SMPS and grounding

[Superdad]

12 minutes ago, jamesg11 said:

Now that JS is closing shop on this epic investigation, for the time being I’ll at least assume, ... who is putting their hand up to write & post the summary/take-away?

 

Self-interest here, because I need to get someone more competent than me to implement, & they’re not going to wade all thru the extensive above.

 

I though @gstew did a really nice job summarizing and giving practical tips in the first half of this post:

https://www.computeraudiophile.com/forums/topic/37034-smps-and-grounding/?do=findComment&comment=735506

 

[Superdad]

2 minutes ago, sonata said:

 

Now that's clever Rob!  Very nice for those who don't mind soldering and using heat shrink.  Thanks for posting. 

[Superdad]

4 minutes ago, chetthejet said:

Thanks, Alex.  So, to confirm, the chain should go Meanwell-->shunt-->LPS-1-->IsoRegen-->Dac??

That is correct. 

[Superdad]

9 hours ago, kboung said:

Regarding combination of iso-regen powered by LPS-1

 

(1) For "tick/dropout" issue with iso-regen, it was recommended to ground downstream of iso-regen

(2) For high impedance leakage with LPS-1, it was recommended to ground Meanwell output negative

 

In this case, both groundings required or (1) will actually cover for (2)?

 

Maybe Alex or John can clarify. Thanks.

 

Hi Kboung:

 

If one is using an UltraCap LPS-1 to power the ISO REGEN, then no, grounding [the zero-volt "neg./grd"] of an SMPS that is "energizing"/charging the LPS-1 is NOT the same thing as grounding the downstream side of the ISO REGEN (the DAC's USB input).

 

This is because the LPS-1 is an isolated/"floated" PS (notwithstanding the newly discovered high impedance common-mode leakage which can creep across the unit's 100pF of capacitance between power domains), and its input and output "grounds" are not common to each other.

By the way, most people do not end up needing to ground the output side of their ISO REGEN.

 

Hope the above clears things up for you.

 

--Alex C.

[Superdad]

1 hour ago, lmitche said:

Oh boy, I guess we are having reading comprehension issues today, LOL!

 

21 minutes ago, Johnseye said:

 

Why wouldn't you answer the question instead of responding like that?

 

Chill guys.  I know it is confusing.  On my call with John last night I brought this up because it seemed at first glance that he was contradicting himself.

First off:

His comment about unshielded CAT cable radiating common-mode leakage was based on actual experience on his test bench where his specialized leakage measurement set up was picking up a large amount of stuff that he could see on his scope as he moved the UTP around.  

Yet of course a normal shielded CAT cable (6A) that has the shield tied at both ends (metal shel around the RJ45) is undesirable for our purposes because it reduces isolation by joining grounds between boxes and giving a path for both other leakage--and potentially noise.

 

The BlueJeans/Belden CAT6A does not tie its shields at all, but that shield may not be doing much for us.  It would be very effective if John's external shield wire loop trick was applied to that (external wire running between the shield ends).

 

But ALL the above is a moot point if one is using an Ethernet switch (such as the Netgear FS/GS units) that have been confirmed to shunt high-impedance leakage (from both an SMPS powering it and from the gear/cables feeding it).  

And that is where John's recommendation to put the Netgear switch at the end with your computers/router/etc.--and just run a single long cable to your hi-fi renderer.  Since the leakage will then not be on the cable (due to blocking by the Netgear), it then does not matter if the CAT cable is UTP or STP!

 

Hope that clears this up.  

[Superdad]

1 minute ago, thyname said:

Oh no!!!! Then we would be buying audiophile air....

 

Well I eat audio-nuts for breakfast so I pass audiophile gas.  It's not wifi, but I guess it is an ether-or situation...

[Superdad]

1 hour ago, austinpop said:

 

BoM and links, please!

