Overall Isolation - network, USB, and power

[fob69]

(sorry for bad writing above : The result is so good that i plan du put a DAC on battery supply too, with no preamp, so that i will only have my power AMP on AC network, and no loops at all).

 

 

Are you sure it makes sense to reclok before the isolation / have an isolation with no reclock after ?

(The isolation is on ES9018 card).

[YashN]

(sorry for bad writing above : The result is so good that i plan du put a DAC on battery supply too, with no preamp, so that i will only have my power AMP on AC network, and no loops at all).

Are you sure it makes sense to reclok before the isolation / have an isolation with no reclock after ?

(The isolation is on ES9018 card).

By all means, re-clock, near the DAC, have the highest quality one possible, and give it clean power.

 

That necessarily means you've isolated from the server computer prior.

[fob69]

In this logic maybe it would be better to use into for isolation, then regen for clean power and reclocking and no isolation on the dac board ?

 

 

Sent from my iPhone using Computer Audiophile

[fob69]

(to use Intona)

 

 

Sent from my iPhone using Computer Audiophile

[austinpop]

Well, it's Nov 18th, and my LPS-1 is "out for delivery," so I am hoping by the time I get home tonight, it will be waiting for me. Thanks, Alex, for adhering so well to your promised time line.

 

My default plan is to replace the el Cheapo Breeze LPS powering my RUR with the LPS-1 and compare. I will definitely do so, and report back.

 

However, I am contemplating another experiment with the LPS-1:

RUR (powered by Breeze) > Vbus2 > Lightning 2G USB data ------- > Ayre Codex DAC

↑

LPS-1 >

Lightning 2G

USB power ----

 

Sadly, unlike the mRendu, the RUR needs 6-9V, so I cannot use a Y-cable to share the LPS-1 with the RUR and the DAC.

 

Do people want to start placing friendly wagers on what I'll hear?

 

I just posted my findings over on the LPS-1 impressions thread.

 

http://www.computeraudiophile.com/f27-uptone-audio-sponsored/ultracap%99-linear-power-supply-1-listening-impressions-thread-30172/index9.html#post606001

 

@Cornan wins the prize for correctly guessing the winning configuration!

[scan80269]

Thanks for sharing your latest findings on that thread, austinpop!

 

Looking at your latest setup diagram, there may be one last leakage loop left, involving the downstream FMC, Aries Mini and the Auralic LPS and Teralink LPS. Replacing the Teralink LPS with an UpTone LPS-1 should break that loop. I have no idea how significant this leakage loop may be in terms of SQ impact, though.

 

For my Aries Femto I have recently placed an order for the Vinnie Rossi MINI PURE-DC-4EVR power supply. My hope is that it can do the same magic for Aries Femto as the UpTone LPS-1 does for the microRendu. I hope the PURE-DC-4EVR supply will show up in time to support the next audition session involving Aries Femto and microRendu.

 

Unfortunately the PURE-DC-4EVR will not work with Aries Mini, as even 12V is not high enough for the Mini. Vinnie's website indicates another version supporting 15V DC and higher is in development to be available later.

 

There seems to be increasing evidence that streamers like Aries can benefit from the trifecta of isolation for USB, Ethernet and power, which has the effect of making our digital audio systems more complicated (and costly)...

[austinpop]

Thanks for sharing your latest findings on that thread, austinpop!

 

Looking at your latest setup diagram, there may be one last leakage loop left, involving the downstream FMC, Aries Mini and the Auralic LPS and Teralink LPS. Replacing the Teralink LPS with an UpTone LPS-1 should break that loop. I have no idea how significant this leakage loop may be in terms of SQ impact, though.

 

For my Aries Femto I have recently placed an order for the Vinnie Rossi MINI PURE-DC-4EVR power supply. My hope is that it can do the same magic for Aries Femto as the UpTone LPS-1 does for the microRendu. I hope the PURE-DC-4EVR supply will show up in time to support the next audition session involving Aries Femto and microRendu.

