Question about 24bit and dithering down to 16bit - Page 3 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #61 of 73 Old 07-26-2012, 04:17 AM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by amirm View Post


There is a twist here though. You can't play floating point values. You must get back to integer PCM samples.

Not necessarily true. Some converters create and play floating point data streams. The ancient Yamaha YM3812 was an example of such a chip.

It is possible to convert a floating point number into a voltage directly using two integer DACs. The first DAC converts the exponent into a voltage, and the second DAC is a multiplying DAC that apples the mantissa to the voltage created by the first DAC.
arnyk is offline  
Sponsored Links
Advertisement
 
post #62 of 73 Old 07-26-2012, 08:20 AM - Thread Starter
Member
 
nateo200's Avatar
 
Join Date: Jun 2012
Posts: 51
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
But how common are systems that can play anything over 24bit? I was under the assumption that even if you play 32bit float you will only be really playing 24bit integer...

Also what is "Sound Pressure"? My headphones have a its Sound Pressure listed as "104 dB (1mW/500Hz)" with a THD of less than 0.1%... what does that mean in terms of dynamic range? I looked up sound pressure and it seamed very similar in concept to dynamic range but with less focus on actual bit integers which was sort of confusing.
nateo200 is offline  
post #63 of 73 Old 07-26-2012, 10:50 AM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 133
Quote:
Originally Posted by nateo200 View Post

So I believe my question went unanswered or I got lost in the flaming but is there an issue with going from 88.2khz to 48khz or 176.4khz to 96khz or whatever combination like that? I know the "multiples of 44.1/48khz" theory for lack of better wording but is it a real factor and if so explain?

Arny already answered that, and in my original post I explained that rendering a mix using more than 16 bits is not useful. The same goes for sample rates higher than 44.1 KHz. More data does not give better sound quality, even though people who sell gear want you to believe otherwise so they can keep selling you more stuff. However, using more data doesn't harm quality either, and there are legitimate reasons that DAW software uses 32-bit FP math for its internal calculations.

I understand people preferring to record at 24 bits (at 44.1 KHz) even though it doesn't sound better, because in theory it offers a safety net. But it doesn't sound better, and only wastes disk space and DAW throughput. So the correct answer is don't use higher sample rates in the first place, and then you don't have to use SRC at all when you render.

--Ethan

RealTraps - The acoustic treatment experts

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Ethan Winer is offline  
post #64 of 73 Old 07-26-2012, 10:52 AM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 133
Quote:
Originally Posted by nateo200 View Post

But how common are systems that can play anything over 24bit?

The Sound Forge software I use can save and read 32-bit FP files. I don't know if any media players can do this, but maybe some can. It's a non-issue because saving audio as 32 bits is an even bigger waste than at 24 bits. biggrin.gif
Quote:
Also what is "Sound Pressure"? My headphones have a its Sound Pressure listed as "104 dB (1mW/500Hz)" with a THD of less than 0.1%... what does that mean in terms of dynamic range? I looked up sound pressure and it seamed very similar in concept to dynamic range but with less focus on actual bit integers which was sort of confusing.

Sound Pressure is an absolute acoustic volume level. The spec you listed is actually for distortion, but it states the SPL (Sound Pressure Level) as a reference since most transducers have more distortion at louder volumes.

--Ethan
nateo200 likes this.

RealTraps - The acoustic treatment experts

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Ethan Winer is offline  
post #65 of 73 Old 07-26-2012, 12:12 PM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by nateo200 View Post

But how common are systems that can play anything over 24bit?

Does it matter? There is no commercial player that can create an analog signal with anything like 24 bit accuracy. There still are no DACs that have anything like 24 bit accuracy.
Quote:
I was under the assumption that even if you play 32bit float you will only be really playing 24bit integer...

24 bit integers only exist in the world of numbers. They don't exist in the world of real voltages or real sound pressures.
Quote:
Also what is "Sound Pressure"?

Seems pretty self-explanatory!

Is Wikipedia broken where you are? ;-)

Also read the article on Sound Pressure Level.
Quote:
My headphones have a its Sound Pressure listed as "104 dB (1mW/500Hz)" with a THD of less than 0.1%... what does that mean in terms of dynamic range?

