A novel way to massively improve the SQ of computer audio streaming

[austinpop]

1 hour ago, KingRex said:

Austinpop, can you  talk more about this chain. I don't quite understand it. I called sotm and they did not mod a switch. My idea of a switch is something like a D-Link or Netgear. I'm also not sure how your cabling your equipment together. The Cyber shaft only has coax in and out.  I see the connect to the USBultra,  but how  do I connect that to my DAC or server which are USB.  I also don't get how the cybershaft is not in a chain. Its almost like it is a loop in and out of the USBultra.  Can you talk to the cabling. 

 

Thanks. 

 

You are likely missing a lot of context from earlier in this massive thread. That's why I took the time and effort to create a handy index in the first post. Please go back and take a look. After @romaz's initial post, I have a summary and an index of links. If you browse through those, you'll get a sense of what the sCLK-EX rationale is.

[austinpop]

To clarify further, here is my latest topology, compared to my older one. My trifecta chain looked like this:

 

 

My new chain with the Zenith SE now looks like this:

 

 

 

Hope this clarifies things.

[KMan]

@austinpop I have been following your informative posts and have a SOTM trifecta and waiting for my SR7. Sorry if you have answered this before, what is the clock cable you use with your Ref. Clock? I am looking at getting Mutec Ref clock. I use Habst with Blu2 now, but do not want to buy another pair for expense reasons.

[austinpop]

57 minutes ago, KMan said:

@austinpop I have been following your informative posts and have a SOTM trifecta and waiting for my SR7. Sorry if you have answered this before, what is the clock cable you use with your Ref. Clock? I am looking at getting Mutec Ref clock. I use Habst with Blu2 now, but do not want to buy another pair for expense reasons.

 

That is a good question. It's a very inexpensive $20 0.5m cable that Hasegawa-san at Cybershaft supplied with my OP-14. I've never actually looked at it closely. Note: it is a 50 Ω cable.

 

The markings say Daiyu Densen RG-58 A/U.

 

I have it on my list to try other cables. Based on the incredibly helpful links posted by @seeteeyou, it looks like the choices for 50 Ω are quite limited. I would love to know what to try.

[KMan]

1 hour ago, austinpop said:

That is a good question. It's a very inexpensive $20 0.5m cable that Hasegawa-san at Cybershaft supplied with my OP-14. I've never actually looked at it closely. Note: it is a 50 Ω cable.

Thanks for the response, it is a pity that OP-14 is rare species or it would have been my first choice. Yes a 50 Ω cable limits the options indeed. I am not much of a DIY person and a pre-built cable is the only option for me. Could you  please share the link to the right post by  @seeteeyou I did not remember seeing  50 ohm options.

[seeteeyou]

2 hours ago, austinpop said:

I have it on my list to try other cables. Based on the incredibly helpful links posted by @seeteeyou, it looks like the choices for 50 Ω are quite limited. I would love to know what to try.

 

Thanks @austinpop.

 

I really dunno very much about 50 Ω cables at the moment, maybe here's a little rundown of what I've been learning so far.

 

First of all, it's just easier to break everything down into two categories, obviously there's one for the audiophile market and then another for the pro market.

 

Regarding the audiophile market, most likely we could only find a dozen choices or so. The price range would go anywhere from mere-mortal-level to @romaz-level ($6.5K) as shown below

 

http://rubi2947.blog.jp/archives/7363072.html

http://www.ojispecial.jp/products/8128xbnc.html

https://item.rakuten.co.jp/miyajimusic/ka-r-062017-wa01/

http://briseaudio.jp/store/product/bnc/bnc50_osa01.html

http://briseaudio.jp/store/product/bnc/bnc50_shin01.html

http://briseaudio.jp/store/product/bnc/bnc50_masa01.html

http://briseaudio.jp/store/product/bnc/bnc50_mura01.html

https://www.thecableco.com/venom-s-pdif-5301.html

https://www.thecableco.com/sigma-clock-50.html

http://goods.ruten.com.tw/item/show?21744272134793

https://almaaudio.com/collections/evolution-acoustics/products/evolution-acoustics-bnc-link-cables

 

Unlike 75 Ω coaxial cables. they weren't making 50 Ω ones for 4K UHD resolutions (i.e. 11.88 Gbps or 12 GHz) and therefore we could only pick something that's thick enough to achieve "exceptional" nominal attenuation.

 

Now let's take a quick look at the comparison between 75 Ω ones and 50 Ω ones (pick either link below and it's quicker to load the 1st one)

 

http://www.canare.co.jp/en/24A/Canare24A_Cables.pdf#page=20

http://www.canare.co.jp/cat/cont/canare_eng/canare_eng.pdf#page=69

 

Focus on 10 MHz for now and we'll see the differences. Obviously L-8CHD should be the winner (1.2 dB/100m) due to its "insane" thickness @ 11.1 mm OD while L-5DFB was simply gathering dust @ 2.5 dB/100m just because Canare never made any kinda "nutty" 50 Ω cables.

