怎麼愈來愈多人曲解SRC

Ares · 文章由 **Ares** » 2006-03-30 19:31

I just wonder that adobe audition(or Cooledit Pro) made the line between dots for visualization,but it didn't apply any fomula to the curve,that is to say,the waveform consists of time information and amplitude,not the formula.In that case,I think figure 10 is ok,although it may not be very precise and suitable for some case.Whether it presents as curve or histogram,the waveform is just amplitude vs time at all...

The reason,I think,which causes the transformed data different from the original should be the main question.

Forgive my poor English.I'm not an expert.If I've been far away from the topic,please ignore me and go on

Peace and cheers.

bennetng · 文章由 **bennetng** » 2006-03-30 23:32

Adobe Audition 也是要 zoom in 到相當程度才會看到點和曲線的, 我猜它應該是在那時才實時計算把曲線描出來, zoom in 得不夠近時也會看到疑似「階梯波」的現象

回想起很久以前初次接觸 wavlab, sound forge, 和 cool edit pro, 我就對 cool edit pro 一見鐘情了, 它簡潔的界面, 詳細的說明文件 (例如很詳細的解釋了 dither 的原理) 實在教懂了我很多

JamesT · 文章由 **JamesT** » 2006-03-31 00:19

Ares 寫:I just wonder that adobe audition(or Cooledit Pro) made the line between dots for visualization,but it didn't apply any fomula to the curve,that is to say,the waveform consists of time information and amplitude,not the formula.In that case,I think figure 10 is ok,although it may not be very precise and suitable for some case.Whether it presents as curve or histogram,the waveform is just amplitude vs time at all...

The reason,I think,which causes the transformed data different from the original should be the main question.

Forgive my poor English.I'm not an expert.If I've been far away from the topic,please ignore me and go on
Peace and cheers.

The argument in your first paragraph is wrong.

Pulse Amplitude Modulation, as a modulation, which means to represent something via another form, is meant to represent continuous signal via discrete pulses. The "continous signal" mentioned above equals the "formula" in your words. The reason why discrete PAM can represent continuous signal is -- the Nyquist sampling theory. Signals of 1/2 sampling frequency or lower can be losslessly represented by means of PAM provided you have absolute accuracy and infinite arithmetical capability.

I suggest you post articles in a language which you handle better than English, lest you cannot accurately express things in your mind and may result in more misunderstanding. And if you intended to be ignored in the first place, you shouldn't have posted anything at all.

Ares · 文章由 **Ares** » 2006-03-31 10:03

JamesT 寫: The argument in your first paragraph is wrong.

Pulse Amplitude Modulation, as a modulation, which means to represent something via another form, is meant to represent continuous signal via discrete pulses. The "continous signal" mentioned above equals the "formula" in your words. The reason why discrete PAM can represent continuous signal is -- the Nyquist sampling theory. Signals of 1/2 sampling frequency or lower can be losslessly represented by means of PAM provided you have absolute accuracy and infinite arithmetical capability.

I suggest you post articles in a language which you handle better than English, lest you cannot accurately express things in your mind and may result in more misunderstanding. And if you intended to be ignored in the first place, you shouldn't have posted anything at all.

Sorry if I made your misunderstanding.I know the Nyquist theory and the restoration of 1/2 Nyquist frequency.Maybe we are telling different stroies.I guess Ssss1 didn't intend to talk about the signal which has been reproduced at the final stage,but he focused on the digital part before the data being regenerated by the DAC.I didn't said anyone is incurrect,but I just want to said maybe Ssssss1's intention is to explain why the different sampling rate will cause those controversial results and he try his best to put everything into a simple explaination.I think a working SRC is the most interesting part in this topic and many prople are willing to see what really happened in SoundCard's SRC and SSRC.

There are many respectful experts in this fourm and I appreciate them.There is no need to have quarrel with each other from a single word.I sincerely hope we can have a peaceful discussion.

I think Shade will agree that too!

Cheers!

PS.I want to type in chinese,but mine is broken

bennetng:
Good to see you again!I was curious whether X-fi implement any dithering or not,In APS driver,dithering can be turned off,do you have any idea?

