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

[Dev]

2 hours ago, kennyb123 said:

I received my second USPCB tonight.  So I am now listening with no USB cable in the USB path.  In other words:  microRendu > USPCB > ISO REGEN > USPCB > Hugo TT.

It's hot tonight - 81 degrees in the listening room - so two fans are blowing.  I will offer initial impressions nonetheless:  knowing what I know now, there's no chance I'll ever want to throw the Audience Au24 USB cable back in.  I've been floored - particularly in light of the cost of the USPCB ($35).  Ridiculous.  

 

I'll share additional findings later tonight after it cools down.

 

That's a great endorsement for the USPCB and makes me want to get one right away. The Audience AU24 USB is an excellent cable which I also use out of trying at least 5 or 6 high priced ones. I am trying to figure if I need a straight or right angled one. Please share more info when you have some time (and off coarse after it cools down :-)).

[Dev]

1 hour ago, austinpop said:

 

ISO-Regen meets Ultra Trifecta - First Impressions

 

With about 70 hours under the ISO-Regen's belt, I decided it was safe to try a first listen. This time, Eric and I met at my place using my headphone setup (see sig) as the reference.

 

We will do another round in a couple weeks when the ISO-Regen should have at least 200 hours of burn-in.

 

Note: 

• All components were powered by LPS-1 PSUs.
• The sCLK-EX modded Zyxel switch (aka Z-switch) and modded sMS-200 were powered by a single LPS-1 using a Ghent Y-cable.
• Galvanic isolation was ON (switch set to position I) on the ISO-Regen in all cases except experiment 5 below.

Key Questions and Associated Experiments

1. Compare mR+ISO-Regen vs Ultra Trifecta - Does adding the ISO-Regen to the mR close the gap relative the SOtM Ultra trifecta?
   • Baseline:          Z-switch > sMS-200 mod > tX-USBultra (Ultra Trifecta)
   • Comparison:     mR > ISO-Regen
   • Result: 
     • Sorry folks, this is still no contest. The Ultra trifecta has significantly better SQ. It will be interesting to see how much further the Ultrarendu closes the gap. As I have maintained, the secret sauce in the trifecta has to be the modded switch. Once more such reclocking switches come along, there should be many more viable alternatives. But for now, the Trifecta rules.
        
2. Replace tX-USBultra with ISO-Regen
   • Baseline:          Z-switch > sMS-200 mod > tX-USBultra
   • Comparison:     Z-switch > sMS-200 mod > ISO-Regen
   • Result: 
     • Very close. WOW. Very, very close! We are so impressed with the ISO-Regen. 
     • Ultimately, we differed slightly in our overall preference.
       • Eric felt the ISO-Regen sounded better by a hair
       • Rajiv felt the tX was ahead by a nose
     • Minor differences aside, it points to the fact that the ISO-R is an insane value for the money.
        
3. Add ISO-Regen before tX in the trifecta
   • Baseline:            Z-switch > sMS-200 mod > ISO-Regen
   • Comparison:       Z-switch > sMS-200 mod > ISO-Regen > tX-USBultra
   • Result: 
     • The best sounding configuration of the day.
     • Again, we differed slightly in our overall preference.
       • Eric felt he could not hear the incremental improvement the tX added over the baseline
       • Rajiv felt the tX added a noticeable increase in SQ - notably low end heft and a tad more resolution
          
4. Add ISO-Regen after tX in the trifecta
   • Baseline:           Z-switch > sMS-200 mod > ISO-Regen > tX-USBultra
   • Comparison:      Z-switch > sMS-200 mod > tX-USBultra > ISO-Regen
   • Result: 
     • Not as good as before the tX
     • This sounded flatter, smaller, and a bit lacking in resolution and texture
        
5. Which sounds better - GI ON or OFF
   • Baseline:           Z-switch > sMS-200 mod > ISO-Regen (GI ON) > tX-USBultra
   • Comparison:      Z-switch > sMS-200 mod > ISO-Regen (GI OFF) > tX-USBultra
   • Result: 
     • This wasn't as cut and dried as I expected
     • Over several tracks, we concluded the GI setting of ON (default) sounded better, but the difference was small
     • On at least one track (Nightingale by Norah Jones on DSD), we felt the GI OFF sounded a tad more natural
   • Editorial: @Superdad - I already ragged you about the labels for the GI DIP switch in the ISO-R thread. Enough said.
      
