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

[Gavin1977]

5 hours ago, Nenon said:

 

LOL :). I did not buy them for business, you know :). Full disclosure - I have six. 

I bought the first at $180 + shipping (almost $200 total). I saw it uses the same board as the Melco and bought another one - the one that went to Rajiv and Roy and now back in my system (Arris8200 --> EdgeRouter --> Buffalo --> Buffalo --> EtherREGEN --> Server). Then I posted about this and a couple of days later they were all gone. I thought: "That's what you get for sharing publicly". Well, the original idea that the Melco uses the same board was from our research master seeteeyou (credit goes to him), but I put the two side by side to confirm and compare the clocks.

Anyhow, I tried to buy more, but all my orders were cancelled, including an order from Walmart, which I found pretty unusual. I think Melco blacklisted my name :)). No, I actually think Buffalo took them out of the market. Not because of our posts, but because they were anounced end of life. The rest of the switches I bought were more expensive. The last one I found had super expensive shipping from Europe, and I paid nearly $400 for it. 

 

I have a grand plan for building the ultimate Buffalo modded switch:

- Replace the clock with a PinkFaun ultraOCXO. 

- Build in a solid aluminum chassis with two rails of Sean Jacobs DC4 LPS. 

- Isolate the transformer with a thick copper plate. 

- Add Gaia isolation feet.

- Find out what set of capacitors work best and add a capacitor board like on the Melco if it sounds better this way. 

- Furutech NCF IEC Inlet.

- Mundorf Silver/Gold wiring.

- SR Orange fuse.

And a few other things I have in mind.

 

The problem is I heard how my server sounds with no network connectivity at all. And I started to think that all these network switches, power supplies, cables, fiber optical isolation, and other network tweaks do is modulate the noise and make it sound better... I feel like they act like tone controls. Don't ask me how that works. But here is an example:

-  If I connect my network cable with all the network tweaks done, the sound changes. One of the things that you can immediately notice is the bass has more authority. But, no, I am not talking about the harsh bass that we get when there is more jitter. I am very sensitive on that. I am talking about clean, detailed, bass with a lot of definition and refinement. 

- You remove the network tweaks, and everything collapses. Literally. It's unlistenable. 

- You disconnect the network and disable the NIC and some of the sparkle disappears, that bass is nearly gone, and first reaction is you are missing something. But then you realize that maybe that sounds more natural. 

And all this is for locally played music, not streaming. That's why at this current stage I think that network tweaks are acting like tone controls. And I have managed to tweak my network, so my system sounds the way I like it. But I am taking a step back and exploring what no network connectivity does. 

 

Also, the recent developments at Taiko got me intrigued. They are into something that made Roon much better. 

This Asus Sage motherboard has KVM access. I can remote to my server while the NICs are disabled. What I have noticed is that disabling the NICs is an improvement.

I think there are two types of network noise. One is the electrical noise that comes over the network cable, which everyone is talking about and trying to clean up, isolate, reclock, etc. The other one is the noise that the computer generates when the Operating system is processing packets. We don't talk much about the latter. And I think that is the more harmful noise (at least after a good number of network tweaks). Every packet that comes to your network needs to be processed. That's why we have discovered that isolating your server on a seperate VLAN sounds better - your server does not have to process random broadcast packets from other devices on your network. This is also one of the reasons why we found that Stylus sounds better than Roon. Stylus caches the track before playing and tries to minimize any network activity. Roon constantly does something with the network. There is noise that is generated every time a packet is processed. Those are also CPU interrupts that are distracting the CPU from its main purpose - music playing. And no switch can do anything about that. I discovered this when I connected to my computer with KVM and disabled the NICs. Here is my original post about this:

https://whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-206#post-655430

Those recent findings made me rethink how much more I should invest in my network. It lowered down my enthusiasm and slowed down the grand plans a little bit. 

The Taiko thread is amazing...

