Re: gEDA-user: Edge ringing filtering

[Karel Kulhavy <clock@xxxxxxxxxxxxx>]

On Mon, Mar 06, 2006 at 04:52:40AM +0900, John Doty wrote:
> 
> On Mar 6, 2006, at 2:31 AM, Karel Kulhavy wrote:
> 
> >On Thu, Mar 02, 2006 at 03:05:35PM -0800, joeft wrote:
> >>If I understand your description correctly, you need to raise the
> >>impedance of the 2.5cm wire at 200 MHz to keep it from conducting
> >>interference to the outside of the box.  Is this wire used for power?
> >>If so, get a ferrite core (toroid)  and put it over the wire; pass  
> >>the
> >>wire through it a few times if you can.
> >>
> >>Also, 100 nF caps may be beyond self resonance at 200 MHz  
> >>depending on
> >>their construction.  Check the manufacturer's data sheet.  It may  
> >>help
> >>to put a smaller cap (100 pF) in parallel with the 10 & 100 nF if 200
> >>MHz is your problem.
> >
> >But together with the other caps they form a complex RLC network.
> >What's the typical ESR of these caps so I can simulate it and look  
> >what
> >happens there?
> 
> Simulating this kind of thing is a tricky applied physics problem.  
> Approaching it from an RLC viewpoint is difficult because the EM  
> fields of interest are loose outside the components (and that's the  
> problem ;-).
> 
> ESR of capacitors is usually negligible in problems like this.  

Isn't - if you neglect the ESR and assume it's zero, then you get that
two connected capacitors will have infinite impedance between their
zeroes and will always fail to filter at that frequency no matter how
hard you try. So if the clock frequency hits that point you can throw
the device away - it will not only enormously radiate but stop working
because the noise on the power line will build up until the gates burn
out from overvoltage.

So with probability of 0.000nothing the device will fail otherwise work,
and become a nightmare of the testing process.

In practice this doesn't happen because the capacitors have ESR.

This remembers me those thought of the type "if I passed the zebra
crossing 2 seconds later, I would be overrun by a car). If I passed
the zebra 1 second later, there would be a 70dm^3 vacuum hole in the air
at the place I actually was, which would collapse and knock the incoming
car apart or something like that, plus I would have to go to hospital
for treatment of decompression from the air dissolved in my body.

Analysing the universe and changing one subtle detail can lead to
catastrophic results in the simulation ;-)

CL<
> There's generally no concentration of magnetic energy in the cap, so  
> to get parasitic inductance you need to consider where the return  
> current flows in the circuit and calculate the inductance of that  
> loop, with the cap considered as just a slug of metal. Beads are  
> tricky to model as well: the impedance of a bead designed for EMI  
> suppression is generally intermediate between an inductor and a  
> resistor over a very wide frequency range. If you're just interested  
> in one frequency an LR network with Q~1 and impedance matching the  
> data sheet might work, but over a wide frequency range the bead acts  
> like a fractal series-parallel LR network.
> 
> Since the stray fields overlap with other subsystems in the circuit,  
> they're involved. Even getting a crude model may force you into a  
> multimode transmission line approach.
> 
> Every engineer I know approaches these problems by experimentation  
> guided by experience and "conventional wisdom". If that fails they go  
> looking for a physicist (me ;-). Usually the "analysis" is a  
> blackboard sketch, a very rough estimate of the main EM effects, and  
> then I'll say "try a bead there", "add a return wire here" or  
> something. The only time I went to the trouble of making a detailed  
> model was for a wiring harness for a spacecraft when I didn't have a  
> full scale prototype to experiment with. A model successfully  
> reproduced the problem and correctly predicted the consequences of  
> changing the cable construction. Twisted pairs and strategically  
> placed beads solved the trouble, but finding this solution through  
> detailed modeling was an enormous job, only justified by the very  
> special circumstances.
> 
> John Doty              Noqsi Aerospace, Ltd.
> jpd@xxxxxxxxxxxxx
>