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

[jean-michel6]

23 minutes ago, sandyk said:

 

 It's not that simple . An Ultracap will act like a short circuit to a normal voltage regulator unless additional current limiting at start up is used , as it tries to draw HUGE amounts of current initially from  the voltage regulator and it's power supply.

Even  a high powered amplifier with large value filter capacitors may momentarily dim a nearby table lamp unless it uses a "soft start"circuit.

Thank for the clear explanation , i will not play wit supercaps for the time being as it is not that easy to implement .

[jean-michel6]

12 minutes ago, flkin said:

 

Big, big difference when components are placed on vibration managed platforms. Years ago when I placed my AR CD-7 and BAT VK-5i pre-amp on an Symposium Isis platform and replaced the existing footers with Symposium solids couplers and Rollerblocks, the change in sound was nothing short of a major system upgrade. Bigger soundstage, more solid presence, clarity of sound. A sceptical friend that helped me assemble the rack was gobsmacked by the change. 

 

Vibration management is so important that I was contemplating atomic microscope active platforms to reduce vibrations. 

Fully agree with you !

The three evil to fight for netter music in our crazy digital systems are : vibration , quality of PSU and related noise , less jitter

[jean-michel6]

1 hour ago, seeteeyou said:

 

The ones from Micron should be a joke because the current draw seemed to be nothing to write home about. Samsung ain't too bad the and lowest numbers of M471B5674QH0 are kept under 500mA. SK Hynix should be fantastic and HMT425S6CFR6A would give us numbers that are belowe 400mA. I just kept looking but Crucial might not be willing to publish any numbers or something.

 

It ain't an exhaustive list by any means, just giving everyone a glance at what kinda DDR3L memory modules we could get on the market right now

 

http://www.samsung.com/global/business/semiconductor/file/product/135V_DDR3_4Gb_Qdie_UnbufferedSODIMM_Rev121.pdf#page=28

 

http://www.samsung.com/global/business/semiconductor/file/2011/product/2011/9/2/412764ds_ddr3_2gb_d-die_based_sodimm_rev14.pdf#page=20

 

http://www.samsung.com/global/business/semiconductor/file/product/ds_ddr3_2gb_c-die_based_1_35v_sodimm_rev131-0.pdf#page=23

 

https://www.skhynix.com/products.view.do?vseq=1740&cseq=75

 

https://www.skhynix.com/products.view.do?vseq=1742&cseq=75

 

https://www.skhynix.com/products.view.do?vseq=1946&cseq=75

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT325S6BFR8A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT325S6CFR8A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT325S6EFR8A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT425S6AFR6A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT425S6MFR6A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT351S6BFR8A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT351S6CFR8A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT351S6EFR8A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT451S6AFR6A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT451S6MFR8A&rk=20&rc=module

 

https://www.skhynix.com/eolproducts.view.do?pronm=DDR3+SDRAM&srnm=HMT312S6DFR6A&rk=20&rc=module

 

And then we've got some "monsters" from a few companies, unfortunately they're just boasting "low power consumption" without bothering to back that claim up with a datasheet of some sort

 

http://industrial.adata.com/us/product/317

https://www.transcend-info.com/Embedded/Products/No-589

http://www.advantech.com/products/memory_module/sqr-sd3i/mod_c55cd924-ea83-43e3-962a-04ef161594e1

There is also an important parameter to consider I think is the latency ?

[jean-michel6]

3 hours ago, seeteeyou said:

 

It's more likely to be a big deal for gamers according to this post

 

https://www.computeraudiophile.com/forums/topic/13948-article-computer-audiophile-pocket-server-caps-v3-lagoon/?page=5&tab=comments#comment-192067

 

Actually Roy was throttling the speed of both CPU and RAM to 800 MHz so they're synchronized to the same frequency. And then we could also tweak the PLL with ClockGen

 

http://www.sisoftware.eu/faq/pll-ic-clock-generator/

http://www.softpedia.com/get/Tweak/CPU-Tweak/ClockGen.shtml

 

Maybe we could dial something down a notch for the sake of lower power consumption?

 

BTW, someone is gonna share something to slim Windows 10 down with NTLite soon

 

https://www.ntlite.com/community/index.php?threads/w7-and-w10-presets-downloads.135/

https://www.ntlite.com/community/index.php?threads/w7-and-w10-presets-questions.154/

 

And then there's also another one here

 

https://www.ntlite.com/forum/discussion/1364/tutorial-for-creating-a-700-mb-windows-10-iso-and-install-in-a-vm

 

In that case, Windows 10 could be installed on SD cards with the lowest power consumption (28mA read / 55mA write) ever

 

https://www.mouser.com/access/?pn=922-603989

You are right in fact on my actual dual PC set up with gigabyte Mobo and celeron processor i have adjusted cpu and ram speed to 800Mhz with good results .

 

Unfortunately on the board i am testing now the super micro X10SBA , there is no provision in the bios to adjust cpu and ram speed 

[jean-michel6]

3 hours ago, hurka said:

Ifi power upgraded to sps500 level.

before use need charge up with special charger.

It look's interesting as it is much cheaper than SPS 500  . Could you provide detail ?

Capacitor type( supercar ? , voltage...) , regulator board (it look's like LT 3045 ) .

 

Which special charger are you using to recharge capacitors ?

[jean-michel6]

1 hour ago, KingRex said:

I am by no means any expert.  I have my opinnions I have developed by personal trial and conversations with a few industry professionals. I have developed a belief that power conditioning is best not used, unless absolutly necessary.  I need it.  I have dirty power at my house.  My Rega Osiris mechanically hummed so loud, it could be heard in the next room.  I took it to two friends houses, it was dead quit there.  I have been forced to seek out power conditioning.   No one wants to spend $1k to $7k and more unless necessary.  I did not. 

 

My CJ also hummed very loud.  I took it apart, floated the transformers on sorborhane and added sound dampener material to the covers.  It still made to much noise.  I tried the Syncro, it took out the mechanical noise but killed the music.  I bought the tripp lite.  Step in the right direction. I got the Topaz and it worked.  