 

There are a bunch of variations on Amazon and elsewhere--though they all should be about 1/8 the price:

https://www.amazon.com/Desco-Adapter-0-164-Diameter-09838/dp/B00HKJNF1K

https://www.amazon.com/GROUNDING-CORD-6-foot-3-prong-gator/dp/B00MNP1I3I

 

http://desco.descoindustries.com/DescoCatalog/Grounding-Hardware/Banana-Plug-Adaptors/09838/#.Wf00BzOZMSx

These take a banana plug. Test leads with banana plugs--and an alligator clip at other end are easy to find.

The above are USA-style pieces, so I had no luck finding them cheap on the Alibaba/AliExpress supermarket. 

 

What I'd like a source for are the particular spade that iFi uses on one of their GroundHog connectors:

It clips nicely to a 5.5mm DC barrel.  I am pretty sure they have it custom made.

 

[Superdad]

1 hour ago, magnuska said:

I have the Vinnie Rossi Lio integrated amp that is actually powered by a 24v SMPS.  I don`t know if this will benefit from grounding the DC output but I will try and see how it goes. The LIO is supposed to be immune to AC pollution since it got ultracapacitor PS with twin banks. 

 

1 hour ago, JohnSwenson said:

If it is the Lio with relays then yes it is probably immune to even the high impedance leakage.

 

It is also possible that the 24V SMPS brick chosen by Mr. Rossi already has continuity between its IEC-C14 inlet ground pin and its DC zero-volts ("ground") output.

 

If so it would simply be due to luck as we don't think anyone else has identified the issue of high-impedance leakage and its simple cure. It is just that some SMPS units happen to already be "shunted," while many others are not.  And you can't ask the manufacturers--they don't know or specify.  But it is of course trivial to check with a multimeter: if it beeps when you contact the above-mentioned pins (with the supply unplugged at both ends), then you are good to go.

[Superdad]

2 hours ago, barrows said:

@JohnSwenson,  How much leakage current are we talking about here?  Suppose I have a SMPS rated at 40 µA  (yes micro) of leakage current, at this rate I would presume that I would not need to worry about it at all?  This SMPS made positive subjective improvement when used to power the router which connects directly to my Rendu(s) with the output floating, maybe just because of reduced leakage current (of course this is a high end SMPS with low output noise characteristics as well).

 

 

John is getting tired of repeating himself (and he happens to be working simultaneously on tight deadline projects for both Sonore and UpTone this week), so allow me to explain and illustrate again:

 

For any SMPS you might have that is rated with leakage of 40µA--aside from that not specifying frequency--that rating absolutely is of the low-impedance leakage, not the high-impedance stuff that John is measuring.  All the "medical" SMPS units available--including the cheap ones--have leakage ratings of 50µA or less.  [75µA is the requirement for "medical" units as they are concerned with "touch-current" for human attached equipment, whereas standard "industrial" units may have leakage ratings about 10x that--0.75mA.]

 

But as John explained [in several of his very long posts on the subject in this thread], and as I illustrated with some of his measurement graphs, the high-impedance leakage is very much a separate component--one that can sail right on through and even show up in the single-ended output of DACs.  As also explained, it is relatively easy to shunt the high-impedance leakage [back to ground] so that it is not a bother.

 

Quote

I have another SMPS (20W, 9V) rated for .78 mA leakage current, would that be enough to cause concern?  Maybe I'll build this one up into a Rendu supply, and apply a switch to ground (or not) the output for testing purposes...

 

So that 0.78mA leakage-rated SMPS you have is likely just like the Mean Wells we have been including (0.75mA)--but of course that is just the low impedance leakage.  And there is tons of it--from below 100Hz well into the MHz range:

 

Nothing can block that (except an LPS-1 which just uses such as a charger).

 

And just to show what a "low leakage" (0.75µA) medical SMPS looks like (this graph is the low frequency range; we have other bandwidths):

 

And again from before, the results of shunting the DC zero-volt output ("ground") to AC ground, which shows the residual low-impedance leakage (over the same bandwidth; there is still a lot of leakage at high frequencies):

 

And finally, the leakage from an LPS-1 "energized"/charged from the same unit:

 

Of course none of this has anything to do with power supply output noise, output impedance, or other PS performance factors.  But it (as well as measures taken by John, Jesus, and Amir at the output of an unbalanced DAC) illustrate how pernicious this high-impedance leakage can be.  And as people who are experimenting in their own systems are discovering (both with Ethernet switches, rendering endpoints, and other gear), getting rid of it is worthwhile sonically.

