DIY Project High Performance Audio PC with high quality wiring

[StreamFidelity]

8 minutes ago, Iving said:

Replace with:   Solarflare X2522 + RJ45 Transceiver   > ordinary ethernet cable > A Side etherRegen > DDC > DAC

 

Why don't you go into the ER with SFP fiber? This is a perfect galvanic separation and the bandwidth of the X2522 is used without interference.

 

Occupying an SFP port with an RJ45 transceiver is, as far as I know, a bad solution. The transceiver can transmit electrical interference into the copper cable.

[Iving]

11 minutes ago, StreamFidelity said:

 

Why don't you go into the ER with SFP fiber? This is a perfect galvanic separation and the bandwidth of the X2522 is used without interference.

 

Occupying an SFP port with an RJ45 transceiver is, as far as I know, a bad solution. The transceiver can transmit electrical interference into the copper cable.

 

Well - I not long bought a pricey Sablon ethernet cable for the A Side. I could switch it over to the B Side and sacrifice the cable that is there. Honestly I hadn't thought of what you suggest and, so, thank you.

 

Can you recommend specific products for your recommendation apart from the X2522? We're talking about less than 1m connection - probably c. 75cm.

 

I should add ... RedNet Dante requires a 1Gb network A Side - I expect this is either Auto or Fix in Device Manager/Network Adapter Settings.

[StreamFidelity]

14 minutes ago, Iving said:

Can you recommend specific products for your recommendation apart from the X2522?

 

Transceiver - Startech 1000BASE-EX SFP -SM LC-40KM with FS 2m (7ft) LC UPC to LC UPC Duplex OS2 Single Mode PVC (OFNR) 2.0mm Fiber Optic Patch Cable work very well for me.

 

The fiber optic patch cables are very inexpensive and must not be too short (at least 2m). This is related to the damping factor. Startech also has good cables. It must be yellow cables (Duplex Single Mode), otherwise the laser in the transceiver will not work.

[Iving]

2 minutes ago, StreamFidelity said:

 

Transceiver - Startech 1000BASE-EX SFP -SM LC-40KM with FS 2m (7ft) LC UPC to LC UPC Duplex OS2 Single Mode PVC (OFNR) 2.0mm Fiber Optic Patch Cable work very well for me.

 

The fiber optic patch cables are very inexpensive and must not be too short (at least 2m). This is related to the damping factor. Startech also has good cables. It must be yellow cables (Duplex Single Mode), otherwise the laser in the transceiver will not work.

 

tyvm

[Zauurx]

@Vivling2021, thanks for sharing information... inventive solutions.
But if I understand the schematic, it's to link the audio server under Euphony, directly connected in USB to the dac.
In the end, it is the USB driver that provides the SQ via the JCAT USB card.
But the quality of the audio processing depends on the latency of Euphony.
This is not a server/client scheme (?)

@StreamFidelity, your stream goes back through a Zyxel switch (with store-and-forward buffer) ?
The final SQ also depends on the switch, its power supply and its link to the streamer.

I left out the switches to test the direct connection to my Allo (which also defines the final SQ).
This solution is not sponsored by Jussy but it works.
I send DSD 256 / gauss-xla or PCM 1.412 to the Pegasus via a USBridge SIG under GentooPlayer.
The server under win10 LTSC is otpimized... as much as I can with the chosen hardware (20 windows processes... for the whole, between 34/37 depending on the set of programs and a driver  latency spread over the 6 cores between 0,03 and 0,07)... and yes, the hardest to stabilize is ndis.sys !
Sometimes, it drifts ! 🤪
Then the ethernet flow is sent in IPV6 without intermediary to the streamer. 
I also chose (listening) a mix of ethernet cable with a MED1005 isolator in the middle.