 

TAS and Stereophile make it appear power conditioners are some sort of magic device that better all our stereo gear. They make it seem every stereo will suddenly have an expanded sound stage, more full bass, extended treble etc.  I don't believe that is true.  It is my opinion most alter the music but not in beneficial ways.  I have personally heard a veil or compression of dynamics from most everything I tried.  Why put a bunch of electronic equipment in front of your music if not needed.  It might not be necessary. 

 

That is a pretty round about way of saying, I don't know why a line isolation transformer sounds good on some components and not other.    I don't know why my amp benefitted, but not my preamp, phonograph preamp or music server.  

 

I will say this, the guy who sold me the Isotek Syncro for the Rega brought over other Isotek components to allow me to plug in all my equipment.  After plugging it all in, he even admitted the effect was negative and only the integrated amp was benefiting.  

 

I am fortunate I am an electrician by trade.  I have had the ability to run multiple dedicated lines from my panel to my rack.  I put the Topaz transformer in my basement with a dedicated feed from it to a quad of receptacles in my wall. I also had a second quad of receptacles with nothing in-between them and the panel.  I was able to switch all my equipment back and forth, then listen.  I ended up only feeding my CJ amp with the Topaz.  It worked great for that unit.  When I sold that amp, the Topaz did not work with my Altec.  They actually got more loud.  They did not like the transformer at all.  They made strange sounds.  That sent me down a new path.  In the end my cord direct from the panel to a custom distribution box with an Akiko was the trick.  All my components perform best when feed this way.  When I unplug the Akiko, it is hard to tell its gone.  At certain times of the day the amps will hum loud.  Plug the Akiko back in and 10 minutes later things calm down and the mechanical noise goes away.  

 

I will close by saying this, if a power cord or interconnect can have a heard alteration in the sound of playback, just think what an active component can do.  

Thank for this interesting info. 

My experience with isolation transformer is that you need one isolation transformer for each piece of équipement in your system. 

This help to clean the trash coming from ac but also helps isolate each component from each other which is also a big source of pollution. 

This was particularly effective for DAC.

For amp i use isolation transformer sized witn a power three time the maximum rated power of the amp. 

But for very bid amp's it is usually better with no isolation transformer .

[jean-michel6]

6 hours ago, seeteeyou said:

 

There's something called BLACKOS-lite from PVD-AUDIO in Russia

 

https://pvd-audio.com/blackos/

 

I checked their manual and the BIOS settings were indeed 800MHz for both CPU and RAM speed

 

https://pvd-audio.com/wp-content/uploads/2017/11/Rekomendatsii-po-nastrojke-BIOS.pdf#page=3

 

Then I also found their server with Asrock H87M-ITX and the dedicated PSU looked totally dope, it's interesting to see how they're connecting the PSU to the 24-pin ATX connector and the 4-pin 12V connector (i.e. pretty much like Innous ZENith Mk.II) separately

 

http://forum.doctorhead.ru/index.php?showtopic=27247&st=200

http://www.asrock.com/mb/Intel/H87M-ITX/

 

 

BTW, there's also this $250 internal regulator module from Mojo Audio if we're looking for more current than LT3045 could provide

 

http://www.mojo-audio.com/internal-regulator-modules/

 

Unfortunately that's so much higher than what we're getting from LT3045 and SR7 etc.

 

https://www.head-fi.org/threads/chord-electronics-dave.766517/page-382#post-13007750

 

Thank for this very interesting and informative post .

 

-Obviously several server manufacturer have understood that running at low CPU speed is good for the quality of the sound ( less electrical consumption gives less electrical noise ) .

 

- The mojo audio PSU is quite expensive for what it is , it is probably a good solution for people with low DIY skills  . Otherwise there is plenty in my opinion of better solutions.

One interesting one i have found in another CA thread is this one http://www.diyaudio.com/forums/vendor-s-bazaar/310402-fs-ultra-low-noise-power-supply-lt3045-based-pcb.html

This german guy has come up with LT 3045 power board who can be either 1,5 A or 3A , i bought from him the double 1,5 A board .

I have used it to power the SSD OS on the player board in my dual PC'S 

It did made some improvement on the lower background noise ( " darker background" ) in  the sound  compare to the normal PSU for this SSD which is a linear PSU with LT1083 regulator .

It is a small improvement but it shows again that when you decrease the noise of the PSU you get better results ...even on powering a ssd which is quite far away from the audio signal itself ...

One very interesting PSU to try i think will be the new uptone audio LPS 1-2 , i have the LPS-1 and like it a lot .

 

- I found extremely interesting your last link https://www.head-fi.org/threads/chord-electronics-dave.766517/page-382#post-13007750

 Roy has again given very valuable information .

 What i find extremely interesting is the information about the SI ( signal integrity of the USB signal ) . It shows that several key factors  are important to insure the proper signal integrity of the USB signal to minimise the decoding work and therefore noise at the input of the DAC  and the jitter is just one part of the problem ...

I think that explains why there is now so many product on the market ( regen , txUSB ultra ...) sold to improve the quality of the USB signal received by the DAC .

 

 

[jean-michel6]

21 hours ago, seeteeyou said:

FYI - it's quite interesting to see what motherboards were chosen for some music servers

 

Innuos ZENith Mk.II = Supermicro X10SBA-L
https://devialetchat.com/showthread.php?tid=3853

http://www.supermicro.com/products/motherboard/celeron/x10/x10sba-l.cfm

 

Antipodes Audio DX Gen 2 = Jetway NF9W
https://www.computeraudiophile.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/?page=74&tab=comments#comment-663416

http://jetwaycomputer.com/NF9W.html

 

Fidelizer Nimitra = Realan LR-J1900T1
https://www.stereo.net.au/forums/topic/126194-fidelizer-nimitra-audio-server-whats-going-on-here/

http://www.minicase.net/product_LR-J1900T1.html

 

Clones Audio HOST Music Server = Asrock J3710-ITX
https://parttimeaudiophile.com/2017/04/30/review-clones-audio-host-music-server/

http://www.asrock.com/mb/Intel/J3710-ITX/

 

Pink Faun AV Streamer = Asrock FM2A88X-ITX+
http://www.6moons.com/audioreviews2/pinkfaun/2.html
https://www.asrock.com/mb/AMD/FM2A88X-ITX+/

 

Sound Galleries SGM 2015 = Asus Z170M-PLUS
http://www.6moons.com/audioreviews2/soundgalleries/3.html

https://www.asus.com/Motherboards/Z170M-PLUS/

Thank for this interesting information. In fact on 4 of the 6 server you mention they are using celeron low consumption CPU which is in my opinion the right direction unless you want to upsample to DSD 256 or 512. 

