R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [All]   Go Down

Author Topic: 0.1 dB @ 16 bits of resolution - a big deal?  (Read 6111 times)

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
0.1 dB @ 16 bits of resolution - a big deal?
« on: August 09, 2007, 01:17:13 PM »

What's YOUR opinion/experience/thought regarding this very matter when it comes to digital levels.

Fishin' some serious opinions about this.


Best Regards
Patrik
Logged

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #1 on: August 09, 2007, 01:54:07 PM »

0.1dB of what?  Output ceiling?

Best regards,
Steve Berson

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #2 on: August 10, 2007, 06:10:28 AM »

No, not ceilings and stuff around the magic barrier. That has been discussed to death.

Gain. Level. The digital mumbojumbo that increases level in any digital way. Fader. Input in a digital limiter. And so on...


Best Regards
Patrik
Logged

bblackwood

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 7036
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #3 on: August 10, 2007, 12:22:38 PM »

Nah, no worries at all. Any decent modern device will do the math at much larger word lengths than 16 bits, even if that is the source file's bit depth.
Logged
Brad Blackwood
euphonic masters

Viitalahde

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1069
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #4 on: August 10, 2007, 12:55:41 PM »

Yep, just today I had to insert an old single song to the album I was working for. Ripped it off the CD, 16 bits. There was no room to do any processing to sonically match the album, so I just levelled it to not stick out too much.

If it sounds good, it is good!
Logged
Jaakko Viitalähde
Virtalähde Mastering, Kuhmoinen/Finland
http://www.virtalahde.com
   http://www.facebook.com/pages/Helsinki-Finland/Virtalahde-Ma stering/278311633180

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #5 on: August 10, 2007, 02:39:14 PM »

bblackwood wrote on Fri, 10 August 2007 12:22

Nah, no worries at all. Any decent modern device will do the math at much larger word lengths than 16 bits, even if that is the source file's bit depth.


Yup, what Brad said.

Best regards,
Steve Berson

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #6 on: August 12, 2007, 12:05:39 AM »

Where I'm driving at is - if you settle a song on some kind of level, at a good balance, and "print" it...

...and THEN adjust it up or down by as little as 0.1 dB and print another copy...

...and THEN take these versions outside the room and into the outside world world...

...will you notice "differences" or not.

(Nevermind what happens around the ceiling - this is just from a balance point of view).

Best Regards
Patrik
Logged

$a1Ty

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 92
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #7 on: August 13, 2007, 06:35:40 AM »

if you can here a difference of 0.1db you must be some kind of god
Logged
Nathan Salt
Hence the nickname - it's an aussie thing

bblackwood

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 7036
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #8 on: August 13, 2007, 07:17:54 AM »

Patrik T wrote on Sat, 11 August 2007 23:05

Where I'm driving at is - if you settle a song on some kind of level, at a good balance, and "print" it...

...and THEN adjust it up or down by as little as 0.1 dB and print another copy...

...and THEN take these versions outside the room and into the outside world world...

...will you notice "differences" or not.

(Nevermind what happens around the ceiling - this is just from a balance point of view).

Well, that's right at the point where people can, in controlled situations, potentially hear differences. I doubt it would be audible to 99.9% of people in 99.9% of listening environments, but I'm not sure one can say that difference would be totally inaudible...
Logged
Brad Blackwood
euphonic masters

AndreasN

  • Full Member
  • ***
  • Offline Offline
  • Posts: 247
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #9 on: August 13, 2007, 10:12:43 AM »

Assuming the signal is level adjusted at high enough calculation bit depth, the change should be in the noise floor/dither.

Adding another TDPF dither to a previous TPDF dither shouldn't do much harm, I guess, but adding two very shaped dither's may give an unpleasant effect. Dunno. Have you tried? Smile
Logged

Jerry Tubb

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2761
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #10 on: August 13, 2007, 11:08:15 AM »

Kinda reminds me of this discussion.

JT
Logged
Terra Nova Mastering
Celebrating 20 years of Mastering!

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #11 on: August 14, 2007, 11:51:15 AM »

AndreasN wrote on Mon, 13 August 2007 16:12

Assuming the signal is level adjusted at high enough calculation bit depth, the change should be in the noise floor/dither.

Adding another TDPF dither to a previous TPDF dither shouldn't do much harm, I guess, but adding two very shaped dither's may give an unpleasant effect. Dunno. Have you tried? Smile


This ain't about how many thousand bits of resolution there might be before the final 16 bits. And in order to make things easy it ain't about any funky shapes of noise either. Let's just relate to TPDF, for the ease of it.

It is just about balance and level in the final 16 bits.


Best Regards
Patrik
Logged

AndreasN

  • Full Member
  • ***
  • Offline Offline
  • Posts: 247
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #12 on: August 14, 2007, 03:42:03 PM »

For clarity: Bit depth grows as the level change is calculated. This should be done at more than 16 bit word length. If not, it's truncated as it's calculated. Hence another round of dither prior to truncation.