 

In the end the only switch is the server with the dual JCAT Net card (IPV4 in, IPV6 out)

[Vivling2021]

13 hours ago, Zauurx said:

@Vivling2021, thanks for sharing information... inventive solutions.
But if I understand the schematic, it's to link the audio server under Euphony, directly connected in USB to the dac.
In the end, it is the USB driver that provides the SQ via the JCAT USB card.
But the quality of the audio processing depends on the latency of Euphony.
This is not a server/client scheme (?)

@StreamFidelity, your stream goes back through a Zyxel switch (with store-and-forward buffer) ?
The final SQ also depends on the switch, its power supply and its link to the streamer.

I left out the switches to test the direct connection to my Allo (which also defines the final SQ).
This solution is not sponsored by Jussy but it works.
I send DSD 256 / gauss-xla or PCM 1.412 to the Pegasus via a USBridge SIG under GentooPlayer.
The server under win10 LTSC is otpimized... as much as I can with the chosen hardware (20 windows processes... for the whole, between 34/37 depending on the set of programs and a driver  latency spread over the 6 cores between 0,03 and 0,07)... and yes, the hardest to stabilize is ndis.sys !
Sometimes, it drifts ! 🤪
Then the ethernet flow is sent in IPV6 without intermediary to the streamer. 
I also chose (listening) a mix of ethernet cable with a MED1005 isolator in the middle.

 

In the end the only switch is the server with the dual JCAT Net card (IPV4 in, IPV6 out)

You are welcome, it is a one box solution, there is only one computer installed with both Solarflare and JCAT USB. I believe my cascade of switches(Buffalo GS2016 and Mikrotik 10G powered by LPSU) connected via optical fiber already provide signals with good isolation and timing to the computer.  While audiophile switches, server switches(two box solution) could provide signals with same good quality, they usually cost 5-10 times more. I think in most cases it is hard to justify the cost. 

[StreamFidelity]

19 hours ago, Zauurx said:

your stream goes back through a Zyxel switch

 

I'm using a Zyxel GS-108B v3 switch modified by a German engineer. In addition to an improved power supply, a Sellarz clock was installed.

 

Source: G-Switch auf Basis des Zyxel GS-108B v3 (German text)

 

I have tested many solutions, including direct cabling without a switch. It always came down to the same thing: the better the switch, the better the SQ.

 

In my opinion, a switch benefits the most from
- an improved clock
- a linear power supply
- a galvanic separation

 

This cannot be achieved with a standard switch.

[Zauurx]

2 hours ago, StreamFidelity said:

In my opinion, a switch benefits the most from
- an improved clock
- a linear power supply
- a galvanic separation

 

This cannot be achieved with a standard switch.

Thank you for this information.

I'm agree.
I also tweaked one of my two Meraki's and then simplified... for now!
I have not added a clock and I think that is the element that may justify putting a switch back into my system.
It's been discussed at length on the French forum I follow.
The more experienced are currently exploring Diretta... I'm sticking with my schema for the moment.

 

You tried a direct solution like mine ? 

JCAT Net : Eth1 --> ROON (IPV4)
                 Eth2 --> (IPV4) HQP (IPV6) --> USBridge

 

[erwinmitg]

On 8/11/2021 at 3:36 AM, Vivling2021 said:

The Solarflare X2522 is powered directly from LPSU... 

 

... It sounds wonderful... This is perhaps the first Solarflare X2522 powered externally by LPSU on this forum, this would reduce noise generated by the card going to the motherboard.  I think it would beat most copper based NIC solutions many times more expensive. 

 

That sounds really, really great. 

Thanks for sharing! 

 

My Solareflare get its power still from the motherboard - and I want to change this... 

 

You mentioned, that you use a riser cable with a sata connector. And this is connected to your hdplex (dc-atx) pcie port. 

But there, at the hdplex, is as I know only 12V, and the sata connector has 12V and 5V. 

 

How did you solve this? 

Is "12V only" sufficient for the Solareflare? 

 

Did you hear a change in SQ? 