 

Unfortunately the PURE-DC-4EVR will not work with Aries Mini, as even 12V is not high enough for the Mini. Vinnie's website indicates another version supporting 15V DC and higher is in development to be available later.

 

There seems to be increasing evidence that streamers like Aries can benefit from the trifecta of isolation for USB, Ethernet and power, which has the effect of making our digital audio systems more complicated (and costly)...

 

Interesting thought. I will cerrtainly try my LPS-1 on the FMC and see how that goes.

 

Regarding cost and complexity: no argument from me. I do think manufacturers are taking note, so I think we will find that these tweaky boxes are a transient phase in the state of the art.

[d_elm]

I am powering a microRendu with a LPS-1 and would like to hear a second LPS-1 in my system powering a TP-LINK mc200cm instead of a battery or iFi iPower. Is there an LPS-1 owner living close to Calgary, Alberta that is also interested in experimenting ? Perhaps we can help each other. Please send a private e-mail if interested.

 

 

Sent from my iPad using Computer Audiophile

[austinpop]

I just posted my findings over on the LPS-1 impressions thread.

http://www.computeraudiophile.com/f27-uptone-audio-sponsored/ultracap%99-linear-power-supply-1-listening-impressions-thread-30172/index9.html#post606001

 

OK, so since I won't really get to try this until next week, let me start a running list of suggestions. Feel free to add. If you do, please retain the original list, so the list is cumulative. I won't promise to try all suggestions, but I am always open to the great ideas you guys have!

 

Rajiv's HW Assignment

1. Remove Intona and RUR. Use split cable to power the Codex from the LPS-1 ( @lmitche )
   
2. Use the LPS-1 to power the downstream FMC instead of the Teralink. ( @scan80269 )
   
3. Replace Meanwell with LPS for energizing supply. ( @YashN )
   
4. Plug the Meanwell AND the upstream FMC's power supply into an outlet on another AC circuit than your P5 ( @gstew )
   

 

Thanks!

[scan80269]

Hi Rajiv,

 

I think forcing Aries Ethernet speed down to 100Mbps from gigabit is worth a shot:

 

 

1. Remove Intona and RUR. Use split cable to power the Codex from the LPS-1 ( @lmitche )
   
2. Use the LPS-1 to power the downstream FMC instead of the Teralink. ( @scan80269 )
   
3. Replace Meanwell with LPS for energizing supply. ( @YashN )
   
4. Plug the Meanwell AND the upstream FMC's power supply into an outlet on another AC circuit than your P5 ( @gstew )
   
5. Reduce Ethernet speed to 100Mbps for Aries. ( @scan80269 )
   

 

I've been using a Netgear FS105v3 switch (just before Aries) to force the link speed down to 100Mbps, but since yesterday I've successfully tested a modified Ethernet cable with two of four twisted pairs insulated, to force two pieces of gigabit capable gear to talk at 100Mbps. See my most recent post in the "Auralic Aries & Sonore microRendu listening impressions" thread for more details.

 

I'm assuming your FMCs are capable of running at 100Mbps.

 

Happy Thanksgiving!

[gstew]

OK, so since I won't really get to try this until next week, let me start a running list of suggestions. Feel free to add. If you do, please retain the original list, so the list is cumulative. I won't promise to try all suggestions, but I am always open to the great ideas you guys have!

 

Rajiv's HW Assignment

<SNIP>

[*]Use the LPS-1 to power the downstream FMC instead of the Teralink. ( @scan80269 )

<SNIP>

Thanks!

