SMPS and grounding

[Superdad]

39 minutes ago, rickca said:

OK here's a really stupid question.  Why don't they make SMPS with negative DC output grounded in the first place?

 

Yeah, I've been wondering that for weeks now since John starting playing with this during measurement sessions.  Probably has something to do with regulatory requirements.  A lot of the European regs are very strict with regards to what goes back into the mains--caring less about what comes out of the DC end.

I sent John a low-leakage "medical" Mean Well for testing a while ago.  It would be funny if the difference with that one was its grounding.  I'll ask him to check it.

 

Speaking of grounding--but not to be confused with this grounding of the output topic--I recently discovered (upon opening one) that although the Mean Well we offer for REGENs and LPS-1s has a 3-pin, IEC320-C14 socket, that ground pin is not connected to ANYTHING inside the SMPS!

(I still prefer to provide that style of unit--versus wall-wart or figure-8 C7 inlet--so that we don't have to stock and keep track of wall plugs/cables for all countries.  Everyone gets a 50cm USA cord that fits in the package and they can toss it or cut the end and put on a local plug--or whatever.)

[Superdad]

1 hour ago, Ryelands said:

 My point OTOH is that he's to blame for all this.

 

[Superdad]

1 hour ago, sadekkhalifa said:

I have two questions regarding to iFI Groundhog:-

 

1- Will the IsoRegen + Lps-1 can benefit from Groundhog?

 

Shunting SMPS leakage currents is most always a good thing.  (And yes, even a bit for an LPS-1; We will present our findings about this at a later date.)

 

1 hour ago, sadekkhalifa said:

2- And does the Groundhog solve the issue of ground loop in some systems when the Isolation is enabled in IsoRegen like what i have?

 

I don't recall your particular issue, but most of the folks that experience issues with Silanna-based USB galvanic isolation (this includes iGalvanic users too) suffer from lack of good grounding (typically on the downstream side), not "ground loops."

The solution to that is generally to ground the downstream USB input (see my long post here: https://www.computeraudiophile.com/forums/topic/31877-iso-regen-listening-impressions-kicked-off-with-some-measurements/?page=26&tab=comments#comment-714912).

While the single wire solution I present is cheaper, the Groundhog solution can do the same thing for the ISO REGEN--but only if you use it on the DC input jack of the ISO REGEN.  Putting it on the DC input jack of the LPS-1 will not help any "lack of ground ticks/disconnects," as the output of the LPS-1 is completely "floated."

 

--Alex C.

 

 

 

[Superdad]

5 hours ago, rickca said:

So that's different than grounding the downstream USB input, right?  

 

No.  The ground (outside of DC barrel) of the ISO REGEN's power input and the shell/ground of the ISO REGEN's USB output jack (and hence the same on the DAC's USB input jack) are common to one another (use a continuity checker and you will see).

So running a wire from the ISO REGEN output/DAC USB input to AC mains ground is the same thing as connecting the ISO REGEN's DC "ground" to AC mains.

[Superdad]

15 hours ago, agladstone said:

Also, if I followed this thread carefully enough, it seems like for my LPS-1 that powers my ISO Regen, I would want to have the ground wire going to the DC input of the ISO-Regen and NOT the DC input of the LPS-1, is this correct?

 

With the present topic, no, you would just ground the DC of the supply that you have feeding your LPS-1.

 

Quote

Also, If I’m clear, in addition to one going to input of ISO-Regen, another ground wire should also go into the input of the LPS-1 from the meanwell SMPS that is powering the LPS-1? 

 

Grounding of the ISO REGEN (its DC input jack is already and always common to its USB 'A' output shell and consequently your DAC's USB input ground) is only necessary in cases where you experience ticks or disconnects (caused by EMI charge build up with the Silanna chip having no place to "drain" to).  Otherwise, doing that is not advised--and there is even a chance you will introduce a ground loop (though it should be harmless since on that, the downstream side of the isolation,  only the ISO REGEN and your DAC-onward are involved; upstream, computer-side is of course isolated).

 

Hope that all makes sense.  Experiment and do what sounds good.

 

--Alex C.