 

http://www.canare.co.jp/en/24A/Canare24A_Cables.pdf#page=18

http://www.canare.co.jp/cat/cont/canare_eng/canare_eng.pdf#page=67

 

Of course we could still look somewhere else and it's fairly easy to find 50 Ω Fujikura 8D-FB @ 11.1 mm OD

 

http://www.aru-densen.jp/html/page8.html

http://www.fujimusen.com/coaxial/index.html

https://item.rakuten.co.jp/musenkiya/8dfb10m/

https://us.misumi-ec.com/vona2/detail/110500112930/?HissuCode=8D-FB-10

http://www.caledonian-cables.co.uk/Coaxia_Cable/50Ohm RF/8D-FB.html

http://www.caledonian-cables.com/product/Coaxial Cables/50Ohm RF Coaxial Cables/8D-FB.htm

 

Here are some 50 Ω BNC plugs that will be good for 8D-FB

 

To-Conne BNCP-8

https://item.rakuten.co.jp/tiyotoku/bncp-8/

http://www.to-conne.co.jp/connector/c_products/bnc/plugs/

http://www.to-conne.co.jp/connector/c_products/bnc/plugs/BNCP-8.pdf

 

Almic BNC-P-8D

https://item.rakuten.co.jp/auc-treevillage/bnc-p-8d/

http://www.almic.net/product/download/BNC_catalog.pdf

 

Amphenol RF 6775

https://www.amphenolrf.com/000-6775.html

https://octopart.com/000-6775-75-amphenol+rf-49742146

https://media.digikey.com/pdf/Data Sheets/Amphenol PDFs/BNC_Catalog.pdf#page=5

 

If we weren't big fans of Fujikura, there's also Belden 9913 in addition to 7810A as I mentioned below

 

https://www.computeraudiophile.com/forums/topic/31857-mutec-ref-10-masterclock/?page=24&tab=comments#comment-773291

 

Of course we could also go totally crazy with 12D-FB @ 15.6 mm OD but so far I haven't seen any kinda 50 Ω BNC plugs that could accommodate coaxial cables with that kinda thickness

 

http://www.caledonian-cables.co.uk/Coaxia_Cable/50Ohm RF/12D-FB.html

http://www.caledonian-cables.com/product/Coaxial Cables/50Ohm RF Coaxial Cables/12D-FB.htm

 

Finally we've gotta thank @d_elm for pointing out that Pasternack also made some 50 Ω cables with impressive numbers that looked good (at least?) on paper

 

https://www.pasternack.com/semirigid-0.250-50-ohm-coax-cable-tinned-aluminum-pe-sr401al-p.aspx

https://www.pasternack.com/images/ProductPDF/PE-SR401AL.pdf#page=2

 

It really does cost quite a bit @ $21.16 per foot but fortunately clock cables should be kept as short as we could manage to, and then its 6.35 mm OD should be an advantage as well.

 

BTW, they also have something thicker and even more expensive as follows

 

10.92 mm OD @ $31.94 per foot
https://www.pasternack.com/flexible-0.430-rg225-50-ohm-coax-cable-ptfe-fr-jacket-rg225-u-p.aspx

https://www.pasternack.com/images/ProductPDF/RG225-U.pdf

 

10.16 mm OD @ $39.81 per foot
https://www.pasternack.com/semirigid-0.118-50-ohm-coax-cable-copper-pe-118sr-p.aspx

https://www.pasternack.com/images/ProductPDF/PE-118SR.pdf#page=2

[flkin]

I have this one but haven't listened to it yet. Double shielded. Its rather thin but construction wise it feels pretty solid and reasonably priced at $34 for a 12 inch length.

 

https://www.pasternack.com/bnc-male-bnc-male-rg400u-cable-assembly-pe3582lf-12-p.aspx

[flkin]

In addition to the cable from 10Mz Clock to the tX-USBultra (in the case of the Trifecta), the other three 24/25 Hz clock distribution cables from tX-USBultra to the switch and sMS200 would probably need attention too.

 

I found these double shielded ones that seem to be better than the stock recommended ones by SOTM-US. The link is for double right angled connections but normal straight ones are available too. $57 for 12 inch lengths.

 

https://www.pasternack.com/smb-plug-smb-plug-rg316-ds-cable-assembly-pe33476lf-12-p.aspx

 

At short lengths such as these, I'm not convinced that signal drop per 100m matters. We're talking about 1-2db drop differences over 100m making drops over 30cm lengths insignificant. Shielding and termination quality I suspect matter more.

 

[tapatrick]

On 30/01/2018 at 5:58 AM, kelvinwsy said:

i am also considering going with a teradak linear psu for mobo and use my corsair ax760 platinum psu for the GPU.. please share your experience ...

 

21 hours ago, Lebouwsky said:

For almost 2 years I’ve bin using 2 Teradak 30W powersupplies. They’re very reliable, don’t run hot and they’re price/performance is good. That said, in terms of given sound quality, they’re “ok”. The Sotm sps500 was a huge step up for me. You get what you pay for.

I have modified the caps on my Teradak DC30W and it compares favourably with the best at a big diet price. With LT3045s after the output to drop down the voltage, it’s hard to beat. 

[hurka]

26 minutes ago, tapatrick said:

 

I have modified the caps on my Teradak DC30W and it compares favourably with the best at a big diet price. With LT3045s after the output to drop down the voltage, it’s hard to beat. 

