2. tech-talk
  3. 1328289757

Shielding components from EMI/RFI... Help please

A recent experiment with a product designed to reduce EMI/RFI left me curious about other ways to reduce EMI/RFI in my system. In the past ten days, I've stepped onto a slippery slope, at the bottom of which is surely some kind of insanity...

I've been experimenting with copper plates in an effort to absorb, deflect, diffract, and block EMI/RFI. I've tried copper plates under components, on top of components, and inside components.

This is the point where you tell me I don't know what I'm doing and I'm likely to short circuit something and/or electrocute myself. Consider me duly warned. This is also the point where you tell me to get some balanced interconnects, or at least to get some shielded interconnects for Chrissake. Consider me duly informed. Moving on...

I'm hoping you can help me make the most of this experiment, and help me avoid killing a component or myself. My strategy so far has been to:

1. Place copper plates at locations that generate a lot of EMI/RFI, e.g., components with switching mode power supplies or high frequency clocks. The system has a total of 3 SMPS and 3 clocks.

2. Place copper plates at locations that are vulnerable to EMI/RFI, e.g., under the amp, near the transformer.

3. Place copper plates inside noisy components -- in particular, my Meridian G68 preamp/processor. I've begun to build 2 partial Faraday cages, one for the SMPS, and one for the analog output stage.

4. Ground the copper plates either to the component chassis (when plates are used inside a component) or to an independent ground point (when plates are used above/below a component).

Has anyone tried this sort of thing?

Bryon
bryoncunningham
almarg's avatar
Ag insider logo xs@2x
almarg
9,670 posts
 
Hi Bryon,

My suspicion would be that the connection from the power supply is connecting the ground on the DC output side of the supply to chassis, thereby connecting the G68's circuit ground to chassis. That is often done, although as you surmise it can result in a ground loop between the G68 and the components it is connected to.

I wouldn't play around with it, though, unless there is a particular reason to do so, and an understanding of why it was done that way. Other possible approaches, btw, besides simply removing the connection, being to connect circuit ground to chassis via either a resistor, a capacitor, or an inductor.

Keep in mind, also, that balanced connections have much less susceptibility to ground loop issues than unbalanced connections. I assume that you have a balanced connection to the Pass amp, and perhaps also a balanced AES/EBU connection from the re-clocker.

Best,

-- Al
bryoncunningham OP
890 posts
 
My suspicion would be that the connection from the power supply is connecting the ground on the DC output side of the supply to chassis...

You may be right, Al, although the ground wire on the power supply is right next to the power INPUT. Here are some pics...

Modded PSU photo 1

Modded PSU photo 2

Modded PSU photo 3

Stock PSU

The power input is the blue/red wires in the top left of photo 1. The ground is the green wire in the bottom left of photo 1.

Can you tell from looking at those pictures whether the ground is on the AC side or the DC side? Can we infer from its physical location that it is on the AC side?
almarg's avatar
Ag insider logo xs@2x
almarg
9,670 posts
 
Hi Bryon,

Yes, you're right. After looking at the photos I believe that the two small blue objects just above the green wire are capacitors that are provided for purposes of filtering high frequency noise that may be present on the incoming AC. One capacitor is probably connected between AC hot and chassis, and the other capacitor between AC neutral and chassis, with the green wire providing the connection to chassis for both capacitors.

Ergo, not a problem!

Best,
-- Al
bryoncunningham OP
890 posts
 
Thanks, Al. I will leave both ground wires in place. There is something else strange to report...

I installed a shielded Cat6 cable today, replacing an unshielded Cat5 cable. I expected to hear no difference whatsoever. Much to my surprise, there was a BIG difference. I know that sounds crazy. I don't understand it at all.

I know that the shielded cable might keep out EMI/RFI, but that doesn't seem to explain what I'm hearing, which is significantly more resolution. It's dramatic enough that it seems difficult to explain as simply less EMI/RFI. The difference sounds similar to the kind of resolution jump you get when you go from 16/44.1 to 24/96. In other words, it sounds like more information is reaching the system. But how could that be, if the ethernet protocol utilizes error detection/correction? I'm confused.

The old configuration was...

iMac -> 50' unshielded Cat5 -> Apple Airport Extreme -> 50' unshielded Cat5 -> Sonos -> S/PDIF -> Reclocker -> S/PDIF -> Meridian G68

The new configuration is...

iMac -> 50' unshielded Cat5 -> Apple Airport Extreme -> 25' shielded Cat6 -> Sonos -> S/PDIF -> Reclocker -> S/PDIF -> Meridian G68

So only ONE link in this chain changed: a 50' unshielded Cat5 was replaced with a 25' shielded Cat6. And the result was a big jump in perceived resolution. I'm scratching my head...

bc
almarg's avatar
Ag insider logo xs@2x
almarg
9,670 posts
 
Hi Bryon,

Presumably the shielded cable is reducing the amount of rfi/emi that escapes from the cable and couples onto circuit points elsewhere in the system, most likely points that are "after" the reclocker circuitry. There are probably a number of ways in which that could occur, despite the shielding you have incorporated into the G68. One possibility would be that it couples onto the power wiring, and makes its way from there onto the grounds within the G68. Noisy grounds in the digital section could (and probably would) result in increased jitter. Noisy grounds in the analog circuitry would have unpredictable but conceivably significant consequences at audible frequencies, even if the noise itself is at rf frequencies (as a result of intermodulation, AM detection, or other effects that can cause rfi to affect audible frequencies).

That's my speculation, anyway.

Best,

-- Al
More to discover