Building a DIY Music Server

[di-fi]

6 hours ago, Marcin_gps said:

the impact of network can be heard simply, because the network connection is active whether we send data or not.

***

In my personal system I've now eliminated wifi completely. I'm running the JPLAY app on M2 Mac Mini which is connected to XACT S1 (switch) and the XACT S1 (switch) is conected to XACT S1 (server/player). 
 

 

 

Finding wifi being worse than Ethernet cables, I realize your listening ritual must be a desktop experience to get the best out of the XACT S1.
That seems a significant compromise for whom prefer listening not from behind a desk.

Would a wired iPad (Pro M2?) with « wifi not connected »  have the potential to be comparable in SQ to the M2 Mac Mini to allow for a more flexible choice of listening position? 

[MarcelNL]

58 minutes ago, baconbrain said:

Perhaps I am missing something around Taiko’s Tacdda solution, but isn’t a DAC only as good as it’s analog output stage? If so, how will Taiko fit such on a PCiE card?

Tacda is not just an plug in DAC card, that is one option...it is designed as a modular direct from CPU output card.

I personally doubt the DAC module is an end game design, but we may be surprised

[MarcelNL]

4 minutes ago, di-fi said:

 

Wifi being worse than Ethernet cables, I realize your listening ritual must be a desktop experience to get the best out of the XACT S1.
That seems a significant compromise for whom prefer listening not from behind a desk.

Would a wired iPad with wifi not connected be comparable in SQ to the M2 Mac Mini to allow for a more flexible choice of listening position? 

When I shut down thw web page interface to Daphile on my laptop of Ipad sound quality improves after like 20 sec or so.

Wired connection sounds like a great option...or a good old fashioned IR extender ;-)

[littlej0e]

6 hours ago, baconbrain said:

Perhaps I am missing something around Taiko’s Tacdda solution, but isn’t a DAC only as good as it’s analog output stage? If so, how will Taiko fit such on a PCiE card?

I think this is part of the excitement with the TACD system. Taiko's built-in DAC supposedly competes with some very expensive DACs:

 

"We're not DAC designers, the innovation is in the superior interface, we've incorporated a Rohm BD34301EKV DAC chip because we thought it sounded best of the available "off the shelf" DAC chips (more "analog" than for instance ESS Sabre) and added a JFET class A output stage, but that's about all we could fit on a PCIe card sized PCB. It may compete with some 10-20k DAC's, but to directly compete with top-level DAC such as Lampinzator Horizon, MSB Select 2, or Totaldac D1-12 is a bit much to ask for!"

 

It also appears they are going to squeeze as many outputs as possible onto separate, swappable modules using the same base PCIe card:

• Analog RCA
• S/PDIF in any variant
• I2S
• AES/EBU
• Other standard Digital formats
• Custom output like MSB Pro ISL
• New Taiko proprietary interface
• Potential other formats

 

Meaning you'll be stuck using single-ended RCAs or one of the other outputs listed above to connect to your amp. You would think they will figure out the physical space and heat with balanced XLR connections eventually (a modified double slot height PCIe card or make the physical connections outside of the server chassis?). 

 

Personally, I am interested to see how they plan to deliver sampling rates above 192kHz on the non-DAC transport card without using USB (on the external DAC side) or without significant modification of the external DACs themselves. Perhaps they have no intention of solving this and intend for people to use lower sample rates or the internal PCIe DAC until manufacturers adopt the Taiko interface/protocol standard? For this reason alone you would think a USB module, even if it's just a glorified signal converter, would have to come to the TACD system eventually.

[Marcin_gps]

9 hours ago, di-fi said:

 

Finding wifi being worse than Ethernet cables, I realize your listening ritual must be a desktop experience to get the best out of the XACT S1.
That seems a significant compromise for whom prefer listening not from behind a desk.

