Jump to content

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

3 hours ago, HumanMedia said:

Go on @cool_chris - take him up on it, what is one more to compare and this might be even better. You are the best person to compare them. Wont you need two, one at each end to match receiver sensitivity?

I am using a pair of the 1318s here and I like the sound, but I do not find large differences between good quality SMF SFPs, but perhaps these are also system dependent, I use the Sonore Signature Rendu SEoptical as my Renderer/NAA and it isolates the different sections of the internal circuitry pretty well, and also uses very sophisticated all linear power supplies and internal regulation stages. 

SO/ROON/HQPe: DSD 256-Sonore opticalModuleDeluxe-Signature Rendu optical--Bricasti M3 DAC--DIY Purifi Amplifier-Focus Audio FS888-JL E 112 sub-Nordost Tyr USB, DIY EventHorizon AC cables, Iconoclast XLR & speaker cables, Synergistic Orange Fuses, Spacetime system clarifiers.                                                       

                                                                                           SONORE computer audio

Link to post
Share on other sites

Was browsing through my Buffalo BS-GS2048's settings because of the post upthread, just to make sure I had things right, and I had already set everything as suggested (or its equivalent) from prior posts by austinpop etc., but I was curious -- do fellow modified Buffalo owners keep Jumbo Frame enabled, or disable it? This was not, to my knowledge, one of the parameters discussed in past discussions about optimal Buffalo settings, so I assume enabled is the answer. 

Link to post
Share on other sites
  • 2 weeks later...

Thank you @PeterSt for directing me to this thread. This one's a gold mine.


Not sure how to search, but any opinions on Si86xx lineup of digital isolators in this thread?


I was looking into s audio focus 2 dac+amp by Nazar (the username), and found this in his blog/writeup. (English translation, post #8).





Link to post
Share on other sites
On 6/6/2017 at 5:30 AM, JohnSwenson said:

OCXO stands for Oven Controlled Crystal Oscillator.


The hierarchy of crystal Oscillators:


XO  Crystal Oscilator

TCXO Temperature Compensated Crystal Oscilator

OCXO Oven Controlled Crystal Oscillator


What all this has to do with is how the frequency of an oscillator behaves as the temperature of the crystal varies. All crystals will change their frequency due to temperature changes. If you map these changes you get a curve that at low temperatures has a large change per small temperature delta, and at high temperatures a large change per small temperature delta, but at some intermediate temperature the "temperature coefficient" (TEMPCO) is zero.


If the operating temperature inside the equipment is not at this point, changes in temperature will change the frequency. The TCXO and OCXO are ways to make this better than a normal "uncompensated" crystal.


Note that these frequency changes are long term changes, sometimes referred to as drift. They are usually NOT the short term variations we call jitter. Thus optimizing for TEMPCO is not necessarily a good thing for audio.


The TCXO is an electronic circuit added to the oscillator. It turns out that you can change the frequency of an XO by changing the capacitance across it. The TCXO contains some form of temperature sensor that feeds a device that changes the capacitance across the crystal, the change in capacitance cancels out the change in temperature thus significantly lowering the TEMPCO, . But in doing so it increases the jitter, that temperature sensor and cpacitance changing circuit generates noise. This increases the jitter of the crystal. Thus TCXOs are usually not a good thing in digital audio, they are optimizing the wrong thing.


The operation of the OCXO is usually described as maintaining the temperature of the crystal so that the frequency doesn't depend on what is happening outside, the oven keeps the temperature stable. But this is not the whole story. The REAL purpose of the OCXO is to maintain the crystal at that temperature where the TEMPCO is zero, not just any old temperature. If the crystal is at the zero TEMPCO point then the temperature stability doesn't matter that much.


There is a big problem with this. The normal way a chunk of quartz is cut is called the AT cut, this has a zero TEMPCO temperature right around room temperature. This is fortuitous for non compensated crystals, but a real pain in the neck for OCXOs. If the crystal temperature is too high the system has to COOL the crystal, if too low heat it. This is quite complex and very expensive to do well.


The solution is to use a crystal cut called the SC cut. This has a zero TEMPCO point much higher than room temperature, so all you need is a heater, which is really easy to do (current through a resistor). This is why it is called an oven, it is always heating.