I have 2 similar regulated psu (8 Rail total), when feed all components like sparky sbc ,iso r,tx usb ultra,singxer,DA,

lost air and precision.But with only DA and tx usb,  / Sparky with ultracaps,/singxer,iso r with sps 500,

The Best sound here!

[romaz]

On 1/27/2018 at 1:29 AM, austinpop said:

Review of the Innuos Zenith MkII SE

 

Here - finally! - is my promised review of the Zenith MkII SE...

 

Hi, I'm back (but probably only for this 1 post) .

 

On top of my responsibilities as a husband, dad, full-time physician and high school varsity tennis coach, I have recently become significantly more active with a non-profit medical volunteer organization and so to maintain my sanity, I stopped following all forum threads some time ago,  It took quite a while to get over my "forum withdrawal" and I was doing so well until Rajiv e-mailed me a link to his review of the Zenith SE (curses, Rajiv, I thought we were friends!!!) and so here I am, sucked back in but only to clarify a few questions and to provide a brief update.  Truth be told, I'm on the tail end of a trip to Europe and presently waiting at Heathrow airport for my plane to depart.  It has given me a chance to draft this post.

 

First of all, a truly excellent review, Rajiv, although I've come to expect this from you for some time now!  I would have to concur, our conclusions on the Zenith SE are very similar and while it may come as a surprise to some that I purchased an off-the-shelf server when I worked so hard to put together my own server, I did it because I have 2 listening rooms (a near-field Voxativ setup in my home office directly driven by my Blu Mk2/DAVE and my Martin Logan Renaissance 15A hybrid electrostats in my large family room) and it became disconcerting to have a really good server in one room and my Mac Mini powered by my HDPlex in the other room.  If SR7 rails grew on trees, I would have just built another server but as many know, SR7 rails are difficult to come by which made the SE a very attractive option.

 

It should come as no surprise to anyone that I have modded my SE (EMI-absorbing paper to the chassis and RAM, Synergistic Research ECTs, SR Blue Fuse upgrade, Bybee iQSE, Pachanko SATA cable) and they have each contributed to an improvement with perhaps the SR Blue Fuse having the most notable impact.  As I posted on Head-Fi, when preceded by my SOtM reclocked switch and followed by my tX-USBultra (with both connected to my REF10 and powered by SR7 rails), this setup is very much to my liking and competitive with the very best that I have heard in my listening room.  

 

Since I purchased my SE, I have taken it apart and have put it back together (I have probably voided my warranty) and I have had conversations with both Nuno Vitorino and Sean Jacobs about its design and so I have a pretty good idea of the efforts that went into this server.  Out of respect to both Innuos and to Sean (who is independent of Innuos), I won't divulge every detail about this server that I know.  As has been advertised by Innuos, it is chiefly the power supply designed and implemented by Sean that separates the SE from the standard Zenith Mk2 and indeed, the implementation of this PSU is very well done.  Having just paid Sean a visit at his home in West Yorkshire while here in the U.K. (regarding a non-server related project), I was expecting to meet an older, crusty, bespectacled "professor-type" engineer only to be met at the train station by a youngish (think Doogie Howser with a beard) mountaineering aficionado who has already scaled the Himalayas and is wise beyond his years (Sean has a PhD in Robotics and appears well versed in many areas of electronics).  As we worked on our "project" in his lab, it was fascinating to observe just how meticulous, exacting, and tidy he is.  Did it inspire confidence?  Yes, absolutely.

 

Comments have been posted about approaching Sean for a custom ATX power supply and I even saw a proposal for a group buy.  While Sean is young and full of energy, he is a "one man" operation and he personally hand-builds each power supply that he sells and so don't expect a group buy discount, however, Sean has told me that he can build a multi-rail linear ATX power supply for the DIYers on this thread who want something substantially better than an HDPlex or Teradak (in fact, he has already built such a supply for others).  Out of respect to Innuos, he will not provide you unfinished boards that you can install into your chassis but he will sell you a completed linear PSU in a separate chassis that will directly connect to the CPU/ATX connectors on your motherboard of choice via an umbilical cable that will include the requisite 12V, 5V, and 3.3V rails.  For those interested in a PSU for a more powerful motherboard intended for upsampling (i.e. Larry), this is possible but depending on the power requirements, a larger chassis with beefier heatsinks and greater cost might become necessary.  Regarding the question about using Mundorf's best (and more expensive) Silver-Gold caps, those are signal caps and according to Sean, they will likely have no appreciable impact in a power supply although because these are custom builds, you can certainly handpick (and pay extra for) certain components.  If I recall correctly, turnaround will be somewhere between 1-2 months with sourcing of the transformer that he prefers being the time-limiting factor.  As for the price, it will depend on how many rails you require and I'm sure Sean will be happy to quote you a price.  As Sean is not active on the forums, you can reach him at [email protected].