 

Have a nice day, 

Erwin 

[StreamFidelity]

2 hours ago, erwinmitg said:

My Solareflare get its power still from the motherboard - and I want to change this... 

 

You mentioned, that you use a riser cable with a sata connector.

 

I would be careful with riser cables. I don't know of any riser cable of reasonable quality. Instead, you get a ribbon cable with a low cross-section, dubious conductor quality and insufficient insulation. I'm afraid you get more interference than you increase SQ with the separate power supply.

 

With a Taiko Audio 500W DC DC-ATX converter, the power of the mainboard and the PCIe cards connected to it are massively upgraded. It is expensive but a very good alternative.

[Vivling2021]

19 hours ago, erwinmitg said:

That sounds really, really great. 

Thanks for sharing! 

 

My Solareflare get its power still from the motherboard - and I want to change this... 

 

You mentioned, that you use a riser cable with a sata connector. And this is connected to your hdplex (dc-atx) pcie port. 

But there, at the hdplex, is as I know only 12V, and the sata connector has 12V and 5V. 

 

How did you solve this? 

Is "12V only" sufficient for the Solareflare? 

 

Did you hear a change in SQ? 

 

Have a nice day, 

Erwin 

Thanks! As you can see from the photos. I use a riser cable with 6 pin 12V power input that takes power directly from ATX Teradak LPSU.  I make sure the riser cable is as short as possible(20cm) and shielded to reduce interference as much as possible.  I think even though the riser cable may pick up some some noise, it is less noise than putting the entire card inside the computer box so close to the motherboard. Notice that the 12V input is directly from LPSU, not from a DC-ATX board which still has a SMPS in it. Of course I would imagine Taiko DC-ATX may sound better, but you would need pay much more.  I haven't done a comparison, as my fanless Streacom box is too small for both PCI-e card inside, and the heat from solarflare may turn it into a oven, degrading the sound quality.

[erwinmitg]

20 hours ago, StreamFidelity said:

 

I would be careful with riser cables. I don't know of any riser cable of reasonable quality. Instead, you get a ribbon cable with a low cross-section, dubious conductor quality and insufficient insulation. I'm afraid you get more interference than you increase SQ with the separate power supply.

 

With a Taiko Audio 500W DC DC-ATX converter, the power of the mainboard and the PCIe cards connected to it are massively upgraded. It is expensive but a very good alternative.

Thanks for the advice.

That lead me to this idea:

 

With s.th. like this 

https://www.ebay.de/itm/PCI-E-16X-Riser-Karte-90-Grad-Pci-E-Pci-Express-16X-Riser-Adapter-Karte-mit-U6S5-/154412970547?mkcid=16&mkevt=1&_trksid=p2349624.m46890.l49286&mkrid=707-127634-2357-0

you don't have (these "evil") cabels.

The capacitor can be easily changed to audiophile ones and through scratching out the not used power lines and soldering pins to the necessary 12V points you have probably the best power connection ever.

 

Next step better would probably going directly ON the Solarflare card (with Superregulators) ...

 

Taiko Audio with LPSU:

with a ripple of 0,1 mV, I guess even a much more cheaper LPSU like a little bit upgraded zerozone (with active rectifier e.g.) directly connected to the Solareflare is probably much better (because also without going over the mainboard = best mainboard is no mainboard) 

 

[erwinmitg]

3 hours ago, Vivling2021 said:

Thanks! As you can see from the photos. I use a riser cable with 6 pin 12V power input that takes power directly from ATX Teradak LPSU.  I make sure the riser cable is as short as possible(20cm) and shielded to reduce interference as much as possible.  I think even though the riser cable may pick up some some noise, it is less noise than putting the entire card inside the computer box so close to the motherboard. Notice that the 12V input is directly from LPSU, not from a DC-ATX board which still has a SMPS in it. Of course I would imagine Taiko DC-ATX may sound better, but you would need pay much more.  I haven't done a comparison, as my fanless Streacom box is too small for both PCI-e card inside, and the heat from solarflare may turn it into a oven, degrading the sound quality.

