Jump to content


  • Content Count

  • Joined

  • Last visited

About latenz

  • Rank

Recent Profile Visitors

1955 profile views


  1. HQ Player
    HQ Player
    1 hour ago, MemoryPlayer said:

    Why ASDM5/ASDM5EC and not ASDM7/ASDM7EC Miska?

    Do you prefer 5 over 7? I like more the sound with ASDM7/ASDM7EC..., but EC is so much for my Apple 2017 MacBook Pro!


    You can use 7 too, but 5 gives somewhat lower ultrasonic noise level on ESS, especially at rates below DSD512. But it also depends on a particular DAC (what kind of analog stage it has) and overall system. Otherwise 7 is technically better and if it works well for you, just keep using it.


    ASDM5(EC) is the most generic approach for ESS DAC at <= DSD256.


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

    I thought I'd provide an update.  I apologize for another War and Peace length post but it's been awhile since I last posted and since I don't know when I'll post again, I figured I'd leave it all on the table.  Like always, the following represents my opinions based on personal observations and so YMMV.  Full disclosure, I have no financial motivations.


    Some may recall that a year and a half ago, I transitioned to a pair of Wilson Alexia Series 2 speakers.  The good news is that these are as good as I hoped they would be.  The bad news is that they are also as bad as I feared they could be.


    These speakers aren't that difficult to drive but are so much more revealing of the qualities of the driving amplifier than what I previously had that with the wrong amplification, they can sound lifeless and dull.  I ran the gamut of amplifiers, basically whatever I could get my hands on from friends, dealers, and directly from manufacturers and just when I thought the Alexias had shown me all they could, an amplifier would come along and I'd realize the Alexias had more to give.  As I went through this exhaustive exercise, I found that the best amplifiers shared 2 very important traits:  control and immediacy. 


    Without control, complex instrument lines blur into one another.  Transients are smeared.  Resolution and transparency are compromised.  As for immediacy, here is a view of Davies Symphony Hall in San Francisco that I hope to never see again while at a performance there.  It is from high up in the 2nd tier and I have fallen asleep from seats like this.  I might as well have stayed home:




    The following is a stock photo but is an actual view from my preferred seats -- the very front row. 




    These are actually not elite seats as it turns out this vantage point is too close and too low for some tastes and so to my good fortune, I find these seats frequently available and affordable.  The point is they provide the immediacy that I crave and even on days when I arrive to these seats after an exhausting day, my engagement is always there.  I can hear the performers' subtlest expressions.  I can hear them take their breaths.  I can more easily discern the timbral variations between the 1st and 2nd violinist and better glean the space between them.  The soundstage is wider and deeper.  There is simply better localization of everything and while it's easier to hear mistakes from this close up, when perfection happens, you appreciate it better.  For someone like me who craves presence, being this close up is enthralling.


    Music servers are the same way.  To my ears, they provide the exact same qualities that a good amplifier provides and the very best music servers I have heard amount to the equivalent of a serious amplifier upgrade.  Given the large price tag of some of the very best amplifiers, this is saying a lot.  




    It has been said that when listening to a really good amplifier, what you are really listening to is a really good power supply.  This is just as true for music servers.  It starts at the power supply and to date, I have yet to personally use a power supply in a DIY build that performs as well as a DR (double regulated) SR7 from Paul Hynes.  My latest SR7 arrived earlier this year in a large Streacom FC10 chassis and this thing must weigh at least 50 lbs.  It includes the Teflon board and Vishay Foil resistor upgrades for all 3 rails and if you zoom in, you can see what the off-white colored Teflon PCBs look like. 





    Like Teflon coated frying pans, they are very smooth and slippery to the touch.  They also have a low dielectric constant leading to a smoother presentation with less smearing, especially in the treble.  Compared against my standard DR-SR7, the presentation also sounds a bit faster.  While not as impactful as going from SR to DR, it squeezes out the last bit of performance from the SR7 that Paul is capable of and I have found it to be worthwhile. 


    With this SR7 as a foundation, I have found my servers to be competitive with anything I have put them up against.  Here are photos of Streacom's FC9 chassis against Aurender's latest W20SE.  





    The Aurender is a very attractive unit and is solidly built as it should be for $20k.  While it's presentation is very refined and is an obvious improvement over the outgoing W20, in comparison to my server, the transients sound too soft and smeared.  There is an obvious downgrade in resolution and the immediacy is not the same as it lacks that startling quality that my server is capable of when called for.  It's as if I'm sitting high up in the 2nd tier at the Davies Symphony Hall.  In isolation and if money is not a concern, I doubt its buyer will complain but in comparison, the W20SE is a disappointment given its mediocre performance and high asking price.  While there are probably a multitude of factors, I believe the transient softness and lack of immediacy I am hearing is largely due to its hybrid battery power supply.  


    For those who have been patiently waiting for their bespoke SR7s, the most encouraging thing I can say is "hang in there."  This is my 4th SR7 and so I know your pain but the wait has been worth it.  It will probably outlast every other component you own because you'll likely never outgrow its capabilities.


