• State TI Thinks Resolved
  • Locked Locked
  • Replies 15 replies
  • Answers 5 answers
  • Subscribers 99 subscribers
  • Views 2832 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AM335X Emission

Eby Jayan
Expert 2115 points
Other Parts Discussed in Thread: AM3356, TPS65910A

HI,

We are using AM3356 for our board with DDR3 ,NAND Flash  ,2xRGMII interface , 4x UART. THE PMIC we have used is TPS65910A. We ahve taken I2C , SPI , DCAN and Analog line to a connector . Wanted to know whether any peripherals or any power rails of AM335X or the PMIC will cause emissions in the range of 40MHz to 80MHz . If any one has encountered any such issue , please reply what cane be done .

Regards 

Eby Jayan

  • 0
    TI__Guru**** 393215 points
    Hi Eby,

    Electromagnetic emissions are a complex product, that is heavily dependent on PCB design, component placement and routing. It's impossible to predict what exactly will happen in your case. You can only verify this by doing the required testing.
  • 0 in reply to Biser Gatchev-XID
    Expert 2115 points
    Hi Biser ,
    We had conducted the Emission test as per EN5022/CISPR22 standard . We had found Emissions in the Frequency range of 40 MHz to 80MHz . Is there any Interfaces /PLL in the Am335X Controller that could contribute to this emission . Is there any thing we can need to look into int the case of PMIC as well ?

    Regards,
    Eby Jayan
  • 0
    Prodigy 45 points

    Hello Eby,

    You should try to improve your analysis using a spectrum analyzer with a near-field probe to research for these frequencies between 40~80MHz... try to see 3rd and 5th harmonics to predict which of then could be the main frequency...

    Another way that could help you to find the source of the noise, try to power off parts of your board (rgmii interfaces, circuits connected to the uarts, etc) and look up if the disturbances in the frequency that you saw has stopped...

    Generally this range of frequency are emitted by dc-dc converters, look if all switching nodes of your power supplies are smaller enough, and if all dc-dc power supplies are working well (low ripple, low spikes, look if they are working in discontinuous mode, these modes could emit a lot sometimes).

    As Biser said, it's a hard job to find source of emissions and electromagnetic interference since this problem involves all the PCB layout and placement. Did you do SI simulation of your board?

    Any other help to be really helpful unfortunately need more informations, schematics, PCB placement, stack-up, etc.
    Hope to have helped! Good luck!

  • 0 in reply to RNJ
    Expert 2115 points
    Hi,
    Thanks for your reply . We are using an external AC -DC converter which gives us 12 V and a 12 to 5V DC-DC converter on board .Can both the converters be significant source of emission . And one more doubt n the "General Hardware design/BGA PCB design/BGA decoupling " note of TI it is not recommended to use Ferrite beads on the Digital power connections .In our design with AM335X we had considered ferrites for VDDSHV1 to VDDSHV6 . Can this cause noise and emission issue in the power rails of the board .

    We had visited a local lab to check with the Spectrum analyzer we saw distinct pulses of 72 MHz near Processor (24 MHz crystal Used ) and 50 MHz , 75 MHZ spikes near Ethernet PHY (25MHz crystal Used ) . These were distinct pulses . But what we saw in the test in Chamber was a band of frequencies between 40-80 MHz which were causing failures .

    Regards
    Eby Jayan
  • 0
    Genius 3685 points

    DDR3 will run with 300-400MHz.

    NAND Flash will typically be quiet during normal operation.

    RGMII is at 125/250MHz.

    The range of 40-80 MHz is often amplified by cables. Try to add filters/capacitors to your connectors.

  • 0 in reply to Wolfgang Muees1
    Expert 2115 points
    HI,
    Thanks for your comments . While testing the board was powered from 12V DC , The cable carrying the 12V supply was routed from the floor below the chamber with the cables about 5-6 Meters long . will Adding filters to this can help ??

    Regards

    Eby
  • 0 in reply to Eby Jayan
    Prodigy 45 points
    Hello Eby, you're welcome!

    About your power supplies, yes, this can be one of the main sources of the noise. Have you checked the AC-DC power supply on the chamber, use a load similar to your PCB profile and look if the noise remains!

    About the Ti's note, I do not think that these ferrite beads are the problem.. They create a LC filter that in fact helps to reduce emission between source and load.

    About this pulses that you saw, they probably are not the source, they can affect, but not be the main reason.. crystals don't have high di/dt to emit a wide band, only spikes in the harmonics frequency. Try to think on that idea that to have high emission you need two important things, dv/dt + di/dt.

