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

[austinpop]

24 minutes ago, Jud said:

 

DSD to DSD goes through PCM in between.

 

Ah OK. I think I was confused by this in the Roon DSP settings:

 

Also, now that I look at this, it suggests it's even more compute intensive.

[lmitche]

FWIW, with the volume control on in Roon, dsd64 is converting to PCM and by default is upsampled to 352 khz. It sounds great.

 

In Roon I can't find an upsample rate for pcm that sounds as good as NOS. The dac is also set to NOS mode, so there should be no additional upsampling.

 

There are sounds on well know tracks, lyrics, deep bass and gong like sounds that I have never heard so clearly before or never even heard before the NUC. Upsampling diminishes or smears these details at any setting.

 

A fanless case arrives on Wednesday, so any additional benefits there will be reported.

 

Ramboot with Audiolinux is easy to setup and sounds great.

[octaviars]

Are these test done right to see how sheilding works with and without JSSG? 

 

https://www.usaudiomart.com/forum/viewtopic.php?t=1469

 

 

[lmitche]

5 hours ago, octaviars said:

Are these test done right to see how sheilding works with and without JSSG? 

 

https://www.usaudiomart.com/forum/viewtopic.php?t=1469

 

 

No, very inexpensive low bandwidth tools were used.

[BigGuy]

5 hours ago, lmitche said:

No, very inexpensive low bandwidth tools were used.

 

Perhaps but aren't the results with the various configurations of interest in a relative sense?

 

Maybe someone with access to better equipment might do similar testing?!

[austinpop]

35 minutes ago, BigGuy said:

 

Perhaps but aren't the results with the various configurations of interest in a relative sense?

 

Maybe someone with access to better equipment might do similar testing?!

 

Hi BigGuy,

 

This thread has been successful by focusing on new findings driven by empirical, listening impressions. Of course, we are very interested to know why we hear what we do. But we are driven by the principle that if enough of us are hearing a positive impact due to a change, then it does exist, even if it is confounding, based on existing measurement regimes. It is the domain of the designers and manufacturers to figure out the why and to identify - and yes, even devise new - measurements to understand and explain what we are hearing. Hopefully that leads to better products, even cheaper products.

 

This thread has also been very successful by steering clear of the measurements discussion, because let's face it - most people here aren't technically qualified to debate this anyway. And it inevitably devolves into arguments and rancor.

 

If you'd like to debate a question like "do these measurements prove or disprove JSSG 360 works/doesn't work," I'd respectfully ask you do it on a separate thread.

[RickyV]

@kelvinwsy i noticed that you are using a Intel X-25E ssd. What is your experience with it? Better SQ and where did you buy it? In other Words is it worth the money.

[str-1]

Enthused by the recent reference clock comparisons undertaken by @limniscate, @auricgoldfinger and @austinpopI have been looking at the specs of both the Mutec REF 10 and the sCLK-OCX10, and with both there is one measurement that has left me wondering.  Are the “Aging” figures quoted of an order that would suggest performance deterioration would be audible within a 5-10 year period?

 

Taking for example the figures for the Mutec:

Aging after 30 days operation: 
< +/-0.0002 ppm (per day), 
< +/-0.03 ppm (1st year), 
< +/-0.2 ppm (10 years)

 

I realise these are very, very small deteriorations given that stability when shipped is quoted as < +/- 0.01 ppm.  But I am wondering how long it would be before I would be able to hear a difference between the aging unit and a fully run-in new one?

 

I’ve followed appropriate links from the index and done a search but have not been able to find any posts in this thread touching on oscillator aging.  Possibly because it is nothing to worry about?

[seeteeyou]

Looks like everyone could benefit from eMMC?

