Filtering switching noise with ferrite beads and linear regulator

[ramicio]

Too late.  I already ordered the circuit board that will be supplying this power and it's tied to the 12 volt line.  This thing already has some kind of LC crap already and it's flawed...causes random spikes on different bands every few seconds.

[c4757p]

Since huge inductors are hard to come by these days, you can use the primary of a spare mains transformer with the secondary left open. That will easily be upwards of 1H, depending on the transformer.

[c4757p]

It's not flawed, it's "good enough". Those spikes were OK to the designers so they left them. A well designed LC filter will easily strip those out.

[ramicio]

This is all going into a 5.25" empty CD drive.

[ramicio]

It's not flawed, it's "good enough". Those spikes were OK to the designers so they left them. A well designed LC filter will easily strip those out.

It's not "good enough."  They occur with nothing even being input to it.  He even said it's a flaw that he intends to fix in another revision.  I've almost had it with this whole project.  It would have been nice to know how inaccurate it was before purchasing it.

[c4757p]

I'm a bit confused - aren't you trying to get the spikes off the rails from a PC power supply? That's what I meant - I thought you were talking about the internal filters in the power supply. If it's the filters on some add-on board that aren't filtering anything, I'd agree, it's probably a poor design. Still, an LC filter is the way to go, you just have to do it right. In fact, it's probably the only thing that can keep up with those spikes - they'll go right through a voltage regulator because they're too fast.

[ramicio]

No, the [audio] spectrum analyzer has a fault in its own power supply that causes spikes to be randomly displayed every few seconds.  I am trying to give a clean power supply and bypass its own power supply.  I guess I will do an LC filter and then a regulator.

[c4757p]

No, the [audio] spectrum analyzer has a fault in its own power supply that causes spikes to be randomly displayed every few seconds.

Oh, that's just awful.  Is it at least a separate module and not integrated into the main board? If so, try adding chokes inline with the wires. You might have luck with a common mode choke (both power and ground on the same core).

[ramicio]

It's just a small IC and some capacitors and an inductor.  I am just removing them all and feeding it with 5 volts from the supply I am making.  As I said, it's too late now to just do some simple choke stuff.  A related board I made and ordered has the output tied to the 12 volt line.

[ramicio]

I'm noticing these inductors don't have much of a current rating.

EDIT: I do have a bunch of these larger common mode chokes around I can pull off of scrap boards.

[ramicio]

I took some toroidal common mode chokes off a scrapped PSU I have.  I tried the small one with a smallish (220µF) capacitor and there are fewer spikes.  I'd imagine with the larger choke and a larger capacitor, like 2,200µF or 3,300µF,  it should be perfect.  I have no idea what the values of the chokes are.

[c4757p]

I don't think a larger capacitor is going to help much. What you need to do is shunt the large capacitor with multiple small, good quality capacitors (I'd say maybe 100pF, 10nF, 100nF ceramic). As for the choke, it's hard to say that a larger one will have a higher value - it could be just a higher current rating. It's really hard to judge the value of an inductor by looking at it - the core material has a huge effect. I've got some toroids that will give me as much inductance in two turns as some entire small wound inductors...

[free_electron]

Yes, I know a capacitor and a resistor are an RC filter

Are you sure ? Using an R and a C one can make a highpass , lowpass  bandpass, differentiator or integrator.. depends on how they are connected...

Saying an Rc is a filter is like saying 'it's a vehicle'.. ok. but what kind of ?  a bike ? a train ? a car ?

some Jim guy

Ouch. Both Bob and Jim are now preparing to unleash an electronstorm that will wipe out all the data on your harddisk...

You say that spectrum analyser thingy shows 'spikes'. you also say its made with an ic and some LED's. LM3914 ?
who says the spike comes in through the power supply? if that cable that brings in the signal picks up a spike along the way it may very well come in that way. fidgeting with the power supply doesn't do any good then.

the lm3914 needs a high and a low reference. i'm assuming you use gnd as low reference. how do you make high refrence. (please don't say using a potmeter derived from vcc... ) there should at least be a zener or an active regulator there.... with appropriate bypass capacitors.

other thing : how do you feed the analog signal in ? is it properly biased. i'm assuming it comes in through a coupling capacitor. did you provide clipping diodes there ? it is possible if a peak is coming in the input couping capacitor gets charged in such a way that on a further change of the input it actually overloads the lm3914... pusing input above vhigh of an lm3914 gives very weird results ....

