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

[Metnoc]

50 minutes ago, austinpop said:

Listening Impressions with custom frequencies in AL v099

 

Like many have found here, I had previously found an uptick in harshness - particularly noticeable in electric guitar - when turning on Turbo/SpeedStep in the BIOS. So for a while now, I've been running at:

• Turbo/SpeedStep off in BIOS
• Extreme2 boot mode
• Governor Max, turbo off

On my Dell server machine, the difference between base (3.4 GHz) and peak (3.9 GHz) frequency is not too great, but in talking to Roy, he is playing with a CPU with a base of 2.4 and peak of 4.0 GHz. Which made us wonder: is there a sweet spot above the base and below peak for harshness and dynamics? 

 

Which led to a suggestion to Piero, to allow us to set custom frequencies. And lo and behold, he has delivered it in v.099! The theory here is that rather than turning off Turbo in the BIOS, you leave it on. Instead, use the Governor custom frequency setting to "dial in" a frequency. Also in theory, the "Idle=poll" setting of Extreme2 causes the current frequency to stay pegged to the current MAX setting. 

 

With all that said, here's what I heard, by varying frequencies on the server. The server is running Roon, while the endpoint (DS-1) is running squeezelite. I have not yet tried this experiment on the endpoint, but plan to at some point. 

 

So on the server - first, with turbo OFF, I varied frequencies from 100% (i.e. 3.4 GHz) down in 10% steps to 60%.

• Functionally, this worked. The governor setting would show the Min and Max frequency set to the desired value. And - going into the monitor menu and checking current frequency would show a match - there was always a small difference (<100MHz) between actual and the governor settings 
• SQ: With each step there was no change in harshness, but a noticeable drop in dynamics. 

Turning Turbo ON:

• Functionally, this was puzzling.
  • On my i7-4770, setting 100% (MAX) only raised the setting to 3.7 GHz, not the 3.9 rated peak. However, lower governor settings were respected.
  • On Roy's CPU, the system did not respect any settings, and stayed pegged at 3.8GHz, which is still not the peak of 4.0 GHz.
• SQ wise:
  • my Dell got much harsher even with the modest rise from 3.4 to 3.7GHz, with no change in dynamics
  • Roy's system got more dynamic with no increase in harshness.

So what does this all mean? Well, it means that Intel systems seem to respond differently to CPU governors (pstate-frequency), based on the BIOS and processor. We need more data points. I plan to try this on my endpoint.

 

If you've been running with turbo/speedstep off in the BIOS, try turning it back on, and give the custom frequency governors a shot. It would be good to know:

1. listening impressions at different frequencies
2. whether you can control frequencies (especially between base and peak) in turbo mode.

It's possible that the end result is that for some setups, turbo mode - either at the max or some intermediate frequency between base and peak - is better sounding than "turbo off."

 

Please give it a try.

 

Haven't done very much testing yet but for the time being I've settled on the following settings for my Roon core running win10 with fidelizer purist/streaming mode. Cpu: i3-8130U. Endpoint: sms200ultra.

Bios: Turn off cpu turbo

Win10: Set cpu freq to a fixed 1,5 ghz

 

This perhaps sounds a little more "analogue" as in less hf noise compared to setting the cpu to 100% which is 2,2 ghz, (turbo is 3,4 ghz).

[numlog]

 

if there is an option for it, have you tried changing voltage settings? 

Setting a  manual core voltage has had a big influence on SQ , the effect is different to clock speed changes.

there is limit to how low you can go before it becomes unstable, the limit is determined by the clock speed.

Important to disable SVID if the option is there

[soares]

Following some good reviews in several fora, I went for a second hand Cisco Catalyst 2960-8TC-L (just 55$, while new was 789$), while waiting for a test between the TLC switch and the yet to come Etherregen.

Having also acquired recently a new amplifier, I thought it would be a good idea do do an overall cleaning and get rid of some Spaghetti, namely the 2 TP-link MC200CM and its LPS’s (LPS-1 and Sbooster), between my server and the router.

Yes, I can testify that it was a clear SQ improvement over the TP links, so now my server (sonic transporter i5)+ URendu + router are all connected to the Cisco. At the same time I took advantage of the variable output of my HDPlex to feed my Router (the other is feeding my server) and again I was surprised with the results (it improved again but the delta was lower).

But the largest surprise come from the improvement of quality of my TV image connected by wifi through an Apple TV. On top of that, when connecting the Apple ethernet cable to the Cisco I got another substantial improvement. So, guys, despite all the doubts surrounding modded switches and routers, it really makes a different, at least on my system.