    For those not in the hunt for their own custom spec'd DR SR7 or are biding their time until their's arrives, there is hope in the form of the HDPlex 800W DC-ATX Converter ($248):



    This has been discussed by others already and so I won't go into it too much.  It's been out for awhile and when I first noticed it, what caught my attention is the low voltage variance of only 0.1-0.6% for the ATX rails that matter and so I felt it was worth the gamble to try it.  In comparison, the 400W DC-ATX converter I was previously using had reported voltage variances of about 1% and for the all important 12V rail which has a reported variance of only 0.1%, this represents a 10-fold improvement!  


    If a power supply can't maintain voltage (i.e. if voltage sags) which can happen when a PSU is pressed, then based on Ohm's Law, the power supply's capacity to deliver current is compromised.  This ability to maintain voltage is one of the things that makes a DR SR7 sound more controlled and more dynamic and this is exactly what I hear with this converter.  While a 19V DR SR7 rail powering this converter sounds otherworldly, I am finding even the 400W HDPlex ATX LPSU powering this converter via its 19V rail to sound very good and much better sounding than the HDPlex directly powering the motherboard via its ATX connector.  So good that you can use this converter to power both the ATX portion of the motherboard AND the CPU and get surprisingly good results without having to devote a 2nd precious rail to the CPU.  All you would need from your 19V rail is adequate headroom and so for those looking for simplicity, high performance for the dollar, and more ready availability, a single 19V/8-10A rail from Paul Hynes Design LTD, Farad, or Keces could work out well.  In a pinch, the HDPlex's 19V/5A rail is proving to be very satisfying.  It should come as no surprise given my experience with my SR7s that I would have a bias towards Paul's supplies but what is further appealing to me with the fairly affordable, off-the-shelf single rail SR7s from Paul Hynes Design LTD is the Streacom FC-9 chassis that's used as this chassis is presently my preferred chassis to build with.  For those who are space constrained and need or prefer to stack the server on top of the PSU or are looking for the aesthetic of a matching server and PSU chassis (like the innuOS Statement), then this SR7 is all the more appealing.


    For those looking to power other peripherals such as Chord's M-Scaler, a JCAT card, tX-USBultra, or network switch, consider one of these:




    Credit to @seeteeyou for bringing this to my attention and I now have one on order.  Combine one of these DXP-1A5S ($99) single stage regulation power supplies with a 19V SR4 and you now have a Hynes DR rail capable of up to 1.5A output between 3.3-15V.


    For those on a tight budget, then consider the DXP-1A5DSC ($149) dual stage regulation model as you can combine it with an inexpensive PowerAdd battery and using that battery's 20V/4.5A output, you now get linear DR output.  I have one of these on order also.




    This component is more important than some people may realize.  With certain servers like the SGM Extreme, the chassis can be the single most expensive part.  Yes, a good chassis is important for aesthetics and can make an audible difference with respect to minimizing unwanted resonances but there is another important practical consideration.  No one wants spinning fans inside their server based on the acoustic and electrical noise that fans create and so a passively cooled server is what most audiophiles strive for.  At the present time, for DIY, this means fanless cases from Streacom, HDPlex, or Akasa.  Soon, JCAT will be introducing their own fanless chassis that will be able to accommodate a full-sized ATX motherboard and so this is very welcome news.  The ability of these chassis to effectively dissipate heat is crucial as it limits what CPU you can use. 


    For example, Streacom claims its FC-9 Alpha's passive cooling design can accommodate a CPU with up to a 95w TDP but recommends a CPU with only a 65w TDP.  I have found this to be true.  With a recent build utilizing an AMD 12-core 3900X (TDP of 105w), I am unable to run this CPU without de-throttling it significantly (down to 2.4GHz) because this chassis can't handle this CPU's heat output.  While CPU temps of 80 degrees C are considered tolerable in gaming rigs and 3D modeling workstations, I find CPU temps this high to result in fatiguing harshness and so with this build, I shot for temps of <50 degrees C.  


    The SGM Extreme, which utilizes a chassis and a fanless cooling system specifically designed for the dual Xeons housed within has average CPU temps of only 40 degrees C and chassis temps of only 35 degrees C during even prolonged use according to Emile Bok.  This is quite remarkable as this should have not only SQ benefits but also longevity benefits.


    Vibration Control


    A few months ago, I was introduced to the CenterStage2 footers (starting at $320 each) from Critical Mass Systems based in my hometown of Chicago. 




    My first thought was "Oh no, not another footer."  I already own a variety of footers and had settled on a certain combination that was working well for me.  With my heaviest gear such as amplifiers, I liked the very even-handed noise reduction I got from the HRS Vortex footers.  Because of their small footprint, I had been using Stillpoints Ultra SS footers under my power supplies.  I used to be a much bigger fan of Stillpoints but I have had to be careful with them under certain gear because they have a tendency to "oversharpen" resulting in a presentation that doesn't sound natural to me but under my power supplies, I generally liked what they did.  Under everything else, including my DAC, server, and even my Wilson Alexia 2s, I had been using high grade G5 Titanium footers ($35-$50 each) that were custom made for me by an audiophile friend from Taiwan who owns a titanium factory.  Titanium resonates at a frequency well outside the range of audibility and so I have found these footers to be quite effective without adding any unnatural coloration.  They are especially effective under my Wilson Alexia 2s.  