 

”The other thing I learned is that this Roon modification made by Emile to my Extreme could be operable to non Extreme but Roon users so stay tuned”

 

I hope we can find out what the tweak is... I may actually try Roon if it’s that good.

 

Also, which software packages can be used to load music (or buffer general network traffic) into memory first?

[Gavin1977]

35 minutes ago, BigAlMc said:

 

I might be missing what @Gavin1977 meant but my understanding is that Emile at Taiko is ditching Roon altogether and building a new music software system from scratch.

 

Long story short they've concluded Roon sound quality has dropped so much that they are not willing to put up with Roon any longer. As a Roon lifetime license owner I'm a bit concerned and unfortunately the guys are Roon labs are just not worried about sound quality at all. They prefer adding totally bullshit features like being able to cast song lyrics to your TV 😠

 

On the bright side in my case Innuos have kinda saved the day with the experimental mode. This uses Roon as Core but plays to an internal Squeezebox player in the Zenith. The Squeezebox player sounded better than Roon did originally. But Innuos have also configured it to playback from RAM which further improves it. So kudos to the team at Innuos for taking recommendations from their users seriously. I believe that Austinpop was one of the many users including myself that badgered Nuno at Innuos about this. (Keep me honest Rajiv, if I'm wrong here).

 

Likewise, kudos to Innuos to being open minded about reclocking USB. I discussed with Nuno and he was initially very sceptical but said he was aware of the discussions here. But after following the progress of threads like this and feedback from users they embraced the reclocking thoery and voila - the Statement was born and the Phoenix followed later. Bravo for taking your users seriously even if you first suspected they might be crackpots 😆

 

Opposite of kudos (whatever that is? rotten tomatoes? cancellations?) to Roon labs. Your disdain for your audiophile users has been noted. And many will vote with their feet. Not that they care as we're just not the main demographic. 

 

Cheers,

Alan

 

 

 

The WBF topic has had a few posts since Nenons reference - Taiko have stated their reasons...developing their own software for ultimate control and sound quality of their ecosystem... great stuff.  I don’t think they’re necessarily doing it to outmode  Roon.  Plenty of folks have unrevealing HiFi systems where I’m sure that differences in software are not that apparent.

 

Nice idea/request floated on there as well of potentially releasing this software as a wider product for non-extreme owners to use.  Definitely up my street if sound quality is #1.

 

I’m still interested in what the Roon tweaks are through.

 

I might try booting an entire image of Windows into RAM and see how that sounds.

[Gavin1977]

On 8/21/2020 at 4:55 PM, hols said:It is not so much as the same as DSD which still gives a more sumptuous body and well harmonized soundstage.

Great feedback @hols  what do you mean by the above sentence... do you mean DSD still sounds better than PCM upsampled at the same rate?  Is the difference large enough to warrant the huge processing requirements for DSD?

 

I ask because I run T+A DAC 8 DSD, I can’t do NOS PCM, and am presently CPU limited (like most) to DSD256 for demanding filters.

 

Elsewhere a report of someone preferring straight PCM: 

 

[Gavin1977]

The 'Plus' version comes with Onload as default I believe - my card:

 

 

 

[Gavin1977]

16 hours ago, BOBO said:

have you guys tried to overclock the dac`s usb port to 1000hz using hidusbf ?

What’s the basis of this?

[Gavin1977]

Software for lossless digital volume control within windows, ideally at the system level.  Are there really no solutions for this?

 

Alternatively via Linux?

 

I used to use HQPlayer for volume control, but overkill since I no longer upsample and HQPlayer volume control will not integrate with a bluetooth/IR remote I believe.

[Gavin1977]

25 minutes ago, MarcelNL said:

there is one solution called LEEDH (unless I misspelled it) but I'm a bit sceptical about that too...digital volume control yes, but lossless??

https://www.processing-leedh.com/copie-de-presentation

https://www.audio “science” review/forum/index.php?threads/leedh-digital-volume.15582/

 

 

Unfortunately not available as standalone software, only through manufacturers.

