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

[austinpop]

On 3/8/2020 at 11:39 AM, austinpop said:

I've been biting my tongue wrt all the EtherREGEN discussion, while my review was pending. It's finally ready and should be live within the next couple of days.

 

The review is up on the main page.

[austinpop]

25 minutes ago, Nenon said:

Is it worth doing that? IMO - no.

Does it ever end? NO!

 

As I said in review, this is a bottomless pit.

 

For most people, a single good switch like the eR with a good PSU is a sensible place to stop. Some might add an oM or second switch, plus another good PSU. Beyond that, it's crazy-town!

[austinpop]

4 hours ago, auricgoldfinger said:

 

@austinpop and I use SOtM via optical to ER.  In the ER review, he preferred the Melco to the other two switches (when cost is not a factor), so why not try it last in the chain?  Only Uptone says the ER must be last.  I haven't seen anyone test that assertion.

 

Despite the review becoming TL;DR for many people, there were sooo many things I didn't try or chose to leave out!

 

Yes, I do think the "best" switch should be last in the chain. I'll try the Melco as the last switch when I get a chance...

[austinpop]

4 hours ago, Cipollae said:

Just arrived!  Found this forum after reading Rajiv's excellent review of the iFi Pro iDSD.  As it happens, that is the DAC I recently purchased and I am now thinking of upgrading the power supply.  The Farad Super 3 looks extremely promising (also based on Rajiv's and some other people's reviews here 😃), but I am a little worried the 3A might not be enough - especially since the Pro iDSD comes with a 15V/4A SMPS and iFi suggests a "minimum of 60VA rating".  However, the same goes for Rajiv's Chord TT2 and he seems to be using the Farad without any issues nonetheless.  

 

Would anyone be willing to comment on whether it might be unwise to go for the Farad?  Would the Paul Hynes SR5 perhaps be a safer option?  Or am I being overcautious because the Pro iDSD will never draw 4A with normal use?

 

Hi, welcome to AS.

 

This is a very good question, and you're in luck - a local friend of mine has just gone through exactly this thought process with this DAC. He really liked the Farad Super 3 15V driving the Pro iDSD, but then realized how much better it got when we tried my 15V/6A SR-7 rail with it. He ended up getting an SR5-15F which has a 5A continuous/30A transient capacity, and this was a big step up from the Farad.

 

It's no knock on how good the Farad is - the DAC use case really benefits from current capacity, so the PSU selection process needs to weight current capacity much more in this use case.

[austinpop]

51 minutes ago, sbenyo said:

Many great reviews for new 2020 Sablon usb. Anyone compared it to Habst Ultra III or Intona Ultimate?

 

 

They are all excellent cables, and honestly the one you prefer comes down to your ears and your system.

 

I did a comparison of exactly these 3 cables and posted my impressions here: https://www.whatsbestforum.com/threads/sablon-usb-cable-2020-model.29912/post-627980. In my system, I really felt the Sablon's were the best in my system, as I described.

 

39 minutes ago, limniscate said:

 

@austinpop and I compared the Sablon to two different Intona cables and a Habst cable, but I'll let him speak to the differences because he probably took notes.

 

More recently - but before we all went into self-isolation, Eric and I did a comparison in his system. We compared the Habst, Intona Ultimate, and Sablon unpowered between the Zenith SE and tX-USBultra. We kept the second leg from tX-USBultra to the DAC (Ayre QX-5) fixed with a Sablon powered cable.

 

In this case, we both liked the Habst the best, followed very closely by the Sablon, with the Ultimate much further. The Habst is a little prominent on the low mids and upper bass, but also has a very clean, dense, saturated presentation. The Sablon is actually more balanced, but in this system, we both found the Habst more satisfying.

 

I suspect in my headphone system, which is so revealing, the smoother signature of the Sablon was why I preferred it over the Habst. It's amazing how system-dependent this stuff is!