    When I first tried the CS2 footers under my DAC and music server, I was warned they would sound horrible during the first 2-3 days, so bad that I would want to take them out.  I was intrigued by this statement because if a product like a footer can result in that large of a delta of bad sound that I would want to take them out, then this tells me just how negatively impactful vibration can be.  More importantly, once this footer settled, could it then result in a similarly large delta in the positive direction?  This, in fact, has been the case.  So much so that I have been slowly replacing my other footers with these.  As you might expect, good vibration control results in improved clarity and focus but these footers do something else I have never before experienced with vibration control devices.  I hesitate to call it a coloration because the tonal color is neither warmer nor cooler yet the signature is "creamier" and more immersive without smearing detail.  It's hard to explain but these are the most musical footers I have yet heard and improve resolution quite dramatically.  What is especially nice about the CS2s is that if there is desire to stack one component on top of another, while this is generally not something I like to do, these footers allow you to do so without apparent detriment, at least none that I can hear.


    If there is a downside to the CS2 footers, they are quite expensive and for those looking for a less expensive solution, there are the Daiza platforms by Taiko Audio. 




    I saw these first hand during my visit to the Taiko Audio factory in Hengelo in The Netherlands and they are attractive with an earthy look to them but also relatively inexpensive (starting at 400 Euros for a platform).  They also come with a money-back guarantee.  They are made of German Panzerholz wood and are designed to couple with your component resulting in a significant dampening of vibrations.  I have not had a chance to evaluate these yet in my system but they were used exclusively in Taiko Audio's listening room and I was quite impressed by the overall presentation and so I feel these deserve further exploration.  For those who are curious, the footers built into in the SGM Extreme's chassis are a hybrid design that include this Panzerholz wood and so I know that Emile fully believes in the vibration dampening qualities of this wood.




    If you recall my last long post, it was quite surprising just how impactful I found the CPU to be to the overall presentation and all things considered, probably the 2nd largest difference maker behind only the quality of the power supply.  How could a stock Celeron NUC sound better than devices like an sMS-200 or ultraRendu that use low power/low noise ARM CPUs and are designed specifically for high quality audio playback with low phase noise clocks and linear regulation?  If you recall, the initial thought was this improvement in SQ was due to a better OS (AudioLinux running resident in RAM) and yet SQ improved further as I transitioned from a Celeron NUC to a more powerful i7 NUC.  While no doubt the OS plays a large role, in my mind, the CPU plays the bigger role.  From the i7 NUC to an 8700T (35w TDP) to an 8700K (95w TDP), dynamics improved but if I wasn't careful, so did harshness.  Fortunately, with effective CPU heat dissipation and higher quality PSUs (HDPlex --> SR7), this harshness went away and what was left was a fuller bodied presentation with superior dynamics and with transients that were more fully and powerfully expressed.  I won't name names but anytime I hear a server today that runs Roon and is powered by weak Celeron or Pentium, the presentation I hear can sound clean but also sounds thin and sterile.  For me, there's just no going back and I suspect in time, once these companies do their own testing, they'll start to incorporate higher power CPUs.  I find that even streamers or endpoints that run RoonBridge benefit from high power CPUs and so it makes sense that someone like Pink Faun would market a server/streamer combo that use equivalent high-power CPUs.




    The questions for me now are Intel vs AMD (which is better?) and cores vs CPU frequency (which is more important?).  If you ask the companies building high-power servers, Jord at Pink Faun prefers AMD and Emile at Taiko Audio prefers Intel.  Both have told me they believe cores to be more important than CPU frequency in a non-upsampling Roon server although in my testing of Intel's 8700T, turning Turbo off which effectively caps CPU frequency to 2.4GHz didn't sound as dynamic as leaving Turbo turned on so probably, in a perfect world, if heat isn't an issue and the power supply is solid, it would be ideal to have both a large number of cores and high CPU frequency.  I know with HQP upsampling to DSD, higher CPU frequency begins to take on greater importance.


    As I dipped my feet into AMD waters, I've been able to test 2 different Ryzen CPUs including an older 2600 (65w TDP) with 6-cores/12-threads and a max CPU frequency of 3.9GHz and a newer 3900X (105w TDP) with 12-cores/24-threads and a max CPU frequency of 4.6GHz.  As stated above, because of heat issues, I was forced to throttle the 3900X's CPU frequency down to 2.4GHz to keep CPU temps down to acceptable limits but even with this de-throttling, the more core-rich 3900X sounded more dynamic, better controlled, and more expansive.  This would support that cores are more important than CPU frequency if forced to choose.