 

16GB eMMC Module Rock64 and Odroid-C0/C1/C1+/C2 …
https://www.amazon.com/16GB-eMMC-Module-Rock64-Odroid-C0/dp/B07C9HKKFN

https://www.pine64.org/?product=16gb-emmc

 

32GB EMMC Module
https://www.pine64.org/?product=32gb-emmc

 

64GB EMMC Module
https://www.pine64.org/?product=64gb-emmc

 

ODROID USB3.0 eMMC Module Writer
https://www.amazon.com/ODROID-USB3-0-eMMC-Module-Writer/dp/B07BHQ615Q

https://www.hardkernel.com/main/products/prdt_info.php?g_code=G152105300286

 

eMMC Adapter
https://www.amazon.com/ODROID-eMMC-Adapter/dp/B01LHJ2O38

https://www.hardkernel.com/main/products/prdt_info.php?g_code=G135415955758

 

eMMC Module Orange
https://www.hardkernel.com/main/products/prdt_info.php?g_code=G150825660587

https://www.amazon.com/8GB-ODROID-eMMC-Module-Android/dp/B0791411FH

https://www.amazon.com/16GB-ODROID-eMMC-Module-Android/dp/B0762XFZ5X

https://www.amazon.com/32GB-ODROID-eMMC-Module-Android/dp/B079ZQRVZK

https://www.amazon.com/64GB-ODROID-eMMC-Module-Android/dp/B079T22228

https://www.amazon.com/8GB-eMMC-Module-ODROID-XU4-Android/dp/B073WL8TW3

https://www.amazon.com/32GB-eMMC-Module-ODROID-XU4-Android/dp/B073WMFTDR

https://www.amazon.com/64GB-eMMC-Module-ODROID-XU4-Android/dp/B073WLXNWB

 

 

Yet another interesting choice with eMMC and it's starting at $89

 

LattePanda 2G/32GB - a Win10 Development Board (without Win10 product key)
https://www.amazon.com/LattePanda-2G-32GB-Development-without/dp/B01KJMDD9S

https://www.dfrobot.com/product-1498.html

 

LattePanda 4G/64GB - a Win10 Development Board (without Win10 product key)
https://www.amazon.com/LattePanda-4G-64GB-Development-without/dp/B01ID4HYE4

https://www.dfrobot.com/product-1404.html

 

MakerFocus Aluminum Alloy CNC Case for Lattepanda Development Board
https://www.amazon.com/MakerFocus-Aluminum-Alloy-Lattepanda-Development/dp/B06XRV32L2

 

Geekworm CNC Aluminum Alloy Case for Lattepanda 2G/4G Board - Black
http://www.dx.com/p/geekworm-cnc-aluminum-alloy-case-for-lattepanda-2g-4g-board-black-456374

 

 

[rickca]

22 minutes ago, seeteeyou said:

Looks like everyone could benefit from eMMC?

If we're going to use ramboot, why does eMMC make a difference?

[kelvinwsy]

2 hours ago, RickyV said:

@kelvinwsy i noticed that you are using a Intel X-25E ssd. What is your experience with it? Better SQ and where did you buy it? In other Words is it worth the money.

It is hard to say, my Samsung 120 gb SSD has Win 10 / Fidelizer 8.1 Pro / Process Lasso and fitted with the Elfidelity SSD III power (only) filter

while my Intel X25E has Win Server 2016Core Mode/Audioptimizer and SOTM II data and  power filter

Same  5v DC Linear Power Supply

So it is like Apples and Oranges

The remainder of the HQPlayer PC and the Quad Core Cherry Minix Neo NAA is the same..

on Ebay u can get a refurb Intel X25 for USD 60 

It is the reloading of the OS that is a pain

and u cannot hv the same OS on the same Hardware id no.. I think.. So 2 Win 10 pro 1 on Samsung SSD and same on Intel X. 25E wont work( u can download a trial copy of Win 10 Pro then u can!)

Finding a power filter for the Intel X25E is diffucult u can try the SOTM web shop ??

