Audiolinux Server configurations, Software, Hardware, and Listening Impressions

[lmitche]

29 minutes ago, austinpop said:

What files are in /etc/systemd/network?

 

22 minutes ago, bobfa said:

 

-rw-r--r-- 1 root root   34 Jan 20 22:06 bridge0.netdev

-rw-r--r-- 1 root root   41 Jan 20 22:06 bridge0.network

-rw-r--r-- 1 root root   38 Jan 20 22:06 eno1.network

-rw-r--r-- 1 root root   45 Jan 20 22:06 enp0s20f0u3.network

 

BORING - either get a room or exchange email addresses. Debug your network bridge somewhere else!

 

Outcomes only please.

[lmitche]

1 hour ago, mourip said:

 

I have been thinking about your reply ever since you wrote it. 

 

I would first like to say that I have switched out both of my Roon endpoints to run AL in ramroot with no internal disks. I was able to pull out 4 SSDs by doing this. I also have now obtained SQ that is on a level with my Dante based system, which is in my mind an excellent achievement. For me AL SQ is not only excellent on an absolute level but is is also huge bang-for-the buck.

 

Now to my point. It is obvious that your involvement with AS(CA) has really enlivened your efforts. It has hopefully also accelerated your business. You are extremely responsive to suggestions and requests, not to mention helping with issues that arise from those of us unfamiliar with Linux itself.

 

I have one issue with your reply above. When asked if the changes you made to allow cpu cores to be dedicated for audio processes would improve sound quality for reply was: 

 

                      "I am waiting for the results of user experimentation. I must find the time to listen myself... "

I find this a bit troubling. Usually audio application creators come up with a potential idea, test it, and then offer it to users as proven. In this case you seem to have come up with a theory that might improve sound quality and then offered it as "live" for users to test for you. For me the issue here is that most users will assume that if you put an idea into production it will have proven and tested merit. They may then spend time working with it and sorting out bugs when there has been no evidence that it will ever give benefit.

 

I think that it is safe to say that everyone here is primarily interested in improvements in sound quality. Many here may also enjoy the chase since AS has a lot of very inquisitive audiophiles and computer gurus. To be clear, I enjoy the chase and see this as new learning however it is also already clear to me that we are beta testers for Audiolinux. I accept that, however some users may be led down a trail of frustration and quit before ever finding out how good your product is.

 

I would like to suggest that you always have two versions of your software. One should be clearly labeled "beta" and the other will be your well vetted final product. Let subscribed users have both versions and ask for feedback. Currently all of your live products are really in beta. You are asking users to let YOU know if new ideas work or not and if they result in an improvement in SQ. 

 

I hope that you will see these comments as helpful. Keep up the excellent work.

 

Paul

Hi Paul,

 

I have no problem with Piero creating optional capabilities to enhance SQ as long as these capabilities don't interfere with existing SQ. Audiolinux has many "tweak-able" settings that I never use. That's cool, as long as they don't get in the way.

 

FYI, I don't plan to test the CPU task pinning functions offered in the new release. If someone else finds it makes a big impact, I may change my mind. In the meantime I'll ignore it unless it impacts SQ with my preferred configuration.

 

Piero has delivered six to eight releases since August and from my perspective has done a great job testing each release. I can't think of a single release where his testing has been a show stopper in any way.

 

As a software company professional, I can say that maintaining multiple releases in the field gets exponentially complicated and is a major burden over time. I'd hate to place that burden on Piero.

[lmitche]

A bit over a year ago most of us were firmly in the “one box” camp. Positive reviews of the Innuos Zenith MKII SE seemed to confirm the “one box” approach and raised my curiosity to learn how this was done. Later, looking at a pictures of the internals, the answer became obvious.  Innuos was using two separate regulators to power the CPU (EPS) and the motherboard (24 pin ATX) of one motherboard. Using a 19 volt Hdplex 250 watt DC to DC power supply for the 24 pin ATX, and by jumping the pins on my EVGA 1600 watt Titanium ATX SMPS to enable the 12volt CPU (EPS) output, a similar dual power scheme was achieved here. This yielded an impressive increase in SQ.

 

This past August the NUCs arrived. The early thinking was that this new two box solution meant that along with network isolation between the server and endpoint side, “bits would be bits”, and therefore server (dirty) side tweaks no longer mattered. The network isolation would shield the clean from dirty. With this idea in mind, a wifi network isolation scheme was configured here. Later the dual power scheme described above was dismantled with the big EVGA power supply used once again for both motherboard ATX and CPU power.

