I came across this interesting Hong Kong forum thread about improving the SQ of Roon server. I’m impressed with the scope and depth of modifications and upgrades people tried to get from their Roon server, regardless of whether it actually makes any SQ difference or is merely expectation bias. If you think audiophile Ethernet cable is absurd or controversial enough, it uses audiophile SATA cable too. In case there is anybody here interested in reading it, use Google translate:
Disclaimer: I post it here only because I find it interesting, it does not represent my own view or my employer’s.
Since you quoted my disclaimer, perhaps I should state my own view: I believe people can simply buy a fanless NUC for use as a Roon server, and use fiber network in Lumin X1 for noise isolation.
Agree completely, although not sure about the fiber connection.
I’m just sorry about the can of worms this is going to open, again.
Having read the (translated) article I think the worms are the least of our worries:
"“The lower computer snail (Corsair HX1000) has less effect”
The original text said he used a power supply with low ripple measurement (Corsair HX1000) and got a very little SQ improvement.
I’ll try and post a comment that doesn’t get flagged by the community…!
Yes, the JCAT stuff is pretty “out there” in terms of their ideas (and prices). There’s quite a lot of info and discussion of JCAT and other “esoteric” computer mods over on computer audiophile (as you and the author of this web page probably know). He seems to be in pretty deep for sure.
Some folks profess to being able to hear an improvement in the most apparently insignificant component in an audio PC, others (especially those who work in IT) think most of this is just complete and utter nonsense. I generally fall into the later camp and think that a lot of these audiophile computer mods (not JCAT specifically) are snake oil but I have no direct experience, so my opinion doesn’t really count.
But looking at it from a slightly different perspective, I believe there are people who are never fully content with their system and just sit back and really enjoy the music but instead always have this nagging feeling in their mind as to whether their system could be “just a bit better”. There are also people who just like to tinker around with their equipment all the time (playing excerpts of tracks instead of listening to the entire song) and derive as much (if not more) enjoyment from the pursuit / journey of hi-fi nirvana as they do from the actual music itself. Conversely my brother in law (who enjoys music more than most people who I have ever met) will happily listen to music on a beat up old portable CD player and still really enjoy the music, irrespective of the fidelity.
One example is audiophile network cables. There are eternal debates as to whether they (just like audiophile SATA cables) make sense or are basically nonsense. These debates will rage on and won’t end any time soon. While I like to consider myself as an objectivist, a part of me sometimes just thinks that at the end of the day, if having a fancy network cable (battery powered SATA disk, cable supports made from unobtainium, esoteric mains fuses, optimised windows settings) enables someone to simply relax, sit back and enjoy the music more, then it has actually served its purpose. Some might call this placebo effect but that’s fine, this effect is real. It doesn’t matter. In the end, this hobby is only about bringing some happiness and enjoyment into your life, whether though your equipment, the music or both.
JCAT do not make their own stuff as far as I can tell. They rebadge stuff and charge a premium. There may be some reworking of the product but essentially they are building on principles that are already being catered for with commercial products. If you are knowledgable enough to separate that from the marketing is an altogether different discussion.
I cannot comment on the other products, but at least for the network card, I’m not able to find an alternative source for the dual-LAN i350 network card with Crystek CCHD-957 and external non-ATX power input. Its brand is also on the silkscreen of the PCB from the photo.
I like the explorative elements of these kind of discussions. But I do distrust a lot of the supposed benefits of these tweaks.
My, now flagged and hidden, posts were trying to promote an uncensored but still rational discussion. Less negative feelings of trying to silence opinions.
In my experience from work and life, and also looking at HiFi the last 30 years — some things that at first get frawned upon or seem insignificant becomes standard practice some years later. I remember online discussions about all DAC chips being equal in early 2000. As long as they successfully converted analogue to digital no chip could ever be better than the other. A simple truth now not true any more.
Same with “clean” sources. Very frawned upon until quite recently. Still by some. But now in a bigger extent proven real by for example measurements.
Personally I try to keep an open mind and look at what areas have potential that warrants further investigation, despite being educated in 0 and 1. This way hopefully I learn something that can contribute in advancing our development and helping users. Otherwise nothing can be improved. If multiple customers tell me the same thing that something sounds better than the other thing, I try to find the reasons behind it instead of dismissing them as expectation bias. Sometimes the opposite can happen too: our in-house golden ear originally claimed that phase inversion cannot be heard (which is a subject of flame war in another forum), but he actually passed blind test 100% correct in identifying the correct phase, surprising himself.
If you look at their ‘about’ page they make it clear they use contractors from other disciplines. And it was pretty much common knowledge their USB isolator was one audiophiles already knew about in its ‘non-audiophile’ guise, albeit with enhancements and their branding. I’d love to try their stuff and see if I could discern difference but I blew my budget on a single card from Pink Faun so it’ll have to wait!
I think anyone working within a field where things constantly evolve (product, science, health care …) needs to have an open mind. Or I guess most people should have, but as humans we get lazy and tend to settle on “truths” along the course of our life. That’s why the “new generations” are often needed to spark energy and new ideas into old industries. Hopefully we will learn the benefits of not getting stuck mentally as we age
The author has summarized his experiences in https://www.hiendy.com/hififorum/forum.php?mod=redirect&goto=findpost&ptid=124902&pid=2912736 translated below:
Disclaimer: This is just strictly FYI and does not represent my position. I believe it also differs from Roon Labs recommendations.
