
-
2020-09-15, 09:43 AM
#6021
-
The Following 8 Users Say Thank You to sam0402 For This Useful Post:
-
2020-09-15, 10:29 AM
#6022
此篇文章於 2020-09-15 10:43 AM 被 tetsuka 編輯。
-
The Following 12 Users Say Thank You to tetsuka For This Useful Post:
andyhang,babymlin,bchsieh,chrischang0624,Deen,gangster.tank,jimmypan,kevintran,pawaslider,psycho,sam0402,stchen
-
2020-09-15, 11:40 PM
#6023
暫時只需要購買最便宜的型號就行了,反正好戲在後頭,明年 CM4 推出的時候還可以計劃如何設計真正發燒級數的載板,既然有 PCIe 的話 CM4 就能夠安裝 JCAT 的 USB Card XE 或 NET Card XE 了:
The Pi Cast: AMA with Raspberry Pi Founder Eben Upton
https://www.youtube.com/watch?v=fTJ5eLn58to
https://www.tomshardware.com/news/ra...support-coming
“The Raspberry Pi Compute Module, CM4, we will support NVMe to some degree on that, because of course it [Raspberry Pi 4] has a PCI Express channel,” he said. “We have a single lane Gen 2 which is used to supply USB 3 on the Raspberry Pi [4]. On the [Compute] Module that would be exposed to the edge connector and we’re likely to support NVMe over that.”
這是 CM3 加上發燒級數的載板:
https://allo.com/sparky/usbridge-signature-pcb.html
To lower the noise of the USB, we had to lower noise on EACH rail. In total, USBridge Sig uses more than 30 LDOs. Even the SD card power rail has a dedicated LDO. There are zero switching elements on board.
仍然有效的優惠:
https://twitter.com/dietpi_/status/821806723484778499
成績有目共睹:
https://www.diyaudio.com/forums/vend...ml#post5987404
When designing the unit we looked at every possible noise incoming , outgoing and generated on board. We filtered , decoupled , and filtered again while having the 10 layers PCB for best EMI.
USbridge Sig was meant to be the best streamer on the market. As a bonus , we did not overcharge the price .
https://community.roonlabs.com/t/rop...ature/79652/27
You‘ll enjoy the USBridge Sig! I‘ve got mine, along with the Shanti LPS. The combo beats my Aurender N100H hands down (this unit costs 3,000 €!)! In my system, Roon sounds very sweet with this unit!
https://community.roonlabs.com/t/rop...ature/79652/48
I can confirm Paul_Jeno’s comment. I also have the USBridge Sig + LPS with a Gustard U16 DDC. It now replaces a $4k streamer… so much so I have put these with more parts and built a streamer around it inside a slick ali case with screen as a single unit. Looking at a OXCO clock and this will play with the big brands.
有興趣的話請看看其他報告:
https://www.tirnahifi.org/forum/view...p=89933#p89933
https://audiophilestyle.com/ca/revie...g-review-r858/
https://audiophilestyle.com/forums/t...ilocks-moment/
https://audiophilestyle.com/forums/t...omment-1071704
這是 Antipodes EX/P1/P2 的售價:
https://www.eliteaudiouk.com/product-page/antipodes-ex
https://www.eliteaudiouk.com/product-page/antipodes-p1
https://www.eliteaudiouk.com/product-page/antipodes-p2
沒有 8820Hz 的普通 piCorePlayer 已經為 CM3 帶來很棒的聲音,明年推出的 CM4 加上 piCorePlayer 8820Hz 就如虎添翼了:
https://audiophilestyle.com/forums/t...omment-1079295
I am running now PiCoreplayer 6.1.0 on the USBrudge Signature and works great. Sound quality fantastic but for now I still use Roon as the server.
-
The Following 6 Users Say Thank You to seeteeyou For This Useful Post:
-
2020-09-16, 03:16 PM
#6024
http://www.myav.com.tw/bbs/showthrea...pagenumber=319
簡單的說, BFS的最大好處, 在於它提供了最小的latency, 也就是允許kernel frequency設定到10000Hz.
兩個非常有意思的適合桌面使用的Linux task排程器: BFS和MuqSS
https://itw01.com/82N7XET.html
適用於桌面應用的Linux調度器BFS/MuqSS
https://www.twblogs.net/a/5db31c12bd9eee310ee66a0d
linux-4.19-ck1, MuQSS version 0.180 for linux-4.19
https://ck-hack.blogspot.com/2018/11...-0180-for.html
MuQSS Scheduler Updated, Linux 4.19-ck1 Drops BFQ I/O Scheduler
https://www.phoronix.com/scan.php?pa...4.19-ck1-MuQSS
MuQSS - The Multiple Queue Skiplist Scheduler by Con Kolivas.
http://ck.kolivas.org/patches/muqss/sched-MuQSS.txt
http://users.on.net/~ckolivas/kernel/
Note that BFS has been deprecated and renamed to MuQSS.