    As for AMD vs Intel, this one is more difficult for me.  With the AMD 3900X (12-cores) vs the Intel 8700K (6-cores) and outputting via USB, the AMD 3900X sounds once again more dynamic, better controlled and more expansive but there is also a mechanical character to the sound that is less natural and more clinical sounding to my ears.  The Intel 8700K, in contrast, sounds more liquid, tonally richer, and more intimate and so there is no definitive winner based on what I'm hearing.  Is this due to the CPUs themselves or are the motherboards/chipset/RAM/etc. playing just as big of a role and is it possible to get the best of all worlds? 




    The chassis plays a big role in the decision for which motherboard to use as most fanless cases cannot accommodate anything larger than a mini-ITX sized board although the new HDPlex H3 V3 ($258) and the Streacom FC9 Alpha ($300) are both capable of uATX boards.  As far as I am aware, the only fanless cases that can accommodate a full sized ATX motherboard are the HDPlex H5 V2 ($298) and the Streacom FC10 Alpha (approx $400).  The problem with both of these chassis is that you are forced to use PCIe extender cables which is less than ideal.  Hopefully, the upcoming JCAT chassis will not have this limitation.


    As previously posted, the gaming motherboards I tested seem to have an edge over non-gaming motherboards presumably due to better VRMs and use of multi-layer PCBs with more copper in the trace paths resulting in better isolation, grounding, and power delivery.  Because VRMs require real-estate, the ideal sized board would be a full sized ATX motherboard. 


    If limited to mini-ITX, for Intel, the best board I have found is the Asrock Z390 Phantom Gaming-ITX/AC.




    For AMD, thanks to @Nenon, it is the ASUS ROG Strix-X470-I. 



    Between the two, the Asrock Intel board has the superior VRM.  I have already discussed the high quality VRM used in the Asrock board in a previous post.  Unfortunately, the large CPU cooler mounts used by AMD CPUs takes up precious real estate on the already cramped mini-ITX motherboard and so it is the VRM that is compromised.  This may account for perceived sound quality differences between the two.  This also suggests that for AMD CPUs, it would be better to use a full sized ATX motherboard which would likely have a higher quality VRM. 


    The block diagram supplied by ASUS suggests, however, that the ASUS AMD board probably has the better architecture where the PCIe slot, M.2 slots, and 4 of the USB slots all bypass the bandwidth limited chipset (PCH).  With Asrock's Intel board, the M.2 slots and USB slots all have to go through the PCH.  To my ears, with the AMD board, the 4 blue USB ports on the back (which connect directly to the CPU) do indeed sound slightly better than the 2 red ones (which connect through the chipset).  Here is the block diagram for the ASUS ROG Strix B450-I which I am told is equivalent to the ASUS ROG Strix X470-I:




    Good quality uATX-sized boards are not so easy to find.  I had purchased the Asrock X470D4U and had high hopes for this board based on the block diagram provided by Asrock.



    Fortunately, before I could open the box, I read @Nenon's unfavorable review of this board which was quite timely as I was able to return it without having to go through the aggravation.  


    The best uATX board may well be the ASUS ROG Maximus XI Gene which is designed for Intel CPUs and used by @StreamFidelity in his build below although I have not yet personally tried this board.  Unfortunately, I know of no good uATX board for AMD.




    I have to compliment @StreamFidelity as this is a masterclass build -- very clean with excellent attention to detail.  I don't think I could do it any better and I especially like the CPU cooling enhancements.




    I won't really go into this except to give kudos to both @Nenon and @Marcin_gps for bringing to our consciousness the Apacer brand.  I bought Apacer ECC DDR3 memory and an SLC compact flash card back in 2017 based on Marcin's recommendations for a build that only momentarily saw the light of day.  Having purchased Apacer's ECC DDR4 memory for a recent AMD build based on Nenon's findings, there is definitely a nice uptick in SQ.






    JCAT deserves its own subheading here as 2 of their products single-handedly salvaged a somewhat unnatural and mechanical sounding AMD build, even with the Apacer RAM. 




    Those who read my last lengthy post know that I found the Femto Net card to sound incredible with my Intel build and it is just as incredible with this AMD build.  This ASUS board's stock Ethernet port pales in comparison.  This time around, I compared it against a 10Gtek SFP+ PCIe card with the Startech SFP transceiver that Emile at Taiko Audio likes.




    This fiber network card immediately brought forth a lower noise floor, a greater sense of resolution with very well defined bass and extended treble but the presentation was bright and thin and didn't sound natural and so I much prefer the JCAT Femto Net card.  It turns out that I misinterpreted Emile as he not only prefers the Startech SFP transceiver but also the Startech SFP PCIe card and so this will require exploration as the greater perceived resolution brought about by the fiber card is desirable.  If Marcin is able to somehow marry the benefits of SFP with his current Femto Net card, this could be worthwhile. 


    This was my first experience with JCAT's Femto USB card.  The card that shipped to me had the older firmware and I was quite surprised by how much difference in SQ a firmware upgrade could make.  While the 4 blue USB 3.1 ports on the back of the ASUS motherboard sound better than the 2 red ones, it isn't saying much because the stock USB ports on the back of my Intel 8700K server still sound more natural.  If it wasn't for the Femto USB card, I don't think I could recommend this AMD build at all and so in this sense, just like the Femto Net Card, this card is a game changer.  