 

I should mention that I have tried 4 LPS-1s in a configuration where I started at:

 

A. Initial setup

1. Downstream FMC powered from Teknet battery with Jameco 5v Linear Adapter charging and plugged into the same outlet as the P10
   
2. R-Pi player powered from a DIY K&K Linear PS also plugged into the same outlet as the P10
   
3. Kali Reclocker powered from a DIY Linear PS powered from the P10
   
4. Seriously modified Mamboberry R-Pi HAT DAC powered from another DIY Linear PS powered from the P10
   

 

Then installed 2 LPS-1s:

 

B. 1st LPS-1 setup

1. Downstream FMC powered from Teknet battery with Jameco 5v Linear Adapter charging and plugged into the same outlet as the P10
   
2. R-Pi player powered from a DIY K&K Linear PS also plugged into the same outlet as the P10
   
3. Kali Reclocker powered from an LPS-1
   
4. Seriously modified Mamboberry R-Pi HAT DAC powered from an LPS-1
   

 

Added the 3rd LPS-1

 

C. 2nd LPS-1 setup

1. Downstream FMC powered from Teknet battery with Jameco 5v Linear Adapter charging and plugged into the same outlet as the P10
   
2. R-Pi player powered from an LPS-1
   
3. Kali Reclocker powered from an LPS-1
   
4. Seriously modified Mamboberry R-Pi HAT DAC powered from an LPS-1
   

 

And finally the 4th LPS-1:

 

D. 3nd LPS-1 setup

1. Downstream FMC powered from an LPS-1
   
2. R-Pi player powered from an LPS-1
   
3. Kali Reclocker powered from an LPS-1
   
4. Seriously modified Mamboberry R-Pi HAT DAC powered from an LPS-1
   

 

Going from the 1st 'all linear PS' setup to the 2nd '2 LPS-1' setup produced a very significant positive difference. Most would call it huge!

 

Going from the 2nd to the '3 LPS-1' setup produced a much smaller difference that while was totally positive after some time listening to it, sounded a bit less dynamic and exciting at first.

 

Going from the 3rd to the '4 LPS-1' setup didn't make much difference in my setup and to my ear. A tiny bit cleaner, but not by much and definitely not justifying adding an LPS-1 just for the FMC.

 

Note that all LPS-1s were powered from separate DIY linear supplies plugged into a separate AC circuit in that room.

 

Your experience will likely be different because my setup is so different than anyone else's here, but I do wonder if powering the FMC from the sole LPS-1 in a system will make much of a difference.

 

Of course, maybe it is the addition of each LPS-1 that makes the difference and if I started with the FMC and worked forward, I'd get the big difference at first with diminishing returns later. Sorry, no time to try that now, need to solder/drill/assemble some more DIY projects (the ones I REALLY got LPS-1 #2 & #3 for)!

 

Greg in Mississippi

[jabbr]

In this logic maybe it would be better to use into for isolation, then regen for clean power and reclocking and no isolation on the dac board ?

 

 

Sent from my iPhone using Computer Audiophile

 

Isolate I2S after USB input then FIFO buffer then reclock. See @Iancanada on DIYAudio

[austinpop]

Hi Rajiv,

 

I think forcing Aries Ethernet speed down to 100Mbps from gigabit is worth a shot:

<snip>

I'm assuming your FMCs are capable of running at 100Mbps.

 

Happy Thanksgiving!

 

Thanks, @scan80269 - I had seen your results on the other thread.

 

But I may have a problem because the FMCs I have (TP-Link MC200CM) are only gigabit.

 

I might try it without isolation with an old 100Mb switch, to test 100Mb vs 1Gb.

[scan80269]

Thanks, @scan80269 - I had seen your results on the other thread.

 

But I may have a problem because the FMCs I have (TP-Link MC200CM) are only gigabit.

 

I might try it without isolation with an old 100Mb switch, to test 100Mb vs 1Gb.

Yeah, I also have a pair of TP-Link MC210CS FMCs that only do gigabit. The TrendNet TFC-1000MGA is a better FMC choice as it can handle 100Mbps on the copper side, and takes in 5V DC so should be easier to power with LPS-1 or battery.

 

Regarding the temporary removal of the optical Ethernet isolation, watch out for SQ loss that may offset the gain from 100Mbps replacing gigabit. It should be an informative experiment, though.

