Jump to content
IGNORED

A novel way to massively improve the SQ of computer audio streaming


Message added by The Computer Audiophile

Important and useful information about this thread

Posting guidelines

History and index of useful posts

Most important: please realize this thread is about bleeding edge experimentation and discovery. No one has The Answer™. If you are not into tweaking, just know that you can have a musically satisfying system without doing any of the nutty things we do here.

Recommended Posts

  • 1 month later...
16 hours ago, Bricki said:

Has anyone tried reducing USB transfer speeds with similar effect?

 

For example if we reduce our USB transfer speeds for music storage on our server, perhaps there could be some gains there.

 

I assume this is possible somehow in Linux but I wouldn't know where to begin 🤣.

 

USB 3.0 controllers do seem to have better signal integrity and sound better but if we can reduce their speed back to USB 2.0 there might be some further SQ gains.

 

Just a thought

 

Do remember that the Intel NUC's clocks between USB, network and CPU subsystems are not "clock locked" (in harmony) thereby simply setting the network to 100mbit/s could change the EMI/RFI patterns from the networking subsystem, hence the differences spotted.

 

So yes, worth a try if possible though in many cases (DACs at least) they'll be negiotating USB 2.0 speeds anyway!

Link to comment

The problem with the multi-core stuff is when used heavily, will they induce their own noise signature (= thereby impact audio performance) ?  I suspect some trial/error (e.g. underclocking the CPU in general) to find this out will be required.

 

I'm personally keeping an eye out on the AMD r1000 embedded series, for this very reason.

Link to comment
  • 1 month later...
13 hours ago, Chopin75 said:

I use Ryzen 2700x, no issue. It has no graphic card so you need a separate video card for setup purposes. If you run headless (as most do here) take out the video card after setup.   Runs smoothly, recommended by the PinkFaun guys who use them to build their super streamers .They like the multicore design, high Cache memory. However they also think Ryzen works better with their I2S bridge so that may be the main reason and I plan to use an I2S bridge at some point, too. I have not used equivalent intel CPU so cannot compare. I think using Ryzen for Euphony is working fine, does upsampling/PCM--->DSD,  M-ch etc.... more than enough juice for them.  I am currently using it to run Euphony single box. I think you are supposed to use the motherboard for Ryzen like x370 etc.. there is a new Ryzen chip 3000 now!  Can I say this is state of the art streamer this way, not sure, as I have never heard multibox designs. 

 

I'm speculating here, but given the Ryzen 3000 series has just moved to a 7nm process to achieve production I do geniunely wonder if these new CPUs are inherently quieter than their predecessors?  Only one way to find out I guess !

 

I've seen a pattern so far that as the generations of chipsets improve their fab processes, we're also seeing improvements as audio streamer/servers (when fed good power etc).  i.e. SoCs are focussing on noise reduction to help with their own performance/stability goals.

Link to comment
  • 2 weeks later...

These posts genuinely beg for a "reference LPSU that can be delivered within a reasonable timeframe" thread.  Lets face it, most if not all gains come simply from high quality power.

 

I have an Farad Super3 on the way to me, so can report on that though haven't a foggiest how it compares to the "chosen ones" SR4 / SR7s.

Link to comment
1 hour ago, bobfa said:

PLEASE and Thank You.

 

I have been comparing Roon +StylusSQ and Stylus against each other on my Xeon.  I have to start some processor speed limit testing next.  I still feel that Stylus sounds better.  There is more dynamics and there is more detail.  I am going to get some other folks over to listen!

 

My son likes Stylus more for the simplicity of using their web interface and the speed of the UI.  I still like Roon for their look and feel.  

 

I will also bring the NUC back up and listen to it I have to validate the differences again.

 

NOTE: using the HDPLEX 400 and an FMC to the endpoint.  I am not trying to change network speed due to FMC.  The Xeon now has an Element-X  USB card.

 

Do you mean Element M card?  I got one of these to try in my Intel NUC, but it requires 12v and 3.3v from the PCIE bus.  The adapter I tried provided 12v.  Will try the JCAT in time (no power from PCIE bus required).  

Link to comment
2 minutes ago, rickca said:

Matrix Element M and Element X are streamer/DACs.  I think @bobfa means the Element H, which is the latest Matrix USB card.

 

Apologies I have the Element H myself (ready to go back to the seller) due to it requiring PCIE bus voltage (as noted above).  In bobfa's case may not be a big deal to be honest given system is already powered by LPSUs, though I'd prefer to isolate power from system with the JCAT card as mentioned.

 

I confirmed this by setting external power and connecting a 12v DC feed, but disconnecting the PCIE 12v feed on the adapter for the NUC; would not power up at all.  Shame.

Link to comment
8 minutes ago, rickca said:

So how did it sound, and what did you compare it to?  It's around $250, so about half the price of the JCAT USB Femto card.

 

TBH I'm not going to even try.  PC "server" is not fully optimised, will require moving into a different room (and dirtying up the rest of the power feed on Hifi).  From past experience of temporarily powering up a JCAT on ATX internal power, it was ... ugly.  External was night and day.

 

If anyone can confirm the Element H can be powered externally only i will happily try it... but I don't think you can.

Link to comment
  • 4 weeks later...
  • 3 months later...
  • 1 month later...
2 hours ago, seeteeyou said:

Unless we absolutely must run x86 version of Windows while having access to PCI-Express slot, IMHO it's a much better deal to go for stuff like Allo USBridge Sig that's "proven" to sound pretty awesome without having to pay so much for an expensive PSU. For instance, @bobfa should have ordered two of them from Allo and then he's actually saying goodbye to that $3,745 Product of the Decade on this site.

