- 第 1 页 -
剛翻譯了SAWSTDUIO 說明書其中一小節有關Bit Resolution 和 Dither文章
可當作重溫, 但內文多少也反映了BOB LENTINI的觀感, 可以作為大家參考。
------------------------------------------------------------------------------------------------
近年來, bit resolution (比特精度) 的討論非常激烈, 討論的結果都模糊不清或有誤導成份。以下內容嘗試跳過這些混亂或至少導出Sawstudio 如何處理音頻數據的精度流向(audio data stream resolution)。
什麼是 Bit Resolution ?
數碼音頻訊號的每個採樣代表著在錄制這個採樣時所取得的瞬間音量。電腦就是以bit來表達採樣的數目值, bit數愈高,所表達的數值愈大。1 bit 表達個兩位數值 (0,1), 2 bits 可表達四個位數值 (0 1 2 3), 計算公式就是2的n 次方, n為 bit 數。
以高bit數表達音頻訊號, 就愈能將原來的聲音表現得仔細和精確。 16bit的數碼音頻訊號理論動態範圍為96db, 20bit為120db, 24bit則更高。當然有其他因素會影響實際上的動態範圍, 例如是擴音器, 溫度, 訊號線等。但無論如何, bit數愈高, 音頻數的結像能力就愈好。
什麼是 Dither ?
Dithering 主要目的是嚐試盡量咸少數碼音頻訊號由高bit數採樣降到較低bit數採樣的失真, 就如一個24bit或20bit的錄音需要降到16bit作為出版CD的用途一樣。
問題是, 當由高bit數降到低bit數的時候, 必須以某種方式放棄低bit數的數據, 這過程叫 “ Truncating”, 會引起量子化錯誤, 做成失真。
Dithering就是將一些特別的噪音以很小音量加到這個由高bit數採樣降到較低bit數的訊號, 這種心理上聽覺的保留方式, 目的在盡量模擬著本來需要由稍高的bit數來表達的聲音。
這種方式經常引起極大的爭議, 一些說很有效, 有些則不喜歡, 很多人甚至覺得沒有分別。當音頻由16bit轉到8bit時, 比較多人能感覺到 dithering 的好處, 但一到16bit這個層面, 爭論就開始不絕。
很多人意識到這個附加噪音其實就是一個低音量的”嘶嘶”聲音,這個附加噪音已經成為完美作品中不能缺少的一部份, 但事實上可能已經超過其實際上的優點了。
在任何情況之下, 我們應該用耳朵作出判斷, 音頻編輯和混音只是創作的過程, 最後仍要視乎個人的偏愛, 沒有對或錯。 Sawstudio 容許大家自由去選擇什麼是適合和較正確的做法, 你可以使用高bit resolution加上dither 去完成一個工作, 又或者以較低 bit resolution 而不用dither 去處理另一個工作, 一切由自己定調。
這個也放到此好讓大家可以輕輕了解一下不同的驅動操作模式
以下是BOB讲一些有关 asio, wdm, mme, dwave 驱动
-----------------------------------------------------------------------------------------------------------------------
ASIO... I don't like the loosness of it... drivers try to control the app... it should be the other way around, in my opinion.... the driver simply needs to move data to and from ram as efficiently as possible... not take control of everything and dictate buffer formats and sizes etc to the app.
我不喜欢 asio 驱动音图控制主应用软件, 它需要做的应该是以有效率的速度由内存记忆体移动数据, 而不是向主软体控制所有 和 指挥缓冲的格式与大少。
Too many formats... too much splitting and merging of data, in all kinds of possible formats... the app needs to try to handle way too many possibilities... its very complex inside an engine like SAWStudio...
太多不同格式, 太多须分道和紧急的数据, 也有太多不同格式需由 主应用软件去处理不同的可能性, 尤其好像 sawstudio 这样非常复杂的引擎。
ASIO is too easily stepped on by Windows and other threads... there is no room whatsoever for any interruptions in data flow... which will ALWAYS happen at the most inopportune times in Windows... when it does... ASIO spits out repeats of the last buffer... horrible sound...
asio 很容易受 windows 或其他工作影响, 没有足够空间给予中途需处理的数据, 造成错失处理数据之时间, 结果asio就会重复上一个缓冲, 这时怪声便会发出!!
I won't touch true WDM... its probably the worst driver model I've ever fooled with...
我不会碰真正的wdm, 它可能是我见过最不济的驱动!
At least with the MME, I can control sending multiple buffers ahead when I know Windows might interfere with data flow... like when its about to repaint 15 windows onscreen... or when the app is about to minimize...etc..