A side effect of the SC cut is that IF it is done right you get lower phase noise. But the inexpensive OCXOs put all the money into making the oven and skimp on the crystal quality, so they don't get particularly great phase noise. The ones that DO put money into the crystal can have the lowest phase noise of any oscillator, period.


So why doesn't anybody use an SC cut crystal without the oven? At room temperature the TEMPCO is SO far from the zero point that it would be almost useless.


I hope this gives a little insight into the world of crystal oscillators.


John S.





Thank you very much! It feels analogous to directly heated tubes.


I have a few doubts. Clocks as I understand are mainly generated as square waves (or ideally that's what we want for digital logic, and possibly also dacs if they use these clock to generate the output, like AND of signal with clock). And crystal oscillators as I understand general something sinusoidal or a distorted sinusoid depending upon the piezoelectric non linearities and mechanical restoring force non linearities (if I'm wrong please feel free to correct). How is this converted to squares? Do we just generate a high amplitude one and send through a logic that clips this (I think something like this + a high pass filter is used to create higher clock speeds)? Or is it done by some other way?


Do these oscillators exist mostly as is or are they supported by a transistor amplifier (or oscillator) for better signal strength or as a buffer? If so what are considerations for the amplifier to be better for high fidelity audio applications?


8.20 in this video. 

if such tiny amount of material can affect the vibration pattern of the oscillator, what about the impact of physical vibrations (due to mild ground shake, or speaker sound vibrating the PCB), how do they affect the performance of an oscillator? Anything about impact of rf noise, or PCB design choices, solder/mounting choices?


I hope I'm not trying to ask too much but one last question. What considerations should go into the design of power supply to the crystal oscillator (or in other words what is the electrical load character of the crystal oscillator) - in terms of output impedance, stability, noise, topology matching etc! I remember something called voltage controlled oscillator, are they ever used for audio? (If they are, then pretty sure a stable voltage source is a must 😅).

Link to post
Share on other sites
59 minutes ago, barrows said:

For the most part, high end manufacturers use oscillators which output a square wave (they have the necessary circuitry internally to filter and convert the rough sine wave output from he crystal into a square wave).  Yes, all clocks are sensitive to vibrations, and components which address this in how they implement their clocks will have an advantage.  Also, paying attention to vibration control of the entire component (good racks, aftermarket footers, etc) can also improve performance.

Note also that OCXO are not necessary for the best performance in audio components.  The oven actually does not help much with performance for audio, the oven actually helps much more with the long term accuracy of the clock-think over hours, or days, rather than over the µS that are important for audio.  Consider, we do not listen to a single note over a period of hours.  For audio performance, low close in phase noise (phase noise at low frequencies) is what is important for audio, so the popular belief that OCXO are best for audio is partly in error.  But, some of the best clocks for low close in phase noise (and not all, this is important, some OCXO are worse than the better non-ovenised oscillators) also happen to be OCXO, so they may be used.  So, be aware that just because a clock is an OCXO does not indicate, necessarily, that it is a really good clock.

For example, in order to get a really good clock for their DAC, Ayre contracted with OCXO maker Morion in Russia, to have them make a really good clock, with an SC cut crystal, but without the oven  This way they got a clock with very low close in phase noise, but without the additional expense of the oven circuitry.  

Thank you very much. I would be interested to know if there is openly available information on how this rough sine wave is converted to square wave, atleast a basic overview of it.


Few links and videos I found to be somewhat relevant:






Trying to find literature on how stess compensated (sc) cut oscillator design is made to have low acceleration sensitivity.




Quite a while ago (half a decade or so) there was this device called teraplayer (crazy expensive but apparently very low noise). It used oscillators from golledge though the designer didn't mention what particular part was used. Anyone else familiar with golledge oscillators? Their products seem to be geared towards high precision applications.

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

Thank you very much. I would be interested to know if there is openly available information on how this rough sine wave is converted to square wave, atleast a basic overview of it.

I’m sure this is an interesting discussion, and not intentional on your part, but you’re taking this thread off-topic. See posting guidelines at the top of the page. This thread needs to stay focused on direct listening experience.

Please create a separate thread to dig into clock design to your heart’s content. 

Link to post
Share on other sites
  • 2 weeks later...

@Nenon That looks amazing! I’m going to be following the progress with great interest.