 

 

To be fair, bobfa didnt try the "turbo" powered version of the Sonore product.  As we know power can make all the difference.  Though not to say the USBridge sig isn't excellent value - I agree actually that for those who are not willing to goto all this "fiddling" and spending good money on a uber PSU that is by far the better way to go.

 

I do have an OpticalRendu + Farad 7v (both demo items with a few months runtime on them) on their way as a comparison to my own tweaked NUC with Industrial RAM and the same Farad at 19v etc.  The Sig Optical SE is an OpticalRendu with their well matched HQ PSU (of differing grades) along with a few other tweaks such as chassis etc.  I did try to obtain a Sig Opt SE (with uber PSU) in the UK for a home demo but quite frankly it wasnt going to happen.

 

This came out of trying Emile's recommended Startech SFPs with their FMCs (and then inside the Cisco 2960G's SFP).  Really worthwhile improvement for what is little outlay and I had a Pink Faun Ethernet Isolator prior as well.   

 

One downside of the recommended Startech FMCs is they only operate in auto-neg mode, and you cannot set 100mbit/s on the ethernet port as the FMC will not handshake unless its auto-neg.  Have spoken to Startech support about this and there's not much they'll do here.  For those who aren't aware the NUC can sound a bit too "bright" and sibilant when left in 1gigabit ethernet port mode.

 

The good news however is the Cisco 2960G sounds pretty good with SFPs and its stock power supply alone.  I decided to try 2 of these, with the network port set to 100mbit in one case.  This proved to strike the best balance so far.  Just waiting on another good 12v HQ PSU before I mod the 2960G's SMPS (very easy as noted on this thread).

 

Being a HQPlayer user at DSD512 rates (T+A's "sweet spot") the difference between ethernet copper vs fibre is night and day.  Poor Jussi, it seems a lot of his good work is quite literally lost in noisy PCs between HQPlayer and the DACs themselves !!

Link to comment
10 hours ago, Ricardo007 said:

Could the Startech SFP recommended by Emile be used between an optical module and an ether regen Aside fiber input ?  To isolate from router ? What about this auto negociation thing ? 

 

I don't know regarding StarTech SFPs and OpticalModules.  In theory they should.  I'll have an OpticalModule here in a few weeks to test, so will report back.

 

Autonegiotate is the way an ethernet port agrees speeds and duplex between each other.  On the StarTech recommended FMC's you cannot bypass this, it just causes the port to stop responding. 

 

I have however found since placing one StarTech FMC with a good 12v on the "sender" end, yet using the Cisco 2960G on the other end improved performance (again lower noise floor, more palpable etc).  Now when I switch back to 1gig on the NUC ethernet port, so far it's not drifted into overly brightness. 

 

Is setting the NUC's ethernet port to 100mbit masking not ideal network feeds?  Will do a bit more testing to conclude.

Link to comment
2 hours ago, guiltyboxswapper said:

Is setting the NUC's ethernet port to 100mbit masking not ideal network feeds?  Will do a bit more testing to conclude.

 

Yes is the answer.  1gig infact now steps things up another step.  Bare in mind i'm a HQPlayer + NAA type, so there's approx 50mbit/s traversing the network vs the typical 1.5mbit/s without it.

Link to comment
  • 2 weeks later...
1 hour ago, Gavin1977 said:

Interesting - anyone else experienced this compared to NUC? 

 

I would be interested on thoughts about locally bridging a NUC (operating as HQPlayer Server) to the OpticalRendu.

I will try soon, only issue is I know my DAC (T+A DAC 8 DSD with a copper mod) certainly has a sweet spot at DSD 512 where I know it best.  It is in a proper fanless case though so I suspect it wont last all that long before throttling kicks in.

 

Bridge mode will require an additional USB ethernet adapter though.

Link to comment
On 1/20/2020 at 10:51 AM, guiltyboxswapper said:

Next steps will be to try and obtain a OpticalModule OEM DIY card to feed it directly from the host server, cutting out any Ethernet to Fibre conversion stages + associated clocks

 

Turns out this is an opticalModule that's hosted within the server/PC chassis, still has to connect back to ethernet port somehow.  

 

Anyone seen an PCI-express SFP card that accepts external power?

Link to comment
4 hours ago, davide256 said:

There are many happy people running AL and/or Euphony on NUC endpoints that used to swear by Sonore rendus. I'm just one of them. It's been about a year since this

thread made me a convert.

 

The person who recommended trying a NUC was infact an ex-ultraRendu owner.  These designs were previous generation and so yes the NUC is an improvement over those, compared to what's now available in the last 12m.

Link to comment
  • 3 weeks later...

Tried NUC i7 with Farad as the HQP Server for a day or so, and whilst it runs at DSD 512 rates (albeit getting hot and certainly closer to the PSU limits), there is a small loss in dynamic range in particular bass.   

 

My HQP Server is far from ideal either (SMPS power, GPU though CUDA not activated, m.2 SSD) and yet the difference is there.  

Link to comment
5 hours ago, Chopin75 said:

The Farad is only art 3 A ? You may need a higher current than 3A to run upsampled DSD512. I use 10A of the HDplex 200W LPS to run HQplayer upsampled to DSD512 but that is only for fun, not really for regular listening. I don't recall it having much improvement of SQ from DSD256.  

 

Yes 3A, however the watts drawn "from the power socket" was ~47w for DSD 512, 35w~ for DSD 256 so well within the limits of the Farad 3 at 19v.  NUC i7 in passive case was stable for hours on end in this configuration.

 

But yes i'm finding the same in this experiment at least - good power on a simple NUC doesnt reap much compared to SMPS driven desktop system, when the system is doing the heavy lifting only (i.e. streamer running NAA on other end).

Link to comment
  • 4 weeks later...

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...