当 windows 有干扰数据的流程时, mme 驱动至少让我可预先送出多个缓冲, 如windows重划图案, 缩小视窗...等等。
With ASIO... there is only one buffer playing... and one filling... any interruptions whatsoever... and you have a glitch... and a nasty one at that in most cases...
asio 只能进行一个缓冲和 一个补充的, 任何干扰都会构成 "短时脉冲波形干扰" 和 任何可能性的噪音
With DWave... everything is different... DWave does not require any communication between the app and driver during playback or record... no thread interruptions... no context switching between Rings... the driver already knows where to look for the data and pulls it at its own hardware rate without needing to interrupt Windows in any way to ask the app for the next buffer... the app fills the buffers as far in front as it wants... when it wants... without any need to talk to the driver directly during playback or record... the difference is beyond better... at least in my opinion.
dwave 是全然不同, 当播放或录音时, 它不须要应用程式和驱动之间作沟通, 不受其他工作干扰.......驱动早已知道数据在那儿, 并以硬件本身的速度提取, 无须经windows向应用软件要求下一个缓冲.....应用软件在播放和录音亦无须通知驱动,以最快速度填补缓冲...dwave与其他的分别是超越了一般好的程度, 至少我认为。
I'd love to see more soundcards supporting it... but unfortunately, until SAWStudio has sales in the tens of thousands, that most likely wont happen.
我希望看到更多声卡支持 dwave, 可惜sawstudio 的市场占有率只有 1%, 看来实现的机会不太大。 (bob 太坦白, 所以其言可信
)
I wish they were more concerned with the possible performance and quality benefits, but as long as they can sell mass numbers of products with lots of advertising and political connections, things most likely wont change.
我渴望他们(厂商) 真的能够多关心声卡可行的表现和有利于品质的因素。 不过, 只要是能经过大量广告宣传和政治关系从而能卖出更多的产品的话, 这局面似乎难以改变。
Thanks to the former Sonorus and Dirk from the former Soundscape for jumping headfirst into making the idea come alive, so at least we could see it work and give those that said this model couldn't fly a chance to see it in the air.
多谢 前 soundscape 的 sonorus 和 dirk 首先令到我的主要能够实现, 至少能让我们知道这是可行的。
With more interest from other soundcard companies, we could really nail down any remaining loose ends and develop it to be what the standards driver models should have been.
若更多声卡厂商有兴趣, 我们就可真正知道dwave尚欠的东西, 从而发展出一个真正应有的驱动标准出来。
Heh.... you asked. (你们一起去要求吧!!)
Bob L
[ 本帖最后由 himhui 于 2006-3-6 10:24 编辑 ]
------------------------------------------------------------------------------------------------
近年來, bit resolution (比特精度) 的討論非常激烈, 討論的結果都模糊不清或有誤導成份。以下內容嘗試跳過這些混亂或至少導出Sawstudio 如何處理音頻數據的精度流向(audio data stream resolution)。
什麼是 Bit Resolution ?
數碼音頻訊號的每個採樣代表著在錄制這個採樣時所取得的瞬間音量。電腦就是以bit來表達採樣的數目值, bit數愈高,所表達的數值愈大。1 bit 表達個兩位數值 (0,1), 2 bits 可表達四個位數值 (0 1 2 3), 計算公式就是2的n 次方, n為 bit 數。
以高bit數表達音頻訊號, 就愈能將原來的聲音表現得仔細和精確。 16bit的數碼音頻訊號理論動態範圍為96db, 20bit為120db, 24bit則更高。當然有其他因素會影響實際上的動態範圍, 例如是擴音器, 溫度, 訊號線等。但無論如何, bit數愈高, 音頻數的結像能力就愈好。
什麼是 Dither ?
Dithering 主要目的是嚐試盡量咸少數碼音頻訊號由高bit數採樣降到較低bit數採樣的失真, 就如一個24bit或20bit的錄音需要降到16bit作為出版CD的用途一樣。
問題是, 當由高bit數降到低bit數的時候, 必須以某種方式放棄低bit數的數據, 這過程叫 “ Truncating”, 會引起量子化錯誤, 做成失真。
Dithering就是將一些特別的噪音以很小音量加到這個由高bit數採樣降到較低bit數的訊號, 這種心理上聽覺的保留方式, 目的在盡量模擬著本來需要由稍高的bit數來表達的聲音。
這種方式經常引起極大的爭議, 一些說很有效, 有些則不喜歡, 很多人甚至覺得沒有分別。當音頻由16bit轉到8bit時, 比較多人能感覺到 dithering 的好處, 但一到16bit這個層面, 爭論就開始不絕。
很多人意識到這個附加噪音其實就是一個低音量的”嘶嘶”聲音,這個附加噪音已經成為完美作品中不能缺少的一部份, 但事實上可能已經超過其實際上的優點了。
在任何情況之下, 我們應該用耳朵作出判斷, 音頻編輯和混音只是創作的過程, 最後仍要視乎個人的偏愛, 沒有對或錯。 Sawstudio 容許大家自由去選擇什麼是適合和較正確的做法, 你可以使用高bit resolution加上dither 去完成一個工作, 又或者以較低 bit resolution 而不用dither 去處理另一個工作, 一切由自己定調。
這個也放到此好讓大家可以輕輕了解一下不同的驅動操作模式
以下是BOB讲一些有关 asio, wdm, mme, dwave 驱动
-----------------------------------------------------------------------------------------------------------------------
ASIO... I don't like the loosness of it... drivers try to control the app... it should be the other way around, in my opinion.... the driver simply needs to move data to and from ram as efficiently as possible... not take control of everything and dictate buffer formats and sizes etc to the app.
