• TI thinks resolved

LM5113: Parasitic inductance and the LM5113 Spice model

Part Number: LM5113

Is TI aware of any issues with the LM5113 Spice model when parasitic inductance is placed in series with the Vcc pin and the Vss pin.  In my model I am attempting to account for parasitic inductance between the decoupling cap and the Vcc/ Vss pins.  The circuit is running at 30 MHz.  What I observe is if I place an inductance of > 2.5nH in series with these pins the output of the device fails to produce a proper square wave.  When the inductance is <2.5nH the circuit works properly.  .Any thoughts on this?  Thanks

Mike Brand

310-647-3297

  • Hi Michael,
    thanks for your interest in LM5113 and for bringing this up.

    There a few things to consider to better understand the behavior:

    1. The spice model is a behavioral model that will give you a good representation of its actual operation when used withing the spec sheet parameters.

    2. The part is actually not capable of running at 30MHz because of thermal limitations; the highest frequency in application has been ~13MHz, but anything above 10MHz is quite difficult.

    3. If you are doing a tight layout typically you can limit the loop inductance with the Vcc bypass cap to <2.5nH total. If I read your email correctly you actually added ~2.5nH per pin?

    In any case I have simulated with increased inductance and didn't find any unexpected behavior: the rise/fall times get longer as the inductance increases.

    I would be happy to investigate further if you could share your setup with me.

    In case you are concerned about some IP, we can take this to a private channel.

    Please advise.

    Thank you and best regards,
    Alberto
  • In reply to Alberto Doronzo:

    Alberto,

    thanks for the reply. Just to make sure I understood you correctly, you ran the LM5113 with 2.5nH on the supply lines and saw no abnormal behavior? Can you tell me what simulator did you use? At what frequency did you run the sim? Thanks

    Mike
  • In reply to Michael Brand:

    Hi Michael,

    I am running my simulation on Simetrix.

    There are a couple of things that might help running the simulation :

    1. ensuring that your voltage supply for both high and low side is stiff (except for the extra capacitance), as at those frequencies it might be difficult to charge up the HB cap.

    2.because of the extremely low time constant at nodes, the simulator might not converge, you might have to detune the parasitic instructors with a parallel resistor.

    Please see below the simulated waveforms for both high gate (wrt the switched node) and the low side gate - the spikes are likely an artifact but there is no way of checking as the LM5113 is not capable of those frequencies. Also see the setup below.

    If you are interested in higher frequency applications, we can have a private conversation about your needs.

    Thank you and best regards,

    Alberto

  • In reply to Alberto Doronzo:

    Alberto,

    Thanks for the information.  I will try to run my circuit on SIMetrix to see if it makes a difference.  I should point out that I am running the high side driver as a ground referenced circuit off of the same 5V as the low side driver.   I am therefore not using a bootstrap capacitor.  I will contact you later about higher frequency operation.  Thanks

    Mike

  • In reply to Michael Brand:

    Alberto,

    I have copied the LM5113 spice model into my SImetrix circuit file.  The transfer went fine.  I used the unencrypted pspice model from the TI web site.  I can not make it work.  The high side driver switches properly but the peak of the waveform only reaches 3.6 volts.  I would expect it to be closer to 5V.  The low side driver output is just noise.  What model did you use and how did you make it work?  I should point out that I have been successfully running the LM5113 model on LTSpice with no problem.  I had to move to Simetrix for other reasons.   Further this one day turnaround to hear back from you is not working.  I need to get answers much quicker.   This modeling issue is severely delaying the completion of my task.   Is their away we can have direct communication by phone.  Thanks

    Mike

  • In reply to Michael Brand:

    Hi Mike,
    I have just invited to a private conversation where we can share direct contact, so that I can support this more effectively.
    Best regards,
    Alberto