6. Finally, we retried a couple of Ethernet experiments. 
   What is the SQ effect of iSO-Cat 6 Isolator
   • Baseline:           generic 6a > Z-switch > Supra 6a > sMS-200 mod > ISO-Regen > tX-USBultra
   • Comparison:      generic 6a > iSO-Cat6 > BJC 6a > Z-switch > Supra 6a > sMS-200 mod > ISO-Regen > tX-USBultra
   • Result: 
     • No impact. Nada. Zip. Zero.
     • We continue to hear no effect from these isolators - first on Eric's system, now mine.
     • Others do. Odd.
        
7. What is the SQ effect of dCBL-Cat7 cable
   • Baseline:           generic 6a > iSO-Cat6 > BJC 6a > Z-switch > Supra 6a > sMS-200 mod > ISO-Regen > tX-USBultra
   • Comparison:      generic 6a > iSO-Cat6 > BJC 6a > Z-switch > dCBL-7 > sMS-200 mod > ISO-Regen > tX-USBultra
   • Result: 
     • Definite improvement
     • Consistent with SQ improvement we heard on Eric's system
     • But magnitude of SQ improvement is not large enough for me to justify the cost of the cable ($500)
     • Might be worth picking up used.

Overall Musings

• The ISO-Regen is a gem. Why am I not surprised! My order for one has already been placed.
• The Trifecta with ISO-Regen (Z-switch > sMS-200 mod > ISO-Regen > tX-USBultra) is true endgame material. At some level I agree with @hols - the ISO-R adds color. But for me, it corrects the slightly thin SOtM house sound with the more mellow and rich Uptone house sound. The end result is magical. But expensive. No doubt. There's no free lunch.
• In decreasing order of performance, taking price into consideration:
  • $3190:    Z-switch > sMS-200 mod > ISO-Regen > tX-USBultra + 3 x LPS-1s
  • $2540     Z-switch > sMS-200 mod > tX-USBultra + 2 x LPS-1s
  • $2940     Z-switch > sMS-200ultra > ISO-Regen + 3 x LPS-1s
    • you can get by sharing an LPS-1 between the Z-switch and the ISO-R, but as has been pointed out, this defeats GI.
• Bottom line: The Ultra trifecta is still - for me - a  real sweet spot in terms of value for money. If you can swing a bit more, the Trifecta+ISO-Regen is just sublime. After that, you can slice and dice into subsets as your budget requires.

 

Thanks for the great report. Much appreciated. Keep'em coming.

 

Is the Z-switch mod same as the D-Link switch mod as listed in sotm ?

 

https://sotm-usa.com/collections/ultra-series-mods/products/sms-200ultra-audio-network-player-mods?variant=41281316812

[Dev]

58 minutes ago, austinpop said:

 

You're welcome.

 

Since I got this done in the early days, right after Roy, my Z-switch is a one-off Zyxel GS108b v3 that SOtM modded for me. This was before they started offering their own modded switch. Search back for my posts - I explained more.

 

Rajiv (I gather that's your name from other posts :-). Sorry I haven't been on this site for months now), thanks for your response and the time you are taking to do these experiments. They are truly appreciate. It opened up quiet some possibility on my end and I am considering replacing the mR with the sms-200Ultra w/ the Dlink switch mod to start with. Maybe add a ISO-R at a later point. 

[Dev]

40 minutes ago, lmitche said:

No I am not dealing with replacement clocks for the moment, or maybe ever.  Just trying to squeeze out the last drops of SQ using clean power, isolation and storage units with the recent innovations from our friends at Uptone Audio and Intel.  I had a big long USB audio chain at one point, but now I am of the mind that less is more, less clocks, less powered devices so less power supplies, and less, or zero cables.

 

The stock clocks in our devices suffer from poor power, and when that is cleaned up magic happens.  I also continue to be surprised at the impact of storage solutions on SQ.  What I am hearing now is jaw dropping!

 

 

I can't agree more on stressing the importance on clean power. Much of the stock devices can improve a lot by feeding them with good and stable power, without which they may be under-performing.

