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

[soares]

2 hours ago, elan120 said:

If this just the board you are purchasing, inventory and supplier choice could be more limited.  Expend search to both NUC7i7DNKE and NUC7i7DNHE kits will yield more price and supplier.

Thank you Dave!!!

Kind regards,

 

Jorge

[sunny_time_99]

11 hours ago, RickyV said:

Sorry @sunny_time_99 to drag you into this but you seem to know more about clocks.

 

Sorry if this is bit to much DIY

 

 

you don´t need to be sorry. 😉

 

i must apologise, that i dont know that much about clocks. what is know has more to do with practical doing - trial and error-

 

Phase noise is only one aspect. Its nice to measure - like Femto-Sec. Jitter - but thats not the point, what makes it a good clock. a better phase noise ratio brings not a big benefit, if you leave the clock were it is. there are a lot more things to do.

 

I did not buy/use the NewClassD Neutron Star because of the better phase noise! I use it because the clock is seperated from the host board. I dont want the noise from all the circuits in the host machine to be connected with the clock circuit. And as a result destroying the clock signal/quality. As is found out, it destroys the clock performance. The Neutron Star uses an isolator to eliminate ground loops.

 

The next thing is "femtosecond-clocks". Only a nice thing to measure and for a superior marketing-trick. it has nothing to do with the performace and the sonic quality of the clock module. Femto-Clocks are not better than the rest. Femtoseconds looks  great in a factsheet  but its not a fancy new technology for building clocks,  its more about the way jitter is measured.

 

The Neutrino/Neutron Star are just  vehicles. Lars uses normal Oscillators to get the wanted frequence. Its adjustible so the main neutrino/Neutron Star-Board is allways the same. That makes the clock module so flexible. I changed the oscillator by myself.

 

The phase-noise of the small neutrino is mentioned here:

http://newclassd.com/index.php?page=152&hv=1

 

But the Neutrino is only the small version. If i compare the sonic impact - Neutron Star is the winner. The difference is realy big - even if the neutrino is using an isolator too. I try the newclassd-psu, that comes with the clock. Even when Lars can bulid that ultra precise Clocks, the UWB2 Regulator, and the wonderful sounding Singularity Amplifier, the psu for the Neutron Star did not even come close to my own Power Supply - even if is use his regulators 🙂

 

Imho if you try to use an external clock, one big benefit is the seperate power supply of the clock, the isolation of the clock and the host board. @RickyVif you use a crystek instead of the existing clock, it will not be a big step. i think you will not be able to hear a difference. You must take out that clock you want to use and power it with a seperate ultra clean power supply and isolate that circuit from the rest to prevent ground loops. that will make a difference. You must keep wires short. The Clock-Signal is very sensitive. 3-4cm maximum diffgerence between neutron star and host board with the active output.

The sonic benefit of a clock mode inside of a computer/nuc is not that big than a modification of a usb/i2s interface. I did that a long time ago. Nice to see that this thread comes closer 😉

 

On the pictures a small part of my audio setup. A CM 6631A usb/i2s Converter from diyinhk.com together with a ideon audio 3R - both modified with Newtron Star 2 Clocks. That is a big step!

I try to modify the board also with crystek 957, becourse in my mind... 4 years ago, i liked the performance of the 957 oscillator clocks. but that was another place and time. if you can compare the sonic quality of these clocks with other, it sounds boaring today. I am happy with my smd-microscope that makes all that possible.

 

regards

 

Sunny

 

[luisma]

Hi everyone, is this a dual JSST cable? 

 

https://audiobacon.net/2019/03/20/its-here-jcats-flagship-usb-and-usblan-conditioner/

 

[LTG2010]

On 3/20/2019 at 8:56 AM, RickyV said:

I have ideas of putting a neutron star on my nuc7i7 in the same case.

That would be the best, I agree with @sunny_time_99 the entire circuit is important especially the power supply.

I think you could use a single Neutron star clock board @25Mhz for the lan and system/usb 24/25mhz, I believe this is the way its done with the DS1 streamer, a single OCXO board.

Heres an alternative to the Neutron star and there is a fitting service if of interest, there are also some nice regulator and power supply designs.

https://www.fidelityaudio.co.uk/c4-low-jitter-uber-clock-fitted-3989-p.asp

[RickyV]

29 minutes ago, LTG2010 said:

That would be the best, I agree with @sunny_time_99 the entire circuit is important especially the power supply.