[Superdad]

6 hours ago, mourip said:

What about a device that has an internal SMPS that is fed via an external IEC socket and connected to a three plug AC cable. Is it most likely that given the fact that there is a ground pin being used that the SMPS has a ground connection to it or is it more likely that the ground pin is just a chassis safety ground?

 

I can probably answer my own question with "It depends on how it was designed" however is there any harm to making sure that the negative output of the SMPS is in fact tied to the IEC ground or would this be ill advised?

 

Hi Paul, really nice to see you here!

 

Indeed the answer is "depends on how it was designed."

While we (UpTone, myself, John S.) are NOT advising anyone to open up their AC powered devices or to stick fingers inside (while standing in a puddle of water in a rainstorm ), it should be rather easy to check.

 

We are talking about determining if the DC output ground of the SMPS is common to the AC mains safety ground.   So all you need do--with POWER OFF please--is put one probe of a continuity checker (the beeping setting of your multimeter) on whichever terminal of the internal SMPS is not the +DC (avoiding calling it "ground" on purpose), and the other probe on the ground pin of the IEC inlet.

 

If you get a beep then the output of the device's SMPS is already grounded.  If not, well then it is up to you as to if you wish to ground it (may void warranty, don't blame us for damage, etc.).

 

Cheers,

 

--Alex C.

[Superdad]

44 minutes ago, Cornan said:

 

+1 I would be very interested to know this as well. My Brooklyn DAC have a SMPS connected to a grounded IEC. Just thinking out loud. If the IEC is connected to the chassi, would'nt a ground wire from chassi to safety ground (or grounding box) do pretty much the same thing as internally rewiring the IEC socket so it is connected to the safety ground?

 

Not entirely following you.  We are talking about grounding the (non-positive) OUTPUT of the SMPS.  So it really depends upon if the ground-plane of the circuit board is common with the chassis (which IS most likely common to AC mains ground).  There is a chance that only the SMPS INPUT is grounded to the chassis/inlet.  You would need to use a meter to check.

 

44 minutes ago, Cornan said:

 

Looking at a picture inside Brooklyn DAC it does look like the SMPS is actually connected to the IEC ground pin.

 

Well again, the SMPS input is most assuredly connected to the IEC ground pin.  But the photo does not give enough detail to tell if the OUTPUT of the supply is.  I see just the two black wires going to the PCB.

[Superdad]

Off topic and argumentative posts deleted.

Take that stuff somewhere else guys!

[Superdad]

9 hours ago, Cornan said:

I will receive two today and will use one the output of my isolation transformer with floating secondary (replacing my 1-pole ditto) and on the output of my balanced isolation transformer with floating center tap.

 

I assume you realize that by not connecting the center tap it is no longer a "balanced transformer."  A hamburger without the meat patty in the middle is just a bun... 

[Superdad]

2 hours ago, flkin said:

Or you can simply use linear power supplies instead of SMPS and hey presto, no leakage current with generally accepted better sound ... 

 

Actually, linear supplies do also have leakage currents.  Just not nearly as much.

[Superdad]

6 hours ago, BigGuy said:

A bit confused now.  Understand the shunt to ground of the minus side of the DC power supply using the umbilical for the high frequency component of the leakage current for which you provided details previously.

IIRC. you mentioned that it was easy to get rid of the low frequency component as well but it needed to be done differently.

 

You are conflating frequency with impedance and that is not correct.  In other words, there is high-impedance leakage and low-impedance leakage--and both have both high and low frequency components.

 

6 hours ago, R1200CL said:

• John

.......and even if we use a LPS-1 and shunt it we won’t get rid of the low impedance ?

(using a FS105/8 before your AE. )

 

I think yes, cause in another thread you have stated the shunt applies to all SMPS including the LPS-1.

No matter what so ever ?

 

Tough in that posts high and low impedance leakage was not discussed. 

 

So yet no final solution how to get rid of both impedance components is such an application ? ( only those two switches used as AE).

 

Actually the LPS-1 already completely blocks low impedance leakage.  And the SMPS grounding trick (I am now trying to source units that are already grounded--they exist--or get Mean Well to supply a custom version for us) is completely effective at shunting (getting rid of) the high-impedance leakage.

Due to our use of transistors (instead of large, expensive, noisy relays) to alternate between banks of ultracaps, there is a very small amount of capacitance across power domains in the LPS-1 (less than 100pF) which is enough to allow some high impedance leakage through.