 

Most people have always assumed that--apart from their output not being particularly low-noise--that SMPS were bad because of what they kick back into the AC mains.  I have always maintained that that aspect is really not the issue--since the certification restrictions on this things require the emissions into the wall to be spread-spectrum and very low level--there really is more other noise crap in your wall than what a single unit adds.  Since the start we have said that leakage is the real evil--and the low-impedance stuff is just as bad and needs to be blocked.  But the events and tests of the past months have lead John to discover the high-impedance leakage component--which is much easier to get rid of.  We really have never seen anyone else discussing or addressing this.

 

Ciao,

--Alex C.

 

 

[Superdad]

6 minutes ago, austinpop said:

Well heck - why do I need a groundhog for grounding RCA plugs. This'll do it: https://www.amazon.com/gp/product/B01HEN66K8/

 

I'd never seen those.  Cool!  

[Superdad]

11 minutes ago, vortecjr said:

Alex, I can easily measure the power supply noise at the analog output of a DAC with various power supplies. I can also see how shunting the negative terminal on the output of the SMPS cleans this up. What I have yet to see is the infamous "high impedance" noise. While it may be present and may affect the DAC somehow it seems that it's importance is minuscule (being so hard to measure and all) in comparison to plan old AC noise. Anyway, I'm sure Johns research will clear all this up.

 

Yet you ARE seeing it!  The improvement that you see in when you compare the output (from your single-ended DAC) with a given supply shunted and not IS THE REMOVAL OF THE HIGH IMPEDANCE LEAKAGE.

[Superdad]

2 hours ago, austinpop said:

Reminds me of the flying toasters screensaver from the early 90s!

 

Late 80s.  And funny that you mention it because in my former life as product acquisitions director for the now defunct Softsel/Merisel (was the world's largest computer s/w & h/w wholesale distributor until Ingram Micro eclipsed it), Berkeley Systems--the developer of the After Dark, flying toaster screensaver--was one of many hundreds of deals I found and signed.  The founders, Joan Blades and Wes Boyd, later sold at the right time and then founded the liberal activist organization MoveOn.org.

[That story is not near as interesting as my turning down a job offer from Paul Allen/Bill Gates in 1981--because I had just moved back to L.A. to do audio with my Hovland buddies and had a serious girlfriend.  I was just 20 years old and Gates had more pimples than me!] 

 

[Superdad]

3 hours ago, d_elm said:

...but not an FS105v3, 12V. 

 

That matches with what I reported last week.  For the FS105/12V, 7.5V works fine but 7V was not enough.

 

Interesting about the GS108.  Guess that the switching regulator (and supporting power circuitry) used in that model tolerates a wider input voltage range.

[Superdad]

13 hours ago, One and a half said:

The solution to that problem was to remove the  ifi Galvanic 3.0 and replace it with an Intona. 

 

Of course I am biased, but I own both the above devices, so I can suggest to you that the most musical solution might be to replace both with an ISO REGEN!   Don't know that you have ever tried one Gary.

--Alex C.

[Superdad]

14 hours ago, Speed Racer said:

I would imagine the iFi iPower would be an excellent power supply if you tied the negative DC output to ground as suggested by John in this thread. Does anyone have measurements of the iFi thusly grounded?

 

Please remember that shunting an SMPS unit's zero-volt output ("ground") to the AC mains ground only takes care of the high-impedance leakage that John has measured. All SMPS (including the iFi iPower) still have copious amounts of low-impedance leakage which this trick does not take care of (but which is blocked by an LPS-1 and of course is not even there with an LPS or a battery).

[Superdad]

10 hours ago, One and a half said:

I have not tried an ISOREGEN, since it needs also needs LPS-1, JS-2 (forget the AC unfriendly Meanwell) to work with it, plus quite a few connecting cables, USB connectors of the inverting kind and now shorting to ground links. When you elected to build the units this way, there was little/no choice to include them in the one box, I did ask at the time of the release of the ISOREGEN. That's why there's been no order to Uptone and went to ifi instead, since their one box uses USB3.0 bus power, one cable in and out.