 

I,would like to report on Supermicro x10sba.

my current set up is dual pc ( gigabyte mobo )with celeron processor 4 ssd os and music each powered by its own linear psu , optical bridge between the two pc's.

pcie sotm USB ,tx USB ultra.

i am running a Linux roon player in one pc an server in the other

There is 4 battery network to power separately each CPU and each mobo. 

This set up is already giving me extremely good musical results. 

 

However I thought that I could make improvement on the mobo I have been using and I have search for quite sometime the right board and I did selected the super micro. 

Today I was able to finalise my projet and use the supermicro board in the player function. The server is still handled by the gigabyte mobo. 

 

The results are just amazing , my system made a huge progress in musicality .

i got a much quieter background ,better bass and much more natural top. 

Mon good recording it is mesmerizing...

 

Ob top of that the super micro was powered by a simple 12v 1,8 A  IFi smps.

With a better PSU  sound quality will raise even more. 

 

One  obvious conclusion this Spermicro x10sba is a stellar motherboard for audio pc. 

 

This is is probably part of the explanation of the very good results of innuos server 

[jean-michel6]

6 hours ago, LTG2010 said:

True, but a big part is the superb power supply

Yes I agree the PSU is also an important part and innuos has done a good job on that. 

But i was amazed of the huge gap between the super micro board and the gigabyte. 

Ifi power is a good power for it's price but of course there are many options to go further. 

In my set up I have a few advantages over innuos choices 

- the two pc set up allow to have the simplest and less heavy software ( light Linux distro for audio with roon bridge ) on one pc for only the player...

- i have zero accessory drawing power from this board , only the board itself and the CPU and only 2gb ram. This allow for very low power consumption. The 12v input with 1.8A psu did the job easily. 

 

 

[jean-michel6]

19 hours ago, sandyk said:

 

 I use a dual +12V to +5V JLH PSU add-on for both my internal SSDs which improves their isolation from each other and the internal  SMPS.

 

Yes you are right use separate psu for each single equipment in the audio chain is one of the golden recipy to achieve better sound. 

It makes the system somewhat more complicated but it is very effective. v

[jean-michel6]

4 hours ago, seeteeyou said:

 

That's a winner for approximately $150 even though BIOS options don't offer anything for running both CPU and RAM at 800MHz.

 

Now there's a fairly important question to ask - could a single unit of LPS-1.2 with 13.2 watts (1.1A @ 12V) provide enough power for running Celeron J1900 or not? Most likely that won't work if all 4 cores were active but only ONE active core could do the trick

 

https://www.supermicro.com/manuals/motherboard/J1900/MNL-1553.pdf#page=75

 

In addition, we could also add Maximum Processor Frequency in Windows 10 Power Options manually

 

https://www.tenforums.com/tutorials/95580-add-remove-maximum-processor-frequency-windows-10-power-options.html

https://www.tenforums.com/tutorials/95574-change-maximum-processor-frequency-windows-10-a.html

 

Other options are also available

 

https://www.reddit.com/r/Windows10/comments/6gdv6z/i_wrote_a_script_to_unhide_all_advanced_power/

 

Basically we should be able to power X10SBA with LPS-1.2 (or DC3 Power Supply from Custom HiFi Cables) while something else like Intel X25-E could be powered by LT3045 accordingly

 

You are right maybe the new LPS 1.2 could power this board. 

I will try to measure the power consumption since the ifipower was lukewarm after 1 hour it is maybe in the range of the 1.2. 

 

You have a very sharp eye :-).  I read several times this X10sba excellent manual and I overlook the capability to run 4 core or 1 . This is probably something to try even though I think it should make the system somewhat slower but maybe sufficient for audio.  

 

Yes no option to slower processor speed .

 

The X10sba-l used by innuos is may be even better as it has some hardware removed from the board ( additional sata controler.... ). I may try this board since I need 2 boards , I will probably try the x10sbal-l in player function and x10sba in server function as I need three sata port for the server. 

I have not yes optimise the bios set up which allow to turn off most of unused board hardware. 

 

 

[jean-michel6]

3 hours ago, austinpop said:

Review of the Innuos Zenith MkII SE

 

Here - finally! - is my promised review of the Zenith MkII SE, referred to in the rest of this review simply as the SE. I know it feels like I've strung you guys along. I would have written this much earlier, had it not been for the untimely failure of my tX-USBultra, just when I was getting ready to do comparative listening at the 200 hour burn-in mark on my SE. Well, I have received my tX-USBultra back (a saga I posted about separately), which finally enabled me to complete this review. A side benefit of this delay is that my SE is truly burned-in now, with well over 300 hours of play time.

 

The SE came to my attention upon reading Roy's review on head-fi. After his public posts on the matter, he and I stayed in touch and he conveyed how impressed he was with it in his system. Finally, when a group of friends and I decided to do a private group buy, the die was cast. In due course, the SE arrived, and assumed a place in my system.

 

I'm not going to waste time and space here describing what the SE is and its feature set. If you haven't already, go read Roy's review on head-fi and Darko's review on his site. The SE is a standalone music player, that can store music locally, and serve them to a DAC directly via USB. For Roon users like me, the SE can function both as a Roon Core, but also as a Roon Bridge. If used as a bridge, the SE functions as a rather expensive streamer/renderer. In Roon Core mode, the box also provides a bridged streamer output, so you can drive an existing streamer. So it's a very flexible box! It also supports other modes like UPnP, Sonos, etc. I did not experiment with these.

 

As you know, my journey to SQ optimization has led me to my current spaghetti solution based on the SOtM trifecta. As a result, my perspective in evaluating the SE is uniquely mine. It's based on the question - how can the SE improve the overall SQ of my audio chain? Can it do so while simplifying my chain? And does it? This review is all about comparative performance. There are enough reports out there that suggest that on its own merit, the SE sounds wonderful. So my focus here is comparison with my trifecta chain.

