Filtering switching noise with ferrite beads and linear regulator

[ramicio]

I need to take the 12 volt rail from a computer PSU and get 5 volts of clean DC out of it.  500 mA is what it will be seeing.  I have been researching this, but I cannot find any examples.  How do I choose a ferrite bead?  How do I know what value capacitors to put on it?  I am going to be using a LM2940 LDO regulator.

[KJDS]

What frequency is the PSU switching at?

[ramicio]

I have no clue, nor do I have the equipment to measure it.  It should be a very broad range, since what I am making will outlast the computer it's going into.  Let's assume 10 kHz to 100 kHz.

[free_electron]

I have to build a ladder, but i don't know how long ot needs to be and i also have tools ..

You'll have to experiment then....


Modern computer psus go up to 500 khz and beyond....
But there is not only yhr hf noise. What about the ripple ?

I would first filter using a big inductor with an iron core. That will stop a lot of garbage and dampen ripple. Then a big fat capacitor shunted by a few small ones.
Then a good linear regulator followed by another output filter.

In order to dimension that filter i need more i formation about the load...
Is it something that switches ? Something linear ?

[ramicio]

I just don't know.  I don't have the equipment nor money to experiment building circuit after circuit until I get something that works, and then have it be able to work in another future setup with a different switching frequency.

The load is an LED audio spectrum analyzer.  All the maker said is that I need something capable of delivering 500 mA.  He's very closed about the design and I'm not impressed with the fact that it doesn't have FFT, but some digital potentiometer to act as RC filters to get the frequency bands.  The thing's own power supply is faulty and creates random spikes that he says can be fixed by removing that supply and powering it with a linear supply.

[SeanB]

5V at 500mA is easily done just using a 7805 on a heatsink, about as big as the ones inside the supply you are using.  Filter is a 1R- 2R2 5W series resistor in the 12V line, and a 1000uF 25V capacitor after it. Right at the regulator input and output you have a 100-470n ceramic sisc capacitor, and a 100uF 16V capacitor at the output right near the regulator. This will work well for most loads and will have about the best low cost noise performance. You can use your LM2940 in place of the 7805, just note that the output and input capacitors must be low ESR types, generally most 105C electrolytics will do here.

[ramicio]

What is this filter resistor you speak of and how would 5 watts be enough?

[SeanB]

Big white wirewound ceramic resistor, the common one you see all around you in equipment.

Like here

[ramicio]

I know what those are, I just wondered how it's a filter and I had no idea what 1R or 2R2 meant.  And how would 5 watts be enough?

[c4757p]

Google "RC filter".

1R, 2R2 = 1 ohm, 2.2 ohm.

How would 5 watts not be enough? I don't see the problem. P = I²R.

[SeanB]

RC filter, resistor value is anything between 1 ohm and 2.2 ohms, as they will work without getting too hot with 500mA through them. Will actually handle 1A, but then the resistor will be running at around 100C above ambient. Will drop around 1-4V across it, but then again otherwise you will be dumping this as heat in the regulator itself.

[vlf3]

Assuming the noise is switch mode derived, just the use of an iron toroidal-core, with common-mode winding should stop the problem... e.g.

http://www.emo.org.tr/ekler/5730ca725508e90_ek.pdf

[free_electron]

I know what those are, I just wondered how it's a filter and I had no idea what 1R or 2R2 meant.  And how would 5 watts be enough?

Give it up. This is beyond you.
You have :
- no money to experiment
- no equipment
- no tools
- no clue what 1r or 2r2 stand for on a resistor
- only vague description on what it is you have to do

yet you bash a design because 'it doesn't use fft'

I suggest you seek a different hobby. Basketweaving is popular these days ...

Topics like this make my hair curl...

[AndyC_772]

I agree completely, this project is the blind leading the blind. Even if you do come up with something that works, it'll be by chance, and probably won't be working for the reason you think it is. You've not even mentioned grounding, which I strongly suspect is at the root of your problem.

A PC power supply switches at anything from 50kHz upwards, but that's before you consider by far the most important aspect of the noise which is the RF component of the switching spikes. The true noise bandwidth will extend from a few Hz (coming from the switching jitter and instability in the control loop) up to hundreds of MHz, and no single technique works across that entire range.

You need to know what noise is affecting the circuit, where it's coming from, and how it's coupling into the circuit in a way that actually affects its operation. Only then can you determine whether a filtered supply from the PC will even make any difference. Sorry to be blunt, but speculation and guesswork isn't going to solve this one quickly or easily.

[ramicio]

I know what those are, I just wondered how it's a filter and I had no idea what 1R or 2R2 meant.  And how would 5 watts be enough?

Give it up. This is beyond you.
You have :
- no money to experiment
- no equipment
- no tools
- no clue what 1r or 2r2 stand for on a resistor
- only vague description on what it is you have to do

yet you bash a design because 'it doesn't use fft'

I suggest you seek a different hobby. Basketweaving is popular these days ...

Topics like this make my hair curl...

Hey, don't comment if you have nothing to say.  I don't have an oscilloscope, how does that mean I have no tools?  Sorry I didn't understand what 1R or 2R2 meant when someone could have just said 2.2 ohms.  I've never seen anyone use that terminology before.  Yes, I know a capacitor and a resistor are an RC filter, I just didn't glean that because I parsed the sentence as English, not electrician.  I'm a hobbyist, not a professional.  Why isn't there a middle finger smiley?

