Shop OBEX P1 Docs P2 Docs Learn Events
Using ferrite beads to clean switched signals and proper grounding? — Parallax Forums

Using ferrite beads to clean switched signals and proper grounding?

rwgast_logicdesignrwgast_logicdesign Posts: 1,464
edited 2013-03-12 11:58 in General Discussion
Ok so I have been putting together a little brain board for my sting ray seeing as I fried my quick start. Anyways I am using a home made DC to DC board to power the thing. I did the best job I could filtering the signals per the data sheet and my ripple voltage is fairly low iirc around 4mV. I found that adding a 10uH inductor in series with the output makes the signal extremely clean, but I'm a bit low on those and they take up board space. Anyways I remember seeing something on the EEV Blog Forums where a member stripped the ferrite beads off a mother board and used it to filter a signal from his switching supply along with decoupling caps of course. He supplied scope shots before and after using the bead, if I remember right his signal had a fairly bad saw tooth, with the ferrite bead it looked as clean as any linear regulator! Unfortunately I can not find this post again :/, anyways I went to the garage and pulled some of these ferrite beads from old mother boards:

bead.jpg


I then hooked one up in series with the out put of my switcher and viewed it on a prop scope. It made absolutely no difference in my signal, I'm not sure if these beads are all made different, if I'm using them wrong or if the prop scope is just not fast enough to catch the garbage these filter out. I was hoping some one could kind of help me understand these little things and how to use them.

Also when I designed this power system I made sure all my voltage rail were only grounded together at the battery. This way I could keep all the 5v and 3.3v noise isolated to it's own channel, I was told this was a good idea in a mixed voltage system. So now that I am building my brain I have a 5v AVR and a 3.3v prop on the same board. Keeping the ground isolated from each other is no big deal, but I started thinking about different scenarios, where this does get a bit complicated. Lets say both prop and AVR are are connected to the same circuit and this circuit includes a pull down resistor, which ground rail do I pull down to? A good example of this, the reset line from the prop plug will be shared, in order for it to function with the AVR it needs to be pulled down. I'm not sure if what i am saying is making a lot of sense or not.

EDIT:
I just thought I would add, I know a lot of this seems like over kill for most hobby projects but I intend on mixing analog in on both power rails and im looking to keep the noise to a minimal so the ADC's won't loose accuracy
1024 x 768 - 56K

Comments

  • LoopyBytelooseLoopyByteloose Posts: 12,537
    edited 2013-03-11 23:26
    Clean just might be too clean. Any time you try to clean up a signal by adding devices, you risk changing the signal. Other than Schmitt triggers, I suspect you might begin to loose data with these modifications.

    Ferrite beads are intended to mainly block transients from induction on long wires.

    If you are going to clean up a signal with passive filters, you will likely have to used capacitors and inductors, AND do the maths. Anything less is a shot in the dark.

    It seems like you are referring to power as signal. A signal is information, power is a steady supply with no information. Getting rid of ripple is not the same as getting rid of noise.
  • rwgast_logicdesignrwgast_logicdesign Posts: 1,464
    edited 2013-03-12 01:00
    Hmmm ok well i am trying to lower my ripple voltage so I can run clean power to the ADC's. This is a really dumb question but im not really sure how to filter the ripple. I understand you would probably use a low pass filter and I know the math for that BUT I don't know what frequency I should be filtering out, the actual frequency of the regulator or the frequency I see when I look at the circuit on a scope. And is there anything wrong with using an RC filter over LC?
  • jmgjmg Posts: 15,173
    edited 2013-03-12 01:08
    I then hooked one up in series with the out put of my switcher and viewed it on a prop scope. It made absolutely no difference in my signal, I'm not sure if these beads are all made different, if I'm using them wrong or if the prop scope is just not fast enough to catch the garbage these filter out. I was hoping some one could kind of help me understand these little things and how to use them.

    Ferrite beads impedance is at the 10's MHz - sub GHz range, so a LF scope will not see much effect.
    The main gain is in removing the RF region energy, ie those very short ringing spikes you tend to get from switching edges.

    You could try a 1N4148 diode to a multimeter down a short shielded cable, as a simple peak-detector.
  • LoopyBytelooseLoopyByteloose Posts: 12,537
    edited 2013-03-12 03:50
    Okay... if you want 'perfectly' clean power to ADCs, you will have to know how much power you require.
    The more power you require, the harder it is to do.

    And then you really have a 'time domain' problem. If you know the frequency of your ripple, you can design either an active or passive filtering system to remove that frequency. If you don't have a particular frequency range, you have to start making assumptions about how a low pass or high pass filter will best serve you.

    My own experience is with removing the 60 cycle hum in audio pre-amps and in class-A audio amps. At 60 cycles, it is very difficult to remove all of it as the components to do so become huge. Fortunately, the gnomes that sell op-amps these days do an excellent job with the pre-amp stages. But that final power stage is really a challenge.

    Try Wikipedia for 'passive filters' and there are a lot of calculators on the internet that will create a filter with a complete component list for you. This used to be a big undertaking for the hobbyist as the calculations are involved. But the computer age has made it all very easy.

    ~~~~~
    BTW, the ADC inputs might also need from shielding from RF noise.. that is usually done in some sort of ground plane creating a 'moat' around the inputs. I've seen some specific suggestions for AVR chips, but don't have a generalized knowledge of all ADCs. Look for app notes for your particular chip.
  • Mark_TMark_T Posts: 1,981
    edited 2013-03-12 05:27
    jmg wrote: »
    Ferrite beads impedance is at the 10's MHz - sub GHz range, so a LF scope will not see much effect.
    The main gain is in removing the RF region energy, ie those very short ringing spikes you tend to get from switching edges.

    You could try a 1N4148 diode to a multimeter down a short shielded cable, as a simple peak-detector.

    Ferrite beads are often characterised with an impedance in ohms - although inductive impedance increases
    with frequency the relative permeablility of ferrites drops with frequency so above a base frequency they tend
    to have a fairly constant impedance (and because they are lossy its partly resistive, partly inductive I believe).

    For power supply use adding 100 ohms of resistance at MHz and above allows the decoupling caps to do a
    much better job of attenuating switch mode noise if its a high frequency switcher. In RF design beads are used
    ubiquitously to prevent RF currents in the wrong place - simple and cheap, can be added on component leads
    directly.

    For high speed logic beads are a common way to reduce the EMI associated with the fast edges by taming
    them (at the expense of rise/fall times of course, although a large common-mode bead can be used on the
    whole cable to avoid this)
  • rwgast_logicdesignrwgast_logicdesign Posts: 1,464
    edited 2013-03-12 11:58
    Ok well I made a simple 50Khz RC filter, which helped.. but caused two other issues. First of all almost 1v drop, Im assuming this is the problem with RC over LC? Second my ripple went down to about 2mv but then huge spikes would happen up to 100mv, it looks as though the cap was discharging. I used a 30 ohm resistor and .1uf cap.
Sign In or Register to comment.