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

[lateboomer]

My existing network topology is :

 

Windows 10 Hqplayer laptop server ----> dlink gigabit router --> fmc fiber optic ---> NAA Linux ---> Usb Dac

 

Both fmc, NAA and router are powered by LPS.

 

So to try out the theory I followed the guide and configured Win 10 IPv4 as:

ip: 192.168.0.104

subnet: 255.255.255.0

default gateway and DNS: 192.168.0.1

 

Then I configured another unused TPlink wireless N router TL-MR3020 which connect to my home mobile internet. I disconnected dlink router and changed the network topology to:

 

Windows 10 Hqplayer laptop server on default ethernet port ----> fmc fiber optic ---> NAA Linux ---> Usb Dac

 

Hqplayer laptop server >>> wireless connection >>>> tplink router

 

Then I went to network adapter settings and clicked both the ethernet and wireless adapters and added them to bridge. I didn't change any settings inside the newly created virtual adapter.

 

With above settings, I could run Hqplayer and play direct to NAA with direct physical connection without dlink router in between.

 

Initially before using wireless to connect, I also bought one 3.0 usb to ethernat adapter, and connect the TPLink router to the ethernet adapter to create another ethernet port for the pc. So then I bridged those 2 ethernet ports together. Unfortunately, Hqplayer won't detect any NAA with this setup and it won't play.

 

I am not sure why bridging has to be between ethernet and the wireless for Hqplayer to work, where else ethernet to ethernet doesn't work. I am not savvy with network. However there is also an issue with this direct connection method. After playing music for about an hour, sometimes Hqplayer would lost connection with NAA and stops to play.

 

 

About the sound, initial impression is I think it sounds better without router in the playback chain. So thank you Romaz for this suggestion. The main improvement is better micro or macro dynamics, with soft and loud passages more distinctive and less stress in music playing.

[lateboomer]

I have followed Clipper settings that ip address set to 10.0.0.5 for direct ethernet connection between laptop default ethernet port of Hqplayer to client NAA. Amid there is FMC fiber optic setup between Hqplayer and NAA.

 

Then I set ip for usb to ethernet adapter which connected to Tplink wireless N router to 192.168.0.100, with default gateway of 192.168.0.1.

 

I able to bridge these 2 lan networks together. However, Hqplayer failed to see NAA and I guessed DHCP of TPlink router didn't assign ip to NAA. So then I connected to the wireless connection of that Tplink router and added it also to the bridge. Only then Hqplayer was able to see NAA and DHCP was working and assigned ip to NAA.

 

So I not sure how to get lan to lan bridging working and dhcp able to assign ip. I believe lan-to-lan bridging should be more stable, compared to lan-to-wifi bridging. As if I use lan-to-wifi bridging, Hqplayer would stop to play suddenly after around 1 hour. I have to continue refresh Hqplayer settings to pick up NAA again to resume play. I didn't suffer stutter and sounding was good. Just that Hqplayer would suddenly stop playing without warning. Anyone encounters similar issue?

 

My laptop is Win 10 Home edition. My NAA is Intel NUC running on Audiolinux.

[lateboomer]

When you establish the lan to lan bridge again, maybe try manually assigning the following IPv4 settings to the network bridge:

IP Address = 192.168.0.85 (assuming that's not assigned to anything else and is also outside your DHCP server's DHCP range; if that's not the case, pick something else that's unassigned and outside DHCP range); Subnet Mask = 255.255.255.0; Default Gateway = 192.168.0.1; DNS Server = 192.168.0.1.

I set IP Address = 192.168.0.85 for the network bridge, but still Hqplayer unable to detect NAA.

 

I also found something interesting, when I did lan-to-lan bridging, the status for the network bridge didn't show IPv4 DHCP ip and IPv4 DNS ip. Then when I did lan-to-wifi bridging, both IPv4 DHCP ip and IPv4 DNS ip were available at network bridge's status, and Hqplayer able to detect NAA.

 

Also, about Hqplayer stopped playing after about 1 hour time, I suspected since my music stored at external usb drive and the usb to ethernet adaptor also usb related, this may cause stability issue which maybe some how Hqplayer related. So I copied one music file into internal laptop drive and played from there in repeated mode. So far it looks quite promising.

[lateboomer]

Nah. Stability doesn't improve even I play the file from internal drive with lan-to-wifi bridging. Hqplayer will still stop abruptly about an hour of playing.

