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

[afrancois]

1 hour ago, Em2016 said:

 

Well this is the point I made earlier - it's not possible he's inspected/measured every linear PSU on the planet. But I assume he's commenting based on a variety of linear PSU's that he has inspected or measured.

 

As he also says, some people actually prefer a slightly brighter sound with different PSU's, thinking they are hearing more 'detail' - when in reality it's more RF getting through causing higher IM distortion, resulting in the brighter sound, mistaken for more 'detail'.

 

This is the tricky thing when our own preferences vary so much. But he does have access to measurement gear that we don't...

Exactly what I have experienced. My system sounded much darker when I had replaced most SMPSes with LPS. It also revealed other shortcomings in my system. 

[austinpop]

As I read over my post last night, I realized I had overstated the improvement, at least in the context of the whole system. I wrote:

 

10 hours ago, austinpop said:

Wow - much improved!

 

Yes, there was an improvement, and in the local context of the upstream network outside of my listening room, it was much improved. But compared to other changes in the audio chain, this was a very modest improvement. See - I am plagued by this relative magnitude trap like anyone else! So let me restate last night's post in better context.

 

The lesson I learn over and over is that despite every attempt made by component designers and us hobbyists, no isolation scheme seems to be perfect. Even with galvanic isolation, line filters, port magnetics, JSGT shunts, the upstream still seems to matter! And for most of us, once you get outside the listening room, and if you have a big enough home network, the picture is pretty grim when it comes to noise-makers like SMPSes and USPes, to name just a couple.

 

My actual experience with upstream optimization has been that the better you do with upstream isolation (more sandbags in the dyke, so to speak), you lower the effect of these upstream noise makers. The magnitude of the nasties is driven down, and we hear this as improvement. But it doesn't go to zero.

 

So in this context, the removal of the UPS powering my network closet had a very modest overall effect, since it was already being attenuated by the existing isolations in place. What is startling is that any of it was getting through at all, and this goes back to my previous analogy of adding sandbags in the path. You can slow the flow to a trickle, but you can't completely stop the leak. Better to kill the noise upstream, than try to isolate it with 99% success!

 

 

9 hours ago, ElviaCaprice said:

 

That Tripplite still has a breaker installed.  Better is this 20A Tripplite that requires hard wiring.  It's what I use.   In fact I wired it directly from my 1KVA Topez direct using 10AWG 3 conductor wire. 

 

Interesting. Do you have personal experience with finding breakers in a power strip to make an SQ difference? I know in the context of my dedicated line, I asked Jim Weil if he had found breaker quality to affect SQ, and he had indicated not enough to worry about.

 

Regarding the strip I have, compared to the effect of replacing a UPS, I feel comfortable that the effect of the breaker in the strip, even if it's there, is minimal. Also, keep in mind that this is not the power feeding my main rig. It's in a different part of the house, on a completely different circuit, and upstream of all the existing isolations. 

 

I would say that this far upstream, heroic optimizations are unnecessary (due to the high cost). Just good, basic hygiene is my goal for the moment. 

 

4 hours ago, afrancois said:

Good you came to the same conclusion regarding SMPS'es I wrote about many posts ago.

 

 

Yes, completely agree. In my defense, I never disputed that this would matter, and in fact had already done a first round of optimizations removing some, though not all, SMPSes. Of course, neglecting the UPS was an oversight, which is what my current report describes.

 

3 hours ago, the_doc735 said:

quote:

"Many have privately approached me (paul hynes) with questions about how best to improve SQ and then stop me once I start talking about power supplies. Their response typically goes like this: "You don't have to convince me of the importance of good power, I already know this because I own an HDPlex." These folks have no idea. No offense to HDPlex but the model I own is nowhere in the same vicinity as the performance of my SR7. I also have a Teradak, Kenneth Lau, Paul Pang, iFi, and various DIY battery supplies that are now collecting dust."

https://www.computeraudiophile.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/?page=30

 

I'm sure you know I'm a big fan of Paul's PSUs, and I even have a monster SR7MR3DRXL in his queue. But SR7 rails are expensive and precious, and to apply them to this application (upstream network gear) falls into the category of unnecessarily heroic optimizations!  