 

To prove to everyone how effective the grounding trick is, here are 3 graphs--directly measuring leakage versus frequency.  (do not try to compare these to anyone else's measurements--scales and units are different; and these are in dBM not dBV, that's 13dB difference right there).

 

Here is the leakage (just up to 1KHz, John has done wider bandwidth measures as well) from a stock Mean Well GST40A:

 

Here is the same Mean Well unit with its DC zero-volt ("ground") tied to the ground pin of its IEC320-C14 inlet this way:

 

What you see remaining is all the low-impedance leakage. (Again this is the leakage measurement of just the grounded GST40A.)

 

And here is the leakage (not output noise; these are all common-mode leakage tests which John can explain) from an LPS-1 being powered by the same modified Mean Well:

You can see how all the low impedance leakage is now blocked.

 

So yes, this is our admission that LPS-1, when used with an SMPS whose DC output is not grounded to AC mains, will let high-impedance leakage though (that's a different graph that I don't presently have from the same test set up).  How did we allow this to happen?  

a) The test set up to see this properly was not made; 

b) Power supplies used during development may have been grounded units;

c) We were not looking at high-impedance leakage or at these frequencies;

d) We were concentrating on other aspects of performance: Isolation, ultra-low noise, ultra-low impedance;

e) While John purposely chose transistors with the lowest possible capacitance, the spec sheets all specify capacitance only with the transistor in its "on" state, whereas in our application it is the capacitance of the part in its "off" state that is letting a little leakage through.  (He has since built a special board/jig to measure these transistors, and while he found a couple of parts that have both a little lower capacitance and meet the various functional current/voltage requirements, there will always be some; Even if we cut the total capacitance in half the high impedance stuff can still get through; The ground/shunting solution is simpler and totally effective.)

 

I hope this clears things up for those who are interested.  And yes, I have egg on my face for all those months during which I insisted that the choice of "energizing"/charging supply would make zero difference to the output of an LPS-1.  Of course noise, output impedance, and other aspects of our "floating" supply's isolation are not affected.  But yes, based on our choice of bundled SMPS for charging, some leakage current gets through--unless you ground/shunt it. Too bad the Mean Well GST25A-07 we chose was not the sort that already was grounded in this fashion (no safety or emissions violations come from it as far as we can tell; and other certified SMPS units are already built that way).

 

Happy Friday.  Have a great weekend everyone,

--Alex C.

 

 

[Superdad]

18 minutes ago, BigGuy said:

That being said, how does one block the low impedance component? 

I have already added the grounding umbilical to the output of the two LPS which I am using per John's guidance.

 

Hi BigGuy:

Can you be more specific?  

a) What power supplies are you using?  

b) What components are they powering?

 

My post was only discussing SMPS (switch mode power supply) leakage and its relation to our UltraCap LPS-1.  Our LPS-1 always blocks low-impedance leakage.

 

And since LPS (linear power supply) units in general do not have much leakage current (high or low impedance), there is usually never a need to ground their outputs. 

There is the somewhat separate topic of Ethernet switches and grounding of a power supply feeding one.  However, if one is feeding an Ethernet switch from an LPS (ours or someone else's), grounding of that supply's zero-volt leg is not useful--unless you are using a specific Ethernet switch (such as the NetGear FS105/108) that John tested and found to also block leakage coming in from other Ethernet devices (by grounding).  

 

As for blocking the low-impedance leakage component of SMPS units, for that you need to either use an UltraCap LPS-1 or ditch the SMPS.  

I defer to John as to if there are actually any low impedance leakage currents making it into a switch from other networked components.  I suspect that the magnetics at every EN jack already blocks that.

 

Sorry this is so confusing.  

--Alex C.

[Superdad]

1 hour ago, gstew said:

Please help me make sure I have understood all of this...

First, linear AC-DC power supplies, have only the low-impedance type of leakage, though at various levels based on their 'different LPS implementation'**. AND all SMPSs, by nature of how they function, have both of the low-impedance & VERY high-impedance types of leakage.

Uptone Audio LPS-1's (and similar non-AC connected supplies, such as pure battery supplies) do not pass the low-impedance leakage through the power supply feed, whether that leakage originates from an LPS or SMPS.