[lateboomer]

For those of you having problems when using a USB to Ethernet adapter:

 

I did a little reading and the problem may be that to be used in the bridge, the network adapter has to be capable of working in promiscuous mode. See Promiscuous Mode.

 

It sounds like some (many?) of the USB to Ethernet adapters don't do promiscuous mode.

 

I guess you can try to find a USB to Ethernet adapter that works in promiscuous mode. See USB to Ethernet Adapter and Promiscuous Mode, for example.

 

I guess it's also possible to try to force an adapter to run in promiscuous mode. See Enable Promiscuous Mode (Manually) in Windows 7.

Thank you for the suggestion, very much appreciated.

 

I tried Promiscuous Mode either for Hqplayer ethernet. I also turned on the mode for the router side ethernet which was not needed. It worked and allowed Hqplayer to see NAA in lan-to-lan bridging. You should get an award for this.

 

However, the downside is the sound now is severely stuttered and more like hanging. Hqplayer takes very long time to process to output sound. So I think using usb-to-ethernet adapter is a flawed method and like Romaz said, have to source for pc that have 2 ethernet ports. I guess.

[lateboomer]

Not wanting to take credit for the hard work of others, here a bit more detail as to my system and the steps I took, with quotes/links to the references used;

 

My music PC is currently running a standard Windows 10 build with a Gigabyte GA-Z170N Gaming 5 Mobo. The motherboard has an on-board Killer E2200 Ethernet NIC, 11ac 867 Mbps dual band wireless controller, and USB 3.1. connectivity.

 

Initially I attempted to set-up a Network Bridge utilizing an DLock 3.0 USB to Ethernet Adapter. Regardless if I set the bridge up with Internet Access via the on-board Ethernet NIC, or the USB adapter (USB 3.0 as well as 2.0), the results were the same. HQPlayer would not recognize the mR in NAA mode and the data through-put (upload from music PC) was in the basement.

 

I then dumped the USB Ethernet Adapter and setup my Internet Access over the wireless controller, leaving the mR disconnected and the on-board Ethernet Port free. I let DHCP assign an IP Address and successfully tested connectivity with full through-put. It should be noted that I placed the wireless controller antenna directly adjacent to my Access Point.

 

I rebooted the PC and then connected the mR directly to on-board Ethernet link and manually assigned an IP and Subnet to the IP V4 properties for the Ethernet NIC (leaving the DHCP assigned wireless settings unchanged) as shown in this thread from Clipper:

 

 

 

Once that was done, I then established the Network bridge as shown here: How to set up and manage a Network Bridge connection on Windows 10 | Windows Central. It should be noted that it is not absolutely necessary to assign a static IP to the bridge as stated in the article. Internet / LAN connectivity will work with the settings assigned by windows when the bridge is established. I chose to use a static IP as that performs better with the settings of my NAS, etc.

 

Last but not least, I rebooted the PC and then verfied the newly assigned IP address for the mR and successfully connected to it via WebGui. Once I was certain that the mR was responding, I started HQPlayer with fingers crossed and voila, it worked!

 

I can now stream via Roon, HQPlayer, Tidal, and local/NAS library files at 192 / 24 with no stuttering, clicks or other annoyances.

 

Hope this helps and good luck.

I also able to do this using built-in WiFi. But it is not fully stable. After about an hour or more of playing, hqplayer will suddenly stop if WiFi suffer connection dropout

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

I have not tried out promiscuous mode with built-in WiFi yet. Could it improve stability?

Even though I put the Tplink N wireless router very near to the music pc laptop, but still dropout would happen.

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

I had the opportunity to do some extended testing last night (3 hours of streaming) and had no issues. Realize that doesn't help with your situation.

 

Your issue sounds to similar to a problem I had with an Auralic Aries. I could stream music for about 20 mins and then it would just stop. You could almost predict it with a stop watch. Unfortunately I never figured out what really caused the problem and ended up selling the Aries.

 

Can you still connect to the mR from the PC which is hosting HQplayer after it stops?

While my client naa is Intel nuc but running audiolinux. After music stopped, I opened settings and saw music device still highlighted and I would press OK to pick it up again naa connection. Then I could play another 1 hour plus before it would stop again. As you said very predictable behavior. If I don't click settings I got to continue clicking the song in Playlist until it plays again

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

Hmmm, have you checked if you have any power / energy saving modes active for the OS or wireless controller?

I already put no turn off monitor or hard disk when plug in power.