 

Also note - we are at a fairly hefty price point here, and definitely in the diminishing returns region of the price-performance curve. I have no illusions on this score, nor am I even suggesting that this kind of expenditure is necessary for good SQ. 

 

Disclaimer

 

I've had people contact me wanting to know if I was affiliated with, or paid by, one company of the other. Let me set the record straight, once and for all. I only review, and plan to continue to only review, gear that interests me for my own system. All the gear is either purchased by me, or lent to me by friends and CA'ers - NOT by manufacturers. Like all of us, I hate paying retail, so will often buy used or discounted products, but not for review considerations. To be honest, I'm pretty small fry, and I prefer it that way! The main reason I write here is because a) I enjoy writing, and b) I feel too few reviews - especially from mainstream publications - take an unflinching, comparative approach, which is what I, and most of the audiophiles I know, really want to know.

 

OK - back to the interesting parts.

 

My Current Chain

 

Let's review the chain I have been using prior to the SE's arrival. The notation I'll use is to indicate the PSU in parentheses, including whether its output is shunted to ground via a JSGT (John Swenson Ground Tweak).  For all the LPS-1s in my chain, the energizing Meanwell SMPSes are all JSGT shunted. Also, unless otherwise indicated, upstream of the chains below is:

• Netgear CM600i modem (HDPlex LPS) > TP-Link Archer C7 router (HDPlex LPS) > Netgear GS105 switch (Breeze LPS, JSGT)
• TP-Link Archer C7 router (HDPlex LPS) > Synology DS916+ NAS (SMPS) 

The modem, router, and NAS, are in a different room, and on a different AC circuit. The Netgear switch is in my listening room, and immediately precedes the chains below. All my audio gear is on a dedicated AC circuit, using the 6 AWG wire and outlet sourced from Jim Weil, that I described in a post a few days ago.

 

Here is my current chain:

• sCLK-EX-modded switch (LPS-1) > sCLK-EX-modded sMS-200 (LPS-1) > ISO-Regen (LPS-1) > tX-USBultra (Hynes SR-4) > Codex
• Cybershaft OP-14 10 MHz reference clock feeding the sCLK-EX board in the tX-USBultra
• This clock topology essentially delivers OP-14 quality to my switch, sMS, and tX, and has really been the jewel in my audio crown, as it were.

SE Functional Quirks

 

As I previously posted, the SE's USB controller does not play nice with the USB controller on my Ayre Codex. So I was never able to run the simple chain:

• Zenith SE > Codex

In trouble shooting, I had several emails back and forth with Nuno of Innuos and Ariel and Ryan at Ayre. While this issue is still not resolved for the Codex, we did establish that it seems to be related to the fact that the Codex USB stage is bus-powered. Thanks to another loan of the QX-5 Twenty by my local dealer, which uses the exact same USB controller chipset as the Codex, but is internally powered, I established that this chain works fine:

• Zenith SE > QX-5 Twenty

Nuno did mention that another workaround is to use the iFi USB iPurifier2 between the SE and the Codex. Mark Sossa, the US Innuos distributor kindly lent me one to try. Sure enough, this does function correctly:

• Zenith SE > iPurifier2 > Codex

Usability

 

While the SE's UI is not geared to power users and tweakers like me, I have to appreciate its elegance and simplicity. For computer- or technology-phobes, this thing is a dream. One of the cool features is the CD ripping tool. This isn't something I expect to use much, as the bulk of my CDs were ripped more than a decade ago. Upon Roy's recommendation, I did try re-ripping a favorite CD, and the experience could not be nicer. Roy felt that the rips from the SE actually sounded a little better to him, but I didn't find that to be the case. Still, for those needing this feature, it's nifty.

 

Sonic Signatures

 

Before I get into comparisons, let me talk about the sonic signatures of both my trifecta and the SE. In the rest of the review, I'll refer to these signatures rather than enumerate them again and again. 

• Current chain based on SoTM Trifecta: As I've described before, the net effect of clock and PSU improvements - starting with the sCLK-EX, the use of LPS-1 and SR-4, and culminating in the addition of the Cybershaft OP-14 - has been in the following areas:
  • Tonal Smoothness, lack of digital glare
  • Sweet midrange
  • Huge and precise image
  • Air - excellent separation of instruments
  • Texture and dimensionality - instruments sound real, and have a palpable dimension to them
  • Resolving - fine details are exquisitely rendered. Brushes and high hat strokes come across very distinctly. A string section sounds like distinct violins (for example) playing together, rather than one homogenous violin, if that makes any sense.
  • As such, this configuration sets a very high bar to surpass!
     
• Zenith SE: the addition of the SE to a chain yields substantial SQ benefits. These are completely consistent with Roy's and Darko's observations, but I'll describe them in my words:
  • Dynamism: it feels like the music has gained some dynamic range, but it's more that dynamic range is is more realistically rendered. This is actually more apparent if you turn down the volume when making comparisons. 
  • Calmness and ease: related to the above, there is a sense of relaxation and ease to the sound, that I attribute to a lower noise floor
  • Richer and smoother tonality, especially the treble
  • Density, solidity: music had more weight and body to it. Instruments seemed more palpable and real.
  • Improved bass

Findings

 

Let me first dispense with the most obvious and financially most attractive comparison: could the Zenith SE standalone replace my spaghetti chain? If so, I could sell my entire trifecta, and recoup most or all of what I paid for the SE. Sadly - the answer is an emphatic NO. While the SE bettered the trifecta in terms of the traits described above, especially dynamism and weight, there was also a profound sense of loss for the air, dimensionality, and resolution that is so characteristic of all the clock improvements. This was not surprising, nor unexpected, as it's what Roy found in his evaluation too.

 

This would be a good time to take a detour to the topic of axes (or dimensions) of optimization. We have ample evidence now that SQ improvements can be gained by making extreme optimization on both the clocking axis AND the PSU axis. There are other axes too: for example, RF/EMI shielding, and mechanical/vibration control. If you've read about the SE, you know that one of the key things that distinguishes the SE from its lower-priced brethren is the use of very high-quality, bespoke PSUs designed by Sean Jacobs, another British boutique electronics designer. First Paul Hynes, now Sean Jacobs. Must be something in the water there! Or the whisky.  

