Network optimization for serious streamers


In my ongoing experiments, now going on seven years, with network optimization for streaming I've discovered a number of optimizations that should work with any ethernet ISP.

 

I've tried a variety of ethernet cables, modems, routers, switches, FMC, ethernet filters, the following is what I've found to be most effective optimizations.

 

I'll start with ISP quality and speed. Recently I discovered 500mbps to be preferable to 300mbps. Along with upgrade in speed, modem capable of 1gb service replaced 600mbps, both have Broadcom chips and powered by same lps. Can't say which more responsible for improvement, speed or modem, presume speed has at least some role in ping time. As for ISP, there is importance in ISP server geographic location to you, shorter distances  means lower ping time. For information as to how ping time affects jitter-https://www.fusionconnect.com/speed-test-plus/ping-jitter-test

 

Now for modems,  modem close to audio system is most favorable, extending coax cable preferable to long ethernet cable. Coax more resistant to rfi and closer positioning to system means one can more easily afford top quality ethernet cable for modem to router connection. The modem should use Broadcom chipset vs. inferior Intel Puma, Broadcom chipset has lower jitter vs the Intel. Modem should be powered via external lps using quality DC and AC cables, lps to power conditioner for ultimate performance.

 

Following close positioning of modem to audio system, router should also be placed near modem in service of same advantage of making highest quality ethernet cable more affordable, in this case, modem to router and router to switches, streamers and NAS. Router should be powered with lps, this lps should be able to provide more amps than router requires in service of providing greater reliability, having lps with reserves of amperage means lps runs cooler, heat is enemy of reliability, longevity. As with modem, quality dc, ac cables and connection to power conditioner.

 

The next finding is new to me, provides very meaningful upgrade to streaming sound quality. Noise from wifi, injected both internally to router and externally with routers sitting close to audio systems has long been a concern to me. I have quality Trifield meter which measures rfi, router with operational wifi manufacture obscenely high levels of rfi, rfi is noise, noise is enemy of streaming at level we're talking about here. And its very likely the more wifi devices one has in home the higher the levels of rfi produced. This noise is then injected into following cables and streaming equipment. One may convince themselves FMC totally isolates this noise, and while correct, it doesn't mitigate the noise and masking going on within router. The only way to eliminate this noise is turning off wifi. And then, how to provide wifi for the many  wifi devices we have at home? The answer is to connect a second router to the primary router. The primary router will only provide ethernet for streamers, switches and/or NAS in audio system, also for the second router.  Second router provides wifi for the home, this scheme keeps vast majority of rfi out of audio system streaming chain. My own measurements find rfi significantly diminished in primary router, more than mulitiples of ten times lower vs wifi enabled. This was seamless install with the Netgear routers I'm using. There may also be value in provisioning higher quality routers. My new primary router, Netgear XR1000 is marketed as a gamer router, claims of lower ping time, latency, jitter vs other routers. Since my old router, Netgear RS7000 didn't have means to monitor ping time I can't provide evidence of this claim. Whatever the case, my XR1000 ping time test measurements are as follows, 25.35ms highest, 16.50ms lowest, this is A+ measurements against objective criteria. Ping time under load is download 25.93ms, upload 37.34ms, idle 17.31ms, this rates as A. My speed of 565gbps rates B grade, likely need 1gb service to get A here. At to how this all pertains to sound quality, adding up the upgrade in ISP speed and the off loading of wifi is without a doubt one of the most substantial, if not most substantial network upgrades I've experienced. While I  long considered my setup as having a vanishing low noise floor, with this setup I heard a new level of vanishing if such a thing is possible. Even more astounding was a more analog like presentation, while I wasn't aware of even the slightest digital presentation prior, this upgrade certainly exposed it was indeed there. It seems logical to conclude there has been some lowering of jitter here.

 

And then we come to the ethernet filter. I suppose audiophile switches can be considered as one, then we have actual filters such as Network Acoustics Muon, my JCAT Net XE and others. I continue to believe these necessary even with the all measures above.

 

Optical conversion is also valid approach post router. While I found generic FMC somewhat effective, at this point I prefer ethernet. On the other hand I've not yet tried optimizing a fiber solution, for example two Sonore OpticalModules, both powered by lps, further upgraded with Finisar optical transceivers.

 

Assuming one has high resolving audio and streaming systems the above network optimizations should provide for substantial sound quality improvements. In my system, perception of performers in room has been taken to a new level of intimacy, meaning a more emotional connection to the performers and performance.

 

At this point, I consider network has been fully optimized, the only upgrade I'm aware of would be ISP upgrade to 1gb.

sns

My EtherRegen is starting to fail on me now after 2 years. I was using this as an FMC. I think this is the 3rd FMC that has failed on me. So, from now on I am going to use simplicity for streaming. A network switch with SPF into a Sonore OpticaRendu or directly into my Lumin X1.

I had lined up a buyer for my Lumin X1 because the streamer in it was not better than my Sonore OpticaRendu (X1 streaming was disappointing). However, just before I sold it, I was getting failures in the streaming. So, I cancelled the sale and started to see what was going on.