 

As to whether I think the Zenith SE is the finest turnkey low-power server there is, to be honest, until I hear the Antipodes DX Gen3, I can't say.  Having communicated with Mark Jenkins about the Gen3, he commissioned ASUS to build him a custom motherboard and apparently he did a lot of testing and listening before deciding on what he wanted on this motherboard.  Of interest, he found that in some locations on the board, switching regulators were better than linear regulators because they had faster response (I presume he's talking about the regulator for the CPU itself).  He also chose not to use any high-end clocks because it is his belief that clocks make no difference in a USB server (obviously, many of us will beg to differ).  Furthermore, he chose to avoid any filtering because he felt that filters constrain bandwidth which can have a negative impact on the immediacy of your music.  It is his belief that with good power delivery and a properly designed motherboard, filtering becomes unnecessary (and unwanted).  Whether this means he has done away with the SOtM tX-USBhubIN card he used in the Gen2 is unclear.  Having directly compared the Gen2 to the SE, I prefer the SE, however, Mark believes the Gen3 in some ways is "night and day" better than the Gen2 even though the Gen3 utilizes the same PSU that is in the Gen2.  I have been waiting for months now for my local dealer to get a DX Gen3 demo unit on hand so that I can compare it to what I have but thus far, this has not happened.  Something tells me that even if it is better than the SE, it probably won't be night and day better.  Something also tells me that Innuos has something even better than their SE brewing in their pipeline.  Regardless, competition is good and I love how rapidly the digital side of audio continues to evolve.

 

As for the DIY server builders on this thread, I still think the biggest challenge remains sourcing a good ATX power supply and perhaps, Sean holds the answer.  The SE has taught me that is preferable to be able to independently power the CPU separate from the rest of the motherboard (although I believe Larry and others have been preaching this for some time).  For the sake of keeping impedance as low as possible, I also believe it's best to avoid high impedance connectors like barrels and to position your final regulator stage as close to the motherboard as possible.  What is odd is that the Zenith SE appears to use a PicoPSU 12V DC-ATX converter.  Having purchased both a PicoPSU and HDPlex DC-ATX converter, I found the HDPlex to sound a bit better but found that both were a step backward in my setup although it's possible the SE is using one that was custom made for them.  While I obviously don't buy Mark Jenkins' claim that high quality clocks don't benefit a USB server, I do agree that too much filtering can have its downsides (too soft a sound) and that designing a proper motherboard may be the key. 

 

While I am content with what I have for now, at some point (as a leisurely hobby rather than as a necessity), I may yet try and design my own.  For those interested, one way of doing it without going to the expense of designing a complete motherboard from scratch is to buy a SOM (System-on-Module) from someone like DFI and these SOMs are relatively inexpensive.  These small boards essentially contain an Intel SoC (System-on-a-chip) which would then contain the CPU of your choosing (from Celeron to i7) and an appropriate Intel chipset as well as RAM slots but nothing else (i.e. no audio, video, LAN, USB connectors or PCIe slots).  This SOM board then plugs into a "Carrier Board" that customers can design themselves and it is the Carrier Board that will house all peripherals and ports. 

 

https://www.dfi.com/Category/Index/2

 

For example, if I chose a uATX-sized Carrier Board, I could design it with 3 PCIe x 1 slots (for dual SOtM tX-USBexp PCIe cards and a JCAT ethernet PCIe card, as an example).  I could also choose to incorporate an integrated CF/SD flash card reader or a high bandwidth, low latency NVMe slot (which is generally not seen on low power Celeron boards).  Depending on the chipset, one could also choose to integrate Thunderbolt 3, a fiber NIC or any interface you can think of as long as there is physical space on the Carrier Board to do it.  In my case, I would specify a multi-layer board with at least 3oz of copper for an enhanced ground plane.  Obviously, I would choose to exclude any unnecessary items found in most off-the-shelf boards like an integrated audio card and its accompanying dirty clock.  It is SOtM's belief that all bad clocks (even if they are not directly in the signal path) will contribute noise to the board and ideally should be replaced.  I would also bypass the typical ATX connectors altogether and replace them with simpler independent but low impedance input connectors for 12V, 5V, 3.3V, and 1.2V (so that I can independently power my RAM).  The beauty of this type of setup is that as newer CPUs and chipsets become available, you can swap out the SOM but still keep the Carrier Board.  Something to keep in mind for those who are ambitious enough.

 

To conclude, I thought I'd sum up my recent experience on the impact of all of these tweaks relative to other things in my audio chain although I suspect I won't be telling anyone anything they don't already know.  Many here already know my penchant for Voxativ speakers and I have yet to hear anything that speaks to me better than my custom Voxativ AC-4X speakers directly driven by my Blu Mk2 and DAVE.  Now that they are finally properly powered, I have really grown to love the giant sound stage and transparency of my large Martin Logans but for reasons I won't get into, I also have recently auditioned the Kii Threes (nearly $15k with stands) and the Wilson Alexia Series 2 ($55k).  As easy as it is to discern the sound quality differences between a standard Mac Mini and a Zenith SE, the SQ differences among these 4 speakers make the gap between a Mac Mini and the SE sound rather minuscule in comparison.  While I am not in the market to buy speakers in the range of the Alexias, if someone were to offer me a pair of Alexias with the stipulation that I could only use a standard Mac Mini as a source, I would take that deal in a heartbeat and never look back.  In the end, it is still the transducer that has the potential to make the biggest difference.

 

As I will be heading home shortly and will be returning to the business of my normal routine, I likely will not be providing a follow up response to this post.  I wish you all well.

