Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[bblackwood]

masterhse wrote on Wed, 22 December 2004 21:54

The thing that I'm having difficulty grasping is why we put all this effort and concern with dithering at 24 bits and beyond if it's all inaudible anyway.

Hi Tom. To be honest, I'm not sure anyone really does put that much effort or concern in dithering, etc. at the 24 bit level. I couldn't care less if something is truncated to 24 bits. I suppose if this happened, I dunno, hundreds of times it may become audible, but in real world use, any number above 20 bits (recording) is just unneeded space.

We've called the last 4 bits in 24 bit 'marketing bits' for years now...

[masterhse]

bblackwood wrote on Wed, 22 December 2004 23:11

masterhse wrote on Wed, 22 December 2004 21:54 The thing that I'm having difficulty grasping is why we put all this effort and concern with dithering at 24 bits and beyond if it's all inaudible anyway. Hi Tom. To be honest, I'm not sure anyone really does put that much effort or concern in dithering, etc. at the 24 bit level. I couldn't care less if something is truncated to 24 bits. I suppose if this happened, I dunno, hundreds of times it may become audible, but in real world use, any number above 20 bits (recording) is just unneeded space. We've called the last 4 bits in 24 bit 'marketing bits' for years now...

I begining to see the light as far as that's concerned on the conceptual level. You'll have to remind me why Digidesign bothered to create a dithered version of their mixing program (or are we just talking about inital recording and not processing?). Funny thing though, I can hear the difference between the two mixers with just a stereo pair going through 1 bus. Not an ear shattering difference mind you, but I'm a subtle-minded guy. On the recording side of the equation I see where dithering is just creating a better version of thermal noise for most A/D converters.

[bobkatz]

masterhse wrote on Wed, 22 December 2004 22:54

Thanks Bob, I appreciate your response. The thing that I'm having difficulty grasping is why we put all this effort and concern with dithering at 24 bits and beyond if it's all inaudible anyway.

I like (as usual) Bob Olhsson's response on the dithering thread: that the distortion accumulates a lot faster than the veil due to the dithering. In other words, lack of dithering to 24, one pass, "probably inaudible". Don't dither that a few times in a row and the grunge will start to exceed the ear's audibility threshold.

This is a separate issue I believe from the issue of "is there sufficient footroom if I reduce my peak level by 6 dB at the top"?

a) You're not getting any significantly closer to the noise floor when you record or mix with the lower peak level, as long as you don't lower it by an enormous amount. For me, that's peaking anywhere between -10 and -3 dBFS, very conservative.

b) But even though -144 dBFS is so inaudible not even a dog can hear it, if you don't dither at the 24th bit, your sound will get colder.

However, I thoroughly agree that with the severe compressor/limiter/clipping distortion in many of today's CD releases, it would take a helluva lot of passes of not dithering at 24 bits to make a difference.

I try not to make 'em that way...   God, I just listened to some of my rock stuff from 1990, it's 3 to 6 dB lower in loudness than what I'm cutting now and boy does it sound good! Tragic....

BK

[Lee Flier]

bobkatz wrote on Thu, 23 December 2004 23:26

However, I thoroughly agree that with the severe compressor/limiter/clipping distortion in many of today's CD releases, it would take a helluva lot of passes of not dithering at 24 bits to make a difference.

ON THE OTHER HAND... if you start with relatively dynamic source material on the individual tracks... including lotsa fades, reverb tails and other fairly quiet stuff... and then you process it with lotsa EQ, edit crossfades, etc... and you don't dither any of it... AND THEN you compress the living crap out of it thus raising the level of the cumulative distortion...

[Bob Olhsson]

I completely agree with Lee. At some point sound quality always breaks.

I see an integral part of our job as making a robust enough product that any downstream breaking point will be well past a radio station or a consumer's digitally controlled receiver. In the case of correlated distortion, my experience has been that cheap gear magnifies problems rather than obscuring them.

[Ronny]

bobkatz wrote on Wed, 22 December 2004 06:42

Level wrote on Tue, 21 December 2004 19:49 I have a recording that I did for the hell of it that peaks at minus 120dB (yep -120dB) and if you boost the gain in 20dB increments and render (until you get to a good level) it sounds and looks fine!! It came in as a fixed point 24 bit file peaking at -120 dBFS? Bill, you're pulling all of our legs. The noise to signal ratio after you gain boost and "render" (what the hell are you talking about, "render"?) will be so negative all you will hear is noise. Remember the laws of physics. They apply to digital audio as well as analog. I received a mixdown from a client that had tremendous dynamic range. Don't you think that I'm going to get a tremendous amount of hiss and console noise if I raise up the low level stuff? Was this a 32 bit float file that someone had accidentily attenuated? That's another story. BK

The file is 32 float. Likely from processing done after the initial recording, if it was initally at 24 bit. Not sure what he means by render either, but what I gather is that he recorded to his DAW a 24 bit recording with the gain down to -120dB. He than increased gain by +20dB 6 times, to bring it back up to the original peak. I dl'd the clip, ran an RMS and peak on it. RMS is 132.92dB L and 132.69dB R, peak is -120.26 L and exactly -120dB R, no visible waveform of course until about a 10x zoom. I than just put a +120dB amplify in 32 float, instead of +20dB x 6 (no difference either way) and lo and behold a nice clean D chord coming from a couple of horns or maybe synth horns and a synth type handclap in the middle. It sounds fine to me. The digital amplify only knows the audio as ones and zero's, IMHO. There is a difference between analog gain structuring and digital gain structuring and it's why I've been telling my recording newsgroup members to optimize the gain of the analog mic pre where it's most crucial and where the noise floor needs to be the lowest that it can be, before the signal is digitized and to not worry where the signal is hitting the ADC at, as long as it's not over of course. Altering the gain structure in the digital domain certainly has it's benefits. If Level did this same test in the analog domain, he would experience generational degradation from the multiple gain boosts. I'll bet you a Gatorade that the original song at this section has a crest factor of -12dB and not a lot of noise on the tracking and mix. I'd also venture to say that if Level, flew the original in again with recording input peaking at -1dB and leveled the -120dB peak clip to -1dB, that he wouldn't find anyone in an A/B test to tell him which clip was the original and which clip was recorded at -120dB.