 

If you insert a Fast Ethernet switch to force 100Mbps, you may want to know that my friend and I heard a SQ difference between two revisions of Netgear FS105 switches (the latest FS105v3 delivered better sound), and replacing the included SMPS wall wart with a Teradak LPS resulted in a very audible SQ improvement. I'm increasingly motivated to use a modified Ethernet cable to force 100Mbps instead of relying on a powered switch like the FS105. Another listening session is needed to hear how effective the Ethernet cable trick is in improving Aries SQ.

[Cornan]

In case you have missed it. Here is a great recent post from John Swenson that might help you to choose wisely!

 

OK OK,

some details on powering more than one device off a single LPS-1.

 

You cannot harm the LPS-1 itself in any way. If you pull more current out of it than it is designed to handle it shuts down, the LED blinks red. It starts up again every 5 seconds to see if the load is still pulling too much current. I have done extensive listening tests at different current levels and have not heard any difference with different current loads. (hook up a microRendu and adjustable resistor, change the resistor to add additional current, no change in sound)

 

There are three issues with running more than one device off a single LPS-1:

 

1) current draw

2) isolation issues

3) noise modulation

 

#1, see the above, as long as the combination is less than what the LPS-1 is designed to handle (1A) there is no sonic degradation

 

#2 isolation issues, this is far more complicated than can be listed in one bullet point, go read the stuff I have written in various places here on CA for the background. If in doubt, draw a diagram of your system, both AC line and digital/audio interconnects and see where the leakage loops can happen. Put an LPS-1 in the appropriate place to block loops. If your system already has isolation in some places, powering both sides of an existing isolation with the same LPS-1 will negate that isolation. For example powering both the upstream and downstream ends of a fiber link with the same LPS-1 shorts out the isolation due to the fiber. That may or may not degrade SQ. You have to try and see (err hear). Powering the end of a fiber link connected to the Ethernet jack of the microRendu with the same LPS-1 as powers the microRendu does not cause a decrease in isolation, it may increase isolation depending on what the other power supply was and how the Ethernet is connected (shields connected etc), it may or may not change SQ for good or bad, depending on #3 below.

 

#3 noise modulation. Any time you have two devices on the same supply, changes in load current from one device can change the voltage on the shared supply, increasing the noise the other device sees. In practice you wind up with both devices changing the noise on the shared wires, so both devices see the noise from the combined load changes of the two devices. The LPS-1 has a very low output impedance which means load changes cause small changes in output voltage, but they are still there. Thus if you have one LPS-1 supplying an FMC and a microRendu, they are both going to be seeing greater noise than either by itself would see, but that still may wind up sounding better than another supply you might use to power the FMC.

 

So in summary, there are no hard and fast rules. You certainly CAN run multiple devices off a single LPS-1 as long as it can handle the combined current. In pretty much any case things will probably sound better if each device was powered by its own LPS-1, but if you don't have multiple LPS-1s go ahead and try putting two or more devices off a single LPS-1, it may sound better than the alternative, it may not, you won't harm anything by trying.

 

The above is assuming the multiple devices can all be run from the same voltage.

 

John S.

[fob69]

I am waiting for an es9018m2k kit for a supplied powered DAC.

 

My ess9018 nfb29 is on repair, and I have made some tests with my lifepo4 battery supplying a squeezebox touch with its iner dac (analogic output), wifi off, IR off, screen off, so with the lowest possible noise.

 

First, i never heard the poor sbt dac sound like this. Huge difference compare to my sbooster linear psu. Sbt seams to be very noise sensible.

 

Then i compared (with lifepo4 supply), the SQ with usb source connected to the sbt, with ethernet with passive isolation (EMO), and TPLink (and EMO after the TPlink).

 

Result is the SQ with usb (no network connection) is closed to ethernet (EMO) Very deep silences and basses, very detailled sound.

