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

[greenleo]

30 minutes ago, MrUnderhill said:

In my case via ebay, following a bit of research.

 

I have favourably compared the ones I have bought to:

Uptone LPS-1;

SBooster; and

iFI SMPS.

 

I have NOT compared them to JS-2; HD-Plex; Paul Hynes or Sean Jacobs. The last two of which I suspect I will explore once I have a setup I feel I will settle in for a few years .....unless it is the DCS NB, in which case I can avoid all this jiggery pokery!

 

Anyway:

 

NUC - arrived;

AudioLinux - bought & downloaded;

Memory - arrived;

USB - arrived.

 

.....and 19v R-Core LPSU bought. I may well have to adjust this down to 18V, to accommodate 15V LT3045s in due course; unless I can find some 19V 1A LT3045s, don't seem to exist.

 

M

 

 

How about LPS-1.2?  I look forward to your review between your 19V R-Core and SR4 as well.  That should be interesting.

[Johnseye]

I'm looking for some understanding of how latency impacts SQ.  I was having a discussion with some folks in the Roon forums and am coming up short in knowledge of what is potentially a major influencer of SQ.  One argument regarding Roon is that RAAT and buffering eliminates any latency issues.  RAAT is only applicable in a Roon Ready device as Bridge doesn't use RAAT.  Our discussions of a NUC have been with it running Bridge, but I'm wondering if the RAAT argument is too focused on network latency.

 

I'm guessing that the endpoint, regardless of whether it's Roon Ready or Bridge is going to benefit from lower latency because the processor and memory will still need to perform after the packets have been received from the Core.

[mourip]

On 10/31/2018 at 5:39 PM, novaca said:

Full of praise for Audiolinux and NUC here.
Has anyone compared Audiolinux with wtfplay?
Yes, the fundamental difference in features but pure sound quality?

 

Also, I can not find a direct comparison of NUC vs PC (based on Supermicro X10SBA motherboard - powering 12V + ATX) with the same OS. For use like NAA or single computer (non-upsampling).
Is NUC really better (as I have understood from this long thread)?

 

Boy. You are brave! Seriously...

 

I have gone to the AudioLinux web page twice full of enthusiasm and come back defeated each time. It may be the answer to our audiophile dreams but whoever put together that web page should know that it is intimidating and confusing even for someone who has worked in IT for 25 years.

 

Linux and "Plug n' Pray" are two ideas that I am never able to fully reconcile. I played with Linux for a while in the 90's and decided that it was a hobby OS. A bit like DOS from hell. It reminds me of the old analogy: A guy comes up to you and says " I am going to give you the best car in the world for free". "Great!" I say. He takes me out to the driveway and I find 10, 000 excellent parts all neatly set out and a full tool set with them. "All you need to do is assemble it" he says!

 

I believe that AudioLinux could be a great but if the creators want folks to embrace it they should rebuild their initial web page to explain how easy it could be and then use deeper links to detail how that might possibly be done.

 

I may still try it but currently I am looking for a machete on Amazon to cut through the website weeds and a good bottle of cabernet to steel my nerves...

[austinpop]

At this point @lmitche and @romaz are still experimenting, as are others, so their focus is there.

 

I echo your concern, of course. My hope is that someone will find time to write up a cheat sheet or primer to navigate.

[greenleo]

1 hour ago, mourip said:

 

Boy. You are brave! Seriously...

 

I have gone to the AudioLinux web page twice full of enthusiasm and come back defeated each time. It may be the answer to our audiophile dreams but whoever put together that web page should know that it is intimidating and confusing even for someone who has worked in IT for 25 years.

 

Linux and "Plug n' Pray" are two ideas that I am never able to fully reconcile. I played with Linux for a while in the 90's and decided that it was a hobby OS. A bit like DOS from hell. It reminds me of the old analogy: A guy comes up to you and says " I am going to give you the best car in the world for free". "Great!" I say. He takes me out to the driveway and I find 10, 000 excellent parts all neatly set out and a full tool set with them. "All you need to do is assemble it" he says!

 

I believe that AudioLinux could be a great but if the creators want folks to embrace it they should rebuild their initial web page to explain how easy it could be and then use deeper links to detail how that might possibly be done.

 

I may still try it but currently I am looking for a machete on Amazon to cut through the website weeds and a good bottle of cabernet to steel my nerves...

 

Actually not that difficult if you're using the GUI version.