Would a wired iPad (Pro M2?) with « wifi not connected »  have the potential to be comparable in SQ to the M2 Mac Mini to allow for a more flexible choice of listening position? 

 

My listening ritual is on a sofa with an iPad in my hands. To be able to 'see' XACT S1 in the home network again I have to connect one of the ethernet cables back. It's just for more critical listening I isolate my 'audio' network from my main home network by disconnecting the cable going to router from XACT S1 switch (this is possible once XACT S1 server/player and Mac get their IPs assigned by DHCP server running on the main router).

 

The first LAN cable on the right connects to router. It's not a big inconvenience.

 

[flkin]

13 hours ago, MarcelNL said:

We'll simply need a shootout...which is difficult since it's comparing apples and oranges real quick.

 

Personally my bet is not on the PF using supercaps, anything I heard so far using those had a grey veil over it. If Taiko releases the TACDDA with I2S and that is able to do I2S over RJ45/utp/stp my bet would be placed there.

 

 

I’m trying to get one of the Pink Faun Ultra USB Card early. They had 2 prototypes completed at the time I heard it and another few more being built. So it’s not just an announcement but an actual device about to be launched. I guess in a few months time?
 

I plan to compare this card against my older v1 with Ultra clock and whatever I can get my hands on from friends. Given it’s probably going to be significantly more expensive it going to have to be a lot better than their previous v1 to be compelling, not just a little better. We’ll see soon. 
 

Many ingredients discussed but the proof of the pudding ..

 

Oh and PF did mentioned they were going to build an I2S card with the same tech too. Personally I think I’d prefer I2S  but my DAC isn’t capable so that will be for another time

[MarcelNL]

Interesting....beware....at this level the differences may be small but typically so important there usually is no going back!

 

 

[Mr Morris]

Quick thanks to @MarcelNL for his regular reminders to turn of SSD in the Bios if not needed. Also @StreamFidelity for giving me the idea to use to an Optane M10. Turning off the SSD has made a very welcome contribution to my Sean Jacobs LPSU based server, which this thread, thanks @Nenon inspired me to build. The difference is a more realistic,  natural sound. 

 

Everything sounding clearer, with more ambient and sound stage information. Very pleased I am now getting what I always felt was missing from digital.

 

I can tell that more is achievable. I e. Bass clarity. But delighted with this new level. 😀 

[di-fi]

On 8/19/2023 at 1:57 AM, Marcin_gps said:

 

My listening ritual is on a sofa with an iPad in my hands. To be able to 'see' XACT S1 in the home network again I have to connect one of the ethernet cables back. It's just for more critical listening I isolate my 'audio' network from my main home network by disconnecting the cable going to router from XACT S1 switch (this is possible once XACT S1 server/player and Mac get their IPs assigned by DHCP server running on the main router).

 

I am just curious because in your post you seem to suggest in your personal setup by disabling wifi and adding a XACT S1 (switch) you virtually eliminated all network noise resulting in better SQ even for files played directly on the server.
Control point - switch - server

 

Then you added disconnecting from the router besides not using wifi while only using a M2 Mac Mini to control the XACT S1 server over Ethernet, I get that. In that scenario I tried to suggest you could have the same control point with your iPad (wifi disconnected) over Ethernet to the switch with the advantage of controlling from your listening position instead.

In both scenarios no router and no wifi involved to virtually eliminate all network noise or is there a difference?

[Marcin_gps]

On 8/21/2023 at 4:35 AM, di-fi said:

I am just curious because in your post you seem to suggest in your personal setup by disabling wifi and adding a XACT S1 (switch) you virtually eliminated all network noise resulting in better SQ even for files played directly on the server.
Control point - switch - server

 

Then you added disconnecting from the router besides not using wifi while only using a M2 Mac Mini to control the XACT S1 server over Ethernet, I get that. In that scenario I tried to suggest you could have the same control point with your iPad (wifi disconnected) over Ethernet to the switch with the advantage of controlling from your listening position instead.