 

The SE and the trifecta chains are perfect examples of systems that have been optimized along these two axes. My trifecta chain is highly clock-optimized, but while I am using excellent PSUs, it could probably sound way better if I used 4+ rails of an SR-7 instead. In contrast, the SE is highly PSU-optimized, but by Nuno's admission in discussion with Roy and me, not particularly optimized on the clock dimension.

 

With this backdrop, I went on a similar path as Roy described in his head-fi posts: could I combine the SE with elements of my trifecta to achieve a new high-water mark for my system's SQ? And what is the cost/benefit tradeoff of the resulting chain?

 

Note: this post is long enough already, so I will not lengthen it by enumerating all the interesting experiments I tried. At different points in the last 3 weeks, I've had loaners of the Ayre QX-5 and Eric's sMS-200ultra and dX-USB HD Ultra, that have allowed me to try all kinds of permutations and combinations. I'll cut to the chase and list only the most interesting findings.

 

Disclaimer: Keep in mind that all the adjectives and qualifiers I use below are relative. Take it as given that each configuration by itself sounds incredible, and without the other to compare, I would be very pleased with the SQ I was hearing. But for comparisons, it becomes necessary to use positive or negative terms to convey relative merits.

 

Finding 1: The best Zenith SE chain outperforms my current reference chain

 

This is the SE-based chain that sounded best in my system:

• sCLK-EX-modded switch (LPS-1) > Zenith SE > tX-USBultra (Hynes SR-4) > Codex
• Cybershaft OP-14 10 MHz reference clock feeding the sCLK-EX board in the tX-USBultra

You'll notice that this is still "spaghetti" but not as complex as my current chain. Why the switch, you ask, since the SE can store music locally? See the next finding.

 

This is the same combination that Roy has settled on in his system too. As good as the SE sounds by itself, the addition of the tX-USBultra, reference clocked by the Cybershaft, elevates SQ to an absolutely magical level. My system has never sounded this good! This combination takes the dynamism, the density, and the rich and smooth tonality of the SE, and adds in the resolution, micro-details, and texture of the clocking enhancements. 

 

If I'm honest, I still find my current reference chain to be a tad more detailed and resolving. This is where system synergy comes in. As I've mentioned in my past posts, I have always struggled with a slight thinness to the sound of my chain. This starts with my headphone setup, which is extremely resolving. Part of this is just inherent in headphone listening compared to speakers. Add to that my HD800's which, while they've been modded to mitigate the infamous 6kHz peak, are still unforgiving of any treble harshness or imbalance.

 

Compared to the SE chain, my current chain sounds a little constricted, and seems to have more treble energy. Switching to the SE chain delivers that "aahh" - a sense of calm and relaxation that one gets from reduced treble glare and hash. The sound is more dynamic, weighty, and richer. More importantly, voices sound more natural.

 

Having said all that, it isn't by any means a night and day difference. On some well-recorded and smooth-sounding tracks, it's harder to tell the 2 chains apart. The more inherently brighter the tonal palette of the recording, the more the difference is noticeable. To illustrate this - to myself, during listening - I fired up Kate Bush's Wuthering Heights. I know - I have dated musical tastes, but this is a song that will have you ripping the 'phones off your head on a sub-par system.  In this comparison, the constriction and brightness of the trifecta relative to the SE was immediately noticeable.

 

Ultimately though, the SE-based chain extends the SQ of my system to a new high-watermark. Is it worth the price? That is highly subjective. For my situation and my ears, the answer is yes, although I still have a few days left in the 30-day return window. I'll talk about this more in the summary.

 

One final note: Knowing the lust and envy that would result, I did also listen to the SE chain driving the QX-5. OMFG! This was a sublime chain. I know in the past I've said that I find DAC SQ differences to be rather small, compared to what I have heard in the digital chain upstream of the DAC. With the SE in the chain, it felt like the DAC difference between the Codex and the QX-5 was more pronounced. I'll talk about this more in finding 4.

 

Finding 2: The reclocked switch minimizes the gap between music on local SSD vs. NAS

 

Since I already have an sCLK-EX-modded switch available, along with an available clock output on my tX-USBultra, it seemed a no-brainer to try this in the chain. Without the reclocked switch, the difference between music stored locally on the SE sounded significantly better than when retrieved from my NAS on the network. Adding in the reclocked switch, the gap narrowed considerably. I have to listen very carefully now to register the slight improvement provided by local files. And of course, the reclocked switch makes a huge improvement for Tidal and other streams.

 

This is really important to me for flexibility. My music lives on my NAS for multiple reasons - it is accessible from multiple systems and devices, and I have a backup strategy in place. Putting music files locally on the SE would then require me to keep my SE copy in sync with my NAS copy. It's not an insurmountable problem, but it is a bit messy. Nuno has indicated they are looking into proving a sync tool within the SE to address this use case. I hope they do.

 

Finding 3: The Zenith SE internal player outperforms Roon

 

This is one of those ironic findings. On the one hand, kudos to the Innuos team for optimizing their internal SqueezeServer player. But as a committed Roon user, this finding actually annoyed me. I have no interest in switching back from Roon to a Squeezebox ecosystem. I still have iPeng on my iDevices from when I used to own a Squeezebox Touch, but I've moved on.

 

So yes - the internal player, controlled by iPeng, does sound better than Roon Core on the SE. From what I understand, much of the improvement comes from allocating a generous memory cache. To be fair to Innuos, they were able to make these improvements since the codebase is open source (I think), while the Roon code is opaque to them.

 

Still, as a Roon user, I would like to see this gap eliminated, and I hope Innuos devotes some energy to working with Roon to achieve that.

 

Finding 4: Improving music player (Zenith SE) has bigger impact than improving DAC (QX-5)

 

One point Darko made in his review about the "Portuguese fella" (the SE - I love his peculiar phraseology sometimes!) is this:

 

For the more seasoned digital audiophile caught between a streaming rock and a D/A converting hard place, know this: starting over with ten grand to spend, I’d take this Innuos and drop whatever cash I had left on a lesser DAC instead of putting the larger percentage of my dollars into a better D/A converter only to feed it ones and zeroes by a standard PC or Mac.

