DIY Project High Performance Audio PC with high quality wiring

[StreamFidelity]

23 minutes ago, dminches said:

Are you still using 2 Keces P8s?

 

Yes. Both driven with HQPlayer DSD 256x48, convolution and EC-Modulator now about 2 x 20V/1,7A. One Keces P8 ist good, two are better. 😄

[StreamFidelity]

4 hours ago, dminches said:

How high does the amperage get on each of the P8s?  I can definitely see how using 2 would be better, especially while running HQP.

 

@austinpop gave a good answer:

On 12/10/2019 at 10:23 PM, austinpop said:

Clearly, massive current reserves seem to be a benefit when driving the CPU (via EPS), and I believe the same to be true for ATX. It really does seem that a music server's current demand is peaky, so a high current supply with tons of headroom has a distinct advantage.

 

Of course, a Keces P8 is enough for me. Then it is currently about 3.4A at 20V. But the load is better distributed with two Keces P8. Two Keces P8 only get moderately warm. SQ (Sound Quality) is increasing - in my system and for my ears. This does not have to be the case in every system.

[StreamFidelity]

4 hours ago, ASRMichael said:

@StreamFidelity where did you get the 4 cable clips from? 2 wire going out of the case. I’ve tried a few but didn’t work. Yours look like they’ll do the job perfectly. 

 

I use JTDEAL cable clips. They come in different sizes.

 

GhentAudio cables are still used in the old reference Audio PC.

 

For our DIY project fis Audio PC, we decided to design our own DC cables consistently. Every plug and socket has an impedance. There is always a risk that contact resistance will form. Different materials are mostly harmful to the sound. The plan is a direct connection to 6PIN Molex or 4PIN Molex. Stay tuned 🙂

[StreamFidelity]

19 hours ago, sandyk said:

There will only be a benefit to doing this with the Molex connectors if both Male and Female use gold plated contacts

 

I agree that this would be optimal. For this reason Bernd wanted to solder the cable directly to the mainboard and to do without Molex plug / socket altogether. Maybe we'll do that too.

 

At the moment it is a compromise. A high-quality cable should at least have a high-quality connector. And we will test whether it works. 🙂

[StreamFidelity]

2.4 Music player/renderer

 

An Audio PC is so flexible that any software, whether operating system or music player, can be tried out and used. The model and software presented here therefore describes only one of the many possible ways. 

 

In a 2 box system I prefer the variant that the highest computing power is provided by the Audio PC. Here the Control PC runs 24h as a music server (NAS) and consumes low computing power. As a Control PC, for example, a somewhat older and perhaps already existing model can be used. For music operation, the Audio PC is switched on and renders the data in best resolution and sends it to the DAC without any interference.

 

 

Roon is the first to be used. The software can be divided into three parts:

 

1. Roon Core
In the model above, the Roon server is installed on the Control PC. Headless, i.e. without graphical display, Roon works in the background and regularly scans the assigned hard disks for music files and automatically assigns them according to the stored database by music genres, artists, etc. Without DSP (Digital Signal Processor), i.e. without upsampling, room correction, etc., the computing load during music operation is extremely low, for example on an Intel Core i7 7700T (4x2.9GHz) with 1% CPU load. The load increases when new music files are scanned.

 

2. Roon Control
The Roon Server must of course be able to be operated. This can be done easily with apps that run on the smartphone, tablet computer or laptop. Which is just at hand. It is important that the music is always streamed from the Roon Server. With Roon Control, only the operation is actually carried out, so that the music transmission is always optimally maintained, even in the event of malfunctions in the Roon app. 

 

3. Roon Output
The music data is sent to the fis Audio PC via a special protocol RAAT (Roon Advanced Audio Transport).

 

4. HQPlayer
The fis Audio PC can render the music data for the DAC in highest quality, for example with the HQPlayer, without being disturbed by other applications. It would also be possible to use the Roon Bridge instead. In my experience, however, the increase in SQ is enormous with HQPlayer, provided the Audio PC has the necessary processing power. 

[StreamFidelity]

I don't want it to be boring. 😄JCAT has released a new USB Card XE. 

 

Some technical highlights:

- Emerald OCXO (oven controlled oscillator) ±0.005 (±5 ppb) stability 
- ASM3142 USB 3.1 host controller
- LT3045 linear regulators and filters eliminate noise interference from a PC

 

 

The USB Card FEMTO installed in the fis Audio PC is priced significantly lower and remains as a product line. 

 

We will try out the new USB card with the OCXO clock.  

