the reason the general public loves the CD over LP is simple convenience and never having to align a cartridge, flip an album or clean records or worry about wearing them out.
|
|
[複製鏈接]
發表於 2012-10-17 00:30
我知好長，但我好多point都勁同意..... 
樓主
