CAS - 終不悔(一) : USB 篇

k500a

引子
CAS 是一場與噪音的持久戰。
有一天跟一位不是發燒友的IT人談起CAS，他一臉不以爲然，為甚麽要花那麼多的時間去用linear filter, SSD,…，他認為最後的data應該是bit perfect ，因為有checksum, error detection, error correction ⋯等等的科技使用在data 的傳輸和接收，所以在DAC 的那一邊收到的音樂檔案跟在storage (SSD, NAS )上面的檔案是一模一樣，所以没有必要浪費金錢和時間弄PC 上的filtering,power supply…。
但他也提到一疑問：「是不是在處理音樂檔案時，電腦會有噪音産生，而這些噪音經由某些途徑而影響DAC 的analog output」。一言驚醒。


於是我便開始看看有怎麼方法把PC產生的電噪音隔絕在DAC 以外。
參考了網上各位先進和高手的經驗便按圖索驥，堆砌出一個解決方案來。

由電腦產生出來的電噪音，我相信是會經由三種途徑傳到DAC 那邉去。
1. USB
2. Ethernet Lan
3. AC power supply

要隔離這三種傳送噪音途徑，我採用以下方法:
1. USB
以前跟大家介紹過的Adnaco s3b。也比較過其他 PCIe filter 方案，還是這個最好 。留意這個要跟PC分開供電源。

絕到溶掉我

2. Ethernet Lan
Lan Fiber converter - 在網上看到其實他師兄也有用這個東東。用光纖把PC 和 LAN 連在一起。通過光纖避免了PC的電噪音對電源的污染。這位師兄介紹的KTI Conveter是一個很高質素的產品，但在香港沒有渠道可購買，最後還是淘了一對另外一個品牌的產品回來，主要是它有支援Single mode 和 duplex. 功耗也少於2.5W這個對下面 3. 有很大影響。


光速的愛


3. AC power supply
要杜絕電噪音回流到電源，最簡單就是用電池供電。但電腦的功耗不能太高，我用的是intel DN2800MT 功耗是十多W而已，平民一點可以用SLA 電池。怕麻煩也可訂購紅酒產品。

如濃酒可醉掉我

最後整體的配置便是如下圖了:



CAS 經過以上處理後噪音的影響可大幅改善了。
還有是Linux MPD 跟 Winodows JPLAY 在聽感上是互相靠攏了。
雖然不是一個非常出色的CAS系統，但目前我是沒有怎麼不滿了。

可以享受音樂了！

frankyhpm

師兄有冇比較過究竟邊一part改變而影響最大呀?另外，PC用fiber output過部分唔係好明

obee

師兄有無玩過 Linux 上的 "Jack"?

frankyhpm

原來係透過PCI-E將USB extend成為外置device，睇完斷片明白晒
http://www.youtube.com/watch?v=2Eq71JC7aXs

Bomb_Dance

看到那個單模光纖就知道是好東西呀……

k500a

frankyhpm 發表於 2013-3-15 23:29
師兄有冇比較過究竟邊一part改變而影響最大呀?另外，PC用fiber output過部分唔係好明 ...

frankyhpm C-hing,
最大影響是1. 其次是 3. 最後是 2.

k500a

brkkh 發表於 2013-3-16 01:37
"光之妙囊"係廿一世紀CAS界十大發明之一

brkkh C-hing,
對呀，謝謝您的補充。第1點的adnaco 的電腦介面轉化為光纖技術產品在其它中文論壇稱之為"光之妙囊"。
在我的配置上幫助很大。

k500a

obee 發表於 2013-3-15 23:39
師兄有無玩過 Linux 上的 "Jack"?

obee C-hing,
在linux 上面我的道行很淺，目前Linux CAS 只玩過 Linux Voyage MPD。0.9-cr1以正式支援DSD over PCM。Linux Voyage MPD在它的網站上有比較詳細的安裝指引，網上找資料比較容易，安裝在普通的PC上也可以。這個方便dual boot 比較MPD 和 JPLAY。

k500a

Bomb_Dance 發表於 2013-3-16 01:47
看到那個單模光纖就知道是好東西呀……

Bomb_Dance C-hing,
好眼力。

feellikecyber

CHing, 借問Mac平台有何見解....小弟Mac Mini2010(內置火牛)用家

無聊問題: 由Mac Mini用Fiber直入DAC米可以吾洗用USB (又好聲D?)

obee

k500a 發表於 2013-3-16 07:14
obee C-hing,
在linux 上面我的道行很淺，目前Linux CAS 只玩過 Linux Voyage MPD。0.9-cr1以正式支援DSD ...