- The computer needs to be physically and electrically isolated from the audio equipment. Very important.
- CPU needs to be high speed. Turns off energy saving and hyperthreading in BIOS. Use Process Lasso to assign specific cores (2 and 3 only) to Roon processes, leave core 0 to OS.
- Optimize OS. On Windows Server 2016, install Fidelizer 8.3. Unplug display cable and monitor when playing music.
- User server grade PCI-E network card and run bridge mode direct to Roon endpoint, bypassing any network switch and router. Do not use motherboard LAN port.
- Use PCI-E SATA card with audio grade SATA cable. Install OS to SSD, music files to HDD. On Windows, set pagefile.sys to HDD.
- Use a high power power supply with low ripple noise. Prefer LPS.
- Must use PCI-E for network interface, SATA interface and USB interface. Avoid using motherboard peripherals and disable them in BIOS.
- Use audio grade LAN cable and power cord.
(Disabling of hyperthreading and point 2 are especially interesting to me from a software point of view. These probably reduce context switches. On the other hand, this is opposite to a different school of thought: some people believe in using the best possible CPU but underclock and undervolt it. I think a Roon server partner offers underclocked server products like this and got into a debate with Roon Labs. For point 3, I think it’s better off just running RoonServer without GUI, I think a reasonable amount of members here accept that not exercising the GPU is better.)
I’m a self professed tinkerer as you know.
I can appreciate the technical mechanisms where #6 and #8 can sound subjectively different - but I’m not convinced this will always be technically better.
I wasn’t always like this. But enough discussions with John Swenson (Master Level Tinkerer himself), Rob Watts and Ted Smith has led me to understand some of the mechanisms at play.
The only way to know if a linear PSU is better than an SMPS is to see measurements of the specific ones.
An SMPS CAN actually perform better (lower ripple AND lower AC leakage currents) than a linear PSU - a vice versa.
Or you can go with gut feel and trust the designer. For example, John Swenson has shared enough technical information about these potential mechanisms that I trust his JS-2 and ultraCap LPS-1.2 are very very good linear PSU options. I also trust the SMPS’s of Rob Watts DACs (even Dave) for the same reasons, when he says they measure as well as his lab linear PSU.
For #8, again John S has discussed at length the potential technical mechanisms where different methods of shielding digital and power cables (which a lot of the expensive audiophile cables tend to vary in) can sound different - but different may not be technically better in some cases…
So I’ve been convinced by these experts that we shouldn’t generalise that all linear PSU’s are better than SMPS’s.
And I’ve been convinced by these experts that not all expensive digital and power cables are automatically actually technically better…
Edit: Andrew Mason (Analogue and Power Engineer at THX) has also shared with me , the same beliefs that SMPS’s CAN be designed to perform as well as the best performing linear PSU’s and better than some not-so-good performing linear PSU’s. There is no straight rule.
In shorter form, I guess the gist of the above is to not generalise absolutely everything… but to consider some things on a case by case basis…
Whoever shared these 8 points above - it would be more helpful if they shared specific makes/models that they are talking about. Which linear PSU’s specifically does he think is best, which digital and power cables, which server grade PCIe network card, which CPU specifically etc etc. Then the discussion about why they think these are the best options and what they’ve compared with etc.
Even if it’s all just subjective discussion (this is all fine and often very fun discussing subjective impressions), the above approach makes for a more interesting conversation (for me) than the (potentially) flawed approach of generalising absolutely everything.
I can fully understand why these sort of ‘tweaks’ make a difference for a source/Roon endpoint, that feeds an analogue signal into a preamplifier etc. I think an LPS made a positive difference to my Lumin D1, when I had it.
I have more difficulty understanding that such ‘upgrades’ can make a difference to a server/Roon core. After all, from the way I understand it the ‘bits’ travel from my big-standard ASUS router, to my server/Nucleus, and then back to my router, to eventually be sent via a bog-standard Cat6 cable to my endpoint. I can’t see how any ‘upgrades’ to the server/core can make a difference with so much cable between the point where data flows into my home, to my network player.
BUT maybe I ‘know nothing’? Probably…
It is often easy to forget why things are the way they are though. SMPS’s were designed to be cheaper and easier to produce as well as be more efficient. When you start to complicate them in order to extract the possible performance you erode a lot of the possible advantages of SMPS’s to begin with. You also didn’t make a distinction between SMPS, linear and the group of PSU’s that the LPS 1.2 fits in. It is not an SMPS, it seems to fit better into the hybrid ‘never connected’ family of PSU’s but with benefits, i.e. the use of supercaps. The reality is, the bog standard Chinese LPS’s are generally better than their SMPS counterparts, mostly because the plagiarised circuits they used for the former came from pretty good sources while their SMPS designs were based primarily on least cost. So while in absolute terms an SMPS can be as good as linear, in real terms we know they tend not to be.
RFI, EMI, and other more controversial explanations.
Most of these points are just fussy little tweaks that can’t possibly make any audible difference. Since I am reliant on the summary I can’t say for certain, but it seems like the originial author simply threw in everything he could think of.
I refer to any point that deals with CPU speed past a minimum, the superiority of PCI-E cards over motherboard peripherals, the recommendation of using ‘server grade’ components, the optimization of OS to eliminate what are mostly suspended threads, etc.
As for the recommendation about cables, Cheese ‘n’ Crackers, not again.