 

Can you elaborate on the recent innovation from Intel ? What are you referring to in terms of SQ improvements on the storage solution ? Something along the lines of M.2, SSD, Sata devices ?

[Dev]

6 minutes ago, lmitche said:

Here you go:

 

 

 

Cool. Thanks.

[Dev]

19 minutes ago, austinpop said:

 

Dev, 

 

You're welcome. 

 

 

This is a very sound approach to take incrementally. I think you'll be pleased.

 

I was thinking earlier that the UltraR might change the equation a little bit but then I read @romaz post where he claimed the sCLK-Ex board has much better clock than the Crystek. Though clocking is one aspect of good SQ but if I am not mistaken and read your post carefully, I think what you are saying is more importantly the synchronized clocks between the sms-200 and ethernet switch reduces timing and bit errors - "the secret sauce in the trifecta has to be the modded switch".

[Dev]

50 minutes ago, Johnseye said:

 

 

 

Additionally, in this post I quoted something from this thread I posted back in April about Optane and it possibly reducing memory access noise.

 

I failed to understand this part - the memory access is the first place the OS will seek, when it faults it loads pages from the secondary storage back to the main memory. Are you saying that the Optane is acting as the primary memory ? If so, then IMO, its a big hammer. Note, I have yet to read up on the Optane tech. Maybe we should take this up on the Optane thread instead of polluting this one... 

[Dev]

5 hours ago, Elberoth said:

Here is my SOtM sMS-200 setup:

 

- SOtM sMS-200

- Jcat USB Isolator

- Wyred 4 Sound Recovery

- 2x Uptone LPS-1

- Uptone JS-2

 

Sounds better than the $19.000 Aurender W-20 I have on hand. 

 

I think I will try Jcat Femto Ethernet card next.

 

 

Wow....

 

By any chance did you try the ISO-R replacing the JCAT USB Isolator and RUR ?

 

From the figure it appears that the JS-2 is powering both the LPS-1 ? How much do you think the JS-2 powering the LPS-1s contributes to the sound ?

[Dev]

If all your energizing supplies for the LPS-1 are behind an isolation transformer, does it really matter if they are smps or lps ?

[Dev]

1 hour ago, austinpop said:

 

 

So just to be clear, I am not yet sold on the need for LPSes for energizing. Until I hear, and can gauge the extent of further improvement, the jury is still out.

 

What I can report is that my approach - of putting the energizing SMPSes on their own small, cheap IT ($100), and modest audiophile power cables (3 x $50) gave me perhaps a 5-10% further boost in SQ. At a total cost of $250 over 3 LPS-1s, that's an additional $80ish per LPS-1, or a 20% price overhead.

 

Is that a good price-performance investment, and would I actually advise people to do it? I'm not sure. I also think it took my system to reach this level of SQ - a large leap from where it was the last time I tested power cables, for example, for these effects to even be noticeable.

 

So my 20 cents - since I do tend to be verbose  - for 95% of LPS-1 owners, this isn't even something to worry or care about. If you're nutty enough to try chains like mine, then well, yeah, by all means explore this too.

 

These are all subjective opinions off coarse. Somebody would be a fool to think that its going to improve the same level if they had exactly followed what you have. I think everyone understands the beauty of LPS-1 on the output side where it totally isolates but my question was rather on the input side of things - does it make the LPS-1 even a better power supply or does your system sound better because the energizing supplies are not able to kick back noises into the power line. As with anything in audio everything probably matters and an isolation transformer is really a good investment if you aren't willing to spring for a JS-2. As for me, I have dedicated 20A line for my main audio gears and all smps (computers, routers, switches, nas, cable modem, etc) are connected to the other normal line through an isolation transformer. So for me the chances of improvements of LPS are subtle at best and I don't think I am going to venture into it.

@SuperdadAlex brings out an interesting point where he suggests to put the main audio rig behind an isolation transformer which kind of makes more sense but you really need a high wattage one and need something that might be audiophile quality as well - which means you are also looking into non current limiting power cords and outlets (all my dedicated outlets are Furutech NCF) - well, that can be a project for some other time :-) 

 

[Dev]

8 hours ago, Superdad said:

 

 

b) When you see $0.50 bridge rectifiers in an LPS, you can know right away that these will exacerbate transformer ringing and lots of harmonics will be kicked back into the line. 

