My background is as a live sound engineer. I first came to America in the early 70's with a band called Gentle Giant (Yeah, I really am that old!) and returned many times with bands you may know a bit better. The Kinks?
Oh hell yes The Kinks. I'm in 30's but my older cousin used to play records to get me to sleep while babysitting when I was 2-3 years old. Can I just say that sound engineers today do not possess half the talent of the old-timers. There's no sweetening... it's all amplitude pumping to constant clipping, distortion, garbage. Very few albums these days take advantage of the dynamic range of the digital medium.
I've done both live engineering and studio... studio is a whole different deal. Both are crafts that take skill, but there are different objectives and different considerations. e.g. With PA you have one chance to get it right. With studio recordings, problems with fidelity are going to be much more noticeable and permanent once the master recording is final.
I certainly, and greatly, respect your work as a live sound engineer... particularly one in an era with a lot more attention to craftsmanship than I see these days, but I'll tell you that digital formats in the mastering world are a very different ball of wax. That being said, let me try to address some of your questions...
When I decided to put all my vinyl in to the loft (or attic ) and re buy my music collection on CD, it was purely for convenience. I still prefer the sound of vinyl ( I think the best explanation was by a Japanese guy I know who uses old valve amps to power his speakers: "It makes me feel warmer") illogical, but I kind of get it.
Again, this goes back to a few tradeoffs. Vinyl doesn't possess the dynamic range of CD audio. It's maybe about 16 decibels short. Ironically, that's bad news for jazz and classical music... the two genres that audiophiles LOVE to listen to on vinyl.
In the early days of digital mastering, more than 20 years ago, many recordings were lacking on the low end of the frequency spectrum. There are a number of reasons for this, but since then both the hardware, software, and engineering techniques have adjusted to compensate for the fact the noise floor itself is virtually absent. The noise floor of the analogue medium does contribute to the warm, fat sound... consider that you've got a constant underbelly of low frequency "fuzz" plaguing the entire recording.
Therefore what you're actually hearing an analogue recording is false... even though it may "sound better". But I'll tell you something, the first time I fired up my optically-linked surround system and played "The Thomas Crown Affair" and heard in the opening dialogue an absolute dead silence between lines, I was elated. The noise floor is so imperceptible that it adds to the ability of any sound system to make a more faithful reproduction that sounds, if properly engineered, like it's right there in the room.
My Hi Fi has a very "Live" sound and consists of large floor standing Bi-Amped speakers powered by professional amps. It runs at around 750 Watts Rms a side. It's LOUD.
I'm a big fan of LOUD for PA... I used a tri-amped (1600 watts total) Peavey Project II system to disc jockey a school dance once. It was the first time I'd ever heard middle school kids complain about music being too loud.
But where sound reproduction from a recorded medium is concerned, I'm a bigger stickler for accuracy of reproduction. I'm glad you mentioned Tannoy. They make GREAT reference monitors. I'm assuming your British since that's where Gentle Giant came from, and the Kinks... Well, the Brits know how to make great speakers. I have some KEF speakers and love the hell out of them.
None of us could hear any difference between AIFF via the Squeezebox and the direct source from the CD player. However, we could all hear a difference with the MP3 files (much to my sons annoyance, who did the double blind tests!) we didn't test the Apple lossless as it wasn't out then.
Well, there's the rub... MP3 is a VERY different codec from AAC (MPEG-4, Part 10 spec). MP3 performs very poorly at low bitrates due to its dependency on older compression techniques and lack of more sophisticated perceptual encoding. AAC, on the other hand, is indistinguishable from CD audio at 128 Kbps. You are absolutely right you can hear differences between MP3 and AIFF... but even without a few beers in your system you won't tell apart 128 Kbps AAC and 16-bit LPCM even with the goldest of golden ears.
Any kind of signal processing has to do something to the sound by definition, surely? So while you are probably correct in saying it is not lossless (technically speaking) why bother processing the signal ie compressing it and then uncompressing it, when there is no need?
Welllll... Let me straighten a few things out. Signal processing is one thing. Digital compression is another. And perceptual encoding is a little bit of both and also neither. As another user put it, compression in the analogue realm means something considerably different from data compression in the digital domain. We're not talking about compressing the amplitude modulation of an analogue waveform. Instead, complex algorithms are used in the decoding hardware and/or software to recognize clusters of data such that the data could be truncated with a shorter data. Information can also be stored in a nonlinear fashion... entirely unlike the analogue realm.
