Reducing EMI on USB micro connector

[Pack34]

I went through some pre-compliance testing today and I ran into a bit of a snag on the USB circuit. It passed, but just barely.   The emissions were clearly related to the amount of usb traffic.

I'm going to rev the design to add a choke on the USB connector, but I need something permanent and professional to use as an EMI gasket. For testing I just took a fabric EMI gasket and cutout a window for the USB connector. This may have worked in testing but it would not be feasible when trying to build these in bulk.

Does anyone have any suggestions? I've seen ones like in the link below but they're for the original USB-B connector and not the micro-B.

http://www.digikey.com/catalog/en/partgroup/97-728-usb-connector-gasket/33442

[wraper]

probably show a part of schematic /andor layout instead (if allowed to do so).

[rob77]

check this one - you can find some hints there:

http://www.ti.com/sc/docs/apps/msp/intrface/usb/emitest.pdf

[Pack34]

The current design has the USB data lines wired straight to a USB switch that then goes to a USB High Speed PHY for an STM32.

Currently, there are no series termination resistors, no choke, and only a bead on the ground of the USB connector.

I'm adding those, but I'd like to also add a gasket. It worked well on a USB-B connector but I'm having troubles finding a similar solution for micro-B.

[T3sl4co1l]

What would you be gasketing the connector to?  Why isn't it grounded to that already?

Please tell me you didn't fall for the  of ferrite beads on supply/ground/shield pins...

Tim

[Pack34]

The gasket is connecting the shield of the USB connector directly to the metal around the chassis. It was suggested by the compliance engineer and it seems to work. The shielding of the connector is directly connected to the ground plane.

When noise is a concern I typically connect the ground and vdd through ferrite beads.

Both of these are supported by the intel USB EMI design guide.

http://www.ti.com/sc/docs/apps/msp/intrface/usb/emitest.pdf

[wraper]

and only a bead on the ground of the USB connector.

The shielding of the connector is directly connected to the ground plane.

Ferrite bead on GND pin or shield  ? GND pin certainly should be connected to the ground plane directly. Separating the shield by ferrite bead is very questionable measure either, and IMO shouldn't be done unless you connect it to conductive enclosure by conductive gasket or by some other means.

[Pack34]

and only a bead on the ground of the USB connector.

The shielding of the connector is directly connected to the ground plane.

Ferrite bead on GND pin or shield  ? GND pin certainly should be connected to the ground plane directly. Separating the shield by ferrite bead is very questionable measure either, and IMO shouldn't be done unless you connect it to conductive enclosure by conductive gasket or by some other means.

The dedicated ground pin on the connector has a bead connected to ground, but the dedicated shielding pads are filled directly to the ground plane.

[wraper]

The dedicated ground pin on the connector has a bead connected to ground, but the dedicated shielding pads are filled directly to the ground plane.

Do you understand that you put a ferrite bead in the middle of return path of the signal? USB is not 100% differential. Quiet likely this ferrite bead is the cause of your EMI problems.

[georges80]

I had a product with 8 USB (4 double connectors) along with 4 HDMI, gigabit ethernet and external 12V DC power. Passed compliance fine. No 'magic' beads or other goat blood needed.

What you DO need to deal with it escaping RF energy from your PCB and from 'slots' in your housing that are transparent to RF energy. One method around the edge of the PCB is fence posting via's. Space them well under the wavelength of the RF you are worried about. Make sure you have ground vias to stitch return energy due to layer transitions of high speed differential signals. I'm presuming this is at least a 4 layer PCB? I presume you've also layed out the USB differential pairs with appropriate impedance matching if they go any significant distance.

Make sure you use the BEST possible USB cables during your compliance test. We had awful precompliance emissions that turned out to be cheap HDMI cables - the emissions were right at the connector end of the cables. I'd take a variety of quality/name brand USB cables along so you can switch them if needed during the test.

After working with an RF/Microwave engineer I began to visualize our PCB layout as possible microwave/waveguides/slots - that is what you need to do to 'find' the possible emission sources.

cheers,
george.

[canada13records]

I agree with George. Every product I work on has a USB connector (or other high speed interface - HDMI) and none of these use strange mesh or whatever you are referencing.

I would start by analyzing the grounding of your system. Make sure that your USB host has strong ground path returns, and the it isn't relying on a couple of vias for a connection to the USB jack. To me it sounds like your system is emitting ground noise through the high impedance USB connection.

I would then make sure you have via stitching on your board, as well as a nice fence around the edges of the board, and the USB differential pair if space allows (probably overkill).

I would also ensure that you have code or a circuit that detects whether or not a device is connected. Take a look at the mbed schematics for an idea. You need to make sure your MCU isn't switching the lined that aren't terminated.

Lastly, make sure you use termination to reduce stub length. Some 33 resistors are pretty standard if you don't have the ability to calculate impedance.

The only time that I use ferrets on a USB connector is when I have a USB powered device, and then I use them to help prevent ground noise from coupling between the two circuits.

[T3sl4co1l]

The only time that I use ferrets on a USB connector is when I have a USB powered device, and then I use them to help prevent ground noise from coupling between the two circuits.

Simple typo I know, but I can't help but contemplate the mental image...



Tim

[Pack34]

Just to swing back in case anyone else found themselves in the same situation.

The EMI gasket I was referring to previously was this: http://www.digikey.com/product-detail/en/wurth-electronics-inc/3020503/732-3102-ND/2626041

I was using it as a "crutch" because a traditional EMI gasket like the one below does not exist for the smaller USB connectors.
http://www.digikey.com/product-search/en?mpart=0097072802&v=903

Using an added EMI chip and an actual shielded USB cable did wonders. I found the difference between the USB cables to be incredible. The monoprice ones ended up performing the best out of all the ones I tried. Although I ended up passing by 10-dB, there's definitely something more I can do to improve emissions. Using a comparable system running at USB-FullSpeed instead of HighSpeed, the peak noise was about 20-dB less.

I'd still like to find a good SMD EMI finger that I can place adjacent to the USB connector so that I have a guaranteed connection between the shielding of the USB connector and the metal enclosure around the cut-out for the USB connector. Just something that could extend off of the board-edge by about 1.0-mm