This is one of the reasons that I like linux-based endpoints (either Roon Ready devices or the linux RoonBridge). At that point it’s a single-function computer and can be easily (relatively) optimized for audio playback and nothing else. All of the optimization products available for general-purpose operating systems work from the standpoint of disabling functionality, but it’s so much easier to get right if the only things loaded are those required make playback work.
This is one of the most important factors of digital audio reproduction and one that is typically overlooked as it’s difficult to boil down into something that the layman will fully grasp. So much of digital signal processing is counter-intuitive that it takes a lot of time and a lot of analogies to connect the dots so that anyone can understand what’s going on. It’s much easier to get caught up in specifications and jargon (SDM conversion vs R2R is one that I always get a kick out of).
Back to the topic at hand… I don’t doubt that different playback methods sound different. I’ve verified this for myself many times over. What I’ve come to realize is that with the proper amount of care taken in designing the overall system the differences between different playback methods start to converge very quickly.
Assuming that output is bit perfect and jitter is a non-issue, I have trouble believing that in a purely non-mechanical system (with SSD) with no NAS, the difference in electrical noise induced from a few percent of different cpu load/cycles due to processing overhead would cause audible differences on the usb output. I also have a USB isolator post-pc that should deal with noise on the USB lines.
Regardless, assuming the above two conditions, then that’s all of these system optimization changes are going to do right? I will try some tests with all system optimization turned off in A+ and see if how much/if any difference it makes. According to the preferences it a) prioritizes the A+ app, and b) disables 3 system sub-services - I am not sure it’s any more convoluted as suggested above. Alternatively, these optimizations can also be made when Roon is running as well. I will try that too.
I am not suggesting that using a separate playback endpoint won’t improve the Roon SQ, I am sure it will help. But in my simple non-mechanical system with storage on the same machine and going directly to USB, it still doesn’t add up in my mind …
If system optimizations are in fact the main difference and do have a large audible difference, why can’t Roon offer them if a user so chooses … ? It will interesting to see what ROCK does to SQ … in theory this highly optimized OS should mitigate any of these differences and more.
[quote=“nquery, post:43, topic:22674, full:true”]
If system optimizations are in fact the main difference and do have a large audible difference, why can’t Roon offer them if a user so chooses … ? It will interesting to see what ROCK does to SQ … in theory this highly optimized OS should mitigate any of these differences and more.
[/quote]R-O-O-N-B-R-I-D-G-E
That’s a second/separate computer/nuc - I don’t need it. My point is why shouldn’t Roon be able to sound as good as A+ in a single box solution for simple setups?
Because Roon was designed to sound better using a distributed architecture. It can’t be any more plain than that.
The Roon core is doing a huge amount of processing in the background regardless of playback status. Surfing through your collection on your iPad while listening? Well, the core is slicing and dicing your database while filling in the blanks with queries to the internet. Other software packages don’t do this and as a consequence have a user interface and functionality which pales in comparison to Roon.
Want the best possible sound out of Roon? Invest a little bit of cash in a dedicated endpoint that’s optimized for Roon playback.
Ok so standalone Roon use more CPU than A+ in my simple use case - lets say 12% vs 2% cpu on my mac mini. So it’s certainly doing ‘more’ of something.
I have also set up Roon in bridge mode. While Roon Core is now using even more cpu on a more powerful MacBook pro, i would assume that this is because of the work it is doing sending data to Roon Bridge. The roon bridge running on the above mac mini now uses almost 0% cpu.
If your theory holds true I should be able to hear an improvement in SQ - I will get back to you after some listening
ROCK is very helpful from a support standpoint as a lot of the issues with using Roon on a general-purpose system are now gone. Now MS updates breaking things, no virus/firewall programs, no strange network drivers. Just a known-good configuration and a very simple OS.
The core can have problems delivering when installed on a machine that is more of a general-purpose computer. Roon uses a lot of resources and when the computer is resource-constrained then Roon starts having problems (usually slow browsing and/or dropouts).
ROCK addresses these two issues (and others) by providing a very streamlined and simple platform on which to run Roon. No management. No OS updates. No fuss.
I think you should use Audirvana as your program for listening to music via a local PC.
Why would you bring so much drama into your live and use Roon on the PC where the Rooncore stays.
This machine runs heavy dutie software, unlike the Audirvana Machine.
The R-O-O-N-B-R-I-D-G-E was a perfect invention, seperating the Rooncore from your listening system.
Thats what makes Roon so nice. You can build a Musicsystem to your preferences. Totally flexible, seperated from any PC.
