Jump to content

Tasso

Building the ideal(ish) Music Server

Recommended Posts

I don't care about Amir or his ego but I do care about his measurements unless someone else owns AP and can measure different figures...anyway might not be relevant for you guys so will stop posting these things

 

On ?9?/?19?/?2016 at 2:47 PM, rmpfyf said:

Meanwhile some nice 24MHz OCXO's went on Taobao and I missed out. Grrrr. Would have been useful.

 

I bought this 3 months ago, if you add this and good ULN regulator you should get excellent solution on reasonable budget but changing anything without having sorted out PSU is waste of money and energy 

Share this post


Link to post
Share on other sites
14 minutes ago, kukynas said:

I don't care about Amir or his ego but I do care about his measurements unless someone else owns AP and can measure different figures...anyway might not be relevant for you guys so will stop posting these things

 

 

I bought this 3 months ago, if you add this and good ULN regulator you should get excellent solution on reasonable budget but changing anything without having sorted out PSU is waste of money and energy 

Daniel

I find the measurements interesting. Please don't stop posting about that. 

Share this post


Link to post
Share on other sites
2 hours ago, kukynas said:

I don't care about Amir or his ego but I do care about his measurements unless someone else owns AP and can measure different figures...anyway might not be relevant for you guys so will stop posting these things

 

I get it, but Amir's not the only person taking measurements. We're not all here being subjective... some of us do actually take measurements. Nor is data visualised an absolute statement of knowledge. 

 

And at the end of the day - this is just a discussion forum. The fun's in the exchange of ideas.

 

2 hours ago, kukynas said:

I bought this 3 months ago, if you add this and good ULN regulator you should get excellent solution on reasonable budget but changing anything without having sorted out PSU is waste of money and energy 

 

I use different components to the same end. 24MHz is flexible as it's native high speed USB, also newer Core CPU PCH's, etc... and that particular OCXO is very nice.

 

For what it's worth I like your approach. Not just in your last post, but generally. Spending moderate amounts sensibly is something I (and I suspect everyone else) respects. No disrespect to those spending big, were just not all able/willing to do so. 

 

I don't think you'll find anyone on this thread that hasn't made efforts to sort out their PSU.

Edited by rmpfyf

Share this post


Link to post
Share on other sites

than it seems like everyone's measuring but nobody posting :D , it's rare to find measurements of streamer dependency to the DAC distortion so I'm glad that there are some, another here from SW perspective even if not with such high accuracy as from AP... 

 

apologize for stressing so much about PSU but people sometimes tend to overlook dependency of certain investments in audio parts and I'm not trying to point it out to those who knows this very well but in general to anyone who reads this thread ;)  

 

 

Share this post


Link to post
Share on other sites


Spectral performance of streamer to DAC arguments at best are fraught with the notion that distortions aren't purely in analogue domains. They're just measured so, and beyond what's easily visible we get into differences not easily acquired with conventional lab equipment. 

 

I get where you're going with that, though anything subscribing to this thread couldn't miss the importance of PSU. It's almost, almost a thread on people's experiences with CA PSU development. It is quite literally a global epicentre for DIY CA PSU exotica. 

 

By comparison I somewhat limp by on a HD-Plex 100W LPSU. Slowly gathering bits for something better; this thread's been much an inspiration. Your posts included.

Share this post


Link to post
Share on other sites

if we are able to measure selfnoise of the DAC and it's output stage than we should be able to measure the added distortion from the source, better or worse or hidden behind DACs noise floor... 

 

agree, I just hope peoples expectations are clear ;)

 

btw, do you plan to use renesas USB solution? even if USB native frequency is 24Mhz I thought that internal PLL is set to 4mhz with multiplier to align different clocks through microcontroller and that's why every manufacture combine external oscillators accordingly i.e. FPGA using 12Mhz etc. so always times 4, too long I haven't looked at it so not sure now but definitely harder/more expensive to find good 24Mhz xtal, that was the reason I chosen ASmedia chip...

Share this post


Link to post
Share on other sites
1 hour ago, kukynas said:

if we are able to measure selfnoise of the DAC and it's output stage than we should be able to measure the added distortion from the source, better or worse or hidden behind DACs noise floor... 

 

Well, that's the start of a discussion, isn't it - just on determining what a baseline condition actually is. There's an entire theoretical debate over whether a comparing a DAC output with no source input is actually a viable baseline (we could keep that one going a while, but let's not).