 

In December of last year the big 200 watt Hdplex arrived. This is the first linear power supply here with the capacity to power a CPU at close to a 100 watts. The separate 12 volt 10 amp rail, and 19 volt 10 amp rail enables many possible dual rail/supply power schemes. Remember the server and endpoint are connected via wifi.

 

Here is what I tested with observations.

 

• EVGA 1600 Watt ATX to 12 volt CPU(EPS), Hdplex 19 Volt rail to Hdplex 250 watt DC to DC for 24 pin ATX power

Notes: So much for the dirty/clean side idea, this sounds much better then EVGA alone. Image density has increased in the mid-range.

 

• Hdplex 19 Volt rail to Hdplex 250 watt DC to DC for 24 pin ATX power and CPU(EPS) 12 volt

Notes: Nice high end density, but the low mid image density and bass seem diminished.

 

•  Hdplex 19 Volt rail to Hdplex 250 watt DC to DC for 24 pin ATX power, Hdplex 12 Volt rail to CPU(EPS)

Notes: this is the one, improved image density almost 3d, big depth increase and improved ambience, just spine tingling!

 

It is likely that another Hdplex will find it’s way here in the next few months. I am anxious to test again with two separate LPSUs.  In the meantime, the last configuration is the most musical presentation heard here.

 

It should be said that in the final configuration, the 12 volt Hdplex output is used for powering the CPU.  This eliminated the possibility of powering my 12 volt network extender with the Hdplex so another 12 volt power supply was used for this purpose. Hence, two things were changed at once. It is possible that the second configuration was compromised with the wifi extender power connection.

 

Thanks to @Rickca for inspiring this effort.

[lmitche]

4 hours ago, Superdad said:

 

So if server side power supply changes are making a difference to a wi-fi connected end-point, then either something is getting "baked into" the wi-fi stream, or there are some other electrical paths (ground, AC, cables or whatnot) between server, endpoint and rest of system.

More research is required! 

Hi Alex, yes indeed, more research is required.

 

No question, there is a common earth ground between the two layers of floating ground ac isolation transformers on each side. On the NUC side three lps1.2s power everything and the NUC has an Iso Regen between itself and the DAC which shares a ground with the amp. The amp has a switch to make it float, but it sounds better grounded so the switch is in that position. All power supplies are ground shunted, although I haven't measured the Hdplex.

 

From the little I know, there can be no ground loops, as even the wifi extender is isolated from the server via an Ethernet isolation transformer. And yes, there may be some leakage current jumping through that isolation. Otherwise the entire system has no loops.

 

It does sound fantastic, so many thanks for the contribution of UpTone Audio isolation technology and the Hdplex multi-rail design.

[lmitche]

3 hours ago, Dev said:

 

Larry,

 

first of all thanks for the excellent report!

 

Is your server also connected to wifi ? Past couple of weeks, I moved my NUC streamer to wifi and couldn't be happier. Not sure if I will ever get expensive audiophile switches and lan cables - I just don't feel the need anymore to do anything on the network side for the streamer. My next experiment will be to move my server to wifi network as well and see if it benefits as well but I don't see any reason why would it not.

 

Looks like the hdplex is a great and flexible LPS. I had their earlier version but it caused issues (which many know about). Their new version seems to be stable and lot better performance-wise. How hot does it get ?

Dev,

 

The server is connected via a wired network with an isolation transformer to the network extender. The extender has two 5ghz radios one for the internet backhaul, the other for NUC and control laptop link. No other devices use the ssid.

 

I look forward to experimenting with the EtherRegen in various places in the current configuration.

 

The Hdplex temperature is around 40 degrees C on the surface of the top plate, so rather cool.

 

Agreed, the 5ghz wifi link does sound great.

[lmitche]

The Hdplex DC outputs are not ground shunted.

[lmitche]

1 hour ago, austinpop said:

Please correct me if I'm way off base here

More like off topic.

 

My write-up was about the $485 Hdplex 200 watt AC to DC power supply used in combination with the Hdplex 250 watt DC to DC supply to power a standard motherboard and processor. The latter is out of production and has been replaced with a 400 watt version.

 

Are there other power supplies that could be used for this purpose? Yes of course, many power supply manufacturers have solutions that could be used for this purpose including the Hdplex solution that you reference. Is it possible that they will sound better? Maybe, I wouldn't know until they are heard in my system.

 

I would be pleased to test and write-up any of these potential solutions. In the meantime the Hdplex 200 watt solution is very musical and enjoyable. With it's multi-rail design this power supply is also a great candidate for powering a NUC server and/or server side peripherals like network gear, or external hard drives.

[lmitche]

29 minutes ago, luisma said:

Thank you Larry for the update. I have been using wifi for streaming from the beginning and I am very pleased with the results. However would like to get validation and your opinion. 