 

It occurred to me that with the loaner QX-5 sitting in my room, I actually had the makings of a very interesting experiment - i.e, compare these two configurations:

• Upgrade the DAC: replace the Codex with the QX-5 in my current chain
• Upgrade the music player: replace my current chain with the SE chain in finding 1, keeping my Codex DAC.

Cost wise, these are in the same ballpark, depending on whether you factor in retail vs. used cost, and the resale value of superseded gear. Not perfect, but an interesting comparison.

 

Well, I agree with Darko. Upgrading to the SE chain yielded more bang for the buck - to my ears - than upgrading the DAC. Now of course, upgrading both was to be transported to a divine plane, where you recline on a cloud, caressed by sunbeams, while angels sing.  

 

Finding 5: The tX-USBultra sounds better than the dX-USB HD Ultra

 

While my tX was en route to and from Korea, Eric kindly lent me his trifecta chain, comprising the sMS-200ultra and dX-USB HD Ultra. Since I had the QX-5 handy, I was able to drive this chain:

• Zenith SE > dX-USB HD Ultra (Hynes SR-4) > (AES) > QX-5
• Cybershaft OP-14 10 MHz reference clock feeding the sCLK-EX board in the dX-USB HD Ultra

The whole time I had it in my system I felt like something was missing, but with so much going on, I didn't really delve into why. When I got my tX-USBultra back, I immediately noticed how much better my system was sounding - and at a level of refinement that matched my recollection. Now curious, I compared the above to this chain:

• Zenith SE > tX-USBultra (Hynes SR-4) > (USB) > QX-5
• Cybershaft OP-14 10 MHz reference clock feeding the sCLK-EX board in the tX-USBultra.

Based on this experiment, the tX sounded significantly better. Now I'll caveat that I don't know per se what the relative merits of the USB and the AES input of the QX-5 are. Still, it is a data point, and suggests that the older dX may not be as refined as the tX.

 

Finding 6: The SR-4 has pulled ahead of the LPS-1 after more burn-in

 

I know I mentioned a couple weeks ago that I found these PSUs essentially head-to-head. Since then, the SR-4 has continued to burn in, and I am now able to compare on a chain of overall higher SQ with the SE. Whatever the reasons, I now find that switching back and forth powering the tX-USBultra, the SR-4 is the more refined PSU. Of course, this finding will be revisited when my LPS-1.2 comes in, and we shall see if the tables turn again with the 1.2. 

 

Summary

 

The Innuos Zenith MkII SE is a pretty remarkable piece of gear, and I recommend it heartily. Whether it's right for you depends on many things, but most importantly where you're coming from. If you're running your DAC direct from your stock MAC or PC, prepare to have your mind blown. As you go up the spectrum of existing system refinement, the SE's value proposition becomes more and more nuanced.

 

For my system, the SE does three things:

1. It allows me to achieve a level of SQ beyond what I previously had
2. It allows me to simplify my digital audio chain, although it does not completely eliminate the spaghetti
3. It allows me to sell parts of my current chain to partially defray the cost. I will soon be putting my modded sMS-200, my ISO-Regen, and 2-3 LPS-1s up for sale

There are some cons too:

1. For Roon users, the fact that Roon Core SQ lags the internal player is annoying
2. The fussiness of the USB controller with certain DACs requires care with system matching

Final Thoughts

 

Some people will look at this review and ask if I now believe "straight USB" is superior to "endpoint-based." I have never been tied to one approach or the other. For me, it was always about SQ and simplicity. Roy said something similar to me in one of his emails:

 

If you look back at my posts on CA, while I did find SQ to be a bit better with my "straight USB setup," a key reason for switching was because I preferred the simpler setup.  It is interesting just how many people have misinterpreted my rationale for leaving the trifecta, it was never for SQ reasons.

 

It did occur to me that an alternate path to achieve the same improvement as the SE provided in my system, was by the purchase of 4 rails of SR-7. Of course, I'd have to wait a year to get one! I discussed this with Roy, and his take was while the SR-7 still seemed like it had better specs:

 

With the Zenith SE more broken in, it has become a monster.  If I am to be honest, dynamics are a bit better than my server driven by the SR7 but I suspect this is due to the long DC leads I am using (1.5m) and the high impedance barrel connectors used (Paul Hynes hates these).

 

What I have concluded from my experience with the SE is that extremely high-end PSUs do make a palpable difference. This goes beyond the table stakes of low ripple noise, linear supplies, low output impedance, and isolation from high- and low-impedance noise. What these extreme PSUs also provide is a deep reservoir of instantaneous transient current delivery. We know this matters in analog amplification, but who knew it was so important in digital transmission!

 

What I would love to see is an SE-like product that married the concepts of extreme system and PSU optimization with clock optimizations. As Roy's experience shows, his (and others' on this thread) custom sCLK-EX modded servers have the potential for doing this. The trick is to incorporate several rails of an SR-7-like supply into the box, while keeping DC power leads as short and as low impedance as possible. There's the rub.

 

I sure hope Innuos's (or SOtM's or Antipodes') future systems give us the best of both worlds.

Thank for this very interesting and informative post. 

The zenith SE is definitely an extremely well designed équipement. 

 

i am in a dual pc set up ( super micro x10 sba ) running Roon ( Linux )and pièces of the trifecta : tx USB exp ultra (pcie) > tx USB ultra > dx USB hd ultra (aes) > DAC AMR do 777 se.

i have tried at home both SMS 200 and SMS 200 ultra ( player/renderer )to replace one of my Supermicro board and the result is very clear , the super micro mobo easily outperform SMS 200 ultra and of course also SMS 200. 

The strength of zenith is to use this Supermicro mobo which is very good with a very good psu. 

 

I agree with you that the tx USB ultra is a really nice add up to a system , when I put it in my system I was very happy with what it did , difficult to explain but it made music listening much more satisfying. 

 

I do not have yet a ref clock as I am waiting for the sotm one to come out. 

 

For peace of mind and easiness the zenith se / txUSB ultra is the way to go. 

