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

[Nenon]

1 hour ago, ASRMichael said:

Hi, I am looking for advice/recommendations/thoughts with regards to new build I am planning. After the TLS issues I want to make sure I get this right this time..... Here are my initial thoughts...

• One box server
• Existing HDplex H5 case or other recommendations?, higher TDP?
• Mother board - Gigabyte Z390 Aorus Extreme MB, 16 Phases IR digital VRM (My understanding is lower noise output)
• CPU - Intel i9900k (run at 3.6gz base clock) (95 TDP)
• Existing JCAT Ethernet Card
• Existing data WD SSD drive
• RAM - 8GB - Apacer 2666 (What model? no idea yet?)
• OXCO Clocks - Modified Motherboard clocks for USB+Chipset by Fidelityaudio UK
• LPS - Custom Sean Jacobs 7 Rails in total for everything+Ghent PSU cables
• I'll be running Audiolinux or Euphony
• I'll be using Roon and HQplayer
• USB output to Innuos Phoenix USB reclocker
• I might add GPU at later date to help HQplayer CUDA offload (if required)

The above if more of less an Innuos statement, but would be more powerful, and 4k cheaper. 

 

Thanks in advance

 

My recommendation:

- Remove the "WD SSD drive" from this configuration.

- Add an Optane memory card for the OS.

- Get a Synology or similar NAS for your library. For example, the DiskStation DS218 models work great.

- Get a USB card.

   ---- Option #1 - Get a JCAT USB card and power up with a SJ rail.

   ---- Option #2 - Get a PinkFaun USB card with ultraOCXO clock and power up with a SJ rail. You may not need the Innuos Phoenix USB reclocker in that case (the only way to find out is to try it out). Essentially the PinkFaun USB with OCXO clock powered up by SJ rail would be similar to the Innous Phoenix. The only essential difference is that the Innuos powers up the OCXO with a seperate rail, while PinkFaun shares the same power supply rail for the USB and OCXO. You can also try the Intona USB isolator with this configuration. 

- Add a good network switch. 

- I have asked my friend who buys a lot of Apacer products for his company to find out what is the best Apacer non-ECC memory for cases like your. Will post when I get more info. 

 

 

[Chopin75]

I suppose u can also use Ryzen AMD that supposedly allow the use of ECC RAM, and it will support pinkfaun I2S bridge if u plan to use that. Intel does not support their I2S bridge. 

Are you booting CPU with separate rail? Perhaps that would work better? 

[ASRMichael]

Nenon thanks for your recommendations. Question what's the benefit of having a NAS over SSD drive? I actually have a NAS which is used for a 2nd backup. Cheers

 

Chopin75 thanks also, Yes CPU would be on separate rail. With regards to the Pinkfaun I2s Bridge, would this work using HQplayer upscaling to 128 DSD? 

 

Thanks Gavin1977

 

Cheers 

[Nenon]

1 hour ago, ASRMichael said:

Nenon thanks for your recommendations. Question what's the benefit of having a NAS over SSD drive?

I always found disconnecting the SSD drive from my computer to be an improvement no matter what the device is and how good the LPS powering it is. Those SATA devices generate noise that is audible. Stopped using SSD or HDD storage last year and never looked back. There are many discussions on the subject here in this thread. Some people found HDDs to be better. Others used HDD in an external enclosure via USB and separate LPS. There were discussions about specific SSD models/types that work better. Apacer industrial devices were used by some people. Others used arrays of SD memory cards. And some people has good result with specialized SATA cables like the Pachanko Sata "PURE Reference". Taiko uses PCIe-based SSD storage.

I personally think that removing any storage devices from my computer sounds best. But since I need to store the OS somewhere, I use an Optane M10 with excellent results. 

 

1 hour ago, ASRMichael said:

With regards to the Pinkfaun I2s Bridge, would this work using HQplayer upscaling to 128 DSD? 

Absolutely not. Go with USB for that.

