ATX Power, Earth Ground inbalance?

[elecard]

hello I hope you are all having a good day.

I have noticed millivoltage readings from the black ground wire of my ATX PSU to the Chasis of my computer and I think it is the cause of some very noticable interference.

I am not sure if this is a Common ground, inbalance issue?. Please forgive my lack of knowledge. I am hoping to learn and maybe share some findings with others.

A full ATX system is installed, the mother board is connected to the chassis via quite a few bolts/standoffs, each
mount has been tested for continuity before connecting up the (Earthed IEC ATX) PSU, so to avoid confusion. So the parts are connected metal to metal. ground plane of board to PC chassis.

A reasonably high quality PSU is then installed to an unpainted aluminum chasis. by the 4 conductive bolts.

test 1:

with PC on , when i measure with my millivolt meter from the black molex ground from the PSU cable, to one of the 4 bolts of the PSU/chassis, i read:

3.3mv

I have interference during the above test.

test 2:

I have disconected the PSU from the chassis, the PSU is still in the exact same place but just not bolted on, everything else is the exact same, just that all the 4 bolts removed, i read:

0.7mv

i then have no interference in the above test. So I have no interference when the PSU is physically disconneted from the chassis.

I know there a a fair few things that can cause audio interference issues. I am quite good at audio and am only listening to the headphone out, with nothing else connected, even when i dicconect mouse, usb devices, monitor = everything disconnected. and just headphones, listenng while standing on insulated card and soft rubber shoes.

However I notice this millivolt reading is high when there is interference, and the interference vanishes when i unbolt the PSU.

I just hope someone may be able to explain what is going on. I would like my PSU to be bolted on. and Earth to chassis.

Perhaps i just need a more balanced PSU? Maybe one that is designed for sensitive work? Maybe electronics engineers have special PSU's for PC oscilloscopes?

Thank you for any help.

 

[Tony Stewart]

If you are getting audio noise from the ground wire current measurement, then chances are your measurement wires are radiating the SMPS pulsing currents that are supposed to be suppressed by the Y caps to ground. There should be no DC voltage unless rectified by measurement errors.

All good ATX PSUs have some sort of line filter that may look like this.

 

[rjenkinsgb]

I am quite good at audio and am only listening to the headphone out, with nothing else connected

Are you connecting the headphones directly to the motherboard, or using the case front link cable & socket?

The case cable may be grounded at the front panel, causing a hum loop?

 

[elecard]

thank you very much for the replies.

it is a PCI slot sound card, the noise is interfernece, like mouse movements, windows being opened, rendering. even with the monitor disconnected.

But all this vanishes and the headphones become nice and clean sounding when the PSU is simply phsically disconnected from the chassis.

"If you are getting audio noise from the ground wire current measurement,..." the noise is from the headphones connected to the audio card's headphone output, but only when the PSU is connected to the chassis.

The PSU has Earth, but also a black wire ground (molex)and so I presumed the millivolt reading may be related. I am sorry if this reply is repeating. I am just trying to rephrase.

I thought perhaps there may be a way to make the two grounds be less conflicting if that is what is happening? tuned, or balanced better?

 

[rjenkinsgb]

That is a very common problem with desktop PCs; you will get that type of noise on most machines when using headphones.

With external speakers, the PC volume can be turned high & the speaker volume lower, which usually reduces the noise so its below ambient level.

There is no solution that I am aware of, other possibly than trying many different motherboard + PSU combinations..


A decent USB sound card or interface is the simplest way around the problem.

A basic, cheap USB one would likely work far better than an internal PCI or mother board sound device - or if you want the ultimate quality, get a Focusrite Scarlett 2i2 or similar; the headphone quality (s/n & distortion levels) on one of those is better than typical "Audiophile" headphone DACs costing £1000+

 

[ZipZapOuch]

If it only happens at high current draw of the power supply, I'm guessing the there is an Current x resistance = voltage created across the wire that connects your 0v (black) "ground" to chassis ground. You may be able to attenuate by reducing the resistance of that connection by adding a second, thicker, wire between the two points. Make sure you have a nice firm connection to chassis because the little star washers typically used in the screw don't promote a lot of current flow.

If you try it and it works, please let me know.

 

[elecard]

Thank you both.

note. This is a very expensive card. Low latency pci-e. and so one would expect it to work inside a PC chasis. I shall not name and shame, just yet. i thnk i am right in holdiong back.:/

zipzap you mention "If it only happens at high current draw of the power supply,..."

it happens also when on the desktop with nothing really going on, just a window open, we hear graphics being drawn. The window open makes interference noises, not really hiss , more machine 01 01 type noises. But it all goes away when the PSU is detached.