[Gavin1977]

2 hours ago, Gavin1977 said:

Unfortunately not available as standalone software, only through manufacturers.

Unfortunately the Leedh Volume Control is lossless only down to -30dB and only under certain conditions (link).  Taiko have also not progressed with Digital Volume Control (DVC) and it's not on their development roadmap.  So, looks like I'm back to looking traditional volume control methods (passive in my case).

[Gavin1977]

Onload is apparently worthwhile, I haven’t tried to though - happy with 8522 as stock in Windows.  You can load onload in audiolinux and Gentoo player.

[Gavin1977]

On 7/26/2021 at 2:47 AM, romaz said:

 

It's been awhile since I've posted on this thread.  I like to post when I feel I have something meaningful to contribute.  Hopefully, some of you will find this to be meaningful.  If you have no interest in the Chord DAVE or in power supplies in general, then feel free to ignore this long post.  It is the storyline of how the ARC6 DC4 for DAVE came to be.

 

In November of 2017, while in he U.K. visiting Rob Watts, he was kind enough to share with me the necessary specifications for building an external PSU for DAVE.  In exchange, he had expressed a desire to hear this PSU during his planned visit to my home in March of 2018.  While still in the U.K., I initially approached Paul Hynes with this PSU project.  He expressed interest in it and was confident DAVE could be improved with a better power supply as he had built outboard PSUs for a Berkeley DAC and TotalDAC and those customers were apparently quite pleased.  But because Paul said he was busy and given his history of delays, as I was eager to make sure this PSU was ready in time for Rob's visit in early March, I reached out to Sean Jacobs.  It was a pleasure interacting with Sean and I had the further privilege of picking up this PSU personally from Sean at his home in West Yorkshire, a 2 hour and 20 minute train ride form London.   

 

 

The above is a photo of Sean preparing to explant the stock SMPS from my DAVE.  This photo was taken on January 29, 2018, just weeks before Rob's visit to my home.  To add to the drama, to our knowledge, no one had ever attempted to design a custom PSU for DAVE before and so other than Rob's recommended specifications (minimum of 2A per rail), there was no precedent.  Until I brought my DAVE to Sean that day, he had never actually personally seen a DAVE before and so he designed this PSU based on photos he had seen on the internet.  DAVE has 3 rails and so it remains a mystery to me how he knew which leads on the Molex connector fed which rail.  What's worse is when I got to Sean's house, while the "surgery" proved to be quite easy, because this PSU was designed for 120V AC and because the U.K. operates at 230V AC, we had no way to test it but Sean was pretty confident that it "should" work. 

 

Fortunately, it did work.  The time and effort it took to pick this up personally from Sean, the expense, and the voiding of my DAVE's warranty -- it was worth it.  First of all, it was a pleasure to interact with Sean.  While in West Yorkshire, we had lunch at a local pub that was at least 500 years old.  Not unusual for for people in the U.K. or people from many other countries, I suppose, but as an American, it was novel and I loved it.  During that conversation, I realized Sean wasn't just a PhD-level robotics engineer, an avid mountain climber, and skilled PSU designer and builder, he was also a genuinely nice human being.   As for the resultant SQ improvement from this DC3 for DAVE, it was splendid, better than I hoped for, and Rob was able to hear it for himself during his visit. 

 

The DC3 for DAVE had been the standard I had been living with until August 3, 2020, when Sean designed and built another "first" for my DAVE, this time, a battery-operated DC4.  I was aware of the standard DC4 but as I already had an SR7 for DAVE on order with Paul Hynes, my desire was for a PSU that would allow me to travel with DAVE to a mountain cabin that we frequently stay at that has no reliable electricity.  At the same time, I was intrigued with the concept of a DAVE powered by low-impedance batteries and what that might sound like.  Here is a photo of the battery-powered DC4 versus the larger DC3 chassis:

 

 

Fortuitously, this lightweight and compact PSU which contained Sean's cascaded DC4 regulators had battery inputs that allowed me to connect 3 separate batteries to power each of DAVE's 3 rails. 