In both scenarios no router and no wifi involved to virtually eliminate all network noise or is there a difference?

with iPad I still need the connection to main router (wifi). 

[di-fi]

On 8/24/2023 at 6:36 AM, Marcin_gps said:

with iPad I still need the connection to main router (wifi). 

Except when utilizing a Lightning / USB C to Ethernet Adapter. Connecting the iPad directly to the XACT S1 switch via an Ethernet cable eliminates the need for a router in the setup (unless JPLAY does not allow for this).

Sorry to persist in my assertion, as in your arrangement, aiming to minimize network interference similar to the Mac Mini, assuming that the iPad's network port does not introduce more noise, all the while utilizing JPLAY.

The iPad's more user-friendly interface could prove advantageous for any (diy) server, especially considering the claimed high sound quality it offers for critical listening. At that point I would gladly accept one more cable in my setup.

[matthias]

On 8/25/2023 at 1:55 PM, di-fi said:

Except when utilizing a Lightning / USB C to Ethernet Adapter. Connecting the iPad directly to the XACT S1 switch via an Ethernet cable eliminates the need for a router in the setup (unless JPLAY does not allow for this).
Sorry to persist in my assertion, as in your arrangement, aiming to minimize network interference similar to the Mac Mini, assuming that the iPad's network port does not introduce more noise, all the while utilizing JPLAY.
The iPad's more user-friendly interface could prove advantageous for any (diy) server, especially considering the claimed high sound quality it offers for critical listening. At that point I would gladly accept one more cable in my setup.

Just for reference:

Emile from Taiko Audio prefers a wifi solution for his upcoming Taiko router. He would never do this if wifi is inferior from a sound quality POV.

 

Matt

[MarcelNL]

16 hours ago, matthias said:

Just for reference:

Emile from Taiko Audio prefers a wifi solution for his upcoming Taiko router. He would never do this if wifi is inferior from a sound quality POV.

 

Matt

if we only could tap into the knowledge behind that ;-)

 

 

[seeteeyou]

Personally I wouldn't even bother with any streaming services at all and therefore the whole point of having a router would be moot, IMHO they're considered nothing more than "second fiddle" when compared to what one could achieve with PGGB 256.

 

OTOH, how about turning Asus WS C621E SAGE into some kinda "network router on steroids" after installing a copy of Win11PE that's under 100MB?

 

http://jplay.eu/forum/index.php?/topic/4410-windows-11-pe-audiophile-creation-guide/?p=63862

Quote

I have removed "iertutil.dll" and "imageres.dll" , contained in \windows\system32, which has allowed me to get a boot.wim size of 99.7 MB, with my ASIO drivers and F2k and BugHead installed. Without ASIO drivers and without audio apps the size is 84.4 MB !!

 

Software should be available for free as shown below

 

https://www.nat32.com

https://github.com/crpietschmann/VirtualRouter

https://sourceforge.net/software/virtual-routers/windows/

 

Actually both the router and the music server could take advantage of this

 

https://apps.microsoft.com/store/detail/killer-intelligence-center/9PCWJKS4JSN1

https://www.intel.com/content/www/us/en/download/19779/intel-killer-performance-suite.html

 

 

Heck, even getting the "Killer" flavor of drivers installed (while we're installing the "regular" flavor of Intel hardware that it) would actually reduce the latency significantly

 

https://post.smzdm.com/p/aekez9lk/

https://tieba.baidu.com/p/6377940051

https://bbs.luobotou.org/thread-45955-1-1.html

https://blog.yazawaniko.com/index.php/archives/229/

https://www.right.com.cn/forum/thread-2336261-1-1.html

https://blog.csdn.net/Nature2022/article/details/123196021

https://blog.csdn.net/u014390186/article/details/103423636

https://forum.gamer.com.tw/C.php?bsn=60030&snA=547688

Quote

目前發現刷驅動後打遊戲延遲降低了，之前3168 延遲在45ms 現在28ms 現在有線 26ms，ping值更穩定 如果不打遊戲，感覺不是很有必要刷

 

