Building a DIY Music Server

[kraftundkultur]

51 minutes ago, Johnseye said:

Both as the existing design is the same for a ULPS. 

 

it would be nice to see Angelo ULPS chassis with internal pictures

[Exocer]

58 minutes ago, Johnseye said:

 

Both as the existing design is the same for a ULPS. 

Would require a new backplate though, with one output vs a motherboard rear panel.

[AngeloVRA]

(DRAFT) Full size option;

HxWxD: internal 150 x 410 x 450

               External: 170 x 480 x 480

 

Pick your poison.

 

I'll proceed with only one design choice if there is enough interest.

This is because costs become prohibitive with quantities less than 10/20.

And since cost had been mentioned as being critical in several posts, my apologies for this spoiler alert but the reality is that a full CNC and thicker more substantial chassis with similar dimensions as Taiko DIY chassis cannot be made anywhere near less than or even equal to Taiko DIY chassis price.

 

Some notes on heatsink:

- a thicker base (now at 10mm) aids in spreading the heat to more fins but any thicker than 10mm increases the thermal resistance.

- our best bet is to locate heat pipe grooves all above the level of the heat pipe at the CPU heat block, and see if we can get a polished surface inside the heat pipe grooves.

 

 

 

 

 

 

[JackJohnson316]

2 hours ago, Dev said:

 

Depending on the price, I am in for a group buy but I am not sure I am following you. Are you talking about AgeloVRA's chassis or some other chassis ? No doubt I would like to have the Taiko atx inside the motherboard chassis as well.

Hi yes, I was referring to the AgeloVRA diy case. My apologies for any confusion.

[Johnseye]

1 hour ago, Exocer said:

Would require a new backplate though, with one output vs a motherboard rear panel.

 

I would probably buy the existing design with PC backplate so that I could use it for both a server case or power supply if I wanted.  Granted the back wouldn't be pretty from a power supply chassis standpoint, but I'd prefer flexible functionality and no one sees the back but me.

[AngeloVRA]

5 hours ago, Johnseye said:

 

I would probably buy the existing design with PC backplate so that I could use it for both a server case or power supply if I wanted.  Granted the back wouldn't be pretty from a power supply chassis standpoint, but I'd prefer flexible functionality and no one sees the back but me.

Though visually the same at first glance, My other chassis is substantially different from my server chassis, and has a different heatsink.  It was designed to mimic the heatsink mounting holes of a Streacom FC-10 but with twice the heatsink fins, so that my existing DR7T regulator modules can be installed in it. 
It has the same front, top and bottom panels as my server chassis. 
Im awaiting Jaeger chassis connectors for it and have to build 2 more Teflon ZFoil regulator modules for it before I can complete the build and share. 


And here are some close up pics that show the surface texture and details of the chassis design. 
 

 

regards

Angelo

[AngeloVRA]

8 hours ago, Exocer said:

Would require a new backplate though, with one output vs a motherboard rear panel.

If the server chassis is repurposed as a power supply chassis, They can use the holes for the Jaeger chassis connector for output.

and cover the I/O panel and PCIE slot openings with a thin aluminum or copper plate from the inside. 

The cover plate for the I/O Panel can then have an IEC connector.  

[Exocer]

4 hours ago, AngeloVRA said:

If the server chassis is repurposed as a power supply chassis, They can use the holes for the Jaeger chassis connector for output.

and cover the I/O panel and PCIE slot openings with a thin aluminum or copper plate from the inside. 

The cover plate for the I/O Panel can then have an IEC connector.  

True! But chassis height becomes a concern, especially if one uses the large Mundorf cap. The 120mm height of the slimline 3U is already almost snug .

[AngeloVRA]

UPDATE on design of chassis for use with dual CPU ASUS Sage server

 

The majority consensus amongst the guys who contacted me was that they preferred to run their system in a 2 box configuration. Hence, the server chassis was designed to accommodate the Asus Sage motherboard  with a Taiko ATX DC-DC converter or any similar device,

 

• Using my current server chassis as a base,  we increased the internal depth from 340mm to 380mm
• Internal height is now 130mm, up 30mm from original 100mm
• Internal width stayed the same at 370mm

The above changes will allow mounting of the Taiko ATX DC converter vertically on the back side of the front panel.

