It's your streamer, not your modem

So many discussions I've seen lately have been about upgrading Internet devices, especially the modems and routers to get the best possible audio.  Audiogoners are talking about installing 10 GigE (10 Gigabits per second) cable for signals that barely need 10 megabits per second.  Three full orders of magnitude more bandwidth than required by hi resolution audio.  (192 k/24 bit)

I've also seen discussions about home Internet getting a little higher latency and jitter.

None of this should matter with a decent streamer.  Let me give you an example.  Because my work requires me to be online with high reliability I have two different Internet providers and a switch that detects failure in one and switches me to another.

It takes the switch approximately 40 seconds to detect the Internet is down and fail over to the other.  40 seconds.  40,000 milliseconds. For this testing I shut the modem off.  In that moment, for the next 40 seconds, I had no working Internet.  Then my back-up 5G Internet took over.  About 3 minutes after that my primary Internet's modem has rebooted and my router has recognized it as available and switched back over.

During the testing I coincidentally had Roon playing a random Jazz selection.

Not once did my audio stop.  Not even a hiccup.

Why?  Buffering.  Roon had gotten the entire song and doled it out to my end point a little at a time. 

Point is, modem quality, router quality, switches, and Ethernet cables don't matter that much.  What does is the size of the buffer and the effectiveness of the anti-jitter circuitry in the DAC.

I do by the way recommend shielded cables, Ethernet isolators and gas discharge surge protectors, but sweat a modem or router?  Not me.


Yup. Best possible streamer first. Digital cable, power cord then Ethernet cable. Unnecessary added clutter with converters, filters, switches etc. is just complicating things, adding its own noise that leads to linear power supply upgrades, adding more cables and clutter, more money dumped down the drain. This cure, in my experience, is worse than the disease. I’d say play with switches and all these gadgets after you have absolutely nothing else to upgrade. 
Just my $0.05 inflation adjusted. 

and isolating all that gigital claptrap on the the non analog leg of the panel with at least a Furman grade power conditioner……

and isolating all that gigital claptrap on the the non analog leg of the panel with at least a Furman grade power conditioner……

Paying attention to where your wall warts are is also important.  My network closet is on the opposite side of the house from my home theater and well isolated.

@erik_squires to confirm your post.  I have had a lot of internet problems lately.  Streaming of TV stopping. Computers disconnecting.  lots of latency and intermittent service.  

Through it all I've been very impressed that my music just kept playing.  A couple of times it did shut down but it just stopped and I restarted it.  Much, much better performance that my other internet components.

For info:  Streamer Grimm MU1.  Roon.


Thanks, @carlsbad2

I’m not saying it’s NEVER the Internet, at all. The performance of a lousy Internet with live streams ( hi Jazz FM 91 Toronto! ) is quite different. Last year it would run for like an hour before quitting. At the same time my browser would also stop responding. That was before I had Internet fail over.

My network provider has a multitude of ways it fails, and that’s why I’ve implemented a hot-fail over solution with ongoing monitoring.  Also, for whatever reason, about 2/3rds of the time when I have an Internet outage it requires a modem reset because it just won't come back on it's own.  Now, will this modem reset work, or am I going to have to wait 6 hours until it does?  I don't know, but that's what my automation is for.

I am saying that the overall quality of your devices in the home or looking for 1 Gig data plans isn’t going to help you.



I completely agree, although I came at it from the listening side of it. I am sure electrical, physical interference, jitter… etc are the cause. The best solution is a good streamer.

I’ll chime in to say that the streamer is incredibly important, but the technologies in the router, switch and Ethernet cables in my chain make a huge, audible improvement as well. Take it for what you will as I am an authorized Synergistic Research dealer, but I purchased their Ethernet Switch UEF long before I was a dealer for them. It replaced an Uptone EtherRegen with a $700 Farad Super3 LPS and $1K power cable, and the SR Switch delivered a quieter noise floor, more relaxed, but naturally discernible Sonics over my previous switch setup. Moreso, I added the Router UEF after I became a dealer, which further lowered the noise floor and  increased the harmonics. All of this is feeding my Aurender N30SA.