 

Stays tuned 👍

[StreamFidelity]

27 minutes ago, The Computer Audiophile said:

What’s your industry affiliation?

I'm a banker. 😉Bernd is a sound engineer.

[StreamFidelity]

8 minutes ago, ASRMichael said:

When’s the JCAT ethernet XE coming out?

I don't know. Ask Marcin from JCAT. 

[StreamFidelity]

Hi guys,

 

like many others I am on forums to exchange ideas with other HiFi crazy people.  During the years of my hobby I have accumulated a lot of knowledge which I like to pass on. But I always hope with the necessary humility, because I am not a technician and I can be wrong.

 

When I presented my first Audio PC here, someone asked me if I could build one for him. I denied, because it was built by cirrus7 at that time. Now I have already built the second Audio PC myself. If one banker can do it, then everybody can do it. 😄That's what this thread is for.

 

But I don't want to exclude the possibility that I will turn my hobby into a profession in the future. I would then officially announce that.

 

Audiophile greetings

 

Gabriel

[StreamFidelity]

The following PCIe card arrived: 10Gtek® 10GbE PCIE network card. I installed them right away, see the picture below.

 

 

The first functional test ran smoothly. Below I have drawn a sketch of the fibre optic cable route. I could do without additional boxes, because the SOtM Switch has two SFP ports. It immediately accepted the SFP transceiver (SF Intel E10GSFPLR SFP+ 1310nm Transceiver Module). Via fiber optics (SF  Single Mode Fibre Patch) it goes to the other transceiver in the PCIe card. I was able to make similar settings in the network adapter as the other NET connections. Very nice.

 

 

The latency test is also absolutely perfect. Exactly zero interference at lowest latencies. I wouldn't have expected that from the start. 

 

 

Hearing tests are next.
 

[StreamFidelity]

A sound comparison of the different Ethernet connections follows. The following were used

 

1. Ethernet (copper) from the mainboard
Chip Intel I219V (1 Gigabit Ethernet Connection)
The network card adapter (driver settings) was not optimized.

 

2. Ethernet (copper) from JCAT NET Card Femto
Chip Intel I350 (1 Gigabit Ethernet Connection)
The network card adapter (driver settings) was optimized.

 

3. Ethernet (fiber) from 10Gtek® 10GbE PCIE network card
Chip Intel 82599ES (10 Gigabit Network Connection)
The network card adapter (driver settings) was optimized.

 

I have heard well-known pieces:

Misa Criolla: Kyrie from Mercedes Sosa
A classic with the choir and the bass drum. And of course with the catchy voice of Mercedes Sosa.

The Moon Is a Harsh Mistress by Radka Toneff
In the Burmester CD version, a low rumbling sound from turntable can be heard at the beginning (if the system is capable of it). 

Who I Am from Victory
Not a very well-recorded piece. I like to take it anyway, because the deep basses tend to boom when the connections are bad and the voice can be sharpened.

 

To make a long story short: The difference between the LAN connection from the motherboard to the PCIe cards was immediately audible. Less spatiality, the voices in the choir were not differentiated and Victory's voice had the unpleasant sharpness. When comparing the PCIe cards fiber optics and copper I would have expected a clear difference in favor of JCAT. What should I say? My relatively inexpensive fiber optic connection (2m) with a power supply from the mainboard and a "normal" 25MHz clock was absolutely on par. I wouldn't have thought so. 

 

My amateurish explanation: The galvanic isolation via fiber optics seems to make a big difference in my system. The clock is probably not that important on the LAN side, because in the end it goes via USB to the DAC. For systems that go to the DAC via Ethernet, it's probably different.  Maybe the higher bandwidth of the Intel chip 82599ES (10 Gigabit Network Connection) has its share. My SOtM switch is limited to 1 Gigabit and so the fiber optic controller has more headroom. As always, my findings and assumptions apply to my system and my ears!

[StreamFidelity]

On 4/8/2020 at 2:56 PM, StreamFidelity said:

I don't want it to be boring. 😄JCAT has released a new USB Card XE. 

 

Some technical highlights:

- Emerald OCXO (oven controlled oscillator) ±0.005 (±5 ppb) stability 
- ASM3142 USB 3.1 host controller
- LT3045 linear regulators and filters eliminate noise interference from a PC

 

 

The USB Card FEMTO installed in the fis Audio PC is priced significantly lower and remains as a product line. 

 

We will try out the new USB card with the OCXO clock.  

 

Stays tuned 👍

 

Today the USB Card XE arrived. As usual from JCAT in simple cardboard box and low & high profile bracket. The latter is pre-assembled.