Yet another one would require specific motherboards, though

 

https://apps.microsoft.com/store/detail/intel®-connectivity-performance-suite/9NSG1HWGCKVM

https://www.intel.com/content/dam/support/us/en/documents/wireless/intel-connectivity-performance-suite-2.0-user-guide.pdf
https://www.intel.com/content/www/us/en/download/738623/intel-connectivity-performance-suite-for-intel-wireless-products.html

Quote

This software only works on specific Intel® Evo™ or Intel vPro® platforms with Intel® Wi-Fi 6/6E (or greater) products.

[matthias]

2 hours ago, MarcelNL said:

if we only could tap into the knowledge behind that ;-)

Indeed,

my point is that it doesn't make any sense to say "wired sounds better than wifi".

 

Matt

[matthias]

On 8/18/2023 at 10:25 PM, MarcelNL said:

I personally doubt the DAC module is an end game design, but we may be surprised

Agree,

I might be sound better than a Roon Nucleus/MSB Select II combo anyway.

 

Matt

[di-fi]

On 8/30/2023 at 4:04 PM, matthias said:

Just for reference:

Emile from Taiko Audio prefers a wifi solution for his upcoming Taiko router. He would never do this if wifi is inferior from a sound quality POV.

 

Matt

 

But  @Marcin_gps seems to suggest here that in his personal setup by disabling Wi-Fi and adding a XACT S1 (switch) plus using JPLAY he virtually eliminated all network noise resulting in better SQ even for files played directly on the server.
Control point - switch - server

 

Contrarily to Taiko, in his setup @Marcin_gpshas to disconnect from the router (besides not using wifi) to obtain the best SQ. Note we are talking files from a NAS or local SSD. Emile does (still) not claim the best SQ in his setup when files are streamed.

 

For playing local files I tried introduce the idea to explore an iPad (no Wi-Fi) connection in his setup to resolve the UI being somewhere on a desk in the room.

 

 

On 8/31/2023 at 8:44 AM, MarcelNL said:

if we only could tap into the knowledge behind that ;-)

yep! (2x)

[MarcelNL]

1 minute ago, di-fi said:

 

But  @Marcin_gps seems to suggest in his personal setup by disabling wifi and adding a XACT S1 (switch) he  virtually eliminated all network noise resulting in better SQ even for files played directly on the server.
Control point - switch - server

 

Contrarily to Taiko in his setup @Marcin_gpshas to disconnect from the router (besides not using wifi) to obtain the best SQ.

 

I tried introduce the idea to explore an iPad connection in his setup to resolve the UI being somewhere on a desk in the room.

 

 

yep!

It'll be an interesting journey discovering what makes a network for audio tick....

 

I usually switch off the user interface for Daphile for serious listening, after 20 sec or so SQ definitely gets better...not by much but still.

 

It would be great to have a forward backward control, hardwired to the server, for critical listening.....

[matthias]

1 hour ago, di-fi said:

 

But  @Marcin_gps seems to suggest here that in his personal setup by disabling Wi-Fi and adding a XACT S1 (switch) plus using JPLAY he virtually eliminated all network noise resulting in better SQ even for files played directly on the server.
Control point - switch - server

 

Contrarily to Taiko, in his setup @Marcin_gpshas to disconnect from the router (besides not using wifi) to obtain the best SQ. Note we are talking files from a NAS or local SSD. Emile does (still) not claim the best SQ in his setup when files are streamed.

 

For playing local files I tried introduce the idea to explore an iPad (no Wi-Fi) connection in his setup to resolve the UI being somewhere on a desk in the room.

 

The approaches are different.

Taiko Audio evaluated both methods and came up with a wifi solution which outperformed wired.

So it is not about the principle but about the implementation.