    Directly compared against my tX-USBultra with the EVOX cap + Ref10 master clock, the tX yields slightly better detail resolution with a greater sense of air and space.  Powered by a DR SR7 rail, it is very dynamic sounding.  Those that know my situation, however, know that I have had a love/hate relationship with the tX-USBultra.  I have pulled it out several times because it can sound thin and so perhaps I would prefer it with copper rather than silver DC cabling.  This is where the Femto USB card, to my ears, is better.  While it doesn't have that last bit of detail resolution and air that the tX has, what it brings to the table is rich, glorious tone and body.  Timbres are expressed beautifully and naturally and it is eminently a more listenable presentation to me.


    But there is a caveat.  This beautiful presentation that is so captivating ONLY occurs when I power the Femto USB card with a DR SR7 rail.  The results are not the same with an HDPlex, LPS-1.2, SR4, or even SR SR7.  What about bus power with a DR SR7 19V rail + HDPlex 800w DC-ATX converter providing that bus power?  With the Femto Net card, this provides excellent results.  If you purely bus power the Femto USB card, it will work but the Femto USB card will fail to pass 5V VBUS to your DAC and my DAC requires 5V VBUS power.  I found a workaround by using an iFI iDefender 3.0.  With this device, I was then able to send my DAC clean 5V VBUS power using an LPS-1.2 and so I was hopeful but unfortunately, this still didn't come close to what I got by externally powering this card with a DR SR7 rail.  


    Operating System + Playback software


    I placed this close to the bottom of my post but it deserves to be at the top.  Along with the power supply and the CPU, the OS and software player have tremendous ability to affect the sound presentation even with bit-perfect playback.  Many already know my preference for the combination of Euphony + Stylus and this preference has not changed. 


    Željko at Euphony has been using Rajiv and me (and perhaps others) to vet Euphony and Stylus updates to make sure his coding changes haven't somehow negatively altered SQ and it has been amazing to witness how even subtle changes did indeed negatively impact SQ.  On more than one occasion, we have been left scratching our heads and so credit to Željko for not being rash and careless with these updates. 


    As for Stylus vs Roon, I don't know what to say.  I spent $500 for a lifetime Roon subscription and given Roon's superior library management capabilities, I would like nothing better than to use Roon but every time I move back from Stylus to Roon for playback, there is a massive loss of engagement.  With complex orchestral music, Roon just sounds horribly controlled to me.  The latest 1.7 update has led to improvements but to my ears, not enough.  If I specifically allocate 12 of the 24 available cores (both real and virtual) that the 3900X offers, control improves but not enough.   StylusEP improves it but once again, not enough.  For my tastes, Stylus remains the best playback software I have heard at home.


    What is interesting is that the SGM Extreme uses Roon and during my brief time with the Extreme, I heard none of the control issues that I hear at home.  Somehow, Emile has figured out how to tame Roon in ways that I have not.  Some are aware that I have placed an order for my own Extreme and so no doubt, once I receive it, I will load Euphony and Stylus (via USB stick) and see how it compares.


    SGM Extreme


    Given the knowledge accumulated through so many hours of testing and comparing, the natural question arises why I would buy an SGM Extreme?  The bottom line is as much as I would like to, I cannot build a server of the caliber of the Extreme.  I don't think anyone but Emile can.  Having communicated with Emile at length over a span of months, it became evident that Emile has spent many more hours than I have with his testing and comparing.  He even quit his day job as a university level IT professional so he can test and compare all day long.  Unlike me, he has access to measuring equipment and has spent tens of thousands of Euros measuring the noise spectra of motherboards, CPUs, chipsets, clocks, memory, storage media, and power supplies to guide his path whereas I am left to random guesses as to which CPU and motherboard might sound best.  I also do not have the gifts that Emile has with respect to hardware and software optimization capability including network allocation.  I don't think most other IT professionals do either based on the fact that no one else has come up with a server like the Extreme.  Here is an example of one of Emile's e-mails to me and I think you'll quickly get the picture as to how Emile views music server design.  I had asked Emile why he felt he needed to use 48GB of RAM in the Extreme when this seemed like overkill and would potentially be a significant source of noise:


    "Well RAM is a topic on its own, to start with, the 2 cpu’s are split into domains (NUMA / SNC), so you really have 2 x 6 dimms, 6 for each CPU, they are not shared. Music services have their own cpu/dimms and the OS has its own cpu/dims. So its sort of a core and endpoint into a single machine going beyond just core allocations for individual processes. These Ram modules are a custom order type, similar to the Apacer types popular in the Jplay forums, but taking it just a bit further. They do create less noise and draw less current then other offerings. If more dimms reduce performance, it typically means your power supply is negatively impacted by the increased current draw. As occupying more memory channels increases bandwidth and reduces wait states, you do get better individual process performance."  "What you really want to do is reduce your hardware active processing times as much as possible. The net effect is much like a class A amplifier, you have a higher baseline power consumption, but power draw does not vary much, and this is very good for a more “natural/relaxed” sound. I hope this makes sense 🙂 But you do need a power supply which is very comfortable supplying the load. You really want the least possible variation in load, and higher cpu power / bandwidth systems are better at that with very low load music playback processes."