[austinpop]

3 hours ago, paulhynes said:

As Rob says, some people like the resulting bright airy sound of noise floor modulation. It is not something I can live with and I much prefer to hear the music unadulterated.

 

I couldn't agree with you more, Paul.

 

But the mystery here is that Rob seems to be attributing the noise floor modulation (and resulting brightness) to LPSes - like yours - whereas most of us here have had the opposite experience. It's the SMPSes that cause the bright, fatiguing sound, and once you replace with an SR4 or SR7DR, the resulting sound is so much more relaxing, effortless, and open.

 

I don't understand this.

[auricgoldfinger]

41 minutes ago, austinpop said:

Of course, neglecting the UPS was an oversight, which is what my current report describes.

 

Have you sacrificed the 'uninterruptible' feature of your network? 

[ElviaCaprice]

50 minutes ago, austinpop said:

Interesting. Do you have personal experience with finding breakers in a power strip to make an SQ difference? I know in the context of my dedicated line, I asked Jim Weil if he had found breaker quality to affect SQ, and he had indicated not enough to worry about.

 

Regarding the strip I have, compared to the effect of replacing a UPS, I feel comfortable that the effect of the breaker in the strip, even if it's there, is minimal. Also, keep in mind that this is not the power feeding my main rig. It's in a different part of the house, on a completely different circuit, and upstream of all the existing isolations. 

 

I would say that this far upstream, heroic optimizations are unnecessary (due to the high cost). Just good, basic hygiene is my goal for the moment. 

 

 

I do not have any direct comparison.  But since you we're referring to JS's previous recommendations, not having any breakers/LED's was also mentioned by him as best power strip for use.

[austinpop]

Just now, auricgoldfinger said:

 

Have you sacrificed the 'uninterruptible' feature of your network? 

 

Yes. With good LTE ubiquitous, I can tolerate the loss of my network in a power outage. In any case, that was a false security, because even if I kept my network gear up, it was often the case the power outage took out an upstream node, so continuous internet was never guaranteed anyway.

 

It does mean my NAS is now subject to sudden power failure if it occurs. I'm assessing how big a risk that is, given I'm using btrfs as the filesystem, and SHR (Synology Hybrid RAID). My NAS doesn't contain critical data that isn't backed up elsewhere, and it's access pattern is read-mostly. So my current working assumption is that the risk of a power failure happening in the middle of an uncommitted write I/O operation is quite small.

[austinpop]

8 minutes ago, ElviaCaprice said:

 

I do not have any direct comparison.  But since you we're referring to JS's previous recommendations, not having any breakers/LED's was also mentioned by him as best power strip for use.

 

Ah, I had missed that.

 

Well, this may be something I look at in my next round of improvements. It never ends!

[paulhynes]

49 minutes ago, austinpop said:

 

I couldn't agree with you more, Paul.

 

But the mystery here is that Rob seems to be attributing the noise floor modulation (and resulting brightness) to LPSes - like yours - whereas most of us here have had the opposite experience. It's the SMPSes that cause the bright, fatiguing sound, and once you replace with an SR4 or SR7DR, the resulting sound is so much more relaxing, effortless, and open.

 

I don't understand this.

 

Linear power supplies are not exempt from being the cause of bright fatiguing sound. Signal processing currents in your audio equipment also flow around the power supply so anything that disrupts the current flow within the power supply is going to imprint itself on the signal is some way. Both types of power supply can have aberrations that can imprint on the current flow but the SMPS also has the added problem of dealing with the high frequencies used for timing switching. Filtering attempts can give mixed results as they often shunt noise pollution into other areas of sensitivity. Some imprints are relatively benign but others are highly fatiguing.

[asdf1000]

5 hours ago, austinpop said:

But the mystery here is that Rob seems to be attributing the noise floor modulation (and resulting brightness) to LPSes - like yours - whereas most of us here have had the opposite experience. It's the SMPSes that cause the bright, fatiguing sound, and once you replace with an SR4 or SR7DR, the resulting sound is so much more relaxing, effortless, and open.