I have an i3 fanless NAA connected with xover cable to a core i7 nuc running roon and HQPE on Ubuntu. The server has wifi on the other leg so it isolated from the network and only connected to the other device. Both are powered by uptone js2. 

This is different that your setup I know but would you consider that's enough isolation? The DAC is a denafrips connected USB to the naa. 

 

Luisma,

 

Good question. Until recently I had a similar setup. Let me explain.

 

My system is on the third floor. With no cable to the first floor cable modem and router, I have to use a network extender to get internet service up there. A side benefit of this is isolation from the cable system running into the house. It sounds like you have that isolation with your setup as well.

 

In the prior one box world, I ran a cable from the extender to the server. Later I learned that having a switch ala Mr. Swenson's Netgear fs105 with a ground shunt added value between the server and extender. I was looking forward to replacing this with the EtherRegen someday.

 

Then the NUC endpoint arrived with a WiFi card included in the kit. So rather than add another cable and bridging server NICs, I merely enabled the WiFi card to talk to the extender at 5ghz. This second, or third if you count the wifi backhaul, bit of isolation has worked well. SQ is awesome.

 

When the EtherRegen arrives it will be interesting to replace the NUC wifi link with a wired connection and to compare SQ.

 

I hope this makes sense. It is a bit complicated.

 

Larry

 

[lmitche]

2 hours ago, tapatrick said:

Hi Larry. Just trying to get clear on your wireless setup.

Is this how it works?

Actually it is much simpler. You have the endpoint right, it's 5ghz wireless to the extender. Now draw a solid line(wired cat6) to the server from the extender. A laptop also connects wirelessly for control purposes. Lastly a second 5ghz radio runs to the router on the first floor for internet.

 

That's it. I have no switch.

 

 

[lmitche]

17 minutes ago, tapatrick said:

Cool. Thanks Larry!

so the wireless just talks to the endpoint 

and I just reread this from your previous post ...“When the EtherRegen arrives it will be interesting to replace the NUC wifi link with a wired connection and to compare SQ.“ 

 

How do you power the extender? Are you using the supplied or have you diy’d a clean voltage?

The extender is powered with a 13.5 volt sigma 11 lpsu and with two 12 volt 1 amp ldovr lt3045 cards connected in parallel to deliver 2 amps of current. It won't run on 1 amp. The extender is a two 5ghz radio Linksys re9000. The 2.4ghz  radio is disabled.

 

Various lpsu configs were tested and this sounds best. FWIW, there is also an Emosystems isolation transformer between the server and extender.

[lmitche]

On another topic, Rickca did some research and learned it was better to power the 4  port Startech USB card at 12 instead of 5 volts.

 

Rick was right. The SQ jump is substantial. If you have one if these Startech cards this tweak is highly recommended.

 

I'm using a Hdplex 200 watt lpsu and it's lt3045 output at 12 volts and running that directly into the card via sata connector.

 

Awesome tweak, thanks Rick!

[lmitche]

55 minutes ago, Dev said:

 

Larry, you completely lost me - what does the Startech card do in your system ?

 

Between, for folks who are using wired Ethernet the GigaFOILv4 with active isolation seems like an awesome tweak. I haven't tried it but in some forums, folks with MSB renderer have noticed a significant improvement with it.  

Dev 

 

On my custom Audiolinux server the 4 port Startech connects the motherboard to my two music sources, a usb 3.5 inch hdd and the internet via my wifi extender. 

 

It also pushes a bit stream to the nuc endpoint.

[lmitche]

1 hour ago, tapatrick said:

When I got home I was trying to visualise this (being in the design business and looking at pictures all day  ) and still am not quite clear.

So is it A or B? Thanks!

Let's try math:

 

Wireless extender = one or more WirelessAccess Point(s) + Switch in one box.

 

Wireless router = Wireless extender + Wired router in one box.

 

A=B

[lmitche]

3 hours ago, tapatrick said:

I see, Thanks.  I'm asking as I'm trying to work out how I can make use of what I have on hand. Technically I have a range extender plus a separate Aqvox switch (which when added gave one of the best improvements I've had).

 

 

Patrick,

 

If you can draw what you have in place today (current state) along with any physical restrictions requiring a wireless link, it would be straightforward to make any suggestions for improvements (future state). If you have unused network gear on hand as well, please provide that list.

[lmitche]

2 hours ago, tapatrick said:

Here you go Larry. The NUCs are powered by Paul Hynes SR4 and the Switch and router by modified Teradak LPSUs.