I found mine onEbay

Just a total OS / SSD Comparison I hv done

Intel X25E / WinServer2016Core/AO wins hands down-- It is quieter and calmer.. esp on crescendo music passages.. Sound images of sounds iike backing instruments in mid/left/right field background stands out more identifiable vs Win10Pro/Samsung

I will buy a Intel X25E for storing 64gb of my listening list and disconnect my 2Tb HDD to see if there isfurther SQ improvement. Doubt it as HQPlayer loads an entire playlist into memory b4 playback 

[austinpop]

3 hours ago, str-1 said:

Enthused by the recent reference clock comparisons undertaken by @limniscate, @auricgoldfinger and @austinpopI have been looking at the specs of both the Mutec REF 10 and the sCLK-OCX10, and with both there is one measurement that has left me wondering.  Are the “Aging” figures quoted of an order that would suggest performance deterioration would be audible within a 5-10 year period?

 

Taking for example the figures for the Mutec:

Aging after 30 days operation: 
< +/-0.0002 ppm (per day), 
< +/-0.03 ppm (1st year), 
< +/-0.2 ppm (10 years)

 

I realise these are very, very small deteriorations given that stability when shipped is quoted as < +/- 0.01 ppm.  But I am wondering how long it would be before I would be able to hear a difference between the aging unit and a fully run-in new one?

 

I’ve followed appropriate links from the index and done a search but have not been able to find any posts in this thread touching on oscillator aging.  Possibly because it is nothing to worry about?

 

Wow, that's a really good question. I've never really paid any attention to this metric. It would be really useful if someone from Mutec could jump in and answer this question.

 

What I'd want to know is if aging is of particular concern for reference clocks, or is a more general trait of crystal oscillators. Absent a compelling argument for the former, my working assumption would be that all oscillators are prone to aging. So if aging is a concern in a Ref 10 (or reference clock in general), then wouldn't it be of concern in most digital gear? Since everything from DACs to switches to USB devices to computers and endpoints have oscillators?

 

In summary, my assumption is that aging of the crystal in a reference clock should not be any more cause for concern than aging  for all digital gear.

 

But I would love to hear an expert veiwpoint.

[lmitche]

50 minutes ago, rickca said:

If we're going to use ramboot, why does eMMC make a difference?

Rick,

 

Yes you are right, the emmc is just used to load the disk into ram. Once done, it's not used. I know this as the orange disk activity light on the front of NUC is dark.

 

There are other alternatives of course including an SDCARD or SSD or USB stick or Optane stick if NVME is available. I don't know how much noise these make when idle. 

 

Nevertheless it is good to have a non-removable store for this as one needs to write the disk contents back to disk before shutdown or history and software updates are lost.

 

[seeteeyou]

Here's what Roy mentioned before regarding dormant drives

 

https://www.computeraudiophile.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/?page=359&tab=comments#comment-860030

On 8/8/2018 at 7:29 PM, romaz said:

While you could argue that I could have used an SSD drive in this situation and that the SSD would sit idle since the OS would be running completely from memory, my contention is that even a dormant SSD still generates noise.  In the end, the proof is in the listening.

 

The key in my post above was the word "everyone" in a sense that not everyone would get one of those (used) SLC SSDs with much lower noise than the others, and then not everyone would dive into Linux while running that entirely from RAM. Finally not everyone would go through the steps to work on FiraDisk / GRUB4DOS / iSCSI / PXE etc. in order to run the VHD image of Windows completely from RAM.

 

For an Intel NUC with Celeron / Pentium it's even more work to keep the VHD image (of 64-bit Windows since WoW64 needs more space) much less than 16GB with DISM / Dism++ / MSMG Toolkit / NTLite / WinReducer EX-100 / Win Toolkit etc. We'll see if Windows 10 Lean were getting any easier.

[rickca]

Yes I remember that comment from Roy.  I guess I am unconvinced that an idle SSD is a significant problem.  I just find limiting the platform to NUCs with eMMC is too big a constraint.