Imagine that you have a sequence of 12301234. Now imagine that you can use a value let's call x to represent the cluster "123". What a codec can do is reduce that string from 12301234 to x0x4... where x is understood by the codec to mean "123". It could go even further to say, in some manner, to truncate x0x4 to essentially say in an even smaller string that: x appears twice, and is interrupted by a zero and a 4.
Another technique, linear prediction, can reduce data in a manner similar to this...
Imgn w rmvd ll th vwls frm th sntnc.
The sentence can still be reconstructed if the algorithm of the decoder knows everything it needs to know about speech synthesis to fill in the blanks. Consider for a moment that your brain already did it... I doubt it was very difficult at all for you to decipher what the above sentence says.
I know it sounds counterintuitive because this is a lot for a system to have to know... but we're talking about reducing data requirements of the medium, not the playback system. Notice the difference here... You're packaging more knowledge into the hardware so that the format can be made more compact.
Then there's the aforementioned ADPCM (from my previous post) which uses techniques like quantization throttling and relative amplitude values to store only the change between amplitude values from one quantization sample to another, and limits the size of each sample to only the number of bits required to store the value given at that sample.
All of these methods, in principle, can be used to reconstruct an analogue waveform with enough resolution that your ears cannot tell the difference between the reconstructed wave and the original. Now keep this in mind:
in the 12301234 example, x0x4 is not what is played back. x0x4 is decoded, and, here's the important part,
reconstructed into 12301234.
Same with the sentence... the sentence is not played back with vowels absent, the algorithm is applied, the string decoded, and the original analogue sentence reconstructed.
It is upon reconstruction where the potential for error can occur... but this potential is mitigated heavily by today's methods of digital sampling and reconstruction, encoding and decoding.
One example is frequency aliasing. If you use 44,100Hz as the digital sampling frequency (again this is not an analogue wave but a representation of the number of times, frequency, a discrete time sample is recorded as digital data... a sample/quantization interval)... then according to Harry Nyquist's ever reliable equation, your frequency response extends to 22,050Hz. Why? Because the bare minimum data needed to reconstruct a sinewave is the peak and the trough of one wavelength, i.e. two digital samples per cycle (1 Hz). 44100Hz was picked because it gives enough overhead past the A-weighted spectrum... the range of human hearing.
But what happens if you sample a frequency HIGHER than 22050Hz... you will get aliasing. Now, some people hear the term "aliasing" and think this means a jagged, staircase like signal... and audiophiles think that this is actually perceptible. Well, yes and no.
First, aliasing is better understood by its dictionary definition which is closer to the truth of what it means in the digital realm... Take a look at
this demonstration to see how sampling a 7000 Hz frequency 8000 times a second will actually create an aliased frequency of 1000 Hz in the recording.
So what happens if we sample a 33kHz signal at 44.1kHz, we'll get an alias frequency because we aren't sampling the amplitude values of a 33kHz signal at the peak and trough... but somewhere else. The result is a frequency reproduced closer to 11kHz.
What's the answer? An antialias filter. In the mastering stage, a 20kHz lowpass filter will knock out any frequencies that would not be sampled properly at the sampling rate of 44.1kHz. Problem solved.
Just as there are scores of techniques used in analogue sampling, recording, mixing and mastering to maximize the potential accuracy of that reproduction... same goes with digital.
The one thing I learned when doing live sound is that you strive to do as little as possible to the original source. It pains me to see these young guys who come out of college courses, piling on EQ to a snare drum, instead of going to the stage and trying to re tune the drum itself.
One of the beauties of digital nonlinear editing is that you can do all kinds of nondestructive filtering, eq-ing, leveling, effects, etc. that do absolutely nothing to the source data... you can always go back to the pristine, unaltered source. But I know what you mean. At the same time, where do we draw the line between one kind of art and another? When it comes right down to it, the acoustic drum and trumpet are still artificially constructed instruments designed to reproduce sounds we first heard in nature. Even though the designs have changed, the instruments are still man-made as they always were.
