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

[Johnseye]

12 minutes ago, lmitche said:

I'm running wifi now. But with the EtherRegen one server nic will be in one port and the NUC in the isolated port. All else empty.

 

 

6 minutes ago, elan120 said:

Below (red dotted lines) is how my network setup looks like at the moment, which I believe is what was described, and is how I plan to use to test when EtherRegen is available.

 

The limitation to the setup above is the required 3 Ethernet ports, which in order to make it fit for server with just two Ethernet ports, a switch before the server will be needed.  For those that use external HDD connected to server via USB, above setup will work fine since NAS connection will not be there, makes a two-port bridging connection as below.

 

 

Larry I'm guessing you're doing something similar to elan120's first drawing.  I don't follow your statement "The idea of the bridge between the server and endpoint is to stop noise generated by the server from reaching the dac".  If you've got a switch between the two it's not a bridge unless what you're doing is different from elan120's drawing.  Unless you're calling your switch a bridge which could be why I'm confused.

 

Looks like you guys are bridging the server but still including a switch between the server and NUC.  Your belief is that the switch provides some kind of filter or isolation between the server and NUC?  And you're not using the switch upstream of the server so you can further isolate the NUC from any noise that's on that upstream switch?

 

[Johnseye]

6 minutes ago, lmitche said:

Study Elan120's bottom picture posted above. If a bridge equals a switch with two endpoints there are three bridges in a line.

 

5 minutes ago, austinpop said:

 

Bridging is just another name for implementing a virtual (software) switch on the server, using 2 (or more) Ethernet NICs. So the key thing to remember is that a bridged machine has more than one Ethernet interfaces.

1. One NIC is connected to the router.
2. The other NIC is connected to the switch (TLS in my case)
3. The endpoint is connected to the switch

Please see @elan120's nice diagram posted above.

 

Guys we need to be careful how we interchangeably use terms. While switches can act like bridges, bridges are not switches. There are also layer 2 and 3 switches which provide different functionality. We should call a bridge a bridge and a switch a switch or we're going to confuse a lot of people. A PC or server can act as a bridge but not a switch. 

[Johnseye]

1 minute ago, lmitche said:

I give up.

 

A for effort. Must be some common east coast slang

 

[Johnseye]

This is my bridge tonight. Brahms 4th and Strauss at Symphony Center with an amazing guest orchestra. 

 

 

[Johnseye]

1 hour ago, austinpop said:

 

John, let's not go down a rabbit hole here. We're talking about a facility (EDIT - a multiport facility) either in software (OS bridge) or hardware (switch) that intelligently allows Layer 2 frames to traverse through them.

 

And a layer 3 "switch" is a router. No confusion there.

 

I just think we should call a bridge a bridge, a switch a switch and a router a router. I deal with buying, configuring and supporting this stuff every day. If I called a layer 3 switch a router I'd confuse a lot of people. Granted this isn't work but I've been confused what people have been talking about here because there is a difference between this stuff. 

 

Your thread though and I'll just go sit in my orchestra seat now and keep quiet. 🤫

 

https://learningnetwork.cisco.com/thread/10642

 

https://learningnetwork.cisco.com/thread/41392

 

[Johnseye]

7 hours ago, rickca said:

Why upgrade the HDMI clock?  What do you use HDMI for?  Why not do the system clock along with USB/ethernet?

 

5 hours ago, austinpop said:

 

Since we don’t really know why the clock mods have an S.Q. benefit - at least I certainly don’t - the premise was to improve ALL clocks on the mobo, since they all share a ground plane. To the extent a low phase noise clock lowers noise, why leave one noisy clock in play.

 

Lee at SOtM always recommends changing out every clock on the board. He believes every clock contributes noise.  To Rajiv's point regarding the ground plane.  IMO if you've got the taps, or the cash, why not.

 

@austinpop were you able to compare an unmodified i7DNBE directly against your modified i7DNBE?

Did they drill holes in your Plato X7D case for the smb connectors?

[Johnseye]

I received some feedback from @romaz as he also sent his NUC into SOtM for a clock mod.  I believe he will write something up in the future.  I'm just going to quote him here so I don't mince words.  For those of us who have modded our PCs or use any of the SOtM ultra devices, we know the impact an sCLK-EX can have on a device.  It appears that is the same with the NUC.

 

"The improvement is HUGE.  The impact of the EVOX caps on the tX-USBultra is much bigger than anticipated.  Dynamics are much improved.  Transients have more bite and the detail articulation has taken a surprising leap.  The tX-USBultra is back in the chain for good and it will only get better once I am able to power it with a DR SR7.

