Reducing switching noise from power supply on 5V rail

[Pasky]

  So let me start by saying I'm not electrical engineer just a hobbyist so my understanding of electrical circuits isn't nearly as high as most of the members here.  I've asked this before and was given some helpful advice to use bypass caps and using a LDO is also a possibility and a bypass capacitor and an inductor (http://en.wikipedia.org/wiki/Capacitor-input_filter) was good enough to relieve my last project of switching noise but now I have a new one.  So let me explain my issue.

  I have a video board that outputs analog video.  I am getting what appears to be switching noise on the video out put and confirmed it was switching noise by powering the board with a linear power supply and the noise was gone.  I tried adding a pi filter to the 5V rail of the video board but this only alleviated some of the noise.

  The switching power supply I'm working with provides a 12V and a 5V output but the video board is powered by 5V.  Using another power supply is not an option.  Now I understand using a LDO can greatly reduce the switching noise if it's placed after the switching power supply.  I don't have room for a 7805 and a heat sink so giving the video board 5V through a 7805 from the 12V rail is out of the question.  So I guess my question is, how do I use a LDO that outputs 5V to my video board, while only being able to feed it 5V?  It's my understanding that you need to feed the LDO a voltage above it's output (roughly 1-2V above the output).  Is this possible?  I'd like to try running a LDO after the power supply and also adding the pi filter after the LDO.  Is this feasible?  Or perhaps there is a better option?

Thank you for any help you can provide!

[T3sl4co1l]

What values did you use in the Pi filter?

What is the nature of your signal source(s) and outputs?  It could be common mode (i.e., noise in the ground lead(s)) as well.

Tim

[Pasky]

Ah, my mistake.  I'm not actually using an inductor, but a ferrite bead.  I was under the impression that these two items were one and the same.  Unless I misinterpreted what I read, they are not.

Here is a crude drawing of the pi filter I used:

[mariush]

Pasky, there are linear regulators that have very little voltage drop, like 50mV at 100mA or 0.1-0.3v at 500-800mA .  It's just a matter of searching for one at sites like Digikey or Farnell, here's for example a Farnell search for fixed 5v ldo regulator : http://goo.gl/20bgSl  - they're sorted by the lowest voltage drop so pay attention to maximum current (feel free to filter by current you need as well)

So your power supply outputs 5v... if you can tweak it somehow to output 5.1-5.5v it would be great, and then the ldo would output 5v to your board. But even without tweaking the voltage, such linear regulators would simply output the input voltage, minus their internal voltage drop, so your board would see 4.8-4.9v or something like that.  I doubt your board really cares so much about having the exact voltage of 5v.

But you have to be careful, as some switching noise may pass through the linear regulator, or the meer proximity of the switching power supply to your board may introduce some noise. 

[Rog520]

I don't know the details of your power supply, but I would do as mariush suggests: try tweaking the feedback on the supply to produce a slightly (~200mV should be more than enough) higher output. Then find an LDO with a low enough dropout to make use of the slightly higher rail voltage and use an appropriate filter on the resulting output. If both the 12v and 5v rails are derived from a common transformer, you'll end up with a slightly higher 12v rail, too.

Your power supply may also be radiating EMI, either from components on the board itself or via the output wires. Try powering the video board from a linear supply while your switching supply is powered up (but not connected tot he video board) to see if this is the case. You might have to load down the switching supply with *something* in order to get it to produce significant EMI. If you can introduce noise by turning on the switching supply -- without it being connected to your video board -- then you'll have to do some shielding of the supply and/or noise suppression right at the output before it has a chance to radiate. Or just make life easier and go with a linear supply to begin with  .

[Pasky]

I see, thank you all for the advice!  I can indeed adjust the 5V rail slightly.  If all it takes is 200-400mV more than that will be no problem.  I will go ahead and order a LDO and give this a shot.  Thank you!

[T3sl4co1l]

Oh, and how much current draw is the video board?

The problem with ferrite beads in power filtering is, they are low values to begin with, and they get much worse under DC current (saturation).  "Inductors" are designed to handle DC, even if they look identical (you can find both multilayer chip "inductors" and multilayer chip "beads").

Your circuit may've helped with the highest frequency noise (>20MHz?), but this also depends on good layout, because such frequencies don't like to stay in wires.  So it's hard to say how effective it was.  Of course, it's hard to say which frequencies are most important to the video board, too.  From your description, it might've been doing exactly what it was capable of, so it could just a matter of needing more.

The problem with LDOs is, their performance at higher frequencies usually sucks, especially if you need that function at low drop conditions as well (the datasheets never show measurements for less than 1 or 2V of drop).  Some devices are specifically designed to achieve good filtering, so shop carefully.  The spec you're most interested in (besides dropout voltage and current rating) is PSRR, graph vs. frequency, being above say 40-60dB at all frequencies.

Probably the best solution will be both combined; and also try an inductor in the ground as well, or a ferrite bead around the power cable.

Tim

[Pasky]

  After the video board and device are hooked up, the current draw is about 5 Amps, so the LDO would need to be able to supply 5A.  I saw this on digikey (http://www.digikey.com/product-detail/en/LM1084IT-5.0%2FNOPB/LM1084IT-5.0%2FNOPB-ND/363556) and from my understanding it will do the trick if I feed it 5.1V  (The power supply is capable of supplying 15A on the 5V rail)  . 

  Would you please elaborate or link me to an explanation (or what I need to search for myself) the benefits of how and why to use an inductor on the ground?  I'd love to attack this from all angles.   I'm unsure how you connect the inductor to ground or what it does.  Do you mean in parallel or is there more to it than that?  Thanks again!

[madires]

I'd add an LC filter. L around 10-20µH and C maybe 100µF per Ampere.

[T3sl4co1l]

Like the stuff at the line input here (but ignore the ground, and the capacitors connected to it): http://seventransistorlabs.com/Images/LED_Light.png

If the inductors are coupled (as shown), you have a common mode filter.  The most basic example of which is a ferrite bead around both wires.

If they are uncoupled, you still get some benefit over having just one inductor at all.

It's still not apparent whether common mode filtering will help you any, so at this point it's just a "cover your ass" option.  If you build the best LCLCLC.. filter and knock off all the (differential mode) noise, but it's still not good enough, that would be a strong indication.

Agree, 10uH sounds like a good starting point.

That LM1084 has pretty poor filtering -- notice how rejection drops sharply above merely 1kHz (Fig.4), and is hardly 20dB at 100kHz -- right where you need it the most (switching noise is usually in the 100kHz to 20MHz range, plus plenty of opportunity for RF noise in the 20-200MHz range, sometimes more..).  20dB is only a factor of 0.1, so it's still doing something, but not much; and if the trend continues, it'll be passing 1MHz+ completely unfiltered.

It also has high dropout (0.7-1.4V), so you need over 6.5V input to guarantee 5V out.

Tim

[Pasky]

Thanks again everyone for the advice!

One last question.  Is there a difference between an inductor that uses a ferrite core as it's main material  such as (http://www.digikey.com/product-detail/en/744772200/732-3765-ND/3073923) and a ferrite bead?

[Phaedrus]

Inductors can use a number of materials for their cores, ferrite being one of them. A ferrite bead is simply a specialized type of inductor, meant for AC applications mostly. General purpose inductors are generally meant for power circuits, like the one you're currently working on.