One interesting post is this one:

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-943#post-897728

You get some nice insights when you do a search on WBF with "wifi" and "Taiko Audio".

BTW, I am sure the JCAT approach with the 12k switch does sound very good too.

 

Matt

[MarcelNL]

ultimately the implementation is what matters..whoever gets that right  rules.

[di-fi]

48 minutes ago, matthias said:

The approaches are different.

Taiko Audio evaluated both methods and came up with a wifi solution which outperformed wired.

So it is not about the principle but about the implementation.

One interesting post is this one:

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-943#post-897728

You get some nice insights when you do a search on WBF with "wifi" and "Taiko Audio".

BTW, I am sure the JCAT approach with the 12k switch does sound very good too.

 

Matt

 

Thanks. I'm aware of the information in that thread, but I missed that part about it. From what I gathered in the discussion, Taiko utilized an extremely low-power 2.4 GHz Wi-Fi setup (we are only talking control UI here) , which yielded excellent results with the MikroTik mAP Lite acces point, or something similar with better integration into the Extreme router.

 

Besides the choice of the OS, the quality of the Ethernet setup and upstream network implementation plays a crucial role in the sound quality of DIY servers. This aspect is often underrepresented in SQ evaluations here, alongside the differentiation between local file storage and streaming services.

[matthias]

10 minutes ago, di-fi said:

It's worth noting that, besides the choice of the operating system (OS), the quality of the Ethernet setup and upstream network implementation plays a crucial role in the sound quality of DIY servers.

Agree,

btw it is also possible and convenient to switch wifi off (for audio transfer) when files have been loaded into RAM. This is super easy for example with a Macbook for streaming from Qobuz and Spotify.

 

Matt

[seeteeyou]

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/438/#comment-897055  

On 11/19/2018 at 8:30 AM, austinpop said:

Some more findings on Zenith SE running AL Extreme

1. Removing the USB stick after boot (the withdrawal method) does indeed sound a little bit better. Thanks, @lmitche for helping me with my premature optimization.

 

Stuff like that might work well for local storage devices, though it turned out to be no good at all when an audiophile in Asia tried that with Ethernet (i.e. wired network) adapter for iOS devices. The Ethernet cable was indeed unplugged from the RJ-45 port but he concluded that SQ was still suffering from the largest source of noise, namely the 3D graphics as follows

 

https://support.apple.com/kb/SP830

Quote

4-core GPU

 

https://support.apple.com/kb/SP869

Quote

8-core GPU

 

He also gave RPi a shot but even low-latency Linux kernel (i.e. Xenomai 4 with EVL) couldn't save the day

 

https://github.com/sam0402/pcp-44.1KHz#pcp-441khz-

Quote

piCorePlayer 8 with 264KHz / 176KHz realtime(EVL) kernel

 

https://forums.raspberrypi.com/viewtopic.php?t=244519

https://www.raspberrypi.com/documentation/computers/processors.html

Quote

Videocore VI GPU (on BCM2711 SoC)

 

https://bakhi.github.io/mobileGPU/v3d/

Quote

Hence, videocore VI depends on the pipeline driver found in vc4 driver. v3d is the name for the 3d engine.

 

Basically he found that most likely the damage was already done in the first place, even if the Ethernet cable were disconnected at some point. The noise (thanks to the "blessings" of 3D graphics) wouldn't be gone unless he's powering the music server off and on.