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

Do you have wireless enabled on your modem and have a wireless router? Sometimes the 2 wireless networks fight each other and changes ip settings. Also try and increase the dhcp lease times on your modem. It might be set to 1 hour. I changed mine to 365 days.

 

 

Sent from my iPhone using Computer Audiophile

The wireless router is Tplink N wireless router. I just connected the music pc laptop to this router via wifi connection with internal built-in wifi of the laptop.

 

I relied on the dhcp of this router and the router was not connected to modem or to internet.

 

Then I setup a bridge between this wifi connection and the direct Lan Ethernet connection between Hqplayer and NAA. So maybe there is settings at the router to improve connection stability which I don't know yet. But reading at latest development, things are getting weirder and weider. Hopefully good solution can be found.

[lateboomer]

I feel weird for quoting myself...

 

Anyway, that USB to Ethernet Adapter works really well (to make up one of the NICs in the bridge).

 

Network Adapter, Anker USB 3.0 to RJ45 Gigabit

 

$16 from Amazon.

So do you enable forcecompatmode mode for this adapter? Is it connected to a router like my case and get dhcp from there? Do you have to enable forcecompatmode mode for direct Lan between Hqplayer and mRendu? Thank you.

 

I also have got one usb 3.0 usb to ethernet adapter, but it only recognized on usb 2.0 port, if I plug it to usb3.0 port, it said hardware unrecognized. Does it matter whether to connect it to usb2.0 or usb3.0 port? I really got lots of questions.

[lateboomer]

I didn't have to force anything with this one. I just plugged it into a USB port and it worked.

 

It worked in both positions (either as the NIC connected to a microRendu or the NIC connected to my switch which is connected to my router in a different room).

 

Once the network bridge was set up with it and another NIC, HQPlayer could detect the microRendu running NAA. This was with HQPlayer running on the same computer as the network bridge and 2 NICs. But I ran HQPlayer on another computer (a mac connected to the same network as the NIC in the bridged machine connected to my switch/router) and it detected the microRendu running NAA that was directly connected to one of the NICs in the machine with the network bridge.

 

I have this USB to Ethernet adapter plugged into a USB 3 port. One of the listed product features is: USB 3.0 enables SuperSpeed data transfer, backward compatible with USB 2.0 / 1.1 standards.

 

Highly recommended if you can't throw Ethernet cards inside your machine!

 

[ATTACH=CONFIG]32553[/ATTACH]

Thank you for the sharing. I need to get one to try it out

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

Thank you for the sharing. I need to get one to try it out

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

I think first I will retry again this weekend using the usb3.0 ethernet adapter that I already have got it. Because I see the chipset is Realtek family which same as mine. Just that previously I couldn't make it recognized on USB port 3.0! But only port 2.0. I think I need to install the driver come with the adapter

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

I tried to reuse again the usb3.0 to ethernet adapter at USB 3.0 port of my music pc laptop but it can only accept the adapter with additional USB multi port adapter. Anyway, to enable 2 way bridging has to turn on adapter router side promiscuous mode. And unfortunately the sound was heavily stuttered and was no go. So the best solution is to get a new USB ethernet adapter suggested by Clipper

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

My current network setup is music pc - - > dlink ac router - - > fmc fiber optic - - > Naa streamer. All are hard wired ethernet. But I see a lot guys here have used a switch between router and a streamer. Could anyone enlighten me why we need a switch there instead of using a router? Of course my router is router switch combo with WiFi feature, is is dlink dir850l ac router

 

Sent from my Redmi Note 3 using Computer Audiophile mobile app

[lateboomer]

I finally get direct connection working between 2 wired Ethernet connections without bridging with existing laptop wifi. I adopted the suggested trick here to add in one additional unconnected usb wireless Ethernet adapter (dlink dwa-137) to the i5 Win 10 Home laptop. The default laptop's Ethernet port is direct connect to fmc converter and further connected to Audiolinux NAA and then to dac. Another one usb port is connected with usb-to-ethernet adapter and further connected to one tplink wireless n router. So I added these 2 Ethernet connections and the one unconnected wifi adapter to the bridge.

 

I set the bridge properties for IP4 and dns to the tplink router. That is it, and everything works fine. Now I let it play continuously to see whether it will stop abruptly after about an hour of playing.

[lateboomer]

I am very glad that this latest connection method works flawlessly. I have put Hqplayer in repeat mode and it plays about 3 hours now without dropping a ball. I would declare the connection is very stable. Thank you guys for this wonderful trick.