 

 

[Marcin_gps]

Hi Guys, 

 

Those of you who are looking at Apacer industrial ECC RAM, make sure to order the brand new model with Samsung IC:

D31.23185S.001  DDR4 4GB ECC UDIMM 2666 CL19 WT

 

It's a clear step up from the previously available model. 

 

BTW - I'm in the process of setting up JPLAY / JCAT Club here at AS where I'll be talking about new exciting products. 

 

Best regards,
Marcin 

[rickca]

Intel has reduced prices on 9th Gen CPUs without integrated graphics (F and KF models) to better compete with Ryzen 3000.

Note that some of the base/turbo frequencies in the chart on page 1 of this article are incorrect.  For example, 9700F is listed as 3.6/4.9 GHz, it should be 3.0/4.7 GHz.  Anandtech is usually pretty quick to make corrections.

https://www.anandtech.com/show/14941/intel-announces-price-cut-for-9th-generation-f-and-kf-processors

[henke]

I just recieved 5 of theese LT3045 boards with 4 parallel LT3045 that I bought from an ebay seller:

 

 To early to say, but I think I like them. Thinking about changing the big cap which is an Jamicon @ 1000uF 30V.

Does anyone have any recommedation for cap change? They will be used in the USB chain mostly (two ISO regen and one Matrix element H pcie USB card) so current draw maybe will be at most 100mA per regulator board. Was thinking about Panasonic OS-CON 1000uF 16V but not sure.

[Superdad]

16 minutes ago, henke said:

Thinking about changing the big cap which is an Jamicon @ 1000uF 30V.

Does anyone have any recommedation for cap change? They will be used in the USB chain mostly (two ISO regen and one Matrix element H pcie USB card) so current draw maybe will be at most 100mA per regulator board. Was thinking about Panasonic OS-CON 1000uF 16V but not sure.

 

The caps on the output are much more critical than big one you are looking at on the input side.  But you don't want to go large there, and you want to stay with extremely low inductance parts.

This looks like another LT3045 board where they did not think to make use of the ILIM pin when paralleling. Too bad, as that's an easy way to cut the output impedance nearly in half.

Just shows that selection of parts is not everything: There is knowledge that goes into implementation for best performance. 

[RickyV]

On 9/25/2019 at 10:30 PM, RickyV said:

 

Till this week I used Corsair Vengeance 4Gb modules till I read jplay thread.

On the jplay site they also liked the Hynix branded modules and by pure luck I had a 4Gb G.skill ripjaws module which uses skHynix memory chips so I tried it for a few days. Indeed the G.skill modules sound better (crispier, clearer) then the corsair’s. We will see how they compare to the apacer module.

A week ago got my Apacer industrial ram module, SODIMM DDR4 2400 4GB CL17 Tc=-40...85°C. I thought I first let it burn in for three days and then compare it with the 4Gb G.skill ripjaws module which was all ready better than the Corsair Vengeance 4Gb module. So the G.skill ripjaws module sounded harsher and more up front. The Apacer module is much better, as said before more dense but I kept thinking that the noise floor seemed lower (cleaner). On top of that I thought it has a better, more linear, tonality and sounds more stable. Well worth the money.

👍

[lwr]

4 hours ago, Marcin_gps said:

Hi Guys, 

 

Those of you who are looking at Apacer industrial ECC RAM, make sure to order the brand new model with Samsung IC:

D31.23185S.001  DDR4 4GB ECC UDIMM 2666 CL19 WT

 

It's a clear step up from the previously available model. 

 

BTW - I'm in the process of setting up JPLAY / JCAT Club here at AS where I'll be talking about new exciting products. 

 

Best regards,
Marcin 

Is it a clear step up in technical specs only, or a clear step up in SQ?  If it’s a clear step up in SQ, is it as big a step up as from consumer grade memory to industrial grade memory? 