If you are looking for an LPS that is a little more "line quiet" (short of our JS-2 design), then at least find one with discrete Schottky diodes.

 

 

These are great points and something to keep at the back of your mind when shopping for cheapo LPS. Does anyone really make them and in prices alike the cheapo ones ?

 

 

[Dev]

For the SOtM Trifecta, I see two configurations

 

1. sms-200Ultra w/ Switch mod + tx-USBUltra

2. tx-USBUltra w/ sms-200 mod & Switch mod

 

For (2), it seems like the sCLK-Ex board in tx-USBUltra is able to use all the 4 clock points but in (1), some of the clock points are wasted in both tx-USBUltra and sms-200Ultra. Also configuration (1) costs more. Which configuration are folks using for the Trifecta ? and what is the difference in audio quality ? Does one sound better than the other ?

 

[Dev]

7 hours ago, austinpop said:

 

To my ears, 1 and 2 sound virtually identical. I highly recommend 2 when anyone asks me, because I consider it the real sweet spot in value for money. Without trying to be immodest, I believe I used the term trifecta first in this thread, and it was to describe this configuration.

 

Config 1 really only makes sense if you can fully utilize all 8 clock points you get from having 2 sCLK-EX boards.

 

Thanks! That's exactly my point as well. At the same time, I am a bit confused as to with option (2) why is the sCLK-Ex board installed in tx-USBUltra instead of sms-200 ? With (2), we are only talking about a single sCLK-Ex board and it would make much sense if the board resides inside the sms-200 chassis. The advantage would be that it only runs 2 x SMB connectors - one to the tx-USB and the other to the switch, instead of running 3 x SMB connectors if the sCLK-Ex were to be installed in the tx-USB, i.e. two to sms-200 and one to switch.

 

 

[Dev]

16 minutes ago, LTG2010 said:

Looking at this a bit closer, the designers have decided to use the atx connectors thus bypassing most of the motherboards regulators.

A 4 pin atx connector (this I presume links to the Celeron processor directly) is fed a clean 12V from the linear PSU. Then the 24 pin atx power connector is supplied by the Pico connector (again from a second rail from the PSU) the third rail to the Hard drive. Clever stuff.

http://www.supermicro.com/products/motherboard/celeron/x10/x10sba.cfm

 

The Supermicro manual says that 4-pin 12v is optional power source and doesn't specifies that it can power the CPU directly. Wonder if the Zenith folks have made any mods to it or to the pico.

[Dev]

23 minutes ago, austinpop said:

 

Hi Jean-Michel,

 

No doubt, one can achieve similar SQ gains going down the DIY path. Finding PSUs of the caliber of the SE may be the hardest to achieve. It will be interesting to see if people are able to source these from Sean or other vendors, without the wait times for the SR-7.

 

The power supply is critical but the optimized s/w that Innuous created is critical as well which this is very evident from your excellent comparison (and thanks for the detail thoughts which I and many other have been waiting for sometime). Is there anybody willing to do a group buy from Sean on a multi-rail psu  ?

[Dev]

4 minutes ago, sandyk said:

 

 You are likely to find, that just like Paul Hynes, Sean is not geared up to do group buys.

Building a PSU like in the photo, would take quite a while for a small company to make just one, unless they are contracted out perhaps ?

 

I don't know Sean's operation, so perhaps you are true but its definitely worth asking, if there is enough interest in this thread ?

[Dev]

7 hours ago, vortecjr said:

Understand that Andrew's music servers are designed primarily for streaming to an endpoint and dedicated to this task so he is not worried about server noise with network streaming.

 

I am not so sure if this is totally true. Over at Jplay, most has been using a dual jplay setup for a while (long before microRendu or the likes came out) - the control PC, which is equivalent to the music server, is also very important in contributing to the SQ as is the audio PC, which is equivalent of a streamer, like mR. If you have to optimize one, the one closer to the DAC must be done first but a noisy music server can also pollute the SQ. Thoughts ?

[Dev]

5 hours ago, lmitche said:

Everything matters:

 

Exactly. With digital audio, every component in the chain has influence to the SQ.