The new 130mm height will also result in a more substantial heatsink. Its structure has been updated to use 2mm fins with 7mm vent space between fins.

 

Here are some pics of the design.

 

In about a week's time, I will be finalizing the design documents that's required by the CNC facility  for them to be able to quote for the quantity that the group requires. Anybody interested to join in can simply message me.

 

 

[altsouza]

Hello,

 

My Roon Server (Asus Tuf/i5/16gb/hdplex h5/hdplex 300psu/Gentooplayer OS) died recently. 

First Option: Buy Aorus B760m Mb+ i7 14700k + ddr5 replacements...

Or buy a MacMini M2 (not the pro version).

Both options will be for the similar price in my Country.

I would very much appreciate if you could recomend on this matter.

Also recomendations on Mb brand and processor type for the first option will be appreciated.

Thanks

Antonio

 

[Nenon]

20 hours ago, AngeloVRA said:

The majority consensus amongst the guys who contacted me was that they preferred to run their system in a 2 box configuration.

 

Sorry to ruin the party but I have to say this - a long umbilical between the ULPS and the Taiko ATX will negatively affect the sound quality. I would keep it as short as possible and make it with as thick wires as possible. I am saying this from personal experience but also had some interesting feedback. 

 

The only one time I've heard someone preferring a regulated linear power supply (the absolute best version of the teflon DR7 PH made before scamming dozens of my friends) over a well built unregulated LPS + Taiko ATX on the dual SAGE motherboard was when a two box solution was used with an umbilical over 1 meter long. I advised that person to redo his build with the Taiko chassis and the ULPS inside the chassis, and he never looked back. 

 

The bottom line is that the umbilical length is crucial and I advise against using two chassis.  

[Todd H]

Hi Nenon, is their an umbilical combination (length/gauge/material) that would lose little to no performance over a one box configuration? Seems a bit counterintuitive to put all the power pieces in the same box as the server. If the only benefit is ULPS umbilical length, can that not be overcome with proximal component location and shortest possible wire length etc.?

 

I hope I'm understanding your post correctly.

 

Todd

[littlej0e]

So it seems the ultimate solution would be to design a two chassis system capable of an incredibly short DC run (perhaps the same, or slightly longer, than a one chassis solution). Or use internal shielding for power components in a one chassis system? 

[Johnseye]

6 hours ago, Nenon said:

 

Sorry to ruin the party but I have to say this - a long umbilical between the ULPS and the Taiko ATX will negatively affect the sound quality. I would keep it as short as possible and make it with as thick wires as possible. I am saying this from personal experience but also had some interesting feedback. 

 

The only one time I've heard someone preferring a regulated linear power supply (the absolute best version of the teflon DR7 PH made before scamming dozens of my friends) over a well built unregulated LPS + Taiko ATX on the dual SAGE motherboard was when a two box solution was used with an umbilical over 1 meter long. I advised that person to redo his build with the Taiko chassis and the ULPS inside the chassis, and he never looked back. 

 

The bottom line is that the umbilical length is crucial and I advise against using two chassis.  

 

1 meter is a pretty long umbilical. By positioning the capacitor connecting to the umbilical close to a corner in the chassis and likewise the DC-ATX module directly next to the server chassis back corner, the umbilical can be short.

 

I'm using the same one you made and even when I had everything inside one chassis I couldn't have shortened it more than a couple inches. What I did have to increase was the length of the cables from the DC-ATX module to the motherboard. I think that was a better trade off than keeping all of the supply components in the case. 

 

Taiko does quiet a bit to shield the supply components from the motherboard, using copper to do so. 

 

You probably have more familiarity with this than anyone Nenon, but with component placement, wire gauge and shielding don't you think a 2 case solution can be viable?