 

I don't understand this.

 

This isn’t a company sponsored forum thread so I wonder if @Rob Watts could chime in.

 

As I mentioned a few times, it’s not possible that he’s seen and measured every single linear PSU on the planet but I assume he based his comments on a variety he has seen (we don’t know which..)

 

Edit: I see paul explains things well above, i.e. some linear PSU’s can perform poorly and some well. i.e. like everything, they should be considered case by case

[asdf1000]

@marce , another qualified person, also said something very similar to Rob Watts, just last week:

 

"LPS's are not the panacea for all evils), slapping a LPS on the front end does not guarantee better quality in fact often it is worse because most noise is in the audio frequency range... But it has become a Audiophile belief/mantra use a LPS!""

 

https://www.computeraudiophile.com/forums/topic/54151-design-criteria-in-dacs/?do=findComment&comment=875816

 

Another example from a qualified expert, that each PSU needs to be looked at case by case, rather than treating all linear PSU's holistically... 

 

Of course, it always goes without saying, that if your ears are happy then that's all that matters. And Rob Watts says the same in the quote I shared above. But it can be technically worse performing.

[seeteeyou]

13 hours ago, Em2016 said:

Lol good questions. I guess it may possible that he relies on the (state of the art?) measured performance of his gear, as a starting point? I don't know though.

 

https://www.stereophile.com/content/chord-electronics-dave-da-processor-measurements

 

Those measurements really didn't tell us anything about the performance of that SMPS inside DAVE whatsoever unless they're "brave" enough to swap it with a worthy LPS (definitely not el cheapo ones with laughable PSRR and impedance etc.) and measure again, we've gotta be able to eliminate the confounding factors or else we can't really single out what the PSU itself is contributing to the entire system.

 

And then he simply dodged this question when I was trying to learn more about Hugo M Scaler's bundled SMPS:

 

https://www.head-fi.org/threads/chord-electronics-blu-mk-2-the-official-thread.831343/page-267#post-14374183

Quote

How about this particular concern regarding negatively impacting other gear? Is the bundled switching PSU theoretically as well as measurably having no negative impact whatsoever?

 

Actually even he admitted that SMPS was doing more harm than good, we could certainly measure anything we want but in the end it's still a matter of listening to the results after the damage is done

 

https://www.head-fi.org/threads/chord-electronics-blu-mk-2-the-official-thread.831343/page-196#post-14103671

Quote

While neither Rob, Jay nor myself agreed on everything we heard during our listening sessions, we were all in agreement here and it was Rob's assessment that DAVE's PSU was likely negatively impacting my other gear.

 

[asdf1000]

31 minutes ago, seeteeyou said:

we've gotta be able to eliminate the confounding factors or else we can't really single out what the PSU itself is contributing to the entire system.

 

I agree. Even marce (in that link I posted above) says:

 

"This is one situation where I think how it sounds would be a better way of determining whether you like a DAC or don't..."

 

With so many variables in a system of components in a home setup, it can be hard to technically predict what interactions will affect what.

 

33 minutes ago, seeteeyou said:

Actually even he admitted that SMPS was doing more harm than good, we could certainly measure anything we want but in the end it's still a matter of listening to the results after the damage is done

 

Interesting. I hadn't see that one. Thanks. Even more interesting would be to know why Rob's guessed Dave's SMPS was negatively affecting the system performance (i.e. what technical mechanisms did he think/guess were at play) but I guess we may never know.