I think you could use a single Neutron star clock board @25Mhz for the lan and system/usb 24/25mhz, I believe this is the way its done with the DS1 streamer, a single OCXO board.

Heres an alternative to the Neutron star and there is a fitting service if of interest, there are also some nice regulator and power supply designs.

https://www.fidelityaudio.co.uk/c4-low-jitter-uber-clock-fitted-3989-p.asp

 

The neutron star 2 has two outputs and you can use them at the same time with coax output being the best, I asked Lars when I modded the iso regen. If the CPU clock is 24Mhz I can't supply it with 25Mhz, can I???

The fidelityaudio clocks don't have output transformers.

I am wondering if the neutrino clock would be an improvement over standard? Hot darn it, but I want the best 😅

[LTG2010]

12 minutes ago, RickyV said:

If the CPU clock is 24Mhz I can't supply it with 25Mhz, can I???

Yes, it depends on the chipset if its compatible, I think thats how its done on the DS1 and that uses a NUC7PJYH but is available with your board as an option. I think @austinpop would know more as he has/had one.

The neutrino should be a big improvement over the standard clocks but if the Neutron star has 2 outputs your good to go

Also Pink Faun have an OCXO clock board that might work for you.

[sunny_time_99]

Hi @RickyV,

it depends. The coax output is only better if you cant mount it very close to the host. the neutrino is an improvement due to the seperate power supply and the isolator but it didn´t come with the temperature regulation like the neutron star got. how many other clock-boards do you know that got that feature??

@LTG2010the fidelityaudio clock would not be my choice, becouse it got no isolator and the rest looks easy made for me. (it got no temperature regulation also) some polymere caps did not make me to click on the "buy "button. what is really nice is a fitting service... just wondering how that thing could weigh 10kg!

 

you can compare the neutrino sq with a sellarz audio

https://www.ebay.de/itm/R-Clock-High-Precision-Low-Jitter-Reference-Clock-with-Ultra-Low-Noise-Regulator/230871603430?hash=item35c10518e6:g:EJAAAOxyyUtScgHh

 

for example. It got a seperate supply and a isolator. neutrino and sellarz are worth what you pay for it. i prefer the neutron stars active input. if you use it to replace a cmos-clock, it will work without adjustment. sometimes when you are using the coax output it can be tricky, and sometimes you need to adjust the signal with a small trimmer. so without the possibillity of measurement impossible to do.

 

regards

 

Sunny

 

[austinpop]

Custom Roon Server - Initial Report

 

As I've described here in the past, while I have upgraded my Roon endpoint to near-endgame with the sCLK-EX-modified NUC7i7DNBE/Plato X7D running AL in RAM, my server upgrade has been pending. In the meantime, I've managed to squeeze (pun intended) out a lot of performance from my generic Dell XPS 8700 desktop. However, it was always my plan to upgrade my server, following shamelessly in the footsteps of @romaz.

 

Roy has been refining his server build, and while he is still tweaking, I acquired his first build to try in my system. I am just going to enumerate the system here, because most of the credit goes to Roy for the thought and effort he put into this. Here are the system details:

• HDPlex H3 chassis
• Asrock H370M-ITX/AC mobo
• Intel 8700T CPU
• HyperX Fury 2400MHz DDR4 8GB
• JCAT NET Card Femto
• HDPLEX 400W ATX Linear Power Supply
  • 19V rail modded to 12V to supply independent EPS input to CPU
  • Ghent JSSG360 ATX-to-ATX and XLR-to-EPS cables

While Roy and I discussed these options at length, it's really he who drove this system selection, and the end result is truly special. Before I describe the sound, here is a brief rationale for these choices. 

• The 8700T CPU has 6 cores/12 threads/12MB Smart Cache, all for a TDP of 35W at a base frequency of 2.4GHz. This seemed to be a good balance of low power and hefty horsepower. 
• The HDPlex 400W Linear PSU has independent rails for the ATX voltages (3.3, 5, and 12V), with no DC-ATX conversion. Plus, by modifying the 19V rail to 12V, this adds another rail solely to power the CPU via the EPS input.
• TDP 35W is well within the capability of the H3 fanless chassis, and we snake the cables into the chassis through a PCIe slot opening on the back. Not elegant, but effective.
• The JCAT Net card was chosen to provide a high quality bridging solution, and as you'll read, boy does this thing sound good!

