Simultaneous Conjugate Impedance Match in AWR Microwave Office

[tec5c]

Note: Certain things below may not make sense if you are not familiar with the AWR software package. I apologise in advance and if so, I can do my best to explain anything that is unclear.

I am attempting to design a 5GHz LNA using a SiGe 0.25um process in AWR, I have the necessary process design kit (PDK) for said process though it does not contain the S parameters for the NMOS device that I am using.

I have a few questions about how one is to go about doing this. I have selected my bias conditions based on an IV curve for the device, then I created a BiasTee circuit (depicted below) which I then plotted the S parameters for. Is this the correct way of obtaining the data?



With the S parameter results I got from the graph, I then tried calculating the Rollet Stability Factor (K) which gave -0.9767 so either my data is incorrect or I am going to have myself a nice oscillator   

Any help is greatly appreciated.

[tipofthesowrd]

I am attempting to design a 5GHz LNA using a SiGe 0.25um process in AWR, I have the necessary process design kit (PDK) for said process though it does not contain the S parameters for the NMOS device that I am using.

First let me say I have no clue about designing this kind of thing in RF IC's. I do have some experience in RF/microwave board level work.

I have a few questions about how one is to go about doing this. I have selected my bias conditions based on an IV curve for the device, then I created a BiasTee circuit (depicted below) which I then plotted the S parameters for. Is this the correct way of obtaining the data?

Well, yes, that is exactly how I would design an LNA based on the physical device model (and not the S-parameters).
Advantage is often that you can choose your optimum bias point.
When you're working with pre-measured S-parameters of the manufacturer you have to be confident you know his measurement setup and bias conditions.
Depends on your application if you can stick with the pre-measured S-parameter model.

With the S parameter results I got from the graph, I then tried calculating the Rollet Stability Factor (K) which gave -0.9767 so either my data is incorrect or I am going to have myself a nice oscillator   

You didn't think you were going to have an unconditionally stable amplifier without any matching components?  O0
At least in board level design after selecting your bias point this is were the actual design fun of the matching/stability/trade off of gain vs noise / vs input match begins 

[tec5c]

You didn't think you were going to have an unconditionally stable amplifier without any matching components?  O0
At least in board level design after selecting your bias point this is were the actual design fun of the matching/stability/trade off of gain vs noise / vs input match begins 

The BiasTee element in the picture is an ideal L network, while the conjugate pair is not perfectly matched this is to be used as a starting point for an LNA design in AWR (according to their own video tutorials).

I am simply trying to see how stable the MOSFET is at 5GHz with the biasing conditions that I have chosen. Using the S parameters that I plotted with this circuit, I calculated the K factor. I then plotted the K factor in AWR and the results were not the same, so I am missing something.

Edit: I am completely aware that having a perfectly matched input and output will improve the overall stability. Though I am new to this so I am trying to follow and understand an example given in a textbook I have. 

Do not even get me started on the art of simultaneous conjugate pair (input and output) matching. I threw together a common source topogly and was playing around with values in order to try to get them simultaneously matched... It did not go well. 

[rfeecs]

Your setup doesn't have any obvious problem.  Try sweeping over a wide range, say 0.1GHz - 6GHz, and look at S11 and S22 on the Smith chart and S21 in dB.  See if they look reasonable.  You try varying the bias, like cutting the device off and at a normal bias point and see if it is behaving as you expect.

Looking at the K-factor doesn't tell you much.  As has been mentioned, a transistor without matching will typically not be stable over a wide frequency range.  By the way, lossless matching does not change K.  But I would not worry about K for now.

[tec5c]

Thanks for the reply.

I was sweeping over 0.1Ghz - 10GHz, everything seemed to be okay. I'll take your advice and not worry about K for now.

Any advice you can give for doing simultaneous matching? Using the tuning tool while observing Zin and Zout just seems like a wild goose chase. Sure, it's helpful to fine tune but I would like to think that there's a more insightful way to approach the situation prior to fine tuning.

Do you match one port so it's ~50% lossless then do the other port? Do you focus on the input or the output first? I have found the iMatch wizard in AWR, though it appears that it can only do one port at a time so by matching the input or output first, the other port is going to throw the matched port off when you begin to adjust it.

Is there any way to do simultaneous matching in AWR?

[tec5c]

Looking at the K-factor doesn't tell you much.  As has been mentioned, a transistor without matching will typically not be stable over a wide frequency range.  By the way, lossless matching does not change K.  But I would not worry about K for now.

"If K < 1 then simultaneous conjugate matching of the input and output is not possible." This is from Pozar's book - 'Microwave Engineering'.
After playing around with my design, the only element that effects K is the source degeneration inductor. Even with this, the highest K value I can achieve is 0.999

Is there a flaw in my design or is this all based on the MOSFET that is being used?

[rfeecs]

Thanks for the reply.

I was sweeping over 0.1Ghz - 10GHz, everything seemed to be okay. I'll take your advice and not worry about K for now.

Any advice you can give for doing simultaneous matching? Using the tuning tool while observing Zin and Zout just seems like a wild goose chase. Sure, it's helpful to fine tune but I would like to think that there's a more insightful way to approach the situation prior to fine tuning.

Do you match one port so it's ~50% lossless then do the other port? Do you focus on the input or the output first? I have found the iMatch wizard in AWR, though it appears that it can only do one port at a time so by matching the input or output first, the other port is going to throw the matched port off when you begin to adjust it.

Is there any way to do simultaneous matching in AWR?

If you are really trying to do an LNA, then you match the input for optimum noise figure.  Usually the best noise match is not a conjugate match.  So there often is a trade off between NF and S11.  Using lossless feedback, like source degeneration, can sometimes improve the trade off.

Your matching is going to depend on the bandwidth that you need.  For a narrow band application, you can usually do a two element match.  You should get to the point where you look at the Smith chart and know what topology you need to do the match.  You also have to think about how to incorporate the bias elements into the match.

You match the output for gain.  You will probably have to add loss for stability.  For an LNA you would add that to the output, not the input.

Microwave office has a very powerful optimizer.  Once you know your topology, you can set some goals for S11, S22, S21, etc. and push optimize and it will adjust the component values to try to meet the goals.  For a simple narrow band circuit you don't need it.  It's better to manually adjust or tune things and get a feel of what is going on instead of using automatic methods.

[rfeecs]

"If K < 1 then simultaneous conjugate matching of the input and output is not possible." This is from Pozar's book - 'Microwave Engineering'.
After playing around with my design, the only element that effects K is the source degeneration inductor. Even with this, the highest K value I can achieve is 0.999

Is there a flaw in my design or is this all based on the MOSFET that is being used?

If K<1 then you need lossy matching for a simultaneous match.  You will be chasing your tail if you try to use lossless matching.  Just adding a small resistor in series with the input or output often brings instant stability.  But it also kills gain, noise figure or power depending where you put it.  You would want to put loss on the output for an LNA so you don't kill your noise figure.