[seeteeyou]

While Intel announced that it's delaying its 10nm ramp until late 2019, their Cannon Lake with i3-8121U @ 15W TDP will be available soon

 

https://ark.intel.com/products/126145/Intel-NUC-8-Home-a-Mini-PC-with-Windows-10---NUC8i3CYSM

https://www.intel.com/content/dam/www/public/us/en/documents/product-briefs/nuc8i3cysm-nuc8i3cysn-product-brief.pdf

 

Block diagram of Intel's latest NUC models NUC8i3CYSM and NUC8i3CYSN here

 

https://www.intel.com/content/dam/support/us/en/documents/mini-pcs/mini-pcs/NUC8i3CYS_TechProdSpec.pdf#page=18

https://ark.intel.com/products/126145/Intel-NUC-8-Home-a-Mini-PC-with-Windows-10---NUC8i3CYSM

https://ark.intel.com/products/128503/Intel-NUC-8-Home-a-Mini-PC-with-Windows-10---NUC8i3CYSN

 

$579 with 8GB of soldered-down RAM (no good for AudioLinux) and 1TB HDD, stuff like AudioPhile Linux / Snakeoil OS / wtfplay are OK since they're so slim

 

https://www.bhphotovideo.com/c/product/1428649-REG/intel_mini_pc_new_boxnuc8i3cysm1.html

 

No fanless cases from Akasa yet, most likely i3-8121U @ 15W TDP should work fine without a cooling fan.

[austinpop]

28 minutes ago, Em2016 said:

 

I agree. Even marce (in that link I posted above) says:

 

"This is one situation where I think how it sounds would be a better way of determining whether you like a DAC or don't..."

 

With so many variables in a system of components in a home setup, it can be hard to technically predict what interactions will affect what.

 

This is where it all comes full circle, especially given the theme of this thread. Ultimately, it's about using your ears to determine the aggregate effect of all these competing factors.

 

Rob has strong opinions, and has earned the right to hold them. But even with that, you'll notice that he ultimately uses his ears to voice his products.

 

28 minutes ago, Em2016 said:

Thanks. Even more interesting would be to know why Rob's guessed Dave's SMPS was negatively affecting the system performance (i.e. what technical mechanisms did he think/guess were at play) but I guess we may never know.

That example was from listening tests during Rob's visit to Roy's house. Roy's full quote might explain it better:

 

Quote

A few hours later, Rob wanted to test the specific impact of this RF mains conditioner on both DAVE and a prototype that he brought along (that was also powered by a switching PSU). To our collective surprise, we found that our overall system's SQ improved when both DAVE and Rob's prototype were powered separate from the other components (the Zenith SE, SR7 and Pass Labs amp). As good as this RF mains conditioner is, SQ is better with either DAVE or Rob's prototype isolated from the other components. While neither Rob, Jay nor myself agreed on everything we heard during our listening sessions, we were all in agreement here and it was Rob's assessment that DAVE's PSU was likely negatively impacting my other gear.

 

[asdf1000]

2 minutes ago, austinpop said:

This is where it all comes full circle, especially given the theme of this thread. Ultimately, it's about using your ears to determine the aggregate effect of all these competing factors.

Yes, as I said, this should always go without saying. But it's always a good idea to repeat every now and then.

 

[asdf1000]

6 minutes ago, austinpop said:

That example was from listening tests during Rob's visit to Roy's house. Roy's full quote might explain it better:

 

Thanks but the part I'm interested in is the end:

 

"it was Rob's assessment that DAVE's PSU was likely negatively impacting my other gear."

 

As I mentioned above, I wonder what was going through Rob's mind as he was trying to guess what technical mechanisms were at play, effecting SQ. We may never know. Or maybe I should just ask him ?

 

[austinpop]

1 minute ago, Em2016 said:

As I mentioned above, I wonder what was going through Rob's mind as he was trying to guess what technical mechanisms were at play, effecting SQ. We may never know. Or maybe I should just ask him ?

 

And if you do, please report back on what you learn.

[beautiful music]

6 hours ago, austinpop said:

 

And if you do, please report back on what you learn.

+1

 

[Johnseye]

On 9/15/2018 at 11:42 PM, austinpop said:

Network tuning update

 

A few weeks ago, I wrote my impressions of the SOtM switch. In that review, I again referred to a recurring theme in my setup, where I said:

 

 

and

 

 

Well, I've been scratching my head at this to understand the source of this harshness. Was this an inherent characteristic of the Zenith SE's ethernet interface, or could there be more going on? After auditing my entire network chain, I realized the answer was under my very nose. And it's a bit embarrassing. To understand, here's a picture of my network topology:

 

 

And for completeness, here's the audio topology:

 

You'll notice that everything up the chain to the cable modem is on an LPS, but what the picture does not show is the AC power source. This is the embarrassing bit. For historical reasons I won't go into, my network closet (top picture) is powered by an APC 350VA UPS. In this latest audit of my system, it finally registered on me (duh!) that this UPS must be generating very noisy AC to the HDPlex LPS. Following @JohnSwenson's advice in the isolation transformer thread, I got myself an unfiltered Tripplite power strip (this one can be folded into convenient shapes).

