671MHz and 805MHz EMI emissions

[dmriker]

Hello,
I'm trying to get through EMC (electromagnetic compliance) testing for a medical product and find that 671MHz and 805MHz frequencies are being emitted from what appears to be the video function. I've tried 4 different video cards (GTX 460, GTX 560, GT 430, HD 6770) all running in 2560x1600 mode, DVI dual-link, two (redundant) monitors. The frequencies show up with all boards tested, and they go away when the boards are not inserted. Likewise, measured emissions go away when the video cables are unplugged from the card. Any ideas?

 

[jrazor247]

Wow that is a technically advanced question. I'd say it is whatever is generating the 75 watts on pci-express bus, but what do I know about emi? Also to consider is the RAMDACs that take image data from the frame buffer and convert them into voltaged video signals be it VGA or DVI. Driving two DVI displays is alot of voltage switching power. If you trying using just one VGA cable, do you still get emissions?

 

[TeraMedia]

It sounds like you're getting a leak from the cables. The pixel signals run at ~1.35 GHz (for digital flat-panel displays with short blanking) or ~1.74 GHz (for CRTs with longer blanking intervals) most likely for your resolution, and I suspect that the 671 MHz is half of the 1.35 GHz pixel bit rate and indicative of a cable EMI leak. Lookup DVI on wikipedia for more info about these topics.

Are you using high-quality shielded cables? If you have any others you could borrow and try, I would suggest that. If you have a way to try a DisplayPort connection, I would suggest that as well.

1) Try one monitor instead of 2 (just to see what the impact is). If one monitor has a larger, non-standard blanking interval, that could be your 805 MHz generator. That should help you learn more about the cause.
2) Try a different image on the screens, such as all-black, all-white, or #AAAAAA in html just to see if it is content-dependent (it may be). That last color alternates the bit values and should result in a different transmission characteristic over the cable.
2) Try different resolutions, if possible, to see if the frequencies change (they probably will, and this will give you more information)
2) Try different cables. You may have a faulty shield or pair-twisting on one of them
4) Try a different connection (e.g. over VGA, DP or HDMI - potentially at a lower resolution), to verify that the radiation goes away (or doesn't...)

One other "idoit" question: You *are* using a closed metal case with an I/O port plate in place on the Mobo, right?

 

[dmriker]

Wow that is a technically advanced question. I'd say it is whatever is generating the 75 watts on pci-express bus, but what do I know about emi? Also to consider is the RAMDACs that take image data from the frame buffer and convert them into voltaged video signals be it VGA or DVI. Driving two DVI displays is alot of voltage switching power. If you trying using just one VGA cable, do you still get emissions?

-----------------------------------------

Thanks very much for the feedback!

Doug Riker
NinePoint Medical

 

[dmriker]

It sounds like you're getting a leak from the cables. The pixel signals run at ~1.35 GHz (for digital flat-panel displays with short blanking) or ~1.74 GHz (for CRTs with longer blanking intervals) most likely for your resolution, and I suspect that the 671 MHz is half of the 1.35 GHz pixel bit rate and indicative of a cable EMI leak. Lookup DVI on wikipedia for more info about these topics.

Are you using high-quality shielded cables? If you have any others you could borrow and try, I would suggest that. If you have a way to try a DisplayPort connection, I would suggest that as well.

1) Try one monitor instead of 2 (just to see what the impact is). If one monitor has a larger, non-standard blanking interval, that could be your 805 MHz generator. That should help you learn more about the cause.
2) Try a different image on the screens, such as all-black, all-white, or #AAAAAA in html just to see if it is content-dependent (it may be). That last color alternates the bit values and should result in a different transmission characteristic over the cable.
2) Try different resolutions, if possible, to see if the frequencies change (they probably will, and this will give you more information)
2) Try different cables. You may have a faulty shield or pair-twisting on one of them
4) Try a different connection (e.g. over VGA, DP or HDMI - potentially at a lower resolution), to verify that the radiation goes away (or doesn't...)

One other "idoit" question: You *are* using a closed metal case with an I/O port plate in place on the Mobo, right?

--------------------------------
Thanks for the feedback, you gave me a lot to look into. Some responses:
1) I'll have to set up time at the test house to see if only 1 monitor affects anything. The monitors are identical.
2) The emissions begin during BIOS boot, and remain regardless of screen content
3) I haven't tried changing resolution yet
4) The cables appear to be good quality (Dell), with embedded ferrites. Will check as related to 1)
5) something to try
6) Yes, enclosed metal chassis enclosing dual-Xenon motherboard, I/O plate inplace (although some gaps/holes in at least one plate. When we installed copper fingers to improve cover mating with chassis, emissions got a little WORSE