明白了, 唔該

cwtim01

obee 發表於 2013-3-15 23:39
師兄有無玩過 Linux 上的 "Jack"?

JACK is most useful in the following applications:
- when you need to synchronize audio between multiple audio applications, important for studio mixing and editing
- when your music server/player doesn't support ALSA directly, so you have to go through JACK (very rare these days)
- when you want to be sure that there is no X-runs under low latency settings

Otherwise, JACK is an additional component in the audio chain, and should be avoided for audiophile use, to get the best sound possible. For example, in my own customed Linux setup:
moc->jackdbus->alsa, minimum latency for 24/192k music is 3ms, otherwise xruns occurs
moc->alsa directly, minimum latency for 24/192k music is 1ms

k500a

cwtim01 發表於 2013-3-16 08:50
JACK is most useful in the following applications:
- when you need to synchronize audio between mu ...

cwtim01 c-hing,
非常清楚的說明，謝謝寶貴意見。

DigitalMana

我都曾經玩CAS玩到头痛（其實系卑CAS玩，唔系玩CAS ），用Linux+MPD，隔離USB，分體電池電源等等。

現發覺簡簡單單加一個靜過電池的iFi iUSB Power系DAC前面已經達到幾好境界(當然computer上的tuning重系需要的，只是電源同隔離簡單好多），大家不妨一試。



http://ifi-audio.com.cn/en/iUSB.html

k500a

DigitalMana 發表於 2013-3-16 13:13
我都曾經玩CAS玩到头痛（其實系卑CAS玩，唔系玩CAS ），用Linux+MPD，隔離USB，分體電池電源等等 ...

DigitalMana C-hing,
有機會一定會試這個iFi iUSB Power。

morel

Ching太靜唔一定係好事!

k500a

morel 發表於 2013-3-16 14:19
Ching太靜唔一定係好事!

morel C-hing,
是甚麼原因呢？請指教。

obee

cwtim01 發表於 2013-3-16 08:50
JACK is most useful in the following applications:
- when you need to synchronize audio between mu ...

我想請教師兄, 師兄所知, 今天為止, 有邊個 Linux CAS 玩法係較好聲? 只談軟件方面, 不考慮電源, 光纖等硬件.

chuck

The whole setup is quite overwhelming to me, especially the number of power supplies involves. As long as the whole set up is giving you big improvement in sound quality from before, that would mean a lot of good experience and personal satisfaction.

Have you considered having a network streaming player as well ?

cwtim01

obee 發表於 2013-3-16 15:20
我想請教師兄, 師兄所知, 今天為止, 有邊個 Linux CAS 玩法係較好聲? 只談軟件方面, 不考慮電源, 光纖等 ...

My vote goes to serial console mode Linux running MOC. For a CAS system, there are 3 things you fight:
1. Noise - the way k500a C-hing has done is a classic good example of minimizing noise
2. Vibration - it's the easy bit, just make sure the PC is fanless, replace all harddisk with SSD, and have a good thick chassis
3. Jitter - that's the part the OS and the software comes to play

Following are some ways to reduce jitter:
- Use custom compiled real-time kernel, assign the highest priority to software components of the audio chain, remove the kernel support for ALL unnecessary hardware and features
- Disable ALL unnecessary hardware, which will contend for system attention and hence potential cause of jitter, as well as creating noise. This is the reason for serial console - because the video card will be removed physically, or at least disabled completely. Clearly the network port have to be disabled as well, so streaming from NAS is not an option anymore.
- Disable ALL unnecessary software, which contend for system resources and generate unnecessary interrupts (for that I compile my own Linux system from scratch, compile only the software components that are absolutely necessary, http://www.linuxfromscratch.org/lfs/)
- Configure the music player software to force a lowest possible latency, but without overflows (or X-runs)
- Disable power saving to be sure that the CPU won't change frequency from time to time
- Force use of high frequency clock "hpet" instead of the normal clock "rtc", force use of highest possible frequency for the hpet clock
- Use ext2 (instead of ext3 or ext4) to avoid journaling and unnecessary disk writes
- the list go on and on ...

MPD server and streaming from NAS can give good sound, but due to the additional network protocol stack required (NFS, SMB ... etc), the need to fetch files from a remote NAS, these will introduce latency and jitter. It would be difficult for an MPD based system to reach a latency of 5ms or below. (Remark: Some audio science engineers say that for latency below 5 ms, the difference is indistinguishable by human ears, so there may be not point to go below 5ms)