The dynamic range in that spec comes from the "THD of less than 0.1%". 0.1 THD suggests dynamic range of 80 dB. Dynamic range is not just about how loud, but also about how clean.

THD of less than 0.1%
I looked up sound pressure and it seemed very similar in concept to dynamic range but with less focus on actual bit integers which was sort of confusing.[/quote]

It sounds to me like you may be getting lost in the numbers.

Dynamic range is not just about how loud, but also about how clean.
arnyk is offline  
post #66 of 73 Old 07-26-2012, 12:53 PM - Thread Starter
Member
 
nateo200's Avatar
 
Join Date: Jun 2012
Posts: 51
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by arnyk View Post

Does it matter? There is no commercial player that can create an analog signal with anything like 24 bit accuracy. There still are no DACs that have anything like 24 bit accuracy.
Just curious, I remember reading a thread about people fighting over whether to capture at 16bit, 24bit or 32bit float and it was just like someone vomitted audiophile all over the place :O Though I realize that there are 16bit DACs out their that are more accurate than 24bit DACs (my receiver is sort of in the category of crappy DAC's but I have essentially no budget at the moment so I can't complain)
24 bit integers only exist in the world of numbers. They don't exist in the world of real voltages or real sound pressures.

Seems pretty self-explanatory!
Is Wikipedia broken where you are? ;-)
Also read the article on Sound Pressure Level.
The dynamic range in that spec comes from the "THD of less than 0.1%". 0.1 THD suggests dynamic range of 80 dB. Dynamic range is not just about how loud, but also about how clean.
THD of less than 0.1%
I looked up sound pressure and it seemed very similar in concept to dynamic range but with less focus on actual bit integers which was sort of confusing.
It sounds to me like you may be getting lost in the numbers.
Dynamic range is not just about how loud, but also about how clean.[/quote]

No I didn't think dynamic range was just about loudness at all. Was just trying to understand how sound pressure and dynamic range are related or converted. I don't know how you came up with a dynamic range of around 80dB from a THD of less than 0.1% with a sound pressure of 104dB...I'm more of a visual learner than a numbers guy so yeah I am lost in numbers. I looked up the wikipedia on Sound Pressure Level before I posted (as I always do, been a moderator on other forums so I've been saying "Search before you post" for a while ;-) Well my headphones sound pretty clean to me and I use them when mixing my surround mix's before I test out the accuracies of the panning, placement of certain sound elements, etc. that can only really be measured in multichannel, although my surround set up is sort of crappy...its basically set up to enjoy multichannel playback in 6.1 but the speakers are all a bit old and compared to my headphones sound like they have no range of sound :/
nateo200 is offline  
post #67 of 73 Old 07-26-2012, 03:06 PM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by nateo200 View Post

I don't know how you came up with a dynamic range of around 80dB from a THD of less than 0.1% with a sound pressure of 104dB..]

Here is my logic.

The definition of dynamic range is( Maximum Undistorted Amplitude)/(Noise+Distortion). IOW very much like SNR except that you also include all distortion

In this case I felt safe assuming that the distortion was much more than the noise.

The distortion was 0.1% which corresponds to -60 dB, Whoops, I misread 0.1 as 0.01. My bad!

Since the SPL is 104 dB this is better described as the Dynamic range @ 104 dB SPL. The SPL was the independent variable.
arnyk is offline  
post #68 of 73 Old 07-26-2012, 03:16 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,082
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 753 Post(s)
Liked: 434
Quote:
Originally Posted by arnyk View Post

Not necessarily true. Some converters create and play floating point data streams. The ancient Yamaha YM3812 was an example of such a chip.
The 3812 was an FM Synthesis chip. I suspect you are thinking of the companion 3014 DAC. (Yes, I had to look them up smile.gif ) That chip was "floating point" but not in the manner you think or imply. It used a format of 10 bits with 3 bits of exponent. That is not remotely the same as 32 or 64 bit floating point math used in computer CPUs. IEEE floating point representation there is 23 bits for the mantissa and 8 bits for exponent for single precision. Double precision is 52 bits for 11 bits for exponent. No way could you have fed those floating pint values to that old DAC or any current one. A bicycle wheel doesn't perform the same function as a car tire even though they are both round.....
Quote:
It is possible to convert a floating point number into a voltage directly using two integer DACs. The first DAC converts the exponent into a voltage, and the second DAC is a multiplying DAC that apples the mantissa to the voltage created by the first DAC.
Anything is "possible" but mentioning them adds confusion to the topic especially when the reference is no applicable to the context being discussed. Lest you want to tell us that resampling algorithms on computers use 10 bits for mantissa and 3 bits for exponent....