Make a nice metal box , make a nice PCB , use feedthrough cap for power in , followed by a nice choke-capacitor lowpass filter, followed by a 7805 with some ceramic caps around it. make sure the case is grounded to the board ground.
Use an input circuit with a lowpass that only lets audio go through , use a shielded cable, maybe a common mode choke there and keep that input line away from noise generators such as the choppers on the motherboard.

There. Happy now ? now you have some more constructive criticism. But i stand by my earlier remarks.
You are gunning blind for a problem that may not even be there. it's like trying to hit a black cat in a dark room , blindfolded... and the cat may not even be there...

It is a bad idea to blindly make a board , slap on some parts and then cry that it doesn't work. it's even a worse idea to raise the middle finger if someone tells you your approach is wrong and that there are easier hobbies.... like basketweaving ( that was a joke by the way )
and that's all i'm going to say about that.

[c4757p]

it's like trying to hit a black cat in a dark room , blindfolded... and the cat may not even be there...

Oh, it's definitely there. And not there, simultaneously in a superposition of - oh wait, you said room...

[ramicio]

You say that spectrum analyser thingy shows 'spikes'. you also say its made with an ic and some LED's. LM3914 ?
who says the spike comes in through the power supply? if that cable that brings in the signal picks up a spike along the way it may very well come in that way. fidgeting with the power supply doesn't do any good then.

the lm3914 needs a high and a low reference. i'm assuming you use gnd as low reference. how do you make high refrence. (please don't say using a potmeter derived from vcc... ) there should at least be a zener or an active regulator there.... with appropriate bypass capacitors.

other thing : how do you feed the analog signal in ? is it properly biased. i'm assuming it comes in through a coupling capacitor. did you provide clipping diodes there ? it is possible if a peak is coming in the input couping capacitor gets charged in such a way that on a further change of the input it actually overloads the lm3914... pusing input above vhigh of an lm3914 gives very weird results ....

Make a nice metal box , make a nice PCB , use feedthrough cap for power in , followed by a nice choke-capacitor lowpass filter, followed by a 7805 with some ceramic caps around it. make sure the case is grounded to the board ground.
Use an input circuit with a lowpass that only lets audio go through , use a shielded cable, maybe a common mode choke there and keep that input line away from noise generators such as the choppers on the motherboard.

There. Happy now ? now you have some more constructive criticism. But i stand by my earlier remarks.
You are gunning blind for a problem that may not even be there. it's like trying to hit a black cat in a dark room , blindfolded... and the cat may not even be there...

It is a bad idea to blindly make a board , slap on some parts and then cry that it doesn't work. it's even a worse idea to raise the middle finger if someone tells you your approach is wrong and that there are easier hobbies.... like basketweaving ( that was a joke by the way )
and that's all i'm going to say about that.

I didn't have anything to do with making the spectrum analyzer.  It's powered by a ATmega64.  It uses a digital potentiometer to get the various bands.  The spikes are an admitted fault from the person who designed/made it.  It's not noise from the audio source.  It happens with nothing connected to the input.  He won't explain the problem in detail.  It's a fault on this specific revision.

It's all going in a gutted 5.25" CD drive, but I might just bend up a piece of thin aluminum plate instead.

[David_AVD]

OMG this thread is entertaining and disturbing at the same time.   

[c4757p]

I didn't have anything to do with making the spectrum analyzer.  It's powered by a ATmega64.  It uses a digital potentiometer to get the various bands.  The spikes are an admitted fault from the person who designed/made it.  It's not noise from the audio source.  It happens with nothing connected to the input.  He won't explain the problem in detail.  It's a fault on this specific revision.

If I had to make a diagnosis, I'd say the design's probably retarded. Not "retarded" as in "that paint color's retarded, I don't like it", but more as in "it's going to hurt itself, put it out of its misery". Why are you averse to modifying it?

He won't explain the problem in detail.

Subdiagnosis: He doesn't have a damned clue what the problem is. Not even a hunch.

[ramicio]

I didn't have anything to do with making the spectrum analyzer.  It's powered by a ATmega64.  It uses a digital potentiometer to get the various bands.  The spikes are an admitted fault from the person who designed/made it.  It's not noise from the audio source.  It happens with nothing connected to the input.  He won't explain the problem in detail.  It's a fault on this specific revision.

If I had to make a diagnosis, I'd say the design's probably retarded. Not "retarded" as in "that paint color's retarded, I don't like it", but more as in "it's going to hurt itself, put it out of its misery". Why are you averse to modifying it?

He won't explain the problem in detail.

Subdiagnosis: He doesn't have a damned clue what the problem is. Not even a hunch.