It is a bit of a PITA to have a fully separate audio network apart from my home network, but the sonic advantages are worth it to me. YMMV. But where Erik mentioned router choices for speed, it’s not about that. Internet devices can be crazy noisy, and in SR’s case, benefits are evident given the way they handle the EMI, reclock the signal, etc.

At the end of the day, I agree a better streamer is a more worthwhile investment, but if you’re already investing in a quality streamer, do not hesitate to explore opportunities for improvement further upstream.  

cable is inherently noisy and using isolators and switches to clean it up has proven effective widely. It isn’t just the streamer’s buffering, EMI/RFI and ground level noise do corrupt signals beyond most DAC’s cleanup capabilities and reclocking makes subsequent stages at both streamer and DAC audibly better. In audio everything matters and monocausal explanations tend to be wide off the mark

All your saying may be true in part But

These pathetic $5 wall warts for router and ethernet switch

Induce a Ton of noise , put linear power supplies on them  it would dramatically 

Lower the noise floor , and decdnt quality cables 

A must if you want a streamer that sounds respectable 

And a good digital power cord .

I purchased a  Audio Art Signature digital power

Cord and sounds much cleaner then more expensive 

High current power cord , very surprising but true

Everything counts.

@audioman58 I hope you are right about the Audio Art Signature digital power.  I ordered one last Friday.

This all gets a bit circular, but have to agree the best place to start is with a highly capable streamer. But after getting that into place, even a decent $30 switch can make a considerable difference. I couldn’t ignore the findings and suggestions from Alpha Audio and Hans Beekhausen - much as I tried! I just have to tweak to see what cheap fixes actually prevail. But seriously, was a more than palatable improvement in sound quality, tonal balance and realism with an upgraded Ethernet cable and a d-link switch before a Pi2design Mercury streamer - which is in no way a slouch of a streamer/transport to begin with. This was even after using a quality filter in that chain. Of course an Ifi wall wart for the switch and a decent linear power supply for the Mercury helped as well.

Certainly imagine that something considerably higher grade like a Grimm audio or upper echelon Aurender might not as much need it. But until then, very very happy with the results.

@erik_squires One time I unplugged the Ethernet cable in the back of the streamer and the band played on.

I agree the streamer is more important than the things that come before it. But the HiFi network switches also help. My PhoenixNET makes a positive difference I can’t live without (although I’ll be trying a Tempus soon to see how it compares).

The real question is, dollar for dollar, if you were to invest in upgrading my streamer vs using those same number of dollars for a HiFi network switch, which will get you further? In my particular case and with my circumstances at the time, the winner was the network switch. I know because I tried a series of more expensive streamers and to be honest the network switch made as much or more of a positive benefit than the streamer upgrades I tried (Innuos Zenith mk3 + PhoenixUSB Reclocker -> Grimm MU1, Aurender N20.



Erik/Nigel, I don’t think most of the respondents got your point, or perhaps, they don’t agree with it. Increased buffering capacity is not an expensive, exotic feature on a streamer, yet the respondents above are mentioning some of the most expensive streamers on the market. Most seem to be clean to their gadgets as well. or, perhaps, I am misreading it.

@bruce19 I see your point.

To be clear, I'm not advocating for a more expensive streamer, just streamers with better buffers, and honestly I have no idea who they are.  It's not a specification I often see published or mentioned in reviews.

Dollar for dollar, room treatment is much more important than an Ethernet switch. 

Is streamer A for $4,000 better than streamer B at $1,000 for managing Internet weather?  I have no idea. In many other areas I've seen exorbitant prices without any guarantee of better performance.  One example is galvanic isolation of USB ports.  It should be standard today but it isn't, even on expensive streamers. Sometimes they have it and sometimes they don't.

I'm doing eveything wrong.

I stream through my Asus ROG laptop computer which is connected to my network through wifi. I hook up the laptop to my Berkeley Alpha USB/Alpha Reference DAC Series 2 MQA, via cheap USB cable. When I play a Qobuz file and I'm careful to play the same version as a CD I own, the sound is identical to the CD played through my Jay's CD 3 Mk III into my Berkeley DAC.