 

For the first time, the JCAT USB Card has a heat sink.

 

As usual, there are two USB ports and one port for the external 5V/1A power supply that I use.

 

Of course I installed the new USB card right away and it worked. The card has been playing for several hours and I listen carefully. Although it burns in my fingers to write now something to the sound. 😄But I hold back as usual. The sound report and a comparison to the JCAT USB Card Femto will follow. 

[StreamFidelity]

2 hours ago, WilliamWykeham said:

My issue is that my CPU temperature is pushing 95C when upsampling to DSD256, ASDM7EC, while StreamFidelity’s temp was reported in the low 50s.

 

Was the thermal paste compound also applied to the solid copper block? I didn't use the paste from HDPLEX. 

 

Maybe it is not a problem of assemblings, but of the settings in BIOS? My overclocking ends at 4.1GHz. Which consumers are still active on maindboard? SATA drives for example draw a lot of power. I have deactivated this. 

 

Such a remote diagnosis is difficult. The temperature should not go that high. 

[StreamFidelity]

13 hours ago, WilliamWykeham said:

I noticed in Task Manger that while a Roon window was active, the “3D” GPU graph would show 50%.  

 

Is the graphics driver up-to-date? How many processes are displayed in the Task Manager? Check Latencymon to see if audio can be executed without interference. 😉

[StreamFidelity]

13 hours ago, WilliamWykeham said:

I have 70 processes running after AO and your recommendations.  

 

It's already looking very good. 👍

 

I prefer the 2 box implementation. Roon consumes very few resources and can therefore be run on a NAS. This brings a lot of silence to my system.

[StreamFidelity]

14 hours ago, WilliamWykeham said:

Im leaning toward a fresh install from the start with Windows

 

You have big problems. NT Kernel can be anything. I'm afraid you'll have to reinstall if you don't want to spend hours looking for the problem. Then try it first without AO.

 

Are there still temperature problems? Does the CPU really sit firmly in the socket? I would go through that again and above all distribute the heat paste correctly.

[StreamFidelity]

JCAT USB Card XE vs. JCAT USB Card FEMTO

 

Here a comparison of the two cards. Above the smaller JCAT USB Card FEMTO and below the JCAT USB Card XE.

 

 

 

I didn't want to describe this sound impression immediately, but let it set first. Especially since Marcin from JCAT recommends to burn in 200h. From time to time I also have to slow down my enthusiasm. 😄

 

What I heard about the JCAT USB Card XE has exceeded my expectations: An audibly cleaner sound image, better three-dimensionality, more precise bass range, very fine resolution in the trebles. 

 

Yesterday I changed to the JCAT USB Card FEMTO again, but I couldn't listen any further. This morning I had forgotten this again and was surprised about the smaller space. I was disappointed. What a great relief when I noticed that I heared with the JCAT USB Card FEMTO. 😃After plugging it into the JCAT USB Card XE it was the magic moment again. A much better location of individual instruments. Very precise and yet very homogeneous. A flowing voice, especially from women, just beautiful to kneel down.

 

The inclined reader may think, how often I read this in the forums. It can't be that there are always new sound enhancements described. Was the system bad before? Certainly not. But like many others, I have the feeling that only a comparison with something better reveals the shortcomings of the old system. And admittedly. My previous sound descriptions of the different Ethernet connections were marginal in my listening test compared to this one. In terms of sound, the different USB connections have a similar dimension for me as my test with the room correction or with different speaker cables.

 

The JCAT USB Card FEMTO is certainly a very good card but the JCAT USB Card XE enhances the sound considerably. And maybe it's a key for the last meter to the DAC regarding the much improved clock and the better power supply.

 

Great job @Marcin_gps

[StreamFidelity]

The thread is "DIY project High Performance Audio PC with high quality cabling". It is time to write about the cables that I have now received. 😉

 

Bernd tried a lot and rejected some. For example, different types of gold-plated Molex pin connectors. Not all of them were really gold-plated. Compared to my previously used DC / ATX cables, the new ones snap in much better and hold up very well.

 

The 24PIN and 8PIN Molex cables have a marked running direction.

 

Bernd rejected a plug connection between the DC cable and the internal Molex PINs. Plugs and sockets have contact resistances. Why not give it up right away? Said and done! There are two DC cables on it, as we are optionally planning a double power supply (from two Keces P8) for the fis Audio PC.

 

It continues with the DC cable for the 5V connection to the JCAT USB and NET cards.