 

Thanks for your reply Gary.  But to be clear, unlike the larger boxes you use, the ISO REGEN (with included USPCB A>B Adapter and light weight to go right into DAC) does not require an extra USB cable. And it also draws less that half the current (from the bus or from whatever PS is used with it) than the competing product (I have one and have measured).  We put the money into the clock and the ultra-low-noise regs, not into elaborate methods to run from bus power.

And with regards to a PS for the ISO REGEN, nobody says one needs to use a premium UpTone model.

 

Quote

What I would appreciate though is to use Uptone technology inside a DAC as a joint venture, where all the cables are hidden and paths are short.

 

Perhaps, but will you then eliminate all the rest of your spaghetti chain?  Your up-to-date profile lists the following as your chain:

USB chain (2017-11-11)
Computer USB 3.0 Port

-> Generic USB3 cable 1m

-> ifi Micro USB 3.0 Powered with iPower 9V 1.5A, 230V AC balanced powered]. Acoustic Revive RUT-1 USB filter

>  Nordost Blue Heaven USB cable 1m

-> Wyred 4 Sound Recovery [powered with iPower 9V 2.5A, 230V AC balanced powered]

-> SAEC SUS480 USB Cable 

-> Intona Standard Galvanic Isolator

-> Nordost Blue Heaven USB cable 1m
-> Icron Ranger 2212 USB Extender LEX 24VDC Powered with Acopian 3A, 24VDC Linear PSU 0V earthed

 -> 25m CAT5e 
 -> Icron Ranger 2212 USB Extender REX 
 -> Acoustic Revive USB-1.0SPS (Separate power and signal lines) (1m) 
-> DAC

 

Wow!  If you were to evaluate an ISO REGEN, I would ask that you do so with the removal of 12(!) items (devices, cables, PS) from your present chain.  

So you see, we ARE trying to simplify things...

 

Cheers to you,

--Alex C.

[Superdad]

16 minutes ago, auricgoldfinger said:

2. John S has said it is one of the noisiest power supplies he has tested. 

 

No, John never said that!  

He has never tested the output noise of the iFi iPower (properly measuring very low PS output noise--across wide bandwidth is not a trivial procedure) and we do not make broad statements against our competitors.

 

What John did measure and say about the iPower SMPS is that its leakage current profile (both high- and low-impedance as at the time he had not sorted the difference) was one of the wildest he has seen. (I have the graphs to concur but it is not appropriate to post them).  And the iPower benefits from John's ground shunt trick just as other SMPS units do.

 

It is a fact--with no intention to denigrate the good value of their piece--that the iPower is an SMPS wall-wart (with switching noise and leakage just as all others) that iFi has added several parts to produce a unit with much lower output noise than a typical SMPS, though they only state doing so across the audio bandwidth.

 

I am uncomfortable--though partially culpable I see--with this discussion veering into commentary about our competitors' products.  We respect iFi Audio and they seem to respect us.  I would like to keep it that way.

Thanks,

--Alex C.

[Superdad]

6 minutes ago, One and a half said:

 

On the USB Chain, I would like to clarify, since ridicule and ignorance befuddle the intent to overcome USB limitations. The Icron and associated gear stays, to isolate an electrically and noisy computer from the listening room.

The USB chain changes from day to day, sometimes less even to the barest bones, so it is under constant review.

 

It appears that the ifi Galvanic3.0 won't be the saviour, due to either design or manufacturing flaws, we'll see. Certainly after your post Alex, a ISORegen won't be on the shopping list either. Daniel from Intona is working on a newer model USB Isolator, so will definitely put this design on the short list. It may take some time, there's lots of music to listen to.

 

Cheers,

 

Very sorry if I offended you Gary, that certainly was not my intent.  You simply confused me with your indictment of the ISO REGEN as something that adds cables and complexity, when it does the reverse.  And when I went to look at your set up I was stuck by the irony of it.