[BigAlMc]

Yet another epic post from @romaz

 

Great stuff. Here's hoping he manages to pop back soon! 

[lateboomer]

13 hours ago, austinpop said:

To clarify further, here is my latest topology, compared to my older one. My trifecta chain looked like this:

 

 

My new chain with the Zenith SE now looks like this:

 

 

 

Hope this clarifies things.

Hi Austinpop,

 

It is embarrass for me to ask this novice question, but I hope you could explain briefly to me. In the first trifecta, the server is supposed bridged and direct connect to its client sms-200. Then may I ask what is the purpose of switch Zyxel in between the server and client? Does this still consider to be direct bridged? And in latest topology, there are even 2 switches precede the Zenith SE, so again why do you need Zyxel? Many Thanks.

[auricgoldfinger]

8 hours ago, flkin said:

In addition to the cable from 10Mz Clock to the tX-USBultra (in the case of the Trifecta), the other three 24/25 Hz clock distribution cables from tX-USBultra to the switch and sMS200 would probably need attention too.

 

I found these double shielded ones that seem to be better than the stock recommended ones by SOTM-US. The link is for double right angled connections but normal straight ones are available too. $57 for 12 inch lengths.

 

https://www.pasternack.com/smb-plug-smb-plug-rg316-ds-cable-assembly-pe33476lf-12-p.aspx

 

At short lengths such as these, I'm not convinced that signal drop per 100m matters. We're talking about 1-2db drop differences over 100m making drops over 30cm lengths insignificant. Shielding and termination quality I suspect matter more.

 

 

Have you actually tried these cables to compare them to those recommended by SOtM?  If so, would you please describe the nature and magnitude of the difference in SQ?

[LTG2010]

old post deleted

[KingRex]

36 minutes ago, lateboomer said:

Hi Austinpop,

 

It is embarrass for me to ask this novice question, but I hope you could explain briefly to me. In the first trifecta, the server is supposed bridged and direct connect to its client sms-200. Then may I ask what is the purpose of switch Zyxel in between the server and client? Does this still consider to be direct bridged? And in latest topology, there are even 2 switches precede the Zenith SE, so again why do you need Zyxel? Many Thanks.

I'm in the same boat too.  Did a little more reading last night, but this chain is huge.  Its hard to find the relevant answer to some questions.  I too don't get why 2 switches either.  I do get a bad switch following a good switch is no improvement.   I also don't understand why the clock only seems to have one connection to the USBultra?  Is it a software directive that tells it to see the clock and use it.  And there is also a jumper bypassing the Zenith from the Zyxel switch to the USBultra.  hmmm.

May at SOTM is saying they don't modify any switches.  I guess the answer is you are doing the work yourself.  Is that correct?  What about a Aqvox switch.  Do they provide any improvement for streaming from the NAS/Tidal.

Thanks

[austinpop]

6 hours ago, romaz said:

Hi, I'm back (but probably only for this 1 post) .

 

Welcome back, Roy. I hope you return now and again!

 

Quote

(curses, Rajiv, I thought we were friends!!!)

 

 I had to give credit where it was due.

[austinpop]

1 hour ago, lateboomer said:

Hi Austinpop,

 

It is embarrass for me to ask this novice question, but I hope you could explain briefly to me. In the first trifecta, the server is supposed bridged and direct connect to its client sms-200. Then may I ask what is the purpose of switch Zyxel in between the server and client? Does this still consider to be direct bridged? And in latest topology, there are even 2 switches precede the Zenith SE, so again why do you need Zyxel? Many Thanks.

 

31 minutes ago, KingRex said:

I'm in the same boat too.  Did a little more reading last night, but this chain is huge.  Its hard to find the relevant answer to some questions.  I too don't get why 2 switches either.  I do get a bad switch following a good switch is no improvement.   I also don't understand why the clock only seems to have one connection to the USBultra?  Is it a software directive that tells it to see the clock and use it.  And there is also a jumper bypassing the Zenith from the Zyxel switch to the USBultra.  hmmm.

May at SOTM is saying they don't modify any switches.  I guess the answer is you are doing the work yourself.  Is that correct?  What about a Aqvox switch.  Do they provide any improvement for streaming from the NAS/Tidal.

Thanks

 

Guys, the index on the first post was my best attempt to make it easier to navigate this thread, but there is only so much one person can do to reduce a monster thread to digestible form. All I can say is - use the index to go to interesting topics and then read a few pages around them for context.

 

Regarding the reclocked switch. Think of its role not so much in networking terms, since it's really not needed at that location for Ethernet switching at all. No - think of it as a kind of Ethernet reclocker, just like the tX-USBultra (or ISO-Regen) is for USB. It doesn't serve any network purpose there.

 

I did a quick search, and found one representative post to reinforce that:

https://www.computeraudiophile.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/?page=92&tab=comments#

 

As for SOtM not modifying switches - that is news to me. I know @Bruce Orr recently got it done. Are you sure you're asking May the right question? There can be a language issue as English is not her first language. SOtM mods switches:

• to accept a 25MHz clock, via an SMB port on the back panel, from an external sCLK-EX board, either in an Ultra component (for example tX-uSBultra) or in a DIY custom server.
• replace capacitor(s)
• replace switching regulators with linear ones.