It was the super-hot EtherRegen that was at fault. When I took the eR out and went direct fibre from my Ubiquiti network switch, the streaming sounded so much better. Now I could hear why the Lumin X1 was so highly regarded as both a DAC and streamer. The streamer is now as good as the OpticalRendu so I am going to sell 2 of my oR’s and use the SPDIF and USB stream out of the X1.

The lesson learned for me with streaming is to use the KISS principle. I think I got lucky with that eR failure.

 

I have just placed an ADOT FMC in front of my EtherRegen. It feeds into the ER's SFP socket, with Ethernet out of the ER's B port

The effect has been transformative, as significant as the original introduction of the ER - maybe even more so. It seems clear to me that the quality of the ER's input doors matter, though this might be a symptom of the Netgear switch which feeds the whole system. 

I did also try feeding the FMC from the ER but it didn't sound great. 

Really, the only 'quality' FMC out there is the OpticalRendu, quality voltage regulation, filtering, clocking. Runs warm, not hot so reliability should be good, hot is enemy of longevity, I'd send back any component that runs hot 24/7, won't last. I do recall conversation about running Uptone and other network devices from lower voltage lps, lets say device can run on anything from 7v-12v, run on 7v for cooler running,  max reliability. But then I also recall some claiming these devices sound best when running max voltages. I've run network devices on a variety of voltages, can't say I've heard any noticeable difference in SQ.

 

So, first listening session with Finisar 1475 in OpticalModule, OpticalRendu, also AfterDark Ref 2m optical cable. First of all absolutely no glitches, drop outs, nothing out of ordinary. On to SQ, sorry if I sound like I'm blathering, but I can only say, WOW! Talk about revealing/resolving/transparency, this set far surpassed my expectations! The lowest level details are now resolved, and the higher level info in mix is more nuanced in that performers have become even more real, breath for singers, technique for all the other instrument, things like use of bow on violins, fingering on guitars, it all becomes more real. Sound stage has even greater precision, size, more organized and individualized performers, greater image dimensionality, each performer within a bubble of 3D air, so natural integration with rest of performers, in other words, no cookie cutter images which is something I've heard with some other network upgrades tried in past. Some think precisely drawn images is a good thing, and symptomatic of best quality clocking. I totally disagree, live performers don't have tightly drawn images, don't want my system reproducing hifi pyrotechnics, I'll take natural sound stage and images any day. Bottom line this level of resolution/transparency is both addicting and revelatory!

 

Now I come to another important aspect of performance I desire from my source, whether analog or digital, this being timing and/or flow. Analog does this well, digital has a harder time, jitter being the culprit. First few listening hours felt things a bit uptight, not terrible, but not what I had been experiencing with stock transceivers. Also things felt a little closed in, not quite as much spaciousness as before. Over the hours spaciousness was first to come back, by last couple hours greater sense of ease imparted. Presume this burn in issue, symptoms typical of lack of burn in. 24 hours, beginning of session, 30 by end. Final judgment awaits with more burn in, also still in final stages of Duelund Cast coupling capacitors burn in in 300B amps.

 

Bottom line on these upgrades. This is far more than incremental improvement, rather its meaningful, profound, one of the greatest improvements I've experienced from network upgrades, could turn out to be number one. I also have to take into account this result after having fully optimized every single link in streaming chain, already had extreme level of resolution/transparency, thus, this change magnified by what had come before. Now, on a possible downside, I could see this upgrade as possibly too revealing for systems with ANY prior defect. If one has some undesired coloration this will expose it without mercy. Keep in mind, I'm not fully burned in so I may detect more forgiving nature in time, but I expect this more forgiving nature will come from the flow/timing thing. I don't expect there's any flavor with this upgrade, no warmth added, just purely revealing, not unforgiving, not forgiving, just is.

 

@thieliste has more experience with various transceivers, perhaps he can offer advice for those who may suspect 1475 may be excessively revealing for their current setup, perhaps the 1318 is better fit. I can only say these modules are like having an entirely new streamer, seems much of OpticalRendu performance hinges on these transceivers!

Hi @sns

Just a point regarding the EtherRegen running hot. I’m no electrical engineer, and it’s 3 decades since I completed my physics, but I would expect the higher voltage to generate less heat.

If we assume power requirements to be constant, then a lower voltage means a higher current.

Power = I * I * R

I run my ER on 12V, and it’s only slightly warm. My cheapo point and click thermometer bought for Covid etc reads at 38C (we don’t do Fahrenheit over here :-). I know it’s on a different scale, but electricity is transformed to high voltage for transfer over distance to minimize heat loss.

@lollipopguild You may be correct, I just seem to recall reading somewhere these devices throw off the unneeded voltage as heat.

 

I just should also contextualize my impressions of the 1475 as not being used exclusively as network device in my setup, use in streamer may impact sound differently vs use in network only devices. I also have the AffterDark optical cable which undoubtedly has it's own impact.