JamesT · 文章由 **JamesT** » 2006-03-31 11:57

To Ares,

ssssssss001 在他第一篇發言(本討論串第3頁)當中，對於數位取樣的描述以及示意，是有一些錯誤的，這是無可否認的。這與他想針對什麼東西做討論並沒有關係。

關於「許多人想知道音效卡的 SRC 與各種軟體 SRC 例如：SSRC、PPHS、KMixer...等，它們到底真的做了什麼...」這一個問題，我想提出一個此問題背後的問題：請問這些人他們想要瞭解這個問題，其動機為何？請問這些人他們想要瞭解這個問題，是想瞭解到哪個程度？

事實上，在這個討論區當中，要將這類涉及工程專業的問題圓滿的呈現給所有閱讀這個討論區的人，是有相當大的困難的。因為每一位不同讀者的需求，存在著相當大的差異，而每一位發言者的切入角度，也都有所受限。

所以，爭執，並不是一件壞事，只要每位參與者都能保持學習的態度，那麼即使在強烈的爭執之後，也能夠幫助到深廣的層面；反之，如果參與的人，並不是抱持學習的態度而來，那麼即使用再怎麼文雅婉轉的辭藻，心中已樹起了成見，背後的意義仍然是不歡而散。

Ares · 文章由 **Ares** » 2006-03-31 12:17

JamesT 寫:To Ares,

ssssssss001 在他第一篇發言(本討論串第3頁)當中，對於數位取樣的描述以及示意，是有一些錯誤的，這是無可否認的。這與他想針對什麼東西做討論並沒有關係。

關於「許多人想知道音效卡的 SRC 與各種軟體 SRC 例如：SSRC、PPHS、KMixer...等，它們到底真的做了什麼...」這一個問題，我想提出一個此問題背後的問題：請問這些人他們想要瞭解這個問題，其動機為何？請問這些人他們想要瞭解這個問題，是想瞭解到哪個程度？

事實上，在這個討論區當中，要將這類涉及工程專業的問題圓滿的呈現給所有閱讀這個討論區的人，是有相當大的困難的。因為每一位不同讀者的需求，存在著相當大的差異，而每一位發言者的切入角度，也都有所受限。

所以，爭執，並不是一件壞事，只要每位參與者都能保持學習的態度，那麼即使在強烈的爭執之後，也能夠幫助到深廣的層面；反之，如果參與的人，並不是抱持學習的態度而來，那麼即使用再怎麼文雅婉轉的辭藻，心中已樹起了成見，背後的意義仍然是不歡而散。

Yes,I know and thank you for your comments.I read the entire argument again and I finally realize what all of you are talking about.I think Sssss1 is talking about AD,and other people are talking about the representation of DA.I guess Figure 10 is not the picture of the final DA result,but the initial process of AD.Maybe that is why make us so confused.

To discuss the mechanism behind the SRC is quite a big mass,but I'm glad to see many people try to share their knowledge and I really learned a lot from them.

Hope the discussion will be moved on.

JamesT · 文章由 **JamesT** » 2006-03-31 13:19

古典的 AD 概念是 sample and hold，所以 bennetng 才會說圖畫錯了，sample and hold 的圖是像 bennetng 後續提供的那樣。對於數位取樣的概念也應該要稍微瞭解工程數學才能比較清楚知道它在幹嘛。

如果要再進一步探討現今實際的 AD 則必須涉入 PWM 與 delta-sigma 以及更多純 digital domain 相關的 signal processing 等議題。我個人認為，把這些東西搬到這個討論區上面，並沒有太大的意義和幫助，反而帶來更多困擾，因為這些東西並不應該歸類於常識，同時也不是生活必須品，過於簡化地去介紹這些東西容易造成誤解，因為它們本來就是建立在一定程度以上的數學基礎之上。

我覺得就使用者的觀點而言，作用、效能、比較、使用細節...等，在討論區上，這些資訊應該會比講理論來得有價值些。對於想深入瞭解的人，應該去找書來看或去修習相關的課程，會比用討論區來得有效。

Ares · 文章由 **Ares** » 2006-03-31 14:42

受益良多,感謝大大指點!!
(在學校圖書館偷上網XD)

bennetng · 文章由 **bennetng** » 2006-03-31 14:50

x-fi 的 record dither 是強制開啟的, 想不用的話, 除非用 asio
另外, creative 網站最近釋出了新版 vienna
===================

說回那個「是否曲線」的問題, 這個概念對一般使用者是有實際作用的, 比如說很多人以為在 foobar 裡不應用 replaygain, 不應用 volume control, 因為數位音訊音量越低, 音質越差。但之前我貼的方波例子卻反映了就是因為曲線的問題, 原本沒破音的音訊在實際播放時其實是會有機會破音的, 遇過 foobar2000 的 console window 在播放時會出現 clipping detected 的訊息嗎? 以網路名曲 orz.mp3 為例

http://zonble.twbbs.org/archives/2004_11/622.php

用 foobar2000 以 diskwriter 用 32-bit float 轉成 wave 後, 可見到 Peak Amplitude 是大於 0dB, 而 clipping 的次數是以十萬計, 但如果用 replaygain 或 volume control 把音量先衰減才播放, 是不會出現 clipping 的

rogerman · 文章由 **rogerman** » 2006-04-07 19:54

小弟用A2ZS感覺比內建要好太多了
對SRC倒沒有這麼在意