Re: Is there a theorem about the anti-GBW of a filter, for some given amount of reactive energy?

Get a copy of my favorite filter book...

"Synthesis of Filters" Jose Luis Herrero & Gideon Willoner

Library of Congress Number: 66-27547

(Peruse the old/used bookstores...

Math intensive, transforms required, not for those who can't already do a hand-math filter design with Laplace (and need to fall back on FilterPro :-)

But excellent for handling depth of stop bands, etc. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142     Skype: skypeanalog  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

What, not the filter book written by me*? ;-)

(*Alas, a guy with a coincidental name, not actually me.)

An example use would be making a filter as small as possible, physically. Say for PWM, or EMI, or... anything. I don't know of any filter design guide that discusses this (all too practical and messy) aspect.

Also, don't know of any transformer / coil design guides that discuss the same. Power density (literally) goes as Bmax^2 / (2 * mu_0), but that says nothing about the actual shape of the core, or any means to relate that to its losses. Sure, you can make a tiny coil, and run it up around saturation, but if it's so tiny... just how hot is that sucker going to get?

Calculations: ran through it on paper, got: N = -ln A for A = gain of the filter in the cutoff region (A

How about being really nasty? Take incoming analog waveform and sample it; real time FFT and digitally remove what you do not want, then FFT back into a DAC.

Williams and Taylor "Electronic Filter Design Handbook" ISBN 0-07-070430-9 for the 3rd edition. There's now a fourth edition

0071471715/ref=pd_cp_b_0

Sadly, it's a handbook, not a theoretical exposition. It points you at the theory which it exploits and makes accessible.

a hand-math filter design with Laplace (and need to fall back on FilterPro :-) But excellent for handling depth of stop bands, etc.

Say for PWM, or EMI, or... anything. I don't know of any filter design guide that discusses this (all too practical and messy) aspect.

Filter design is already a low volume market and there was no guarantee tha t it would be blessed by something as good as Williams and Taylor.

Transformer/inductor design is an even smaller pool and hasn't done remotel y as well.

same. Power density (literally) goes as Bmax^2 / (2 * mu_0), but that says nothing about the actual shape of the core, or any means to relate th at to its losses.

There's no necessary relationship. A non-linear relationship between curren t and magnetic field is no guarantee of power dissipation in the core.

t's so tiny... just how hot is that sucker going to get?

It's going to get hot enough that convection, conduction and radiation can dissipate the power being deposited in the core and its windings.

If you know the power being lost in the inductive element, and it's thermal resistance to its environment, you know how much hotter it has to get than its environment.

int, and it's when the cutoff frequency is about 1/3 the frequency you nee d attenuation at (give or take rolloff, which will be noticeable at small N), with N given by the attenuation as above.

370kHz. Lower f_0 and, though you save stages, the components have to be bigger; more f_0 and you don't save much on size, but N keeps growing.

And bigger N usually means tighter-tolerance components.

--
Bill Sloman, Sydney

Ok, so W&T not so much, H&M more promising.

says nothing about the actual shape of the core, or any means to relate that to its losses.

Why not? It is to Steinmetz. But real materials suck. I've seen powdered iron materials with a B exponent as low as... 1.7ish I think. No material is even on frequency, always another exponent. In fact, the last ferrite I looked at (EPCOS/TDK N49) has p_c ~= Bpk^3 or so. Now that's a nasty exponent, I might even want to calculate a harmonic or two for that one.

Thanks, Dr. Tautology. You've been so helpful.

Yes, which would be valuable design information.

If you're concerned with the flatness and sharpness around the stop band.

In this case, I couldn't give two s**ts, I just want asymptotic behavior. Accordingly.. it doesn't matter what profile I choose, as long as the asymptote starts as soon as possible (Butterworth) or stays as low as possible (Cheb.). Also, even at a factor of e above cutoff, Bessel just sucks. ;-D

Also considering a notch filter, which isn't always acceptable, but may have applications.

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

he theory which it exploits and makes accessible.

the same. Power density (literally) goes as Bmax^2 / (2 * mu_0), but th at says nothing about the actual shape of the core, or any means to relate that to its losses.

rent and magnetic field is no guarantee of power dissipation in the core.

I - for one - could do with some exposition of the physics involved.

powdered iron materials with a B exponent as low as... 1.7ish I think. No material is even on frequency, always another exponent. In fact, the last ferrite I looked at (EPCOS/TDK N49) has p_c ~= Bpk^3 or so. Now that's a nasty exponent, I might even want to calculate a harmonic or two for th at one.

f it's so tiny... just how hot is that sucker going to get?

an dissipate the power being deposited in the core and its windings.

Your own contribution assumed a great deal more physics than I happen to kn ow. You didn't articulate your concerns in a way that made any kind of phys ical sense, so restating the obvious was a way getting back to basics.

mal resistance to its environment, you know how much hotter it has to get t han its environment.