 

 

This allowed me to feed DAVE's digital (5V) rail with any DC source between 14-18V and DAVE's analog (+/-15V) rails with any DC source between 24-30V.  This new platform allowed for lots of experimentation and it especially caught @ray-dude's interest.  Because batteries are known to be "high-impedance" DC sources compared to AC/DC rectification, @Nenon was initially skeptical that a battery-powered DC4 could sound as good as a standard AC-powered DC4 but as Ray came up with the brilliant idea of combining highly volatile and combustible lithium batteries in parallel to achieve peak instantaneous discharge capabilities of up to 1000A, @Nenon had to eventually concede that this type of battery array could outperform a standard DC4 and so he jumped in.  With @austinpop safely monitoring our progress from his fire-safe home in Austin, here is what we learned:

 

1.  Low output impedance is just as important as low noise with respect to SQ when designing a PSU and is more difficult to achieve.  If low noise is all that mattered, then why not just use batteries for everything?  Try battery powering a DAVE with any ol' cheap battery that you can buy from Amazon such as a PowerAdd that are designed for recharging cell phones and what you'll get are soft, slow, and smeared transients, a flat sound stage, and weak dynamic contrasts.  As you start to lower the output impedance from the power source, music starts to take form and comes to life.  It just becomes more 3D and palpable and low output impedance is what differentiates a DC4, SR7, Farad, or LPS-1.2 from the rest.  It's amazing how few ultra-low output impedance PSUs there are that are being sold for audio.  Chances are that if the manufacturer fails to mention "output impedance" or better yet, fails to state the output impedance of their PSU in their ad, that PSU is probably not a low output impedance PSU.  Despite the long wait times for PSUs from Sean Jacobs and Paul Hynes, there's a reason people have queued up for these devices.

 

With respect to our battery experiments, while we had no way to directly measure the impedance of a battery, we could approximate it with an internal resistance meter like this AideTek that you can purchase on Amazon and so this device was especially useful for making relative comparisons:

 

 

If you connect battery cells in series, voltage output goes up but if you connect them in parallel, peak current discharge capability of that array of batteries goes up while the internal resistance of the array as a whole goes down and so we used the peak discharge capability of a battery array as a surrogate for output impedance.  The problem with this from a practical standpoint is the number of batteries that you would need to combine in parallel becomes quite large.  Here is an array that @ray-dude put together.  Imagine having something like this by your DAVE in your listening room:

 

 

This led @Nenon to design capacitor arrays since adding energy storage capacitance in parallel to the batteries effectively drops output impedance.  @Nenon and @ray-dude experimented with capacitors of various sizes and configurations:

 

 

 

We even contemplated these giant paint-can sized caps that were expressly designed by Mundorf for Emile at Taiko Audio:
 

 

What was initially an experiment about batteries soon became more about capacitors.  With capacitors, it was now possible to achieve output impedances of <1 milliohm.  With the aid of these capacitors and at this ultra low level of output impedance, the standard DC4 which @Nenon was kind enough to make available to us for direct A/B comparisons was now soundly being outperformed by batteries.  