They've been doing this for quite a while now. Their modding policy, at least as far as I was aware, was that they will not mod another competitor's audio product. So certain things - like ISO-Regen, or SU-1 - were not accepted.

 

See this link from Crux Audio. Kamal Bekkari who runs that site is an SOtM reseller or distributor in the US and Australia, for example: https://sotm-usa.com/collections/ultra-series-mods/products/sms-200ultra-ethernet-switch-mod

 

Regarding the clock distribution. I got to this link from the index:

https://www.computeraudiophile.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/?do=findComment&comment=722622

 

[Bruce Orr]

Guys, I can confirm that May/SoTM did indeed modify a Netgear GS-108 for me. I received the modified switch over Xmas.  It was modified three ways:

- bypass its internal clock and create a new in-port to accept an available sCLK tap from the tx-USBultra. The ultra was modified by creating an sCLK out-port. A skinny little interconnect cable was provided to couple the two. 

- I also asked SoTM to modify any capacitors, etc. they deemed would be an upgrade

- it was also modified to enable it to be powered by a 7V source. 

 

I have no idea whether they have changed their service platform, but it would surprise me. 

[AmusedToD]

6 hours ago, romaz said:

To conclude, I thought I'd sum up my recent experience on the impact of all of these tweaks relative to other things in my audio chain although I suspect I won't.  While I am not in the market to buy speakers in the range of the Alexias, if someone were to offer me a pair of Alexias with the stipulation that I could only use a standard Mac Mini as a source, I would take that deal in a heartbeat and never look back.  In the end, it is still the transducer that has the potential to make the biggest difference.

 

Great to see you back, @romaz!

 

When we met at the Hi End 2017 show in Munich, Germany and discussed our impressions from visiting different audio rooms, I told you I was absolutely mesmerized by what I heard in the TAD room and as well is in the Simaudio room (they demonstrated the 888 flagship mono amps on a pair of Wilson Alexx speakers). Neither TAD nor Simaudio/Wilson had fancy music servers with linear PSU’s and/or superior clocks. Both used simple MacBook laptops directly powered from the wall (!) and  directly connected via low cost USB cables to their respective DACs (D1000 mk2 in the case of TAD, and 780D in the case of Moon). Yet, both setups were absolutely mind blowing in terms of their presentations of music.

 

To conclude, I agree with you 100% and thus I could never decide to spend big money on a server. I could have kept my Dynaudio speakers and invested in a digital server, yet after Munich I made a decision to do the exact opposite - ditch the Dyn’s, get a cheap server (I moved from an iMac to an even cheaper NUC this month) and put my money on speakers (the entry level TAD’s in my case). And I am convinced this is the right way to go, especially after some experimentation with a couple of SOtM products.

 

[onlychild]

For anyone interested, I needed to cancel my Innuos Zenith SE order so my dealer, Audio Doctor, has one left with no wait time (actually just received it today from Innuos). He is one of the authorized dealers in the U.S. found on Innuos's website.

You can reach him at http://www.audiodoctor.com/index.html

[lmitche]

3 hours ago, austinpop said:

 

Regarding the reclocked switch. Think of its role not so much in networking terms, since it's really not needed at that location for Ethernet switching at all. No - think of it as a kind of Ethernet reclocker, just like the tX-USBultra (or ISO-Regen) is for USB. It doesn't serve any network purpose there.

Here are two experiments that would be interesting:

 

1) Separately clock the tx-USBultra and network switch to learn if the synchrony matters.  One would need two clk boards and potentially two reference clocks for this experiment.

 

2) Modify a Netgear fs-10x with the clock mod and jsgt to learn if two switches are really necessary or one will do.

 

It makes sense to me that a USB or Ethernet re-clocker improves with a better clock.  It is the synchrony between components that has me curious.

 

To my mind, the answer to this synchronized clock question is the last frontier we need to conquer.

[lmitche]

9 hours ago, romaz said:

 

Hi, I'm back (but probably only for this 1 post) .

 

On top of my responsibilities as a husband, dad, full-time physician and high school varsity tennis coach, I have recently become significantly more active with a non-profit medical volunteer organization and so to maintain my sanity, I stopped following all forum threads some time ago,  It took quite a while to get over my "forum withdrawal" and I was doing so well until Rajiv e-mailed me a link to his review of the Zenith SE (curses, Rajiv, I thought we were friends!!!) and so here I am, sucked back in but only to clarify a few questions and to provide a brief update.  Truth be told, I'm on the tail end of a trip to Europe and presently waiting at Heathrow airport for my plane to depart.  It has given me a chance to draft this post.

 

First of all, a truly excellent review, Rajiv, although I've come to expect this from you for some time now!  I would have to concur, our conclusions on the Zenith SE are very similar and while it may come as a surprise to some that I purchased an off-the-shelf server when I worked so hard to put together my own server, I did it because I have 2 listening rooms (a near-field Voxativ setup in my home office directly driven by my Blu Mk2/DAVE and my Martin Logan Renaissance 15A hybrid electrostats in my large family room) and it became disconcerting to have a really good server in one room and my Mac Mini powered by my HDPlex in the other room.  If SR7 rails grew on trees, I would have just built another server but as many know, SR7 rails are difficult to come by which made the SE a very attractive option.