 

And IMHO what we're seeing in spectra for everything that's not an RF scope is really just accounting for power issues. Jitter effects (assuming your power's sorted) are going to tough to see without some serious 'scope speed, resolution and intelligent use. And most of us don't have a $25k scope sitting around all the time. At some point, for reasons of cost, time, complexity, ineptitude, etc... data acquisition falls to a glass of wine, music and if we're lucky, good company. And IMHO that's OK.

 

1 hour ago, kukynas said:

btw, do you plan to use renesas USB solution? even if USB native frequency is 24Mhz I thought that internal PLL is set to 4mhz with multiplier to align different clocks through microcontroller and that's why every manufacture combine external oscillators accordingly i.e. FPGA using 12Mhz etc. so always times 4, too long I haven't looked at it so not sure now but definitely harder/more expensive to find good 24Mhz xtal, that was the reason I chosen ASmedia chip...

 

Sure, which is why we see 12MHz etc clocks in existence and others. Occasionally there's some old stock of very nice OCXO that were made for a custom order at some weird frequency (e.g. 24MHz - who needs that, other that some obscure audiophile anoraks or people selling to them). When it's surplus stock it's cheap to have a little fun and see what's what. I was amused to see that stock mentioned appear and disappear pretty quickly; I can't imagine anything other than USB having an immediate use. 

 

I'm frankly still working out the influence of what these xtals do on an async USB connection (happy to have this explained, I'm pretty sure some people would know e.g. WA crew). For all the clocking in the world USB still sends packets which need to be reconstructed into a stream at the receive end, so clock beautifully and there's still inherent delays pre getting to I2S. So buffers etc help, and we'll always find something new to... spend money on :)

Edited by rmpfyf

Share this post


Link to post
Share on other sites

Few days ago, we did a shootout between Shaar Deluxe vs Sotm with the additional clock module. The cost for Shaar Deluxe is approx $300, and Sotm with clock module is approx $1000. All are the approximate from the exchange rates at time. As always - there are 3 SNAers at the gtg, and whether they will provide their observations on the day here...?!? Furthermore - the venue wasn't mine and so the server, audio setup, hence would be inappropriate of me posting any photos without consent. So no photos this time. My observations: Shaar has better bottom ends' definition, timing, and articulation. The mid was very engaging, nice timbre, with weight and body without compromise the clarity nor transparency. The micro-dynamic and detail, I found to be on the parity which was concurred to another member, however, one member did not. I must stress, without a side by side, the critical evaluations at pretty much in same session, and under the one setup, I could live with either.

Share this post


Link to post
Share on other sites


@Tasso, some very interesting experimentation as per usual; if i may:

 

  • What prompted the use of the sound absorbing material?
  • Did adding it to the system make a noticeable difference to sq and if yes, can you articulate it?
  • Re the grounding, same question as above and
  • How did you know how to ground the components? I'd love to do this myself but don't have the first clue... Is there a guide you've used?

 

Has anyone tried the SOtM clock upgrade for their motherboard (not the pcie card) and if so report on what it's done for the sound?

 

In other news I've just bitten the bullet and purchased an i7 6700T Skylake -@rmpfyf I blame you for this! I've gone all Rambo with cpu choice for a couple of reasons:

 

  • 8mb cache instead of 4mb (biggest it seems you can get within the thermal and financial envelopes I want to remain)
  • Same tdp as what I've currently got but should be able to decode 4k and run drc without pain
  • with the grunt offered i can play with underclocking now and see how that might impact sq


Some questions to the group, all answers welcome:

 

  • Does Hyperhtreading hurt/improve SQ in your experience?
  • How does one go about making sure all of the audio tasks get their own exclusive core?
  • Any other observations on CPU related stuff impacts SQ?

 

Thanks guys - this threads really helped me to improve the musical capability of my system.

Edited by realysm42

Share this post


Link to post
Share on other sites

@realysm42

 

Ha! @acg is more Rambo than both of us combined with his CPU choice :) (it's very nice).

 

2 hours ago, realysm42 said:

 

  • Does Hyperhtreading hurt/improve SQ in your experience?
  • How does one go about making sure all of the audio tasks get their own exclusive core?
  • Any other observations on CPU related stuff impacts SQ?

 

Hyperthreading (HT)

Switch it off and recompile your kernel to not support it. HT requires your OS to include a scheduler for HT, so there's a bit of your OS that has to stop and think of which instruction goes where next. This is not the same bit of code responsible for working on multiple cores, which you will need. You're introducing latency and limiting resource availability, and seriously you don't need more than 4 cores for audio (and often less). I'm paraphrasing here so... just get rid of hyperthreading.