[Bruce Orr]

Different Question:  is 50 or 75 ohm optimal for use with my REF-10?  Most of what I read in this forum or others seems to suggest 75 is better. Here’s what May/SOtM told me:

 

-> Normally 10MHz signal use 50 ohm cable. And word clock or SPDIF signal use 75 ohm cable. If you want to connect audio signal or clock signal in long distance like CATV signal, the 75 ohm is better. But, for most cases, audio application is not needed to connect in long distances. So 50ohm cable is better than 75 ohm cable.

 

Do you agree with this?

[auricgoldfinger]

7 minutes ago, Bruce Orr said:

Different Question:  is 50 or 75 ohm optimal for use with my REF-10?  Most of what I read in this forum or others seems to suggest 75 is better. Here’s what May/SOtM told me:

 

-> Normally 10MHz signal use 50 ohm cable. And word clock or SPDIF signal use 75 ohm cable. If you want to connect audio signal or clock signal in long distance like CATV signal, the 75 ohm is better. But, for most cases, audio application is not needed to connect in long distances. So 50ohm cable is better than 75 ohm cable.

 

Do you agree with this?

 

I think 50 ohm is more common in Asia, while 75 ohm is used elsewhere.  I have never seen any independent claims that one sounds better than the other.  You will find a much wider selection of 75 ohm cables, which is the real advantage as far as I'm concerned.

[Bruce Orr]

Thanks!  I will be using Habst which is available in either 50 or 75 so that won’t be a consideration with me, but I appreciate your reply. 

 

In your opinion there there is no real difference. Would you feel the same if a 75 cable is used in a 50 ohm REF-10 output to a 50 ohm input?

 

thanks. 

[auricgoldfinger]

6 minutes ago, Bruce Orr said:

Thanks!  I will be using Habst which is available in either 50 or 75 so that won’t be a consideration with me, but I appreciate your reply. 

 

In your opinion there there is no real difference. Would you feel the same if a 75 cable is used in a 50 ohm REF-10 output to a 50 ohm input?

 

thanks. 

 

You need to keep the impedance matched.  Otherwise, you will degrade the sound quality.  If your device has a 50 ohm input, then get a 50 ohm cable and use a 50 ohm REF 10 output.

[d_elm]

6 minutes ago, Bruce Orr said:

Thanks!  I will be using Habst which is available in either 50 or 75 so that won’t be a consideration with me, but I appreciate your reply. 

 

In your opinion there there is no real difference. Would you feel the same if a 75 cable is used in a 50 ohm REF-10 output to a 50 ohm input?

 

thanks. 

The bigest single mistake is to mix impedance of cables and connectors or the wrong cable system impedance for the device ports being connected. ie do not put a 75 ohm connector on a 50 ohm cable and visa versa, and do not use a 75 ohm cable assembly to connect 50 ohm ports.  In a pinch you can do any of this but the sound might be strange but do not plan a mismatch.

[Bruce Orr]

Thanks on the cable debate -50 v. 75 ohm- question asked and answered!

[tapatrick]

 A warning for those in the UK if you are looking for a fanless case for your NUC7CJYH don't buy from amazon as its 2x the price of the same one here: https://www.scan.co.uk/products/akasa-newton-jc-fanless-low-profile-case-8th-generation-intel-nuc-june-canyon-25-hdd-ssd

 

NUC7CJYH can be found here.. for £109

https://www.laptopsdirect.co.uk/intel-nuc-kit-nuc7cjyh-celeron-j4005-mini-barebone-pc-boxnuc7cjyh3/version.asp

 

[misterspense]

On 10/30/2018 at 6:17 AM, romaz said:

 

For a Roon Server NUC, the more power the better and so this NUC looks like it would be better than the i7 NUC I have.  The following NUC with an i7 that has a TDP of 100w may be better still but it costs more money:

 

https://www.intel.com/content/www/us/en/products/boards-kits/nuc/kits/nuc8i7hvk.html

 

If a small form factor isn't crucial, it may be better to go with an i7 or Xeon on a mini-ITX board and place it in a fanless Streacom case.  You could then power it with HDPlex's latest 400w DC-ATX converter:

 

https://www.hdplex.com/hdplex-400w-hi-fi-dc-atx-power-supply-16v-24v-wide-range-voltage-input.html

 

I'm still not sure I know where the sweet spot is.