 

 

 

[austinpop]

4 hours ago, mourip said:

 

Every time I have introduced/re-introduced fiber into my system in any of a number of possible positions I end up removing it. To be clear I have not tested some of the more advanced fiber options but I just prefer copper ethernet. The combination of the JCAT FEMTO card in my server and the ER set up as recommended has given me the best sound yet. Very natural and realistic sound to my ears.

 

I have had the same off and on experience with fiber. I currently happen to like it between the sNH-10G and eR, but only after I switched to the Startech - and now Planet Tech - transceivers. Ultimately, you just have to experiment.

[austinpop]

7 hours ago, bodiebill said:

 

For some time my setup has been:

modem/router => MikroTik SFP switch => fiber => Intel fiber NIC => server PC with HQPe

same MikroTik => fiber => Intel fiber NIC => audio PC with NAA

 

The last few weeks, after changing to a Denafrips Terminator DAC, many things improved, but I was not totally happy with the sound for classical music. High strings were a little thin and strident.

Inspired by this thread I changed the setup to:

modem/router => MikroTik SFP switch => fiber => Intel fiber NIC => server PC with HQPe

modem/router => copper ethernet => audio PC with NAA

and I like that better. Strings are fuller. The bloom is back.

I do not think it is the fiber as such but maybe the Intel fiber NIC in the audio PC is adding noise/jitter?

 

I could go all-copper, but maybe keeping the fiber upstream to isolate the audio system from the rest of the house is better? Like in:

modem/router => MikroTik SFP switch => fiber => Intel fiber NIC => server PC with HQPe

MikroTik => fiber => 2nd MikroTik => copper ethernet => audio PC with NAA

 

Any ideas?

 

(PS I just ordered a Cisco 2960G-24TC-L to experiment with.)

 

 

I don't have experience with those switches and NIC, so in a general sense, the only advice I can offer is to continue experimenting.

 

In my experience, I find that the fiber vs. copper difference is a subtler, second order effect. This is just a suggestion for you to consider - your efforts might yield more fruit by first cleaning up the PSUs in the chain, starting with your audio PC and working backwards.

 

Are you willing to consider an etherREGEN? If so, I'd suggest it instead of the 2nd MikroTik, exactly as you've shown: fiber in, copper out to audio PC.

 

Finally, transceivers do make a big difference, and I've benefitted from the findings of others here who've recommended the Planet Tech MGB-TLX transceivers. It specifically adds density and weight to SQ, so may be useful for your thinness/stridency observations.

[austinpop]

Server Update

 

I just wanted to jump in with a server update and some insights. These are not new - @romaz and @Nenon along with others have described similar experiences. For me, these reinforced my experiences to date.

 

I have in hand a 2nd server that is essentially the same as my old server, with 2 important differences:

• CPU is i7-8700K vs. my current 8700T
• Motherboard is Asrock Z390 Phantom Gaming ITX/ac vs. H370M-ITX/ac

For the rest of the system details, see my signature.

 

I've been able to compare both servers powered by the same HDPlex 400W Linear ATX PSU I've been using. You might think these servers should sound very similar, but you'd be wrong. Key observations:

• Using Euphony OS to set the frequency to 4GHz in both cases, the Phantom Z390/8700K  is notably better than the H370M/8700T. The biggest impact is on dynamics, as you'd expect, but there is a tonal richness that is very alluring. Transients are crisper, and imaging is bigger in all 3 dimensions.
• I've been using my 8700T basically unconstrained in frequency in Euphony, so it's been maxing out to 4.0GHz. If I run the 8700K at this same frequency, the 8700K does run a little hotter by 2-3º C, which is notable, but not sufficient to cause problems. 
• At full bore (unconstrained by Euphony), the 8700K maxes out at 4.4GHz, and at this point, the harshness I've described in past posts when using the HDPlex, rather than the SR-7 DR, creeps in. CPU temps climb up to near 50º C. The nice thing is I can dial this back with Euphony so easily. I find 4GHz to still be the sweet spot in my system the way it currently is. When I first did this experiment some a year or so ago, I was finding I had to set my 8700T max frequency to between 3.2 and 3.6 GHz. I attribute this change in the optimum to 4GHz to improvements in the rest of my chain.
• I haven't yet tried it with the SR-7DR powering it, but I fully expect to hear the same Roy did, since this new machine is basically the same as his 8700K server.
• As if I needed reminding, but I did do an experiment after a long time: I compared Euphony max frequency set to 4GHz (optimal SQ for me) to max frequency set to 800MHz. If anyone ever needs convincing how much a powerful CPU matters to SQ, try this experiment. The system at 800MHz felt like it had shrunk to half its size. The dynamics were shockingly poor, and the SQ was dull and unexciting.