 

I must acknowledge @seeteeyou that you have the most awesome googling skills.  Respect.

 

 

[lmitche]

1 hour ago, rickca said:

Yes I remember that comment from Roy.  I guess I am unconvinced that an idle SSD is a significant problem.  I just find limiting the platform to NUCs with eMMC is too big a constraint.

 

I must acknowledge @seeteeyou that you have the most awesome googling skills.  Respect.

 

 

The July NUC announcements include 8th generation $500 NUCs with a quad core i7 processor capable of 4.5ghz at 28 watts and Optane capable. More than enough for hqplayer upsampling. Combined with an emmc NUC and an EtherRegen this could be a nice solution in the way Roy was suggesting.

 

I have to say that having a system in a 4x4x1.5 inch cube is liberating. I could fit 8 to 12 NUCs in the metal case holding it's linear power supply.

[seeteeyou]

1 hour ago, rickca said:

I just find limiting the platform to NUCs with eMMC is too big a constraint.

 

That's the thing about NUCs for now, we just dunno if there were any kinda "magic sauce" that might be baked right into its design. The idle storage device itself might (or might not) make a significant difference but fortunately it shouldn't be that challenging to test that out since standalone eMMC drives could be purchased from Hardkernel.

 

If impedance were actually a big deal because of a (much) smaller motherboard, let's do a comparison together

 

http://www.supermicro.com/products/motherboard/celeron/x10/x10sba-l.cfm

Quote

170 × 170 = 28900 mm²

 

https://www.intel.com/content/dam/support/us/en/documents/mini-pcs/nuc-kits/NUC7xJY_TechProdSpec.pdf#page=52

Quote

101.6 × 101.6 = 10322.56 mm²

 

http://download.udoo.org/files/UDOO_X86/Doc/UDOO_X86_MANUAL.pdf#page=15

Quote

120 × 85 = 10200 mm²

 

https://www.lattepanda.com/blog-2446.html

Quote

110 × 78 = 8580 mm²

 

https://www.lattepanda.com/products/1.html

Quote

88 × 70 = 6160 mm²

 

https://www.zotac.com/jm/product/mini_pcs/pi225#spec

Quote

95.4 × 63 = 6010.2 mm²

 

http://www.up-board.org/ai-edge/up-core-plus/

Quote

56.5 × 90 = 5085 mm²

 

http://www.up-board.org/upcore/specifications-upcore/

Quote

56.5 × 66 = 3729 mm²

 

https://wiki.solid-run.com/lib/exe/fetch.php?media=ibx:microsom:docs:sr-usom-ibx-assy-rev1.0.pdf

Quote

52.8 × 40 = 2112 mm²

 

https://wiki.solid-run.com/lib/exe/fetch.php?media=ibx:carrierboard:docs:sr-ibx-solidpc-datasheet.pdf#page=3

Quote

100 × 80 = 8000 mm²

 

SolidPC Q4 is the tiniest but it's just the system on module (SOM) part, its official carrier board turned out to be much larger while it's gonna cost a fortunate to design our own board.

 

Next in line is UP Core from Aaeon (a member of the Asus group) as shown below, I could find one clock on the left plus two more on the right

 

https://www.mouser.co.uk/images/microsites/upcore-3qtr1bd1.png

 

SOtM tX-USBhubIN could be added via USB 2.0 pin header, Ethernet is also available on this carrier board

 

https://up-shop.org/up-core/202-up-core-carrier-board-high-speed.html

 

Alternatively, 4 Gigabit LAN ports could be added with this $40 USB adapter that's powered externally via its 5V/2A DC input

 

https://www.amazon.com/IO-Crest-SY-HUB24047-Gigabit-Network/dp/B01N16C75R

 

The board itself costs just under $130 if 2GB of RAM were OK for a SUPER slim OS, replacing the clocks with better ones and we should be good to go

 

https://www.mouser.com/search/refine.aspx?Ntk=P_MarCom&Ntt=159430045

 

As usual it's even better to figure out what voltage regulators could be replaced with LT3045, though we might end up destroying the board or something.