Just wanted to share with you that this couple of changes resulted on a substantial improvement of audio and video on my system (yes video!).

By the way, when feeding a Qutest, the Sbooster MKII is the winner among LPS-1, IFi (ipower); HDplex and the Stock one. Unfortunately, I did not test the LPS-2.  

Cheers

[rickca]

On 2/8/2018 at 10:54 PM, rickca said:

I got that Startech 4-port dual channel USB 3.1 Gen2 card and I'm in the process of sourcing a DC jack to SATA power connector cable with 5V for it.  I will try using an LPS-1.2 directly to power this card before implementing the 6V ==>5V LT3045 approach.  The DC jack will be mounted on an expansion slot bracket.

I just found out from Startech that this card (PEXUSB314A2V) uses 12V from the optional LP4 molex or SATA power connector, not 5V.  This makes sense because the card is designed to work with power from the PCIe slot only ... and PCIe supplies only 3.3V and 12V.

 

[numlog]

6 hours ago, rickca said:

I just found out from Startech that this card (PEXUSB314A2V) uses 12V from the optional LP4 molex or SATA power connector, not 5V.  This makes sense because the card is designed to work with power from the PCIe slot only ... and PCIe supplies only 3.3V and 12V.

 

Was the goal to provide clean 5V through the USB ports or power the USB chips? you can still achieve both with that card

[rickca]

 

39 minutes ago, numlog said:

Was the goal to provide clean 5V through the USB ports or power the USB chips? you can still achieve both with that card

How can you achieve both?  Here is what I got from Startech

 

[numlog]

1 hour ago, rickca said:

 

How can you achieve both?  Here is what I got from Startech

 

the original idea involved cutting the PCIe power pins, but you did mention before that you wouldnt like to do that.

Instead of cutting these pins you could try putting tape over them.

the 12V input uses a switching converter to get 5V and  introduces noise to the rest of the card even if you dont use the internal 5v output  so for the best results you need to stop the 12V getting into the card at all by taping the PCIe power.

To see if the tape is working test the green fuses at the ports with multimeter to check that the internal 5V power is disconnected.

you can wire a DC jack to the fuses for the external supply.

know that the card will give a false  ''current overload'' warning if there is no 5V supply at all.

(I discovered with the 2 port version if you dont want/need a 5V supply from the card you can connect the 3.3V supply to the disconnected 5V output on the ports and it will trick the USB chips into not giving this error, the reason is the 5V output is connected to ''current sense'' pins on the chip)

 

Powering the chips is not easy, especially with the 4 port version.
The ASM1142 USB chips need 3.3v and 1.2V, the 2 port card used PCIe 3.3V directly and surprisingly there was linear regulator on board that drops that down to 1.2V.

 

 problems with 4 port version:

there are more switching voltage converters visible on the 4 port version, i guess it might use it to convert 3.3V to 1.2V, not as ideal but it would still be worth using an external 3.3V. there is one advantage, the 1.2V switching converters would draw much less current than a 1.2V linear regulator so load on the 3.3V supply is lessened.

 

There are no big 3.3V input filter caps visible on the 4 port card (unless they are on the other side) that you could solder the external supply to, you would need to scrap away some of the solder mask on 3.3V supply traces to connect it.

[numlog]

this is what I mean about connecting 3.3V supply, the trace for 3.3V ont he back of the card is very big so its not too hard. dont forget there is a single 3,3V pin on the front too.

Its not necessary to do it on ground plane for GND, the SATA and MOLEX sockets have ground, I prefered the neater wiring (also the wires I used are way too thick).

You could even connect wires from the unused 5V and 3.3V input on SATA socket to this trace and USB port fuses and have a neat an easily disconnected external sata supply.

 

also see in the pic in the previous post, most of the unlabeled pins arent connected to the card, tape over these too, the bigger surface will help the tape stay in place.

 

[lmitche]

Gee, I just connected a straight SATA connector at 12 volts from the Hdplex lt3045 output to the Startech card. This uses Gotham cable from end to end. This card is connected to a 3.0 USB Realtek NIC and Startech 3.5 inch drive enclosure.

 

Happily I can report a very nice bump up in SQ, especially high end extension and clarity. While tempting, I don't think I will be doing any more radical surgery. I may order a 15 volt to 12 volt 2 amp lt3045 chain as a second regulator.

 

Thank you Rick!

[rickca]

@numlog your expertise is impressive!  Unfortunately, I am unable to do anything requiring fine motor skills like soldering due to my shaky hands.  I really appreciate you taking the time to document your solution.  I hope another user will benefit from your guidance.