 

By the way, I highly doubt that the fact that a i7 NUC has a higher TDP (thermal design power) will result in better sound. The 'higher power' should be read as 'higher computing power'. So it is able to do more computations than the Celeron processor. TDP or energy consumption will most likely only have a a negative impact, the switching voltage converters on the motherboard produce more noise at higher currents.

Note also that TDP is a sort of maximum energy consumption, not the consumption during idling. An i7 that is only running a stripped down Linux and some audioplayer like Roon bridge is probably only consuming 5W or something (for instance: https://www.anandtech.com/show/11198/gigabyte-gbbki7ha7500-kaby-lake-brix-review/6)

 

[simonklp]

1 hour ago, greenleo said:

 

Actually not that difficult if you're using the GUI version.

 

Exactly. Even if you use the Command Line version of AudioLinux, with the advice or support from Perio, it should not be a mission impossible. Cheers.

[greenleo]

13 minutes ago, simonklp said:

 

Exactly. Even if you use the Command Line version of AudioLinux, with the advice or support from Perio, it should not be a mission impossible. Cheers.

Is the Command Line version the headless version?

[romaz]

1 hour ago, misterspense said:

 

By the way, I highly doubt that the fact that a i7 NUC has a higher TDP (thermal design power) will result in better sound. The 'higher power' should be read as 'higher computing power'. So it is able to do more computations than the Celeron processor. TDP or energy consumption will most likely only have a a negative impact, the switching voltage converters on the motherboard produce more noise at higher currents.

Note also that TDP is a sort of maximum energy consumption, not the consumption during idling. An i7 that is only running a stripped down Linux and some audioplayer like Roon bridge is probably only consuming 5W or something (for instance: https://www.anandtech.com/show/11198/gigabyte-gbbki7ha7500-kaby-lake-brix-review/6)

 

 

My observations tell me differently.

 

For a NUC NAA running RoonBridge, absolutely, a Celeron or a CPU with a low TDP is ideal and I have been a proponent of this concept from the beginning.

 

For RoonServer, my recent observations suggest the opposite.  The more computational power, the better the SQ and this is a concept that is foreign to my way of thinking but a concept that is being championed by the likes of Pink Faun, Sound Galleries, Baetis, and others.  I was initially skeptical of this approach but I am now convinced without a doubt that this is the way to go with RoonServer.  I look at TDP as a sort of throttle to guide the computer builder in the proper cooling measures a CPU will require.  If you limit an i7 with a TDP of only 15 watts, while that CPU may be capable of much more, that low TDP will throttle it so that the CPU doesn't generate more than 15 watts of heat.  In some instances, Intel will allow you to downrate or uprate the TDP of their CPUs for those looking for greater computational power or less heat generation. 

 

With regards to computation power and SQ, greater power leads to lower latency but more noise and so it becomes important to figure out at which point too much noise or too high a latency becomes overly detrimental.  With my i7 NUC as a Roon endpoint, this NUC has an i7 with a TDP of 15w and yet it is sounding better than my Celeron NUC which has a TDP of only 10w.  This is not a result I would have intuitively guessed a few months ago and yet, this is what I am hearing.

 

I was criticized by a certain someone about making such a fuss about a microSD card but this is another good example.  As many know, I have 2 Mac Minis.  One Mac Mini runs it's OS off a low noise but high latency SD card and another Mac Mini runs it's OS off a high noise but low latency PCIe SSD.  The Mac Mini with the SD card results in a quieter and calmer presentation but presence and immediacy are lacking.  The Mac Mini with the PCIe SSD brings with it an annoying harshness that results in fatigue in the longterm but there is greater "joie de vivre" in the short term.  What this NUC combo provides is really a best of both worlds.

[kyrill]

Does nuc properties accept 15V?

 

[Monge]

@austinpopThanks for a great thread ?

And thanks to Larry and Roy for sharing their findings on NUC and many other things ?

I ordered a Intel NUC7CJYH and some Ballistix ram.

Now I only need an USB memory stick and Audiolinux. But which one for the headless Audiolinux and which size?

Thanks In advance.

Cheers

 

 

[romaz]

Right now, I'm too busy to respond to some of the questions asked of me or to provide more details about recent findings but I will provide this one detail.  With Piero's help, I was able to run headless AudioLinux on my Mac Pro and I was able to compare it against headless Linux on my HP workstation.  In both instances, AudioLinux is running completely in RAM and so the PCIe SSDs on both these machines are running idle and not being utilized.  I know this because my drive lights are not coming on at all.