[rickca]

58 minutes ago, lmitche said:

The July NUC announcements include 8th generation $500 NUCs with a quad core i7 processor capable of 4.5ghz at 28 watts and Optane capable.

Yes I read this article and I was really intrigued.  I think it's the most feature-rich NUC yet.

https://www.anandtech.com/show/13113/intel-officially-launches-bean-canyon-nucs-with-coffee-lake-u-processors

 

I don't think your going to see eMMC on a machine with a 28w CPU.    

[rickca]

40 minutes ago, seeteeyou said:

That's the thing about NUCs for now, we just dunno if there were any kinda "magic sauce" that might be baked right into its design.

I'm skeptical about exactly that speculation.  NUC motherboards are nothing special that I can tell.  They certainly don't have the sophisticated power design or BIOS options of bigger motherboards.  I suspect most of the secret sauce is the low latency Linux with ramboot.

 

Now if only Apple cared about audio, they could really do something wonderful with the new Mac Mini.  But they don't.

[seeteeyou]

17 minutes ago, rickca said:

They certainly don't have the sophisticated power design or BIOS options of bigger motherboards.

 

The BIOS part should be nothing to write home about since the latest release seemed to be kinda messy. However, I really dunno much about the sophisticated power design of bigger motherboards since everything is coming with switching regulators by default. Other than spending an arm and a leg on a custom-made motherboard, the best we could do might be finding something that's similar to Supermicro X10SBA-L etc.

 

https://www.computeraudiophile.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/?page=313&tab=comments#comment-813712

On 4/26/2018 at 4:25 PM, nvitorino said:

Hi everyone,

 

I can provide a bit of feedback on this based on our experience with the Supermicro X10SBA-L. On this board there is a significant difference in sound quality when powering both the ATX and the 12V Input compared to 12V alone. If you connect the ATX supply, the board will not use it's own switching regulators to convert to 5V and 3.3V from 12V. The interesting part is that it still uses the 12V rail on both the ATX and the 12V input separately if the 12V input is available. We've tested 3 configurations:

 

1. ATX only (using a Pico-PSU with a clean 12V input)

2. 12V only

3. ATX + 12V

 

1. and 2. sound similar in our tests. 3 sounds significantly better. At first, we were expecting 2. to be the best, as it removed the Pico-PSU from the equation. But seems the board itself will do the same job as the Pico and doesn't seem to do it any better than that.

We've fthen further tested with separate rails using the ATX only and compared to using separate rails for ATX + 12V input. The result is the same, adding both ATX and 12V results is clearly superior sound with this board. If you don't want / can't get separate rails for ATX and 12V, getting at least another 12V supply and connecting a Pico-PSU will still yield benefits compared to feeding the 12V alone. If you are using a separate power supply for the SSD, 1.5A on ATX and 1.5A on 12V is enough for this board.

 

Nuno

 

And then it's also a mystery when it comes to the noise patterns, though measuring lots of motherboards ourselves could cost even more than designing a custom-made one

 

http://hifipig.com/innuos-zenith-mk2/

Quote

“We had to go through a very detailed oscilloscope for measurements. We have literally tested dozens of system boards and the noise patterns you get vary wildly between them. Some are absolutely dreadful and should never be used for audio.

 

Once EtherREGEN is released, maybe more of us would become interested in the "Server + Endpoint" approach since a noisy server ain't such a big deal thanks to the network switch. If that were the case, we could possibly consider rebooting our project for crowd-funded motherboards (with eMMC?) and aim for something very simple with 1 PCIe slot for USB card + another slot for Ethernet card respectively.

