How do I design a Klopfenstein Taper? 1

[bpelec]

I need to impedance match between a 50 Ohm channel and a 100 Ohm channel, and the channel has a bandwith around 2 Ghz.

It has been suggested to me that I use a "Klopfenstein Taper", as it is apparently a near optimum design.

My understanding is that I will essentially be custom designing a transmission line to sit between my two channels, and the Klopfenstein Taper has a near optimum geometry. For example, I guess I could also use a triangular taper or an exponential taper.

Here's my question:

I am new to impedance matching using tapers, and am finding the material I have found on the Klopfenstein Taper quite complicated. I guess it comes down to how to work out the starting and finishing width of the taper, and the length of the taper, as a function of the impedances and bandwidths involved... Does anyone have any experience with this kind of design, and could anyone point me in the direction of some easy to digest material?

Thank you for your help.

Best regards,

BPELEC.

 

[VEBill]

Lots on Google:

http://www.google.com/search?q=Klopfenstein+Taper

Why not use a 2:1 transformer?

Also, a 2GHz bandwidth is a big deal at L-band (~100% BW), but much less so at visible light wavelengths. What is the center frequency?

 

[Warpspeed]

The optimum approach to this problem depends heavily upon the absolute frequency range. There is a very big difference between matching from say dc to 2Ghz, and matching in a narrow band from 50Ghz to 52Ghz.

Tapered transmission lines are ideal for single frequencies, or narrow (relative) bandwidths. But there will be a lower operating frequency below which they crease to be effective.

For wider bandwidths, transmission line transformers probably offer a better solution. If you can accept the attenuation loss, even a resistive matching pad may be worth a passing thought.

 

[bpelec]

Thank you both for your replies and the information.

The channel is passing high-definition video (HDSDI) and requires a data rate of around 1.5 Gbit. I have been told that I need the pass band to be from about 1 Mhz to 3 Ghz. We thought about a transformer, but I don't think you can get transformers with that kind of wide bandwidth?

I think that HDSDI can tolerate even up to about 20db of loss, so a transmission line transformer might work. However, I must admit that I am not entirely sure what a "transmission line" transformer is in comparison to a normal transformer. Can you give a simple explanation or point me towards some documentation?

Best regards,

BPELEC.

 

[Warpspeed]

There is a lot of information on transmission line transformer theory, try google. Here is just a small sample:

http://rfic.eecs.berkeley.edu/niknejad/ee217sp05/lect10.pdf

1Mhz to 2GHZ may be possible, as the power levels in a video signal are very low. A twisted transmission line wound with very fine wire wound onto a small very high permeability ferrite bead may do the job. There may even be something commercially available, I have no idea.

If your circuit can stand 20 db loss, a resistive matching pad may be a lot less trouble, and probably end up having better bandwidth characteristics.

 

[Warpspeed]

Oops, try this:

http://rfic.eecs.berkeley.edu/~niknejad/ee217sp05/lect10.pdf

 

[nbucska]

Perhaps you could use an actice circuit ( a buffer with
Z1 input and Z2 output impedance )

Plesae read FAQ240-1032
My WEB: <

http://geocities.com/nbucska/>

 

[potcore]

bpelec,

"I guess it comes down to how to work out the starting and finishing width of the taper, and the length of the taper, as a function of the impedances and bandwidths involved"

i think you've hit the nail right on the head. I think its best to get some copper clad board and some etchant and make some tapers of different dimensions etc and test and tabulate/graph results....it will become obvious just which geometry you need to go for from viewing your results.

By the way, was it a typo saying "1MHz" tO 3GHz?.....as i think you said your BW was 2GHZ....and i think you mean 1GHZ to 3GHZ(?)

i once worked in a company where they were doing these tapers...for wideband matching, and there comes a point when its best to put down the books and start thinking along the very logical and to-the-point, practical lines that you have obviously done with your (quoted above) words.

 

[nbucska]

The K. taper will have to be longer with increasing BW
-- in your case it would require an unreasonable length.



----------------------------
Please read FAQ240-1032
My WEB: <

http://geocities.com/nbucska/>