Thanks for the informations. 

 

That are really good news and makes it easy, only 12V are needed. 

 

Have fun... 

[MarcelNL]

ATD-link make some nice quality risers, no not high end audio grade 'full paranoia options on' but still way better than the 40 cent production cost crap you usually see.

[ASRMichael]

@StreamFidelity Any ideas the diameter of HDPlex heat pipes? 

[Dev]

2 minutes ago, ASRMichael said:

@StreamFidelity Any ideas the diameter of HDPlex heat pipes? 

 

6mm.

[StreamFidelity]

There is a new major update in my streaming solution:

 

First of all: I have no business relationship with Afterdark. I bought the Afterdark Switch used.

 

Hardware

 

The switch comes from the first production series (April 2021) and has a 10 MHz output to synchronize other devices with the clock. For reasons unknown to me, the new switches no longer have this option.

 

As the name of the product suggests, two industrial Buffalo BS-GS2016 switches are cascaded. The clocks were replaced by a Giesemann OCXO clock. This has three variably adjustable outputs, two of which clock the Buffalo BS-GS2016 synchronously and the third provides the 10 MHz externally. The specification of the "entry-level clock" Giesemann PRINCE installed here has slightly better values (phase noise -131 dBC at 10Hz) than the highly regarded Pink Faun Ultra OCXO clock (phase noise -130 dBC at 10Hz).

 

The connection between the two switches is ideally established via a fiber optic cable. But it is also possible via a LAN cable, in which case, logically, there is no galvanic separation.

 

The Giesemann Clock is supplied with power via a linear power supply built into the device. Instead, the two switches receive the power of 12V / 5A from an external power supply unit (for me from a Keces P8 12V / 8A).

 

The dimensions are 43cm width x 30.5cm depth x 9cm height (without feet). The front and back are made of carbon fiber.

 

 

The back contains all the connections. The picture shows the power connections with DC 12V (left) and AC 230V (right). All LEDs are lit when booting.

 

 

There is still a lot of space inside. On the left is the Giesemann clock with the power supply, in the middle the two Buffalo BS-GS2016 on top of each other and on the right the oversized MIFLEX audio capacitor (70 mm x 40 mm) is placed on a DC power unit.

 

 

My test circuit can be seen below. The signal (from the router) is fed into the upper switch. The signal is then forwarded to the lower switch, electrically isolated. From there the network is split up. The LEDs at the bottom left only show the connections of the upper switch. With this division, the lower switch is perfectly galvanically isolated and also has no LEDs that may be annoying. I think that's pretty clever.

 

 

Software

 

What I find particularly interesting is the configuration option. At Buffalo there is administration software for Windows that can be downloaded. First of all, the two switches can be found, which is done automatically.

 

Somewhat disappointing was the fact that both switches firmware were from 2014! The firmware update was easy. I had already saved the latest firmware on the NAS and was able to load it in this menu.

 

 

In addition, a few settings had to be made. For example, Flow Control must be activated. Only then could I connect my Solarflare NIC X2522 to the switch using a fiber optic cable.

 

 

Build quality

 

While the internal structure is convincing, I have already seen better quality in the case. The housing is sharp-edged. The feet are of poor quality. Some of the fixings look provisional.

 

 

This is my ethernet topology

 

 

Sound

 

The sound is sensational and the better SQ is immediately audible. A holographic space and a deep black background ensure excellent instrument localization. Even at low volume, the bass is physical and very rhythmic. This is the best switch I've ever had. A dream with the Solarflare NIC X2522.

[Topk]

On 8/9/2021 at 5:33 PM, StreamFidelity said:

 

A fan with an external power source is a good idea because electronic pollution from PWM is avoided.