    As for the Extreme being a core and endpoint in a single chassis, this was interesting for me as well.  You basically have 2 CPUs with each CPU having its own dedicated RAM bank (24GB each) and so there is a genuine distribution of tasks between 2 machines just like dual Pink Faun 2.16Xs.  While I very much like what I heard in the Aries Cerat room at Munich this past May where dual 2.16Xs were playing, this configuration costs north of $30k, has fewer cores, uses a noisy SSD, consumes more than 200 watts, and capably functions as a room heater.  I find the Extreme to be a more elegant and practical solution and at least on paper, I believe it is the most technologically advanced music server at this time.  My brief listening experience in Taiko Audio's listening room did nothing to dissuade me from this opinion.  


    Happy holidays.   


  3. Building a DIY Music Server
    Building a DIY Music Server

    The yellow wire you see in my previous post is Mudorf silver/gold. I will be using this wire exclusively in this build. I estimated I need 6 meters for the internal wiring and 5 meters for the umbilical cords. I will be using the 15.5AWG wire where possible. Some of the boards cannot take 15.5AWG wire. I will be using 18AWG there. I thought about using terminals and thicker wire but I did not want to introduce another connector and material in the path. 


    CPU installed:



    I opted for a lower power Intel i9 - the 9900. Heat management is important. If the temperature in the chassis gets too hot, the sound becomes harsher. And given that I also have 6 regulators generating heat in the case, I decided to go with the 65W TDP Intel i9-9900 instead of the 95W TDP Intel i9-9900K. I think we get very similar performance for audio but a few degrees lower temperature would be beneficial. It's all about making the right compromises, and I will try to point out the compromises I am making and explain why. 


    Passive cooling installed as well:





    Done with the messy thermal compound paste that goes on the CPU, copper pipes, and all over you and the case!!! :).

    I used the ARCTIC MX-4 compound. If I was installing a 95W TDP CPU I would have used the Thermal Grizzly Kryonaut. 


    BTW, you might have noticed that I removed the WiFi module from the motherboard. That's why there is an empty slot next to the 8-pin EPS connector. WiFi is not needed for this build. I typically disable it in the BIOS but decided to completely remove it in this case. 


    Time to make some connectors. I am using gold plated pins (Molex part number 45750-1212). Thanks to @Volfram for bringing the gold plated ones to my attention. 




    Stay tuned...


  4. Hi filter basis
    hqplayer resampling filter setup guide for ordinary person

    I've been trying to avoid too much technical jargon in the manual, but balancing it is quite difficult. I'll try to explain things a bit further, but please ask if you would like to have some areas covered further.


    First a bit explanation on time and frequency domain, please excuse me for some technical jargon. Frequency is signal change as function of time. Thus a signal has presentation in both frequency and time domains. "Linear phase filter" is a filter where all frequencies pass with same time delay. "Minimum phase filter" is a filter where all frequencies pass through as fast as possible, higher frequencies faster than lower ones. Longer/steeper filters change faster from passing frequencies to not passing frequencies as function of frequency. Shorter/gentler filters transition more slowly or "gently" from pass to stop as function of frequency. More accurately the filter wants to detect frequencies and transition pass/stop faster, longer time the filter has to "look" at the signal. This has side effect called "ringing" or rather "time blur". On the other hand, extremely short filter like a one that looks only at single moment cannot filter anything at all, because it sees only single point of time at once without any history or future (so it cannot detect any frequencies as those are a change over time). Linear phase filter takes equal amount of history and future into account during calculation. The problem in this is that it is kind of unnatural for something that is going to happen in future to affect already the present. Minimum phase filter on the other hand considers only from present to past, so it doesn't reflect things that are coming in future. This "ringing" is already in most RedBook recordings, since in most cases the ADC has gone through down-conversion and possibly another round at mastering from 24/96 or similar to RedBook. "Apodizing" filter is one that replaces or modifies this original ringing with it's own - that can be less than the original. All the filters explained below are more or less "apodizing" unless otherwise noted.


    Why is "filtering" needed? Because otherwise upsampling/oversampling produces alias (distortion) components in frequencies above the original one. In down-conversion case it is even worse, because those components are produces below the original ones. D-A conversion also produces these components above half of the sampling rate frequency, and those are then removed by the analog reconstruction filters. Higher the sampling rate seen by the D-A conversion stage, simpler the following analog filter can be. Digital filters can easily outperform analog ones. Removing those spurious frequencies by filtering is called signal "reconstruction".


    So if I go from left to right on the main window...


    First is the filter selection, most of these can perform either up or down conversion, depending on what is needed.

    - So "IIR" the first one is how a steep analog filter would sound like, I don't recommend using it for anything else than upsampling and only at 2x or 3x ratios, although it can do higher ratios or down-conversion too. I think this is mostly useful to hear how "extreme analog" would sound like. Some DAC chips have slightly similar output stages.