Adding two uncorrelated sources at the same level gives a ~3dB rise. It's hard to believe this -90dBFS noise could affect balance and level. Smile
Logged

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #13 on: August 15, 2007, 07:49:46 AM »

AndreasN wrote on Tue, 14 August 2007 21:42

For clarity: Bit depth grows as the level change is calculated. This should be done at more than 16 bit word length. If not, it's truncated as it's calculated. Hence another round of dither prior to truncation.

Adding two uncorrelated sources at the same level gives a ~3dB rise. It's hard to believe this -90dBFS noise could affect balance and level. Smile


I must be extremely unclear...


Logged

escape

  • Full Member
  • ***
  • Offline Offline
  • Posts: 125
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #14 on: August 15, 2007, 12:45:53 PM »

my opinion is I can't hear it, as long as the gain is done in 16 bit res. so it's not a big deal to me. the higher resolutions that require dither for coming back to 16 bits are for compression and EQ plugs, but just for gain I believe 16 bits is fine.

my experience agrees with what I've stated above.

my thoughts are that, mathematically even, this makes sense and it looks good on paper. ultimately I still have to come back to the sound quality.

therefore, I have no problem doing gain adjustments at 16 bit res and often do, post dither, if that is all that is needed.
it's hard for me to justify going to 20 bits or 24 just to change gain, especially when it's as high as 0.1 dB, and then having to add another layer of dither. your gain adjustment may be more accurate to the 0.1 dB in the higher bit depths, but I seriously doubt you are going to hear it. I never have. regardless, when using higher bit res for changing gain you have to dither again when coming back to the final resolution and that adds some chaos to the mix so it might not be as much more accurate to the 0.1 dB as you would instinctively suspect.

96 dB divided by 32768 is 0.0029296875, so unless you're making a gain change of 0.0029 or less I wouldn't worry about it.

Logged
ERIC JENSON
MASTERING ENGINEER
ACOUSTICS ENGINEER(ALMOST)

E-Scapes Mastering Services
Miami, FL

AndreasN

  • Full Member
  • ***
  • Offline Offline
  • Posts: 247
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #15 on: August 15, 2007, 02:01:27 PM »

Patrick; My apologies if misunderstanding. Your question is what happens when changing a signal by 0.1dB at 16 bit. This can be broken down to two separate effects.

First, the practical implementation. In this case, digital. It's not mumbojumbo. :) It's a calculation with associated word length expansion, dither and truncation back to 16 bit. There's no distortion or altered frequency balance - except for the added dither noise. Unless something is broken. This is the issue I feel the original post was about..?

Secondly, there's the intended 0.1dB change. This is a separate issue from the actual means of accomplishing the task. Even if it may somehow be detectable in clinical conditions, it can't have any practical value. That's the least of worries in audio, especially compared to the side effects of doing the gain change (truncation distortion, dither noise or analogue noise).


Eric; all DSP's are created equal and have the unalienable right to avoid truncation distortion. Even gain! Hehe.. An example: Multiplying, say, 300 by 1,023293 is 306.9879. Multiplication have a habit of creating more decimal numbers. If the operation does not have enough footroom, the sum will truncate as the result is calculated. Therefore longer word length in the calculation and another round of dither. It's the only option to truncation distortion! Dither does not add chaos to the mix, it's actually the straight opposite. Truncation gives (non musical) spurious frequencies, some relatively high in level, while dither spreads this energy across the spectra as a noise floor. Word length expansion is not only for comp and EQ. :)

The sound forge program springs to mind. The version I tried had the nasty habit of opening and editing all files in their original bit depth. To avoid truncating these results, the user had to manually select a higher bit depth, do the calculations and then select the proper way to reduce bit depth with dither prior to saving. I don't know how many accumulative truncations I did before discovering that little fact. It may not bother much in a single truncation, but hey, this is mastering! Details count. :)


Cheers,

Andreas Nordenstam
Logged

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
Re: 0.1 dB @ 16 bits of resolution - a big deal?
« Reply #16 on: August 15, 2007, 02:23:02 PM »

AndreasN wrote on Wed, 15 August 2007 20:01

Patrick; My apologies if misunderstanding. Your question is what happens when changing a signal by 0.1dB at 16 bit.


Nope, still wrong.

It's all about two different versions of a FINAL 16 bit file. One at one level and the other where some digital gainstage has increased or lowered this very level.

Hopefully at a resolution higher than 16 bits.

It's all about how the end resolution appears to sound. To the ear. If one version is the "perfect" one and the other is just a tiny little bit higher or lower, do you fancy that one as well? Or not?

No science. Just two versions. Simple as that.


Best Regards
Patrik

Logged
Pages: 1 2 [All]   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.1 seconds with 19 queries.