Re: [LAU] ground-loops -- was Re: USB audio interfaces >= 8

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: <linux-audio-user@...>
Date: Wednesday, May 16, 2012 - 6:01 pm

Replying to a digest, sorry if that screws up mail threading.

From: Robin Gareus

You have found that there are ways to design balanced inputs improperly.
The Audio Engineering Society devoted an entire issue of the journal (June
2005) to such issues.

> A multimeter shows a constant 1mA current and ~3-4mV AC potential.

Should not be a problem for properly designed balanced inputs and outputs,
can be for improperly designed equipment.

> /me is pondering to cut the ground-wire from the screen..

Could cause other problems. Probably your choices are either to modify
the equipment so it is not quite so improperly constructed, use a
transformer to isolate the equipment from the shield current, or possibly
to construct a cable which works around the offending equipment.

From: Fons Adriaensen

Probably what happens is that when the screen is not connected, there is a
high impedance voltage difference between the devices, and when the
shields are connected together (through the ammeter) current actually
flows, but is so low that the ammeter does not measure it correctly.

When the VGA monitor is connected, the monitor has low enough impedance
leakage path that it can source a couple of milliamps through the shield,
and the USB interface or speaker (or both) has a common impedance path
for that current to flow on the reference potential node of a high gain
stage and amplifies the noise current.

From: Robin Gareus

Verify the wiring, make sure that there are no broken conductors or broken
solder joints in the shield connections of the cables between the audio
interface and the speakers.

Are you using TRS-TRS cables, or TRS-XLR cables for the connection from
audio interface to speakers?

Which of the equipment (video monitor, computer, active speakers) have
three wire power connections with safety ground connected, and which (if
any) have only two wire connections. Equipment in the US is typically
double insulated and so does not require a safety earth connection, but I
don't know if that also applies on equipment shipped for use on European
240V power distribution.

I ask because you need to give the ground current someplace to flow to
complete the circuit which does not flow across the reference potential
node (sometimes called the "ground connection") of a high gain stage.
Could be in the output of the audio interface, or the input of the
speaker, or both, where that current is being converted to audible noise.
One way to do that is to make sure the shield of the cable connects well
to the shield of the connector. You generally have to connect the cable
that way in a TRS connector, but XLR connectors do not by default have a
connection between pin1 and the connector shell. The equipment should
connect pin 1 to a low impedance equipment shield connection internally,
but many designs do not. In that case connecting pin 1 to the connector
shell inside the connector can sometimes help.

If some of the equipment has an earthed chassis and some does not,
sometime making an external connection (using wire or copper braid)
between the different chassis can reduce the potential difference enough
that the current flowing on the audio cable becomes low enough to be

Something like one of these would probably help:

But good transformers are not inexpensive, it might be the same cost to
get an audio interface which did not inject so much noise current into the
output. If the speakers are causing the problem and not the audio
interface, then you may be able to make some modifications to the input
connections of the speaker interface to solve the issue. Depends on how
the amplifier assembly is physically constructed.

Chris Caudle

Linux-audio-user mailing list

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
Re: [LAU] ground-loops -- was Re: USB audio interfaces >= 8, Chris Caudle, (Wed May 16, 6:01 pm)