When people implement all this esoteric stuff to their digital infrastructure, do they ever compare the sound to a CD? Does anybody ever get to the point where streamed music sounds better than the same song/version played from a CD? How bad did the streamed music sound to begin with?

I don't see how a streamed file with all of the complicating issues that have been described in this thread can ever sound better than the same file played from a CD through a good quality transport into the same DAC.

I don't get it.

@erik_squires I guess buffering could be done either on the music server, or the streamer itself. For instance, I have an M1 Mac mini that serves as my roon hub, and it delivers the stream to various endpoints on my home network. I suppose it might be possible, even probable that the endpoints have some degree of buffering built into them. My hunch is that they do. But you’re right I never see this mentioned or discussed in tech specs. I would guess that buffering at the endpoint is not as critical as where the stream comes into the home, but that could be a legitimate point of discussion.


I'm doing eveything wrong. I stream through my Asus ROG laptop computer which is connected to my network through wifi. I hook up the laptop to my Berkeley Alpha USB/Alpha Reference DAC Series 2 MQA, via cheap USB cable.

Well, certainly not wrong in my book! As a previous apartment dweller, I know how bad Wifi signals can be in a densely populated apartment building, but in a home with sparse neighbors it's perfectly serviceable.   There are always two things I worry about when using a PC or laptop as the source:

  • Digital ground loops
  • Power supply noise

If those two issues are mitigated I think they're perfectly fine.  One way to test for this with a laptop is to let your laptop discharge to around 50% and then plug your power supply into the AC. See if you hear any noticeable difference in playback quality.

To prevent these issues I keep any computers outside of the clean side of power conditioners and use a USB isolator.  The further away your PC is from the stereo, the more important a USB isolator becomes, as ground loops are more likely to occur.

Otherwise, I never bother getting very esoteric.

I like this discussion... I think Eric, et-al are correct from the standpoint of buffering, however there are issues I’ve found with poor Ethernet regarding noise, etc... I just finished a response to a recent discussion on streaming services which I believe applies to this discussion...And Ethernet bandwidth is usually NOT the issue.  Here are parts of it which I believe apply.

.I had issues with streaming music versus the EXACT same album loaded on my Innuos ZENith MK3 with a Phoenix USB re-clocker feeding my Denafrips Terminator-Plus DAC. I use Qobuz which in my listening opinion sounds slightly better than Tidal. No matter what time of day though did streaming sound much better until I added in an Uptone-Audio EtherREGEN device between my Ethernet modem and the ZENith. Finally the music streamed was very close to that loaded onto the ZENith. I believe re-clocking the Ethernet signal with high-quality clocking and noise isolation from the EtherREGEN made all the difference no matter what time of day. It’s the most cost-effective $-for-$ I’ve spent for digital sonic improvement. My present cable internet service is ~300mbps. I will soon be going to fiber-optics with another ISP and their minimum service is ~500mbps. So, with this change I’ll know if anything improves, as I will keep both running and do a direct comparison before cancelling the cable service. I’ll let y’all know what I find ! In my opinion, I don’t think for audio you would need any more than what I have presently for internet speed as sonically the streamed music is extremely close (splitting hairs) to that stored on my music server. Finally, I understand about buffering and the moving of that data from the music-server to the DAC and theoretically it shouldn’t matter what’s going on upstream, however in practice, cleaning up the Ethernet signal in terms of noise, clocking and jitter obviously has made a difference in my system and listening. A positive thing for me that’s come out of this and a few other recent discussions is that I really need to look at the modem wall-wart as being a culprit in noise, etc. not only on the AC line, but in causing issues in the signal path as well... I have the modem wall-wart on a separate AC line from my listening space, but that doesn’t eliminate other signal-related anomalies created by it.

Buffering and caching are two entirely different concepts.
Every device that intakes a data stream from network or ISP will buffer that data stream. It is done to create a stable consistent flow of data irrespective of the incoming speed. If it’s too fast, the buffering regulates it down to the data flow that meets the requirements of the downstream processing. Same with slower speeds…it will buffer the data up until the requirements are met. Buffer size us typically not that large and is dictated by the design and the needs of the downstream components/processors.The data is not maintained in the buffer for too long.

