R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: [1]   Go Down

Author Topic: 32bit Float to 24bit fixed?  (Read 2837 times)

KHStudio

  • Newbie
  • *
  • Offline Offline
  • Posts: 9
32bit Float to 24bit fixed?
« on: May 30, 2005, 10:17:36 PM »

 I’ve been using Nuendo for over 5 years.  I just realized that I’ve been truncating from 32-bit within the program to my 24-bit converters (2x) RME 8 i/o and Tango 24).  Nowhere in the manual or website did anyone warn that the program doesn’t handle this correctly.  I hardly ever mix in the box.  I send 24 channels out to an analog mixing console almost exclusively.  To me it just sounds better.  I am aware from reading Nika’s paper that there are two separate issues concerning float-point dithering.  With that said, the only one that concerns me is how Nuendo doesn’t give me the right tools to dither the outputs prior to DA conversion. They really need to add more dithering options to address this issue.  

I recently applied dither plug-ins to my final output busses in Nuendo (post fader) following information I read from Bob Katz and Nika (24-bits of triangular dither, no shaping, and 2-bits of dither).  Myself and many others have heard huge differences in the way the audio sounds coming through the mixing board:  more dynamic, better separation, more alive and less harsh, etc...  The best plug-in out of the bunch was the Izotope Ozone, all other functions were turned off except the DC offset was turned on prior to dithering (it seemed to make it even better).  The only problem is using 12 of the Ozone plugs in Nuendo eats up a lot of processing.  I’m using a 3 gig P4 and a gig of RAM.

My questions are:

Could you guys recommend any dithering plug-ins that you’ve found to be efficient and also sound good for this purpose?

What are your opinions on the sound quality of Nuendo as opposed to Samplitude for my way of working?  Personally, I found Samplitude to sound very, very good and also gave me excellent tools for handling the output dither to my 24-bit converters. Nuendo also sounded nice with the dither on the outputs & very flat & annoying with it off, the sound that I’ve been fighting with for years and didn’t know it.

I would especially like to thank Nika for sharing his time & knowledge with the rest of us along with Bob Katz and others who have helped me over the years.

Thank You,
Kevin Hogan
www.KHStudio.net

PS. THIS THREAD IS NOT ABOUT FLOAT BEING BETTER THAN FIXED POINT!
A lot of us use float-point and would just like to get the best out of our system as possible. Any advise on how to do so better is very much appreciated.
Logged

KHStudio

  • Newbie
  • *
  • Offline Offline
  • Posts: 9
Re: 32bit Float to 24bit fixed?
« Reply #1 on: June 01, 2005, 04:34:51 AM »

I’m sorry I posted this topic on a few other forums & was notified that it was bad etiquette. I was just trying to get some diverse help & opinions from several sources.

If you would like to follow an ongoing discussion on this topic go here:
http://forum.nuendo.com/phpbb2/viewtopic.php?p=36140&sid =db664eca2cb0feae66c9b29a250c771f#36140
Logged

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728
Re: 32bit Float to 24bit fixed?
« Reply #2 on: June 01, 2005, 10:15:19 AM »

Speaking frankly -
I'm not sure I've ever truly heard the difference between 32bit (either fixed or floating) truncated just only one single time to 24bit vs. 32bit dithered to 24bit.  24to 16bit is really obvious though - but the LSB at 24bit is a heckuva quiet point.  

Anyone ever run blind tests on this??

Best regards,
Steve Berson

jazzius

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 458
Re: 32bit Float to 24bit fixed?
« Reply #3 on: June 01, 2005, 11:02:06 AM »

KHStudio wrote on Tue, 31 May 2005 03:17

 
Myself and many others have heard huge differences in the way the audio sounds coming through the mixing board:  more dynamic, better separation, more alive and less harsh, etc...  


Perhaps you imagined a huge difference....wouldn't be the first time, won't be the last.

Bob Olhsson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3968
Re: 32bit Float to 24bit fixed?
« Reply #4 on: June 01, 2005, 12:53:00 PM »

TotalSonic wrote on Wed, 01 June 2005 09:15

Speaking frankly -
I'm not sure I've ever truly heard the difference between 32bit (either fixed or floating) truncated just only one single time to 24bit vs. 32bit dithered to 24bit.  
If that were all that was ever going to happen to the audio I doubt you'd hear much difference. In the real world, a LOT is going to happen to that signal before it reaches the listener's ear so it makes no sense to allow truncation distortion because it builds up with subsequent processing.

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728
Re: 32bit Float to 24bit fixed?
« Reply #5 on: June 01, 2005, 02:23:07 PM »

Bob Olhsson wrote on Wed, 01 June 2005 17:53

TotalSonic wrote on Wed, 01 June 2005 09:15

Speaking frankly -
I'm not sure I've ever truly heard the difference between 32bit (either fixed or floating) truncated just only one single time to 24bit vs. 32bit dithered to 24bit.  
If that were all that was ever going to happen to the audio I doubt you'd hear much difference. In the real world, a LOT is going to happen to that signal before it reaches the listener's ear so it makes no sense to allow truncation distortion because it builds up with subsequent processing.


But that's my exact point - is this one time truncation of the returned summ figure before final output  the real reason to be blamed for degradations or is it a result of the way math is handled throughout different places in the entire internal signal processing path?  The thing to me is not to bark up the wrong tree.  There's a definite reason why I've chosen the DAW app I use (SAWStudio) over others because of this exact reason.

Best regards,
Steve Berson

Bob Olhsson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3968
Re: 32bit Float to 24bit fixed?
« Reply #6 on: June 01, 2005, 04:38:01 PM »

I don't like to assume my 16 bit CD is the final signal processing the audio will receive before people hear it.

