SMPS and grounding

[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.

 

 

[rickca]

6 minutes ago, Superdad said:

 I am looking for pieces for $8-15 in quantity, not $22-28 or more.

OK, thanks Alex.  I understand your criteria, but even $25 is half the price of the iFi Groundhog which doesn't even include an IEC power cord and requires an extra outlet.  

[BigGuy]

20 hours ago, lmitche said:

John, WOW what a great explanation of the cause of the mysterious SQ changes we hear so frequently.   I even get it.  Many thanks!

 

Larry

+1  Starting to understand this whole issue of impedance noise and how to deal with it.

[jabbr]

41 minutes ago, Superdad said:

 

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

 

 

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

[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? 

[jabbr]

Haha ok (there are mosfets in the cpld/FPGA)... my own power supplies are not intended stand alone, rather for project s I am working on. The published circuits are examples that folks may modify to their own needs. I’m fine with any of bipolars, jfets and mosfets ... the particular MOSFET shown has particularly good specs  for the application. 

 

The point I am trying to make ... which applies to all solid state devices ... is that the impedance has both a parasitic and inductive component and hence leakage current will tend to be higher at the node which will be at a higher frequency.

 

Ths point has come up with isolation transformers — why would 0.0005 pF interwinding capacitance be better than 0.005 — where at 60 Hz allows negligible current yet at higher frequencies, one can measure a difference...

[charlesphoto]

6 hours ago, Superdad said:

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.

 

I have one of these Indeed ultra low noise adapters currently powering the modem into the house. It has the double insulated symbol and Class II rating on the case, and in the description says it's grounded. Unfortunately I don't have a three wire continuity checker. $17.99 on eBay, 12V only. 

 

http://vi.vipr.ebaydesc.com/ws/eBayISAPI.dll?ViewItemDescV4&item=250734255819&t=1457554608000&tid=10&category=56172&seller=indeed-hi-fi-lab&excSoj=1&excTrk=1&lsite=0&ittenable=false&domain=ebay.com&descgauge=1&cspheader=1&oneClk=1&secureDesc=0

 

I would say it's most likely based on this one, which is half the price:

 

https://www.amazon.com/CAIXING-Switching-100-240V-Transformers-Security/dp/B0735G19BP/ref=sr_1_2_sspa?s=electronics&ie=UTF8&qid=1509318785&sr=1-2-spons&keywords=12v++caixing&psc=1

 

I will say, that the ground shunt trick has totally transformed my system. And even putting one on the Roku (which is powered on a different circuit from the hifi but lives next to it) I could swear the picture is brighter and sharper and the sound better. So far I've put them (ten in all) on most everything in the house that connects with the wired ethernet (Mac Pro, newer AE, and Xbox not included) and so I'm not particularly sure which is doing the most good, or if it's a cumulative effect. Whatever, it's almost like somebody delivered a bigger better system (esp the amp) to my house, for under $50 at that. Thank you John Swenson! 

[JohnSwenson]

10 hours ago, jabbr said:

@JohnSwenson nice investigation  

 

Help a simple country boy out please: the term “high impedance leakage current” is making me a little dizzy as I try to drink my coffee on this Sunday AM. Impedance isn’t a term normally used to qualify a current so perhaps we should rename this before it takes on a life of its own — you mean “high voltage/impedance” = current ... what is a better term? 

 

Also these currents are going to be way more important at higher frequencies. I think if a circuit were made available to precisely illustrate the excellent point you are making here (this is all about parasitic capacitances and inductances) this frequency point will be better illustrated. (Yes— high impedance probes are essential when working with RF)

All current comes from some form of voltage between two points. No real voltage source is perfect, there will always be SOME impedance in series with that voltage. It might be 1 ohm it might be 1 mega ohm it might be 300 mega ohms.

 

This impedance shows up in what happens when you connect a resistance across the voltage source. For example if you have a 1K impedance source and put a 1 ohm resistor across it, you will get a very large voltage drop. If you put  a 1 mega ohm resistor across the source you will barely have any change. If the source is 300 mega ohms and you put a 1 mega ohm resistor across it, the voltage drop is large.

 

In the case of SMPS leakage there seems to be both a high impedance component and a low impedance component. At low frequencies the impedance was very high, about 300 mega ohms. I determined this by trying different resistors in series and to shunt to ground.  But it wasn't behaving correctly. I put a resistor in parallel that should have dropped the level by 80dB, but it didn't it only dropped by 30dB. Then I tried putting a medium range resistor in series, and it dropped down to below the instrument floor. The only way this can happen is if the source has at least two sources, one very high impedance and one much lower.

