No, they (especially Wadax, what a joke) just do a great job of using fancy-sounding but completely meaningless language to convince you that you should pay tens if not hundreds of thousands of dollars to achieve result either equal to or even worse than what you could get for a couple grand.
Introducing noise into the signal path is pretty easy, really, but noise in digital sounds very peculiar. For fun, scratch a CD, and then rip it ignoring errors. There will be no āless blacknessā or āveils loweredā or āsoundstage collapsingā (advice ā turn the volume down if you try it).
You know what is the strongest (analog) noise that goes into the DAC? The digital signal itself. Somehow DACs do not seem to have a problem with that. And if a DAC doesnāt filter mains hum and needs an external device for thatā¦ well, itās just a crappy DAC.
Those are actually called āresolvingā.
Almost missed this
No the reason is that there are enough people with money but no knowledge of what passed for high school physics where Iām from, and it is extremely easy to sell them pricey geegaws. No different from multi-thousand dollars power cables, cable risers, or maghic crystals,
AudioQuest (not even the worst offender) will happily sell you a directional ethernet cable. They literally say āOf course all AudioQuest Ethernet cables honor the directionality inherent in all analog and digital audio cables; arrows on the jackets indicate the direction (from source to destination) for the best audio performance.ā Seriously.
This is all the audio equivalent of a Nigerian prince wanting to give you lots of money if you just help him out s bit.
Yup. And you need to have golden ears to hear it, too.
If all else fails, Mr. Swenson could have a second career being an MC for IgNoble Awards.
Of course if he actually believes what he wrote thereā¦ thatās scary.
Those electrons have absolutely no sense of direction. Thatās why they line up the grain in the wire to avoid confusing them
I agree. But Iām not talking about that. Iām talking about the network stack on the streamer and how network jitter could influence things like interrupts and higher load on the CPU / PCI (or other) bus which could then make subtle differences in sound quality.
Iām going to link to some work @David_Snyder did a long while back to look at RTTs using various OSs on the same streamer hardware. To his ears, at least, lower RTT tended to provide an overall better SQ. Iāve also heard differences with different kernel versions when using a CM3 and thatās with no network changes. So, just looking at network isnāt the whole story here. And, believe me, Iām a network engineer of a few years and I can argue all day long about why the network doesnāt matter hereā¦ but then Iād be ignoring how the OS and applications function which is actually more complicated these days.
I could also theorize all day long why things like network jitter influences sound quality on specific OS / hardware combinations but I just donāt have the time to test any of those theories so I wonāt waste anyoneās time.
Link to postā¦
I just wonder, where do all those electrons, blindly following the arrow, end up, and what would happen if one day they break loose!
But how would different load on the digital source affect sound on a downstream device?
All those cute graphs are very nice but absolutely useless. If there were graphs of sound (or even electric signal coming out of the DAC) and they looked different for different OSs running on the source, then sure, weād have something to talk about (and a few Turing awards, and probably Nobels as well to share).
I also did not see any mention of the subjective sound quality (granted, I only scanned the thread) being evaluated in a blind ABX test. If it was sighted, well, other than a few pictures of network throughput, which might be of interest if you were trying to build an AWS competitor out of RPies, it is about as useful as a green marker for improving sound of CDs.
Iāll say, Iāve never seen a Wireshark capture used to explain differences in SQ before. Too bad an audio analyzer wasnāt also used to correlate an objective measure (RTT) to other equally objective measures of the actual audio output (e.g. THD or jitter). Audiophiles would be quick to dismiss audio measurements as irrelevant, but RTT may get a free pass.
It wouldnāt and Iām not talking about load on the source. Iām talking about what happens at the streamer; the thing feeding the DAC (internal or external).
Streamers sound different, they just do. If youāre using a Raspberry Pi then the kernel you run can sound different. I donāt know why. I can speculate, and a part of that speculation is kernel latency and how optimized the network stack is. What the graphs show is variation in how long it takes the kernel, network stack, to ACK the TCP packets used for RAAT. This can be an indication of increased delay or cycles within the kernel to process the incoming packets. However, it does not give any insight into the actual cause of the increased RTT so it could be completely meaningless to what the tester was hearing. Itās a good data point to dive deeperā¦ of which Iāve not seen anything published here doing that.