[lateboomer]

10 hours ago, elan120 said:

@Bamber, G-10 is an excellent material for this application, and hope you don't mind me offer couple tips in cutting (hopefully not something you already know)...1) G-10 is a hard and abrasive material, using regular saw blade will dull the blade rather quickly, and it is best to use a carbide cutter and set the feed speed slow.  2) The dust generated from machining contains very fine fiberglass, breathing these powder is extremely harmful to the lung, and contact with skin will make you itch easily, so wear good quality personal protection gears while fabricating, and I high recommend using respirator, long sleeve closing, and surgical gloves.  Although the second tip may look to be excessive, but it is not especially when working with materials like G-10, FR4, carbonfiber...etc.

 

Hope this helps, and let me know in case you have questions about getting your back plate cut.

Hi,

 

Could you provide link or pic where is this G-10 material being mentioned? I have not heard about this material before, so I would like to find out more.

 

[lateboomer]

Thank you. It is helpful.

[lateboomer]

On 9/4/2017 at 7:10 AM, rickca said:

You can use this same approach to galvanically isolate your music files.  Just substitute an external SATA SSD like a Samsung T5 or a bus-powered USB3 hard drive enclosure (with 2.5" drives) for the USB3 ethernet adapter.  I haven't tried this yet.  I'm a bit concerned whether sufficient current will get through the iGalvanic3.0 to satisfy these bus-powered devices.

 

Many thanks to @lmitche for coming up with these creative ideas with his Adnaco S3B solution.  I have adapted them to accommodate a network end-point.

Hi Rickca,

 

 Based on this input side idea, is this the correct configuration? :

 

Music server's Mobo eth port <-- Startech eth adapter <--  iGalvanic3.0 <-- SATA SSD or USB3.0 HDD enclosure

 

Have you compared iGalvanic3.0 --> Startech eth adapter --> mRendu output setup to ethernet port --> FMC fiber optic setup --> mRendu setup? Which you think will derive better sound quality? 

 

 

[lateboomer]

Thank you. Your connection description is very clear now, and I realized my mistake.

 

Anyway, my music files setup in my system:

USB3 port => audioquest jitterbug (or earlier Regen versions powered by lps) => USB3 Minipro enclosure with 2.5" hard drive (powered by diy lps)

[lateboomer]

7 hours ago, romaz said:

Sorry for my delay in posting.  Hopefully, I can post my more definite findings soon.

 

Just to add to the discussion on clocks, it is unclear what clock frequency Ayre uses in its AX-5 Twenty but I suspect it won't be a 10MHz clock and you cannot compare phase noise measurements between clocks if they operate at different frequencies.  This is according to Chris Peters, CEO of Mutec.

 

When I asked him to compare for me the differences between the clock used in their well-regarded MC-3+USB vs their REF10, here is what he had to say:

 

"The problem is that comparing the output signals of the REF10 and MC-3+USB is nearly impossible because of the different clock frequencies of both devices. The phase noise depends directly on the clock frequencies. We have measured the MC-3+USB with the nearest clock frequency of 11.2896 MHz to the REF10. The difference is approx. 30 dBc. That means the phase noise of the REF10 is approx. 30 dBc lower as this one of the MC-3+USB."

 

When I asked him why I have been underwhelmed by my experience with certain "atomic clocks" in the past, here is what else he had to say about the REF10 and atomic clocks:

 

"The phase noise difference e.g. between the REF10 and the Antelope 10M or the 10MX is approx. 40 dBc. That is the reason why people do almost not hear any difference when connecting the 10M to the MC-3+USB. The MC-3+USB is better the so-called Atomic Clock of Antelope."

 

Now is a difference of approx 30dBc audible?  Based on my experience in Munich, very much so and enough of a difference that I went ahead and bought the REF10.  

 

Here are the phase noise tracings for Mutec's REF10:

 

So, how do you interpret this plot?  Not so easy, it seems.  There is a knowledgeable and well-respected poster here on CA (who I shall not name) who comments on the importance of phase noise measurements below 1Hz.  Well, when I showed these plot tracings to Lee in Munich, he told me the important measurements to look at are between 10Hz and 100Hz and it is between these frequencies that a really good clock will separate itself from other clocks.  In this regard, he felt the REF10 was stellar, perhaps the best he had ever seen and so it was clear to me that this is the frequency range he would be targeting with SOtM's master clock.  Here is what Chris Peters had to say about this (I have intentionally X'd out the name of the CA poster in question):

 

"I assume you are referring to the discussion in the CA forum regarding the audio-relevant measurement areas. And I think you are referring to "XXX" when talking about the range lower than 1 Hz. I honestly speaking do not know where or how "XXX" got this information. My beta testers have started trying to experience for audio most relevant frequency range since 2013 when we released the first version of MC-3+. So far we experienced audible differences between various oscillators when their specifications where different in a range between 1 Hz and 100 Hz distance from the carrier frequency of 10 MHz. Differences in this frequency range between the oscillators were audible best for my beta testers. So we optimized the REF10 for this frequency range specifically."