This report is just my initial listening impressions after a couple of days. For now, I've located this server adjacent to my Dell (in another room), so I can compare them head to head. For simplicity, the endpoint is configured with Roon Bridge. 

 

From the first note, there's really no contest. I've talked about how I was living with some harshness ever since I resumed using my Dell as the Roon Server. Well, the lack of that harshness is what hits you first. The sound is so much more natural, relaxed, and yet, much more dynamic. I don't have the ability to compare the NUC i7 as a server with this 8700T server, but I strongly suspect this is superior.

 

Happy as I was with this, I next ran a 50ft Cat 6a generic cable from the second JCAT Net port from this server to the SOtM switch, to which my endpoint is connected. Wow, this was another uptick in SQ. I think this credit goes to bridging (in general) but more to the JCAT Net card in particular. Roy had already raved about this card to me, but I had to hear it for myself. The biggest improvement is in tonal richness and density. There is just more texture and bloom (in a good way). Most importantly, and personally very exciting, for the first time in my system, I heard the same album on Qobuz sound as good as local files on NAS. Kudos to @Marcin_gps et al. This is an outstanding product.

 

I have a lot of interesting things to try next. I plan to move this server into my listening room, where it can benefit from better cabling and better power. We'll see how much of an uptick that provides. This is where the fanless chassis comes in really handy. There are some PSU experiments to try, but Roy has already explored these, and I'm inclined to trust him on these. I'll describe these later. Finally, there is all this buzz about Euphony, so at some point I should give that a whirl.

 

I will say, this may be as close to endgame in the digital streaming chain as I've ever felt I've gotten!

[rickca]

21 minutes ago, austinpop said:

Well, the lack of that harshness is what hits you first.

To what do you attribute this?  Did you compare this config to your Dell with the HDPLEX 400W ATX LPS?  Or your Dell with the JCAT NET card?

 

Before you added the SOtM switch, how were the server and endpoint connected?

[austinpop]

47 minutes ago, rickca said:

To what do you attribute this?

 

Well, the hope was that all of these would contribute:

• low noise LPS
• better regulation due to separate rails for each ATX voltage
• CPU on separate rail, per findings by Larry and others
• No grunge from DC-ATX converter
• lower TDP == lower heat == lower noise
• JCAT Net Card Femto reclocking
• fanless

The net result is certainly spectacular, but I can't even begin to establish causality. It would take a huge amount of experimentation, for which I have no patience at this point. All I want to do now is kick back and listen!

 

47 minutes ago, rickca said:

Did you compare this config to your Dell with the HDPLEX 400W ATX LPS?  Or your Dell with the JCAT NET card?

 

No. I had zero motivation to do so. Sounds like a lot of work, for a slower i7-4770 CPU with close to double the TDP.  I wanted a ground-up low noise solution.

 

47 minutes ago, rickca said:

Before you added the SOtM switch, how were the server and endpoint connected?

 

Currently (this is temporary), the configuration is as follows. Keep in mind the new server has 4 bridged ports, 2 on the JCAT card, and 2 on the mobo.

 

R

O    <---------> (JCAT port 1) new server (bridged) (JCAT port 2) <--------->  Dell server

U

T

E    <-------> switch (SOtM or TLS) <--------> NUC i7 (clock modded) <---> tX-USBultra <----> DAC

R

 

The bridged config for today's experiment looks like this:

 

R

O    <---------> (JCAT port 1) new server (bridged) (mobo port) <--------->  Dell server

U                                                             |

T                                               (JCAT port 2)

E                                                             |

R                                                               <-------> switch (SOtM or TLS) <--------> NUC i7 (clock modded) <---> tX-USBultra <----> DAC

 

The final configuration will look like:

 

R

O <--------->  Dell server

U   

T

E 

R    <---------> (JCAT port 1) new server (bridged) (JCAT port 2)  <-------> switch (SOtM or TLS) <--------> NUC i7 (clock modded) <---> tX-USBultra <----> DAC

[Chopin75]

1 hour ago, austinpop said:

Custom Roon Server - Initial Report

 

As I've described here in the past, while I have upgraded my Roon endpoint to near-endgame with the sCLK-EX-modified NUC7i7DNBE/Plato X7D running AL in RAM, my server upgrade has been pending. In the meantime, I've managed to squeeze (pun intended) out a lot of performance from my generic Dell XPS 8700 desktop. However, it was always my plan to upgrade my server, following shamelessly in the footsteps of @romaz.