"I'm guessing the there is an Current x resistance = voltage created across the wire that connects your 0v (black) "ground" to chassis ground. You may be able to attenuate by reducing the resistance of that connection by adding a second, thicker, wire between the two points"

Do you mean to add a thick copper wire from black molex to chasis, I have tried that with a firm connection and it was still the same interference issue

rjenkins you mention "There is no solution that I am aware of, other possibly than trying many different motherboard + PSU combinations.."

Owch. that could be expensive!

I hoped to use the headphone port as it was just one reason I bought this expensive card.

thank you all for your help.

 

[ZipZapOuch]

Thank you both.

note. This is a very expensive card. Low latency pci-e. and so one would expect it to work inside a PC chasis. I shall not name and shame, just yet. i thnk i am right in holdiong back.:/

zipzap you mention "If it only happens at high current draw of the power supply,..."

it happens also when on the desktop with nothing really going on, just a window open, we hear graphics being drawn. The window open makes interference noises, not really hiss , more machine 01 01 type noises. But it all goes away when the PSU is detached.

"I'm guessing the there is an Current x resistance = voltage created across the wire that connects your 0v (black) "ground" to chassis ground. You may be able to attenuate by reducing the resistance of that connection by adding a second, thicker, wire between the two points"

Do you mean to add a thick copper wire from black molex to chasis, I have tried that with a firm connection and it was still the same interference issue

rjenkins you mention "There is no solution that I am aware of, other possibly than trying many different motherboard + PSU combinations.."

Owch. that could be expensive!

I hoped to use the headphone port as it was just one reason I bought this expensive card.

thank you all for your help.

One last thing would be to not use the chassis and connect your black as close to the green ground where the cord enters the ATX box (if you can do it safely). The steel of the new chassis is very thin and not a good conductor.

Not that even a small screen update can draw a quick pulse that may be audible click as your PC ground shifts.

Also, you could try a larger wattage, higher frequency power supply.

 

[rjenkinsgb]

I hoped to use the headphone port as it was just one reason I bought this expensive card.

I've not yet been able to get a desktop PC to give perfect noise-free headphone audio, despite building them since 286 days & using more sound systems than I can count, since the Soundblaster etc. first appeared.

For anything other than basic desktop sound via small speakers, I use external interfaces exclusively.

The on-board sound system can often be better than an add-on card, as the motherboard designers can compensate for ground currents to some extent.


Just a thought - your plug-in card may work better if you isolate the rear bracket from the case? eg. Cover it in Kapton tape, including through the screw slot, and use a nylon washer on the retaining screw to prevent that cutting in to the tape.

That should eliminate any ground loop between the motherboard & case via the card?

 

[elecard]

Thank you both again.

zipzap you mention: ".The steel of the new chassis is very thin and not a good conductor." Good point, I never thought that it might not be! it is quite thin.

"to not use the chassis and connect your black as close to the green ground where the cord enters the ATX box (if you can do it safely)."

i did connect it to the bolt near the IEC with thick copper to no avail. Do you mean to open the atx psu and solder a thick copper wire from the black wire (or pad from the PSU board) to the green Earth of the PSU inside the PSU? But leave earth connected to the PSU chassis also?

"to not use the chassis" , do you mean to float the motherbaord and all parts off the chasis completely. but leave the 4 PSU bolts to allow Earth connection to the PC chassis?


rjenkinsgb you mention: "Just a thought - your plug-in card may work better if you isolate the rear bracket from the case?"

yes indeed. and thank you, this also works,(like disconecting the PSU does) I just hoped there would be a way to solve the issue without moding the card itself with tape, I know this sounds weak of me, but it would devalue it to have to sell it with that description. Normally i would not mind. I just hoped to see if anyone had a solution like a power conditioner, perhaps? or a special ATX PSU that has some type of ferrit core thing or is balanced at the factory for sensitive gear.

"Also, you could try a larger wattage, higher frequency power supply" that might be worth a try.

It seems it is down to luck, if the PSU and motherbaord play well together, however

one might think that a manufaturer would produce a combination for people who like to use audio.
headphones and unbalanced. gigabyte partnership with corsair or alike.

it might help with other devices too, to have a clean PSU ground. i wonder if it might help with overclocking or other sensitve card and adaptors?

thank you all.

 

[Nigel Goodwin]

rjenkinsgb you mention: "Just a thought - your plug-in card may work better if you isolate the rear bracket from the case?"

yes indeed. and thank you, this also works,(like disconecting the PSU does) I just hoped there would be a way to solve the issue without moding the card itself with tape, I know this sounds weak of me, but it would devalue it to have to sell it with that description.

If isolating the rear bracket works, then it's most probably down to an earth loop - so that's probably the only cure.

How about making a plastic bracket to replace the metal one?.

 

[rjenkinsgb]

I just hoped there would be a way to solve the issue without moding the card itself with tape,

A layer of kapton tape on the metal bracket could just be peeled off again, it you ever want to sell it? It's not a permanent change.

Or, tape around the inside of the case rear slot (plus nylon washers or a plastic screw), to give the same isolation?