Grounding the negative side of the output of SMPSs effectively eliminates the majority of the high-impedance leakage, preventing it from being passed through to the driven piece, whether to an Uptone LPS-1 or another piece of equipment, such as a network switch.

AND grounding the negative side of the power input into one of the tested & confirmed switches (I won't list them here, that list may grow over time) ALONG with proper Ethernet cable connection (currently using alternating jacks, but that also may change over time) will prevent high-impedence leakage from being passed through the switch to downstream connected gear.

Have I got it right so far?

 

Yes, you got it perfectly!  And your clear summary stands well and very helpful alongside what John and I wrote.  This sort of stuff should be preserved somewhere to be useful to others.

 

 

1 hour ago, gstew said:

Next, a few questions....

 

All GREAT questions and ideas Greg!  

It's been a long day so answers and I-don't-knows will have to wait a day or so.  

As usual, you are on the right track.  And we are waiting on photos of your wild web of PS wizardry! 

Happy Sunday my friend,

--AJC

[Superdad]

3 hours ago, jabbr said:

or is it this: the solid state “relays”/MOSFET switches in the power supply have very high impedance when “off” and connected to the input SMPS so transmit this leakage?

 

yeah I think these are parasitic capacitances on the package, mosfets etc and really do pass much more current at higher frequencies — and grounding the SMPS does provide a low impedance path to ground.

 

MOSFETs where?  There are none in the LPS-1.

[Superdad]

34 minutes ago, BigGuy said:

IF the LPS or SMPS is grounded with a 3-prong plug, does this mean a grounding umbilical on the output is superfluous?

 

No!  Just because a PS may have a 3-prong (and actually have that pin connected on the primary side), it does not mean that it is also connecting that ground to its DC zero-volt output side.

 

An example that is the 40 watt Mean Well I posted graphs and a photo of:  Unlike the 25W series MW (okay the 7V version is labeled 22W) the GST40A units do come with their C14 AC input ground pin connected to the primary side.  But they still need the mod (green wire in my photo or external accomplishment of the same) to shunt the high-impedance leakage back to ground as discussed.

 

See also my overdue reply to @rickca in my next post.

[Superdad]

On 10/27/2017 at 5:49 PM, rickca said:

This is promising.  Is there an SMPS product line that's already grounded with a variety of volts/amps that might help me replace my other SMPS (the ones not powering an LPS-1).  Since I don't do DIY and find iFi's $50 kit obscenely priced, this would be great.

 

Hi Rick:

Well the thing is, this is not something that any SMPS manufacturer specifies.  And it seems to vary some even within a brand's offerings.  So each one has to be tested (VERY easy: Just put a continuity checker across the ground pin--has to be a 3-wire unit of course--and the outer barrel of the output and hope for a beep.)

 

I have bought a few brands through normal distribution but don't thing we can draw any conclusions about who does and does not offer what we seek.

I confess that most of my sourcing research is focused on Chinese brands that have low cost models which meet our requirements (or for which the factory is willing to customize for us).  I am looking for pieces for $8-15 in quantity, not $22-28 or more.

 

But if anyone comes across units that display continuity (as above), please e-mail me with the brand and model.  Even if you don't discover a brand whose whole line is correct for us, maybe there will be a model or two.  It's worth knowing.

[Superdad]

14 minutes ago, jabbr said:

Haha are you sure? Of course there are

What “relays” are you using? Solid state or mechanical?

 

Quite sure!    No mechanical relays and no MOSFETs!  MOSFETs have much too much capacitance.

 

 

[Superdad]

1 minute ago, jabbr said:

Some do ... but trust me there are some MOSFETs ... including likely inside the solid state relays... even without there are parasitics 

 

Lets take this offline Jonathan.  But just to be clear, there are no "solid state relays" on the LPS-1, no MOSFETs.  John chose very low capacitance bipolars for the task.

 

Are you still working on your ultracapacitor LPS (which we saw uses MOSFETs)?  Did you ever get it fully working? 

[Superdad]

1 hour ago, jabbr said:

Perhaps I am missing something? Would it be possible to draw out a proposed schematic that would illustrate these two distinct, components (each having their own impedance/frequency profile)?

