How to find optimal load of the amplifier?

[sze wen]

I am doing it at 3 GHz.
I am trying to find the Zout of the RF amplifier. Not the Zout=50 ohm into the matching network. How should it be done? Is it by varying 50 ohm from 10 ohm to 100 ohm and look at the gain and output power?

 

[BigBoss]

Z_OUT of the amplifier?? What type of amplifier ?? Power Amplifier (my guess) or Linear Small Signal Amplifier??
So, you desire a Power Matching or Conjugate Matching ??

 

[sze wen]

Z_OUT of the amplifier?? What type of amplifier ?? Power Amplifier (my guess) or Linear Small Signal Amplifier??
So, you desire a Power Matching or Conjugate Matching ??

You are right that it is power amplifier. Power matching to the load.

--- Updated Apr 21, 2025 ---

I am using Cadence software. How to find the optimal load?

--- Updated Apr 21, 2025 ---

I wanted to select the right power of the design first first before I did the conjugate matching. Is it right to do so?

 

[BigBoss]

You are right that it is power amplifier. Power matching to the load.

--- Updated Apr 21, 2025 ---

I am using Cadence software. How to find the optimal load?

--- Updated Apr 21, 2025 ---

I wanted to select the right power of the design first first before I did the conjugate matching. Is it right to do so?

No, you have to do Load Pull simulation to find Optimum Load Impedance for max. Delivered Power.
See Cadence Spectre application papers/guides for more information. There is a Load Pull Tuner Component (or similar), and you can find optimum values by using it.

 

[sze wen]

No, you have to do Load Pull simulation to find Optimum Load Impedance for max. Delivered Power.
See Cadence Spectre application papers/guides for more information. There is a Load Pull Tuner Component (or similar), and you can find optimum values by using it.

After finding the maximum from the Smith Chart. What's next?

--- Updated Apr 22, 2025 ---

No, you have to do Load Pull simulation to find Optimum Load Impedance for max. Delivered Power.
See Cadence Spectre application papers/guides for more information. There is a Load Pull Tuner Component (or similar), and you can find optimum values by using it.

Should I set certain input power for it like -10dBm or I always assume that the input is 20dBm? My input is from -10dBm to 20dBm

 

[BigBoss]

After finding the maximum from the Smith Chart. What's next?

--- Updated Apr 22, 2025 ---

Should I set certain input power for it like -10dBm or I always assume that the input is 20dBm ? My input is from -10dBm to 20dBm

Next is to design a appropriate Matching Circuit.
Optimum Load Impedance is variable against Delivered Power. Take +20dBm Input Power as a reference then check your PA for other Input Powers.

 

[sze wen]

Next is to design a appropriate Matching Circuit.
Optimum Load Impedance is variable against Delivered Power. Take +20dBm Input Power as a reference then check your PA for other Input Powers.

Noted. I just done a load pull. I got gamma=0.5 and theta = 190 at the PORT1 50 ohm load termination for maximum PAE.
Or should I put the probe at the Port terminal instead of PORT1 50 ohm load terminal at cadence
Since my design is to support 3 GHz. How should I proceed from here?

 

[mtwieg]

View attachment 199140

I am doing it at 3 GHz.
I am trying to find the Zout of the RF amplifier. Not the Zout=50 ohm into the matching network. How should it be done? Is it by varying 50 ohm from 10 ohm to 100 ohm and look at the gain and output power?

This looks backwards. Usually the matching network is designed such that the load (50 ohms in your case) is transformed into Zopt at the amplifier output, where Zopt is determined by load pull simulation. The Zout in your diagram will not be 50 ohm (unless you add an isolator on the output).

 

[sze wen]

This looks backwards. Usually the matching network is designed such that the load (50 ohms in your case) is transformed into Zopt at the amplifier output, where Zopt is determined by load pull simulation. The Zout in your diagram will not be 50 ohm (unless you add an isolator on the output).

Got it. Does that mean we could choose Efficiency and Output power after output impedance is matched?

--- Updated Apr 22, 2025 ---

This looks backwards. Usually the matching network is designed such that the load (50 ohms in your case) is transformed into Zopt at the amplifier output, where Zopt is determined by load pull simulation. The Zout in your diagram will not be 50 ohm (unless you add an isolator on the output).

I haven't added a DC block yet. When and how do we determine the value DC block needed at specific frequency?

 

[mtwieg]

Got it. Does that mean we could choose Efficiency and Output power after output impedance is matched?

First you do load pull to determine the optimal load impedance Zopt. The details of this procedure depend on the simulation tool you use.

After Zopt is chosen, then you design a matching network to transform the actual load impedance (50 ohms, for example) into Zopt.

 

[sze wen]

First you do load pull to determine the optimal load impedance Zopt. The details of this procedure depend on the simulation tool you use.

After Zopt is chosen, then you design a matching network to transform the actual load impedance (50 ohms, for example) into Zopt.

after that? If i noticed the S22 is very low at 2 GHz instead of 3 GHz.

--- Updated Apr 22, 2025 ---

First you do load pull to determine the optimal load impedance Zopt. The details of this procedure depend on the simulation tool you use.

After Zopt is chosen, then you design a matching network to transform the actual load impedance (50 ohms, for example) into Zopt.

Let's say the Zopt is 20ohm + 10j ohm.