 

Wow - much improved! I can't believe the crud the UPS was putting into the network gear made it across the entire topology. I've only done an evening's worth of listening, but at this point, the gap between network playback and local playback on the SE has shrunk. Not completely disappeared, but definitely smaller.

 

Encouraged by this, I have a plan to continue cleanup of my network. Referring back to my network topology, here's the plan:

1. I still have SMPSes in my network closet for things like the NAS, assorted adapters (like my Directv broadband adapter, a broadband alarm adapter, etc). I plan to replace all these SMPSes with an HDPlex 200W LPS. This still does not eliminate SMPSes for the ASUS WAP and a couple more switches. These are in other parts of the house but wire back to the network closet. I will address these with isolation. See next 2 points.
2. I plan to replace my DGS2205 switch with a JS approved Netgear GS108, powered by an existing HDPlex rail, and shunted with a JSGT.
3. The Archer router will only have a direct connection from my listening room on one port, and the Netgear GS108 switch on another.
4. All other equipment: NAS, adapters, other switches, the ASUS, will all connect to the GS108. Hopefully, this will block the bulk of both low-impedance SMPS noise (via the port magnetics) and the high-impedance noise, via the JSGT, from my audio topology. Don't forget, my TLS switch is also ground shunted, but I don't know if the underlying Dlink switch, on which it is modded, has isolating port magnetics.

Once I've done all that, I'll do another round of listening tests, and report back.

 

Wouldn't an isolation transformer like the Topaz filter that upstream noise?  Or are you saying the noise is coming through over the ethernet, bypassing an IT?

[austinpop]

32 minutes ago, Johnseye said:

 

Wouldn't an isolation transformer like the Topaz filter that upstream noise?  Or are you saying the noise is coming through over the ethernet, bypassing an IT?

 

Exactly. That’s the working assumption.

[look&listen]

46 minutes ago, austinpop said:

1 hour ago, Johnseye said:

1) Wouldn't an isolation transformer like the Topaz filter that upstream noise? 

 

2) Or are you saying the noise is coming through over the ethernet, bypassing an IT?

 

Exactly. That’s the working assumption.

Which one is assumed?

[Johnseye]

4 minutes ago, look&listen said:

Which one is assumed?

 

I believe both are assumed.

[austinpop]

Yeah sorry about the short cryptic post - I was on a mobile device.

 

Since the problem I'm trying to affect is the disparity between local playback from SSD and remote playback from the network, I am less suspicious of the AC line, for 2 reasons.

1. If it were the AC, it would affect both paths equally
2. my AC path is already pretty highly optimized, with a dedicated circuit, and a PS Audio P5 regenerator.

So the working assumption is that "noise" is coming over the ethernet. This is based on @JohnSwenson's findings that were posted here on CA.

 

Summarizing in a couple of sentences - SMPS noise is presented to a switch from 2 places: it's own SMPS, and the SMPSes on all upstream gear connected to the ports of the switch. This noise can be "stopped" from propagating (to your audio system connected to the switch) by 2 methods:

1. by shunting (connecting DC negative to AC ground) the PSU input
2. by using switches with known "good" port magnetics that block upstream SMPS noise from propagating downstream.

To this, I added my third: ensuring at least basic AC hygiene on the network - i.e. no UPSes.

 

I use "stopped" in quotes, because as with all other isolation in other spheres of digital audio, despite manufacturers' earnest and diligent efforts, these mechanisms don't appear to be 100% effective. This is why I am pursuing the fourth approach: by using LPSes upstream where possible.