    You said a good new information, in your CISPR22 test you didn't saw spikes, you saw a large band of frequencies bellow 100MHz with a high dBuV scale right?
    In my opinion, looking to this kind of trouble i guess that are noises generated by power supplies, it's very common.. Have you checked it your switching node and the ground path of your board?

    Since I don't have all informations, I will guess that your 5Vdc-dc supply is a regular step-down converter.. if possible try to see with the spectrum analyzer your switching node (nearest possible of the flyback diode or lowside FET). In my opinion this can probably be the source of your noise.. look with an osciloscope the switching node.. any overshooting on this line can emit A LOT!! A good placement of your schematic and a snubber well calculated it's more than enough to reduce significantly noises bellow 100MHz.

    Good luck!!!
  • 0 in reply to Eby Jayan
    Prodigy 45 points
    Eby, just to complement, Wolfgang's tips it's a good idea, this won't solve the source of the problem but can attenuate several dBs, there is a lot of ferrites for cables with that purpose, in some cases the source of the noise (PCB or AC-DC power supply) use a bad ground path as an antenna to couple more noise and transmit more and more..
  • 0 in reply to RNJ
    Expert 2115 points
    HI,
    Thank you so much for the information. We will check the power supply the next time we visit the lab . One more thing the Lab staff suggested was using a battery to avoid switching noise of Power supply .

    Regards

    Eby
  • 0 in reply to RNJ
    Expert 2115 points

    Hi ,

    We had visited a local lab having spectrum analyzer and measured with near field porbe . We observed noise near the flyback diode of a DC-DC converter in our board . We changed the DC-DC converter IC with another make and observed reduction in noise by about 10-15 dbm . Could this save us ?

  • 0 in reply to Eby Jayan
    Prodigy 45 points
    Hello Eby,
    It's hard to compare numbers between near field probe and a anechoic chamber test. You can have ideas from where are coming the noises. Lower dBms on spectrum analyzer not necessarily means lower dBuV in CISPR22 tests. You can mistake your analysis with near field probe using different Video/Res bandwidths.
    Assuming that the noise that you observed in the beginning are sourced in your power supply, just verify if you have a good switching node without overshooting. Check always the ground path return to all high speed signals of your board. There is a lot of tips on the web that could help to improve EMC performance, some tips that I recommend to at least check, are:

    Improve your pcb stackup.
    Shielding the PCB borders with vias.
    Use of ferrites or RC circuits to attenuate square signals.
    Use of common mode chokes.
    Ground path return to all critical signals.

    I can provide more help if you need, contact me in my personal e-mail (noronha.ricardo@gmail.com).
  • 0 in reply to RNJ
    Expert 2115 points

    Hi ,

    Thanks for the reply . i have one more question .

    We tried with different chips 

    1. Richtek - showed overshoot in Fly back when viewed in CRO Showed the least emission in Spectrum analyzer

    2. TI - LM 2374 - Did not have any overshoot but is showing more emission than richtek 

    These two are having better performance in spectrum analyzer than from the original board .

    which one of the above tow will be a better solution ?

    Regards

    Eby

  • 0 in reply to Eby Jayan
    Prodigy 45 points

    Hello Eby,

    It's not possible to choice between PN1 or PN2.. as was said it before, EMC/EMI depends a lot of the PCB and placement. Any decision will be a blind guess.. I think you should first try to understand all your scenario before decide to replace the powersupply as the main guilty of the noise.

  • 0 in reply to RNJ
    Expert 2115 points
    Hi ,
    We tried testing the following DC-DC converter
    1. Richtek - showed overshoot in Fly back when viewed in CRO Showed the least emission in Spectrum analyzer

    and the result was that we cleared all the previous failed frequencies but we had failure in a new frequency 32.7 MHz
    Is it worth checking the other option

    2. TI - LM 2374 - Did not have any overshoot but is showing more emission than richtek

    or we have to change the layout of the board . The layout change would be the last thing we would be looking in to .


    Regrads
    Eby Jayan
  • 0 in reply to Eby Jayan
    Prodigy 45 points

    Helly Eby,


    I spent a time away from the e2e.. considering that you did not solve your problem yet, I guess you should try to do more tests in the anechoic chamber with both options of your power supply.

    Before doing that, just to ensure that you are not making any primary mistake, ask Ti for a review of your schematic and layout, and richtek for their version, stay alert about do a revision on all your circuits that could emit, (RF circuits, Power supplies, clock lines, highspeed buses, etc).

    Hope you have success in your CISPR22 test!