[asdf1000]

2 hours ago, lmitche said:

The July NUC announcements include 8th generation $500 NUCs with a quad core i7 processor capable of 4.5ghz at 28 watts and Optane capable. More than enough for hqplayer upsampling. Combined with an emmc NUC and an EtherRegen

 

If the eMMC NUC is for use as a HQP NAA/endpoint (or RoonBridge), I've been using an Allo USBridge with eMMC for a few months with great results.

 

Handles DSD512 with both NAA and RoonBridge really well. The USBbridge would have a much lower noise output than a NUC... specs and some measurements are published on Allo's website.

 

I need to keep things fanless since my Roon Core and HQPe server are in the listening room (not idea but has to be that way for now). I haven't seen a fanless NUC build yet that can do all the poly-sinc-2s filters to DSD512. We might be a year or two away, but it will come. This one can though, 35W max:

 

https://www.atlastsolutions.com/fanless-tx-series-pc-core-i7-7700t-kaby-lake-8gb-250gb-860-evo-ssd-asus-h110t/

 

[lmitche]

EM2016,

 

My guess, and it is a guess, is that Akasa will build a fanless case for the 28 watt NUC. The 8th gen NUCs were just announced in July so it may be a while.

 

In the meantime the $220 spent on the June Canyon NUC and $67 on the Akasa case seems more than reasonable to me, especially as it can be repurposed as a NAA DSD512 or Roon bridge device. I'll test the two PC solution using my 6700k upsampling machine and June Canyon NUC in the next few weeks. I have never had a two PC Hqplayer solution work successfully here, so I am dubious.

 

I don't know much about the Allo product. It looks like a RPI that has a second board to fix the shared USB/NUC problem. Is that right? Will it run roonbridge or NAA at dsd512 rates?

[taipan254]

On 1/2/2017 at 2:49 AM, austinpop said:

b Roon inside the DX is to have Roon Core talk to Roon Ready directly. Think of this scenario as Roon playing out the server-c

 

I'd like to thank everyone on here, especially @romaz and @austinpop, for sharing so much useful knowledge. 

 

I'm very late to this party - I recently bridged my RasPi with Allo DigiOne to my main media PC, which is feeding a Chord 2Qute, LFD LE V amp and Kudos X2's as well as a headphone rig comprised of a Schiit Audio Lyr 2 driving beyerdynamic DT 1770 Pro headphones. Sadly my power conditioning is limited, with a simple Furman strip (though I love my cheap ifi iPower PS for the RasPi!). 

 

I'm thrilled with the results provided by this simple and cheap tweak (cost = $50 for a new NIC). 

 

I guess I need to start saving so I can build a low-power media server with the SOTM SCLK-EX connected to one of their USB cards! And an isolation transformer! And a few nice linear power supplies for upstream components! And new speakers! And (eventually) a new DAC! ? Or maybe I should learn how to build everything myself?!?

[str-1]

19 hours ago, austinpop said:

 

Wow, that's a really good question. I've never really paid any attention to this metric. It would be really useful if someone from Mutec could jump in and answer this question.

 

Also posted a slightly extended question on the Mutec REF 10 thread - https://www.computeraudiophile.com/forums/topic/31857-mutec-ref-10-masterclock/?do=bookmarkComment&comment=867182&csrfKey=b2011515f303be04fa0de62e7bef68b1&isButton=

 

[CJH]

Just wanted to give Hurka some props for his never ending efforts for great sound and sharing.  Been using an AGM battery (12V) regulated by Paul Hynes board (7V) to power my mRendu, IR and Micro iDSD black. My assumption was that powering the PH from a low impedance, high current, low noise battery had to be better than from rectified AC. Wrong! I replaced the battery with another PH regulator this one powered by rectified AC. So basically this is my version of a double regulated PH. The sound is better in every way. Even my wife provided an unsolicited comment on the improvement. Also wanted to add my amazement at the iFi Micro BL that continues to sound better each time I make an improvement in front of it---a bargain at $600. YMMV.

CJH