22 replies to this topic

[bodiebill]

I just changed something in my dual PC setup with optical isolation, just because I was curious whether I could hear the difference:

 

(1) previous setup:

router => long ethernet cable => control PC NIC1

control PC NIC2 => short ethernet cable => FMC => fiber cable => FMC => short ethernet cable => audio PC

 

(2) new setup:

router => long ethernet cable => FMC => fiber cable => FMC => short ethernet cable => control PC NIC1

control PC NIC2 => short ethernet cable =>  audio PC

 

where obviously the blue bit contains the optical isolation.

 

I originally chose for (1) as it seems to make more sense to isolate the down-stream audio PC from the control PC and everything further up-stream to minimise electrical pollution.

However, I was surprised to find that (2) sounds better, so I guess it is a keeper.

 

Anyone has similar – or different – experiences with this?

[Marcin_gps]

I'm not surprised. I gave up on fiber optic isolation completely in playback path. 

[bodiebill]

I'm not surprised. I gave up on fiber optic isolation completely in playback path. 

 

I am not giving it up yet :-)

Best SQ for me so far (see above) is (2); then (1), and only then, at some distance:

 

(3) setup without optical isolation:

router => long ethernet cable => control PC NIC1

control PC NIC2 => short ethernet cable =>  audio PC

[Thuan]

I've been happily running the following setup:

Router => FMC => long fiber cable => FMC => short Ethernet cable => NIC2 (USB Ethernet adapter, enhanced by a powered iDefender) on controlPC

NIC1 on controlPC => short Ethernet cable => audio grade Ethernet switch (powered by linear PSU) => short Ethernet cable => audioPC

HySolid receives commands/instructions from iPad via NIC2 then spits sound bits to audioPC/DAC via NIC1. After explicitly separating subnets, SQ took a leap higher. Everything fleshed out, black blacker, lower noise floor so I could and wanted to raise the volume a few notches louder. There's a noticeable increase of wholeness of the playback.

[bodiebill]

I've been happily running the following setup:

Router => FMC => long fiber cable => FMC => short Ethernet cable => NIC2 (USB Ethernet adapter, enhanced by a powered iDefender) on controlPC

NIC1 on controlPC => short Ethernet cable => audio grade Ethernet switch (powered by linear PSU) => short Ethernet cable => audioPC

HySolid receives commands/instructions from iPad via NIC2 then spits sound bits to audioPC/DAC via NIC1. After explicitly separating subnets, SQ took a leap higher. Everything fleshed out, black blacker, lower noise floor so I could and wanted to raise the volume a few notches louder. There's a noticeable increase of wholeness of the playback.

 

Thanks thuanb, that looks somewhat like my favorite setup (2), although I don't have the extra's (iDefender, switch).

[Thuan]

Basically, yes. You may want to consider long fiber cables vs long Ethernet cables. Also, an audio grade switch between controlPC and audioPC does help SQ quite a bit. iDefender enhancement is icing on the cake.

[bodiebill]

Basically, yes. You may want to consider long fiber cables vs long Ethernet cables. Also, an audio grade switch between controlPC and audioPC does help SQ quite a bit. iDefender enhancement is icing on the cake.

 

Locally offline (on the control PC, that is) JRiver converts all files to DSD128 and put them on a SLC usb stick. I then play them with f.i. Hysolid.

Logic could tell me that the type of ethernet cable to my control PC would hardly matter in that setup; as I can even remove the cable while playing.

But I have been wrong before :-) And I also realize that this would not explain why having the ethernet connection from router to control PC optically isolated make a difference in the first place.

Perhaps the optical isolation gets it closer to 'no cable' and perhaps the long fiber cable could bring it even closer.

Intriguing stuff.

[Thuan]

Ah, SLC. I've just ordered a SLC ssd to be used as OS disk. Soon I'll find out how it performs to improve SQ.

By the way, how do you do about reducing the services in Task List?

[bodiebill]

Ah, SLC. I've just ordered a SLC ssd to be used as OS disk. Soon I'll find out how it performs to improve SQ.

By the way, how do you do about reducing the services in Task List?

 

On the audio PC (core) I am down to approx. 16 services/processes. On the control PC  (GUI) I have over 40. But I am no longer paying much attention to this. It was useful, but became a goal in itself, and now I just use the good old ears.

[bodiebill]

And today I noticed that even the power supply for the upstream FMC (that is, the FMC furthest from the audio PC) makes a noticeable difference. I had it powered by an Anker battery, with the nuisance that I have to recharge that every 2 days, which I sometimes forget; and I measured its output voltage as only 3.8V rather than the required 5V. So I tweaked my Teradak U9AV linear PS so that it delivers 5V and replaced the Anker by the Teradak. After a few listens (back and forth) it was very clear: the sound was much richer and more involving with the battery. So I am back to the Anker for serious listening. Good thing is that I now have the Teradak as a backup should I forget to recharge the battery.