 

So, to a large extent, Chris Peters is agreeing with Lee of SOtM.  When looking at phase noise measurements of a clock, look at the measurements between 1Hz - 100Hz.  What is interesting is that with better, clocks, you definitely hear an improvement in bass definition but improvements are also clearly heard in the midrange and treble.  The best explanation I have for this is that if you improve the signal at 20Hz, for example, you would likely also improve its harmonic frequencies as well (ie 20Hz, 40Hz, 60Hz, etc).  As an example, I am currently testing Synergistic Research's "Black Box."  This is a $2k device that I had once assumed was just a glorified overpriced bass trap but it's impact is so much more.  As I am presently testing potential speakers in my large listening room with cathedral ceilings, while my room is not a resonant nightmare, there are clearly nodes in my room resulting in boomy bass in different areas.  With the introduction of this small box (which contains specially tuned passive resonators) into my room, not only does the bass boom disappear but midrange and treble clarity are greatly enhanced.  This Black Box is definitely staying put.

 

Anyway, back to clocking, as you further assess phase noise measurements, remember that with the REF10, phase noise measurements are taken from the BNC outputs and not from the clock.  Taken from the clock, the measurements would probably be even better.  I suspect any clock measurements Ayre might report will be provided by the manufacturer meaning the phase noise at the Ayre's outputs will most certainly be worse.  It remains unclear what SOtM's new clock measurements are based from although I have posed this question to Lee himself.

 

As a further example of how important this is, I have been testing clock cables of various price points and length.  Using various inexpensive DigiKey clock cables from the same manufacturer of various lengths, as you go from 20 to 40 cm in cable length, the SQ degradation is clearly audible.  This is why I had to send my gear back to SOtM.  Because they didn't have the really short clock cables in stock, they ended up using much longer clock cables in my build which I ultimately deemed as unacceptable.  Moreover, as I have tested identical length clock cables with my REF10 from companies like Pasternack ($40), Blue Jeans Cables (<$20), and Black Cat ($250) against the 700 Euro Habst clock cables that I purchased with my REF10, unfortunately, the differences are quite significant with respect to HF harshness and a very flat sound.  Not that the cheap cables sound horrible but when you replace them with the Habst, there's simply no wanting to go back to those cheap cables.  This is where those external clock doubters have a leg to stand on when they make their claims that external clocks don't add anything.  Cable length and cable quality DEFINITELY matters.

 

As to the impact of these clocks to signal timing, I don't believe it has anything to do with it.  Referring back to John's Swenson's post on the REF10 thread that Rajiv provided a link to, the function of the clock in devices like the sMS-200, microRendu, Iso-Regen and tX-USB is not to time the signal (which is what a word clock is responsible for) but for "processing" the signal. How I interpret this is these clocks are necessary for the "functioning" of these components and so a good clock allows a component to function better.  I'll use my production line analogy once again.  When a production line operates smoothly and in a timely manner, less mistakes are made and less time and energy are wasted to correct any mistakes.  This would be similar to how Audiophile Optimizer improves SQ, by removing unnecessary background processes, you get fewer software errors and fewer latencies which also translates to less current draw and less noise being generated in the ground plane.

 

Now, does a better clock guarantee better SQ?  Not always and if there is SQ improvement, this improvement can be variable.  For example, having replaced the stock Crystek clock in the ISO-Regen with the clock in the REF10, is it now on equal footing as the tX-USBultra (which is now also being clocked by the REF10)?  The simple answer is no.  While the Iso-Regen definitely is improved on the REF10, my tX-USBultra (which is also being clocked by the REF10) is still the better component to my ears.  In essence, and I have suggested this before, the best that a clock can do is to allow a component to perform at its very best but a good clock can never make a component perform beyond its physical capabilities.  Absolutely, it is the circuit as a whole that matters and the clock plays just a small part of it.