 

Instead of messing with iOS devices, any Intel Alder Lake motherboards with something like i3-12100F etc. would be fine as long as we're underclocking the CPU. 800 MHz should be OK and then we'll go for an Intel Optane with the lowest power consumption ever

 

https://www.ebay.com/itm/115728184536

https://www.ebay.com/itm/125829350920

https://www.ebay.com/itm/134619691333

https://www.aliexpress.us/item/3256804772234177.html

https://www.aliexpress.us/item/3256805499373767.html

https://www.amazon.com/Intel-Optane-Memory-PCIe-80mm/dp/B06XSMTN31

https://ark.intel.com/content/www/us/en/ark/products/series/132776/intel-optane-memory-m10-series.html

Quote

2W

 

Of course the crucial component really is all about getting one of those 2D-only VGA cards, granted either one would cost quite a bit

 

https://www.ebay.com/itm/284959070279

https://www.ebay.com/itm/304413120787

 

Though we've got yet another option that's only 10 bucks or so

 

https://www.taobao.com/list/product/ast2400.htm

https://www.taobao.com/list/item/719585058819.htm

Quote

理卡顯卡，256M超低功耗3W超級省電！

 

That's pretty much like what we've got here

 

https://servers.asus.com/products/Servers/Server-Accessories/ASMB9-iKVM

https://www.amazon.com/ASUS-ASMB9-IKVM-C621E-SAGE-Workstation/dp/B07M71HSMT

https://www.servethehome.com/asus-ws-c621e-sage-motherboard-search-incredible-over/asus-ws-c621e-sage-asmb9-ikvm/

https://forums.servethehome.com/index.php?threads/is-asmb8-9-compatible-with-motherboards-that-supports-asmb7.34813/#post-320232

Quote

ASMB7-iKVM = AST2300
ASMB8-iKVM = AST2400
ASMB9-iKVM = AST2500
ASMB10-iKVM = AST2600

 

While @Nenon also mentioned that module before

 

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-206#post-655430

 

Though in this particular case, we're interested in nothing more than the VGA output of ASPEED AST2400 

 

https://www.htesemi.com/product/aspeed-ast2400

https://www.datasheetcafe.com/ast2400-datasheet-server-management-processor/

Quote

2D Video Graphic Adapter with PCIe bus interface

 

YMMV as usual, though personally I'm trying my best to stay away from "spying" operating systems such as Android / iOS / macOS etc.

 

---

  

9 hours ago, matthias said:

Agree,

btw it is also possible and convenient to switch wifi off (for audio transfer) when files have been loaded into RAM. This is super easy for example with a Macbook for streaming from Qobuz and Spotify.

 

FYI - specific implementations might require the corresponding option(s) as shown below

 

https://pinkfishmedia.net/forum/threads/long-tracks-jumping-on-qobuz-with-lms.270754/#post-4764780

Quote

By default LMS/Qobuz ("Normal streaming" mode) downloads a small burst of data and then continuously pulls down what I presume is just enough to keep a local small in-memory buffer from emptying. Towards the end of a track this trickle-fill stops for a short while, while the track is still playing. Then there's another small data burst (again while the track is still playing). This is presumably the new track. Then a continuous trickle starts. Then the new track starts playing.

In "Persistent mode" the traffic looks the same as in "Normal streaming". I think the difference is that if the connection breaks it gets re-tried whereas I think normally it is dropped. This mode has worked for me so far.

In "Cache HTTP(S) streams on disk" mode, LMS/Qobuz behaves just like the official Qobuz apps from a network traffic PoV and downloads the complete track in one burst.

 

Obviously stuff like Euphony would make our lives much easier, though I wouldn't even bother with anything that really needs to phone home simply because of Murphy's Law

 

http://jplay.eu/forum/index.php?/topic/4656-jplay-and-murphys-law/

Quote

• Muttering under my breath, put the wine down, ease myself out of the chair and head over to the Control/Audio PC.
• Swear profusely, as I discover on my Audio PC in red text "Jplay licence has expired......" &^#&#$ Murphy's law, why now, why tonight!!!!!
• Knowing what I have to do in putting the Audio PC onto the internet to reactivate (no small matter, have to reset Audiophile 3, add external USB/Ethernet hub, find ethernet cable etc.... I swear profusely again, drink the rest of my wine, switch everything off saying 'I'll get to you in the morning',
• Switch off the lights, head to bed and load up an audiobook to try and de-stress again......

