You'll have to check the IEE 802.3 spec for a definitive answer, but I suspect 47pF is likely to be too much - have you tested it at 100Mbps with a long cable? If you can't get hold of the IEE spec I can email you a copy but it's
15MB.
Looking at the photos from your previous post, it's not clear of the RJ45 jack on the SBC is being used directly through the side of the case. If you can, use a short male -> female RJ45 extender lead to take the connection out of the case, and fit it with the biggest (solid) ferrite bead you can get. This won't have any effect on the differential signals in the cable, but will attenuate any common-mode RFI produced by the board. I had to do this on a previous product which used an SBC. It does mean you lose the LINK and ACT LEDS but we decided we could live with that.
Getting products based on these SBCs to pass EMC regs is, as you are discovering, an absolute bastard. Best of luck...
I have only tested it with a 4M cable since I felt it was probably too much capacitance anyway.
I found it in 5 sections but Acrobat reader won't open any of them. If you don't mind e-mailing it, my mailbox should accept 15MB but I often use yousendit.com for large files. It's free and you don't have to supply the sender's email address.
There is a small PCB holding the back panel sockets. It just has 2 RJ45s connected together and a short cable (white in the pic) goes across the SBC to it's RJ45.
I have tried a large clip-on ferrite. At the end closest to the case it makes less than 1dB of improvement. At the SBC end it makes the problem slightly worse. A ferrite bead on the cable outside gives about 4dB improvement.
The internal cable is just flat 6 core. I did try UTP cable in there but there was no difference. I didn't think there would be but had to try. :-)
Disconnect the internal RJ45 lead from the SBC (but leave it in roughly the same position) and repeat the test. If the emissions are still bad this tells you the length of cable which passes over the SBC is picking up the RFI rather than it being conducted out of the port.
Wrap the internal RJ45 lead in copper tape (make sure you get the type with conductive adhesive) and ground it to the case, by the shortest possible route, at both ends. If this helps then shielded cable may work, but only if you gen get a low-impedance connection to ground at both ends.
Try (2) in combination with the ferrite.
Try disabling the Ethernet peripheral in the BIOS, if this doesn't give any change then the problem is other sources of RFI coupling through onto the port.
If you can get the EMC test report from the SBC vendor, compare their test setup with yours. Different lengths of test cable (and how they are terminated) can make a big difference to the results.
A highly recommended book is "EMC for product designers" by Tim Williams, ISBN 0-7506-4930-5. It mainly deals with CE regs but the tests called up in most other standards are similar.
Already did that. There is almost no change when disconnected from the SBC end.
I tried aluminium tape but could only get a good ground at the SBC end. It made a small difference. I didn't put too much effort into that test as I don't know where to get very short (about 150mm) shielded cables.
Ok.
Also tried. This made no change.
I have their FCC and CE reports. The cable is very similar as far as I can tell from the photos.
I just looked it up and RS Components Australia has it. I'll place an order in the morning. The LAN cable is only one of my problems. The Audio RCAs and speaker wires are the next worst. I already had 1nF capacitors to ground on each speaker line and will try adding series ferrite beads next.
The audio RCAs are a bigger problem unless I can find some aluminium foil washers to fit between them and the case or a 4 way PCB mount RCA block which can be grounded at the case.
In message , dated Wed, 23 Aug 2006, Mike Warren writes
You cut up cable from a reel and strip the ends.
--
OOO - Own Opinions Only. Try www.jmwa.demon.co.uk and www.isce.org.uk
2006 is YMMVI- Your mileage may vary immensely.
John Woodgate, J M Woodgate and Associates, Rayleigh, Essex UK
I was thinking a plug-in card would come with its own port and thus bypass the internal cable (and perhaps dodgy layout in the existing motherboard) all together. But if adding a card is not practical the point is moot anyway.
We had the same problem with our SBC, with every #*?#%$! cable that came off the board. It didn't help that the product was in a plastic housing, but we ended up having to put the SBC in its own custom-made screening can, with ferrites on every cable that passed through it. This also helped in that it provided a solid ground to attach cable shields to.
Even with all this we barely scraped through at 1dB under the limit! The PHB decreed that this was OK. Fortunately we didn't ship very many of them...
[snip]
It might be worth considering partitioning up the inside of your case - basically build a Faraday cage around the SBC, and filter every connection to/from it with ferrite beads. We used copper sheet, soldered at the seams, but tinplate would be just as good and cheaper. Not very production friendly, but there are companies that will build you a custom screening can with clip-on lid (like the ones on VCR tuners etc), once you have a scheme that works.
In message , dated Wed, 23 Aug 2006, Mike Warren writes
Sorry, I don't. I admit to not realising you were using CAT-5 cable. Can you use standard CAT-5 cables with an added copper-braid overall shield?
--
OOO - Own Opinions Only. Try www.jmwa.demon.co.uk and www.isce.org.uk
2006 is YMMVI- Your mileage may vary immensely.
John Woodgate, J M Woodgate and Associates, Rayleigh, Essex UK
I didn't catch your earlier post, but I will say that we had a similar problem which turned out to be a single test pad that was enabled when it shouldn't have been. The trace couldn't have been much more than 1 or 2cm from under a BGA to a test point, but it was from the PHY, and it caused the ethernet port to radiate at the 6th harmonic of the test signal. It was almost impossible to get rid of with shielded cables, copper tape, etc, but turning off that one point dropped 30dB at the back of the box.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.