RF filter design questions

1. Get yourself a copy of:
   formatting link
   (I guess it's available freely online, at the moment; seems kind of weird to me. In any case, print is available too.)

1. Read through the Theory, and Use, sections. See how the examples turn coefficients into real parts values.

The jist of it is:

- Filter tables are calculated for unit frequency and impedance.

- sqrt(L/C) is the network impedance. Scale L and C proportionally to achieve the desired Zo (L up, C down to raise Zo).

- 1 / sqrt(L*C) is the network frequency (within a constant). Scale everything proportionally to achieve the desired Fo (L down, C down to raise Fo).

Tables are calculated based on 1 ohm and 1Hz (unit ohm, unit frequency) because it's the easiest form to start with. You must apply these ratios to calculate real component values.

There are a great many types of filter designs in this book. There are your standard lossless Bessel/Butterworth/Chebyshev kinds; and there are coefficients that have been adjusted to compensate for assumed losses in components; and there are tapered impedance (Zin != Zout) coefficients; and there are single-loss (Zin finite, Zout zero or infinite) coefficients; and on and on.

There is a restriction: if the filter network is lossless, then there must be a loss element either at the source, load, or both. If there is no loss element, then energy circulates forever, and the whole thing is a terrible mess. (This is easier to illustrate in SPICE than in real life, because there are no truly lossless components in real life!)

A filter works by reflecting* undesired energy, so that the excess is either returned to the source, or absorbed by the source's own resistance.

If you don't know what you need, you should probably choose an ordinary Butterworth design (with or without loss adjustments), and design your circuit so that the reflected (undesired) power is dealt with somehow (often, something as crude as a -3dB pad).

*You can make an absorbing (constant-resistance) filter by adding a damping network and massaging the other component values. An occasional resident here derived some examples:

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com 


 wrote in message  
news:17023b99-4859-44ae-b593-f53744ddb663@googlegroups.com... 
> Could some electronics?RF design guru please help ? 
> 
> I was looking over some online tutorials on RF filter 
> design, and some questions came to mind. 
> 1. Why is the source and load resistances always set 
> to 1 Ohm that is for a maximally matched filter, why 
> could the source and load imppedances be greater than 
> 1 Ohm ? 
> 2. What is an unit element, and what esactly is its 
> functionality ? 
> 
> Basically, it is all fine to start with a normalized 
> filter and read in filter coefficients from a table, 
> but is there any restriction that no higher value source 
> or load impedances be used ? 
> 
> Any hints/suggestions/pointers would be of immense help. 
> Thanks in advance.

Filter tables are usually normalized to R = 1 ohm and W = 1 radian/second. You can scale them to your required impedance and frequency.

if C' is the new cap value and L' is the new inductor, and R' is your new impedance

C' = C/(R'*W')

L' = L * R' / W'

where W' = 2*pi* new frequency.

There are also asymmetric filters, the most common being 1 ohm on the input and infinite on the output. Or vice versa, since LC filters are reversible.

Get the Williams and Taylor filter design handbook. An old edition is fine.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Then there is:

AFAIK the nag screen goes away for those who buy the AADE LCR meter. Which supposedly is quite good if you need one.

--
Regards, Joerg 

http://www.analogconsultants.com/

The LC meter is very good. The guy died, and I don't know what will happen to his business.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

You may also want to get a copy of "Experimental Methods in RF Design" by Wes Hayward et all, and sold by the ARRL

I wouldn't be surprised if it didn't have filter tables for amateur bands and real-valued components, filter design software, or both.

--
www.wescottdesign.com

I bought one of his last ones last April--he was only selling it as a kit at that stage. Works great.

RIP Neil.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Link is table of contents and preface only.

That's a good text to have. It has tables, and software for design.

Best regards Frank IZ8DWF

As already pointed out, the 1 ohm value normalizes the tables. This was for practicality when working with tables. Users simply scale as needed. Norma lization is not important or visible when using software filter generators.

The purpose is to make distributed filters practical. A unit element is a "redundant transmission line."

1. Physically separate transmission line stubs
2. Transform series stubs into shunt stubs, and vice versa
3. Change impractical characteristic impedances into more practical ones

See the so-called Kuroda identities. A decent filter book will give simple examples that illustrate the benefit.

Doubly terminated and "matched" filters have the lowest component sensitivi ty. The 2 terminations on doubly terminated bandpass filters can be quite different, and still meet the matched condition because the bandpass can al so serve as an impedance transformer.

It is a very specialized field with few experts. It consumes more like a c areer than a hint. lol.