[numlog]

Just now, rickca said:

@numlog your expertise is impressive!  Unfortunately, I am unable to do anything requiring fine motor skills like soldering due to my shaky hands.  I really appreciate you taking the time to document your solution.  I hope another user will benefit from your guidance.

sorry to hear that 

[elan120]

My recent network journey and experience.

 

This past year was a busy but exciting year for me upgrading different sections of my system.  I went from a two-box setup to single server, concentrate in upgrading mobo, USB, and network clocks so I can utilize my existing reference clock, improving server CPU power to run HQPlayer’s non-2’s filters when upsampling from redbook to DSD512, refining DC power feeding to mobo, HDD, network card, and USB card.  All these efforts result in great sonic improvements that I feel content for a short while until the AL/NUC heatwave.

 

After reading and following Roy and Larry’s NUC related posts, I also gave AL/NUC a try, and like many of you have experienced, I too, liked the result better than my one box server.  This new setup has me busy again since late last year, trying to figure out the best setup for my system.  First, I have NUC endpoint setup in wifi, streaming music files from my server wirelessly, this worked nicely for a short while until I gave my OCXO clocked/LT3045 powered switch a quick comparison test and decided to stay with the wired setup.

 

After my network scheme is more settled, my time mostly spent testing different AL configurations, but recently, I started to see an area could be improved further, that is the ethernet cable area.  In the beginning, I didn’t think ethernet cable can be a bottleneck, I tried various generic CAT6, 7 cables with no obvious difference, then I tried Ghant JSSG ethernet cable with slight improvement, later test a couple generic CAT7 cable with JSSG360 treatment with slight improvement as well, but my feeling towards ethernet cable changed from my recent experience with Sablon Audio.  I mostly get to know their ethernet cable based on positive inputs from different forums, some online reviews, and discussions privately with few members here.  I was conservative at first, ordering just one cable to replace the one from switch to NUC endpoint, surprisingly, after inserting this cable, right away, I detect a positive change that isn’t subtle.  With 4 days of burn-in, the cable is settling down nicely, and clearly makes sound stage a bit wider but much deeper.  Overall, I sense an increased overall detail, very neutral and smooth, music is tighter, denser, bass layering is more defined, and piano sound more delicate, crisp, and controlled.  As a result, I placed a second order to replace the cable between server and switch.

 

As a side note, while ordering the second Sablon ethernet cable, based on the positive experience from the first, I decided to give their digital cable a try as well, and this cable, as expected, didn’t disappoint at all.  I tested this cable from sCLK-OCX10 to txUSBUltra that is providing the clocking signal to my Singxer SU-1, and from sCLK-OCX10 to sCLK-EX clock board that is providing the clocking signal to server mobo system clock, network clocks, and JCAT Net Card clock.  This cable is very quiet, took resolution, focus to another level.  The improved soundstage and ability to detect elements in soundstage is quite amazing.  In the end, I am now working on another order for two more of this digital cable.

 

It has been a lot of fun making changes to improve my system, and I do plan to test more components in this section, that includes sNH-10G, EtherRegen when available, and OCXO clocked/LT3045 powered FMC.  Hopefully, there are room for more improvements. 

 

For reference, below is my current network configuration:

[numlog]

With full size desktops its typical for MoBo ground to be connected to the case in many places, it could be good idea to isolate it and have the case like a floating JSSG shield.

 

PSU mounting screw, MoBo standoff screws and PCIe bracket screws are some of the main connection points.

 

 

[Metnoc]

I've borrowed Ansuz A2 ethernet cable and tried it instead of my own Chord Shawline. The results where a baffling expansion of the soundstage and the noise floor lowered as well (which made it easier to distinguish decay and a experience of room information for a more immersive experience) . I feel that the sound had a heightened immediacy with snappier transients, all quite spectacular. However the last "information" in the uppermost tweeter frequency felt just a little bit lacking. My theory is that is because of copper only cable, which usually cant convey that type of information as silver cables. But if that's a problem or not depends on your system and tastes I think. The sound improvement where the equivalent upgrading to a better power cord. I was quite stunned by the scale of the improvement.

My scheme:
Router -> normal cat 7 cable -> cheap switch with ifi iPower -> supra cat 8 (replaced this with my chord while I used the Ansuz, It might betterd the sound a little bit but I didn't test this back and forth so can't tell for sure) -> Zotac 620 (Roon core, win10/fidelizer, double ethernet ports so direct connection to my endpoint) -> ansuz a2 -> sms200ultra (lps 1.2 power) -> anti cables usb -> Aqua la scala dac.


