Quote:
Scnr, but instead of going for greatest bandwidth you should concentrate on recording stuff so that it doesn't clip.
There's nothing wrong with recording at higher sample rates and it is de facto standard to use higher bit depths, but such formats are not needed for playback. As I've said
before, the stuff up there is extremely low in level, easily masked, mostly noise and most importantly: inaudible.
How did you determine that these harmonics were not bursts of noise and there was no constant HF noise? Because I just checked a 24/96 recording of a harpsichord and the HF rms amplitude is -80 dB on average to well below -120 dB minimum. It IS constant HF noise plus bursts of noise.
You can repeat your claims of the differences you hear between DSD and CD all you want, it's not going to change anything. Again, do you have anything to back up your claims other than anecdotes?
Seriously, where are you pulling those claims out of? The dynamic range of concert halls doesn't exceed 80 dB (Eargle, John; 2005; Handbook of Recording Engineering) and all the recordings I checked of Mahler's 2nd do rarely exceed 70 dB.
16 bit has clearly enough dynamic range. Again a comparison to vinyl: "Under no legitimate circumstances will the dynamic range of vinyl ever exceed the dynamic range of CD, under any frequency, given the wide performance gap and the physical limitations of vinyl playback."
PCM recording is usually done with 24 bits and possibly higher sample rates (88.2, 96 or even 192 kHz). You do not have to record directly to CD-R. All you need is a (quality) resampler and dither for the conversion to the CD format.
Yes, editing and even recording is done at higher bit depths than 16 bit (DSP uses as high as 80 bit floating point). Yes, you can hear stuff below the dither noise floor, you can for example hear a 1 kHz sine wave at -100 dBFS quite clearly on a properly dithered 16-bit file.
First things first. Of course I ALWAYS try to record without clipping. That sole extraordinary incident would happen under given circumstances to anybody not familiar with that extreme composition. Some 10 or so dB above anything Beethoven wrote for piano is REALLY loud - much more so than any other piano music I ever heard. I simply did not know it is possible to play piano so loud.
The dynamic range of concert halls is - rather vague. What do I mean by that? It is no problem at the loud extreme, it is loud as it is loud. Today, it is very problematic to find QUIET place - lighting system(s) hum ( 50 Hz in Europe, 60 in USA ), air conditioning is not totally quiet, etc, etc. I have constant trouble with musicians who for some reason always turn on every electric light available - and a single humming/buzzing light can drop your dynamic range by 20 dB ! During recording, I try to put out each and every light that is not absolutely necessary for the pereformers - yet, there are so called "fire lights" that have to be on because of safety all the time - and they do buzz. Since I record mainly from distance, this humming/buzzing unfortunately is a factor; close miking does not have this problem, but simply does not sound natural to me.
Mahler's 2nd: you said you were comparing/measuring RECORDINGS, not the real thing. Recordings of 2nd to date have almost all been subjected to
some kind of compression - either with mike positioning not to capture that much of tympani in the finale ( it IS louder than everybody else combined - live, not on compressed recordings ) or by compressors. Depending on hall, it can approach ear pain level - that is about 120 dB - and there are moments of total silence, which should be equal to the noise floor of the hall. Even if that is 40 dB (really loud/bad hall), that still makes dynamic range of 80 dB. One recording I know that does stand out as far as dynamic range is concerned is Telarc's
http://www.amazon.com/Mahler-Symphony-Minor-Resurrection-incomplete/dp/B00006879J
but even this has troubles with resolution of 44,1/16 and intelegibility of softly sung chorus as a result. I am sad Benjamin Zander's version with Philharmonia Orchestra is not available - most others are
http://benjaminzander.com/recordings/philharmonia,
DSD recordings available as SACDs.
It is true that LP, as GENERALLY known, does not meet or exceed dynamic range of CD. The maximum dynamic range of conventional LP is 78 dB, under best possible conditions - that is to say having master lacquer with - 60 dB ref 0 dB noise ( unfortunately VERY rare ) and cutting at the absolute upper limit of + 18 dB. Very few, if any, LPs have achieved this - or even approached these figures. There is one aspect of performance in which LP can exceed the capabilities of CD - rise/fall time, or in other words, frequency response. Half speed mastering, developed initially for quadrophonic systems requiring HF carriers up to 45 kHz, can put flat response up to at least 50 kHz on disc - making change of loudness per specified amount of time twice faster than possible with CD. Recent developments have reached even better figures - and top styli and cartridges have no serious problems with that frequency range. Which means LP can reproduce sudden bursts much better - CD can have greater dynamic range on paper, but can realize that only with slower sound sources - on percussion, properly done LP will always win.