 

Assume nothing :-)

[Thuan]

Never assume anything, can't be truer. Experimenting and trust your own ears are best policy. Thanks for sharing your finding.

Related to your 1 message earlier, I too was able to get down to 13 services (audioPC) and 18 services (controlPC), both in 2016 Core. But do you know what I've found? Musicality was gone, the setup sounded very detailed but thinner than I like. So I decided to go back to the AO exclusive configuration, still not logging in the Windows. Music is back, detailed, holographic, separation, reasonably full and warm when the album asks for. Lesson I've learned: reducing services/processes to minimum isn't everything in getting the most in SQ from your PCs.

[bodiebill]

Never assume anything, can't be truer. Experimenting and trust your own ears are best policy. Thanks for sharing your finding.

Related to your 1 message earlier, I too was able to get down to 13 services (audioPC) and 18 services (controlPC), both in 2016 Core. But do you know what I've found? Musicality was gone, the setup sounded very detailed but thinner than I like. So I decided to go back to the AO exclusive configuration, still not logging in the Windows. Music is back, detailed, holographic, separation, reasonably full and warm when the album asks for. Lesson I've learned: reducing services/processes to minimum isn't everything in getting the most in SQ from your PCs.

 

Interesting: another counter-intuitive result. It reminds me of the time I went to Server 2016 core on the control PC and liked it less than Server 2016 GUI (which surprisingly was confirmed by one other member of this forum). So now I am back to GUI for the control PC.

 

One thing I like about this computer audiophile thing is that it puts the experience ('the goosebumps') first, and that it makes you see that you never 'know' anything just because it seems logical. Our (auditory) sensitivity is much more fine-grained than our ratio.

[Thuan]

Reminded me of the time I ran both PCs in 2012r2. When both in Core, they didn't sound as good as when controlPC was in GUI. Afterwards, I went all out on audioPC, then controlPC. Trust me, equally optimized, they sound crazily good when both in Core, 2012 or 2016.

[BernieK]

The findings reported in this and other threads continue to make it clear that system differences are important with respect to perceived outcomes following optimization steps. At the risk of stating the obvious, what works well in one system may not be beneficial or even possible in another. But optimization involves experimenting and tailoring to the individual setup. That's part of the fun, n'est-ce pas?

[bodiebill]

The findings reported in this and other threads continue to make it clear that system differences are important with respect to perceived outcomes following optimization steps. At the risk of stating the obvious, what works well in one system may not be beneficial or even possible in another. But optimization involves experimenting and tailoring to the individual setup. That's part of the fun, n'est-ce pas?

 

oui, c'est très vrai!

[bodiebill]

Reminded me of the time I ran both PCs in 2012r2. When both in Core, they didn't sound as good as when controlPC was in GUI. Afterwards, I went all out on audioPC, then controlPC. Trust me, equally optimized, they sound crazily good when both in Core, 2012 or 2016.

 

Maybe one day I will try control PC with core again... However, for more casual listening I am using the (ancient) FlacWavLoader tool which quickly and easily converts flac files to wav and puts them on my SLC usb stick, and this tool does not work on server core. Also, I understand that Java (needed for minimserver) is not obvious on core.

[Thuan]

Right, I couldn't get Java to work on 2016 core so any attempt to install MiniServer will have to wait. HySolid sounds great anyway. It's understandable to stay with GUI 2016, especially now when Bernie and several others have been able to streamline the OS to its minimum.

I can't wait until AO 2.20 beta 5 is out. Are you using Panasonic SLC USB stick?

[bodiebill]

Right, I couldn't get Java to work on 2016 core so any attempt to install MiniServer will have to wait. HySolid sounds great anyway. It's understandable to stay with GUI 2016, especially now when Bernie and several others have been able to streamline the OS to its minimum.

I can't wait until AO 2.20 beta 5 is out. Are you using Panasonic SLC USB stick?

 

Indeed I was inspired by Bernie's streamline techniques with CAD optimization and such, but I noticed more effect of that on the audio PC than on the control PC. Maybe that will be different now that the optical isolation is between the router and the control PC.

 

I use a MX-ES SLC usb stick. I did not compare different brands. This one was affordable, makes a sturdy impression and sounds good.

[Thuan]

Good to hear your findings on both optimization scripts and affordable SLC flash drive.

P.S. In anticipation of the AO beta 5 (software) and iFi iGalvanic3.0 (hardware), a Swissbit SLC ssd comes tomorrow to have a clean 2016 core installed on. May and June are going to be exciting time.

[Pachanko]

I use the smarthub from Waversa Systems (1500€), which is also a huge improvement:

 

http://www.waversa.com/#wsmarthub