This whole thread was started by OP asking why two different bits of streaming software would sound different on the same hardware / network. I responded that Roon is about 2x traffic on the network. Some people think that makes no difference. I think it does and especially on wifi. Increased jitter / RTT can have an influence on audio quality if that jitter is caused by the streamers kernel / network stack. And I know that kernel latency can be exacerbated by network jitter; especially as you approach network saturation. Thatās all Iām saying. And, honestly, thatās not saying much.
Without audio measurements and/or blind tests, this is not a fact, so no explanation is needed.
Maybe they do (especially if they have different DACs). But somehow it is interesting that if you ask how that difference had been found for systems that should sound the same, you get either 1) āA valued advertiser sent us an expensive piece of equipment for review, of course it sounds great!ā or 2) "I just paid $1000 for a USB cable, of course it improves the sound, do you think Iām stupid?!: or 3) āI am Joe Schmoe from UpYours Systems, hereās a whitepaper explaining how flux capacitance of network cables needs to be aligned with quantum tunnels of directional OFC copper strands to minimize weak interaction between quarks of logical gates in your DAC, which might cause local fluctuations of time-space at the output stage, resulting in soundstage shrinkage.ā
When thereās a reasonably well designed, reproducible demonstration of those sound differences under blind test conditions, then thereāll be something to look into (every major electronics manufacturer will jump on it, not to mention research universities and labs, thereād be some major discoveries to be made there. Until then, this is all only about some people and their money being soon parted.
Iāve done blind tests. They can differ. Not a lot, not always.
You all welcome to come over. I can load different kernels on my Raspberry Pi4. Sorry I have no test equipment here. But you can listen to whatever you want. I can even load a kernel on my CM3 which destroys the sound stage width because the highs are a mess.
And my system, to most āhigh-endā standards, isnāt high-end.
When I say āstreamerā I mean digital in as network and digital out as bitstream to a DAC.
Not true in my experience.
Set-up your system so you have both sound stage width and depth. Imaging should good enough to identify seperation of instruments. You should not be able to locate where the speakers are from what you are hearing. Speakers need to disappear. There is a set of test tracks I have where the singers start at 3 feet, then go to 6 feet, then 9 feet. You should be able to hear how far away they are in each track. Then, weāll swap kernels on a Raspberry Pi4 and youāll hear a difference. Yes, a $45 streamer with different, free, kernels. Youāll hear a difference. But youāve got to start with a properly set-up system. This is not an uber expensive experiment in money only time.
How long does it take to change the kernel? And can you do it blind?
If not, itās an interesting exercise, but does not really prove anything.
I think my system(s) are pretty decent (itās more dependent on the recording being good). I can definitely get good soundstage out of them (although I do need to find better subwoofer placement for one of them, it works great everywhere, except on the couch where I actually want to sit bass nicely cancels itself out with backwall reflection ) In one of them I have replaced an integrated streamer/amp (NAD) on WiFi with a combo of DAC/Amp (Peachtree) fed by an old laptop on ethernet. Nope, canāt hear any differences in soundstaging, veils, or anything. I could say that Peachtree produces better bass (itād better, having double the power of NAD) but without blind testing even that would be a stretch.
Seriously, if you can demonstrate that 2 identical RPis, with different kernels, running exact same software with same settings, and connected to the same DAC in a same way produce obviously, noticeably different sound in blind ABX testing, it will be best investment of $45 youāve ever made.
The easy way is a stack of microSD cards. Power down, swap card, boot new. Some config tweak may need to happen in Roon to get up and running with the reboot.
It can be blind as long as you donāt know which cards contain which OS / kernel on them.
Itās not that easy. Audio memory is very short, in the order of seconds, so itās not the same as flipping a switch between A and B. Then, you need to repeat the test a sufficient number of times to get a relevant result - say 20. Thatās a lot of swaps.
Honestly, that takes too long to make any meaningful SQ comparison, and if you have to tweak the config files, you will likely see which kernel youāre running on.
I donāt doubt that you perceive a difference. I could perceive a difference from applying a green marker, when I wanted to. Psychology is funny like that. But it does not mean that the physical difference exists.