I would never use a PC to listen to Music. To much noise.
I have an A+ license too. I dont use it because of some other serious shortcomings (NAS mounting problems and etc…). But when it allows playing Music via a networked setup, I give it a try.
I’ve found PCs need to be tweaked for good sound. Buying the best low noise ATX power supply like the fanless Seasonic X-460FL. Buying top end gaming motherboards boards when I building PCs because they have numerous user bios settings.
For example, turning off Spread Spectrum avoids base clock modulation (deliberate system jitter), setting power phase functions to extreme avoids power fluctuations and PWM frequency switching throughout the system and gives a steady supply of power. Turning off CPU power saving modes, like, Intel SpeedStep will keep the CPU at a steady frequency. Turning off intel CPU turbo mode avoids frequency switching/power fluctuations which can generate noise and, again, keeps CPU at one frequency. The more stable and constant the system the better. Sound is less fuzzy and edgy, much more focused.
Well, people have been using PC’s to listen to music for a long time, with fair success. I am not arguing that dedicated playback endpoints, pc’s or not, with isolation between architectural layers/components, won’t sound (much) better. As a software engineer, I think the Roon platform is a fantastic piece of work - very well architected, stable/solid, and excellent UX.
And I really want to keep using it. That’s why I am just trying so hard to figure out why, apples to apples, there is a difference in SQ between Roon and some other players (like A+) on a single pc install. All else being equal, is it just because Roon eats up more cpu and thus generates more electrical noise? Sounds interesting but debatable. Jitter control? Maybe not. Or maybe A+ actually colors the music with DSP to make it sound different? Or is it the system optimizations that A+ makes? If so, then turning them off should equalize things., or ROCK should mitigate these things. But there must be a reason more technical/specific than ‘Roon was built for a distributed environment’ …
@AMP is not wrong. There may be an SQ improvement with ROCK on a single computer compared to a full OS, but that is not what it is designed for. It is designed as a Network “appliance” device to run RoonServer and send audio to RoonBridge or RoonReady devices.
There are many ways to improve audio on a single computer. Pick how much you want to spend and someone will take it in order to help you on that quest. Or you can not play that game and just use a network connection to an inexpensive optimised thin client.
One of the most important comments I’ve ever read about computer audio was the following by Chris Connaker in the CA thread regarding his SonicOrbiter review:
I setup five different computers as Roon servers to send audio to the Sonicorbiter SE.
Windows 8.1 noisy, non-tweaked (hardware or software) PC with 6TB spinning drives of local storage.
Windows 2012 R2 server SOtM / AO server with highly tweaker hardware and software.
OS X El Capitan MacBook Pro
OS X El Capitan / Windows 10 (Boot Camp) MacBook Pro retina
OS X El Capitan iMac 5K
I couldn’t hear a hint of difference between the sources when sending audio over Ethernet to the RoonReady Sonicorbiter SE. This combined with the fact I have no idea how the source could possibly effect the sound of the Ethernet endpoint in this situation, lead me to believe the source has no effect on sound quality.
Note: I completely respect the opinions of others who will try similar tests as more RoonReady endpoints become available. In addition, my conclusion here has absolutely nothing to do with locally attached server / DAC combinations not using audio over Ethernet.
Ever since I read that I’ve regarded one computer setups as quaint expensive anachronisms.
Let me support this with a more general observation about system architectures, beyond the world of audio.
I have been a software architect for 45 years. We used to think of systems running on “a computer”. But no more. And certainly not in the next 45 years. All modern system architecture is disaggregated and distributed. Computers in a network. Computers connecting to the cloud. The systems running in the cloud. (Cloud services do not run on “giant servers” as suggested in a Spotify thread, they run on a vast quantity of cheap commodity servers.) A car has hundreds of computers. Each “computer” is really a network of separate processors, not just the cores we know but specialized processors, some high powered for advanced work, some miserly with power consumption. Power is a very big deal, both because of battery life and because of cooling. We can’t make computers faster because they melt, but they are cheap so we can have many of them. “Many” doesn’t mean tens, it means thousands and millions.
So we have a recursively distributed system. Nowhere is there a trend to tight coupling, toward a single box.
Having many systems sounds expensive and difficult, to buy and configure and manage. But distributed goes hand in hand with small and single-function.
Small and cheap, easy to configure and manage and service because single-function, easy to replace because loosely-coupled.
These are the trends everywhere. Cloud, internet of things.
The SonicOrbiter and MicroRendu are harbingers of the new world. And ROCK.