This reinforces to me that systems with more powerful CPUs, with more cores, and bigger caches DO sound better, EVEN though the CPU utilization with the non-upsampling workloads of Stylus or Roon is vanishingly small. The challenge with such systems of course, as Nenon is detailing in his excellent DIY server thread, is getting a power supply that is both outstanding quality AND high-current.

 

For now, this new 8700K server will take residence in my system. I still have a few experiments I want to do with my 8700T server, but after that I'll be done with it, so if anyone has an interest in buying it, PM me. Just know you'll need to separately buy a PSU, or use an existing one.

[austinpop]

8 minutes ago, StreamFidelity said:

 

I fully support this fine report.

 

Coincidentally, the 4GHz is exactly the limit for smooth operation of the EC modulators (DSD) in HQPlayer. 🙂I set my system to 4.1GHz to have some headroom.

 

Thanks. I’ve been enjoying your build thread - my bad for not mentioning it!

[austinpop]

3 hours ago, lmitche said:

It would be good to know not only the OS but also the playback software configuration used in your experiment.

 

Hi Larry,

 

Sure. I was running Euphony OS 20200316 in RAM Root, with Stylus as the player. Buffering and caching flags were On.

 

3 hours ago, lmitche said:

I do wish that we knew what is going on here.

Completely agree.

[austinpop]

I started to write about how I had seesawed at least twice between distributed and standalone topologies, then thought it would be fun to create a chronology of my journey. Maybe this will become the seed of an article at some point. But anyway, for your amusement and edification, here is a chronology of my 10-year journey in computer audio streaming.

 

Note: for simplicity, I focused on topology, so this doesn't address all the other improvements and tweaks to get better SQ.

 

Keep in mind that there is an implied increase in SQ as you go from top to bottom. This core question is what has driven me throughout: what's the next step to better SQ? I started my journey firmly believing in the distributed model being superior. Over time, I learned to shed my dogmas, and just follow my ears.

 

Time Period	Music Storage	Music Server (HW/SW)	Music Endpoint (HW/SW)	Standalone Player (HW/SW)	Control (HW/SW)
2010-2015	Synology NAS	Synology NAS/LMS	Squeezebox Touch		IOS/iPeng
2016-2017	Synology NAS	Synology NAS/MinimServer	Aries Mini/Lightning		iOS/Lighning DS
2017	Synology NAS	Desktop/W10, Roon Core	Aries Mini/Roon		iOS/Roon
2017	Synology NAS	Desktop/W10, Roon Core	sMS-200/Roon		iOS/Roon
2017	Synology NAS	Desktop/W10, Roon Core	SOtM trifecta (switch, sMS, tX)/Roon		iOS/Roon
2018	Local SSD on ZENith SE			ZENith SE/Roon, Roon+SL	iOS/Roon
2019	Synology NAS	Desktop/W10, Roon Core	NUC7i7DNBE+Plato/AL, Roon		iOS/Roon
2019	Synology NAS	Desktop/AL, Roon Core	NUC7i7DNBE+Plato/AL, Roon		iOS/Roon
2019	Synology NAS	Desktop/AL, Roon Core	NUC7i7DNBE+Plato+SOtM clock mods/AL, Roon		iOS/Roon
2019	Synology NAS	Desktop/AL, Roon Core	NUC7i7DNBE+Plato+SOtM clock mods/AL, Squeezelite		iOS/Roon
2019	Synology NAS	i7-8700T/AL, Roon Core	NUC7i7DNBE+Plato+SOtM clock mods/AL, Squeezelite		iOS/Roon
2019	Synology NAS	i7-8700T/Euphony OS, Roon Core	NUC7i7DNBE+Plato+SOtM clock mods/Euphony OS, Squeezelite		iOS/Roon
2019	Synology NAS	i7-8700T/Euphony OS, Roon Core	NUC7i7DNBE+Plato+SOtM clock mods/Euphony OS, StylusEP		iOS/Roon
2019	Synology NAS			i7-8700T/Euphony OS, Roon+SL	iOS/Roon
2019	Synology NAS			i7-8700T/Euphony OS, Roon+StylusEP	iOS/Roon
2019-2020	Synology NAS			i7-8700T/Euphony OS, Stylus	iOS/Stylus
2020	Synology NAS			i7-8700K/Euphony OS, Stylus	iOS/Stylus