For the adventurous and diy the way I am going is fun and full of improvements path ( i am in the process to have my mobo clocks upgraded by sotm ) but as costly as the Zenith. 

 

 

[jean-michel6]

I do have a dual pc set up with x10 sba board. 

I am using only the 12v connector to power them. 

All the accessories have their own dedicated psu. 

I have been able to power the board with simple 12v 1.8A IFI power psu. 

The results are already very good. 

You want really to avoid the pico psi which is so noisy. 

I used to have a gigabyte set up with 4 battery psu 2 for the CPU and 2 for the pico psu. 

Results were far away for the X10 sba. 

I will later power them with sotm sps 500

[jean-michel6]

On 02/02/2018 at 11:33 PM, rickca said:

I'd love to understand why the Supermicro board sounds so much better than the Gigabyte.  You are powering the two boards the same way, right?

 

On 02/02/2018 at 11:43 PM, Lebouwsky said:

Can you be a bit more specific about how you power the mb and “accessoires”,  I don’t quite understand. Do you mean you power the the entire mb with cpu and drive with just the 12v dc input?

There has been already several posts to talk about this super micro board. One key point I think is that the board is low electrical consumption and this equates to less electrical noise , less emi-rfi , this is a good thing for sq. 

The processor on the gigabyte was also a celeron similar to the j1900 on the Supermicro but tdp is 53w vs 9w on supermicro...

The processor is soc which is also a good thing for sq. 

 

Also from other forums , I read several time that Intel and Supermicro boards designed mainly for professional world usually have better sq , may be better components ?

 

It can run only on the 12v connector , you can do away with the noisy DC-DC pico psu 

Since it is not power hungry less than 1,5 A for the board itself and the CPU , you have a great choice of psu. 

M'as a temporary set up i am using the ifi power with already very good sq results. 

 

In my 2 pc set up accessories are 4 ssd's , on txUSB exp . They each have their own dedicated psu : a sps-500 for the txUSB and a diy using lt3045 reg's. Therefore those accessories do not draw any current from the mobo. 

[jean-michel6]

7 hours ago, ElviaCaprice said:

Personally, I find the renderer concept a dead end as an added unnecessary component of complexity and possible generation of noise, let alone cost.  Everything can be done at the server, with great care and design taken place.  The server can do anything a renderer can do and so much more, especially on the software end.   The future is in the server design.

 

Really keeping it simple!

sCLK-EX server=>DAC=>Speakers

The one board server-renderer approach is the route you have taken with great results. 

However my experience is that you can go even further in sq by using i agree  a quite complex 2 pc approach one for server and one for renderer . This is more complex but sonically superior. 

I have to specify I am using roon core and roon bridge on light Linux os.

Roon is very clearly saying that a separate  server and renderer/player approach is the prefered one to get the best sq from their software. 

 

 

 

 

 

 

[jean-michel6]

8 hours ago, vortecjr said:

people reading this should know that if they use a Supermicro board for a Roon server that it will be short on performance. Roon recommends:

• Intel Core i3, Ivy Bridge+
• 4GB RAM
• SSD boot drive
• 1440 x 900 Resolution

I can tell you from speaking with Andrew that he recommends an i5 at minimum and prefers an i7. The difference being that you want an i7 with a large music collection and with DSP. His i7 is only 11 watts so nothing to obsess over regarding noise. Professional use could mean better parts or longer runs, or any number of things.

 

As for the noisy DC-DC converter there is no free lunch you now have those right on the board. And if you use the 4 pin power source input and an ATX power source you will have 2x the number of noisy DC-DC converter:) 

 

Also, you said that the tX-USBexp does not draw current form the motherboard. You might want to double check this with SOtM. I asked them about this a long time ago, but I can't remember with certainty what they said. 

The Zenith SE is using this board both for server / player  and is running Roon According to several reviews their owners have excellent results from sq perspective. 

As long as you don't do heavy DSD over sampling (128,256 ). The super micro board has no problem. There is an indicator of CPU usage in roon which shows you how intensive is your CPU usage. 

 

Whatever the powering method ATX or 4 pins you are faced with DC-DC converter unless you have an excellent  linear ATX psu ...which are very expensive and hard to find .

With the 4 pins and a very good 12v psu , you have a very good solution as the dc-dc converter on the board are perfectly sized for the function they should power , additionally you can stop in the bios most of the unused hardware. 

With ATX you have to go through the very noisy pico psu. 

 

Romaz i think has done a test with an other asrock board who could be powered either from 4 pin or ATX and they have found the 4 pin to be always superior in term of sq. 

 

As for the txUSB exp pcie card , it has its own dedicated psu. Sotm recommend as a safety to also power it with the internal flat 4 pin connector found in pc's. 

This is a safety measure to protect the card who could get damaged if the computer is running and the card has no power. 

Since i have found this to be sonically inferior , I am using only the external power. 

 

[jean-michel6]

3 hours ago, seeteeyou said:

 

 

They're also making many server boards for Xeon processors, this particular Micro ATX motherboard caught my attention because all 48 PCI Express lanes are fully utilized

 

https://www.supermicro.com/products/motherboard/Xeon/C420/X11SRM-VF.cfm

https://www.supermicro.com/manuals/motherboard/C420/MNL-1967.pdf#page=18

 

The cheapest Xeon W-2123 would cost 300 bucks or so

 

https://ark.intel.com/products/125036/Intel-Xeon-W-2123-Processor-8_25M-Cache-3_60-GHz

 

Of course we've gotta use ThrottleStop to lower speed / voltage etc. because of its 120W TDP

 

https://beebom.com/how-use-throttlestop-control-cpu/

https://www.techpowerup.com/download/techpowerup-throttlestop/

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/

https://www.notebookcheck.net/How-to-Lower-Temperatures-Stop-Throttling-and-Increase-Battery-Life-The-ThrottleStop-Guide-2017.213140.0.html

 

---

 

What is OCuLink?

http://blog.fosketts.net/2017/06/22/what-is-oculink/

 

OCuLink (just ignore their typo) also seemed to be mighty interesting to me since they're directly connected to the CPU. That means we should be able to get one of these cables from Supermicro

 

http://store.supermicro.com/55cm-oculink-sff-8639-u2-cbl-sast-0956.html

http://store.supermicro.com/75cm-oculink-sff-8639-cbl-sast-1011.html

 

Then convert each OCuLink into PCI Express x4 slot that's far away from the motherboard

 

http://www.microsatacables.com/u2-sff8639-to-pcie-4-lane-adapter-sff-993-u2-4l

 

 

Granted they're meant for those storage devices such as PCI Express SSDs in the AIC (Add-in Card) form factor, though stuff like JCAT NET Card FEMTO, SOtM tX-USBexp, or professional sound cards (e.g. Lynx Studio and RME etc.) might also work.