 

The 4 clock replacements on the NUC (system, Ethernet, 2X HDMI) improve SQ in an expected and welcome fashion.  Noise floor is lower resulting in better perceived resolution, space, clarity, timbre along with a deeper and richer tonal bouquet.  Bass heft was already really good with the i7 NUC but bass definition is at a level I've not heard before.  What is surprising is how wide my sound stage has become. 

 

Before, the REF10 made a decent improvement with the tX-USBultra but now with the combo of NUC + tX-USBultra being clocked by the same sCLK-EX board, the improvement with the REF10 is just much more prominent.  Along with master clocking my sNH-10G, I'm really feeling like my REF10 has become a must have."

[Johnseye]

8 minutes ago, rickca said:

I'm confused by a couple of things @romaz said.

 

He says 4 clock replacements on the NUC (system, Ethernet, 2X HDMI).  But @austinpop said he did 3 clock replacements ... USB, ethernet and HDMI.  I think Roy and Rajiv are using the same NUC.  Rajiv, why did you do HDMI instead of the system clock?

 

Why didn't Roy replace the USB clock?  How can he have done 4 clock points if the NUC + tX-USBultra are being clocked by the same sCLK-EX board?  If the board is in the tX-USBultra, it only has 3 available clock points for other devices.

 

There are 4 clocks but you can use the same sCLK-EX tap for the HDMI ports.  Here's Roy's words again.  This was the same situation with my motherboard.  I used a single tap for both ethernet adapters.

 

"With respect to my i7 NUC board, which is an SoC board, there are 4 replaceable clocks (system, Ethernet, HDMI x 2).  All the other clocks are subclocks (dPLL) that cannot be replaced.  The 2 HDMI ports each utilize a clock of the same frequency and so 1 tap from the sCLK-EX board can be used for both and so this allows the 3 free clocks on my tX-USBultra to clock all 4 clocks on the board."

[Johnseye]

15 minutes ago, rickca said:

I'm confused by a couple of things @romaz said.

 

He says 4 clock replacements on the NUC (system, Ethernet, 2X HDMI).  But @austinpop said he did 3 clock replacements ... USB, ethernet and HDMI.  I think Roy and Rajiv are using the same NUC.  Rajiv, why did you do HDMI instead of the system clock?

 

Why didn't Roy replace the USB clock?  How can he have done 4 clock points if the NUC + tX-USBultra are being clocked by the same sCLK-EX board?  If the board is in the tX-USBultra, it only has 3 available clock points for other devices.

 

You also raise a good point as to why he didn't mod his USB clocks.  I'll check with him.  I know he also wanted to test out SQ over HDMI at some point in the future.

[Johnseye]

Just now, rickca said:

OK good answer, thanks.  I forgot you can daisy chain clocks of the same frequency.  

 

If all the other clocks are subclocks (dPLL) that cannot be replaced ... is the USB clock a subclock?  If so, how did @austinpop replace his USB clock with a tap from his tX-USBultra?

 

The USB clocks aren't subclocks afaik. 

[Johnseye]

2 hours ago, rickca said:

I wouldn't think so either.

 

Got the info from Roy.  I asked him why he didn't mod the USB clock.

 

"Because there is no USB clock to modify.  If you look at Intel's white papers on this i7 board, you'll note that the USB subsystem is on the diecast of the SoC and not separate.  Typical of SoC (System on Chip) boards, there is one system clock that governs all the other clocks (DPLL) on this chip and that one clock is the only clock that is replaceable.  Because board manufacturers like ASUS, Asrock, etc., want control of the types of video ports offered (HDMI, DisplayPort, DVI, VGA, etc), these specific ports are generally never integrated onto the SoC which is why they require separate clocks."

[Johnseye]

7 hours ago, austinpop said:

I was trying to keep from mentioning @romaz since he's staying off the forums, but since @Johnseye spilled the beans... 

 

Indeed Roy and I colluded and both sent our i7 NUCs and tX-USBultra to be modded at about the same time. It looks like John's already clarified the clock question. @rickca I did indeed misspeak. It's the system clock, not the USB clock, as Roy explained.

 

Copyrighted and approved by the author before posting on his behalf.

