I've studied the Meridian papers and the 518 manual. In 1995, the 518 was addressing certain very real problems that to some extent have now been addressed in other ways. Specifically, their major concern is optimising the transfer function across boundaries where bit-depth changes and preserving/improving the "subjective dynamic range" (their term) across boundaries of equal bit-depth.
Where DSP functions such as gain control act on digital signals they introduce errors (noise) in the LSB's. For a signal well below full-scale (say 30dB - a very quiet section in a piece of classical music) the MSB carrying musical signal is only a few bits away from the noise floor. Errors introduced in the LSB will be more audible here. So they advocate, amongst other things, raising the gain of a track so that it peaks at full scale (or even occasionally clips) in order to maximise the gap between the LSB errors and the music. This is perfectly logical, not voodoo and indeed what I and 99% of people who have ever mixed/mastered anything would do - although there are many different ways of doing it. For dealing with very brief clippping on individual tracks for instance, I would manually edit the handful of samples involved to reduce their level to full scale - effectively a manual compression operation, albeit rather time consuming and fiddly. In practice this would enable a master mix at near full-scale peaking without having to leave 3-6dB headroom across the track as a whole. One could also use soft-clip digital (or analogue) limiters to achieve the same thing - this can be abused (loudness wars). Anyway, all of the above is very pertinent to what happens prior to dithering down to 16-bit for CD manufacture. Bear in mind that in 1995 full-chain 24/96 recording was not available... and that 20-bit DACs were not common in domestic settings. The term "subjective dynamic range" is important... they are talking about keeping the dsp-generated noise in the LSB as far away as possible away from the signal. So, it's really a discussion about SNR. I think this is where we are talking at cross-purposes. AUDIBLE dynamic range is the difference between the max/min SPL actually produced by your speakers (excluding the noise floor post-DAC) - i.e. what you hear! A DAC signal has a fixed, finite theoretical "dynamic range" determined by the number of bits (96/144). In practice, this is limited by noise in the LSB's (from a variety of causes) and the real-world 20-bit limitation of audio ADC's. Using Meridians own graph, the scope of human hearing is 19-bits. This is why a 20-bit replay chain can outperform a 16-bit one with 20-bit source material. However because of auditory masking, the EFFECTIVE (ie the difference between the loudest and softest sounds you can actually discern at the same time) "subjective dynamic range" is more like 60-75dB (10-13 bits). This is why good FM radio, analogue master tape and even vinyl can sound great! The LSB errors that Meridian are (rightly) concerned with are MUCH less important with a 24/20 bit chain - as they are at/below the scope of human hearing. Also, the MSB-LSB gap is greater at all times, even on quieter passages so the LSB noise remains further from the music. This is why many people love tracking at 24-bit - you can run with (say) 18dB of headroom for the occasional peaks without needing compression and not worry too much about the noise floor - you simply wouldn't do that with 16-bit - you'd run as near full scale as you dare and use a hard-knee compressor for safety - but then the compression is irreversible and the original dynamics have been permanently altered... So returning to the original topic of this thread, what level of digital attenuation in an SD device is deleterious to the sound? The OP's opinion is "any digital attenuation is bad" (but presumably analogue attenuation POST-DAC is fine?) My position is that both analogue and (properly designed 24-bit or greater) digital attenuation are audibly indistinguishable on decent downstream equipment... until you reach the point where the EFFECTIVE (ie resolvable) subjective dynamic range is compromised. IME this doesn't begin to happen until you get down to below 13 "live bits" on 16-bit source material (19 on 24-bit material) So, incorrect gain staging can induce this. Stay below 18dB of digital 24-bit attenuation on 16-bit material, use a 24/20 bit DAC and all should be fine... On 24-bit material, up to 30dB of 24-bit digital attenutaion should be fine. And remember, once something is so quiet you can't hear it, it doesn't matter HOW it go to be that quiet... For the last time... if you can't hear a marching band mixed into Brahms Lullaby in the bottom 3-bits of a 16-bit full-scale peak recording, you can't hear the bottom 3-bits of the Lullaby either!!!! If ANYONE can pass this test (or indeed if anyone can tell me how to cheat the test - and there are enough clues in this post to answer that conundrum) we can talk further.. By the way, I was looking at some very quiet passages on 16-bit classical stuff and the lowest level I could find was 40dB (7 bits)... and that was right at the very beginning of a quiet track. So even losing 3 bits is not fatal to the music... although for that brief period the SNR will only be 24dB or so... Does it sound bad - no, just quieter. Can you hear the worsened SNR? - not at normal listening volumes. If you boost the post-DAC gain to speaker-threatening levels and press your ear to the tweeter... yes. I wouldn't advise this. Can any of this be measured? - yes. A final point to all of those worried about "loss of resolution", which I take to mean they think that bits of music signal are being damaged... use ADM and prove it to yourself. How far do you have to attenuate before the null difference is not just white noise? That's your homework... now play nicely. -- Phil Leigh You want to see the signal path BEFORE it gets onto a CD/vinyl...it ain't what you'd call minimal... Touch(wired/XP) - TACT 2.2X (Linear PSU) + Good Vibrations S/W - MF Triplethreat(Audiocom full mods) - Linn 5103 - Aktiv 5.1 system (6x LK140's, ESPEK/TRIKAN/KATAN/SEIZMIK 10.5), Townsend Supertweeters, Blue Jeans Digital,Kimber Speaker & Chord Interconnect cables Kitchen Boom, Outdoors: SB Radio ------------------------------------------------------------------------ Phil Leigh's Profile: http://forums.slimdevices.com/member.php?userid=85 View this thread: http://forums.slimdevices.com/showthread.php?t=77725 _______________________________________________ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/audiophiles