 

2.  Low output impedance is important for analog but our testing with DAVE suggests low output impedance is more important for digital.  I think there are many smart engineers, designers, and manufacturers out there that do not believe output impedance matters for digital because it shows in their designs.  These people believe with digital, it's purely about low noise.  For example, with my Oppo UDP-205, it's interesting to note that in stock form, this device has a transformer-based PSU with low ESR caps for the analog side and a basic SMPS for the digital side.  When I opened up a dCS Network Bridge, I noticed the miniscule transformer inside.  As I opened up an Innuos Zenith SE Mk2 music server, the 400VA transformer used by Sean Jacobs was much more massive in comparison despite the fact that the Innuos Zenith uses a low power Pentium CPU and has no high power requirement.  I don't think it's a coincidence that the Innuos sounded much more dynamic than the dCS Network Bridge in my A/B testing:

 

https://www.head-fi.org/threads/review-comparison-of-5-high-end-digital-music-servers-aurender-n10-cad-cat-server-totaldac-d1-server-auralic-aries-audiophile-vortex-box.787020/page-69#post-13801299 

 

With respect to our experiments with DAVE, applying a lower output impedance power source to the analog rails made a notable difference with respect to dynamics but that difference was much more profound on DAVE's digital rail.  If all we did was apply ultra low output impedance power to DAVE's digital rail, it would be already be a significant win.

 

3.  Batteries are ultimately NOT what is best for digital.  Yes, digital is more susceptible to HF noise and digital also gives off more HF noise and so to isolate it from your other components by powering it with batteries makes sense in theory but in practice, it does not necessarily sound better.  Even using a large array of batteries in parallel with approximately 5 milliohms of impedance, while these batteries can discharge quickly, they do not seem to have the agility that transformer-based power sources are capable of.  Since my DC4 has the ability to connect to any DC source, battery or otherwise, I was able to compare various batteries against an SR4T from Paul Hynes.  According to Paul the SR4 has an output impedance of <3 milliohms from DC to 100kHz but when you combine the impedance of the DC cable and the barrel connector (which are horribly high impedance), then the output impedance of both sources are roughly the same and while the batteries and the SR4T achieve roughly the same dynamic contrasts (which is to say both are excellent), the batteries in comparison sound slow, soft, and smeared around the edges.  The tiny SR4T does far better as an energizing source for DAVE's digital rail than a giant LiFePO4 battery.

 

4.  The ARC6 DC4 - A new king is crowned. 

 

 

The ARC6 that @austinpop had in his possession made it to me on Thursday and I have been able to directly compare it against the latest hybrid DC4 that I currently have (which was also built by @Nenon) which uses an SR4T + cap array to power DAVE's digital rail and 2 giant LiFePO4 batteries + cap array to power DAVE's 2 analog rails.  While I preferred the SR4T over batteries on DAVE's digital rail, I had no high-quality transformer-based supply that could output 24-30V for DAVE's analog rails and so I was forced to use batteries there but I found the characteristic "slower" nature of batteries on the analog rails to be to its advantage as the resultant sound was immersive, liquid, non-fatiguing, and pleasingly musical.  I thought the character of the batteries worked really well on the analog rails and so I was very interested to know just how the ARC6 would compare.  

 

The ARC6 manifests the best of everything we learned through our months of experimentation.  It utilizes Sean's cascaded regulation which is in both the ARC6 but also my unit.  It incorporates 594,000 uF of Mundorf capacitors for DAVE's 3 rails but my unit has this, too. 

 

 

The ARC6 uses @Nenon's preferred Mundorf Silver/Gold DC wiring.  Having done careful A/B comparisons against this more expensive Mundorf wire and Sean's less expensive Neotech OCC copper option (which is what comes stock), I have found myself preferring the Mundorf wire and so the unit I have has the same wire.  The only difference is the ARC6 uses @Nenon's and Sean's best transformer-based answer to a giant battery as an energizing source for the capacitor arrays.  With the giant capacitor array, did the energizing source (transformer vs battery) matter that much and if so, which was better?