其實 CK 與 MuQSS 兩者之間的分別跟 piCorePlayer 8820Hz 有任何直接/間接關係嗎?最後更新的 kernel 版本是 5.7,而 4.19 就在這裡:
https://github.com/ckolivas/linux/tree/4.19-ck
https://github.com/ckolivas/linux/tree/4.19-muqss
另外還有一種方法可以減低雙機之間的 latency,問題是樹莓派 4 無法安裝 Mellanox 或 Chelsio 那種 RoCE/InfiniBand/iWARP 特殊網卡:
https://www.nas.nasa.gov/assets/pdf/...AS-2014-01.pdf
One of the desirable features associated with InfiniBand, another network fabric technology, is its Remote Direct Memory Access (RDMA) capability. RDMA allows for communication between systems but can bypass the overhead associated with the operating system kernel, so applications have reduced latency and much lower CPU utilization. This results in much faster network performance rates than traditional TCP/IP.
https://www.hwchiu.com/rdma-introduction-i.html
RDMA 到底能帶來什麼樣的優勢?
- low-latency:
- high-throughput:
- low-CPU usage: 透過 RDMA 減少CPU對於網路功能的消耗,讓 CPU 能夠處理更多其他的事情。


既然傳統的 TCP/IP 比較不濟,學日本的高手那樣乾脆跳過 kernel 的 TCP/IP stack 就行了:
https://www.symphonic-mpd.com/forum/...?こちら/300
一つは「カーネルのTCP/IPスタックをバイパス」すること。
連 IP header 和 UDP header 兩者也一同省卻掉:
https://www.symphonic-mpd.com/forum/...?こちら/302
直結に限定すれば、IPヘッダ・UDPヘッダを省くことも技術的には可能です。
確實是用一種比較獨特的方法來演繹『斷捨離』哩,哈哈。
近來還發現了外國的高手如何大大減低雙機之間的 latency,就是用雙 CPU 的單機了,結果是非常正面的說,因此估計 JPLAY 也能夠達到 HQPlayer 那樣的效果:
https://audiophilestyle.com/forums/t...omment-1075198
Actually when I installed my HQplayer on to extreme there are warnings that some of the processes are not carried out so I approached Emile and see if is that a problem. And he kindly helped to assure me that it is OK and he further offered to help me install a NAA in the same extreme computer making use of the two CPUs available. And I certainly would like to give it a try. So next up is the extreme HQ+NAA. And that's a WOW. The performance just leaps to another level. The presence , the focus as well as the soundstage all goes to another level. the string quartet now just sits right in front of you.
https://audiophilestyle.com/forums/t...omment-1079668
The last experiment I've done, inspired by @austinpop and @romaz was to install HQplayer with affinities on one CPU and NAA with affinities to the other CPU. It's good to have two processors! I also created a loopback interface with ipv6. And I used ipv6 via the loopback interface to stream from the HQP to NAA. No upsampling, just using HQP as a bit perfect player. This brought the server that was sounding so great already to yet another level. Everything sounds so effortless now.
一般的交換器與 PCIe 網卡是 1Gb/s,大概是 PCIe 1.0 上限的一半,此外還會受到線材方面的制肘:

可是兩顆 Xeon CPUs 之間的 UPI 就完全不需要任何線材,飛快的速度也不容忽視:

超微 X8DTL-IF/3F 主機板現在的售價是 ¥300 以下,40W TDP 的 Xeon L5630 就只要 ¥6 一顆,屆時 Daitron RFS50A 正式推出的時候它也可以滿足 L5630 的需要,亦肯定不用擔心性能過剩的問題了:
https://ark.intel.com/content/www/tw...intel-qpi.html
Xeon Scalable 的價格還是高一截,僅僅用來測試的話就只需要買一些『出古文物』CPU。
-
The Following 8 Users Say Thank You to seeteeyou For This Useful Post:
-
2020-09-17, 10:24 AM
#6025
 作者: tetsuka
像日本音響BLOG主他在購買線上數位音樂檔回來之後,會把檔案去掉TAG轉為WAV後丟到Google drive後再下載回來。
根據他的說法購買回來的線上音樂檔聲音都不夠好,必須使用上面的步驟經過Google drive過水後的檔案聲音才會比較好。
所以他之前在測試小弟的DAC時發現原始數位音樂檔跟過水後的檔案,在聲音上幾乎沒有差別時才會很驚訝。
既然 piCorePlayer 8820Hz 的力量如此強大,不知它是否能夠同時扮演 RIP(應該是『實時播放』才對)與及過水兩者的角色呢?