 

Unfortunately, the other drawbacks of a fan remain. Even if a large case fan with 200 rpm makes almost no noise, the inevitable vibrations can seriously disturb the electronic components working in the high frequency range. That doesn't bother with word processing. But jitter or other electronic interference are problems in audio processing.

 

Fans tend to collect a ton of dust. Some people wonder about the large openings in the fis Audio acrylic glass cover. Since with passive cooling, the heated air moves from the bottom up, most of the dust remains outside.

 

The fis Audio PC is constructed in such a way that there are no moving parts. 😉

What if you put 5V USB 120mm low noise fans on rubber feet, placed outside the case, powered with power bank, blowing air towards the side heatsinks of the H5 case (therefore away from the MB and all the sensitive components)?

 

I am doing this now, and it dropped the CPU temperature of about 10 degrees Celsius resulting in improved SQ: I found out that lower CPU temp = fuller, more stable sound.

 

I bought this and it’s very low noise at the lowest speed setting. I cheaped out since it was a test, and might upgrade to Noctuas later on. I found out that - generally speaking - heat is the enemy of SQ in an audio PC (and for the LPS too).

 

[ASRMichael]

On 8/21/2021 at 2:43 AM, Topk said:

What if you put 5V USB 120mm low noise fans on rubber feet, placed outside the case, powered with power bank, blowing air towards the side heatsinks of the H5 case (therefore away from the MB and all the sensitive components)?

 

 

@Topk do you think there’s negative effects by blowing air through the mobo? Until I seen your post I have been blowing air through the case, but your comment has got me rethinking what I’m doing. 

[StreamFidelity]

On 8/21/2021 at 3:43 AM, Topk said:

I found out that lower CPU temp = fuller, more stable sound.

 

I agree. Then all that remains is the problem of the dust being raised.

 

I try to keep the CPU temperature low by undervolting, among other things.

[StreamFidelity]

Warning: this post may be a bit to long. 😉

 

New Fanless Chassis of HDPLEX H5 Version 3

 

Disclosure of my financial interests: I use HDPLEX products in the fis Audio PC!

 

Larry gave me the permission to briefly introduce the product innovations. The new heat sink made of solid copper will be able to cool Intel's 12th generation core. The new dimensions are very well described on Igor's LAB:

 

Source: Igor's LAB - Intel Socket LGA-1700 and LGA-1800 in detail - Exclusive data and drawings for the new CPUs from Alder Lake

 

Here are the new heat sinks from HDPLEX. All Pictures are from HDPLEX.

 

 

 

As before, the heat pipes are connected to the copper heat sink with an aluminum cover and massive screws.

 

 

The heat pipes dissipate the heat through the side wall. The new side wall increases the cooling surface with fine-toothed grooves. The picture also shows a section of the new cover with optimized cooling openings.

 

 

Intel's Core 12000 with E-Cores and P-Cores and Windows 11

 

After Intel with its 14nm production fell behind AMD with its modern 7nm production, a new processor generation is finally about to be released: the  12th generation Core (Core 12000) or also called Alder Lake. Provides a good overview of the various Intel CPU generations: Wikipedia Intel Core Comparization.

 

Hybrid CPU with E-Cores (Efficient) and P-Cores (Performance)

 

Efficient processor cores (E-Cores), as previously known from Atom, are supported by a varying number of large  performance cores (P-Cores). A maximum of 16 cores are possible, which are divided into 8 E cores and 8 P cores. With hyperthreading of the P-cores, i.e. the distribution of the load over two virtual cores, a total of 24 threads are possible.

 

Source: Overview of Intel Alder Lake (Image: Intel) published by https://www.computerbase.de/2021-08/intel-alder-lake-architektur/

 

Intel Manufacturing Process Intel 7

 

The Intel marketing department came up with something nice. The name Intel 7 suggests a 7nm production. In fact, it is a so-called "Enhanced SuperFin 10nm production". 😄

 

Intel says in its defense that the specification in nm (nanometers - billionths of a meter) with regard to the gate length, at which the original designation was aimed, has not been correct since 1997. In the future, Intel will only give the performance per watt.