 

John may be able to (though he is pretty busy on one of our  unannounced new products right now), but I think the three measurement graphs I posted (https://www.computeraudiophile.com/forums/topic/37034-smps-and-grounding/?do=findComment&comment=734822)

do a good job of proving the existence of the two distinct forms of leakage.  The SMPS before and after ground shunting graphs shows the elimination of the high impedance stuff, and then the graph going through the LPS-1 shows the elimination of the remaining low impedance stuff. (Not presented was the graph showing  the un-shunted SMPS powering the LPS-1: Much of the high impedance leakage can be seen getting through--until the simple input ground shunting is put in place.)

 

[Superdad]

40 minutes ago, Speed Racer said:

 

Why would you think it could? It requires 12VDC.....

 

I have the FS105 (12V version) here and just tested.  7.5V is its lower limit for operating.  7V is too low.  I found out by trying to power it with an LPS-1 set first at 5V and then at 7V--neither worked, but 7.5 from the Mean Well I was [the LPS-1] charging with works fine.

So there you have it. 

[Superdad]

52 minutes ago, Speed Racer said:

There is a reason they use a 12VDC power supply if that is what they spec.

 

Yes, there is a reason:

So they can supply them with a cheap 500mA $1 wall wart instead of a higher current supply.  Nothing in the box runs from 12V--the chips run 3.3V or even 1.1.   All these things just use switching regulators that run from a range of input voltages.  Give them lower voltage and they just draw a little more current.  That's all.

[Superdad]

1 minute ago, chetthejet said:

Should the shunt go in at the isoRegen connection or at the connection between the LPS-1 and the Meanwell???

 

Ground ONLY the Mean Well's zero-volt ("ground")!  Do NOT ground the output of the LPS-1 or you will defeat both its "floating"/isolation feature and the galvanic isolation of the ISO REGEN. 

[Superdad]

15 minutes ago, Speed Racer said:

 

Then your LPS-1 at 5VDC should work just fine, right? But it doesn't......

 

Wow, you really want to argue this?  The fact that it is not happy with 5V input simply means that with the synchronous buck regulator chip used, combined with whatever surrounding caps and inductor, 5V falls outside the input range.  There really is very little sophistication in the PS of these cheap switches.

[Superdad]

By the way, over the weekend John made another very interesting (and provocative to some I'm sure) post in thread that may have been overlooked.  I think a lot of you here might enjoy it:

 

 

While we won't yet reveal the new products we are working on, I can promise that all of what John has been posting lately--and some of the pieces that many of you are experimenting with--is germane to the directions we are going in.  

[Superdad]

41 minutes ago, Speed Racer said:

 

So, it was designed for a higher voltage input....wow....they didn't use a 12VDC supply JUST because they were cheap.

 

This is all quite ridiculous. We don't need an LPS-1 on every DC powered device connected to our audio systems. Just ground the damn zero voltage line after it leaves the SMPS and be done with it.

 

Okay, it is becoming clear you may be among the group that likes to twist things around to prove some conspiracy. 

Let's clear this up in  hurry:

a) The person asked if he could use a lower voltage supply to power his switch.  He said nothing about an LPS-1 and I was not promoting such.  All I said was that I tested the range of the MP1482DS buck regulator as implemented in the Netgear FS105 switch and found that it works fine down to about 7.5V.  Could handle up to 18V I suspect.

 

b) As with laptops and many other computer devices, the widespread use of wide-input range synchronous buck converters allows the manufacturer to supply a higher voltage, lower current PS--which are cheaper and more efficient.  That IS why they do it and that is why you see 18-19V laptop power supplies.  There are no logic chips running from 12V--few even run from 5V.  Lower voltage, higher current. 

 

c) We have never promoted the use of our UltraCap LPS-1 for Ethernet switches (though people are free to do whatever they wish with out popular, ultra-low-noise, "floating"/isolated supply).

 

d) "Earthing" the zero-volt ("ground") output of an SMPS only removes its high-impedance leakage component.  It does not deal with any of the following: low-Impedance leakage, output noise/ripple, high output impedance, or anything else.

 

Since you seem to just now be joining this thread, I suggest you scroll back and read John's long treatise on the issues of leakage: