Bridging Grounds with Ferrites? TERRIBLE IDEA!!!

แชร์
ฝัง
  • เผยแพร่เมื่อ 20 ธ.ค. 2024

ความคิดเห็น • 34

  • @hkn308
    @hkn308 11 หลายเดือนก่อน +13

    Please, make video tutorial series about High Speed ADC/DAC Interfacing with FPGA, (JESD and LVDS), and also mezzanine, VPX, etc... how to interface an RF transceiver with an FPGA for example... Thank you Zach for your all amazing tutorials ! also thank you Altium for these video series !

  • @Graham_Wideman
    @Graham_Wideman 11 หลายเดือนก่อน +1

    7:40 "If you were trying to route a single-ended digital signal alongside that ferrite and thereby use the ferrite as part of the return path, you'll have a high impedance return path and the signal will radiate." Rather worse than that, for this case where a single-ended digital signal communicates between ground domains, if you route the return current through an inductor the resulting voltage drop (ie: voltage pulses at current-intensive signal transitions) adds to the intended signal at the receiver, and thus the receiver may see additional unwanted transitions. Also receivers of any single-ended analog signals crossing the domains will see the inductor-voltage-drop noise added to their intended analog input signals.

  • @theondono
    @theondono 11 หลายเดือนก่อน +4

    Anyone who has worked in EMC knows that whenever you think you have a good reason to split grounds, you're almost surely wrong!

  • @sajjsamm
    @sajjsamm 11 หลายเดือนก่อน +1

    Thanks for the explanation Zach but this video only addresses switching mode powersupplies in a pcb design. When we are allowed to bridge sparate grounds together with a ferrite bead? Maybe in a mixed signal design when we want to route analog signal from analog region to digital region in a pcb? or always just use a single ground plane?

    • @gsuberland
      @gsuberland 11 หลายเดือนก่อน

      Stick to a single ground plane, and ensure your board stackup has a solid ground reference plane under your signals and power planes (ideally SIG/PWR GND GND SIG/PWR). If you're seeing noise on your ground plane due to high dI/dt, you need to assess the specific layout and stackup problems rather than trying to patch it with a ferrite, which will generally make it worse. If you need ground isolation between two devices then use a proper galvanic isolation method like an opto.

    • @Zachariah-Peterson
      @Zachariah-Peterson 11 หลายเดือนก่อน

      TL;DR answer: never unless you can prove you need it. There are some specialty cases of low-SNR low-current DC/low frequency measurements where you need it to control the path of noise currents, but these are really corner cases and it's not a good idea to extrapolate those practices to all other types of mixed signal systems. Most mixed signal systems do not low-SNR low-current measurements.

    • @mohammedabdeltawab9882
      @mohammedabdeltawab9882 11 หลายเดือนก่อน

      Never connect your return path by a FB, during EMC tests the high frequency current flow will make a high drop in the FB
      this will make a big voltage difference between your return path sections before and after the FB. so your functionality will be interrupted and signals will be miss interpreted

  • @caglauzun3454
    @caglauzun3454 4 หลายเดือนก่อน

    How should we separete analog and digital grounds in mixed signal circuits? Some advices are ferride beads, schottky diodes and zero ohms. Which ones should we used and how we decided?

    • @Zachariah-Peterson
      @Zachariah-Peterson 3 หลายเดือนก่อน

      In almost every case, you should not separate those grounds, this is based on 40-year old design guidance from semiconductor manufacturers which is no longer relevant in today's electronics. Use a ground plane on the layer adjacent to the signals. Using a thinner layer between signal and ground helps ensure the return paths are confined near signal traces. Probably the only exception is below kHz frequency, but even that case can be solved by placing components and carefully following the return path so that the analog signal does not intersect any digital signals. Separation of analog and digital grounds is a great way to fail EMC testing; even if the board would function correctly, you would not be allowed to sell it in most markets because of EMC testing failure.

  • @rb8049
    @rb8049 11 หลายเดือนก่อน

    Designers even separate grounds in ICs. This requires ESD protection between the grounds and additional analysis. But level shifters are normally not immune to common mode noise. One requires differential level shifters. Just more and more complexity.

  • @lesterlan422
    @lesterlan422 11 หลายเดือนก่อน

    Do common mode chokes also have same issues?
    I ever heard a doctor doesn't recommend to use any common mode choke.
    But I still see people use common mode chokes in some practices, such at the power input port, using a common mode choke to bridge the power rail.
    Or that's a different situation?
    And also, will this kind of GND bridge degrade the ESD protection?
    From my perspective, it makes the inrush ESD current difficult to go through to Earth GND.
    Thanks.

    • @Zachariah-Peterson
      @Zachariah-Peterson 10 หลายเดือนก่อน +1

      I didn't bring up chokes in this discussion because the post presented in the video was not about chokes. Also chokes are not used to bridge ground reference planes, they are used on a pair of signal lines that require filtering. Anyways these components are used to remove parts of a signal, while the ferrite is claimed to be used for passing some signal, it is the opposite implementation. The GND bridge can degrade the ESD protection because the ESD protection will now be limited to the ferrite's breakdown voltage (whatever value that happens to be). For example, if you require 4kV isolation, maybe the bead only has 2kV isolation, then your galvanic isolation is only 2kV because the bead becomes the failure point.

  • @Graham_Wideman
    @Graham_Wideman 11 หลายเดือนก่อน

    Thanks Zach for producing this video, it's certainly a worthwhile topic, plus its general conclusion agrees with me :-). That said, I feel like it got a little off target, and of course there's more to say on the topic.
    First, I don't think anyone is confused that the ferrite somehow magically confers "_galvanic_" isolation that automatically should warrant a DC-DC isolated converter, so that seems like a red herring. Of course isolated DC-DC converters have their place when galvanic isolation is explicitly needed.
    Instead, I think the "isolation" attempted in the sample schematic is poorly-considered "isolation of noise" -- where there's been some experience of "dirty" ground messing up some other signals. This scenario can arise where the system has (a) some noisy high-current load and (b) some analog or sensitive digital circuitry. A pedestrian example would be an embedded controller, such as a 3D printer controller or similar that drives motors (significant switching currents, 12V or 24V supply) and also handles low voltage analog signals like temperature sensors, all ostensibly on the same ground net. AND such assemblies may have an MCU offering AGND pins associated with the ADC implying that you SHOULD have some separate analog ground net.
    A ferrite is still the wrong way to handle it. Instead, manage the PCB current paths for the high-current-switching loads with careful application of "star ground" principles. (Use net tie components on the schematic to allow separate net names for the branches.) This avoids having the load-switching-current PCB-trace voltage drops in the path of the reference ground supplied to the sensitive analog ICs, relative to which their analog signals are measured and produced (or analog sections of ICs with separate analog ground terminals). Similar consideration applies to digital IC ground nets, though these generally have signal reference ground and power ground as the same terminal. So for these, again avoid them seeing the load-switching ground-voltage noise. But these chips of course also generate ground noise that can be minimized by the usual digital ground principles, and kept away from analog signals again with some star ground thinking.

  • @NathanSweet
    @NathanSweet 11 หลายเดือนก่อน

    What do you think about the AFBR-S50MV85G datasheet recommending a bead between grounds to isolate the noise from a laser?

  • @Graham_Wideman
    @Graham_Wideman 11 หลายเดือนก่อน

    @Zach: The link in the description for LinkedIn, for me, only shows your page if I'm not logged in to LinkedIn. If I am logged in, LinkedIn shows a 404 error for your page, and you don't appear in search results. It's acting like your account has blocked me on LinkedIn, which is quite odd, since I've never interacted with you on LinkedIn. Of course you're free to do that, but I'd appreciate if you could check whether that's the case so I don't need to troubleshoot this further.

  • @Dilipkumar-pc9rk
    @Dilipkumar-pc9rk 11 หลายเดือนก่อน

    Apart from the video .. I’ve one question on ferrite beads implementation..
    What frequency of a ferrite bead I should choose when my switching noise is around 1-5 MHz ?

    • @Zachariah-Peterson
      @Zachariah-Peterson 10 หลายเดือนก่อน

      What are you trying to use the bead for? Where will it appear in the circuit? Are you trying to use this to filter switching noise?

    • @Dilipkumar-pc9rk
      @Dilipkumar-pc9rk 10 หลายเดือนก่อน

      @@Zachariah-Peterson thanks for your reply..
      I’m working on industrial automation product.
      Here for one of the inputs is Digital input frequency mode from 0 to 20 KHz.
      When a signal applied with 1us or less rise I can see the disturbance on the other Channel.
      Even after having 100MHz beads on the return side input.
      No issues when rise time increases. But in field we don’t know the signal rise time from the customer.
      How I can mitigate this ?

  • @mohammedabdeltawab9882
    @mohammedabdeltawab9882 11 หลายเดือนก่อน

    I agree this is the worst design habit to add a FB in return path. But, i think it's miss-understood why the guy used this FB.
    it's common in juniors understanding that connecting two grounds (usually Analog and Digital) through FB will block digital noise from interrupting analog circuit.
    the guy intention is not galvanic isolation.
    so the best answer here,
    - function wise, you may get problem as you add HF impedance in your functional current path
    - EMC wise this FB will interrupt your system functionality whenever any conducted immunity test is done "EFT, Surge" as there will be a hight voltage difference between two points that are supposed to be at the potential.

    • @Zachariah-Peterson
      @Zachariah-Peterson 11 หลายเดือนก่อน +1

      Are you referring to me specifically not understanding? I think I addressed both points (isolation in a power supply and isolation of signal/noise in general) in my LI comments. It was most of the other commenters who referred to using this for an isolated power supply, which would of course imply a requirement for galvanic isolation.

    • @mohammedabdeltawab9882
      @mohammedabdeltawab9882 11 หลายเดือนก่อน

      @@Zachariah-Peterson
      of course not referring to you, we learn from your posts and videos. I rephrased it and let me thank you more for the valuable info you share

    • @Zachariah-Peterson
      @Zachariah-Peterson 11 หลายเดือนก่อน

      @@mohammedabdeltawab9882 No worries, thank you

  • @87Spectr
    @87Spectr 11 หลายเดือนก่อน

    I use ferrite beads near input connector (and after tvs-diode, of course) - in "+" and "-" nets... to protect from common noise... instead of common mode choke (CMC)... Of course, CMC is ideal variant to eliminate common noise, but it's expensive. I think that FB help to eliminate some common noise - if i am not correct - please, tell me...

    • @Zachariah-Peterson
      @Zachariah-Peterson 10 หลายเดือนก่อน +2

      A pair of ferrite beads placed on adjacent signal lines is basically a weakly coupled inductor, or a very poor common mode choke. If you need to have strong filtering of common mode noise, then use the appropriate common mode choke component.

    • @87Spectr
      @87Spectr 10 หลายเดือนก่อน

      @Zachariah-Peterson thank you very much for reply!)

  • @portblock
    @portblock 11 หลายเดือนก่อน

    Forgive me, I am not good at pcb layout, I am good at circuit design, but not the little details of a circuit.
    My question is, I use a little switching regulator, with simple In/Out/Ground, and I put a little ferrite at the output before my cricuit, is this a bad idea?

    • @user-tc2ky6fg2o
      @user-tc2ky6fg2o 11 หลายเดือนก่อน

      Not an expert answer, but at first, try to answer the question: "why?". What is the goal (and how much)? You should know your regulator capabilities, know the load, and examine the ferrite datasheet you intend to use, finally put your oscilloscope at both ends of the ferrite. You will get the answer to your question if it is a bad idea or not.

  • @rb8049
    @rb8049 11 หลายเดือนก่อน

    Just make sure currents coming from the noise source don’t cross the sensitive circuit.

  • @petersage5157
    @petersage5157 11 หลายเดือนก่อน

    tl;dw: This is the exact opposite of good practice.
    This reminds me of an old hifi schematic that I may have mentioned in a previous comment; for some reason, it had a 10 ohm resistor between the phono preamp ground and the rest of the circuit. In that case, I think it was something along the lines of the damping resistor sometimes used with an LC power supply filter, but it made very little sense to put it on the ground rail.

  • @rb8049
    @rb8049 11 หลายเดือนก่อน

    The BOSS had a wire go a long distance from our lab to the outside of the building to earth ground to solve a noise issue in the lab. 🤦‍♂️ No arguing with the boss.

  • @Daniel-p3g8q
    @Daniel-p3g8q 11 หลายเดือนก่อน

    Ferrites can be a trap for youngster for sure. I know I did it once in the past.

  • @electronichome1153
    @electronichome1153 11 หลายเดือนก่อน

    was fun

  • @سهيلثابتقدم-ذ5خ
    @سهيلثابتقدم-ذ5خ 11 หลายเดือนก่อน

    👀

  • @hadighorbani355
    @hadighorbani355 11 หลายเดือนก่อน

    like