Just for folks to know where to post regarding this... while this project comes to fruition, please discuss over on the main DIY server thread: https://audiophilestyle.com/forums/topic/58164-building-a-diy-music-server/page/


Once builds are completed, by all means post your listening impressions and comparisons with other servers here on this thread.


Link to post
Share on other sites
30 minutes ago, simon68 said:


even better than the Diamond??

then I have to try one for real,.... do you know which model I have to ask and how much is the cost for one?

Yes I sold my Diamond & replaced it with Sablon. Not sure currently prices, just one model. Latest & greatest I suppose. 

email Mark at Sablon for current prices. 

If you buy them you’ll never need to upgrade! 

Link to post
Share on other sites
On 1/9/2021 at 10:35 PM, ray-dude said:

we’re done, no need to do any more comparisons. 


Same experience for me as the ever insightful @ray-dude


I haven't tried the power cords but the Sablon USB and LAN cables have simply stopped me wondering about 'better' or suffering cable envy. My next USB/LAN cables will probably be the Sablon 2022. (I'm hoping Mark skips a year in deference to my wallet 😉).




Synergistic Research Powercell UEF SE > Sonore OpticalModule (LPS-1.2 & DXP-1A5DSC) > EtherRegen (SR4T & DXP-1A5DSC) > PinkFaun modded Buffalo BS-GS2016P (Farad Super3 & LPS-1.2) > (Sablon 2020 LAN) Innuos Zenith SE server > (Sablon 2020 USB) Innuos Phoenix > (Sablon 2020 USB) PS Audio Directstream DAC > PS Audio M1200 monoblocks > Salk Sound Supercharged Songtowers 

Link to post
Share on other sites
28 minutes ago, bit01 said:

I thought I would mention here how seemingly natural and relaxed the music plays using this setup. When 'tuning' it I usually just listen to familiar recordings for a lenghty period between tweaks rather than the A/B comparisons back and forth. When the music/SQ/PRaT etc hits me  'right' -  that is when I try to document where it is all at! At some point I  will compare some recordings on vinyl of the same mastering for the fun of it. Once again, many thanks to the great contributors in this forum.


In the current listening I find that the bass goes really low and has that power reserve! The clarity on voices is excellent. The noise level is surprising low, the quietest sounds and the full note decays being nicely audible. Piano notes are 'solid' and overall instrumental timbre and body of the intruments is better than I have heard previously. Many kudos to the power supplies and all the other components! I found that the AC cord to the SR7T and the DC wiring to the ATX/ and EPS had influence and enabled me to tweak to my preferrences.


Of note -

-WiFi on this board kills the SQ by introducing 'hardness' - so not used.

-The LAN controller is turned off too. I am using the TP-Link UE300 USB/RJ45 adapter for the ethernet.

- SATA too is off (according to notes!).

-One USB MOBO port that I find excellent, and which I am using- the USB 3.2/Gen 2, 2nd one below the RJ45 port.

-The PHD SR7T I have has two 19VDC/10A rails. I am using one to power the HDPLEX 800W DC-ATX using Ghent cables .

-The second rail is being used to power the CPU (EPS 8pin Molex) via the Sean Jacobs 12V regulator+dual current booster setup. This is wired with Mundorf Ag/Au 16 AWG. I have wondered how this would compare with a direct SR7T/12V rail (which would be more efficient/shorter length but less the added regulation)? Irrespective, powering the EPS this way lowered the perceived noise floor appreciably, revealing more detail compared to the HDPlex DC-ATX converter.

-The CPU turbo speeds tried ranged from 3.7-5.0 Ghz. 3.7 to 4.5 seemed preferrable. I am at 4.5GHz currently with the memory at 3200 MHz. Core temps are around 40 deg C or less, when playing music.


The setup:




The server build is here:




@bit01 - Thanks for sharing! I am having a bit of FOMO since I debated going with the 10900k w/ full ATX board vs sticking with my z390 ITX board and picking up an i9 9900k at blackfriday special prices... I chose the latter.


Anyway, I agree with your feedback on the SR7T. It lowered the noise floor immediately. I am addicted to re-listening to my favs and finding joy in genres I could not fully appreciate until now, since the system is so capable of recreating the sense of being there now. What a wonderful power supply :)


If it weren't for the Xeon dual CPU stuff coming up I could happy consider this setup endgame.


Happy listening!

Link to post
Share on other sites

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...