[RickyV]

37 minutes ago, Superdad said:

This looks like another LT3045 board where they did not think to make use of the ILIM pin when paralleling. Too bad, as that's an easy way to cut the output impedance nearly in half.

Can you tell us how one should  do that with LT3045’s? 

 

1 hour ago, henke said:

I just recieved 5 of theese LT3045 boards with 4 parallel LT3045 that I bought from an ebay seller:

 

 To early to say, but I think I like them. Thinking about changing the big cap which is an Jamicon @ 1000uF 30V.

Does anyone have any recommedation for cap change? They will be used in the USB chain mostly (two ISO regen and one Matrix element H pcie USB card) so current draw maybe will be at most 100mA per regulator board. Was thinking about Panasonic OS-CON 1000uF 16V but not sure.

The LT3045’s on my ldovr.com modules have a different writing on them.

[Superdad]

3 hours ago, RickyV said:

Can you tell us how one should  do that with LT3045’s? 

The technique is hiding in plain site in the data sheet.   John has wired LT3045s both ways and measured about half the output impedance when it is done right.  For some reason Alexy at ldovr does not believe it. Oh well...

[lmitche]

1 hour ago, Superdad said:

The technique is hiding in plain site in the data sheet.   John has wired LT3045s both ways and measured about half the output impedance when it is done right.  For some reason Alexy at ldovr does not believe it. Oh well...

Hi Alex,

 

FYI, the board pictured above does not look like one of Alexey's Ldovr.com brand boards. It is likely a Chinese knockoff of his design.

 

Larry

 

 

[Superdad]

2 hours ago, lmitche said:

FYI, the board pictured above does not look like one of Alexey's Ldovr.com brand boards. It is likely a Chinese knockoff of his design.

Sorry, I didn’t intend to confuse anyone.@RickyV mentioned his Ldover modules and it reminded me that Alexey has a couple of times questioned our assertion that the alternate paralleling technique results in significantly lower output impedance (it does).

It is not hard to implement, and we are surprised that almost nobody does so.

[Energy]

On ldovr.com LT3045 boards the ILIM is connected to GND since monitoring is not used. All the methods on the data sheet also suggest connecting it to GND.

 

If there was a way to cut the output impedance by half I’d do it to all of my LT3045 boards. Please enlighten us. I spent half an hour looking at the data-sheet

[marce]

9 hours ago, Superdad said:

 

The caps on the output are much more critical than big one you are looking at on the input side.  But you don't want to go large there, and you want to stay with extremely low inductance parts.

This looks like another LT3045 board where they did not think to make use of the ILIM pin when paralleling. Too bad, as that's an easy way to cut the output impedance nearly in half.

Just shows that selection of parts is not everything: There is knowledge that goes into implementation for best performance. 

I agree, plus the layout is lacking...

[Gavin1977]

10 minutes ago, Energy said:

On ldovr.com LT3045 boards the ILIM is connected to GND since monitoring is not used. All the methods on the data sheet also suggest connecting it to GND.

 

If there was a way to cut the output impedance by half I’d do it to all of my LT3045 boards. Please enlighten us. I spent half an hour looking at the data-sheet

Yes - I'm interested too - please spill the beans :-) Could be used to power components individually and interconnected for higher amperage?

 

I also found this on  https://dxpwr.com/

 

Modules with Input Output connectors and On/Off Switch, can be used stand alone or interconnected with other modules.

LT3045-1A5IO : 3xLT3045 Input/Output Board 1V-15V, 1.5A
LT3045-3A5IO : 6xLT3045 Input/Output Board 1V-15V, 3.0A
(coming soon)
TP4701-2A0IO :  2xTPS7A4701 Board 1.4V-20.5V, 2A
(under development)

 

[Gavin1977]

12 hours ago, lwr said:

Is it a clear step up in technical specs only, or a clear step up in SQ?  If it’s a clear step up in SQ, is it as big a step up as from consumer grade memory to industrial grade memory? 