Averse to modifying what?  I have to come ask for help in designing something to clean up noise in a power line, you think I could actually do something like design and code a spectrum analyzer?  I told ya it was retarded.  I don't even know why someone would think of using a digital potentiometer to be varied very fast to get bands instead of using some open source FFT library.  Just about all LED spectrum analyzer kits have done something like switching between frequency filters to feed a bar graph driver whose output goes through a column selection section.  This thing has 20 columns and 20 rows.  That would be 40 bar graph drivers.  It would probably work flawlessly and not have a visible refresh rate pulse to it, but the circuit would be massive thanks to the 3915 being only through-hole DIP.  It would also need 401 signals to interface between boards instead of 40.  It was a choice between a fast response and low accuracy or high accuracy and something that would probably not be so responsive on an ATmega64.

He likely does have an idea, he is just very closed and secretive about the design and source of the thing.  It was not there in previous revisions.  It's a product for the guy, not a project.  He also likely didn't tell me because it would just go over my head anyway.

[c4757p]

Not to be rude - but you don't even have an oscilloscope. How do you even know the noise is from the power supply? Hell, for all we know the spikes are being generated in software by an erroneous algorithm. You are likely to be barking up the wrong tree.

[ramicio]

Because it's not noise.  It's a single spike that occurs every few seconds.  The guy said to use a linear supply and bypass the one on the board.  If you think this is all wrong, go to youritronics.com and try to get it out of him what the deal is.

[c4757p]

To me that really says it's not on the power supply. What power supply has sub-Hz spikes?? Spikes over 100kHz, sure, but below 1 Hz? Nope. I'm guessing it's a stupid software error - some buffer is overflowing periodically, the watchdog timer is expiring, he forgot to bang his head on the desk at just the right angle before coding - whatever, but I highly doubt a "spike every few seconds" is coming from the power supply.

[ramicio]

The go over there and tell him that he's full of it.

[c4757p]

If you think this is all wrong, go to youritronics.com and try to get it out of him what the deal is.

The go over there and tell him that he's full of it.

If you cannot get a straight answer from him yourself, you should not bother to do this. Why are you even doing it, anyway?

I'm trying to help you figure this out because I wanted to help you. I really couldn't be bothered to go tell this guy how much he sucks, and I doubt he'd give me any more information than he'd give you. I've got much more interesting problems to work on. You've reached a dead end unless you can get this guy to move his ass and fix his design, and I'm not going to keep dancing around his inabilities for you.

I promise this isn't power supply related. No power supply hiccups like that. But there's no more I can do for you at this point.

[oPossum]

Info from the mfg:

http://www.youritronics.com/forum/post2626.html#p2626

And one thing to consider is that the current revision of the spectrum analyzer motherboard suffers from a hardware bug that manifests itself by randomly showing a high signal on a random column(a spike) for a fraction of a second like there was a signal on that particular frequency . This effect is not visible and it does not disturb the use during normal operation (ie with audio signal being applied) because the "spikes" are practically covered by the real signal coming in and being displayed. The effect is however visible when no signal is applied to the spectrum analyzer and all the columns are down at minimum: at random intervals 3-5 seconds a random column will spike for a fraction of a second.

http://www.youritronics.com/forum/post2633.html#p2633

Unfortunately this is a hardware/design issue with the onboard power supply circuitry. The problem manifest itself on all devices
and can only be fixed by a hardware revision. The SMPS circuit (ST1S10) behaves strangely and gives out spikes on the 5V rail, probably due to internal
circuitry changing operating modes maybe to maintain high efficiency or even possibly due to the output capacitor ESR. In the next version of the spectrum analyzer the problem will be addressed.

In the meantime the only fix is to remove the ST1S10 and supply the 5V externally from a clean linear power supply, preferably a linear regulator like a 7805.

http://www.youritronics.com/forum/post207.html#p207

I guess one of the biggest problems of V2 was the noise coming out of the inductor. The noise was being generated by the inductor reaching Isat together with the rather low frequency converter. The ST1S10 used on V2.1 solves this issue by working on a higher frequency thus requesting lower value inductors which have higher Isat.

[ramicio]

There it is.  I forgot he said that, it was so far back.  And it actually does spike during audio being played.  One nice thing about this was that he told me about this spiking crap after I paid.

If it was from the onboard SMPS switching modes, why would it happen with a completely different supply with that whole thing removed?  He's basically stated that he has no interest in improving this thing other than this spiking issue, despite almost 3 years ago saying there is going to be a version 3 with FFT.  There is someone else who basically makes the same thing, but with FFT, but it's one giant board.  That won't work for me.  And in the videos the thing seems pretty slow.  Since it just sits there in the video, I have no idea if it's the camera that is slow or the device.