Building a DIY Music Server

[rico25]

Thank you Angelo :)

[AngeloVRA]

@rico25 Found a pic of my experimental silk tubing covered DC cable.

[rico25]

Interesting that you prefer your DC cable without geometry (parallel conductors) because in theory, jssg360 (or twisted pair or starquad) are better for inductance. But the ears are the judges :)

[bit01]

@AngeloVRAMeticulous work - inspiring! Thanks for posting!

[Vivling2021]

On 4/4/2021 at 1:34 AM, Topk said:

Oh I completely agree that *dual* Xeon *with* Asus Sage C621 sounds better according to Emile. The combination.
 

Single Xeon was never tried that I’m aware of vs AMD Ryzen, and only dual XEON was tried on that dual CPU Sage c621 MB (which happens to be very expensive and industrial grade). So it’s a package.
 

In fact, I actually did test different MB on same CPU and RAM and will report soon (always only 1 variable change at a time!). Hint: significant impact of MB on sound quality. 

 

On the performance upgrade side, I did try CPU upgrade with, again, 1 variable change at a time (including going from 12 to 16 cores with all other variables unchanged) so that point specifically can actually be concluded based on actual listening test with a sample of 1 :-) .... with same architecture. And the importance of locking the CPU ratio to stock (or above) in my step by step undervolt process 

 

Another variable not mentioned is that USB card (eg Jcat) is critical to SQ with AMD (which Emile did NOT use when Xeon was selected since he tried the USB direct from the MB).

Usb direct from MB certainly will be less tha. optimal solution.  taiko has developed a pci-e usb card available only to extreme users as upgrade.  So having a pci-e usb is essential for optimal performance 

[Vivling2021]

On 1/13/2021 at 10:56 PM, guiltyboxswapper said:

I have pondered whether its worth buying another one of these Solarflare cards and seeing if we can externally feed it DC, given how good it is from just motherboard power alone.

I have found a good way to externally power the solarflare card. just buy a pci-e extender(without chips) with an external power input. given that the extender is not very long and is without chips, I assume it would not have too much of a negative effect on sound quality, the benefit is that it allows me to externally power the NIC using LPSU. That could potentially elevate my Solarflare 2522 into new heights. Stay tuned, I will update with my findings.

[MarcelNL]

Nice idea !

 

What power do you need to feed the Solarflare? I happen to have a nice 5V linear PSU lying around.....

[AME]

Anyone perhaps has a Solarflare X2522 working under Windows 10 LTSC?

 

After installing the card, the computer starts up fine, but after installing the windows drivers (Win Server 2019), I get a blue screen.

When removing the card everything works fine.

 

 

[Vivling2021]

5 hours ago, MarcelNL said:

Nice idea !

 

What power do you need to feed the Solarflare? I happen to have a nice 5V linear PSU lying around.....

I feed the solarflare x2522 using 6 pin 12v output from my ATX LPSU. This line goes directly from the LPSU itself, not from motherboard 

 

 

I tried to connect the 2522 to my Buffalo GS2016 and it doesn’t work.  Compatibility issues. I will buy a 25g router and maybe try SFP28 transceivers. Any suggestions?

[MarcelNL]

the only issue I can think of is when the SFP is incompatible with the router, or that you are crossfeeding the fiber channel. I struggled a bit when connecting my Zyxel router, swapping the fiber channels at one end and firing up the switch again did the trick.

 

12V..hmmm that voltage I do not have lying around...

[Tatomek7]

Following @Nenon list of parts, I did LPS to power server. 

 

The first one was based on vintage parts and I was more than delighted with the results. My server with single Xeon and Euphony sang beautifully.  Unfortunately I was not able to power C621 server through HDPlex 800 or Taiko ATX converter. In fact I burned two Hammond chokes. So, I was forced to use awaiting parts and build variation of V2 @Nenon’s unregulated LPS. I ordered the second pair of Hammonds as a replacement of the burned ones and voila

 

 

Rectifier board is based on Saligny active rectifier, Mundorf AG 33000uF 40V bypassed with 3,3uF Supreme Evo Oil, 
The LPS is still settling on, but after 4 days I already find it to play more modern way vs my “vintage” LPS. Bass has less authority vs “vintage” one, but Mundorfs HC needs much more time to express full potential, so now I have to be patient. 
Mundorf HC was bypassed with the same vintage 0,1uF capacitors except the last one, which is Mundorf Supreme Silver Gold Oil. If I will miss some warm, I will try Miflex KPCU, which is waiting for a chance to be applied. I didn’t have Vcap Odam’s and at the moment do not want to buy them, as I’m tired with contacts with customs ( VCaps are not available within EU zone ). 
I didn’t try to power C621 server with the latter LPS. I will try it, when capacitors will have at least 200h on the counter.