 

I don't think these results are surprising.  We went through this exercise with the sCLK in the past and I was waiting for confirmation with these NUCs before pulling the trigger myself.  I've had the sNH-10G on loan for a while now and while it is a slight improvement (with master clock using REF10) over the TLS switch, the cost of that switch is 3x the TLS.  Now that I can use the taps from the sCLK in that switch I have further justification for buying it and modding my NUC.  I'm still waiting for Uptone's switch with bated breath.

 

I also just upgraded my Persona 3F speakers to the 9H, and my projector to the new JVC 4k.  The speakers are a substantial, or in Roy's words HUGE, improvement and run at 96db which opens up some doors if I'd like to try some other stuff out.  Yes, I am gushing over these speakers.  It's my belief that speakers have the biggest impact on sound and these should be my endgame there.

[Johnseye]

2 hours ago, mourip said:

 

Sorry to be dim but...

... is the tX-USBultra  modified to provide an external clock output back to the NUC which is also modified to accept that?

 

Also has anyone compared the tX-USBultra to a Mutec 3+ USB/REF10 combo or is this apples to oranges?

 

12 minutes ago, tims said:

I believe the Ultra and NUC were both sent to sotm for modification to do the clock upgrades.

 

Correct, they were sent to SOtM for mods.  The sCLK-EX board which is included in the tX-USBultra has extra clock taps which can be leveraged for other devices.  There are internal cables run from the sCLK-EX to connectors mounted into the chassis of the tX-USBultra.  Then cables are run from those connectors to similar connectors mounted into the NUC case.  Cables run from each clock tap of the NUC board to those connectors mounted into the NUC.

[Johnseye]

43 minutes ago, mourip said:

 

Thanks. Any idea how much the mods cost?

 

The last I knew enabling each sCLK tap was $50 and each SMB connection was $70.  I've emailed May for pricing and @austinpop should know if that's current.

[Johnseye]

8 minutes ago, mikicasellas said:

Hello Gents,

 

Does "audio" in BIOS should be disabled?

 

Yes. That should disable your onboard audio chipset.

 

[Johnseye]

1 hour ago, tboooe said:

No truer words have ever been spoken....and this is precisely why I won't even consider dabbling in vinyl even though my daughter keeps trying to tempt me with her setup which does sound very good.  BUT. MUST. RESIST.

 

 

Vinyl's my benchmark for sound.  Nothing sounds  better than an AAA sourced album to me.

I've invested much more into my digital equipment in terms of time, effort and money.  It doesn't take that much to get a decent TT.

[Johnseye]

28 minutes ago, austinpop said:

 

It was a pleasure to visit Ray's amazing setup. See the picture he poster here: https://audiophilestyle.com/forums/topic/55723-article-my-quest-for-a-new-dac-part-5-chord-electronics-hugo-m-scaler-hugo-tt-2/?do=findComment&comment=931977

 

 

I don't often get to do this, but on the occasions where I've been able to meet an AS'er in person after interacting the forum, it's always been delightful. Ray was no exception, and was a very gracious host. The big Voxativ 9.87 with the Pi bass unit was awe-inspiring, and sounded way better than I remember it from RMAF. However, I was most taken by those Omegas, especially listening nearfield (3-4'). Once I get my HMS/TT2, this may be my next (non digital) upgrade.

 

Awesome.  I'd love to hear those speakers. I'm going to search em out at Axpona.  Steve Hoffman's hilarious.  Listen to him at the :56 mark.  "I like the sound of life".

 

 

[Johnseye]

4 hours ago, tboooe said:

Ok thank you.  I am very curious because I am also considering the X10 for it's SFP port that I would use with my NAS.   In my case, the NAS is upstairs and my music system is downstairs connected to a wireless bridge.  Since I am streaming the music wirelessly I don't know what kind of benefit I would get with the SFP for my NAS.

 

What NAS are you using that has a fiber port? Qnap?

How do you like it?

[Johnseye]

whoops.

 

 

[Johnseye]

On 12/3/2018 at 3:26 PM, austinpop said:

 

No, I'm serious. I'm paranoid about opening up someone else's device. The last time the SR-7 was in my setup, it started to suddenly blow fuses one after the other! Turned out, Paul was fine with us going up to a higher amperage, and things are fine since.

 

What fuse ended up working with the SR7?

[Johnseye]

On 2/25/2019 at 5:58 PM, austinpop said:

 

 

Since SR-7's are custom builds, I don't believe there is a standard rating. Eric's SR-7 has 2 DR rails, each 6A, rated 3-13V and 9-19V. Paul had sent it with a 3.15A fuse, I believe. On Paul's advice we ended up replacing it with a 5A fuse. I'm talking standard cheap fuses from Fry's.