 

One exclusive core per audio stuff

Are you Linux or Windows? I've no experience on the latter.

 

On the former you need to isolate the CPU cores everything's allowed to run on - you do this at boot as part of your GRUB configuration ('isolcpus' kernel parameter). This is the start of it as it'll isolate user-space threads from running on said CPUs, kernel tasks can still wander a bit. There are ways to patch this. 

 

Then you need to assign whatever you want to run on a dedicated CPU to that CPU. You can do this with affinity calls or numactl (e.g. sudo taskset -c CPUNumber CommandOfInterest etc).

 

Verify in top or htop. You can repeat this for different tasks.

 

You will find, on a 4-core CPU (or greater) that which cores you allocate processes to does affect net latencies: think of where each are located on the CPU die, and work from there. Or just have a play. (Or copy someone else's settings, PM me if you get lost). 

 

Other observations

Many. 

 

  • Don't let the CPU shut down the cache - so limit C states to C3. You can get more advanced than this for SQ but it's already a good start. This is pretty important for a CPU with a nice cache (kind of defeats the point otherwise). You should be able to do this in your BIOS, otherwise there are ways to ensure it in your OS. 
  • Run your CPU down on speed. You should be able to hit 800MHz without hard modding. You'll need to run the userspace governor for this. Those with K-spec CPUs are running down to 400MHz no problems. I ran Kodi at 1080P with no dropped frames and DRC full speed on a dedicated core at 800MHz with an i3, so really don't worry about increasing speed. 
  • Experiment with your GPU etc speeds. Consider that the same VRM circuit powers your CPU and GPU (if integrated) and whilst you can run them at separate speeds, you're requesting completely different things in terms of power, according power circuit noise, etc. I found a nice (and surprisingly audible) advantage running them at multiples of each other. (yes, yes, smug audiophile types will wonder why run the graphics at all, and they're right, but WAF wins here). 
  • Found better results on RAM at full speed at low voltage. Different power regulation, it's a lesser effect. Sounded better with a RAM power filter in the second slot, that's just my mileage.

 

There's a lot else you can control in the OS (clock buffer sizes, scheduler, etc). If lost, download Snakeoil. If you want to build your own it's not impossible.

 

Motherboard clock upgrade was a nice touch for me. There are solutions out there at USD$400 which are ridiculously priced; have a few others in the works that are a little better and considerably lower priced. Bear in mind some motherboards run oscillators you can reasonably access (e.g. through-hole mount), some are damn near impossible (surface mount), some run independent clocks for different systems (PCH, secondary Ethernet, secondary USB)... on your motherboard the PCH clock can make a difference, the rest not really so (unless you're running USB out of a secondary controller - you'd sooner spend the money on a dedicated card anyway - see @kukynas suggestion for a cheaper way out if you're keen to tinker), and WA people's SHAAR experience otherwise (that's kinda hard to beat on price). I think for Skylake you'll be on a 24MHz oscillator, better for USB, harder to find otherwise in a nice OCXO. What's your PCH chipset (H87/H97/Z97/Z97 or something later?)

 

Hope it helps.

Share this post


Link to post
Share on other sites


On 9/22/2016 at 11:27 PM, rmpfyf said:
1 hour ago, rmpfyf said:

Ha! @acg is more Rambo than both of us combined with his CPU choice :) (it's very nice).

 

 

 

Well, if I was Rambo before I am not sure what I am going to be soon.  I am about to try Broadwell-EP, 20 cores, hyperthreaded to 40 cores, 50MB cache with X99 motherboard.  I have the parts here but have to find time to put it together and test it.

 

Win10 run from 32GB single ranked ram (no hard drive), passive cooling, headless (no graphics, mouse, keyboard) a true memory player controlled over the wired network from another computer via RDC blah blah blah.

 

Edited by acg

Share this post


Link to post
Share on other sites

@acg Bloody hell! Keep us posted. I'd wanted to try Broadwell-DE; you're in a bigger universe altogether :P

Share this post


Link to post
Share on other sites
Guest

@acg

This one would do :P (little 2 core with a modded sound card, DC powered)

You definitely went overboard with yours, but, this is you

Excellent stuff

 

PS

@rmpfyf thanks for the sata cap card,,,,,,,,,,,doesn't seem to do much, but look cool :)

003.JPG

Matts desk top.JPG

Share this post


Link to post
Share on other sites


Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...