Everyone else, you've been helpful and patient.  Thank you.  I'm new to this area of "the hobby."  The only information I've found on this was some Jim guy talking about a ferrite bead.  vlf3 helped by posting a nice article that gives examples.

[c4757p]

Hey, don't comment if you have nothing to say.

Just because it's a harsh opinion doesn't mean it's nothing to say.

I don't have an oscilloscope, how does that mean I have no tools?

In the same way that if you lacked a wrench you could not change a tire with string, chewing gum and wishful thinking. That is the tool for this job.

Sorry I didn't understand what 1R or 2R2 meant when someone could have just said 2.2 ohms.  I've never seen anyone use that terminology before.

Pay more attention. It's quite common. The resistors themselves are likely to be marked that way.

Yes, I know a capacitor and a resistor are an RC filter, I just didn't glean that because I parsed the sentence as English, not electrician.  I'm a hobbyist, not a professional.

You know a cap and a resistor form an RC filter, you were told to use a cap and a resistor and even told that it would be a "filter" and you didn't recognize that it's an RC filter?

Why isn't there a middle finger smiley?

Because the correct response to constructive criticism is "Thank you", even if you don't like the answer. Reality is under no obligation to be nice. The only place for "fuck you" is in jest or in response to a complete asshole, and nobody here is joking or appears to be an asshole.

[ramicio]

His entire response was that of an asshole's, with his assholish basket weaving BS.  His criticism was not constructive, it was "go away because you're too poor to afford this hobby."

[marshallh]

I have received verbal whacks upside the head, proper response is to actually address the issues instead of having hurt pride... Suck it up and take the opportunity to teach and better yourself

[c4757p]

I thought the basket weaving comment was a joke, but either way it doesn't make a difference. Even if it was meant to be an insult, one insult doesn't invalidate everything else somebody says. "Go away because you're poor" wouldn't be very nice, but unfortunately things cost money. You can't do some things without an oscilloscope, and whining that you're too poor for one isn't going to make those things doable without it.

[TerminalJack505]

Hey, don't comment if you have nothing to say.  I don't have an oscilloscope, how does that mean I have no tools?  Sorry I didn't understand what 1R or 2R2 meant when someone could have just said 2.2 ohms.  I've never seen anyone use that terminology before.  Yes, I know a capacitor and a resistor are an RC filter, I just didn't glean that because I parsed the sentence as English, not electrician.  I'm a hobbyist, not a professional.  Why isn't there a middle finger smiley?

Everyone else, you've been helpful and patient.  Thank you.  I'm new to this area of "the hobby."  The only information I've found on this was some Jim guy talking about a ferrite bead.  vlf3 helped by posting a nice article that gives examples.

I'm guessing this is that "Jim guy" you are referring to:



That's the late Jim Williams.  You'd do well to following any advice Jim gives regarding electronics.  He's very well respected.  Here's the app note he refers to.  At the very least read appendix A.

It sounds like you just need to do a little reading on ferrite beads.  Manufacturers will always have material to help make a sale you out.  Don't know who manufactures ferrite beads?  Guess who does: Digikey, Mouser, etc.  I bet Digikey even has a Product Training Module on ferrite beads.  (Here's one.)

All you probably need to worry about is choosing beads that have the proper current handling capability.  Basically beads designed for DC power supplies as opposed to bead designed for signals, for example. 

The only other characteristic you might need to worry about is the ferrite's impedance at 100MHz.  I'd just pick one around 200 ohms and be done with it.  Put these in your circuit as described by Jim (figure 4 of the app. note) and this will be about the best you can do considering you have no test equipment.

[ramicio]

I'll look for a scope at some point.  I'm not in a hurry with this project.

[AndyC_772]

The only other characteristic you might need to worry about is the ferrite's impedance at 100MHz.  I'd just pick one around 200 ohms and be done with it.  Put these in your circuit as described by Jim (figure 4 of the app. note) and this will be about the best you can do considering you have no test equipment.

Many ferrite beads are quoted @ 100 MHz, and in many cases the manufacturer's data sheet includes a plot showing the impedance across a much wider frequency range. The proper way to choose one requires knowledge of the noise frequency that actually exists and which is causing the problem - for example, following a trip to an EMC lab.

If the product is an audio spectrum analyser, it might be sensitive to audio frequency noise (where no ferrite bead will have any useful effect whatsoever), or it might be affected by much higher frequencies. There's no way to know without experimentation.

[veryevil]

This may be a silly question but why can you not use the 5V rail on the computer PSU? Why does it have to be the 12V rail?

[ramicio]

This may be a silly question but why can you not use the 5V rail on the computer PSU? Why does it have to be the 12V rail?

Because the circuit needs a clean 5 volts DC.  I don't know any way of starting from 5 volts, cleaning it up, and getting 5 volts back out.

[c4757p]

Because the circuit needs a clean 5 volts DC.  I don't know any way of starting from 5 volts, cleaning it up, and getting 5 volts back out.

A giant LC filter? Lots of L, lots of C, and good quality C (bypass with small ceramic types). That's how it was done in the old days. I have an ancient function generator (vacuum tube days, before simple linear regulators) that manages to get nearly perfect DC straight off the rectifier with only two 56uF capacitors, because it has one bigass choke between them.