 

 

It was time to finish building the fis Audio PC. It was very difficult to attach the back wall with the three PCI Express cards. Overall, I'm very happy with the look.

 

More importantly, how does it sound? Patience, it's time for a burn-in phase. 👍

[StreamFidelity]

18 hours ago, Nenon said:

This is turning out really nice. 

 

Thank you! 

 

I sawed off the metal web above on the back plate, so the PCIe cards can be replaced quite easily. 😁

[StreamFidelity]

56 minutes ago, Gavin1977 said:

Any Intel equivalent for @StreamFidelity's 9900k?

 

On ASUS mainboards there are settings for the maximum power:

 

- I have the power consumption at 120W with max. 160W limited

 

Other possible settings support this:

 

- Set all cores to a CPU fixed frequency, for me 4.1GHz has proven itself.
- The CPU i9-9900K can up to max. 100 ° C operated. Way to high. I have a max. Temperature set at 95 ° C.
- Possibly undervolting with BLCK Aware Adaptive Voltage.

 

Look my BIOS settings.

[StreamFidelity]

23 minutes ago, Nenon said:

If you have chance, can you give us your subjective opinion which cable is most important?

 

That's a good idea. 👍I think I can report later this week.

[StreamFidelity]

Test Preparation!

 

To prepare for the hearing test with Bernd I have prepared the standard computer cabling and standard DC cables, which will be compared to fis Audio cables.

 

PC / ATX Cable

 

This affects the computer wiring inside. Left to right:
1. DC input 20V
Female DC (7.4 / 5.0) --- Molex 6P
2. ATX 24 PIN
Molex 24P --- Molex 24P
3. CPU EPS 8 PIN
Molex 8P --- Molex 8P

 

DC Cable

 

These are all connected to Keces P8. Left to right:
4. DC output 20V
Male DC (2.5x5.5mm) --- Male DC (7.4x5.0mm)
5. DC Output JCAT USB 5V
USB Type-A --- Male DC (2.1x5.5mm)
6. DC Output JCAT NET 5V
USB Type-A --- Male DC (2.1x5.5mm)

 

 

It is then connected as follows:

 

 

[StreamFidelity]

On 4/29/2020 at 6:39 PM, ASRMichael said:

I can’t find anything on Asus website 

 

Unfortunately I cannot answer this question either. The CPU i9-9900K doesn't support ECC.

[StreamFidelity]

Sound Report : Test PC/ATX Cable and DC Cable

 

Please be careful. This report is getting a little longer. 😄

 

Test requirements

 

When performing a test, it is always important to describe the requirements and the test procedure. The claim was clearly not a scientific one. No blind test was carried out either. Bernd and I describe a subjective listening impression. The fis Audio PC can sound completely different in another room with another system. We heard with the following devices:

 

 

First I did the hearing test alone and took notes. The next day Bernd did the same hearing test without knowing my notes.

 

We heard a small selection of well-known music.

 

 

All tracks were sent by Roon (in the Control PC) in native resolution to the HQPlayer (in the fis Audio PC) via Ethernet. In HQPlayer the conversion to DSD 256x48 was carried out with the extremely precise ASDM7EC modulator. Poly-sinc-ext2 was selected as the filter and poly-sinc-mqa-lp for HiRes. The room correction (convolution) was activated.

 

We carried out the cable test in the same order:

 

1. Test fis audio cable computer and DC completely
2. Test standard cable computer and DC completely
3. Exchange of individual cables
3.1 ATX 24 PIN (Molex 24P --- Molex 24P)
3.2 CPU EPS 8 PIN (Molex 8P --- Molex 8P)
3.3 DC Input 20V (Female DC (7.4 / 5.0) --- Molex 6P and DC Output 20V Male DC (2.5x5.5mm) --- Male DC (7.4x5.0mm))
We could only carry out this test with two cables, since the fis Audio DC cable is continuous:
DC Output 20V Male DC (2.5x5.5mm) --- Molex 6
3.4 DC Output JCAT USB EX 5V (USB Type-A --- Male DC (2.1x5.5mm))
3.5 DC Output JCAT NET FEMTO 5V (USB Type-A --- Male DC (2.1x5.5mm))

4. Test fis audio cable computer and DC complete with 2 Keces P8 (20V / 8A & 12V / 8A)

This is the ultimate expansion stage, with a separate power supply for the CPU.

 

Test execution

 

And now we finally get to the listening results. I would like to say in advance that I am a layperson compared to Bernd. Bernd is a sound engineer and therefore knows what to look out for. Our auditory impressions largely coincided independently of one another! I give the key words listed separately.

