Suggestions on Broadband Matching in Power amplifier

[circuitking]

Hi All,

I am designing a broadband power amplifier 1-20 GHz. I found optimum load impedance for all frequencies with a step of 4 GHz (1G:4G:20G). Now I am confused as how to find matching network using these different load impedances. I would like to know what different techniques availble for broadband matching. Thanks in advance

 

[BigBoss]

If the Optimum Load Impedances are far away from each other ( probably they are ) , you will find a Impedance Region that more or less satisfies those Optimum Impedances then you can design your matching circuit in this way.The Amplifier will not give you "the best performance" all over the frequency band but a compromised result may satisfy..

 

[circuitking]

you will find a Impedance Region that more or less satisfies those Optimum Impedances

Do you mean, I can choose any one impedance arbitrarily from this impedance region which may be close to all those impedances.

By the way, my teacher mentioned that I can extract real and imaginary parts from the impedances and then find polynomial from them using curve fitting. With the help of polynomial, I can find one impedance with which I can do matching. I am not sure what this method is and how it is called. I am actually searching for this procedure.

 

[BigBoss]

Do you mean, I can choose any one impedance arbitrarily from this impedance region which may be close to all those impedances.

By the way, my teacher mentioned that I can extract real and imaginary parts from the impedances and then find polynomial from them using curve fitting. With the help of polynomial, I can find one impedance with which I can do matching. I am not sure what this method is and how it is called. I am actually searching for this procedure.

Optimum Load Impedances are generally take place on line or curve.You can try different matching circuit to obtain your better result.You can also multiple matching circuit in cascade that match the amplifier for each impedance point but it's really difficult.There is another method Real Frequency Technique it's an analytical method but implementation is pretty difficult.

 

[cariban]

Broadband impedance matching is quite different from narrow-band matching. There is a theoretic bound for returnloss if the matching network must be lossless. You can not obtain better result than the bound with finite number of lossless components.
Usually people solve the problem from different approach:
1. Use numerical optimization to search for "good" enough lossless matching network. In reality you may just need the returnloss to be better than a certain level (for example 10dB) in the whole band. You may be lucky to find a matching network just meeting your requirement;

2. Use RF switch to divide the whole band to several sub-band. For you case, you may divide the band to 1-4, 4-8. 8-12, 12-16, 16-20. It is much easier to design the impedance matching network with narrower bandwidth;

3. Use lossy impedance matching network. If you could accept a certain level of loss, then you can introduce resistors into the matching network. As a rule of thumb, 1 dB power loss leads to 2-3 dB returnloss improvement.

 

[circuitking]

2. Use RF switch to divide the whole band to several sub-band. For you case, you may divide the band to 1-4, 4-8. 8-12, 12-16, 16-20. It is much easier to design the impedance matching network with narrower bandwidth;

so If I visualize this, I would have 5 matching networks in cascade. That is going to be a big network.

 

[cariban]

It is true that the filter bank will increase the complexity significantly. Try your luck to design a wideband impedance matching network first.

 

[pancho_hideboo]

Consider Bode-Fano Criteria.
https://designers-guide.org/forum/YaBB.pl?num=1207009725/2#2

 

[circuitking]

I found negative image modeling presented in this paper "Efficient technique for ultra broadband, linear power amplifier design by ahmed sayed, sebastian preis and georg boeck". I think this is gonna fit for my purpose.

 

[circuitking]

Consider Bode-Fano Criteria.
https://designers-guide.org/forum/YaBB.pl?num=1207009725/2#2

As I understood Bode Fano criteria gives a maximum limit on bandwidth that can be achieved based on load impedance for different matching networks. In other words it gives a minimum gamma that can be achieved.

But in my case I have different loads at different frequencies, do I not need to take all of them into consideration?

By the way how did you get the networks between Term 5 and 6 and Term 7 and 8 in your post in that link? Did you use smith chart utility?

 

[pancho_hideboo]

By the way how did you get the networks between Term 5 and 6 and Term 7 and 8 in your post in that link?
Did you use smith chart utility?

Yes...........