Caching is typically a much larger data store (could be in memory or SSD) that caches the result set and stores it. Depending on the design, most of the critical processing may actually happen from the cache and that could result in cleaner downstream processing because some of the impacting entities like noise, etc had already been taking care of. Reading and processing data from cache in this case is similar to reading a CD. You can perform the remaining processing and further purify the data when it is converted to either the USB or SPDIF outbound signal that the DAC will understand.

If you pull the Ethernet cable out of your streamer that utilizes buffering only, you will probably get about 30s to a minute of play.
In the case of caching, depending on the defined size of the cache and what’s been cached, you might be able to listen to the entire album or the entire playlist.

In my experience, the impact of Ethernet tweaks is less evident with streamers that use caching. For example, with the Aurender you can A/B two different Ethernet cables by caching the same song twice, version A with cable A and version B with your second cable. It would be impossible to do with the buffering only streamers. Data in the cache will remain until it rolls of due to capacity or you clear the cache manually.

Comparing a ripped file to Qobuz or Tidal might also be a fool’s errand. You can never possibly know and ensure that the two versions are the exact same master. So the difference in sound may be attributed not only to CD vs. Streaming but also to the version of the streamed album you’re comparing your cd or cd rip to. The most critical link in the chain is immediately before the DAC. If you’re using a server, what happens on that server as long as it’s properly sized, is far less critical.

Quality of the streamer is extremely important. You can never make a mediocre streamer sound like a great one by adding gadgets, no matter how expensive they are. Just like you can never soup up a Civic to perform like a proper sports car, no matter what you change or add to it (not knocking on Civics…great cars).

Just adding my $0.05 here…


Let me help you.  Go to Wikipedia and look up "Transmission Control Protocol" and  then in there search for "flow control." I'd link it but A'gon's firewall is blocking URLs

You are confounding  "buffering" with "flow control."  There is almost no buffering at all in network devices between say Netflix and your TV.  None.  What there is, which you describe, is flow control, which limits the amount of data on the wire at any given time to prevent packets from being dropped.  This is entirely negotiated by the endpoints.  There are no mini-caches strategically placed around the Internet just in case your Internet provider is congested. 

Buffering is to use a cache for the sake of preventing interruptions via playback.  This is entirely an endpoint thing.


Buffering and caching are two entirely different concepts.

They are different, but not in the way you explain.

@erik_squires is correct. Here's the link he's suggesting.

Who said anything about caches being strategically placed throughout internet? I was talking about a cache in the streamer. 

From the same Wikipedia article….

Flow control: limits the rate a sender transfers data to guarantee reliable delivery. The receiver continually hints the sender on how much data can be received. When the receiving host’s buffer fills, the next acknowledgment suspends the transfer and allows the data in the buffer to be processed.

I didn’t word it exactly as the wikepedia article did but the concept I explained is exactly that. You are splitting hairs. 

@audphile1 Funny, you are the one trying to describe buffering and caching as two different things. 


Why? Buffering. 

Finally, someone understands how it works. I learned this stuff in the '80's when I was setting up computer networks between mainframes, mini computers and POS systems. IOW, this isn't a new concept. I don't know where all this new talk of latency and jitter is coming from. The transmission of data is not where those things would happen. It seems audiophiles always have to invent a new problem in order to have an excuse to buy more equipment to fix it.

@jssmith - I could see lesser streamers with poor designs being more subject to upstream issues.  The writing into and the reading from the cache has to happen without the system tripping over itself, and then there is also decompression of FLAC, ALAC or mpx (mp3, mp4, etc) formats.  Writing non-blocking IO code for embedded systems really is an art form, so I could see a hastily put together system with limited CPU resources and very little stress testing performing exceptionally poorly, but my home router isn't going to make things much better.

To be clear, a Raspberry Pi 3 could do this really well, so this should not require server class hardware to do this.  It does require engineering attention.