 

Yes - I realize that LPS vs SMPS is a fraught topic, and there are bad LPSes and good SMPSes. But looking at the cheapo SMPS warts driving my cheap routers/switches/NAS, I feel comfortable that a good quality LPS (even if not a Paul Hynes) should be better.

 

I know HDPlex gets a bad rap here, and I know there are better supplies to be used on audio critical gear. But for that far upstream, I don't think the LPS needs to be heroic - the term I used a few posts ago. Just my opinion. 

 

I will report back here once I have tried some more experiments.

 

 

[lmitche]

1 hour ago, austinpop said:

Yeah sorry about the short cryptic post - I was on a mobile device.

 

Since the problem I'm trying to affect is the disparity between local playback from SSD and remote playback from the network, I am less suspicious of the AC line, for 2 reasons.

1. If it were the AC, it would affect both paths equally
2. my AC path is already pretty highly optimized, with a dedicated circuit, and a PS Audio P5 regenerator.

So the working assumption is that "noise" is coming over the ethernet. This is based on @JohnSwenson's findings that were posted here on CA.

 

Summarizing in a couple of sentences - SMPS noise is presented to a switch from 2 places: it's own SMPS, and the SMPSes on all upstream gear connected to the ports of the switch. This noise can be "stopped" from propagating (to your audio system connected to the switch) by 2 methods:

1. by shunting (connecting DC negative to AC ground) the PSU input
2. by using switches with known "good" port magnetics that block upstream SMPS noise from propagating downstream.

To this, I added my third: ensuring at least basic AC hygiene on the network - i.e. no UPSes.

 

I use "stopped" in quotes, because as with all other isolation in other spheres of digital audio, despite manufacturers' earnest and diligent efforts, these mechanisms don't appear to be 100% effective. This is why I am pursuing the fourth approach: by using LPSes upstream where possible.

 

Yes - I realize that LPS vs SMPS is a fraught topic, and there are bad LPSes and good SMPSes. But looking at the cheapo SMPS warts driving my cheap routers/switches/NAS, I feel comfortable that a good quality LPS (even if not a Paul Hynes) should be better.

 

I know HDPlex gets a bad rap here, and I know there are better supplies to be used on audio critical gear. But for that far upstream, I don't think the LPS needs to be heroic - the term I used a few posts ago. Just my opinion. 

 

I will report back here once I have tried some more experiments.

1. 
    
2.  

 

 

Rajiv,

 

If you have a physical connection between your cable modem and music system, on a copper cable system there is a ton of noise carried into the house from the street. Carrying RF signals to your house is what the cable system is designed to do.

 

If you have a fiber based cable system, I can't help you as I have no experience here. 

 

So if copper based, you need to find a way to isolate your music server, either with fiber, or a wireless link. I like wireless best as you only need one access point (WA)P to pull this off. Use a WAP and a spare wireless router for a dedicated subnet where the second subnet is created in the new router.

[austinpop]

5 minutes ago, lmitche said:

Rajiv,

 

If you have a physical connection between your cable modem and music system, on a copper cable system there is a ton of noise carried into the house from the street. Carrying RF signals to your house is what the cable system is designed to do.

 

If you have a fiber based cable system, I can't help you as I have no experience here. 

 

So if copper based, you need to find a way to isolate your music server, either with fiber, or a wireless link. I like wireless best as you only need one access point (WA)P to pull this off. Use a WAP and a spare wireless router for a dedicated subnet where the second subnet is created in the new router.

 

Been there, done that Larry.

 

Back in the Stone Age, when this thread began, I had FMCs fronting my audio gear, but I actually found bridging to sound better. Wireless bridging has its own issues, and had intermittent dropouts at higher resolutions. Since I plan to stream Qobuz, which is streams as high as 24/192, I'm not looking forward to the dropouts.

 

Fair point on the ingress CATV line. I do have a DC blocker on ingress, but not sure if there is an appropriate filter to attenuate noise. Since the cable line is carrying legitimate data at GHz frequencies, not sure what specs to use.

 

But to your point, it sounds like shunting the cable modem, in addition to powering with an LPS may be another safeguard.