[seeteeyou]

Virtually nothing would actually beat both conductivity and impedance of soldering / hardwiring (i.e bypassing the male + female connectors altogether) at all, otherwise @romaz wouldn't have posted all about them

 

https://audiophilestyle.com/forums/topic/29458-ultracap™-lps-1-listening-impressions-thread/page/4/#comment-589391

 

https://audiophilestyle.com/forums/topic/29458-ultracap™-lps-1-listening-impressions-thread/page/7/#comment-595070

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/720/#comment-1149407

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/721/#comment-1149634

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/721/#comment-1149674

 

---

 

While @Nenon might be happy with the impedance of Jaeger connectors (metal with relatively poor conductivity that is) back in 2019, obviously things evolved and he's also foregoing connectors even though the impedance might seem to be fairly low

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/593/#comment-987770

https://www.sealconusa.com/images/connectors/pdf/standard-rapid-series.pdf

Quote

< 0,0015Ω ⌀ 2 mm
< 0,001Ω ⌀ 3 mm
< 0,0007Ω ⌀ 4 mm

 

Then @romaz found something else that's quite interesting as well

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/723/#comment-1150466

https://www.helifreak.com/showthread.php?t=56967

Quote

I happen to have a sourcemeter at work, which is basically a ten thousand dollar multimeter. Using a 4-wire measurement (more accurate than using only two), i measured 130 micro ohms on a new Deans plug.

 

I also posted these RCA jacks and plugs with impressive impedance and conductivity

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/723/#comment-1150480

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/724/#comment-1156474

 

---

 

Whenever a 2-case design with umbilical cords are involved, both conductivity and impedance will be "ruined" as soon as connectors are added. Obviously those connectors couldn't be avoided in practice unless we're actually planning to put both cases in a permanent position, while actually having the "luxury" to solder that umbilical cord between them.

 

Even if we're somehow managing to find the "best" female and male connectors in terms of both connectivity and impedance, it's still inevitable that at least some of the performance should be left on the table whenever we don't go for something else that's tried and true already. A situation like this would remind me how to make the right choices in our own lives

 

https://nextbigideaclub.com/magazine/lessons-longevity-longest-lived-people-around-world-bookbite/46835/

Quote

The vast majority of premature death and disability is preventable with a healthy enough diet and lifestyle. There was an editorial published in the leading anti-aging journal Rejuvenation Research entitled, “Finally, a regimen to extend human life expectancy.” Was it some exotic new stem cell treatment or gene therapy? No, it was a reference to a Harvard analysis in which more than 100,000 men and women were followed for decades, and just a few basic lifestyle behaviors appeared to translate into 12 to 14 years of extra lifespan for the average 50 year-old. Even at age 50, you may be able to add an extra dozen years to your lifespan. Even at age 70, there could still be about ten extra years on the table. And if you clean up your life before age 50, nearly 18 years may be up-for-grabs based on simple common-sense behaviors.

 

Spending so much money and time on building the ULPS while "handicapping" everything afterwards, are you kidding me?

[Nenon]

14 hours ago, littlej0e said:

So it seems the ultimate solution would be to design a two chassis system capable of an incredibly short DC run 

 

Agreed. That's would be the best solution. 

 

A few years ago when I was considering a cost no object server there were two things I wanted to engineer. One was a two chassis solution so when they are stacked there is a special plug on the top panel of the bottom chassis and on the bottom panel of the top chassis. This way you don't need an umbilical between the two. I had a pretty awesome sketch that I drew on a piece of paper. When I showed it to a manufacturer, he said "Oh no!!!" because it was super complicated to manufacture. 

The other one was PCIe DAC that eliminates the need for USB, similar idea as Taiko's XDMI.  

 

Those are probably extreme cases, but if you can design the two chassis to minimize the umbilical length, that will be awesome. 

[NewOldman]

A small tip

NEUTRIK SPEAKON Connectors
Series of Speakon connectors for professional speaker connection technology. Robust design, latching device, all-insulated contacts, cable inlet with self-adjusting clamps.