 

[kyrill]

I don t see the  1800 US$  Ansuz A2 ethernet cable  to be in the same league as the anti usb cable.

I will not be surprised the Ansuz A2 ethernet cable  needs a more ehh refined usb cable?

 

Nice DAC by the way

[Metnoc]

5 hours ago, kyrill said:

I don t see the  1800 US$  Ansuz A2 ethernet cable  to be in the same league as the anti usb cable.

I will not be surprised the Ansuz A2 ethernet cable  needs a more ehh refined usb cable?

 

Nice DAC by the way

I've actually compared the Ansuz a2 usb vs my anti usb.
The a2 sounded warmer (again due to copper I guess), perhaps more analouge in a classical sense.
The anti however where way more resolving of the upper most frequencies.
Did quite a few comparisons with both cheaper and more expensive dacs.
Aqua la scala however just has a sense of "thereness" that few dacs and compete with.
This probably due to the non-digital-filter solution, because if I upsample (which automatically includes a digital filter in Roon at least) the Aqua lose this "thereness" it then sounds more distant and not really special anymore. 

[auricgoldfinger]

UPDATED COMPARISON:  sMS-200ultra vs. NUC7PJYH

 

Today I upgraded my sMS-200ultra to the latest Eunhasu image v.0.4.56 for the purpose of comparing this new and improved version to my NUC7PJYH.  By way of reminder, my sMS-200ultra has some unique attributes that are described in my initial comparison to the NUC7PJYH.

 

 

Since that time, I have made the following changes to my chain:

 

1. I removed the FMCs. 

 

2. I upgraded the tX-USBultra to the “SE” version with the Evox capacitor.  In addition, I further improved the internal DC cable from the SOtM silver wire upgrade to a cable made from a pair of 20 AWG solid core Neotech silver wires with JSSG 360 shielding. (The standard SOtM silver DC upgrade uses unshielded 22 AWG stranded silver of lesser quality.)  Although it is outside the scope of this review, I would like to note that the tX-USBultra now sounds like a completely new device which is vastly superior to the original model.  The shielded 20 AWG solid core silver Neotech internal DC cable has lowered the noise floor, which has the effect of magnifying the improved dynamics and musicality arising from the Evox capacitor.

 

3. I purchased a Sony SR-64HXA “low noise” microSD card for the sMS-200ultra. 

 

I began by listening to the sMS-200ultra.  The latest image has made a nice improvement to the sound quality.  Detail and dynamics are better than I remembered them.  My initial reaction was that the gap with the NUC7PJYH had been closed.

 

When I switched to the NUC7PJYH, however, I heard an obvious difference in sound quality.  I only needed to listen for a few seconds to know that the NUC7PJYH still outperforms the sMS-200ultra in terms of detail and dynamics.  I also feel like the NUC7PJYH has more body, better tonality, and sounds more natural overall.

 

[tboooe]

Is anyone using Server 2012 R2 or Server 2016 with NUC7PJYH?  I am thinking of getting one to use as my NAA but was worried about the NIC driver compatibility.  Last year I bought a NUC7i3BEH but had to return it because I could get the NIC drivers to work on Server 2102 R2.  From the Intel support site, it only shows Windows 10 support for the NIC but after some digging around it seems the NUC7PJYH uses a Realtek RTL8111 NIC.  According to this site  there is a driver for Server 2012 but not sure if the Windows 10 driver will work with Server 2016.  Any insight would be greatly appreciated.

[Lebouwsky]

Windows Server 2016 with Audiophile Optimizer has served me very well. After some reading on the internet I decided to give it a go and to expand the AO with Fidelizer. I did a clean install of WS2016 including drivers and software, everything worked. Then “fidelized” Jriver and then optimized it with AO.

 

To be honest, I’m not convinced this is a step up, more of a step sideways at max and maybe even a step down. Ofcourse I need to give it a little more time, but mostly my first impression is right 90% of the time.

 

Anyone else with some experience in this area?

[Bricki]

Audiolinux isolated CPU cores listening impressions

 

I have got the isolated CPU core function working on my roon server with 5200u CPU and 8 GB of RAM in ramroot. I also have the isolated CPU core function working on my NUC7CJYH running roon bridge with 4 GB of RAM in ramroot. The nuc is also network bridge to the server.