 

Source: New designations for Intel's production stages (Image: Intel) published by https://www.computerbase.de/2021-07/nanometer-rennen-intel-fuehrt-neue-node-namen-wie-intel-7-und-intel-4-ein/

 

Intel Thread Director

 

The better the core assignment of tasks works, the more efficiently the CPU works. The new  Intel Thread Director monitors the threads and the status of the individual cores in the nanosecond range.

 

Operating systems like Windows tend to screw everything up. Therefore, in the HQPlayer, for example, there is the option of not leaving the core assignment to Windows. The conversion to DSD256 with the EC modulator would not run under Windows 10 if the HQPlayer did not take command. This should no longer happen with Windows 11, as it should work as the first optimized operating system for Intel Alder Lake.

 

Behind this is a hardware-based solution that Windows dominates in thread assignment. As an intermediate step, the Intel Thread Director already roughly assigns the tasks to the cores, which Windows then implements. This includes, for example, changing from large to small cores.

 

Source: Intel Alder Lake (Image: Intel) published by https://www.computerbase.de/2021-08/hot-chips-33-intel-alder-lake-steht-und-faellt-mit-dem-thread-director/

 

New Socket LGA 1700

 

LGA 1700 is designed as a replacement for LGA 1200. LGA 1700 has 1700 protruding pins to make contact with the pads on the processor. Compared to its predecessor, the LGA 1200, the LGA 1700 will have 500 more pins. Therefore the base is 7.5 mm longer: 37.5 mm x 45 mm. 

 

Summary

 

With Intel's new i9-12900K with 8 E-Cores (Efficient) and 8 P-Cores (Performance), the new Windows 11 operating system, which supports the new Intel Thread Director and the new fanless chassis from HDPLEX H5 Version 3, I have it in winter start a new project. I am going to build a new fis Audio PC i9-12900K Core and hear whether a further increase in SQ is possible with it. I am looking forward to.

 

[MarcelNL]

that copper CPU block and clasp look like a slightly better design that what I have waiting for my AMD build. Loooking forward to hear your experience with actual TDP of that HDplex as I expect that the current chassis is not capable to the Ryzen 9 with 105 W TDP.  

[MarcelNL]

for cabling (discussion about USB interlinks a bit earlier) I can also recommend Jupiter copper in cotton, saves a whole lot of time insulating UPOCC and the stripping ain't no walk in the park either,

 

I just finally reinstated (read repair) my RJ45 I2S link, and am again convinced that wire matters also in the digital domain.

 

 

[Topk]

On 8/23/2021 at 7:36 PM, ASRMichael said:

@Topk do you think there’s negative effects by blowing air through the mobo? Until I seen your post I have been blowing air through the case, but your comment has got me rethinking what I’m doing. 

Well… my idea is that since I’m blowing air towards the side heatsink of the case, I don’t create vibration from the flow of air on the cards, components, clocks etc (for example there is a clock on the Pink Faun I2S card and we know those clocks are sensitive to vibrations at the microscopic level). I just like to keep the mobo and inside of the H5 case as calm as possible.

 

I just want to emphasize again that the colder, the better the sound, everywhere, including CPU, chipset, LPS…etc. Generally  speaking, hotter=thinner, harsher sound. While I used to run fully passive, I’m now using strategically placed fans (not directly on the CPU of course like a CPU air cooler), and that temperature reduction improved the sound.

[Topk]

On 8/23/2021 at 9:35 PM, StreamFidelity said:

 

I agree. Then all that remains is the problem of the dust being raised.

 

I try to keep the CPU temperature low by undervolting, among other things.

One of my recent finding is that there’s a threshold where low CPU voltage does impact SQ. I would suggest you raise the voltage just as a test and listen, just to make sure you did not dip too low into that threshold….