 

Both LPS's described above gave serious increase in SQ and it's worth to try it. 

 

I already ordered two hand made chokes to replace Hammonds, so in few weeks I should have more information if it was right decision. Next step will be to improve wiring and to use better connectors, I will also add next Mundorf HC 47000uF to be closer to V3 version.

 

[Exocer]

28 minutes ago, Tatomek7 said:

Following @Nenon list of parts, I did LPS to power server. 

 

The first one was based on vintage parts and I was more than delighted with the results. My server with single Xeon and Euphony sang beautifully.  Unfortunately I was not able to power C621 server through HDPlex 800 or Taiko ATX converter. In fact I burned two Hammond chokes. So, I was forced to use awaiting parts and build variation of V2 @Nenon’s unregulated LPS. I ordered the second pair of Hammonds as a replacement of the burned ones and voila

 

 

Rectifier board is based on Saligny active rectifier, Mundorf AG 33000uF 40V bypassed with 3,3uF Supreme Evo Oil, 
The LPS is still settling on, but after 4 days I already find it to play more modern way vs my “vintage” LPS. Bass has less authority vs “vintage” one, but Mundorfs HC needs much more time to express full potential, so now I have to be patient. 
Mundorf HC was bypassed with the same vintage 0,1uF capacitors except the last one, which is Mundorf Supreme Silver Gold Oil. If I will miss some warm, I will try Miflex KPCU, which is waiting for a chance to be applied. I didn’t have Vcap Odam’s and at the moment do not want to buy them, as I’m tired with contacts with customs ( VCaps are not available within EU zone ). 
I didn’t try to power C621 server with the latter LPS. I will try it, when capacitors will have at least 200h on the counter.

 

Both LPS's described above gave serious increase in SQ and it's worth to try it. 

 

I already ordered two hand made chokes to replace Hammonds, so in few weeks I should have more information if it was right decision. Next step will be to improve wiring and to use better connectors, I will also add next Mundorf HC 47000uF to be closer to V3 version.

 

Very nice build (s)!

 

It is fair to say that the benefits of the Unregulated LPS carry over to single CPU builds as well. My V2 is getting better by the day. What I will say is the break-in process is real! What started out as an immediate incremental improvement is starting to edge more and more in the direction of massive improvement as things settle in. 

 

I'm sure many would be interested in sourcing a similar rectifier board for their builds. Also your V2 uses two chokes. That being said, I plan to try V3 soon myself :). I do want to give V2 at least 600 hours before adding the additional components though.

 

My build:

 

Cheers,

-Rob

[Tatomek7]

21 minutes ago, Exocer said:

I'm sure many would be interested in sourcing a similar rectifier board for their builds.

I bought Saligny standard bridge through diy forum, but it’s also available on evotronix page. I saw also, it’s used in ARC6 build by @Nenon, @romaz, @ray-dude

[RickyV]

1 hour ago, Tatomek7 said:

VCaps are not available within EU zone

https://www.eltim.eu/en/capacitors-for-audio/vh-audio-capacitors/g-10000339

[Nsxturbo]

Here’s my version 3 of Nenon’s unregulated LPS powering my Taiko DC/ATX and my Asus WS-621 Sage. Still breaking in and obviously, some bypass cap differences, Mundorf Supreme silver/golds and Jupiter copper foil beeswax. All of these were ones I happened to have on-hand. Also one Mundorf 47000mF  AG with the 2- 47000mF HC’s. Plus the Saligny mosfet rectifier.

My impression after listening to version 2 for a couple of weeks, falls inline with Nenon’s description. Version 2 was super fast and wonderfully resolving but without any digital glare. I really loved it.
 