我不喜欢 asio 驱动音图控制主应用软件, 它需要做的应该是以有效率的速度由内存记忆体移动数据, 而不是向主软体控制所有 和 指挥缓冲的格式与大少。
Too many formats... too much splitting and merging of data, in all kinds of possible formats... the app needs to try to handle way too many possibilities... its very complex inside an engine like SAWStudio...
太多不同格式, 太多须分道和紧急的数据, 也有太多不同格式需由 主应用软件去处理不同的可能性, 尤其好像 sawstudio 这样非常复杂的引擎。
ASIO is too easily stepped on by Windows and other threads... there is no room whatsoever for any interruptions in data flow... which will ALWAYS happen at the most inopportune times in Windows... when it does... ASIO spits out repeats of the last buffer... horrible sound...
asio 很容易受 windows 或其他工作影响, 没有足够空间给予中途需处理的数据, 造成错失处理数据之时间, 结果asio就会重复上一个缓冲, 这时怪声便会发出!!
I won't touch true WDM... its probably the worst driver model I've ever fooled with...
我不会碰真正的wdm, 它可能是我见过最不济的驱动!
At least with the MME, I can control sending multiple buffers ahead when I know Windows might interfere with data flow... like when its about to repaint 15 windows onscreen... or when the app is about to minimize...etc..
当 windows 有干扰数据的流程时, mme 驱动至少让我可预先送出多个缓冲, 如windows重划图案, 缩小视窗...等等。
With ASIO... there is only one buffer playing... and one filling... any interruptions whatsoever... and you have a glitch... and a nasty one at that in most cases...
asio 只能进行一个缓冲和 一个补充的, 任何干扰都会构成 "短时脉冲波形干扰" 和 任何可能性的噪音
With DWave... everything is different... DWave does not require any communication between the app and driver during playback or record... no thread interruptions... no context switching between Rings... the driver already knows where to look for the data and pulls it at its own hardware rate without needing to interrupt Windows in any way to ask the app for the next buffer... the app fills the buffers as far in front as it wants... when it wants... without any need to talk to the driver directly during playback or record... the difference is beyond better... at least in my opinion.
dwave 是全然不同, 当播放或录音时, 它不须要应用程式和驱动之间作沟通, 不受其他工作干扰.......驱动早已知道数据在那儿, 并以硬件本身的速度提取, 无须经windows向应用软件要求下一个缓冲.....应用软件在播放和录音亦无须通知驱动,以最快速度填补缓冲...dwave与其他的分别是超越了一般好的程度, 至少我认为。
I'd love to see more soundcards supporting it... but unfortunately, until SAWStudio has sales in the tens of thousands, that most likely wont happen.
我希望看到更多声卡支持 dwave, 可惜sawstudio 的市场占有率只有 1%, 看来实现的机会不太大。 (bob 太坦白, 所以其言可信
) I wish they were more concerned with the possible performance and quality benefits, but as long as they can sell mass numbers of products with lots of advertising and political connections, things most likely wont change.
我渴望他们(厂商) 真的能够多关心声卡可行的表现和有利于品质的因素。 不过, 只要是能经过大量广告宣传和政治关系从而能卖出更多的产品的话, 这局面似乎难以改变。
Thanks to the former Sonorus and Dirk from the former Soundscape for jumping headfirst into making the idea come alive, so at least we could see it work and give those that said this model couldn't fly a chance to see it in the air.
多谢 前 soundscape 的 sonorus 和 dirk 首先令到我的主要能够实现, 至少能让我们知道这是可行的。
With more interest from other soundcard companies, we could really nail down any remaining loose ends and develop it to be what the standards driver models should have been.
若更多声卡厂商有兴趣, 我们就可真正知道dwave尚欠的东西, 从而发展出一个真正应有的驱动标准出来。
Heh.... you asked. (你们一起去要求吧!!)
Bob L
[ 本帖最后由 himhui 于 2006-3-6 10:24 编辑 ]