ADV7280-M Video chip EMI emmissions

Hi,

We're using an ADV7280-M chip to do our SD video to CSi interface and it's being run in progressive mode with the de-interlacing turned on. During some EMC testing it was noticed that there was a huge amount of noise around the centre frequency of the CSI link (214Mhz) , as shown in the image below. Ignoring the two large spikes either side, as they are from known sources. The rough shape of the noise changes if the camera is pointed at different things making me suspect it's possible the back end CSI link that's generating a lot of it.  We're actually monitoring emissions from cabling attached and the signal is coming either from video lines or the power lines.

Has anyone experienced this kind of situation before? Ideally with ideas of how to help mitigate it?

Best regards,

Marc.

  • 0
    •  Analog Employees 
    on Apr 29, 2016 5:10 PM

    Hi,

    We have asked an expert on this part to have a look and reply.

    Best Regards,

    Jeyasudha.M

  • 0
    •  Analog Employees 
    on May 3, 2016 2:23 PM

    Hi Marc,

    This is unusual, I have not come across issues regarding EMI emissions from the MIPI interface of the ADV7282-M being problematic before.

    1) Is your backend processor terminating the MIPI signals correctly ? Are you getting picture out of your system?

         Systems that are not properly terminated could result in RF reflection and result in large EMI output?

    2) Can you double check that it is the MIPI signals that are causing this EMI spike. Could you write 0x80 to the register 0x00 in the CSI Map. This will stop the MIPI output from the ADV7282-M. Does the EMI spike disappear ?  Writing 0x00 to the register 0x00 in the CSI Map will cause MIPI output to resume.

    3) Please do not insert common-mode chokes, capacitors or any other filtering on the MIPI traces. Doing so will affect the D-PHY termination and will most likely result in loss of picture.

    4) The MIPI D-PHY specification has very specific slew rate limits. These are specifically designed in order to limit EMI emissions. The ADV7282-M meets these slew rate limits.

    5) Are you following the layout recommendations I wrote in the applications note AN-1337? See last page of link below. MIPI is designed to be a chip-to-chip interface so traces need to be as short as possible. Usually under 30 cm.

    I also advise that traces should be kept on the top layer and not go through vias.

    Link to AN-1337:

    http://www.analog.com/media/en/technical-documentation/application-notes/AN-1337.pdf

    6) Would you mind probing the MIPI traces with an oscilloscope? Do they look like the sample traces in AN-1337?

    Thanks and Regards,

    Robert Hinchy

    Senior Applications Engineer,

    Analog Devices Inc.

  • Hi Rob,

    Thanks for the reply.

    For the various points you mentioned:

    1) As far as we know yes, the receiver is working correctly. It's a Technexion iMX6 module and the image we see looks clear and free from artifacts when viewing live data and also the free running colour bars.

    2) We're waiting for another chance with the EMI test gear to try this test, it and the general PWRDWN, of the chip were going to be our next investigation avenues. As well as running with and without the I2P module enabled. Currently we have the I2P running to get a better picture.

    3) The link is a short one, via about 5 cm of FCC cable, there is nothing in-line.

    4 & 5) Are vias likely to introduce a huge problem? Due to space requirements the video chip is on the back of the board and the connector is on the top side, directly opposite it. So both clock and data lines go though vias. We're also wondering about power supply filtering as the Eval board has separate filters on each of the 1v8 lines into the chip. Out first design did not have as much filtering and we retro fitted as much as we could (t-filters etc) and there was a change in the EMI. We'd need to re-spin our carrier card to include all the filtering as per the eval board so one possibility if we can be confident it would improve things.

    The suggestion we had in house was noise was coming back up the power/gnd rails from the device and travelling though the switch mode that is feeding things at the 5v level. The modulation on the large EMI bump looks to be 1.3Mhz which matches our switch mode’s frequency.

    Best regards,

    Marc.

  • 0
    •  Analog Employees 
    on May 5, 2016 9:12 PM

    Hi Marc,

    • Before you go back to an EMI test house it might be a good idea to run a near-field probe over your system to hunt for the source of this noise.
      • This could show if it is your switch mode power supply or MIPI transmissions are the source of the EMI.
    • When you do go back to the EMI test house could you run the ADV7282-M in a progressive free-run mode. This causes the clock speed to double. You can see if this affects the EMI capture frequency.
      • Please also do the experiment where the MIPI D-PHY of the ADV7282-M is powered down.
    • FCC is an unshielded ribbon cable correct ? A bend etc in this cable could be a source for EMI. I used shielded SMA cables to transmit the MIPI output from the ADV7282-M to a protocol analyzer.
    • I advise placing ferrite beads and decoupling capacitors on each power supply.
    • I am far from an expert in switch mode power supplies but I have designed board using them where I had to minimize output ripple etc. I found the ADIsimPower Votlage Regulator Design Tool and the ADP232x Buck Regulator Design Tool were really useful in achieving this. These tools can be found here:
    • Also the following article may be useful:

         http://www.analog.com/en/education/education-library/raqs/raq_jb_lock_down_that_noise_issue37.html

    If you post a EZ question in the Power Management section I am sure you will get help regarding you switch mode power supply question.

    Regards,

    Robert Hinchy

    Senior Applications Engineer,

    Video Section,

    ACE Department,

    Analog Devices Inc.

  • Just a follow up to this, to solve some of the mysteries.

    We got a near field probe set and scope rented and after a lot of exploring it looked like some kind of interaction between the ADV7280-M and the main 12-5V SEPIC DC converter. Adding some extra filtering and changing some inductors seems to have lowed the huge bump of broadband noise significantly.

    Toggling the enable bits for parts the the chip had some effect, but nothing major. However de-powering the chip had the biggest impact. Which helped point to the Supply itself being the problem.

    We do see emissions from the MIPI FCC however their power levels seem low enough that this will not be a worry, they seem very narrow band and don't radiate far with the near field probe testing.

    The kit is down at the test labs again today so hopefully things are a lot better.

    Thanks for all your help on this, much appreciated!

    Marc.