With tplink (powered by lifepo4 too), i lost the deep silence and basses, the sound is much more "digital" And i have "the" 2xwider soundstage.

I did not do the test without EMO but i am noy sure it is bery efficient.

 

I have the TP link soundstage difference with the linear too psu but not the "not digital" sound due to lifepo4 supply.

 

I thougth the wider soundstage with tp link was due to the noise reduction. I think now that this bigger soundstage is caused by NOISE from the tplink.

 

HF noise could cause phase problems or enhance hf that would create an artificial large sound stage ?

 

I had the same effect with Tripath class T amps two years ago.

 

I then open the tplink and check the supply.

There first a DC/DC switching converter with possibly an huge noisy output, followed by a CMOS regulator with 0,2mv output noise.

 

I conclude that TP link possibly does nothing that generating a "clean digital noise" upstream the server, with an artficial sound stage enhancement.

 

Only the way its iner supply is built makes sypplying it with lps1 a no sense.

 

(and i was wrong)

 

 

 

 

 

 

Sent from my iPhone using Computer Audiophile

[mmerrill99]

Yea, stage depth rather than stage width is a sign of lower noise - as you say width can often be the result of added noise.

[gstew]

<SNIP>

I then open the tplink and check the supply.

There first a DC/DC switching converter with possibly an huge noisy output, followed by a CMOS regulator with 0,2mv output noise.

 

I conclude that TP link possibly does nothing that generating a "clean digital noise" upstream the server, with an artficial sound stage enhancement.

 

Only the way its iner supply is built makes sypplying it with lps1 a no sense.

 

(and i was wrong)

 

<SNIP>

 

I have both Trendnet and TPLink FMCs along with several Baaske Ethernet filters. I did perceive an improvement with the FMCs powered by linear supplies and additional improvements adding the Baaske, but I have not tried the Baaske along.

 

I too spent some time opening the FMCs and found the same as you did... in another thread (http://www.computeraudiophile.com/f22-networking-networked-audio-and-streaming/optical-network-configurations-24641/index28.html#post582630) I said

 

"...the TPLinks use a 34063 switching step-up/step-down regulator as the main regulator for 3.3v... that's why it works on 5v-9v (spec for the device is 3v-40v!). That feeds some of the chips directly, but also feeds a couple of RT9166A linear regulators, one producing 2.5v and the other 1.2v. A handy DIY'er COULD LIKELY bypass the 34063 and feed linearly-produced 3.3v directly into the device for some additional power supply noise reduction. If the Trendnet device is the way to go for me, I'll report on how it's power setup is configured after I receive them."

 

Since writing this I've been inside of my Trendnet units too and found the same regulator.

 

At some point I'll DIY small regulators to bypass the 34063s & see what difference that makes.

 

Interesting results, can you point to or PM info on the configuration changes you did to your SBT? I have one and have NEVER used it, going forward with R-Pi's running Squeezelite. Now I'm curious how it will sound configured like yours and powered by an LPS-1.

 

TIA

 

Greg in Mississippi

[fob69]

I will by pass the inner switching tplink psu and give a feed back.

 

Cinsidering optical decoupling is the only way to guaranty a full isolation and that i never hear music at more than 192k, i am working on a simple toslink + fifo + reclock dac feed.

Does any one tryed this ?

 

 

Sent from my iPhone using Computer Audiophile

[scan80269]

I have both Trendnet and TPLink FMCs along with several Baaske Ethernet filters. I did perceive an improvement with the FMCs powered by linear supplies and additional improvements adding the Baaske, but I have not tried the Baaske along.

I too spent some time opening the FMCs and found the same as you did... in another thread (http://www.computeraudiophile.com/f22-networking-networked-audio-and-streaming/optical-network-configurations-24641/index28.html#post582630) I said

"...the TPLinks use a 34063 switching step-up/step-down regulator as the main regulator for 3.3v... that's why it works on 5v-9v (spec for the device is 3v-40v!). That feeds some of the chips directly, but also feeds a couple of RT9166A linear regulators, one producing 2.5v and the other 1.2v. A handy DIY'er COULD LIKELY bypass the 34063 and feed linearly-produced 3.3v directly into the device for some additional power supply noise reduction. If the Trendnet device is the way to go for me, I'll report on how it's power setup is configured after I receive them."