• Contact rating: 30 A_RMS
• Contact resistance: ≤ 3 mΩ
• Operating voltage: 250 V˜ (music signal)
• Self-extinguishing according to UL 94V-0
• Spring copper with silver plating contacts and nickel plating zinc die-cast shell

 

speakON® - Neutrik

[SK8]

1 hour ago, Nenon said:

 

Agreed. That's would be the best solution. 

 

A few years ago when I was considering a cost no object server there were two things I wanted to engineer. One was a two chassis solution so when they are stacked there is a special plug on the top panel of the bottom chassis and on the bottom panel of the top chassis. This way you don't need an umbilical between the two. I had a pretty awesome sketch that I drew on a piece of paper. When I showed it to a manufacturer, he said "Oh no!!!" because it was super complicated to manufacture. 

The other one was PCIe DAC that eliminates the need for USB, similar idea as Taiko's XDMI.  

 

Those are probably extreme cases, but if you can design the two chassis to minimize the umbilical length, that will be awesome. 

I also like the idea of PCIe DAC and I did modify my DIY chassis for full height PCIe cards and expected the launch of the Taiko PCIe DAC card more than a year ago. Now there is no hope for this card from Taiko, even the Olympus I/O with DAC would not be available to the DIY community. It's quite a disappointment. 

 

I am not giving up this idea and just started pursuing it as my next project - a PCIe DAC card using available components from the market. Hope I can get to a stage so that I can share more with the DIY community here.

[AngeloVRA]

2 hours ago, Nenon said:

 

Agreed. That's would be the best solution. 

 

A few years ago when I was considering a cost no object server there were two things I wanted to engineer. One was a two chassis solution so when they are stacked there is a special plug on the top panel of the bottom chassis and on the bottom panel of the top chassis. This way you don't need an umbilical between the two. I had a pretty awesome sketch that I drew on a piece of paper. When I showed it to a manufacturer, he said "Oh no!!!" because it was super complicated to manufacture. 

The other one was PCIe DAC that eliminates the need for USB, similar idea as Taiko's XDMI.  

 

Those are probably extreme cases, but if you can design the two chassis to minimize the umbilical length, that will be awesome. 

This is where I am at the moment in finding a solution to shortest possible umbilical in a 2 box config

 

   

 

 

 

 

design notes:

 

-  No big hole for EMI to leak in/out

-  would look nice from the front… no visible wires/cables

To disengage server from a ULPS chassis below:
- disconnect molex from Taiko ATX
- ⁠push molex into the hole
- ⁠lift server chassis

 

To install:
-  unscrew ULPS link tube from server 
- position ULPS cable
- ⁠place server chassis on top, threading ULPS cable through the hole
- ⁠install ULPS link tube
Doing it this way will avoid the ULPS link tube from scratching the top of your ULPS chassis

 

A separate cover plug plate will neatly cover ULPS port if not in use. 

 

At least that’s how it’s envisioned at this point.

 

 

 

[Nenon]

On 4/11/2024 at 10:32 AM, AngeloVRA said:

This is where I am at the moment in finding a solution to shortest possible umbilical in a 2 box config

 

 

Screen Recording 2024-04-11 at 21.49.16.mov 45.69 MB · 17 downloads   Screen Recording 2024-04-11 at 21.52.10.mov 29.44 MB · 8 downloads  

 

 

 

 

design notes:

 

-  No big hole for EMI to leak in/out

-  would look nice from the front… no visible wires/cables

To disengage server from a ULPS chassis below:
- disconnect molex from Taiko ATX
- ⁠push molex into the hole
- ⁠lift server chassis

 

To install:
-  unscrew ULPS link tube from server 
- position ULPS cable
- ⁠place server chassis on top, threading ULPS cable through the hole
- ⁠install ULPS link tube
Doing it this way will avoid the ULPS link tube from scratching the top of your ULPS chassis

 

A separate cover plug plate will neatly cover ULPS port if not in use. 