 

In fact, I am running two PC setup myself due to the need to upsample to DSD256 with ROCK running on both.  The Upsampling server is an i7 (@ 65w tdp) w/ ASUS mobo powered with Streacom ZF240 smps and the endpoint is Jetway low power mobo with JCAT femto USB card and powered with Mojo Illuminati. My library is stored in the music server. I haven't played much with power on the upsampling server yet but changes to the local music drive from spinning to SSD has impact on the SQ - this tells me that the music server is not immune to network audio. Since I run ROCK, I am currently inclined to change the music server to an i5 or i7 NUC and power it with LPS - the NUC consumes less power than a full fledged i5/i7 based mobo and can make +ive contribution to SQ. Any other ideas ?  

 

[Dev]

3 hours ago, Johnseye said:

Our most recent discoveries have shown us that SQ can be improved without the endpoint when specific components are used. 

 

By any chance, have you experimented with two identical server with those recipe (which eliminates the need of endpoint) and connected them through network - one acting as server and another acting as endpoint and found the SQ did not improve ?  

[Dev]

12 minutes ago, lmitche said:

I'd consolidate everything to your i7 and simplify your system. From there you can upgrade cabling, PSUs, add an ISO Regen . . . etc.

 

Originally I had everything in i7 with the JCAT card powered separately and rest powered through smps. Adding another low power Jetway based mobo (with excellent psu) made a significant improvement to SQ. As much as I would like a simple setup, from experience, it seems like managing a high power system is much more difficult and complex than breaking it down to two box solution - the biggest challenge is the power supply rails for powering mobo, cpu, ssd, usb card, etc. Obviously getting an SGM server is a solution but way way over my budget. I would like to hear what are folks doing, at least for the psu, when they need to upsample ?

 

 

[Dev]

I have recently loaned a tx-USBUltra from a friend (who is also using a sms200ultra) to see if I get any improvement in my streamer which has a JCAT Femoto USB card. Unfortunately, my experience has been very different than others - the tx-USBUltra degraded the SQ. I tried with two DACs and both has the same effect. This is also true for my friend who was testing it along with his sms200ultra. Though I am not using the sms200ultra in the chain, I always thought the tx-USBUltra would bring something +ive to the table and am really perplexed as this is very different from the trifecta experience folks have over here. Not sure if the tx-USBultra is defective or not.

 

[Dev]

4 minutes ago, vortecjr said:

If I recall correctly that is a UPnP based system where the audio PC is a renderer and actually playing the content. Anyway, not according to tests I have done here. In fact, in order to exaggerate the test I had a microRendu play from memory and then pulled the network cable. No measurable difference.   

 

Don't think so. You can use Roon running on both control and audio PC and use Jplay on audio PC in windows environment.

[Dev]

Anybody getting the Supermicro (one used in the Zenith) mobo modded from Sotm yet ?

[Dev]

1 hour ago, seeteeyou said:

 

Sometimes the SQ (or lack thereof) could very well depend on how we're powering tX-USBultra since sCLK-EX is super sensitive to its power source. Maybe you guys could also contact SOtM and see if they're able to give you any suggestions?

 

1 hour ago, austinpop said:

 

+1.

 

@Dev, what were you powering the tX-USBultra with?

 

We both (myself and my friend's setup) tried 4 different power supply with similar results.

 

1. Sonore Signature PSU (originally sold for microRendu)

2. Mojo Audio Illuminati v2 (best of the lot)

3. SoTm sps500

3. LPS-1

 

We also tried different power cords on these PSU fed from two different power conditioner - Audience aR6-TSSD and SR 10UEF powercell connected to 20A dedicated circuit with Furutech gtx-d(R) wall outlet.

 

[Dev]

2 hours ago, vortecjr said:

Sounds like you are using Mikes DAC? If you look at the post about performance you will get some insight on why this might be the case. I don't think it's noise when you are streaming over the network. There is no specific mention about sound quality since it his more to do with performance. Using an SSD in your case make sense because they are much faster compared to spinning drives. This is why Andrew at Small Green Computer uses them.   

 

Yes, I use Mike's DAC which is optimized for DSD256 playback. I also have Directstream which I rotate time to time. Good point on the performance part - I don't know if the improvements are coming from performance vs noise. If its due to performance, then the control PC could be powered with the worst psu and still the outcome will have the same result.