--- Updated Apr 23, 2025 ---

First you do load pull to determine the optimal load impedance Zopt. The details of this procedure depend on the simulation tool you use.

After Zopt is chosen, then you design a matching network to transform the actual load impedance (50 ohms, for example) into Zopt.

How do I design the transformation network using Pi or T etc?

 

[BradtheRad]

The aim of impedance matching is to convert what you've got, to a combination of V & A which is optimally usable to the next stage of your circuitry.

If impedance matching is done properly then power (Watt= VxA) conveyed is kept to a maximum.

Generally speaking you'll use one or more LC stages. Their configuration are determined by the purposes of your filter.
Small capacitor Farad values are associated with small Ampere levels and/or higher frequencies.
Small inductor Henry values are associated with larger Ampere levels and/or higher frequencies.

 

[sze wen]

The aim of impedance matching is to convert what you've got, to a combination of V & A which is optimally usable to the next stage of your circuitry.

If impedance matching is done properly then power (Watt= VxA) conveyed is kept to a maximum.

Generally speaking you'll use one or more LC stages. Their configuration are determined by the purposes of your filter.
Small capacitor Farad values are associated with small Ampere levels and/or higher frequencies.
Small inductor Henry values are associated with larger Ampere levels and/or higher frequencies.

Given at high frequency, the capacitor values and inductor values are small. How do we determine the equations for the PI network etc?
When should we include the DC blocking capacitance so that it operates at frequency of interest.

 

[BradtheRad]

Given at high frequency, the capacitor values and inductor values are small. How do we determine the equations for the PI network etc?
When should we include the DC blocking capacitance so that it operates at frequency of interest.

To pass AC while blocking DC...
The RC time constant comes into play here. It helps if you know what impedance is your incoming signal. In other words its combination of voltage and Amperes. As a rule you'll start with a series capacitor of a large-ish Farad value, then reduce the value while watching how it affects your incoming signal. When it's at a small enough value it charges and discharges just rapidly enough when exposed to 3 GHz, so that your AC portion is unattenuated...
Yet lower frequencies (including DC) get blocked, or filtered out.

I lack experience working with your frequency range. Someone knowledgeable would ask what signal strength emerges from your initial circuit, so he can gauge the incoming impedance.

 

[sze wen]

To pass AC while blocking DC...
The RC time constant comes into play here. It helps if you know what impedance is your incoming signal. In other words its combination of voltage and Amperes. As a rule you'll start with a series capacitor of a large-ish Farad value, then reduce the value while watching how it affects your incoming signal. When it's at a small enough value it charges and discharges just rapidly enough when exposed to 3 GHz, so that your AC portion is unattenuated...
Yet lower frequencies (including DC) get blocked, or filtered out.

I lack experience working with your frequency range. Someone knowledgeable would ask what signal strength emerges from your initial circuit, so he can gauge the incoming impedance.

I want to design it from 2 to 4 GHz broadband.

 

[FvM]

I want to design it from 2 to 4 GHz broadband.

Good you finally mentioned after 6 posts! Expect that it's not feasible to realize exact matching for the frequency range, only an approximation.

 

[sze wen]

Good you finally mentioned after 6 posts! Expect that it's not feasible to realize exact matching for the frequency range, only an approximation.

How to do the broadband matching? Is there any standard way of doing so?

 

[mtwieg]

If i noticed the S22 is very low at 2 GHz instead of 3 GHz.

S22 usually isn't relevant.

Let's say the Zopt is 20ohm + 10j ohm.

Ok.

How do I design the transformation network using Pi or T etc?

There are lots of straightforward tools for designing LC matching networks, I use this one sometimes: https://home.sandiego.edu/~ekim/e194rfs01/jwmatcher/matcher2.html

Of course that only gives the desired matching at a specific target frequency.

Also the load impedance at harmonic frequencies is also often important.

How to do the broadband matching? Is there any standard way of doing so?

Broadband matching is a whole topic in itself....

In a sense, it's adding another dimension to the optimization problem, which means all your other figures of merit (PAE, P1dB, Gain, IM3, Popt, etc) are going to be functions of frequency. You will need to choose a matching network design which gives a good enough compromise of these parameters over your operating bandwidth. There's no "standard" approach AFAIK.

 

[sze wen]

S22 usually isn't relevant.

Ok.

There are lots of straightforward tools for designing LC matching networks, I use this one sometimes: https://home.sandiego.edu/~ekim/e194rfs01/jwmatcher/matcher2.html

Of course that only gives the desired matching at a specific target frequency.

Also the load impedance at harmonic frequencies is also often important.

Broadband matching is a whole topic in itself....

In a sense, it's adding another dimension to the optimization problem, which means all your other figures of merit (PAE, P1dB, Gain, IM3, Popt, etc) are going to be functions of frequency. You will need to choose a matching network design which gives a good enough compromise of these parameters over your operating bandwidth. There's no "standard" approach AFAIK.

How to match it to broadband bro?

 

[BigBoss]

How to do the broadband matching? Is there any standard way of doing so?

You dream too much. you cannot obtain a good set of PAE, Power Gain and Maximum Delivered Power simultaneously. No way ..
The found values by Load Pull technique are only and only for single frequency, not 2 octaves broadband.
If you have already found a Z_OPT for a certain purpose, S₂₂ can be anything. It doesn't matter.