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #69 of 73 Old 07-26-2012, 05:23 PM - Thread Starter
Member
 
nateo200's Avatar
 
Join Date: Jun 2012
Posts: 51
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Ethan Winer View Post

Arny already answered that, and in my original post I explained that rendering a mix using more than 16 bits is not useful. The same goes for sample rates higher than 44.1 KHz. More data does not give better sound quality, even though people who sell gear want you to believe otherwise so they can keep selling you more stuff. However, using more data doesn't harm quality either, and there are legitimate reasons that DAW software uses 32-bit FP math for its internal calculations.
I understand people preferring to record at 24 bits (at 44.1 KHz) even though it doesn't sound better, because in theory it offers a safety net. But it doesn't sound better, and only wastes disk space and DAW throughput. So the correct answer is don't use higher sample rates in the first place, and then you don't have to use SRC at all when you render.
--Ethan

Well I'm still going to use 48khz for practical reasons. I just don't like 44.1khz because film audio is almost always 48khz. The Dolby Digital spec allows for different sample rates but 48khz is the most common and I believe the only allowed one for DVD and Blu-ray...regardless I always like knowing that 48khz can theoretically capture 24khz according to the Nyquiest theory(sp?) where as 44.1 can capture 22.05khz, and while we can't here past that for the most part I can hear up to 22khz (Just barely, but I'm younger biggrin.gif) and while its a more annoying frequency than beneficial if its present I want to here it! None of my family can hear the dog whistle but me so when they blare it I'm the first one to complain meanwhile my dad looks at me as if Im crazy! biggrin.gif As for 24bit, I think I'm going to forget capturing at 32bit float and capture at 24bit, I can't go down to 16bit, I don't know what it is but 24bit gives me piece of mind that is worth more than disk space, 32bit float isn't worth disk space for me now that I think about it especially since Im ALWAYS low on hard drive space.
Quote:
Originally Posted by arnyk View Post

Here is my logic.
The definition of dynamic range is( Maximum Undistorted Amplitude)/(Noise+Distortion). IOW very much like SNR except that you also include all distortion
In this case I felt safe assuming that the distortion was much more than the noise.
The distortion was 0.1% which corresponds to -60 dB, Whoops, I misread 0.1 as 0.01. My bad!
Since the SPL is 104 dB this is better described as the Dynamic range @ 104 dB SPL. The SPL was the independent variable.
Ah ok makes much more sense! Thanks a bunch! So how "good" performance is that for my headphones? Poor?
nateo200 is offline  
post #70 of 73 Old 07-26-2012, 08:02 PM
AVS Special Member
 
DonH50's Avatar
 
Join Date: Feb 2010
Location: Monument CO
Posts: 6,087
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 159 Post(s)
Liked: 262
Dynamic range is defined differently by different folk...

SINAD = signal to (noise and distortion) ratio; basically the ratio of signal to everything else.

SNR = signal to noise ratio; some include distortion, some don't. For those that do not include distortion (a popular choice) SNR will usually be higher than SINAD.

SFDR = spurious-free dynamic range = ratio of signal to highest peak (noise, distortion, whatever); rather than integrating (RSS'ing) all terms, SFDR just looks at the signal and single highest spur so is typically higher than SINAD or SNR.

For an ideal ADC or DAC, quantization noise sets SNR = 6.021 * N + 1.76 dB (SINAD is the same if there is no distortion) and SFDR ~ 9*N dB for a sinusoidal signal. SFDR is higher than SINAD because SFDR is a spot measurement of the signal to the single highest spurious component; SNR and SINAD is the ratio of signal to the sum (technically the root-sum-square, RSS value) of all other (non-signal) components.

The ideal SNR is not too bad to derive mathematically; the derivation of SFDR for an ideal data converter is a bit painful, at least for a hairy-knuckled engineer like me (Bessel functions and other fun stuff).

The IEEE (Standard 1241) uses SINAD to define ENOB, the effective number of bits: ENOB = (SINAD - 1.76) / 6.021 bits.