And you are complaining about not being able to calculate the power being d issipated in the core while not bothering to spell out the way (or ways) in which the core is absorbing power.

Accordingly.. it doesn't matter what profile I choose, as long as the asy mptote starts as soon as possible (Butterworth) or stays as low as possibl e (Cheb.). Also, even at a factor of e above cutoff, Bessel just sucks. ;- D

ave applications.

Notch filters are useful to the extent that they have sharp, deep notches, which take even more tuning/close tolerance components.

--
Bill Sloman, Sydney

Oooh, DC bias curves. Cool.

That would be interesting. But probably not as valuable as it sounds. These materials tend to be rather arbitrary. It would be nice to say DC bias doesn't matter, or that the exponents will always be nice integers like 0, 1, or 2, or whatever.

Another part: the B-H curve is pretty much useless, because it spans from saturation to saturation; and smaller loops, near zero magnetization, have disproportionate loop areas, as do loops at nonzero magnetization (but away from saturation).

There's probably a broad range of ugly time and temperature and history* dependencies to all of these, besides. I can just imagine magnetic domains oozing like goo....or popping like geological fault lines (

).

*History being not just magnetic state, but thermal and mechanical as well. Like how NiZn cores usually come with the advice "saturation or mechanical shock may cause permanent changes in core properties".

Also, supposedly you can magnetize hard steel merely by striking it parallel to the Earth's magnetic field lines. Seems unlikely, but with crystals this nonlinear, who needs common sense?

Well, point being, if you make a coil arbitrarily small, but its Q remains constant (at a given frequency), when does reality intervene? Or if Q doesn't remain constant (which is usually the case),

Uh... it gets hot because of fields, and magic unicorns. Probably.

Scratch that, not magic unicorns. Magnetic unicorns. Totally on topic. Wouldn't want to miss that opportunity.

In any case, it seems impossible to a) describe all the mechanisms by which a ferr[i/o]magnetic material might dissipate (and thus, the possible expressions relating relevant variables), and b) to come up with a sufficiently descriptive theory to predict those losses (even assuming you could select the relevant mechanisms enumerated in (a)). So you're SOL either way and stuck with the empirical data.

But you have to generate those data yourself, because even if the manufacturer provides some, it's only valid for the core and fixture they measured it on.

Which... brings us back to tautology, actually, because it seems the only way to prove your component has known losses is to measure it under those conditions, then wind exactly the same part and call it 'proven by theory'.

Exact, in this case, meaning, replicating whatever the core sees. So you can still go from transformers to energy-storage inductors, in any arbitrary number of turns. So you have some freedom in that space...as long as it's on the same exact core. ("They can have any color they want, as long as it's black"?)

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

You were talking about heat dissipation in the core as if it related to hy sterisis in the magnetic field. There was a suggestion on the Wikipedia pie ce on Steinmetz that he'd worked out some such connection, but nothing I fo und spelled out what that connection might be.

t if it's so tiny... just how hot is that sucker going to get?

n can dissipate the power being deposited in the core and its windings.

know. You didn't articulate your concerns in a way that made any kind of physical sense, so restating the obvious was a way getting back to basics.

s constant (at a given frequency), when does reality intervene? Or if Q d oesn't remain constant (which is usually the case).

When the wire gets too fine to wind? Either because it breaks, or because t he insulating lacquer becomes so much thicker than the copper that it isn't worth winding the coil in the first place?

thermal resistance to its environment, you know how much hotter it has to g et than its environment.

g dissipated in the core while not bothering to spell out the way (or ways) i n which the core is absorbing power.

Probably not. Are you planning on running on some kind of John Larkin vanit y ticket?

ich a ferr[i/o]magnetic material might dissipate (and thus, the possible e xpressions relating relevant variables), and b) to come up with a sufficie ntly descriptive theory to predict those losses (even assuming you could s elect the relevant mechanisms enumerated in (a)). So you're SOL either wa y and stuck with the empirical data.

It is true that transformers and inductors is not a subject where there's a well organised and accessible body of knowledge that helps you work out wh at's going on.

There are bit and pieces that do help. John Chan's model of a hysteretic in ductor in LTSpice does work, and I've found it helpful.

Wasting bandwidth on complaining about what doesn't work is less helpful.

manufacturer provides some, it's only valid for the core and fixture they measured it on.

way to prove your component has known losses is to measure it under those conditions, then wind exactly the same part and call it 'proven by theory'.

Actually, that's "proven by experiment".

can still go from transformers to energy-storage inductors, in any arbit rary number of turns. So you have some freedom in that space...as long as it's on the same exact core. ("They can have any color they want, as long as it's black"?).

That's comforting. Of course real windings have intra-winding capacitance, so some of the current you stick into the winding isn't going along the wir e, but rather skipping from layer to layer, and doesn't have the same effec t on the core as the current that stays in the wire.

--
Bill Sloman, Sydney