    - Then there are three types of traditional "FIR", these are similar in construction to those ones used inside most DACs, the "asym" one being somewhere between linear and minimum phases, only taking "near future" into account. So a traditional design made as good as possible.

    - "FFT" is a special kind in that it performs it's work in frequency domain and is also fairly steep. This is technically closer to how audio codecs work than how upsampling is traditionally done. I don't know if any hardware oversampling implementation would use similar technique.

    - "poly-sinc-*" these are the ones I use most and recommend the most, these can perform conversion from most input rates to outputs rates in a single pass and with a very low CPU load. Single pass approach maximizes the filter precision. (those who will eventually ask, these are synchronous converters)

    - "sinc" is a true asynchronous converter and can perform conversion practically from any rate to any other rate. Although it is quite high quality, it has fairly high CPU load too and not recommended unless the "poly-sinc-*" ones cannot do the needed conversion.

    - "polynomial*" is not a filter as such, but just polynomial interpolation approach to upsampling. These look only at small number of samples to calculate a new one and thus don't "ring", but on the other hand the filtering performance is poor too. These kind of filters typically also cause premature treble roll-off (roughly 3 dB or so at 20 kHz for RedBook material, starting from ~10 kHz). These are the controversial upsampling "filters" some people like a lot while others don't like at all. (non-apodizing)

    - "minringFIR" this is a single-pass filter that is very similar to the polynomial interpolators above in that it is really short and looks only at very brief period of time, while still performing better at filtering and not having such treble roll-off issues. Not recommended for other than 2x/4x/8x/etc ratios. (non-apodizing)


    Then to the next item, dither and noise-shaping. This is needed whenever any processing is performed. Reason is that calculations can lead to results that have more precision than can be expressed in the resolution supported by the DAC. Just truncating or rounding the result to fit the DACs precision causes distortion that is directly related to the signal. Dither hides this rounding error into very low-level non-audible constant noise (a bit like thermal noise) - then it's no more related to the signal. Noise-shaping takes this further by moving this noise to less- or non-audible frequencies. Especially multi-bit converters but to some extent others too also benefit from noise-shaped upsampling in improved linearity.


    I don't recommend any noise-shaper for 44.1/48 kHz output rates, because there is no proper frequency space available where to park the noise.


    There are number of noise shapers:

    - "NS1" is a first-order shaper, just tilting the noise floor so that it increases towards higher frequencies and it has a bit of extra "against-the-wall" high frequency noise too. Not really recommended for anything, but included for completeness sake.

    - "NS4" is fourth order shaper that has a gentle step to move a bit of lower frequency noise to ultrasonic frequencies. The only shaper that I would say is useful at 88.2/96 kHz rates.

    - "NS5" is fifth order shaper that has been designed to be used at 352.8/384 kHz output rates or above. This one moves aggressively roughly 40 dB worth of noise from low frequencies to ultrasonic range.

    - "NS9" is ninth order shaper variant for use with 176.4/192 kHz, the step from low to higher noise is more clear, but otherwise similar to the "NS5"

    - "RPDF" this is just plain white noise, not really recommended, but also included for completeness sake.

    - "TPDF" is industry standard flat triangular dither, good for any case, especially for 44.1/48 playback cases. Doesn't generate practically any CPU load either.

    - "Gauss1" is Gaussian noise dither, should be more "perfect" than TPDF, but also loads the CPU more. Works for all cases too.


    Third selection is set of available output sampling rates, computed based on what the hardware and selected filter are capable of, in combination.


    Generally, I recommend choosing between "poly-sinc-*" filters and using highest possible sampling rate. Dither or noise-shaper chosen based on above description, "NS9" for 192 kHz output, "NS5" for 384 kHz output and "TPDF" or "Gauss1" for any lower rates...


    To be continued, I'll make two other posts. One for the DSD->PCM conversion and maybe other one for PCM->SDM (DSD) conversion.


    Hope this helps... :)

  5. Hi filter
    HQ Player
    13 minutes ago, AnotherSpin said:

    Miska, I am sure this question was answered already, but what is the the difference between apodizing and non-apodizing filters in a plain, non-technical words for those, who do not understand special terminology?) In my set closed-form gives more natural sound, while ext2 gives very sweet, but slightly artificial (in comparison with closed-form) sound. And, which other filters are non-apodizing besides closed-form?


    Since all ADCs that produce something like 44.1k output are oversampling, there's a decimation filter to convert the higher rate down to lower one. In addition, these days is quite common that recording is made for example using 96k sampling rate and then converted down to 44.1k using another decimation filter.


    Point of apodizing filter is to replace original decimation filter's impulse response with another one. This allows altering time- and frequency domain behavior of the original filter. You can get generally shorter "ringing" one, with something like poly-sinc-short or longer one with something like poly-sinc-ext2. Or you can change to a minimum-phase one.