 

Since writing this I've been inside of my Trendnet units too and found the same regulator.

 

At some point I'll DIY small regulators to bypass the 34063s & see what difference that makes.

 

Interesting results, can you point to or PM info on the configuration changes you did to your SBT? I have one and have NEVER used it, going forward with R-Pi's running Squeezelite. Now I'm curious how it will sound configured like yours and powered by an LPS-1.

 

TIA

 

Greg in Mississippi

 

Greg, I'm very interested in how you plan to replace 34063 switching regulators in FMCs. I looked inside my TrendNet TFC-100MGA and found a GMT G5752A SOIC-8 switching regulator IC, which is apparently similar to 34063 in function. Since this G5752A appears to be the only switching regulator in this FMC, I'm interested in replacing it with a low-noise linear regulator.

 

Let me know which linear regulator model you select for your mods.

[fob69]

there are two small holes for 3,3V bus supply close to the electro capacitor (mark TP3V3 / TPGND2).

 

 

http://m.ebay.com/itm/TPS7A4700-Ultralow-noise-4-VRMS-LDO-linear-regulators-1-4V-20-5V-LM317-replace-/252533525965?hash=item3acc2bcdcd%3Ag%3A18wAAOSwgmJX1PKh&_trkparms=pageci%253A28c61375-b3b9-11e6-a07f-74dbd1804d9e%257Cparentrq%253A9fece22a1580a6a8d0d46acdffec4314%257Ciid%253A1

 

I have a lot of heat on dissipator. Id did not cut the pcb track and i think i have leaks througth the original regulator.

Real improvement but not neutral.

(i will go on passive isolation)

Sent from my iPhone using Computer Audiophile

[d_elm]

What about 3.3v in, perhaps from LPS-1, direct to board, bypassing the input regulator ?

 

 

Sent from my iPad using Computer Audiophile

[fob69]

Anyway there is still a problem because there are a least two level voltage : 3,3 and 2,5V.

I did supply the 3,3V bypassing the DC/DC switching regulator but there is still the other "0,4mv noise" regulator (RT9166) that produce the 2,5V... So a clean bypassing would need at least two linear regulators (or a 3,3V lifepo4 battery + a linear regulator for the 2,5V).

[d_elm]

Here is an image of a TP-LINK MC200CM board. I do not know much about this and am hoping others can examine the hi-res image and determine if the regulators used are acceptable for this device to be considered "low noise". The regulators used are RT9166A and BR B772D.

 

[scan80269]

Here is an image of a TP-LINK MC200CM board. I do not know much about this and am hoping others can examine the hi-res image and determine if the regulators used are acceptable for this device to be considered "low noise". The regulators used are RT9166A and BR B772D.

 

[ATTACH=CONFIG]31124[/ATTACH]

The B772D at Q3 is a bipolar PNP transistor that is part of a switching regulator circuit, with the SOIC-8 chip at U3 being the switching controller. In my MC-210CS, U3 is a 34063 switching controller. The RT9166A are low-dropout linear regulators downstream from the switcher, and generate 2.5V and 1.2V respectively.

 

RT9166A is not particularly low noise, but replacements may be tricky to find due to the very low dropout spec. If the switcher generates 3.3V (needs confirmation) then a replacement for RT9166A needs to have similarly low dropout voltage (and same pinout) or it will not work. The switcher can be replaced like fob69 did.

 

I looked inside my TrendNet TFC-1000MGA FMC, and found a G5752A switching controller and a LD1117 LDO regulator. It looks like these FMCs have only one switching regulator in their designs, along with one or more linear regulators.