Glossary:

 

Desktop: Vanilla Dell Desktop system

AL: AudioLinux

SL: Squeezelite

W10: Windows 10

LMS: Logitech Media Server

i7-xxxx: custom servers using xxxx processor

[austinpop]

4 hours ago, mourip said:

One thing that really stands out is that you have nearly always stayed with a NAS for storage. Any ideas why this would consistently sound better than a local drive? Probably the only thing that I have never tried was a single box plus NAS setup.

 

Yes, that is true. This is an area where I admit I've done less direct experimentation, and just gone with the experience of others I trust.

• I know Roy had found NVMe SSDs to sound harsh, so I avoided that path.
• I know some prefer HDD with clean PSU rails for both the drive and the data, while others have used SATA SSDs, with filters like the SOtM, again with clean PSU rails
• When I started down the NUC/AL path, I did experiment with comparing the SQ of files stored on NAS vs. local Optane vs. RAM disk. Indeed, local Optane sounded better, but Optane just does not have the capacity I need. The more interesting result was that RAM disk sounded the best of all.
• Ultimately, this is what has persuaded me to stick with NAS:
  • Due to my investment and care in the network switches preceding the server, I am attenuating (I'd be the last to claim fully isolating!) the noise impact of the upstream NAS
  • I don't have to devote precious LPS rails to power local drives
  • With the Stylus player and Euphony OS, using the option to pre-buffer the music, I am getting the benefit of RAM disk quality during playback. 
  • To me, this is the best of all worlds.

However, like with most things, I would not be surprised if there is more SQ to be had in this area.

 

4 hours ago, mourip said:

Great post...

 

Thanks! I figured laying it out as a chronology would be useful.

[austinpop]

On 4/21/2020 at 2:01 AM, austinpop said:

I still have a few experiments I want to do with my 8700T server, but after that I'll be done with it, so if anyone has an interest in buying it, PM me. Just know you'll need to separately buy a PSU, or use an existing one.

 

I'm done with all my experiments, so this 8700T server is up for sale in the Buy/Sell section. PM me if interested.

[austinpop]

3 minutes ago, ASRMichael said:

@austinpop New purchase coming? 

Yes, as mentioned in this post.  

[austinpop]

8 hours ago, ASRMichael said:

Sorry didn't know where to post. Anyone else have an issue with email notifications from threads? I've had none since Tuesday?

Site issues go on this thread below. There does appear to be an issue with some email providers...

 

 

[austinpop]

On 3/28/2020 at 1:22 PM, Nenon said:

 

I tried Melco --> Buffalo BS-GS2016 (modified) --> EtherREGEN as I mentioned in my post below.

It sounded better than Melco --> EtherREGEN, which sounded better than just an EtherREGEN. 

Diminishing returns but it gets better.... 