 

The source impedance DOES vary with frequency, but no where near close to what it would be if it was just a capacitor from AC line to output. It seems to be way more complex than that.

 

John S.

[JohnSwenson]

20 hours ago, gstew said:

John & Alex,

Thanks for the great explanations & education on how AC power supply leakage works, how & why it was characterized, why that characterization missed the 'high-impedance' type of leakage, and what you can do to diminish its effects on our audio systems.

The wealth of information you've both shared here (and embodied in your products) has made a world of difference in the sound quality of many audio systems, mine VERY much included.

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

Your first section is all correct. Now on to the questions:

 

#1:  linear supplies still have leakage and shunted SMPS (not going through a LPS-1) still have low impedance leakage. You can still have effects from this. This is where my posts awhile back on leakage loops can come into play. Particularly the parts about keeping the impedance on the AC line between devices as low as possible.

 

In your specific application there is no easy way to tell whether SMPS or LPS is going to sound better in your amps, it very much depends on the implementation of both.

 

#2: I did a bunch of tests and this and did not see ANY increase in leakage when using a switching DC-DC converter.  They CAN produce all kinds of radiated noise (primarily from unshielded inductors and poor PCB layout) and high frequency noise on both the output and input. All of these can be very effectively dealt with, shielded inductors and good PCB layout dramatically reduce emissions, and proper filtering can almost eliminate the electical noise. Putting a good linear reg after the DC-DC converter can essentially eliminate the noise on the output. In particular the LT3042/LT3045 are particularly effective in getting rid of noise from switching converter. The combination is a very useful tool, the switching converter can handle large differences in input and output voltages, producing a voltage just a little bit above the output voltage and the linear reg can clean up the output without generating too much heat.

 

#3: Grounding the output of a linear supply may help some. It depends a lot on the leakage coming out of the LPS. I can't say it won't help at all, but it is not going to be anywhere near as dramatic as with the high impedance of the SMPS. I measured about 4 LPS and 2 of them had no difference at all when grounding, one made a small difference and one was a little bit larger difference.

 

#4: Correct, leakage is usually independant of other properties of a linear supply.

 

The ps section:

Un-damped linear supplies can have their transformer resonances excited by high frequency noise on the AC line. This can come from other LPS or especially from SMPS. The broadband high frequency noise from an SMPS can easily excite a non-damped linear supply. SMPS don't have these resonances so don't have this issue. This resonance exciting noise is high frequency so it CAN be blocked by a Topaz or similar transformer. A Topaz blocks noise starting at around 400Hz, so it is great at blocking this high frequency noise, but it does NOT block the low frequency part of leakage (60, 120, 180 etc).  This may contradict the "put everything on the same strip rule", if the leakage from the SMPS is already dealt with, then putting on the other side of a Topaz from the component with a linear supply can make a big difference.

 

pps:  In a linear supply the leakage mostly comes from the transformer and relates to how it is physically built. I have not spent any time trying to make generalizations based on transformer construction and leakage, I'm sure they are to be had, I just haven't done the testing.

 

John S.

 

[jabbr]

2 hours ago, JohnSwenson said:

In the case of SMPS leakage there seems to be both a high impedance component and a low impedance component. At low frequencies the impedance was very high, about 300 mega ohms. I determined this by trying different resistors in series and to shunt to ground.  But it wasn't behaving correctly. I put a resistor in parallel that should have dropped the level by 80dB, but it didn't it only dropped by 30dB. Then I tried putting a medium range resistor in series, and it dropped down to below the instrument floor. The only way this can happen is if the source has at least two sources, one very high impedance and one much lower.

 

The source impedance DOES vary with frequency, but no where near close to what it would be if it was just a capacitor from AC line to output. It seems to be way more complex than that.

 

Can you draw out a circuit? I am assuming there is a shared ground, and if there is a shared source i.e. there is a single circuit, then there is no physical way to separate currents into a low and high impedance component *except* by frequency. 

 

Typically capacitance will predominate at lower frequencies and then inductances at higher frequencies (package, leads etc), so that there will be a dip at some point of minimum impedance -- below this frequency, capacitance predominates, above inductance.

 

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)?

[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.)