 

To elaborate on this topic further, while a good clock improves how a component operates, clocks contribute more than just timely functioning, they also contribute noise which can be measured in dB.  Since there is no such thing as a perfect clock, all clocks contribute some degree of noise, it's just a matter of how much.  When one bad clock in your digital pathway can contribute more than 30dB more noise compared to the REF10, imagine the cumulative noise impact of 8 noisy clocks in that same pathway.  To some extent, you can mitigate that noise by throwing an sMS-200ultra into the pathway between your server and your DAC but as many of us have experienced, this doesn't completely fix the problem since adding a reclocking switch before the sMS-200ultra and adding a tX-USBultra after it results in further improvement.

 

As I have now replaced all the clocks in my digital upstream from router onward, it is amazing how each clock replacement adds improvement but what I have also found is that the subsequent clock replacements seem to have less impact than before.  What I am trying to say is that with the clocks replaced in my server, the impact of the ISO-Regen and tX-USBultra is now considerably less.  

 

I know there are people out there who are critical of the "spaghetti solution" that we call the SOtM trifecta.  As an owner and originator of this trifecta, I have to agree it is a bit unsightly and cumbersome, especially when you incorporate the very inflexible SOtM dCBL-CAT7 into the mix (which I find to be indispensable).  This is one reason I went away from it but what I can say is that it doesn't have to be this way for those who are opposed to it.  With a low noise server (meaning a low power CPU, minimal RAM, avoidance of noisy storage drives, replacement of noisy clocks and driven by a clean, low-impedance PSU), this "spaghetti solution" is not only no longer necessary, but it is actually an inferior approach.  Not to suggest that those that have this spaghetti solution should move away from it (because this spaghetti solution probably still sounds better than most things out there) but that there are other pathways to achieving similar (and better) results.

 

Once again, while in Munich, Lee himself told me he considered the sMS-200ultra and tX-USBultra as his mid-level products.  He reserved his highest praise for his sMS-1000SQ server and the Ultra version incorporates the sCLK-EX and also his very best card, the tX-USBexp.  With this server, you can request to have the system board, Ethernet port and the tX-USBexp reclocked and everything fits nicely in one chassis.  Where the sMS-1000SQ probably can be improved (based on what people have told me) is with its PSU.  For those looking for an elegant, no-fuss, turnkey solution, I would suggest the sMS-1000SQ Ultra paired with a multi-rail SR7 and either SOtM's new master clock or the REF10.  Can this be improved upon?  Yes, I know it can and I will provide details of this in a further post.

 

 

Thank you for sharing your finding, it is very eye opening. I am particularly very interested with this quote "With a low noise server (meaning a low power CPU, minimal RAM, avoidance of noisy storage drives, replacement of noisy clocks and driven by a clean, low-impedance PSU)". As I just got myself an asrock j3455b-itx mobo recently to use as a music server. It is 1.5Ghz CPU and I like it's atx connector type power supply rather than DC supply, with compare with my previous Atom mobo.

 

So I don't expect it to be a world beater as I am very green in this exploration, and I would be grateful if someone can point me the link to more definitive thread on what it means by low power CPU, min ram etc.

[lateboomer]

Thank you. I will comb through the threads carefully.

[lateboomer]

I am glad you ask. I am using coolmaster 80 plus 400w normal power supply for it. I will consider seasonic 750w prime titanium suggested by Lmitche in the future as I maybe consider migrate to Asus z170 plus mobo with i7-7700 to enable up sample to dsd256. Currently Asrock can only do dsd128 running under Daphile server client setup.

 

Currently, the client is nuc 3815 powered by diy regenerative and lps. I don't think onboard DC supply is better than atx power supply. However, lps for atx  power supply is too challenging to implement. That is why I would consider Seasonic titanium. 

 

My config now:

Minipro ext HDD with Jitterbug >> asrock j3455b >> Ethernet >> drink dir-850l router with lps >> fmc and sfp fiber with lps for both FMC >> nuc 3815 with lps >> Regen Amber with lps >> USB DAC

 

For pcie lane, I closely follow just launched jcat net femto card. If I use Asus z170 as server, then I could swap this Asrock as client and consider jcat USB femto card.

 

But with FMC fiber implementation, I not sure do I still need to consider those femto cards or not.

 

One thing, current memory is 4GB x 2, I thought dual memory can help me to get to dsd256, it is just wishful thinking.

[lateboomer]

I read that jcat femto net card is developed from Intel i350t nic card. 

 

I wonder we could hack the Intel nic card to provide lps to it.