 

If that were no go, maybe we could also take a look at another choice that's going from OCuLink to SFF-8643

 

http://www.cs-electronics.com/product/12g-internal-hd-mini-sas-sff-8643-oculink/

http://www.microsatacables.com/mini-sas-hd-sff-8643-to-pci-e-4-lanes-slot-adapter

 

If either combo were working, then we should be able to save ourselves from these troubles that might happen in some cases

 

PCIe Extension Ribbon Side-Effects
http://www.tomshardware.com/answers/id-1779164/pcie-extension-ribbon-side-effects.html

 

Riser cable for SOtM - defect :-(
https://www.computeraudiophile.com/forums/topic/31499-riser-cable-for-sotm-defect/

 

Question: USB card installations and risers, ribbons etc

https://www.computeraudiophile.com/forums/topic/31662-question-usb-card-installations-and-risers-ribbons-etc/

Thank's for sharing all the good info. You are an encyclopedia about computer audio ?.

[jean-michel6]

46 minutes ago, tims said:

Hi @jean-michel6

 

I also want to feed my tx-USBexp card with an external psu but I will have to open up my SOtM server to disconnect the tx-USBexp cards internal  supply as you have done.  This may invalidate my warranty doing this so do you recall if the SQ was markedly better when you disconnected the internal supply of the tx-USBexp card?

 

Thanks

Hi Tims 

 

Yes the sq gain is significant if you already have a resolving system. 

I have two friends who did the same and experienced a nice sq boost. 

The sotm pcie card requires 6,5 to 9 volts external supply and if you do not use the USB 5 volt , 1A or less is fine. 

However you MUST follow this powering sequence :

- start : power to sotm pcie then power on to mobo

- stop : power off mobo then power off the sotm pcie 

 

Me and my friends have been running like that for months with no issue. 

[jean-michel6]

4 hours ago, vortecjr said:

I’ll ask Roon and Andrew to comment and report back.

 

Your assuming the on board Dc to DC converters are better than the ATX source. Why assume when there are solutions with on board liner regulation. 

 

Not it sure how many boards can use a dual source you have to check the manual to be sure the 4 pin connector is not for the processor. Regardless you will have common grounds. 

 

No one ever powers both on the card. 

I am listening right now , SQ  and emotion is ??

[jean-michel6]

8 hours ago, tims said:

Great, thanks for the information. 

I have a PH SR7 on order that will eventually supply the mobo and the txUSB with separate feeds.   I wonder if I disconnect the internal supply from the txUSB (as you have done) and power up/power down the SR7 to turn the server on and off that should be OK?

I will talk to May from sotm about it.

 

 

I think it should be ok but it is better to ask May. 

 

I have written in the past to them and they do not recommend to disconnect the internal power supply because it is a safety. 

However using only the external power and the right sequence of powering and shutdown is fine with a nice sq boost. 

[jean-michel6]

10 hours ago, Dev said:

 

I am not so sure if this is totally true. Over at Jplay, most has been using a dual jplay setup for a while (long before microRendu or the likes came out) - the control PC, which is equivalent to the music server, is also very important in contributing to the SQ as is the audio PC, which is equivalent of a streamer, like mR. If you have to optimize one, the one closer to the DAC must be done first but a noisy music server can also pollute the SQ. Thoughts ?

Fully agree with you . 

In my dual pc system ( Roon - Linux) , everything that is done on the PC server side can be heard  ( better PSU , shielding RAM with 3M emi paper , changing sata cable for better one  and the list goes on .... )

[jean-michel6]

22 hours ago, Johnseye said:

 

Two identical servers where one was the core and the other the endpoint?  If that's what you're asking, no.  I wouldn't invest in duplicate hardware to test that.  I did test a server (without modifications, tX-USBexp, sCLK-EX, tX-USBultra) and endpoint (microRendu & sMS-200) against the same exact one (with modifications, tX-USBexp, sCLK-EX, tX-USBultra) but without endpoint and it was without a doubt, no question a significant improvement in SQ.  I still have the sMS-200 and adding it instead of the USB output, but still gaining the sCLK-EX utilization on the motherboard system and NIC clocks does not sound as good as with the tX-USBexp and tX-USBultra.

 

With so many variables and possible combinations it tough to test them all.

Which modifications is it ? 

The server clocks driven by the sCLK ex ?

[jean-michel6]

21 hours ago, Dev said:

Anybody getting the Supermicro (one used in the Zenith) mobo modded from Sotm yet ?

I should have that in a few weeks !

[jean-michel6]

6 hours ago, rickca said:

If the computer is running but the tX-USBexp has no power, why would the card get damaged?  I'll ask SOtM about why it doesn't use PCIe slot power.

 

@jean-michel6 since you brought this up, can you please explain?

Ricka

 

I do not really have an explanation. 

It is sotm which says that if you use external power for your pcie txUSB card it is safer to power it also the moles I ternal power connector. 

They say that if the computer is powered with no power to the usb pcie card it may get damaged. 

Myself i use only the external power as I have found that the use of the internal molex power degrade the sound. 

 

If if you need further info only sotm can answer.  

[jean-michel6]

16 hours ago, str-1 said:

My apologies if this question has been addressed in earlier posts.  I’ve followed one or two promising links from the index but couldn’t find what I was looking for.  When using a USB regenerator like the tx-USBultra or ISO Regen, are the USB cables before and after of equal importance in terms of their impact on sound quality, or would the payback on investing in better quality be greater with the cable after the regenerator than with the one before it (or vice versa)?

I have this set up and by far the last cable between txUSB ultra and DAC is the most important.