KHStudio

  • Newbie
  • *
  • Offline Offline
  • Posts: 9
Re: 32bit Float to 24bit fixed?
« Reply #7 on: June 02, 2005, 02:09:51 AM »

jazzius wrote on Wed, 01 June 2005 11:02

KHStudio wrote on Tue, 31 May 2005 03:17

 
Myself and many others have heard huge differences in the way the audio sounds coming through the mixing board:  more dynamic, better separation, more alive and less harsh, etc...  


Perhaps you imagined a huge difference....wouldn't be the first time, won't be the last.




Im sorry, I didn't mean huge. Subtle. But to me, ANYthing that makes my music sound better is great.
Logged

James Perrett

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 339
Re: 32bit Float to 24bit fixed?
« Reply #8 on: June 02, 2005, 06:40:31 AM »

So far no-one has pointed out that, with a normal 32 bit float, you only get 24 bit resolution anyway (provided the exponent stays at zero). So Nuendo may not be doing anything bad to your sound provided the intermediate internal calculations are handled properly.

Cheers.

James.
Logged
James Perrett - JRP Music - Audio Mastering and Restoration.
http://www.jrpmusic.net

KHStudio

  • Newbie
  • *
  • Offline Offline
  • Posts: 9
Re: 32bit Float to 24bit fixed?
« Reply #9 on: June 02, 2005, 02:34:13 PM »

Not true. It's a very complex issue. You'll have to do some research to see what I mean.

There are 2 issues concerning dithering:
One within the app.= Mixing/Plugins

& one leaving the app.= Mastering/Playback

They both concern me but the 2nd is what I've been refering to.
Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: 32bit Float to 24bit fixed?
« Reply #10 on: June 04, 2005, 12:16:17 PM »

TotalSonic wrote on Wed, 01 June 2005 10:15

Speaking frankly -
I'm not sure I've ever truly heard the difference between 32bit (either fixed or floating) truncated just only one single time to 24bit vs. 32bit dithered to 24bit.  24to 16bit is really obvious though - but the LSB at 24bit is a heckuva quiet point.  

Anyone ever run blind tests on this??

Best regards,
Steve Berson


Yes, it's often a tossup blind. One generation of 32 bit float to 24 bit fixed truncation is extremely mild. BUT, it depends a lot on what kind and amount of processing precedes the truncation. The more complex the calculation, especially if there is a lot of filtering, the more likely you will start noticing the losses. We're reaching the point where we have to start evaluating cumulative degradation, e.g., if two or three in a row start to sound a bit cold or gritty or "small", then one of them is probably contributing just a little bit to that inevitable slide. Therefore, for safety's sake, dither all those truncations... And for those who say, "what if that grit is what we're looking for" I'd rather introduce my grittiness purposefully and have it be predictable than the result of a bad accident and fragile, unpredictable results.

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

KHStudio

  • Newbie
  • *
  • Offline Offline
  • Posts: 9
Re: 32bit Float to 24bit fixed?
« Reply #11 on: June 08, 2005, 07:49:50 PM »

I have been doing a lot of testing & have some new information regarding dithering that I‘d like to share. First, I apologize, standard dithering (1 bit without DC offset removal) is very hard to hear & most wouldn’t hear the difference. When I brought this topic up & said it was a huge difference I was referring to my settings:

OZONE Plug-in

DC offset = ON
Dither = type 2 (triangular on this plug)
Shaping = OFF
# of bits = 2
Auto blanking = ON

DC OFFSET = Turned on, 24 outputs to analog consol;
Was almost HALF of the difference I was hearing. I don’t know where the DC came from, probably a little from everywhere – processing & pre-amps? A little here & there adds up. Also, it is highly recommended to remove DC offset before dithering.
That said, when I tested Samplitude with the exact same settings minus the DC offset & auto blanking, the dithered song sounded better on playback.

BITs = I’ve been applying 2 bits of dither. Most dithering plugs don’t even give you an option & you’re stuck with 1.
 This is where I feel I made a pleasant mistake. I don’t care if it adds more noise – it sounds really good – adding a smoother/richer feel & nice separation to the contents of the mix. To me it’s a lot more real & “analog” sounding.
Who knows, I might have come up with a killer new trick. For me & my setup I’ll be using it a lot. BTW, the more you turn up the mix, or compress/limit for final mastering, it becomes very obvious – in or out of the computer.

Here is what Nika said concerning # of bits:

Quote:

KHStudio wrote:
Am I on the right track with the Ozone plug & settings I mentioned earlier?

Nika wrote:
I think so. But it's 2 bits of dither? That's not right. It won't do any harm other than add more noise than necessary. It should be 2 quanta of dither.



Quote:

KHStudio wrote:
Nika,
If you used Nuendo, knowing that it was 32 bit float & your converters were 24 bit:

Would you apply dither to the final outputs or not?

Nika wrote:
If I were a Nuendo engineer and I had to design this product I would always have dither on the trip out to the converters and would also have dither always on when bouncing to 24 bit fixed point files. The dither I would use for this would be TPDF dither with a maximum amplitude of 2 quantization steps and a bit depth of something like 24 bits (so that dither is added to bits 24-48 of the fixed point version of the signal prior to truncation).

If I were an audio engineer using this product and this was not done automatically then I would find a tool that could do it and I would add that tool at the output or the bounce to 24 bits.

KHStudio wrote:
And why?

Nika wrote:
In order to maintain linearity.

------------------------------------------------------------ --

Hopefully the information in these threads will be beneficial to someone besides me.
I have a better understanding of what’s going on & have learned a lot.

Thanks
KHStudio
Logged
Pages: [1]   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.056 seconds with 19 queries.