Initially version 3 was a bit disappointing because it definitely lacked the high frequency detail of version 2, but Wow, the bass had such weight, timber and authority and the soundstage was both wider and deeper! The more it broke-in and the more my ears adjusted to it, the more I thought there was no going back! I’m hopeful when the full complement of V-cap bypass caps arrive, that I’ll regain some of the additional “sparkle” that version 2 had. All-and-all, I couldn’t be more pleased with how well the Taiko DC/ATX with Nenon’s spec’d LPS are performing. My system has never sounded better. It’s nice to have such a resolving system that any new tweak is easily discernible.

 

 

[Vivling2021]

Follow up on the externally powering the Solarflare x2522.

I bought a Microtik CRS305 and connect it to the Solarflare via 10G SFP+ AOC. And it works beautifully under Euphony. I learned from StreamerFidelity to change it to Ultra low Latency under BIOS.  It sounds wonderful. I can cook an egg on this Solarflare(very hot), so I installed it outside the Streacom box. This is perhaps the first Solarflare X2522 powered externally by LPSU on this forum. I think it would beat most copper based NIC solutions many times more expensive.

 

My current streamer is Buffalo GS2016(LPSU)-fiber-MicroTik(LPSU)-fiber-Solarflare(externally powered by LPSU)-i7-10700(Teradak LPSU)-Apacer 8G ECC-JCAT USB XE(JS2)-Uptone Isoregen(JS2)-Schiit Ygg A2-Microzotl 2.0(LPSU)-Line Magnetic 219IA-Audio Note AN-E SPE/HE.

 

I currently consult with Realhifihelp from youtube and he gave me a lot of valuable advices. Any advice is welcome.

 

 

[Exocer]

3 hours ago, Nsxturbo said:

Here’s my version 3 of Nenon’s unregulated LPS powering my Taiko DC/ATX and my Asus WS-621 Sage. Still breaking in and obviously, some bypass cap differences, Mundorf Supreme silver/golds and Jupiter copper foil beeswax. All of these were ones I happened to have on-hand. Also one Mundorf 47000mF  AG with the 2- 47000mF HC’s. Plus the Saligny mosfet rectifier.

My impression after listening to version 2 for a couple of weeks, falls inline with Nenon’s description. Version 2 was super fast and wonderfully resolving but without any digital glare. I really loved it.
 

Initially version 3 was a bit disappointing because it definitely lacked the high frequency detail of version 2, but Wow, the bass had such weight, timber and authority and the soundstage was both wider and deeper! The more it broke-in and the more my ears adjusted to it, the more I thought there was no going back! I’m hopeful when the full complement of V-cap bypass caps arrive, that I’ll regain some of the additional “sparkle” that version 2 had. All-and-all, I couldn’t be more pleased with how well the Taiko DC/ATX with Nenon’s spec’d LPS are performing. My system has never sounded better. It’s nice to have such a resolving system that any new tweak is easily discernible.

 

 

Wonderful build. I am going to order another choke now and the remainder of the parts later this week. I can't wait to try V3. I am only several days into breaking in V2 at this point and it just keeps getting better. For you, when did V2 plateau and fully settle in?

 

I've had the chance to try 4 different power cords with V2. Currently, I own no endgame Power Cords but am hoping to go all out next year as I shift my focus!  I suggest everyone to try and find what matches their tastes, as V2 will absolutely reveal the differences in power cords. My experiences varied widely. As good as the V2 is, the wrong power cord ruined the experience for me. Coincidentally after trying 4 different cords I settled on the same cord I was using with the SR7T. It is the VH Audio Flavor 4 (unshielded, star quad) with Viborg v512 Rhodium plated copper contacts. I purchased this cable cheap on the used market with wattgate plugs and it has been great to experiment with. I also use the Fi-09 R NCF IEC inlet so perhaps there is some Rhodium synergy going on there. Who knows.

[Tatomek7]

16 hours ago, Nsxturbo said:

unregulated LPS powering my Taiko DC/ATX and my Asus WS-621

Did you power it with one 8 pin EPS connection, or you've added also 4 pin connection? At the moment I didn't succeed to move my C621 with unregulated LPS, but the reason could be my EPS 8 pin cables. I wonder if one 8 pin connection is enough to power Sage C621.