 

As previously stated, my HP workstation utilizes an 8-core Xeon with a base frequency of 3.4GHz and 25MB of SmartCache and a TDP of 150w.  My Mac Pro utilizes a 12-core Xeon with a base frequency of 2.7GHz and 30MB of SmartCache and a TDP of 130w.  I can tell you that the Mac Pro is sounding better.

 

Is this because more cores is more important than faster CPU clock speed?  Is the bigger SmartCache responsible?  Both have monster TDPs but is the lower TDP of the Mac Pro's Xeon contributory?  I'm not sure but if I were to hazard a guess, I'm thinking that more cores do make an impact and the avoidance of higher clock speeds is resulting in lower noise.  Obviously, having a larger SmartCache can't hurt.

[romaz]

25 minutes ago, kyrill said:

Does nuc properties accept 15V?

 

 

Most Intel NUCs accept 12-19V with some of the commercial Intel NUCs accepting 12-24V.  In every instance where I have tested it, feeding a NUC a higher voltage results in better dynamics.

[lmitche]

18 minutes ago, romaz said:

Right now, I'm too busy to respond to some of the questions asked of me or to provide more details about recent findings but I will provide this one detail.  With Piero's help, I was able to run headless AudioLinux on my Mac Pro and I was able to compare it against headless Linux on my HP workstation.  In both instances, AudioLinux is running completely in RAM and so the PCIe SSDs on both these machines are running idle and not being utilized.  I know this because my drive lights are not coming on at all.

 

As previously stated, my HP workstation utilizes an 8-core Xeon with a base frequency of 3.4GHz and 25MB of SmartCache and a TDP of 150w.  My Mac Pro utilizes a 12-core Xeon with a base frequency of 2.7GHz and 30MB of SmartCache and a TDP of 130w.  I can tell you that the Mac Pro is sounding better.

 

Is this because more cores is more important than faster CPU clock speed?  Is the bigger SmartCache responsible?  Both have monster TDPs but is the lower TDP of the Mac Pro's Xeon contributory?  I'm not sure but if I were to hazard a guess, I'm thinking that more cores do make an impact and the avoidance of higher clock speeds is resulting in lower noise.  Obviously, having a larger SmartCache can't hurt.

It is also possible that the Mac is just a better engineered machine than the HP.

[elan120]

On 10/30/2018 at 12:06 PM, auricgoldfinger said:

 

Thanks for all the information and links!  I will wait to hear if the Molex Micro-Latch connectors you ordered are the correct ones.

Molex Micro-Latch connector is confirmed NOT to be the right one.  It is very close but just won't work, so you might want to check with May to find out if you plan on using the same connector to make your new cable.

 

As a side note, take a look at the picture below to see size comparison between crimp terminals for ATX, SPOX, and Micro-Latch (terminal in picture is the correct size for SOtM stock cable/connector). 

 

To work with Micro-Latch terminal is quite challenging due to its small size, so consider replacing SPOX header as one viable option.

 

Start from the left:  ATX pin, SPOX pin/connector, Micro-Latch pin/connector, stock SOtM cable/connector:

[auricgoldfinger]

2 minutes ago, elan120 said:

Molex Micro-Latch connector is confirmed NOT to be the right one.  It is very close but just won't work, so you might want to check with May to find out if you plan on using the same connector to make your new cable.

 

As a side note, take a look at the picture below to see size comparison between crimp terminals for ATX, SPOX, and Micro-Latch (terminal in picture is the correct size for SOtM stock cable/connector). 

 

To work with Micro-Latch terminal is quite challenging due to its small size, so consider replacing SPOX header as one viable option.

 

Start from the left:  ATX pin, SPOX pin/connector, Micro-Latch pin/connector, stock SOtM cable/connector:

 

Thanks for following up!  I'm sorry to learn that it's not the right connector.  I'll give your suggestions some thought.

[MrUnderhill]

NUC setup and running headless via RAM - music pouring forth.

 

I'll report back in due course.

 

Gents, thx for the hard work.

 

M

[rickca]

3 hours ago, lmitche said:

Is this because more cores is more important than faster CPU clock speed?

Have you looked at @PeterSt's experiences on his Stealth Mach III using from 8 to 22 cores?  I think he's saying that higher core count sounds better, but of course he is running XXHighEnd.