 

[BigGuy]

14 hours ago, charlesphoto said:

I will say, that the ground shunt trick has totally transformed my system. And even putting one on the Roku (which is powered on a different circuit from the hifi but lives next to it) I could swear the picture is brighter and sharper and the sound better. So far I've put them (ten in all) on most everything in the house that connects with the wired ethernet (Mac Pro, newer AE, and Xbox not included) and so I'm not particularly sure which is doing the most good, or if it's a cumulative effect. Whatever, it's almost like somebody delivered a bigger better system (esp the amp) to my house, for under $50 at that. Thank you John Swenson! 

charlesphoto's post raises a question for me.

I understand that the PS for a router should have DC grounded but is it important to go upstream to the modem as well?

 

IF so I need to get a power extender strip with a lot more outlets...or find a source for inexpensive grounded LPS of varying output voltages for all these accessories! 

[JohnSwenson]

3 hours ago, BigGuy said:

charlesphoto's post raises a question for me.

I understand that the PS for a router should have DC grounded but is it important to go upstream to the modem as well?

 

IF so I need to get a power extender strip with a lot more outlets...or find a source for inexpensive grounded LPS of varying output voltages for all these accessories! 

I'm sorry, I guess I didn't explain this well enough. IF the last switch before your audio endpoint (computer, renderer etc) is one of the named switches (FS105 and FS108) AND you ground the power supply driving THAT switch, you do NOT, repeat NOT have to ground the SMPS from any other upstream network devices, period. Grounding the PS to one of those switches will get rid of the leakage from all the other network devices you have. This is why I am singling them out, if you have one of these as the last switch, (and ground its PS) then you do not have to worry about leakage from anything else on the network, period.

 

John S.

[thyname]

3 minutes ago, JohnSwenson said:

I'm sorry, I guess I didn't explain this well enough. IF the last switch before your audio endpoint (computer, renderer etc) is one of the named switches (FS105 and FS108) AND you ground the power supply driving THAT switch, you do NOT, repeat NOT have to ground the SMPS from any other upstream network devices, period. Grounding the PS to one of those switches will get rid of the leakage from all the other network devices you have. This is why I am singling them out, if you have one of these as the last switch, (and ground its PS) then you do not have to worry about leakage from anything else on the network, period.

 

John S.

 

Great explanation!

 

What about the scenario when no switch is involved (audio streamer connected directly to router), would grounding the router would take care of everything in the house, including another location that has a Switch? Or does this just work with the two FS105 and FS108 models you tested? Thanks!

[Speed Racer]

I'd use the GS series switches instead...

 

GS105NA and GS108NA

[jabbr]

14 hours ago, Superdad said:

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.)

 

The best I can formulate this is that leakage currents arise from circuits which include the ground. Many of these circuits are formed by parasitic capacitances (ie are not intended by the schematics). 

 

There may be multiple such leakage circuits. Each with a characteristic impedance, current and frequency profile. Current will tend to flow across the path of least impedance and this providing an alternate low impedance path to ground  will tend to shunt current that would otherwise flow in a higher impedance leakage circuit.  

 

Note that the shunts must be through low impedance (not just low resistance) connections to ground. When mitigating otherwise low impedance leakage circuits need to focus on low inductance shunts to ground as inductance becomes more important as frequency rises.

[charlesphoto]

Interesting. I'm currently using a GS108NA with the ground shunt and powered by an "El Cheapo" Lps. It has other traffic on it though than just the hifi. I have an eBay FS105NA for $10 coming so will isolate the streamer and server on that with a ground shunt and see what happens.

 

As far as I'm concerned the grounding I did upstream hasn't hurt, esp as I had a bunch of extra adapters and it's really easy to do. I'm quite sensitive to EMF so getting the overall level down is always good. That said I wouldn't go out of your way or break the bank (i.e. silver wire nonsense) for these. But I do notice a difference when I unplug the ground shunt from the HDPLEX 160 which powers the downstream FMC and the LPS-1 and an iSPDIF Purifier for the TV. Shunts for all three run into the same socket which is plugged into a dedicated circuit along with the HDPLEX. Unplugged lots more sibilance.  

[BigGuy]

35 minutes ago, thyname said:

 

Great explanation!

 

What about the scenario when no switch is involved (audio streamer connected directly to router), would grounding the router would take care of everything in the house, including another location that has a Switch? Or does this just work with the two FS105 and FS108 models you tested? Thanks!