[bit01]

A query for those building the unreg PS for TAIKO DC DC/ATX What output voltages are you using? Can you give us an idea  the amount of voltage dip in your builds/usage?. Thanks.

[Exocer]

1 hour ago, bit01 said:

A query for those building the unreg PS for TAIKO DC DC/ATX What output voltages are you using? Can you give us an idea  the amount of voltage dip in your builds/usage?. Thanks.

Without a load, my early measurements got really close to 37V but seemed to settle closed to 36-36.5V.

With the machine up and running, and unregulated LPS attachhed, measurements came in at around 35.75V

[bit01]

1 hour ago, Exocer said:

Without a load, my early measurements got really close to 37V but seemed to settle closed to 36-36.5V.

With the machine up and running, and unregulated LPS attached, measurements came in at around 35.75V

That looks amazing- .75V dip but probably more than that during boot transients? This is with your i9-9900K (no OC) /NIC/USB-IC/NVME SSD/Euphony server without up sampling if I read that correctly from your previous posts? 

Thanks

ps I was getting about .7V dip with the SR7T/HDplex 800 during boot just on the CPU (i9-10900K no OC) rail. Have not measured with TAIKO but likely much more efficient

[Exocer]

4 minutes ago, bit01 said:

That looks amazing- .75V dip but probably more than that during boot transients? This is with your i9-9900K (no OC) /NIC/USB-IC/NVME SSD/Euphony server without up sampling if I read that correctly from your previous posts? 

Thanks

Probably more during boot, yes. I have not measured that yet.

 

i9-9900k OCed to 4.6ghz

Nic internally powered

One NVME internally powered

 

I have a USB card and OS NVME externally powered via LPS.

 

Euphony OS without upsampling, correct.

 

Cheers!

-Rob

 

 

[Gavin1977]

On 8/3/2021 at 2:56 AM, Vivling2021 said:

Follow up on the externally powering the Solarflare x2522.

I bought a Microtik CRS305 and connect it to the Solarflare via 10G SFP+ AOC. And it works beautifully under Euphony. I learned from StreamerFidelity to change it to Ultra low Latency under BIOS.  It sounds wonderful. I can cook an egg on this Solarflare(very hot), so I installed it outside the Streacom box. This is perhaps the first Solarflare X2522 powered externally by LPSU on this forum. I think it would beat most copper based NIC solutions many times more expensive.

 

My current streamer is Buffalo GS2016(LPSU)-fiber-MicroTik(LPSU)-fiber-Solarflare(externally powered by LPSU)-i7-10700(Teradak LPSU)-Apacer 8G ECC-JCAT USB XE(JS2)-Uptone Isoregen(JS2)-Schiit Ygg A2-Microzotl 2.0(LPSU)-Line Magnetic 219IA-Audio Note AN-E SPE/HE.

 

I currently consult with Realhifihelp from youtube and he gave me a lot of valuable advices. Any advice is welcome.

 

 

Can you detail the use of the  Microtik CRS305, perhaps some photos of your chain?

[Tatomek7]

15 hours ago, bit01 said:

What output voltages are you using? Can you give us an idea  the amount of voltage dip in your builds/usage?

Server connected, but not powered 33,2V,

during boot - 31,2 - 31,5V,

operating 32,2V.

Transformer output 24VAC, 400VA

Server is based on X99 MB, Xeon e5 4610 v3 - TDP 105W, 8x4GB RAM, 1TB nmve storage and 256GB nvme OS powered from MB, NIC and USB cards powered externally.

[bit01]

8 hours ago, Tatomek7 said:

Server connected, but not powered 33,2V,

during boot - 31,2 - 31,5V,

operating 32,2V.

Transformer output 24VAC, 400VA

Server is based on X99 MB, Xeon e5 4610 v3 - TDP 105W, 8x4GB RAM, 1TB nmve storage and 256GB nvme OS powered from MB, NIC and USB cards powered externally.

Thank you very much for your data. I have a 500VA/24VAC Primrose tranny lying around, I am encouraged to give the unregulated PS a try.