While working briefly at Benz Micro Switzerland back in the early 90s, I was able to measure the probably fastest phono cartridge ever made - prototype for what later became known as Benz Ruby. Its rise/fall time, using CBS STR 112 test record, was 3 usec ( in words - three microseconds ). That is about three times faster than most other, regularly available Benz included, phono cartridges. The cartridge had far too low output to be deemed useable for general public - 0,03 mV/5cm/sec. One listen to this cart in the reference system used ( Maggies with ribbon tweeters good to 40 kHz and Class A ultra low noise amplification good to ??? KHz ) would change your opinion regarding LPs for good.
Reasonably achievable dynamic range on conventional LP is 70 dB - which are available at twice the loudness change per amount of time compared to that of the CD. Just compare once a good CD and good LP version of the same
analog recording on the system that overall supports at least 40 kHz bandwidth.
There was a very short lived compressor/expander system called CX for LPs developed by CBS Columbia - it allowed for 20 dB reduction of noise and rumble compared to conventional LP record. Ultimately, 90 dB dynamic range is achievable from LP using this system. For all practical purposes, this equals the dynamic range of CD on loudspeakers, slight vestiges of noise/rumble compared to CD are audible with headphones. Unlimited analog resolution, of course. CBS shot itself in the foot by insisting to claim CX encoded LPs can be played without appreciable sound change without cx decoder - there are even reports artists demanded and achieved withdrawal of CX encoded discs from the market due to this incompatibility. It would be fair to say this compatibility does not exist - and system might have had a chance. Needless to say, immediately after Sony acquired CBS, they pulled back from the shops each and every record bearing CX designation they possibly could - for the fear of system catching on and delaying or even preventing adoption of CD. Getting any CX encoded LP record today borders on mission impossible as a consequence - it is both sad and ironic that massed background violins accompaniying Julio Iglesias on
http://www.discogs.com/Julio-Iglesias-Momentos/release/928092
when played with appropriate cx decoder are among the best sounding violins on any recording yet made - whereas serious classical music playing is marred by vynil noise and rumble in conventional LPs. Those few releases of serious stuff really border on mission impossible these days as far as availability is concerned.
It should be no problem for you to calculate or measure the rise time of a CD. It is what it is - on 20 kHz bandwidth with brick wall filtering required by RBCD that is all that will ever be possible with CD. And that is (too) slow.
Of course, one can record digital PCM in greater depths and higher sampling rates than CD's 44,1 kHz/16bit. The higher the better - but even 192/24
exibits for PCM typical ringing on square wave - DSD, particularly DSD at 5,6 MHz or DSD128, is far better here, presenting square wave response much similar to analog, with uncomparably less ringing. That gives it tremendously precise pulse response.
Even DXD, at sampling rate of 384 KHz / 24 (or is it 32 ? ) bit, by the admission of its most outspoken proponent, can not reach full value of pulse - IIRC, figure is 84 %. CD is here far, far worse still - it is simply too slow.. DSD can reach full amplitude of pulse.
It is true that DSD recordings are most usually mastered in PCM (at highest sampling frequencies and greatest bit depths with floating decimal possible ) - but that is not inherent incapability of 1 bit DSD system, just hardware and software
to perform mastering in native 1 bit enviroment is still very expensive and all but commonly available. Vast majority of SACDs available have been mastered in PCM enviroment - but this is now changing relatively fast.
What are the prime audible benefits of extended bandwidth ? Recreation of acoustics of the space music was recorded in - those few seconds before actual music starts is all that it takes to convience one that such recordings are simply better, much closer to the real thing. To the listener accustomed to CD these high resolution recordings may well appear as noisy - that is because CD does not have good resolution at low levels and higher frequencies and does not support anything above 20 kHz at all; feed from microphone, analog recording and higher resolutions of digital will always sound more "noisy" than CD version - and that is not CD's quality but actually drawback.
Regarding 96/24 recording of harpsichord - yes, noise in PCM is constant, about the amount you stated. 96/24 recording should be pretty linear to approx 40 KHz, depending on the recorder - but first requirement are microphones and mic preamps to be able to capture above 20 kHz sounds.
The lack of harmonics above 20 kHz observed is most probably due to mic set up - should be there.
Aichille's heel of DSD is noise above 20 kHz - it starts getting pretty loud just above 20 kHz and increases in amplitude with rising frequency.. It is here that double DSD at 5,6 MHz or DSD 128 shows main advantage over 2,8 MHz or DSD64 - the spectrum of noise is about the same, difference being 20 dB lower noise and therefore by the same amount better S/N above 20 kHz for double DSD.