 

16 minutes ago, charlesphoto said:

Interesting. I'm currently using a GS108NA with the ground shunt and powered by an "El Cheapo" Lps. It has other traffic on it though than just the hifi. I have an eBay FS105NA for $10 coming so will isolate the streamer and server on that with a ground shunt and see what happens.

 

As far as I'm concerned the grounding I did upstream hasn't hurt, esp as I had a bunch of extra adapters and it's really easy to do. I'm quite sensitive to EMF so getting the overall level down is always good. That said I wouldn't go out of your way or break the bank (i.e. silver wire nonsense) for these. But I do notice a difference when I unplug the ground shunt from the HDPLEX 160 which powers the downstream FMC and the LPS-1 and an iSPDIF Purifier for the TV. Shunts for all three run into the same socket which is plugged into a dedicated circuit along with the HDPLEX. Unplugged lots more sibilance.  

Great minds think alike!  I have cable service moden feeding a Linksys 1000 router that feeds my music server PC.

 

What is the purpose of a "switch" vs my Linksys router?

 

Do I just connect the switch between the existing router and PC?

[lmitche]

The purpose of the new Netgear switch, properly grounded, is filtering low and high impedance leakage currents from your music PC.

[BigGuy]

11 minutes ago, lmitche said:

The purpose of the new Netgear switch, properly grounded, is filtering low and high impedance leakage currents from your music PC.

I see that there is a screw GND terminal on the back of the Netgear switch. 

 

Does that terminal need to be connected to ground using John's umbilical (modified with spade connector, etc.) IN ADDITION TO grounding the Netgear power supply?

 

At this point I think the only thing not grounded is me!  ;-)

[lmitche]

I should have written that the new Netgear switch, properly grounded, is filtering low and high impedance leakage currents created by your SMPS and the rest of your network from your music PC.

 

I believe that John wrote that the ground strap on the switch is irrelevant.

[Forehaven]

You know what so weird, one of my prev. working SMPS ground device - Johns design.  Now it's not powering the LPS.  The stock dc cable powers it up fine.  I have 3 other grounded SMPSs that all are working.  Isn't this plain bizarre?

[JohnSwenson]

4 hours ago, thyname said:

 

Great explanation!

 

What about the scenario when no switch is involved (audio streamer connected directly to router), would grounding the router would take care of everything in the house, including another location that has a Switch? Or does this just work with the two FS105 and FS108 models you tested? Thanks!

What I have found is that most wired Ethernet devices (switch or otherwise) do NOT block leakage from other connections passing through them, even if their power supply is grounded. The exception seems to be a few switches, that when their power supply is grounded block leakage from other devices.

 

Thus grounding the PS to a router is probably NOT going to block leakage from other devices plugged into it.

 

If you cannot use one of the specified switches as the last point in front of your audio endpoint, then your best bet is to add one, they are not very expensive, just make sure you ground the power supply going into it. This arrangement will make sure that whatever leakage is going on in your network will not make it through into your audio system.

 

John S.

[JohnSwenson]

3 hours ago, BigGuy said:

I see that there is a screw GND terminal on the back of the Netgear switch. 

 

Does that terminal need to be connected to ground using John's umbilical (modified with spade connector, etc.) IN ADDITION TO grounding the Netgear power supply?

 

At this point I think the only thing not grounded is me!  ;-)

The ground screw terminal on some Netgear switches is JUST connected to the metal box, it is NOT connected to the groundplane on the PCB. The groundplane is what needs to be grounded, so just grounding the screw does NOT improve leakage. The only way to be sure you are grounding the groundplane on the PCB is to ground the negative of the power supply, this is always connected to the groundplane.

 

John S.

[austinpop]

32 minutes ago, JohnSwenson said:

If you cannot use one of the specified switches as the last point in front of your audio endpoint, then your best bet is to add one, they are not very expensive, just make sure you ground the power supply going into it. This arrangement will make sure that whatever leakage is going on in your network will not make it through into your audio system.

 

I have done exactly this. The FS105NA is now acting as an Ethernet isolator, essentially, immediately upstream of my chain.

 

I know John is saying that with this arrangement, you are essentially immune to noise from whatever is upstream of the FS105 switch. I would love for that to be true, as it would be nice to create a barrier (Build A Wall!) immediately upstream of your audio gear, and not worry about grounding every single network device in your house, or putting them all on LPSes!

 

I just have not tested this hypothesis. Yet. Coming from John's measurements, I have high hopes that it will pan out.

 

Now I'd love to see how his experiments with clocking and signal integrity on Ethernet go!