 

Roy has been refining his server build, and while he is still tweaking, I acquired his first build to try in my system. I am just going to enumerate the system here, because most of the credit goes to Roy for the thought and effort he put into this. Here are the system details:

• HDPlex H3 chassis
• Asrock H370M-ITX/AC mobo
• Intel 8700T CPU
• HyperX Fury 2400MHz DDR4 8GB
• JCAT NET Card Femto
• HDPLEX 400W ATX Linear Power Supply
  • 19V rail modded to 12V to supply independent EPS input to CPU
  • Ghent JSSG360 ATX-to-ATX and XLR-to-EPS cables

While Roy and I discussed these options at length, it's really he who drove this system selection, and the end result is truly special. Before I describe the sound, here is a brief rationale for these choices. 

• The 8700T CPU has 6 cores/12 threads/12MB Smart Cache, all for a TDP of 35W at a base frequency of 2.4GHz. This seemed to be a good balance of low power and hefty horsepower. 
• The HDPlex 400W Linear PSU has independent rails for the ATX voltages (3.3, 5, and 12V), with no DC-ATX conversion. Plus, by modifying the 19V rail to 12V, this adds another rail solely to power the CPU via the EPS input.
• TDP 35W is well within the capability of the H3 fanless chassis, and we snake the cables into the chassis through a PCIe slot opening on the back. Not elegant, but effective.
• The JCAT Net card was chosen to provide a high quality bridging solution, and as you'll read, boy does this thing sound good!

This report is just my initial listening impressions after a couple of days. For now, I've located this server adjacent to my Dell (in another room), so I can compare them head to head. For simplicity, the endpoint is configured with Roon Bridge. 

 

From the first note, there's really no contest. I've talked about how I was living with some harshness ever since I resumed using my Dell as the Roon Server. Well, the lack of that harshness is what hits you first. The sound is so much more natural, relaxed, and yet, much more dynamic. I don't have the ability to compare the NUC i7 as a server with this 8700T server, but I strongly suspect this is superior.

 

Happy as I was with this, I next ran a 50ft Cat 6a generic cable from the second JCAT Net port from this server to the SOtM switch, to which my endpoint is connected. Wow, this was another uptick in SQ. I think this credit goes to bridging (in general) but more to the JCAT Net card in particular. Roy had already raved about this card to me, but I had to hear it for myself. The biggest improvement is in tonal richness and density. There is just more texture and bloom (in a good way). Most importantly, and personally very exciting, for the first time in my system, I heard the same album on Qobuz sound as good as local files on NAS. Kudos to @Marcin_gps et al. This is an outstanding product.

 

I have a lot of interesting things to try next. I plan to move this server into my listening room, where it can benefit from better cabling and better power. We'll see how much of an uptick that provides. This is where the fanless chassis comes in really handy. There are some PSU experiments to try, but Roy has already explored these, and I'm inclined to trust him on these. I'll describe these later. Finally, there is all this buzz about Euphony, so at some point I should give that a whirl.

 

I will say, this may be as close to endgame in the digital streaming chain as I've ever felt I've gotten!

Mine was similar to yours, except it was local drive all in one box HDPlex + external LPS with euphony OS, no network, ryzen 2700x and pinkfaun USB bridge instead, direct USB —-> DAC, no need galvanic isolation. SQ is much improved, similar to your observation. HDPlex does make a fine setup.

[rickca]

@Chopin75 I thought you were using the HDPLEX 200W LPS/400W DC-ATX combination rather than the 400W ATX LPS?

 

Interesting, Ghent has some new cables that could be used with an HDPLEX 200W LPS/400W DC-ATX configuration ... DC42 and DC-PC01.  I'm not sure who asked Ghent to start making these.

[Chopin75]

23 minutes ago, rickca said:

@Chopin75 I thought you were using the HDPLEX 200W LPS/400W DC-ATX combination rather than the 400W ATX LPS?

 

Interesting, Ghent has some new cables that could be used with an HDPLEX 200W LPS/400W DC-ATX configuration ... DC42 and DC-PC01.  I'm not sure who asked Ghent to start making these.

Oh yah, too lazy to type 200W, (great memory!) not sure if that would make much difference. If I were to drive my CPU hard then maybe, like upsampling to DSD512 or M-ch DSD upsampling with HQplayer. The cables you mean the LPS cables to connect the LPS to PC etc..?  Or internal cables from the LPS connector to the DC-ATX

[rickca]

4 minutes ago, Chopin75 said:

Oh yah, too lazy to type 200W, (great memory!) not sure if that would make much difference. If I were to drive my CPU hard then maybe, like upsampling to DSD512 or M-ch DSD upsampling with HQplayer. The cables you mean the LPS cables to connect the LPS to PC etc..?  Or internal cables from the LPS connector to the DC-ATX

The 400W ATX LPS is quite a different thing than the 200W LPS/DC-ATX combination.  It's not just 400W vs 200W.

DC42 is a replacement for the Neutrik XLR to DC (7.4mm x 5.0 mm) cable supplied with the HDPLEX 200W.  DC-PC01 is a 24-pin cable to use between the DC-ATX and motherboard.  Of course, the quality of these Ghent cables far exceeds what HDPLEX provides. 

[valveboy]

3 hours ago, rickca said:

Of course, the quality of these Ghent cables far exceeds what HDPLEX provides. 

how and why?

please provide explanations/examples to back up your statement.

[rafa]

On 3/5/2019 at 8:38 PM, numlog said:

Have you ever experimenting with underclocking and noticed changes from that?

Sorry for delayed answer. Yes I have. Its same effect as changing timings and other bios settings. You can slightly hear it. 

[rickca]

3 hours ago, valveboy said:

how and why?

please provide explanations/examples to back up your statement.

I've responded to you via PM since @Jiffi32 has marked your question as OT.  I imagine he just doesn't want this thread to turn into a cable discussion.

[Jiffi32]

13 minutes ago, rickca said:

I've responded to you via PM since @Jiffi32 has marked your question as OT.  I imagine he just doesn't want this thread to turn into a cable discussion.

 

Just didn't want to see an objectivist vs subjectivist argument start in this thread, but sure @austinpop would moderate if it did.

[Dutch]

I had been on the fence regarding the JCAT NET card since March 2017 (yes, before it was even released) but you @austinpop just nudged me in the direction of actually ordering it. So thanks a lot! 😉😂

 

It’s currently sold out but new stock should arrive coming week.

 

I don’t have any audiophile switch yet and part of me being on the fence was the upcoming release of the etherregen. I will install the Jcat card in my music server which is in bridged config now using one on board NIC and an add-in i210 PCIE NIC which connects to a Roon end-point AL i7 NUC using an AQ Diamond RJ/E cable. I’m thinking I can always move the JCAT NIC to the end-point using a M.2 -> PCIE 4x converter/riser if the etherregen does the theoretical fantastic job of fully making server optimizations unnecessary.

[Johnseye]

8 hours ago, valveboy said:

how and why?

please provide explanations/examples to back up your statement.

 

5 hours ago, rickca said:

I've responded to you via PM since @Jiffi32 has marked your question as OT.  I imagine he just doesn't want this thread to turn into a cable discussion.

 

4 hours ago, Jiffi32 said:

 

Just didn't want to see an objectivist vs subjectivist argument start in this thread, but sure @austinpop would moderate if it did.

 

I don't think it's OT and I'm sure Rajiv would point @valveboy to all the discussion regarding JSSG 360 benefits found in this thread.  Rajiv certainly doesn't need to back up his statement.  When someone asks about the benefits of JSSG they need to do their own research, much of which is in this thread.

[Johnseye]

14 hours ago, austinpop said:

Custom Roon Server - Initial Report

 

As I've described here in the past, while I have upgraded my Roon endpoint to near-endgame with the sCLK-EX-modified NUC7i7DNBE/Plato X7D running AL in RAM, my server upgrade has been pending. In the meantime, I've managed to squeeze (pun intended) out a lot of performance from my generic Dell XPS 8700 desktop. However, it was always my plan to upgrade my server, following shamelessly in the footsteps of @romaz.

 

Roy has been refining his server build, and while he is still tweaking, I acquired his first build to try in my system. I am just going to enumerate the system here, because most of the credit goes to Roy for the thought and effort he put into this. Here are the system details:

• HDPlex H3 chassis
• Asrock H370M-ITX/AC mobo
• Intel 8700T CPU
• HyperX Fury 2400MHz DDR4 8GB
• JCAT NET Card Femto
• HDPLEX 400W ATX Linear Power Supply
  • 19V rail modded to 12V to supply independent EPS input to CPU
  • Ghent JSSG360 ATX-to-ATX and XLR-to-EPS cables

While Roy and I discussed these options at length, it's really he who drove this system selection, and the end result is truly special. Before I describe the sound, here is a brief rationale for these choices. 

• The 8700T CPU has 6 cores/12 threads/12MB Smart Cache, all for a TDP of 35W at a base frequency of 2.4GHz. This seemed to be a good balance of low power and hefty horsepower. 
• The HDPlex 400W Linear PSU has independent rails for the ATX voltages (3.3, 5, and 12V), with no DC-ATX conversion. Plus, by modifying the 19V rail to 12V, this adds another rail solely to power the CPU via the EPS input.
• TDP 35W is well within the capability of the H3 fanless chassis, and we snake the cables into the chassis through a PCIe slot opening on the back. Not elegant, but effective.
• The JCAT Net card was chosen to provide a high quality bridging solution, and as you'll read, boy does this thing sound good!

This report is just my initial listening impressions after a couple of days. For now, I've located this server adjacent to my Dell (in another room), so I can compare them head to head. For simplicity, the endpoint is configured with Roon Bridge. 

 

From the first note, there's really no contest. I've talked about how I was living with some harshness ever since I resumed using my Dell as the Roon Server. Well, the lack of that harshness is what hits you first. The sound is so much more natural, relaxed, and yet, much more dynamic. I don't have the ability to compare the NUC i7 as a server with this 8700T server, but I strongly suspect this is superior.

 

Happy as I was with this, I next ran a 50ft Cat 6a generic cable from the second JCAT Net port from this server to the SOtM switch, to which my endpoint is connected. Wow, this was another uptick in SQ. I think this credit goes to bridging (in general) but more to the JCAT Net card in particular. Roy had already raved about this card to me, but I had to hear it for myself. The biggest improvement is in tonal richness and density. There is just more texture and bloom (in a good way). Most importantly, and personally very exciting, for the first time in my system, I heard the same album on Qobuz sound as good as local files on NAS. Kudos to @Marcin_gps et al. This is an outstanding product.

 

I have a lot of interesting things to try next. I plan to move this server into my listening room, where it can benefit from better cabling and better power. We'll see how much of an uptick that provides. This is where the fanless chassis comes in really handy. There are some PSU experiments to try, but Roy has already explored these, and I'm inclined to trust him on these. I'll describe these later. Finally, there is all this buzz about Euphony, so at some point I should give that a whirl.

 

I will say, this may be as close to endgame in the digital streaming chain as I've ever felt I've gotten!

 

I've come from a slightly different, but similar direction.  Like several here we had targeted a single server solution and I believe mine was very similar to Roy's.  My server is using an sCLK for its clocks and I thought Roy had been doing the same thing but I could be wrong in that he was only using the tX-USBexp previously and not an sCLK.  That doesn't really matter except that Roy has been the ground breaker in terms of research, experimentation and implementation.  We are all eager to try out his discoveries.  I know a few others here who have similar single servers as mine.  sCLK modified mobos, powered by an SR7.  That was an excellent foundation for a long time.  Only the NUC as an endpoint improved upon it.  Using the NUC allowed me to pull the tX-USBexp out of the server, but I had still been powering it by the SR7.

 

Top notch power for the server has always been a challenge.  While my SR7 was feeding the server itself I did need the HDPlex DC to ATX converter.  What impact that converter has on the SR7's power I don't know exactly.  The other SR7 rail is powering the sCLK and tX-USBultra.  It isn't easy to get an SR7, much less an SR4 and it takes a long long time, not to mention $$.  I'm about 2 years out with Paul right now but I want to use my 2 SR7 rails further downstream where they have a bigger impact.  I have an HDPlex 400w LPSU arriving Monday.  That will allow for this:

 

Server - HDPlex LPSU

Server and tX-USBultra sCLK-EX - SR7 9v rail

NUC - SR7 19v rail (LPS-1.2 rail for sCLK)

LPS-1.2 - sNH-10G switch (LPS-1.2 rail for sCLK)

LPS-1 - TLS switch

sPS-500 - Router

 

In an ideal world SR7's are powering everything.  Unfortunately that's not an easy task so the above is the best I can do...for now.  I can also move the SR7 9v rail from the server/tX-USBultra sCLK to the NUC/sNH-10G switch sCLK.  I'm still not sold on the tX-USBultra in the chain but will wait to hear it with the EVOX caps.

 

My server, NUC and switch will all be master clocked by the REF-10.

 

My server can currently upsample to DSD512 and PCM 384 without breaking a sweat. This is critical for me because I really like what I hear with HQPlayer and the Spring L3 R2R DAC.  Maybe the new Hugo with scaler will be better.  I had the Dave without the scaler for a while and it was pretty darn close.

[elan120]

14 hours ago, austinpop said:

Custom Roon Server - Initial Report

 

As I've described here in the past, while I have upgraded my Roon endpoint to near-endgame with the sCLK-EX-modified NUC7i7DNBE/Plato X7D running AL in RAM, my server upgrade has been pending. In the meantime, I've managed to squeeze (pun intended) out a lot of performance from my generic Dell XPS 8700 desktop. However, it was always my plan to upgrade my server, following shamelessly in the footsteps of @romaz.

 

Roy has been refining his server build, and while he is still tweaking, I acquired his first build to try in my system. I am just going to enumerate the system here, because most of the credit goes to Roy for the thought and effort he put into this. Here are the system details:

• HDPlex H3 chassis
• Asrock H370M-ITX/AC mobo
• Intel 8700T CPU
• HyperX Fury 2400MHz DDR4 8GB
• JCAT NET Card Femto
• HDPLEX 400W ATX Linear Power Supply
  • 19V rail modded to 12V to supply independent EPS input to CPU
  • Ghent JSSG360 ATX-to-ATX and XLR-to-EPS cables

While Roy and I discussed these options at length, it's really he who drove this system selection, and the end result is truly special. Before I describe the sound, here is a brief rationale for these choices. 

• The 8700T CPU has 6 cores/12 threads/12MB Smart Cache, all for a TDP of 35W at a base frequency of 2.4GHz. This seemed to be a good balance of low power and hefty horsepower. 
• The HDPlex 400W Linear PSU has independent rails for the ATX voltages (3.3, 5, and 12V), with no DC-ATX conversion. Plus, by modifying the 19V rail to 12V, this adds another rail solely to power the CPU via the EPS input.
• TDP 35W is well within the capability of the H3 fanless chassis, and we snake the cables into the chassis through a PCIe slot opening on the back. Not elegant, but effective.
• The JCAT Net card was chosen to provide a high quality bridging solution, and as you'll read, boy does this thing sound good!

This report is just my initial listening impressions after a couple of days. For now, I've located this server adjacent to my Dell (in another room), so I can compare them head to head. For simplicity, the endpoint is configured with Roon Bridge. 

 

From the first note, there's really no contest. I've talked about how I was living with some harshness ever since I resumed using my Dell as the Roon Server. Well, the lack of that harshness is what hits you first. The sound is so much more natural, relaxed, and yet, much more dynamic. I don't have the ability to compare the NUC i7 as a server with this 8700T server, but I strongly suspect this is superior.

 

Happy as I was with this, I next ran a 50ft Cat 6a generic cable from the second JCAT Net port from this server to the SOtM switch, to which my endpoint is connected. Wow, this was another uptick in SQ. I think this credit goes to bridging (in general) but more to the JCAT Net card in particular. Roy had already raved about this card to me, but I had to hear it for myself. The biggest improvement is in tonal richness and density. There is just more texture and bloom (in a good way). Most importantly, and personally very exciting, for the first time in my system, I heard the same album on Qobuz sound as good as local files on NAS. Kudos to @Marcin_gps et al. This is an outstanding product.

 

I have a lot of interesting things to try next. I plan to move this server into my listening room, where it can benefit from better cabling and better power. We'll see how much of an uptick that provides. This is where the fanless chassis comes in really handy. There are some PSU experiments to try, but Roy has already explored these, and I'm inclined to trust him on these. I'll describe these later. Finally, there is all this buzz about Euphony, so at some point I should give that a whirl.