    Another maybe more important point is to clean up aliasing band at the highest frequencies that happen due to "modern" half-band ADC and DAW decimation filters that have pretty much no suppression at the fs/2 (Nyquist) frequency, and thus content exceeding that in the higher rate source data folds down into lower frequencies. For example original CD release of Pink Floyd DSOTM didn't have this dirt band at top, but the latest remaster does. This applies primarily for source content at 44.1/48k rate, and to much lesser extent to hires content.


    closed-form is not really a filter, but instead interpolator, so it is non-apodizing due to that. poly-sinc-hb is a non-apodizing half-band filter a bit like the ones in modern ADC and DAC chips, but just better (higher precision and stop-band attenuation). minringFIR is another non-apodizing half-band filter with fairly slow roll-off, so it will let more images through too. You could compare poly-sinc-hb vs poly-sinc and minringFIR vs poly-sinc-short. Also poly-sinc-ext2 vs poly-sinc-xtr (xtr is quite a bit less apodizing than ext2).


    When you use non-apodizing filter the results largely depend on what kind of decimation filter was used for the source content, because all it's faults come through as-is. While apodizing filters give more consistent performance across the board by correcting faults of the original decimation filter and giving same impulse/frequency response across the board regardless of source content.


  6. Switch Cisco
    Router and Ethernet Switch suggestions
    14 hours ago, User471 said:



    I have just acquired a refurbed 2960-8TC as the first part of my attempts to build a fibre network for audio


    A couple of quick questions if I may



    What SFP module(s) do you use?



    What settings in terms of QoS aand so on do you use, if not defaults?




    this cable, get more than you need it's cheap  https://www.fs.com/products/50108.html


    It's plug and play, you shouldnt need to make any changes to your settings.


    Happy new year!

  7. Naa direct not recommended
    HQPlayer's Network Audio Adapter
    On 7/28/2019 at 2:06 AM, jabbr said:

    No need for an audiophile switch. Direct connect server to NAA is not recommended. The SQ of “not working” can’t be very good. It’s possibly that a really cheap switch could transmit noise (I don’t know, never had a bad switch). Just get a high quality switch. 


    And a switch will likely emit less noise noise by itself than a PC...


    A good switch in metal enclosure with a built-in PSU using proper IEC power cord is a good starting point.


    I personally have three switches on the way from my server to a NAA. Two room specific switches and one central switch at the patch bay.


  8. Audiolinux built
    Pink Faun 2.16
    On 8/7/2019 at 8:21 AM, EliteDJ said:

    Here's a link to the SPDIF Bridge https://www.pinkfaun.com/shop/bridge/45-2880-spdif-bridge.html, my dac's best sound quality is provided by its SPDIF inputs, usb is almost as good.  I know if I spent the $12,000 on a 2.16x another household energency would occur right after I received the unit, things always seem to work that way. :)  I have an easier time justifying $4,000 in parts.  



    I have the SPDIF bridge on my 2.16, as my DAC only has SPDIF RCA/XLR inputs.  With this setup, it's a relief to no longer need two very expensive USB cables, a USB conditioner, Mutec MC-3+ usb reclocker, 10M external clock.  I don't miss those items at all! 

  9. audiolinix ram
    A novel way to massively improve the SQ of computer audio streaming

    AudioLinux Guide 4




    Environment: AudioLinux already launched in GUI mode



    1. Open the Start here folder

    2. Open the Expert folder

    3. Open the Ramroot folder

    4. Run the Ramroot enable

    5. Warning may come out.  Don't worry.

    6. Click the power at the bottom bar.

    7. Click Reboot

    8. In the UEFI menu, choose the USB disk

    9. Answer yes

    10.  Wait for quite some time.


    Done.  RAMboot complete.












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

    AudioLinux Installation Guide 2:



    ** Continue from the end of Guide 1.  **

    1. Unplug the USB disk.  Note: After the burning, Windows can't recognize it and hence can't eject it.

    2. Insert the USB disk to a PC (power off) which you want to run the AudioLinux GUI mode

    3. Power on the PC, get into the UEFI, from the boot menu, choose the just created USB disk as the boot device and boot.

    4. Wait a relatively long time until you see a Windows environment (see the image below).  In the process, some text comes out from the screen.  Error message may come out but don't worry.  Wait until Windows environment comes out.


    Steps to mount a local drive

    1. Click the Start here folder (it may lag few seconds depend on the speed of your computer and the USB disk)

    2. Click the local Drive that contains your music files

    3. Input the password audiolinux0 as the password to mount

    4. Check the mount is complete.

    5. (As an example) Launch HQPlayer

    6. Add you library, set the settings of HQPlayer and play your music by HQPlayer.


    End of guide 2.  Enjoy!



    1. After reboot, the mount of the local drive is gone.

    2. Linux is pretty stable, you may leave for tens of days.  Hence 1 won't introduce much problem.

    3. Auto-mount of a drive is possible but needs using terminal, issue commands, and editing the file fstab.  It's better for the users to read the instructions and follow.  The boot may not be possible if fstab is wrong (as stated in the official site.)

    4. In principle you may do the RAMBoot.  However it's better to set up everything properly before using RAMBoot.







  • Create New...