 

The isolated CPU core function brings a clearer sound. Background details in music are more easily discerned. Electric guitar sounds less glassy and digital and more natural and real. Individual sounds are not as flat and they have more of a 3 dimensional presence. In the live version of dire straits"you and your friend" the crowd at the beginning of the track creates a real concert atmosphere and individual crowd members can be heard in the distance. Before the CPU core isolation the crowd had less individual sounds and came across as more of a whole sound.

 

I am very pleased with the increase in SQ it has come as quite a surprise.

 

I highly recommend other NUC/AL owners give it a try 👍🔊😎

[luisma]

19 hours ago, tboooe said:

Is anyone using Server 2012 R2 or Server 2016 with NUC7PJYH?  I am thinking of getting one to use as my NAA but was worried about the NIC driver compatibility.  Last year I bought a NUC7i3BEH but had to return it because I could get the NIC drivers to work on Server 2102 R2.  From the Intel support site, it only shows Windows 10 support for the NIC but after some digging around it seems the NUC7PJYH uses a Realtek RTL8111 NIC.  According to this site  there is a driver for Server 2012 but not sure if the Windows 10 driver will work with Server 2016.  Any insight would be greatly appreciated.

I read somewhere here people are using already server 2019 and if I recall correctly they are seeing improvements, my plan is to have 3 naa's for comparison simultaneously. 

 

Allo USB Bridge naa 

Celeron fanless nuc stripped down from ssd etc running audio Linux ramroot 

Core i3 fanless nuc running server core 2019 

 

As stated before I believe the simplicity and headless convenience of Linux in general is awesome but want to hear by myself the difference that windows does, aka better decay and imaging based on other users experience 

 

I'm getting new speakers soon so I need to hurry before this arrived 

 

I will post when can compare the different naa's 

 

[austinpop]

4 hours ago, Bricki said:

Audiolinux isolated CPU cores listening impressions

 

I have got the isolated CPU core function working on my roon server with 5200u CPU and 8 GB of RAM in ramroot. I also have the isolated CPU core function working on my NUC7CJYH running roon bridge with 4 GB of RAM in ramroot. The nuc is also network bridge to the server.

 

The isolated CPU core function brings a clearer sound. Background details in music are more easily discerned. Electric guitar sounds less glassy and digital and more natural and real. Individual sounds are not as flat and they have more of a 3 dimensional presence. In the live version of dire straits"you and your friend" the crowd at the beginning of the track creates a real concert atmosphere and individual crowd members can be heard in the distance. Before the CPU core isolation the crowd had less individual sounds and came across as more of a whole sound.

 

I am very pleased with the increase in SQ it has come as quite a surprise.

 

I highly recommend other NUC/AL owners give it a try 👍🔊😎

 

Thanks, Bricki!

 

Did you play with the various isolation options? And is what you described above with the "all but one" option?

[tboooe]

44 minutes ago, luisma said:

Allo USB Bridge naa 

Celeron fanless nuc stripped down from ssd etc running audio Linux ramroot 

Core i3 fanless nuc running server core 2019 

 

As stated before I believe the simplicity and headless convenience of Linux in general is awesome but want to hear by myself the difference that windows does, aka better decay and imaging based on other users experience 

 

I did a similar experiment earlier this year.  I compared the Allo against a Rasperberry Pi 3 and an i3 based fanless PC (running both Server 2012 R2 and Audiolinux).  In the end, the preferred the Server 2012 R2 based NAA because the music sounded fuller to me.  This is similar to my experience running HQP on Ubuntu versus Server 2012 R2.

[luisma]

8 minutes ago, tboooe said:

I did a similar experiment earlier this year.  I compared the Allo against a Rasperberry Pi 3 and an i3 based fanless PC (running both Server 2012 R2 and Audiolinux).  In the end, the preferred the Server 2012 R2 based NAA because the music sounded fuller to me.  This is similar to my experience running HQP on Ubuntu versus Server 2012 R2.

Wait, running HQP you mean the server not the naa sounds better in windows than Linux? 

[tboooe]

21 minutes ago, luisma said:

Wait, running HQP you mean the server not the naa sounds better in windows than Linux? 

Sorry for not being clear.  I did a separate test on my server PC which runs HQP with both Ubuntu and Server 2012.  In the end  I preferred Server 2012 though I admit the differences were small.  I have been meaning to revisit this test and try Audiolinux as well but have not gotten around to it.  

 

For the NAA test which was what I was referring to in my original post, I also preferred a Windows based device.  The differences in this was more obvious to me with a fuller sound.  For whatever reason, the Raspberry Pi and Allo  sounded thin.  Even the i3 fanless sounded better running Server 2012 than with Audiolinux.  I suspected it was because of the drivers and buffer settings I could adjust using Server 2012.