• TalkBass has been independent since 1998. Add your voice.
    Create a free account to reply to discussions, view embedded media, and browse with fewer display ads.
    Join free Log in
    Want zero display ads or expanded classifieds tools? Compare plans.

An electrical and electronical engineering question

#29
icarus88 said:
Most of the GPU noise is being directly injected into the ground plane.
Click to expand...
Having a ground plane may be part of the issue. Designing gear with sensitive analog circuitry and dirty stuff is not at all new - I designed some mic mixers, which have gains on the order of 90 dB (voltage wise, and the signal at the outputs has current gain on top of that), compared to the input signal - anything out of place, and you know about it right away. While we didn't have digital circuitry in those products, we did have a power supply that ran off line voltage - lots of nasty currents from rectifiers, flowing into and through capacitors, etc.. The trick in keeping noise from messing up analog stuff is to think differently - that current has to flow from one place to another, and while it may need to connect with the analog grounds at some point to nail the voltages together, you don't want that current flowing in the dirty path to run along with the analog signal on the same conductor for any distance at all - you essentially set up two ground systems, and then carefully connect them at one (well chosen) point from each system. A ground plane approach helps keep the ground path low resistance wise, but connecting the analog ground(s) at random points within a ground plane that has digital circuitry connected to it is pretty much a recipe for a problem.

When you go to lay out a pc board for this kind if thing, you need to run off the autorouter when doing the ground stuff - YOU need to think about where the dirty currents are flowing - leaving it up to the machine to connect all the points on one circuit node means it'll mess it up big time. I would also lay it out as two separate ground systems, and (on a prototype) use a wire to connect them together - that'll let you learn where you want to do that. You can try to guess, and I could probably guess correctly, but if you've never done this before, your guess won't be right the first time.

And no, software can't fix this kind of problem. It has to be fixed during the design of the gear - which may be too late for your situation.
 
Last edited:
  • Like
Reactions: WRM and icarus88
#30
micguy said:
Having a ground plane may be part of the issue. Designing gear with sensitive analog circuitry and dirty stuff is not at all new - I designed some mic mixers, which have gains on the order of 90 dB (voltage wise, and the signal at the outputs has current gain on top of that), compared to the input signal - anything out of place, and you know about it right away. While we didn't have digital circuitry in those products, we did have a power supply that ran off line voltage - lots of nasty currents from rectifiers, flowing into and through capacitors, etc.. The trick in keeping noise from messing up analog stuff is to think differently - that current has to flow from one place to another, and while it may need to connect with the analog grounds at some point to nail the voltages together, you don't want that current flowing in the dirty path to run along with the analog signal on the same conductor for any distance at all - you essentially set up two ground systems, and then carefully connect them at one (well chosen) point from each system. A ground plane approach helps keep the ground path low resistance wise, but connecting the analog ground(s) at random points within a ground plane that has digital circuitry connected to it is pretty much a recipe for a problem.

When you go to lay out a pc board for this kind if thing, you need to run off the autorouter when doing the ground stuff - YOU need to think about where the dirty currents are flowing - leaving it up to the machine to connect all the points on one circuit node means it'll mess it up big time. I would also lay it out as two separate ground systems, and (on a prototype) use a wire to connect them together - that'll let you learn where you want to do that. You can try to guess, and I could probably guess correctly, but if you've never done this before, your guess won't be right the first time.

And no, software can't fix this kind of problem. It has to be fixed during the design of the gear - which may be too late for your situation.
Click to expand...

I'm talking theoretically, I should probably devote all my time to this, too much a recipe for a problem.

So which brand of gpu have the design you mentioned? Nvidia? ATI? old intel?
 
Last edited:
#33
icarus88 said:
A simpler approach, Does downvolting and downclocking the GPU reduce background noise? Or is it unnecessary?
Click to expand...



tested on cpu, gpu would be similar ( my gpu is intel hd3000, not accessable)or more. it has an effect but it is negligible. But this can have different results in different systems. It may have a significant effect in your system, it may not work for me.
 

Attachments

  • 1727309556080.png
    1727309556080.png
    32.5 KB · Views: 8
  • 1727309565976.png
    1727309565976.png
    42.2 KB · Views: 4
Last edited:
#34
Wow, this is a truly bizarre thread. Brutally using the ground connection? That never came up when I was working as an EE :whistle:

I have no idea what computer you are using but my personal experience has been that many Windows machines have grounding problems when using USB audio interfaces. Once I was using a Windows laptop and got ridiculous ground loop noise so I disconnected the AC supply and ran it on battery power…voila, the noise was gone. Macs are better behaved when it comes to proper grounding, but I have experienced ground loops sometimes even with a Mac.

An isolation transformer usually does the job but you still have to figure out what needs to be isolated by trial and error, disconnecting devices until the noise goes away, checking grounding of AC outlets, etc.
 
  • Like
Reactions: Crater and icarus88
#35
brianrost said:
Wow, this is a truly bizarre thread. Brutally using the ground connection? That never came up when I was working as an EE :whistle:

I have no idea what computer you are using but my personal experience has been that many Windows machines have grounding problems when using USB audio interfaces. Once I was using a Windows laptop and got ridiculous ground loop noise so I disconnected the AC supply and ran it on battery power…voila, the noise was gone. Macs are better behaved when it comes to proper grounding, but I have experienced ground loops sometimes even with a Mac.

An isolation transformer usually does the job but you still have to figure out what needs to be isolated by trial and error, disconnecting devices until the noise goes away, checking grounding of AC outlets, etc.
Click to expand...
He’s looking for a way, in software, to snip a lead or a trace on his hardware, from some component that he thinks is brutalizing the ground path.
Do not miss the software part!
 

Latest posts

Top Bottom
Back