 

[ZipZapOuch]

I would add a fat wire from the IEC connector to the mess of black wires (bottom of board. Get a nice form connection so the current isn't flowing through the sheet metal to get to the IEC connector. Typically, the "black" 0V connects to the chassis with the mounting screws in the corners of the PCB. This is typically a poor connection. Since there will be very little current ( current pulses) going though sheet metal, you can leave the thin little wire already connected to the chassis in place and not change it for a larger wire.

Ideally, you can remove the four mounting screws on the bottom (that hold the PCB in place and make the PCB to chassis contact, put a piece of black electrical tape to prevent those contacts and reassemble. You may have to drill out some holes or be creative by scraping copper off the perimeter of the PCB (up to you). In any case, all of these individual contacts are creating multiple patches to ground and a possible ground loop. It would be better to have just one.

 

[Tony Stewart]

It's hard to visualize which carrier frequency is interfering and where it gets injected but the effect is to demodulate the RF into the audio band often from diode law in front-end overvoltage protection rectifies the AM noise to audio.

Some things to try

- Short audio inputs enabled.
-Replace the black wire with various values of RLC using anything you have for L like a transformer and various C's ( @JRW's Kapton solution would be xx nF..)
- These combinations will tell us some idea from impedance attenuation.
- Video cables normally have Ferrite Baluns on them to reduce unbalanced 75 ohm radiation, so proximity of mic or headset wires can also serve to locate the source of the noise. If you have a high mu torroid core , you can experiment with that on the headset cable.
- We are assuming your earth ground is also OK to provide a much lower impedance than free space. - The ground leakage currents if in spec ought to be much less than 2.5 mA over the whole spectrum of SMPS noise.

- Using RLC variations instead of the black wire, allows us to imagine the coupling impedance for half volume and adequate suppression. Then you might figure out a better solution.
- be sure to rule out radiated noise before assuming it is purely conducted noise. They can be related.

Anecdotal.
I once had a similar noise problem using the front panel jack headset. The problem turned out to be the SMPS radiating to the front panel which had no good ground connection to the hinged front door. Adding a short flat braid to the front panel cured that noise.

 

[rjenkinsgb]

It's hard to visualize which carrier frequency is interfering and where it gets injected but the effect is to demodulate the RF into the audio band often from diode law in front-end overvoltage protection rectifies the AM noise to audio.

It's not as complex as that.

The design of the sound card, with its metal bracket connected to the edge connector ground, means it is providing an alternate ground return path for the (tens of amps) motherboard power.

As the CPU & other current varies, the point to point voltage across the card vary and some of that is being superimposed on the analog stages that feed the outputs.

The effect is reduced or eliminated just by breaking the sound card chassis return path to the PSU.

It's a problem that's been known for decades & the card is just badly designed, with the mounting bracket being connected to the audio ground rather than being electrically isolated.

 

[Tony Stewart]

I have not heard of this "decades-old" problem before. Surely the designers would not share Pwr Gnd with Agnd for the DAC output.

I wonder which card, PSU and case are we talking about?
I assume it is a plated steel case, not aluminum.
Links would be helpful.
Which ground did you use for the -ve lead?

audio card?

I've not yet been able to get a desktop PC to give perfect noise-free headphone audio,

@rj Can you record your audio spectrum on Audacity's spectrum display? or on an audio file? I was wondering what your expectations of noise-free are. > 60 dB SNR, > 80 dB? or?

 

[rjenkinsgb]

Can you record your audio spectrum on Audacity's spectrum display?

Not easily, as Audacity on this machine will record from the DAC feed not the analog socket.
(I consider an acceptable noise level to be when it's imperceptible with the volume at a comfortable listening level).

For serious stuff I use external interfaces - S/PDIF optical for my media player PC, Presonus gear for my music workstation, or a little Focusrite interface for testing gear & portable stuff.

 

[elecard]

Thank you all for your great responses.

yes, we suggested they should separate the card from the chassis. with either a sleeve or a plastic bracket.

I am not keen to share the card's manufacturer name until they respond. I suppose, that seems fair.

Thank you for your suggestions on modding the PSU i wonder why the PSU manufactures do not solder IEC earth to black ground, so they are the same? That makes sense.

I imagine the audio card designer was working from a bench environment, with a high quality PSU with only one ground and amazingly did not test a good selection of ATX PSU's and chassis's to see what effects their would be :?

So, we are still awaiting a response from them.

We very much appreciate your responses and they have helped us allot and we shall respond with news, as I am sure other people have similar issues with other cards, maybe not just audio cards? Just stability in general?

Thank you all.

 

[ChrisP58]

A simple way to test this would be to temporarilly remove the bracket from the card. You'll need to be careful when plugging cables in and out to not dislodge the sound card from it's socket, but it should be stable enough to verify if the chassis ground loop is the culprit.