 

 I suppose I could replace the cable between Router and NUC Roon server with the wireless range extender, but don't see how I can send wireless to the Endpoint without a more sophisticated Wireless extender hanging off the switch. There aren’t any other physical restrictions. Thanks so much for having a look!

 

I must say the SQ is superb with recently adding bridging and the latest AL updates using CPU/core isolation on the server.

 

 

Hi Patrick,

 

Well one can take network design to great lengths.  Here are several simple approachs and I am sure others may have different ideas for you.

 

1) You could duplicate what is happening here and replace the Aqvox switch with a 5ghz wifi extender with one radio connected to your virgin router, and the second connected to your NUC and laptop all at 5ghz, (if your virgin router supports 5ghz). This assumes your NUC has a WiFi card or that one can be made available. This assumes you disconect the cable between your server and router. The server is cabled to the extender in this design.

 

2) Simply get a WAP for the server and setup a link to the router, hopefully at 5ghz. Older routers often can be setup in this mode.

 

3) Get a second Aqvox or someday a EtherRegen and place it between the router and server.

 

All three of these designs rid your system from the noise coming from that virgin router and cable from the street.

 

Make sense?

 

 

[lmitche]

42 minutes ago, tapatrick said:

Thanks Larry! Indeed there are many ways to skin this cat..

 

I'll have a think and see which direction I will try out. Probably no. 2 as I'm going to wait for Uptone switch before buying any more network gear and I already have the TP Link range extender which will fit with my router.

 

I did have the switch between the router and everything else up until recently when I tried bridging on server after noting @austinpop reminder recently. When I initially added the Aqvox switch a year or so ago it was a major improvement over all. Moving it to where it is now (bridging server and endpoint) didn't really degrade the sound but improved it (probably because of the bridging), so I'm curious what a wireless link between router and server can do. To add, when I changed the power supply on the Router this also had a significant improvement. My whole system is on a big Topaz isolation transformer.

Option 2 will eliminate the "street noise" from the server and another potential  ground loop which should give you a nice bump in SQ again.

 

Plus it is a free option.

 

Once done, it is easy to swap the extender for the switch, again with the assumption that you have a WiFi card for the endpoint. The Linux setup for wifi is the tricky part with this approach.

[lmitche]

28 minutes ago, Bricki said:

I have now managed to get the isolated CPU cores working with 8gig of RAM with ramroot still working and running roon server. The CPU in my sonicTransporter i5 is the 5200u.

 

Initial impressions are a higher resolution sound with better image focus and textures, more space around the image and less compressed. It also sounds lower noise, smoother and more engaging. There is less bleeding between images and everything sounds cleaner in a natural way not a sterile way.

 

There is also more depth and you are able to better discern each sound to it's location in a seemingly larger sound stage (perhaps it's the same size but just seems like there's more space in there because the sounds aren't bleeding together)

 

I'm very very happy with this improvement 🤓

Bricki,

 

How did you allocate the cores?

[lmitche]

1 minute ago, Bricki said:

I selected half...the 5200u is only a dual core

Lol, thanks!

[lmitche]

1 hour ago, rickca said:

Good news for anyone with a 28W Bean Canyon NUC like NUC8i7BEH.  Akasa has announced a fanless case model called Turing.

https://www.anandtech.com/show/13943/akasa-turing-passively-cooled-chassis-for-intel-bean-canyon-nuc

 

I'd still be cautious until someone gives this a try.  The 28W NUCs have problems with thermal throttling in the original Intel NUC case.

Very interesting Rick, thanks for sharing!

[lmitche]

7 hours ago, Bricki said:

I have the isolated CPU cores working on my NUC7CJYH running roon bridge in ramroot with 4gigs of RAM with similar SQ results to the server.

 

If you haven't tried this already I highly recommend you put it on your to do list. I am very pleased with the results. 🔊😁👍

 

 

On my to-do list, thanks Bricki!

[lmitche]

3 hours ago, Poldi said:

Which storage for NUC-server with AL?

 

Hi,

 

I'm relatively new to this forum and have been reading the threads on the NUC with AL as endpoint and as server with great interest as I am using a sms-200 right now and this new approach seems to offer superior sound quality within a reasonable price limit.

 

Forgive me in advance if I'm asking something obvious (I'm not a computer expert) or if I missed something on this thread. So here is my question

 

What kind of storage do you guys use with your AL-NUC-servers?

 

1. internal storage? (wouldn't that be a possible cause of pollution? I mean, that's why the diskless USB-ramboot-option is advised, correct? Or does the diskless-approach only apply for the AL-NUC as an endpoint?)

2. external self-powered storage connected via USB? (Could a mini pc with PCIe-slot and an audiophile PCIe-USB-card give better results than a NUC?)