 

@littlej0e also got royally screwed in a similar way

 

https://audiophilestyle.com/forums/topic/58164-building-a-diy-music-server/page/215/#comment-1243827

Quote

For example, let's say you happen to live high up in the mountains and lost power to your rig while you were away on a multi-week business trip. When you finally returned, you poured a glass of whiskey and sat down in your favorite listening chair, ready to relax and listen to some music. You press play, and after 20 seconds playback cuts off because your installation reverted back to the trial version while you were away, rendering your server completely useless - M0th#$% F#*@ING $#&T B@#!s A#@ P&$^!!!

 

IMHO one could easily understand why Taiko Audio might never consider THAT kinda product activation since we're only humans, we mere mortals aren't exactly comparable to deity by any means. In other words, we just won't get things right 100% of the time while some companies (but not necessarily all of them) do care about customer satisfaction more than anything else.

[matthias]

2 hours ago, seeteeyou said:

FYI - specific implementations might require the corresponding option(s) as shown below

For Qobuz switching off wifi works with MBP (4GB RAM) and Audirvana (no upsampling) for a few tracks, the same with Spotify desktop app.

 

Matt

[seeteeyou]

https://www.mouser.com/ProductDetail/607-BE200.NGWG

Quote

Multiprotocol Modules Intel Wi-Fi 7 BE200, 2230, 2x2 BE+BT, vPro

 

https://www.mouser.com/ProductDetail/607-BE202.NGWG.NV

Quote

Multiprotocol Modules Intel Wi-Fi 7 BE202 (Gig+), 2230, 2x2 BE+BT, No vPro

 

https://www.intel.com/content/dam/www/public/us/en/documents/pdf/wi-fi-7-and-beyond.pdf#page=5

Quote

▪ Lower latency due to access to multiple links in parallel

 

ROG MAXIMUS Z790 DARK HERO

https://rog.asus.com/motherboards/rog-maximus/rog-maximus-z790-dark-hero/spec/

Quote

2x2 Wi-Fi 7 (802.11 a/b/g/n/ac/ax/be)

 

---

 

https://d86o2zu8ugzlg.cloudfront.net/mediatek-craft/documents/Wi-Fi-7-MLO-White-Paper-WF7MLOWP0622.pdf#page=10

Quote

Figure 12 demonstrates latency benefits of MLO when subjected to different network loading. Measurement is done with 95% CDF. The result shows the latency of a Wi-Fi 7 MLO STA can be improved from 145ms to 18ms (more than 85% improvement) when compared to a Wi-Fi 6 Single STA when network loading is 70%. It also shows that STR has the lowest latency because the two links do not interfere with each other. That also explains why NSTR or EMLSR have higher latencies as its transmission may be blocked by IDC interference or running out of spatial streams.

 

https://arxiv.org/pdf/2111.09281.pdf

https://ieeexplore.ieee.org/abstract/document/9838765

Quote

In this paper, we provided an experimental study of latency for IEEE 802.11be MLO. Using the WACA dataset, which contains real-world channel occupancy measurements in the 5 GHz spectrum, we cast light upon the latency performance of two MLO channel access modes, namely (i) MLO-STR, where two radio interfaces are operated independently, and (ii) MLO-NSTR, where one interface acts as primary and the other as secondary. We showed that when both channels are on average equally occupied, both MLO modes can reduce the 95th percentile latency by nearly one order of magnitude as they avail of a second radio interface. In contrast, in asymmetrically occupied channels, we surprisingly found the use of MLO-STR to be detrimental and cause even higher latency values than SLO. To overcome this issue, we proposed an alternative implementation, denoted MLO-STR+. By running in parallel as many backoff instances as interfaces, MLOSTR+ allocates each packet to the interface whose backoff expires first, thus achieving a significantly lower latency.

 

 

 

 

---

 

Could any of that stuff be applied to wired networking by any chance?