 

As @auricgoldfinger says, 1.3x to 1.6x is standard for SR Blue fuses.

 

However, there is a part 2 to the story of blowing fuses in Eric's SR-7. We had another episode very recently, when we were adjusting the voltages of both rails, and again, even the 5A fuse blew out a couple times. We hypothesized there was a loose connection in one of the terminal blocks that was getting perturbed when using the little screw to adjust voltage. I went through and tightened every screw in every terminal block in the entire PSU (a LOT). After that, no blow outs. So it's possible the original 3.15A fuse was actually fine all along.

 

My SR7 and Eric's are nearly if not completely identical although one of my regs was rated at 3A.  I'm checking whether the 19v was 3A or 6A.  That's the one that failed and likely related to what Paul mentioned in my thread.  When he replaced it last July he did so with a 6A.  My question is why use a 3.15A fuse with a 6A supply?

 

 

[Johnseye]

18 minutes ago, austinpop said:

 

3.15A is at the input 120V AC end of the transformer. It’s way more power (VA) than 6A DC at the rail voltages.

 

Make sense?

 

Yup, just my lack of understanding of how fuses work in power supplies.  Thanks for clarifying.

[Johnseye]

4 hours ago, sgr said:

While I applaud the interest and information, I feel these threads get too long and disorganized as they continue. There needs to be a way to break it into categories or subtopics.  Not being critical as I don’t think there is a way to prevent this in the format we’ve invented. 

But it would be nice if each topic could be separated out under the broad umbrella then bridging, NUCs, etc 

 

And to address the point of the post you deleted...

 

The index of this thread is a blessing and not many posts have it.  You can also bookmark individual posts yourself which not all forums have.  If you consider the amount of time and effort involved in maintaining not only the content of the thread, but the people who visit it with their opinions and input, it's a huge time sink.  This isn't anyone's full time job.  Someone shared some of their personal findings which have influenced countless people's enjoyment of music.  That person didn't have time to tend the interest that blossomed and so @austinpop graciously donates his time.  I wouldn't want to do it and I only contribute when I can.

 

I realize you meant to try and help improve the thread.   I don't read any malice in your post and you tried to address the topic as delicately as possible.  Maybe Rajiv would allow someone like yourself to help improve things if you were willing to put the time in.  Not sure if it's even possible to share that authority.   You could always copy the index out, restructure it and submit it to him for an approval or suggestion.

[Johnseye]

3 minutes ago, davide256 said:

You can drive yourself crazy by trying to insist on textbook reasoning explanations vs having faith in the empirical experimentation results of those willing to put in sweat equity.

 

The main result I see from AL is that it shows the biggest problem we had been facing with computer audio was not the hardware but rather that the OS and player software we had been using created severe limits. How else to explain why testers can now  chuck  expensive proprietary HW implementations for server /endpoint devices and use off the shelf NUC's instead to achieve a much higher level of playback performance.

 

The AL RAM part matters  at endpoint.

 

The AL software is critical at the server... RAM loading maybe not so much.

 

If you forced me to use AL at one end only, I would choose the server.

 

So I use AL loaded into RAM for both server and endpoint.  I find that both matter, but more significantly at the endpoint.  I've used Windows as the server and AL for the endpoint.  My choice would be AL as the endpoint if I had to choose, but prefer both.

 

HQPlayer can also load an NAA into RAM.  I use HQPlayer and find Jussi's NAA close to AL but I prefer AL.

 

I think this newer line of NUCs has had the biggest impact because I can stream from Windows to Jussi's NAA and get a marked improvement.  AL just takes things further for the positive.

 

To each their own.  There are 31 flavors of course

[Johnseye]

Does anyone have experience with the NUC8I7BEH or its motherboard, the NUC8i7BEB?  It has an i7 proc and 28w TDP.  That's 13w more than the NUC7i7DNBE and while both will run at 12-19v it has a little more demand.  I read a few posts where that extra power did pose a challenge.

 

I'm currently powering my NUC7i7DNBE with 12v from my SR7.  It isn't perfect.  There are times when the HDMI won't display.  Then magically it does.  I can power it from my 19v rail but I'll then have to power my server with an SMPS until I get another LPSU.  I'm going to experiment a bit and see how much impact the SR7 has on the server vs an HDPlex SMPS.  I have heard upstream power's influence before so am not thrilled with this option.  Keep in mind, the server is behind the NUC and a switch in the chain.  I'm currently not using my tX-USBultra, but that will change when I get the caps upgraded.