HTH - Don

"After silence, that which best expresses the inexpressible, is music" - Aldous Huxley
DonH50 is online now  
post #71 of 73 Old 07-26-2012, 11:26 PM - Thread Starter
Member
 
nateo200's Avatar
 
Join Date: Jun 2012
Posts: 51
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
I realize this probably isn't proper forum edaquet(sp?) but I have some questions regarding sound mixing that I know a bit about but Id appreciate it greatly if ANYONE with experience could lend me a hand. I feel as though its impossible to get my foot in the door of sound engineering and I'm trying to really grasp some more advanced stuff as well as stuff that I may have read before but need clarified. I made a proper separate thread in the link below.
http://www.avsforum.com/t/1422015/question-about-dolby-digital-ex-and-dts-es-matrix

I do short films and while I understand the concepts of the picture, sound is half the film and I want to create the most impressive soundtracks I can and I have a gut feeling I can do it as Ive made short skids with picture quality and a rock solid multichannel track that impressed.
http://www.avsforum.com/t/1422015/question-about-dolby-digital-ex-and-dts-es-matrix

Quote:
Originally Posted by DonH50 View Post

Dynamic range is defined differently by different folk...
SINAD = signal to (noise and distortion) ratio; basically the ratio of signal to everything else.
SNR = signal to noise ratio; some include distortion, some don't. For those that do not include distortion (a popular choice) SNR will usually be higher than SINAD.
SFDR = spurious-free dynamic range = ratio of signal to highest peak (noise, distortion, whatever); rather than integrating (RSS'ing) all terms, SFDR just looks at the signal and single highest spur so is typically higher than SINAD or SNR.
For an ideal ADC or DAC, quantization noise sets SNR = 6.021 * N + 1.76 dB (SINAD is the same if there is no distortion) and SFDR ~ 9*N dB for a sinusoidal signal. SFDR is higher than SINAD because SFDR is a spot measurement of the signal to the single highest spurious component; SNR and SINAD is the ratio of signal to the sum (technically the root-sum-square, RSS value) of all other (non-signal) components.
The ideal SNR is not too bad to derive mathematically; the derivation of SFDR for an ideal data converter is a bit painful, at least for a hairy-knuckled engineer like me (Bessel functions and other fun stuff).
The IEEE (Standard 1241) uses SINAD to define ENOB, the effective number of bits: ENOB = (SINAD - 1.76) / 6.021 bits.
HTH - Don
Thanks a bunch! Definitely a useful post for me. I will most likely be coming back to this thread over and over for the various informative posts as well as links to videos, pages, etc. that I'm looking forward to reading. Glad I joined this forum! I always love having people above me in knowledge and above all learning from them.
nateo200 is offline  
post #72 of 73 Old 07-27-2012, 10:24 AM
AVS Special Member
 
DonH50's Avatar
 
Join Date: Feb 2010
Location: Monument CO
Posts: 6,087
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 159 Post(s)
Liked: 262
My mixing days are mostly behind me. There are a lot of folks here with current SOTA skills. Please start a new thread, in this sub-forum is probably as good as any. We have members who do professional film mixing for a living that could certainly help. There are also tons of books on mic'ing, mixing, etc. and several schools and classes that would be worthwhile taking. There is an art to it, but a lot of basics that lay the groundwork for the art.

One of the worst wake-up calls in my life was taking a graduate course on acoustics, and spending the first few weeks wading through nasty integral-differential wave equations insteading of learning how speakers worked and which knobs on the console to turn... smile.gif

Glad you found my post helpful, thanks! - Don
nateo200 likes this.

"After silence, that which best expresses the inexpressible, is music" - Aldous Huxley
DonH50 is online now  
post #73 of 73 Old 07-27-2012, 12:25 PM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 133
Quote:
Originally Posted by nateo200 View Post

I'm still going to use 48khz for practical reasons. I just don't like 44.1khz because film audio is almost always 48khz.

Yes, if you make music and soundtracks for video that goes on DVDs, definitely use 48 KHz. However, video that's meant for YouTube etc can be 44.1 KHz with no re-sampling needed.

--Ethan

RealTraps - The acoustic treatment experts

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Ethan Winer is offline  
Reply Audio theory, Setup and Chat

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off