 

It should come as no surprise to anyone that I have modded my SE (EMI-absorbing paper to the chassis and RAM, Synergistic Research ECTs, SR Blue Fuse upgrade, Bybee iQSE, Pachanko SATA cable) and they have each contributed to an improvement with perhaps the SR Blue Fuse having the most notable impact.  As I posted on Head-Fi, when preceded by my SOtM reclocked switch and followed by my tX-USBultra (with both connected to my REF10 and powered by SR7 rails), this setup is very much to my liking and competitive with the very best that I have heard in my listening room.  

 

Since I purchased my SE, I have taken it apart and have put it back together (I have probably voided my warranty) and I have had conversations with both Nuno Vitorino and Sean Jacobs about its design and so I have a pretty good idea of the efforts that went into this server.  Out of respect to both Innuos and to Sean (who is independent of Innuos), I won't divulge every detail about this server that I know.  As has been advertised by Innuos, it is chiefly the power supply designed and implemented by Sean that separates the SE from the standard Zenith Mk2 and indeed, the implementation of this PSU is very well done.  Having just paid Sean a visit at his home in West Yorkshire while here in the U.K. (regarding a non-server related project), I was expecting to meet an older, crusty, bespectacled "professor-type" engineer only to be met at the train station by a youngish (think Doogie Howser with a beard) mountaineering aficionado who has already scaled the Himalayas and is wise beyond his years (Sean has a PhD in Robotics and appears well versed in many areas of electronics).  As we worked on our "project" in his lab, it was fascinating to observe just how meticulous, exacting, and tidy he is.  Did it inspire confidence?  Yes, absolutely.

 

Comments have been posted about approaching Sean for a custom ATX power supply and I even saw a proposal for a group buy.  While Sean is young and full of energy, he is a "one man" operation and he personally hand-builds each power supply that he sells and so don't expect a group buy discount, however, Sean has told me that he can build a multi-rail linear ATX power supply for the DIYers on this thread who want something substantially better than an HDPlex or Teradak (in fact, he has already built such a supply for others).  Out of respect to Innuos, he will not provide you unfinished boards that you can install into your chassis but he will sell you a completed linear PSU in a separate chassis that will directly connect to the CPU/ATX connectors on your motherboard of choice via an umbilical cable that will include the requisite 12V, 5V, and 3.3V rails.  For those interested in a PSU for a more powerful motherboard intended for upsampling (i.e. Larry), this is possible but depending on the power requirements, a larger chassis with beefier heatsinks and greater cost might become necessary.  Regarding the question about using Mundorf's best (and more expensive) Silver-Gold caps, those are signal caps and according to Sean, they will likely have no appreciable impact in a power supply although because these are custom builds, you can certainly handpick (and pay extra for) certain components.  If I recall correctly, turnaround will be somewhere between 1-2 months with sourcing of the transformer that he prefers being the time-limiting factor.  As for the price, it will depend on how many rails you require and I'm sure Sean will be happy to quote you a price.  As Sean is not active on the forums, you can reach him at [email protected].

 

As to whether I think the Zenith SE is the finest turnkey low-power server there is, to be honest, until I hear the Antipodes DX Gen3, I can't say.  Having communicated with Mark Jenkins about the Gen3, he commissioned ASUS to build him a custom motherboard and apparently he did a lot of testing and listening before deciding on what he wanted on this motherboard.  Of interest, he found that in some locations on the board, switching regulators were better than linear regulators because they had faster response (I presume he's talking about the regulator for the CPU itself).  He also chose not to use any high-end clocks because it is his belief that clocks make no difference in a USB server (obviously, many of us will beg to differ).  Furthermore, he chose to avoid any filtering because he felt that filters constrain bandwidth which can have a negative impact on the immediacy of your music.  It is his belief that with good power delivery and a properly designed motherboard, filtering becomes unnecessary (and unwanted).  Whether this means he has done away with the SOtM tX-USBhubIN card he used in the Gen2 is unclear.  Having directly compared the Gen2 to the SE, I prefer the SE, however, Mark believes the Gen3 in some ways is "night and day" better than the Gen2 even though the Gen3 utilizes the same PSU that is in the Gen2.  I have been waiting for months now for my local dealer to get a DX Gen3 demo unit on hand so that I can compare it to what I have but thus far, this has not happened.  Something tells me that even if it is better than the SE, it probably won't be night and day better.  Something also tells me that Innuos has something even better than their SE brewing in their pipeline.  Regardless, competition is good and I love how rapidly the digital side of audio continues to evolve.

 

As for the DIY server builders on this thread, I still think the biggest challenge remains sourcing a good ATX power supply and perhaps, Sean holds the answer.  The SE has taught me that is preferable to be able to independently power the CPU separate from the rest of the motherboard (although I believe Larry and others have been preaching this for some time).  For the sake of keeping impedance as low as possible, I also believe it's best to avoid high impedance connectors like barrels and to position your final regulator stage as close to the motherboard as possible.  What is odd is that the Zenith SE appears to use a PicoPSU 12V DC-ATX converter.  Having purchased both a PicoPSU and HDPlex DC-ATX converter, I found the HDPlex to sound a bit better but found that both were a step backward in my setup although it's possible the SE is using one that was custom made for them.  While I obviously don't buy Mark Jenkins' claim that high quality clocks don't benefit a USB server, I do agree that too much filtering can have its downsides (too soft a sound) and that designing a proper motherboard may be the key. 