 

@Nenon had kindly sent me another modified Buffalo BS-GS2016, as I was very curious to try it in my system and compare it with both the Melco S100 and the etherRegen. He had warned me it took  a long burn-in period to sound good, so I used it as a regular switch in my home office for a couple of weeks before listening.

 

Comparison between the Buffalo BS-GS2016 and the Melco S100

 

As @seeteeyou first alerted us, the Buffalo BS-GS2016 appears to have an identical board to the S100. The difference is that the Buffalo BS-GS2016 has an inbuilt SMPS. @Nenon's mod was to remove the SMPS, and expose a DC input. Another difference is that the Buffalo's management UI is accessible, while the S100 does not appear to expose any UI. Based on the Chinese forum that @seeteeyou linked to, I set the Buffalo to the following settings:

• top row of ports off
• Ports 2/4/6/8 at 100Mbps, 10/12/14/16 at 1Gbps
• LLDP off
• 802.3az off (aka EEE) on all ports
• ADP (Auto power down) off on all ports
• Flow Control is ON - this is off by default. 

After the extended burnin, I first attempted to answer the biggest burning question in my mind: does the Buffalo match the SQ of the Melco? To my ears, the answer was no. Let me be clearer. The Buffalo had all of the other great attributes of the S100: density, saturation, body, and depth. The family resemblance was very clear. The big difference to me was that the Buffalo had a slightly fatiguing harshness, or sharpness that the Melco did not. And this was with both devices powered by an SR-7 DR rail.

 

My optimum switch chain since I published my eR review has been:

• copper > 'B' eR 'A' > fiber > S100 > server. 

In the above experiment, I was interchanging the S100 and Buffalo as the last switch in the chain. I also tried reversing the order like so:

• copper > (100) Buffalo (1G) > fiber > 'A' eR 'B' > server,

and comparing the Buffalo and S100. Unfortunately, that did not change the outcome. With that said, the Buffalo, priced at just $200ish, is a bargain compared to the S100, and perhaps in a different system, the harshness I was hearing may not be as noticeable or bothersome.

 

At this point, I packed up both S100 and Buffalo and sent them over to Roy, as he has been keen to try these. He posted his impressions in another forum here: https://www.whatsbestforum.com/threads/network-improvements-and-their-impact-on-sound-quality.30455/post-647211

 

Obviously, in his system, the Buffalo was much more tolerable than in mine. This could be down to many things, but ultimately it just shows that beyond a point, you really have to use your own ears to truly evaluate gear!

 

 

[austinpop]

42 minutes ago, Dutch said:

Oops, I see I missed @austinpop’s post right above Nenon’s about the Buffalo switch. Thank you as well!

 

I did not notice an increase in harshness deploying the Buffalo where I deployed it..but it replaced a cheap and internal SMPS powered D-Link switch in my case! 😉

 

That's funny - I had someone else also ask me what I thought of the Buffalo switch, because they missed my post above @Nenon's.

[austinpop]

Just now, Dutch said:

I think that’s because he linked to his post directly from the WBF forum.

 

I really want to move the discussion back here on this thread where, frankly, it belongs! I don't think the core audience of WBF have the understanding and the context the readership of this thread has.

[austinpop]

One possible reason I can think of for why the Buffalo's slight harshness bothers me more than @romaz, despite his system being far more refined than mine, is one word: headphones.

 

Headphones - even my very smooth, neutral Meze Empyrean's in my highly optimized chain - are relentlessly unforgiving of even the hint of harshness. Could this be the reason? Or is it just down to differences in personal preference? Who knows.

 

But that said, at $200ish, the Buffalo is otherwise a killer value, assuming you have the wherewithal to mod it the way @Nenon did.