 

I will say, this may be as close to endgame in the digital streaming chain as I've ever felt I've gotten!

@austinpop, thank you for sharing and write up.  I plan on using some ideas here to see if it will help improve my current server that has much higher TDP due to my current desire of using HQP to upsample to DSD515 non-2s filters, but this build recipe is great to learn and copy in case my upsampling desire changed.  While going through upgrades with my present server, I had my JCAT Net Card clock replaced by a  sCLK-OCX10 connected sCLK-EX clock output instead with good result. Wonder if you plan to do clock modification to your new server and perhaps JCAT Net Card?

[elan120]

35 minutes ago, Johnseye said:

 

I have an HDPlex 400w LPSU arriving Monday.  That will allow for this:

 

Server - HDPlex LPSU

Server and tX-USBultra sCLK-EX - SR7 9v rail

NUC - SR7 19v rail (LPS-1.2 rail for sCLK)

LPS-1.2 - sNH-10G switch (LPS-1.2 rail for sCLK)

LPS-1 - TLS switch

sPS-500 - Router

 

In an ideal world SR7's are powering everything.  Unfortunately that's not an easy task so the above is the best I can do...for now.  I can also move the SR7 9v rail from the server/tX-USBultra sCLK to the NUC/sNH-10G switch sCLK.  I'm still not sold on the tX-USBultra in the chain but will wait to hear it with the EVOX caps.

 

My server, NUC and switch will all be master clocked by the REF-10.

 

My server can currently upsample to DSD512 and PCM 384 without breaking a sweat. This is critical for me because I really like what I hear with HQPlayer and the Spring L3 R2R DAC.  Maybe the new Hugo with scaler will be better.  I had the Dave without the scaler for a while and it was pretty darn close.

Looking forward to your impression on changes from HDPlex 400W LPSU, as I also hope this power supply has enough to power my server.

 

I recently had my tX-USBUltra upgraded with Evox cap, liked the result quite well, and would like to hear your impression once you had your tX-USBUltra upgraded.  In fact, with the positive result from the 3 Evox cap upgrades I have thus far, I plan on adding the Evox cap to the server sCLK-EX soon when parts arrived, hopefully the positive trend will continue.

 

 

[RickyV]

1 hour ago, elan120 said:

Looking forward to your impression on changes from HDPlex 400W LPSU, as I also hope this power supply has enough to power my server.

 

I recently had my tX-USBUltra upgraded with Evox cap, liked the result quite well, and would like to hear your impression once you had your tX-USBUltra upgraded.  In fact, with the positive result from the 3 Evox cap upgrades I have thus far, I plan on adding the Evox cap to the server sCLK-EX soon when parts arrived, hopefully the positive trend will continue.

 

 

 

Hi Kevin, would it be possible to share type of Evox cap they are using? 

 

Btw I have found the Ethernet xo, unbelievable of thousands of components on the mobo the one I need had the drop of glue on it to hold the battery. 🤯 thanks again.

[austinpop]

4 hours ago, Johnseye said:

 

 

 

I don't think it's OT and I'm sure Rajiv would point @valveboy to all the discussion regarding JSSG 360 benefits found in this thread.  Rajiv certainly doesn't need to back up his statement.  When someone asks about the benefits of JSSG they need to do their own research, much of which is in this thread.

 

Let's just put this to bed. Roy and I ordered the JSSG360 cables from Ghent purely on the premise that we've found JSSG360 to be beneficial with other cables. For the ATX umbilical, Ghent said he had not found any suitable OFC wire that could be practical, so we opted for the "18awg silicone copper" with JSSG360 treatment. For the separate XLR-to-8-pin EPS cable, Ghent used a Canare 4S6G(OFC) cable.

 

Is the difference between these cables and the stock cables Larry supplies dramatic? No. But it does seem to be a small improvement, so it was worth it.

 

Also, while I am on the topic of cables and power, I forgot to mention how the JCAT Net Card Femto was being powered. I'm using the 5V XLR output from the HDPlex. In his experiments, Roy found a big improvement stepping up to an SR-4, but I don't have one to spare. Unfortunately, the card needs 2A, so exceeds the capacity of the LPS-1.2.