 

While I am content with what I have for now, at some point (as a leisurely hobby rather than as a necessity), I may yet try and design my own.  For those interested, one way of doing it without going to the expense of designing a complete motherboard from scratch is to buy a SOM (System-on-Module) from someone like DFI and these SOMs are relatively inexpensive.  These small boards essentially contain an Intel SoC (System-on-a-chip) which would then contain the CPU of your choosing (from Celeron to i7) and an appropriate Intel chipset as well as RAM slots but nothing else (i.e. no audio, video, LAN, USB connectors or PCIe slots).  This SOM board then plugs into a "Carrier Board" that customers can design themselves and it is the Carrier Board that will house all peripherals and ports. 

 

https://www.dfi.com/Category/Index/2

 

For example, if I chose a uATX-sized Carrier Board, I could design it with 3 PCIe x 1 slots (for dual SOtM tX-USBexp PCIe cards and a JCAT ethernet PCIe card, as an example).  I could also choose to incorporate an integrated CF/SD flash card reader or a high bandwidth, low latency NVMe slot (which is generally not seen on low power Celeron boards).  Depending on the chipset, one could also choose to integrate Thunderbolt 3, a fiber NIC or any interface you can think of as long as there is physical space on the Carrier Board to do it.  In my case, I would specify a multi-layer board with at least 3oz of copper for an enhanced ground plane.  Obviously, I would choose to exclude any unnecessary items found in most off-the-shelf boards like an integrated audio card and its accompanying dirty clock.  It is SOtM's belief that all bad clocks (even if they are not directly in the signal path) will contribute noise to the board and ideally should be replaced.  I would also bypass the typical ATX connectors altogether and replace them with simpler independent but low impedance input connectors for 12V, 5V, 3.3V, and 1.2V (so that I can independently power my RAM).  The beauty of this type of setup is that as newer CPUs and chipsets become available, you can swap out the SOM but still keep the Carrier Board.  Something to keep in mind for those who are ambitious enough.

 

To conclude, I thought I'd sum up my recent experience on the impact of all of these tweaks relative to other things in my audio chain although I suspect I won't be telling anyone anything they don't already know.  Many here already know my penchant for Voxativ speakers and I have yet to hear anything that speaks to me better than my custom Voxativ AC-4X speakers directly driven by my Blu Mk2 and DAVE.  Now that they are finally properly powered, I have really grown to love the giant sound stage and transparency of my large Martin Logans but for reasons I won't get into, I also have recently auditioned the Kii Threes (nearly $15k with stands) and the Wilson Alexia Series 2 ($55k).  As easy as it is to discern the sound quality differences between a standard Mac Mini and a Zenith SE, the SQ differences among these 4 speakers make the gap between a Mac Mini and the SE sound rather minuscule in comparison.  While I am not in the market to buy speakers in the range of the Alexias, if someone were to offer me a pair of Alexias with the stipulation that I could only use a standard Mac Mini as a source, I would take that deal in a heartbeat and never look back.  In the end, it is still the transducer that has the potential to make the biggest difference.

 

As I will be heading home shortly and will be returning to the business of my normal routine, I likely will not be providing a follow up response to this post.  I wish you all well.

Nice write-up Roy! Thanks for making the time for this contribution.

 

Keep an eye out for the results of my testing with separate PSUs on my upsampling machine. It may be a few weeks as I've hit some unexpected snags.  Nevertheless the results, while preliminary, are very promising.

 

I will look into Sean's LPSU solutions.

 

As regards a minimal motherboard, at least one of the DFI boards looks to have the power for hqplayer upsampling to dsd512.  I don't need any motherboard connections outside of a 24 pin atx and 8 pin 12 volt  power and one 4x pcie direct to cpu connector.  A second direct to cpu pcie or m.2 nvme slot is best.

 

With 3 USB chains(DAC, nic and hdd) and a USB card, all lt3045 powered, only the motherboard and CPU are dependent on atx PSU smps power. With titanium level supply ripple at 5mv, things sound way good with the two separate 12 volt rails. More to come.

 

Thanks again.

 

Larry

[Johnseye]

10 hours ago, romaz said:

As easy as it is to discern the sound quality differences between a standard Mac Mini and a Zenith SE, the SQ differences among these 4 speakers make the gap between a Mac Mini and the SE sound rather minuscule in comparison.  While I am not in the market to buy speakers in the range of the Alexias, if someone were to offer me a pair of Alexias with the stipulation that I could only use a standard Mac Mini as a source, I would take that deal in a heartbeat and never look back.  In the end, it is still the transducer that has the potential to make the biggest difference.

 

 

What I quoted is something to always keep in mind as we dump time and money down other rabbit holes.

 

I suspect Sean will have his hands full after this exposure.  I'm very tempted to request an ATX PSU.

 

[RickyV]

I am also interested/ Curious in such a atx psu but i am affraid the price will also be interesting.

i will probably diy one.