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.

IWR1443: Dielectric constant discrepancy for Rogers 4835 on IWR1443BOOST reference design

Part Number: IWR1443

Hi,

In the reference design for the IWR1443BOOST, the layer stackup calls for Rogers 4835 with a process dielectric constant of 3.48. After reading the datasheet for the Rogers material and consulting with our board shop, it seems that the correct constant should be the design constant specified as 3.66. 

Is this an oversight in the design of the eval board or is there a reason to use the 3.48 dielectric contstant?

Thanks,

Kevin

  • After further discussion with our board shop and reviewing the documentation I have some more questions.

    It seems the that fab drawings for the Rev A board call out the impedance for the antenna feed lines while the Rev B fab drawings do not. Since there are released gerbers for the Rev A board, are those gerbers the ones provided to the board shop or are they the as build gerbers? I need to know if our board shop should adjust the feed line width to hit the 52 ohm impedance specified in the drawings or leave the traces alone.

    Thanks,

    Kevin
  • Hello,

        RF traces are designed to 50 Ohm impedance with the GCPW topology using Rogers Material. 

    Do not refer to the Dk values present in the stack-up as a requirement of controlled impedance for the RF traces.

    These numbers are not accurate and they don't calculate Dk values at mmWave frequencies, also there are many other factors that affects the Dk value such as topology used, surface roughness, type of Cu used, operating frequency etc...   

    Typically these are comprehended in 3D EM simulator (Such as Ansys HFSS, CST or any equivalent 3D EM simulator) tools and from that PCB geometry's are optimized to get 50 Ohm transmission line impedance matching. You could refer to below "TI mmWave Radar sensor RF PCB Design, Manufacturing and Validation Guide" app-note" for more detail, page 3 provides the dimensions for the RF traces. 

    http://www.ti.com/lit/an/spracg5/spracg5.pdf

    It's important to hit the GCPW Dimensions as mentioned in the application note, that ensures required impedance. These values are derived from 3D EM simulator.

    You could mention in the FAB note to adhere to these targeted dimensions for the PCB fabricator.  

    Thanks and regards,

    CHETHAN KUMAR Y.B.

  • Hi Chethan,

    I have read that document and understand it. What I am worried about is that the reference design was designed with the wrong Er from the Rogers datasheet. The striplines are very close to the 50 ohm impedance with standard calculators if using the Er that you have specified in the stackup but that seems to be the wrong Er to use for design with the Rogers 4835 material.

    If the Rev A board was built as specified in the fab notes the board shop would have modified the width of the feed striplines to achieve the specified impedance.

    We measured one of the eval boards and it appears that the board shop did this modification to hit the specified impedance. The traces appeared to be 6.5 mils which is what they calculated as the correct width instead of 7 mils.

    Thanks,

    Kevin
  • Hello Kevin,

    Dk value present in the stack is not correct, hence it should not be used to correct the RF trace width to meet the 50 Ohm. Please follow recommended dimension given in the application note hit design target mean values.

    PCB Fabricators will have etching tolerance which would be typically +/-0.5 mil variation. Hence it should be ok.

    Thanks and regards,
    CHETHAN KUMAR Y.B.
  • Hi Chethan,

    Did the simulation that was used to created the dimensions given in the app note use the correct values for the dielectric?

    Yes, this is within the tolerance of etching for most fabricators but it is still important to start at the correct value so that the tolerance comes out as close as possible.

    Thanks,

    Kevin
  • Kevin,

       For most accurate Dk value please use the Rogers Calculators from their website.

    For example RO4835LOPRO  Datasheet say Design Dk of 3.55, However it de-rates across frequency. 

    Hence it's best to use from the Calculator at the desired frequency, which is in this case 3.46. And same value was used in arriving dimension in app-note. 

    Thanks and best regards,

    CHETHAN KUMAR Y.B.