 

1. Test fis audio cable computer and DC completely

Me: defined deep bass, lots of spatial information (3D), 0.00% sharpening, great transients, lots of fine information

Bernd: Flawless, the sound is great, very good tonality, in this room it was the best one so far.

 

2. Test standard cable computer and DC completely

Ich: Bass thickened, imprecise, no spatiality, sharpening

Bernd: slow fat bass, no precision, hard sound on the piano

 

Interim conclusion

 

In this clarity we didn't expect the difference between standard and fis audio cabling. We completely agreed that we didn't want to listen to music with the standard cables. We estimate SQ (Sound Quality) increase with high-quality cables at a high +30% to + 40%!

 

Now we were very interested in which cables represent an essential sound brake. The Audio PC was again completely equipped with fis audio cable and then replaced individually with the standard cable.

 

3.1 ATX 24 PIN

I: sharpen

Bernd: there is a clear lack of musicality and speed, little space

 

3.2 CPU EPS 8 PIN

Me: less space, unsightly transients

Bernd: massive loss of dynamics, piano has hardness, less space, queasy bass

 

3.3 DC input 20V

Me: digital hardness, less space, less timbres

Bernd: very strong loss of dynamics, crooked sound, space ok but limited

 

3.4 DC output JCAT USB EX 5V

Me: digital hardness, sharpening, less space

Bernd: significantly less space, piano blaring, annoying, speed and audibility ok

 

3.5 DC output JCAT NET FEMTO 5V

Me: slight deterioration in sound

Bernd: a touch of space and dynamism is lost

 

Interim conclusion

 

While Bernd and I were again clearly apart in terms of the sound effects, there were different assessments regarding the degree of sound effects. I saw the strongest sonic effects equally with the following cables:

- 3.3 DC input 20V and
- 3.4 DC Output JCAT USB EX 5V

Then in descending order:

- 3.2 CPU EPS 8 PIN,
- 3.1 ATX 24 PIN
- 3.5 DC output JCAT NET FEMTO 5V

With the last cable Bernd and I noticed the least impact. Here I would like to point out the special feature of our setup. We transmitted the data to the DAC via USB. In systems with data transmission via Ethernet to the DAC, this can lead to a completely different classification.

 

Now to Bernd's order. He heard the biggest impact:

- 3.1 ATX 24 PIN

Closely followed by:
- 3.3 DC input 20V and
- 3.2 CPU EPS 8 PIN.

The least impact was:
- 3.5 DC output JCAT NET FEMTO 5V

And the cable
- 3.4 DC Output JCAT USB EX 5V
said Bernd, that runs outside of the competition. Anyone who buys such a high-quality card and then saves on the power supply is wasting a lot of SQ!

 

4. Test fis audio cable computer and DC complete with 2 Keces P8 (20V / 8A & 12V / 8A)

 

We removed for this wiring
- 3.2 CPU EPS 8 PIN
and added a second fis audio cable instead
- DC Output 12V Male DC (2.5x5.5mm) --- Molex 8 PIN

 

The circuit diagram shows the connections.

 

 

Me: a little more of everything, more space

Bernd: more space, more dynamics, more pulse speed, more resolution, very good voice.

 

Interim conclusion

 

We were both very excited. Bernd even more than me. I estimated the SQ increase at + 10% to + 20%, Bernd even at + 20% to 30%! Maybe I have already got used to the good sound, since I have only been listening to two Keces P8 for days. It was completely new for Bernd.

 

Summary of all tests:

 

It shows again how important a good power supply is. The electricity has to be transported by cable. The effects of good quality cables in an Audio PC are greater than we previously suspected. Bernd and I rated the cables as particularly relevant in terms of SQ

 

- DC input 20V and
- CPU EPS 8 PIN
 

With the cable

- ATX 24 PIN

 

our opinions differed somewhat in the prioritization, but not in the sound effects as such. There are 24 wires with different voltages for the mainboard, which require high quality.

 

Anyone who uses a high-quality JCAT USB CARD XE in the Audio PC should definitely choose the best cabling

- DC Output JCAT USB EX 5V

 

for the external power supply.

 

A direct power supply to the CPU, bypassing the DC / ATX converter, again brings a significant increase in SQ.

 

Everything as usual without guarantee. 😉

 

[StreamFidelity]

2 hours ago, Downtheline said:

I have a quick question, how many amps did the EPS  12v use?

 

Thank you so much! The power consumption is as follows:

Keces P8 20V = 1.5 - 1.6A
Keces P8 12V = 2.5 - 3.5A