Also I notice that you have the ICRON Ranger USB extender at the far end--as the last piece into your DAC.  From the standpoint of sonics I am convinced that is holding your entire system back and that with the ICRON as the last element you will never realize the benefits of improved USB signal integrity, impedance match, or clocking.  Might I suggest that you put a nice USB regenerator AFTER your ICRON--as the last piece before your DAC?  You should achieve a better result.

Best,

--Alex C.

[Superdad]

Okay guys, let's not give @One and a half too hard a time.  And let's not divert this thread into a discussion of power supply performance. (Lots of fine choices on the market these days.)

 

Back to SMPS and grounding for network switches, etc. 

[Superdad]

21 minutes ago, ted_b said:

So here is a weird one:  I installed a JS shunt on my LPS-1's Meanwell.  A couple days later I am hearing a slight buzz in my speakers (no music), on and off, cycling every few seconds.  I go and find my Meanwell power supply green light is dim and cycling on and off at the same rate as the buzzing.  I unhooked the shunt and the Meanwell green light got brighter and stayed on; buzzing cycle gone.  Then added the shunt back in and all is well...so far.

 

What was that?

 

Hmm... (or should I say bzzz?).  We will look into this.  

[Superdad]

48 minutes ago, R1200CL said:

Wrap it in...

 

Moo metal? 

 

 

[Superdad]

24 minutes ago, Cornan said:

 

If you want to optimize it you should consider silver plated wires for the JSGT. It makes a noticable improvement over copper wires.

 

Sorry Cornan, but while I appreciate your enthusiasm and the pleasure you get in doing tweaky things, I am going to call BULL on the notion that silver wires improve the ground-shunt of an SMPS.  That is just a bridge too far.

 

John's trick of shunting an SMPS' high-impedance leakage to ground makes a  measurable difference--and for some an audible one. But using a silver wire to do it is not going to do anything.

 

You can do these things (though replacing the 3+ Gopher SMPS units in your system would give real benefits) but I feel the need to say something here as I really don't want John's based-on-engineering mod to get too mixed up in hoofy-spoofy stuff.  We have enough of a time already with skeptics calling our actual products "snake oil."  

 

All good wishes to you,

--Alex C.

 

[Superdad]

I love you guys...

[Superdad]

1 hour ago, charlesphoto said:

You may want to check out the HDPLEX 200w new version. I have the 100w and like it as it’s so flexible, esp with the 5-19v adjustable rail.

 

24 minutes ago, JohnSwenson said:

You should be able to drive 5 LPS-1s from that supply (4 off the 12V and 1 off the adjustable set fro 9V or so).

 

The UltraCap LPS-1 does not tolerate "energizing"/charging input above 12V.  Even a short spike above 12V can fry one of its input regulators--turning the board into scrap.  Four of the six destroyed LPS-1 boards we have in a box here came from people using an HDPlex power supply. 

Now I do not know the vintage of the offending HDPlex units owned by those who have suffered, and while HDPlex seems to offer good value (and I know they offer good support), at this time we must advise against using that brand as a charger for our LPS-1.  Perhaps one set at 9V would be safe, but that is no guarantee. 

 

It is easy for us to inspect a failed LPS-1 board to determine if an over-voltage killed it:

 

 

Such is not covered by our warranty.

 

As this thread--and the graphs that we published made clear--there is no advantage to powering an LPS-1 with a linear power supply versus the Mean Well we include, once the Mean Well's output is shunted.

 

 

I am still trying to make arrangements with a quality SMPS manufacturer to supply us with units that already have their zero-volt output ("ground") tied to the AC mains ground pin.  There are some that do so already, but we have multiple factors to consider in making our choice before I make the required 1,000 unit stock purchase.

 

Thank you all for your enthusiasm, patience, and understanding.

--Alex C.

[Superdad]

7 minutes ago, auricgoldfinger said:

I think what they mean is each rail is grounded separately from the others.

 

No, it can not be both ways.  Either all the rails share a common ground, or they are all floated from mains ground.