 

Because I'm capable of assessing the digital and analog rails separately, that's what I did.  On the digital rail, I already knew a transformer is better than a battery and so no surprise, that's what I found.  What wasn't clear was how much better, if at all, the ARC6 was compared to the SR4T and the answer is it's unquestionably better, A LOT BETTER.  I guess I shouldn't be too surprised since the ARC6 is utilizing a giant 600VA transformer with a newly improved recap board along with a giant choke that much more effectively filters rectification noise without stunting dynamics like the older recap board did with the original DC4.  With the SR4T, I'm connecting this energizing device to the DC4 chassis via a DC cable terminated with a high-impedance 2.5mm x 5.5mm barrel connector.  With the ARC6, current paths are much shorter and devoid of high impedance connectors.  Comparing ARC6 vs DC4, both result in excellent dynamic contrasts in large part due to the capacitor array but the ARC6 is just incredibly good with respect to noise floor and clarity.  The "blackness" of the ARC6 over the SR4T in this instance is just eerily striking.  There's just tons of space.  Even when I compared the ARC6 to a giant LiFePO4 battery, most people would assume the battery would be quieter but in this case, it is not.  Batteries create their own noise and the ARC6 is good proof that a properly designed and implemented transformer-based, AC-to-DC rectified circuit can sound cleaner.

 

Regarding ARC6 vs batteries on DAVE's analog rails, this is where things got interesting for me.  It turns out the ARC6 is sensitive to power cords and more specifically, it is sensitive to bad power cords.  Quite randomly, I picked out the first spare power cord from my cabinet that I could find and it happened to be the stock Transparent 12awg power cord that came with my D'Agostino HD preamp that I never previously used and so as far as I know, this power cord had virtually zero hours on it.  For those that have experience with Transparent cabling, their house sound is "fullness" but what I first heard with the ARC6 on the analog rails compared against the batteries was a "leanness" that sounded anything but musical.  At some point, a brightness developed and that is when I realized this changing sound was possibly due to a power cord that wasn't broken in.  After a swap of power cords, it was a relief to find that a different power cord made all the difference.  With an inexpensive TG SLVR power cord I've had for years, it became clear that the ARC6 is a full-bodied and rich sounding PSU and even more emotive and satisfying than the batteries.  There was tremendous dynamics with a giant expansive sound stage that was bigger than the batteries and with a blackness that was just as eerily good here as it was on the digital rail.  From bass to treble, definition was better on the ARC6.  Micro details were much more easily gleaned with amazing clarity.  There was also the most incredibly natural sounding decay I've heard from my DAVE with the ARC6 that lingers in the air longer and gradually disappears.  With the batteries, it lingers then then more abruptly cuts off.   Yes, the batteries still have their appeal but the overall appeal of the qualities of the ARC6 is greater.  

 

If you own a DAVE with it's stock PSU, it's hard to accurately convey just how big a step change the ARC6 brings.  It really is transformational and all credit goes to @Nenon and Sean Jacobs.  At the very least, imo, you owe it to yourself to get a DC3 because swapping out DAVE's PSU is among the highest value things you can do to your DAVE and it's easy to do.  I'm sure when Paul Hynes' SR7 for DAVE finally comes out, it will transform your DAVE if you are in line for one.  But with regards to the ARC6, coming from a baseline of a DC3 and then DC4, which are both excellent in their own rights, it's amazing how the ARC6 is just ridiculously better.  

Those mundorfs are great, but what about ultra caps?  I would have thought they’d be preferable when trying to achieve ultra low impedance.  Also smaller and cheaper.

[Gavin1977]

On it's way to me.... https://micro-audio.com/store/product/smps440-tk/

[Gavin1977]

5 hours ago, drjimwillie said:

Hi Gavin

do you have any update, please?

are you using this with the Taiko DC DC ATX?

Not yet - still being made.  I only have HDPLEX 800w DC-DC.

 

I will post when I have feedback 

[Gavin1977]

Price reduced if anyone is interested:

 

 

[Gavin1977]

Just thought I’d ask if anyone has revisited Ian Canada’s DIY RaspberryPI streamers lately and compare against some of the latest breed of network streams (e.g Innuos, Aqua LinQ etc…).

 

He’s been busy churning out upgrades…