Mouser (UK's APacer distributer) are unable to identify this memory.  D31.23185S.001 DDR4 4GB ECC UDIMM

 

Marcin - Are you able to provide a weblink?  Also, any non-ECC version and feedback on sound?

[bit01]

4 hours ago, Energy said:

On ldovr.com LT3045 boards the ILIM is connected to GND since monitoring is not used. All the methods on the data sheet also suggest connecting it to GND.

 

If there was a way to cut the output impedance by half I’d do it to all of my LT3045 boards. Please enlighten us. I spent half an hour looking at the data-sheet

probably page 26 of the datasheet?

[tapatrick]

A tale of SQ updates.

 

I have followed this thread since the beginning and implemented many of the discoveries kindly shared here - especially spending a lot of time tweaking and testing PSUs and NUC/AL setups. Eventually I settled on an NUC/AL Roon endpoint with IsoRegen out to a Sonore Ultradigital giving I2S into my PS Audio Junior DAC.

 

I went back to my Antipodes DXe Server as my source because I was having too much trouble with dropouts etc on an NUC/AL roon core and ran out of patience....  plus my wife gave up trying to listen to music!

 

Anyways I have been satisfied with this setup as I had achieved what I was seeking - great detail and micro detail, natural sound, wide deep soundstage and ambience, without harshness and have been happily listening to great sounding music without hitches for some time. Until an unexpected upheaval occurred and to the point of this post.

 

Being and owner of a PS Audio Dac we get regular software updates, so I loaded up the latest version 'Windom' on the weekend. Everything changed. After 24 hours or so I had to re-evaluate everything about my system as the SQ had changed so much. The software revealed much more more detail from the the music and from my system - exposing a level of harshness that I could not stand.

 

The software upgrade is remarkably revealing and is now exposing limitations elsewhere. After much experimenting I simply ended up with just a USB cable from my server to the DAC, restoring once again a very full, natural SQ with all the detail - everything the NUC and add ons etc were supplying.

 

I'm not sure where this leaves me and I will be doing more experimenting but this has certainly been an unexpected shift.

 

[henke]

16 hours ago, henke said:

I just recieved 5 of theese LT3045 boards with 4 parallel LT3045 that I bought from an ebay seller:

 

 To early to say, but I think I like them. Thinking about changing the big cap which is an Jamicon @ 1000uF 30V.

Does anyone have any recommedation for cap change? They will be used in the USB chain mostly (two ISO regen and one Matrix element H pcie USB card) so current draw maybe will be at most 100mA per regulator board. Was thinking about Panasonic OS-CON 1000uF 16V but not sure.

 

Got me wondering if the supposed LT3045 on my board are legit. It says:

"LGYP

N912

827"

On them not "LT3045" as my MPaudio board. 

 

Have anyone seen theese type LT3045 package that I have?

 

[Nenon]

26 minutes ago, tapatrick said:

After much experimenting I simply ended up with just a USB cable from my server to the DAC, restoring once again a very full, natural SQ with all the detail - everything the NUC and add ons etc were supplying.

 

By "my server" do you mean the NUC or the Antipodes DXe?

[tapatrick]

5 minutes ago, Nenon said:

 

By "my server" do you mean the NUC or the Antipodes DXe?

Antipodes Dxe

[Nenon]

24 minutes ago, tapatrick said:

Antipodes Dxe

 

And that sounds much better than your NUC directly connected with USB to your DAC? And it sounds better than your NUC USB --> IsoRegen --> DAC? I am just trying to understand if your USB started to sound better than you I2S or your Antipodes Dxe sounds better than your NUC. How do you power the NUC?

[RickyV]

57 minutes ago, henke said:

 

Got me wondering if the supposed LT3045 on my board are legit. It says:

"LGYP

N912

827"

On them not "LT3045" as my MPaudio board. 

 

Have anyone seen theese type LT3045 package that I have?

 

Usually I google these kind of numbers and most of the time you can get a data sheet but not this time. Even with the number + regulator.