3. Storage on a NAS?

 

Thanks in advance!

I have a strong preference for option 2 having extensively tested the other options over the years. My findings have been duplicated by at least 5 other regular readers of this thread. To be clear, the solution is to use a 3.5 inch hard disk in an USB 3 enclosure into a USB 3 port.  Over here, the USB 3 enclosure is powered with a 12 volt high quality lps with double regulation.

[lmitche]

1 hour ago, Poldi said:

@bobfa @lmitche thanks for sharing your experiences!

 

I actually had a similar experience with my sMS-200. It sounded better with a more or less (see below) self-powered external disk attached via USB than with the music coming from my DS116 NAS via Ethernet. More relaxed and there is more weight to the sound.

 

However, I wanted to use a external SSD instead of HDD because I didn't want to have the noise in the living room. Although I found an enclosure for SSD that offers external powering (sorry link is in German):

https://www.amazon.de/ICY-BOX-Externes-Festplatten-Aluminium/dp/B01BHR3VPA

I noticed the device stays on even when the power supply is unplugged due to the fact that the single SSD I put in it uses so little power that the sMS-200 is able to power it via USB. So in the end unfortunately it's not really self-powered. That's why I thought about getting an Aqvox power supply:

http://www.myhifishop.de/Devices/AQVOX-USB-Low-Noise-Power-Supply-5V-1A-USB2-0-USB3-0-A-A-B-B-m::49.html

Unfortunately I would need a USB 3.0 B plug which they don't offer. But perhaps they'd customize it for me. Otherwise I would need an adapter.

 

Of course with a 3,5 external HDD you don't have this problem as it needs an external power supply in order to run.

@lmitche Or is there a different reason why you would prefer a 3,5 external HDD (to a 2,5 SSD)?

And would you advise to use an "audiophile" usb-card for the server? (SOtm, JCAT...)

NUCs don't offer a slot for PCIe-cards, right? Could something like this be an alternative to a NUC then?

https://shop.tx-team.de/Industrial-PC/Box-IPC/Fanless-IL1000::6.html

 

 

Poldi,

 

SSDs and all flash memory devices make an enormous amount of noise that corrupts SQ, so avoid at all costs. 

 

A server doesn't connect to a dac so a usb 3 port is what by you want anyway. I am not aware of a usb 3 audiophile pcie card. Startech has a nice usb 3.2 card that can be powered externally.

 

Lastly the nuc USB 3 ports sound terrific so no need for unnatural connections. 

[lmitche]

5 minutes ago, Bricki said:

After removing the internal os SSD in my server I was reluctant to add any local storage because I was finding it was just adding more noise... I had a Seagate 2tb portable HDD (not self powered) and I was finding that tidal streaming sounded better when there was no local storage on my server at all....tidal doesn't have a lot of my favorite music so I tried adding some local storage that didn't add any noise to my tidal streaming.... I tried a Samsung 128g USB 3.1 USB stick with very fast io speeds... even just having the stick plugged into my USB port on my server tidal streaming took a backwards step.

 

I am now using a micro SD card (obviously not much storage for an entire library) plugged into a USB 3.0 hub and I find that I can't tell the difference when it is plugged into my server or not. I have tried a tiny micro SD to USB adapter that plugs straight into the USB port but because it wasn't USB 3.0 (only USB 2.0) it sounded duller and flatter than the USB 3.0 hub...so @Superdad is definitely right that USB 3.0 hubs have better signal integrity. It was the same micro SD card

 

Thanks Larry for the suggestion of the self powered USB 3.0 enclosure with the HDD... I will try that in the coming months.

 

@lmitche have you found a difference between 5400rpm HDD and 7200rpm HDD? I am thinking of repurposing an older 5400rpm drive for this experiment so I don't want to hamstring the experiment from the start. Also I have run out of LPS's so I'm going to have to add a smps 😕....hmmm lots to think about and more money to spend 😅

I don't think there is an audible difference between 5400 vs. 7200 RPM 3.5 inch drives. Having said that I use a 7200 RPM drive and backup to 5400 RPM drives.

 

2.5 inch HDDs sound terrible.

[lmitche]

3 hours ago, Poldi said:

And aren't these "audiophile" PCIe-cards USB 3.0?

https://jcat.eu/featured/usb-card-femto/

https://sotm-usa.com/products/sotm-tx-usbexp-audiophile-pcie-to-usb-audio-card

Hi Poldi,

 

Yes, it looks like the latest versions of these cards support USB 3.

 

I stand corrected.

[lmitche]

4 hours ago, Superdad said:

9V, the other to 12V—giving you 19V

21?