 

At least that’s how it’s envisioned at this point.

 

 

Great. That should work.

[StreamFidelity]

Perhaps it is already known here, but if not then the following information may be of interest: 

Quote

The Olympus Server utilises the latest available computer hardware with the following differences to the Extreme:

-The Extreme has 2 Intel Xeon Scalable processors with 10 CPU physical CPU cores each for total of 20 physical CPU cores (40 virtual), 48Gb DDR4-2400 Memory, PCIe gen3 interfaces, uses PCIe gen3 M.2 music storage drives and runs Windows 10
- The Olympus has 1 AMD Epyc 4th gen processor with 4*8 physical CPU cores for a total of 32 physical CPU cores (64 virtual), 128Gb DDR5-4800 Memory, PCIe gen5 interfaces, uses MCIO (PCIe) gen5 U.2/U.3 music storage drives and runs Windows 11

 

Taiko Audio is therefore moving away from the Intel 2 core concept and is probably using one of these AMD cores:

 

AMD EPYC™ 9374F	32	64	Up to 4.3GHz	4.1GHz	3.85GHz	256MB	320W
AMD EPYC™ 9354P	32	64	Up to 3.8GHz	3.75GHz	3.25GHz	256MB	280W
AMD EPYC™ 9354	32	64	Up to 3.8GHz	3.75GHz	3.25GHz	256MB	280W
AMD EPYC™ 9334	32	64	Up to 3.9GHz	3.85GHz	2.7GHz	128MB	210W

 

[Nenon]

Met with Jord @Pink Faun at Axpona yesterday, and we talked in depth about his new USB card, and all the development and challenges he ran into before releasing it. 

 

It has supercaps that take about two minutes to fully charge. But you need to have power on startup, so the computer can initialize the chip and then you need to switch to supercaps power so quickly that the chip cannot detect it (think nanoseconds). 

[JackJohnson316]

49 minutes ago, Nenon said:

Met with Jord @Pink Faun at Axpona yesterday, and we talked in depth about his new USB card, and all the development and challenges he ran into before releasing it. 

 

It has supercaps that take about two minutes to fully charge. But you need to have power on startup, so the computer can initialize the chip and then you need to switch to supercaps power so quickly that the chip cannot detect it (think nanoseconds). 

Is this new pcie card full height or half height? Thanks 

[seeteeyou]

1 hour ago, StreamFidelity said:

AMD EPYC™ 9374F	32	64	Up to 4.3GHz	4.1GHz	3.85GHz	256MB	320W
AMD EPYC™ 9354P	32	64	Up to 3.8GHz	3.75GHz	3.25GHz	256MB	280W
AMD EPYC™ 9354	32	64	Up to 3.8GHz	3.75GHz	3.25GHz	256MB	280W
AMD EPYC™ 9334	32	64	Up to 3.9GHz	3.85GHz	2.7GHz	128MB	210W

 

Even 210W might be considered somewhat challenging for passive cooling during the boot-up process.

 

Wouldn't 8324PN @ 130W TDP be quite a bit more realistic for the constraint of a fanless chassis? Its wide-temperature feature might also work in a similar way when compared to extended temperature range of Apacer RAM with superior SQ

 

https://www.amd.com/en/products/processors/server/epyc/4th-generation-9004-and-8004-series.html

Quote

Explore AMD EPYC™ 4th Gen 9004 & 8004 server processors with enhanced performance, security, and efficiency for data centers.

 

[Exocer]

4 hours ago, Nenon said:

Met with Jord @Pink Faun at Axpona yesterday, and we talked in depth about his new USB card, and all the development and challenges he ran into before releasing it. 

 

It has supercaps that take about two minutes to fully charge. But you need to have power on startup, so the computer can initialize the chip and then you need to switch to supercaps power so quickly that the chip cannot detect it (think nanoseconds). 

Super interested in hearing this someday. It is well out of my price range but surely it’s up there with the best in terms of performance.