這次又要拜託 sam 兄的幫忙了:
piCorePlayer and ext USB CD drive
https://forums.slimdevices.com/showt...l=1#post924898
Install pcp-cdda2wav.tcz from the extension page. Just make sure to select the piCorePlayer Repository.
ARM64 版本的 cdrom-5.4.51-pcpCore-v8.tcz 和 pcp-cdda2wav.tcz 擴充套件如下:
http://picoreplayer.sourceforge.net/...pcpCore-v8.tcz
http://picoreplayer.sourceforge.net/...p-cdda2wav.tcz
關於擷取 ALSA 輸出的 arecord(alsa-utils)和 snd-aloop(alsa-modules):
https://superuser.com/a/597230
I'd suggest using the ALSA loopback driver.
已確認 alsa-modules-5.4.51-pcpCore-v8.tcz 內包含 /usr/local/lib/modules/5.4.51-pcpCore-v8/kernel/sound/drivers/snd-aloop.ko 這檔案:
http://picoreplayer.sourceforge.net/...pcpCore-v8.tcz
http://picoreplayer.sourceforge.net/...alsa-utils.tcz
http://picoreplayer.sourceforge.net/.../libasound.tcz
http://picoreplayer.sourceforge.net/...cz/ncurses.tcz
理論上 loopback driver 該是可行的,希望測試的結果也沒有問題吧,謝謝。
-
The Following 5 Users Say Thank You to seeteeyou For This Useful Post:
-
2020-09-17, 11:27 AM
#6026
高級交換式電源,照樣會相互污染!
此篇文章於 2020-09-17 11:31 AM 被 psycho 編輯。
-
The Following 7 Users Say Thank You to psycho For This Useful Post:
-
2020-09-17, 11:54 AM
#6027
 作者: psycho
[恕刪]
以下實驗為了避免地迴路干擾,所有電供通通沒有接地線!!!
[恕刪]
教授,不用傷腦筋,因為原因就在您文章內上面這句話。
高階交換式電源的內建濾波,一定需要接地才能發揮預期效果。
AC濾波器也是一樣的,一定要接地才會有跟規格書內的相符的效能。
-
The Following 9 Users Say Thank You to bchsieh For This Useful Post:
-
2020-09-17, 12:06 PM
#6028
 作者: bchsieh
教授,不用傷腦筋,因為原因就在您文章內上面這句話。
高階交換式電源的內建濾波,一定需要接地才能發揮預期效果。
AC濾波器也是一樣的,一定要接地才會有跟規格書內的相符的效能。
真的傷腦筋,我當然測試過有接地線跟沒有接地線的情況,結果相差不多。因為怕被你批評有地迴路,所以只寫出拔掉地線的部分並且加重語氣,沒想到應該是接地線才對?我這個文組真是會昏倒....
沒關係,反正兩種情況我都有測試,拔掉地線的效果好像好一點?但是最重要的電源隔離效果完全沒有改變。請大魔王再看看還有什麼問題??
-
The Following 6 Users Say Thank You to psycho For This Useful Post:
-
2020-09-17, 01:25 PM
#6029
 作者: psycho
真的傷腦筋,我當然測試過有接地線跟沒有接地線的情況,結果相差不多。因為怕被你批評有地迴路,所以只寫出拔掉地線的部分並且加重語氣,沒想到應該是接地線才對?我這個文組真是會昏倒....
沒關係,反正兩種情況我都有測試,拔掉地線的效果好像好一點?但是最重要的電源隔離效果完全沒有改變。請大魔王再看看還有什麼問題??
教授您好,
小弟所謂的「接地」,是接真正的大地,而不是接到接地盒。
除了接地以外,您的實驗並無法排除兩個不同的交換式電供對DAC和擴大機線性電源的影響。
線拉得越長,交互干擾一定是越小。
但是如果很長的線用的等級不夠好不夠粗,交換式電源的聲音就會變差。
而這條線也可能變成瓶頸,讓兩顆不同等級的交換式電源的差異變小。
-
The Following 11 Users Say Thank You to bchsieh For This Useful Post:
andyhang,babymlin,chrischang0624,Deen,gangster.tank,kevintran,pawaslider,psycho,seafood,tetsuka,yuntree
-
2020-09-17, 03:25 PM
#6030
 作者: seeteeyou
seeteeyou兄您好,
您列出的rip所需的檔案,直接用 piCorePlay內建的擴充套件就可以直接安裝了。
但都是用下指令的方式去 rip CD,這樣意義也不大了。
-
The Following 7 Users Say Thank You to sam0402 For This Useful Post:
發文規則
- 您不可以發表新主題
- 您不可以發表回覆
- 您不可以上傳附件
- 您不可以編輯自己的文章
-
討論區規則
|