Andy Johns and Peter Collins have both used miking techniques to enhance the liveness of the drums on Led Zeppelin, Van Halen and Rush tracks. Kashmir by Led Zeppelin would not be the song that it is without the stereo phaser effect that Bonham threw on the drums... that one effect changed the vibe of the whole mix. When do we realize that a drum machine is simply a different kind of instrument than a live drum kit (I'm partial to real percussion but only because I practiced more on a real kit and that's what my mind is partial to).
The problem isn't that the instruments are changing. The problem is that some people are just lazy. What you have to do is get out ahead of the technology and decide that you're going to master the new instruments in a way that no one else has... Talvin Singh is a hell of a drum programmer but he is also a phenomenally accomplished tabla player.
The same goes for the recording medium. Digital has changed the game and we can sit and whine about it or we can figure out how to get out in front of it and take advantage of everything it has to offer in the way of reproductive accuracy.
I've since set up the Apple TV and it works just as well as the Squeezebox sound wise, the big difference though is that I can see what I'm doing. The other gentleman who said the Squeezebox screen is large, must be younger than me and with better eyes! I can see all the artwork as well.
Agreed. This is my gripe about the Squeezebox... It's designed well for someone who just wants a listening room with no television but most people these days have an integrated home theater of some sort. It's a lot easier to navigate from 10 feet away with the AppleTV than even looking at a tiny screen in my hand.
In the end I've never met anyone who can explain technically why JBL speakers sound better on Rock music and Tannoy sound better on classical. They do though!
Amar Bose probably could. He contributed greatly, despite what the snobs want to say, to the advancement of sound reproduction. Though his goals are more oriented toward efficient output rather than accurate reproduction, this psychoacoustics professor at MIT from my native India knows a tremendous amount about how the brain perceives sound and could surely identify why JBL is better oriented to rock and Tannoy to classical. Although my immediate reaction is that rock tends to be heavy in bass and treble, and JBL makes simple speakers that produce tremendous SPL output in two ranges of the spectrum, whereas Tannoy follows my philosophy of smaller and more drivers to tend to the full range.
If you've ever sat in a 2001-2005 Mercedes S-class and listened to the 13-speaker Bose Beta 2 system you'll know what I mean. The system uses smaller drivers, but a large number of them, along with custom developed acoustic waveguides modeled specifically to the Benz cabin's acoustics, to saturate the cabin with sound without exploding your eardrums.
One thing JBL doesn't do very well is produce full-range response at very low sound pressure levels... They're mostly a PA speaker. Tannoy, KEF, Paradigm, B&W, etc. do a phenomenal job of maintaining frequency response at low sound pressure levels. This is critical in both instrumental recordings and feature film soundtracks, which make fuller use of dynamic range than the typically amplitude-pumped, distorted to hell rock recordings that tend to have two volumes... off and eleven.
This is why I'm a big advocate of 24-bit Linear PCM audio for critical listening... the dynamic range of 16-bit CD's is only about 96.7dB whereas 24-bit LPCM extends up to around 140dB dynamic range with an amplitude resolution of (sorry I understated this in my previous post) 16.7 MILLION possible amplitude values per sample interval. I've done some instrumental recordings where CD audio just KILLS your ability to hear simultaneously quiet and loud sounds or abrupt shifts. Note that amplitude resolution, and not frequency response, is the real reason behind the inability to record cymbals very effectively. Audiophiles love to use cymbal heavy recordings as an example of the inferiority of certain digital formats. But they fail to understand that sampling frequency is really not the issue... Cymbals hover somewhere between 7 and 10kHz, both well within the limit of a 44kHz format such as AAC or 16-bit CD audio. But look at the analogue wave of a cymbal... the frequency is pretty constant, what's erratic is the amplitude! So erratic that 16-bit resolution just doesn't cut it.
For critical listening I would like to see formats like DVD-Audio gain popularity... I know they'll always reside in the minority, but bandwidth and storage in nonlinear computer-based systems is getting cheaper and cheaper, and 24-bit LPCM shouldn't be hard to support going forward.
That being said... For most listening, AAC does just fine... because the source recordings were all mastered to 16-bit CD audio anyway. Nothing more than 128 Kbps AAC, maybe 192 at best... is really needed for the vast majority of sound recordings out there.