[austinpop]

3 hours ago, Blackmorec said:

In my experience, identifying sources of harshness is valuable in that it almost always leads to a deeper understanding of and improvement to certain performance aspects of a component. Harshness is a distortion, where certain narrow frequency ranges are overemphasised relative to the rest of the signal. That extra energy comes from somewhere and the most logical place is from the signal itself. A slight loss of detail resolution doesn’t lose that detail’s energy, its still there, just not separated from other parts of the signal, so what you hear is the pure signal, plus the unresolved detail as overlying distortion. The acid test would be a comparison in the absolute high frequency detail level of the 2 switches.....if the S100 seems to have a ‘tick’ more detail, that is where the harshness is coming from. So then the question would be, why does the Buffalo revolve slightly less than the Melco? From the dialog, you mentioned that Nenon had modified the Buffalo, which begs the question how many hours it had to since run in post mods. As we know, components like capacitors and cables take a while and Nenon’s favoured Mundorf Silver/Gold DC cable takes a really long time....like SEVERAL weeks. One of the reasons the switch may have sounded different in your vs. Romaz’s system could be down to the length of time for burn in. I have also found that burn-in isn’t a linear process and that the biggest changes often occur after an extended period of switch-off and cool down.  I often wonder if we don't extend the burn-in time required by leaving things switched on 24x7 instead of constantly putting them through power cycles?  This is of course all pure speculation on my part but it could be a reason. 😗

Roy and I listened to the exact same Buffalo and Melco units. I detached them from my system and shipped them to him. He didn’t subject them to additional hours of burn in before listening.

[austinpop]

2 hours ago, Johnseye said:

 

Are you referring to the capacitors in the Melco SMPS?  The SMPS in the Buffalo got removed in favor of an external LPSU with Nennon's but I believe you kept yours but regulated it.  Forgive me if I'm not accurate, and please clarify.

 

I also own a modified Buffalo thanks to Nennon.  The firmware out of the box was not current, I updated it.  I'm not sure if @austinpop or anyone knows what the Melco out of the box settings are, but mirroring those on the Buffalo as that switch is highly configurable, should be like for like.  I'm guessing the Melco manual will provide some level of detail.  There are other settings on the switch such as setting it as a layer 2 or layer 3 switch, QoS, and turning off SNMP which could have an influence.

 

I am still burning mine in and will post impressions in time.

 

Regarding the settings. I didn’t do careful A/B between settings. I used the following rationale for the settings:

- LLDP OFF - from the Chinese forum

- EEE (802.3az) OFF - inspired by early eR issues with EEE

- APD - looked like another energy management control, so thought it might impact SQ

- flow control (802.3x) ON - I’ve seen this recommended here on AS for years for managed Cisco switches.

 

Since you have the switch too, please compare these to default settings.

[austinpop]

Apologies. This line was in error. Corrected (changes in bold) below, and in the original post:

 

- flow control (802.3x) ON - I’ve seen this recommended here on AS for years for managed Cisco switches.

[austinpop]

17 minutes ago, fpalm69 said:

Before I read through all 655 pages.....

Has anyone been able to improve SQ using a Roon Nucleus and SOtM sMS-200 Neo both attached to the same stock Zyxel switch?  All powered with an HDPlex 200.

I cannot find any info on bridging these two together.

 

Sorry, I'm not sure what your parameters are.

 

Are you asking if there is a way to get better SQ without any change of equipment or expenditure? Short answer is - possibly. It depends on whether ROON ROCK now supports bridging across 2 Ethernet interfaces.

 

Or is it a more aspirational question? Short answer is - YES!

[austinpop]

47 minutes ago, bodiebill said:

Converging to the SQ that can only be had by... no network at all.

 

This is an excellent insight, and is a great reminder of what all the network tuning is trying to achieve. This also explains why some still eschew streaming altogether, and find better SQ by spinning silver discs on transports.

 

Ultimately, many of us want the rich user experience or remote control, and the incredible musical variety of streaming, which is why we don't want to go back to a monitor/keyboard/mouse to control our music. Still, it is always another path to ultimate SQ!

[austinpop]

8 minutes ago, ASRMichael said:

my new build

 

Looks great! Can you post pix and details of the PSU, and what are the specs for each rail? Are you upgrading the EPS rail only from 12v/5a to 12v/10a?