Here’s the page we think you wanted. See search results instead:

 

  

Main Support:       Knowledge Center  > ADS Support Home   >   ADS Documentation (all releases)
Documentation:   ADS 2009 Update 1   >  ADS Loadpull DesignGuide   >  Loadpull DesignGuide

This document contains references to Agilent Technologies. Agilent's former Test and Measurement business has become Keysight Technologies. For more information, go to www.keysight.com.


Skip to end of metadata
Go to start of metadata

Loadpull DesignGuide

Loadpull simulation is frequently used by power amplifier designers to determine which load impedance to present to a device or amplifier in order to achieve a particular power delivered, power-added efficiency, intermodulation distortion level, adjacent-channel leakage ratio, and other specifications.

The Loadpull DesignGuide has a utility to import measured Loadpull data into ADS and several setups to run the Loadpull simulations. Click here for more details on Load Pull Measurement Data Import Utility.

From Analog/RF Schematic window, Select DesignGuide > Loadpull where you can see different options in Loadpull window.

The following two loadpull simulation setups included here are:

  1. Fixed available source power- This setup shows how power delivered and other specifications vary with the load reflection coefficient.
  2. Constant power delivered - This setup shows contours of various specifications, all with a constant power delivered to the load.

The constant power delivered simulations are achieved using an optimization, and include one-tone, two-tone, and WCDMA input signals.

One-Tone Simulation

Select One Tone, Constant Power Delivered Loadpull Simulation to copy HB1Tone_LoadPull_ConstPdel schematic and corresponding data display into your project. This setup sweeps the load reflection coefficient in a circular region of the Smith Chart and optimizes the source power level for each load reflection coefficient until the desired power is delivered to the load. The data display shows contours of constant PAE, bias current, gain, and gain compression. The input reflection coefficient is also shown for a particular load that you specify. This allows you to pick the optimal load that produces the best PAE, gain, gain compression, or bias current, or make trade-offs amongst these specifications.

The SmallSignal and Sweep3 parameter (lower left corner of the above image) sweeps are only used to obtain output powers when the device is being driven with a small signal. These output powers are used as references in the gain compression computations.

You need to specify multiple things while using this schematic, such as:

  1. Replace the device with your device or amplifier.
  2. Set the bias voltages or modify the bias network, as needed. However, the data display calculates the DC power consumption assuming current probe Is_low is connected to supply voltage node Vs_low and current probe Is_high is connected to supply voltage node Vs_high. If you delete any of these or re-name them, you must update the Pdc equation on the corresponding data display so the DC power consumption is computed correctly.
  3. Specify the center and radius of the circle of the reflection coefficients.

    These are set by the s11_center and s11_rho variables (s11_rho is automatically reduced in order to keep the reflection coefficient < 1.) If the device or amplifier is potentially unstable and the circle of reflection coefficients that you specify includes the unstable region, the simulation may run into convergence problems. This would be due to the device wanting to oscillate. A solution to this problem is to add stabilizing components at the input, output, or in parallel with the device. You may want to use a simulation setup for this purpose, DesignGuide > Amplifier > S-Parameter Simulations > Feedback Network Optimization to Attain Stability. Another solution is to specify the circle of reflection coefficients such that the unstable region is avoided.
    The Reflection Coefficient Utility is a data display file that displays circle of reflection coefficients that will be simulated when you specify a particular s11_center and s11_rho variable.
  4. Specify the number of reflection coefficient points to be simulated, pts, and the reference impedance, Z0.
  5. Specify the nominal and allowed range of the available source power, Pavs.
    Figure: Prior to starting on-screen editing

    Figure: While performing on-screen editing

    During the optimization, Pavs is adjusted within the limits until the power that you want is delivered to the load. The nominal value of Pavs does not matter that much, since it is used as the initial value for the first optimization. Depending on how high a power you want delivered to the load and the gain of the device, you may have to adjust the maximum allowed value of Pavs.
  6. On the SmallSignal Parameter Sweep, you should make sure the value is small enough that the device behaves linearly.

The power delivered to the load with this "small signal" available source power is kept as the reference for the gain compression computation. You specify the desired power to be delivered to the load in OptimGoal1.

In this example, the power delivered is between 25 and 25.1 dBm. With this value and a maximum Pavs value of 15 dBm, it is specifyied that the lowest transducer power gain accepted is 10 dB.
You may also specify different load and source impedances at the harmonic frequencies and (for the source) at the fundamental frequency.

To launch the simulation, click the optimization icon (if using ADS 2009 Update 1 or later.) If simulation is started by hitting the F7 key or by selecting Simulate > Simulate, then optimization is not executed and simulation results are not displayed in data display.

After running the optimization, HB1Tone_LoadPull_ConstPdel data display shows the results.

To see the contours effectively, you may need to change the CurrentStep, PAE_step, Gain_step, and GainCompStep variables. These set the step sizes between the contours. As stated above, if you have modified the bias network, you will have to edit the Pdc equation on the Equations page.

Also, the bias supply current calculations only include the current in the probe Is_high. If you change the name of the current probe, you must edit the BiasCurrent equation on the Equations page.

The upper Smith Chart shows contours of constant gain and gain compression. The lower left Smith Chart shows contours of constant bias current and power-added efficiency (PAE), as well as the simulated load reflection coefficients and the corresponding input reflection coefficients. The lower right Smith Chart shows the same data on a Smith Chart with a different reference impedance.

In the red boxes on the left side are data that correspond to a particular optimal condition such as minimum bias current, maximum PAE, or minimum gain compression. However, ensure that the desired power delivered was actually achieved. For example, at the load or reflection coefficient that gave the minimum bias current, the power delivered was < 23 dBm. This minimum bias current load is very close to 0 Ohms, and it is very difficult to deliver any power to it. The gain (transducer power gain) and gain compression with this load are also unreasonable.

The load that corresponds to the maximum PAE is very close to the one that corresponds to the minimum bias current, but the power delivered meets the 25 dBm requirement and the gain and gain compression values are much more reasonable.

You also have the option of selecting any of the simulated load reflection coefficients with marker m1. The following data appears in a separate box:

This enables you to see potential trade-offs. For example, for this load impedance, the PAE is worse, but the amount of gain compression is much less. Also the input reflection coefficient now has a positive real part, which aids stability.

Two-Tone Simulation

Select Two Tone, Constant Power Delivered Loadpull Simulation to copy HB2Tone_LoadPull_ConstPdel schematic and corresponding data display into your project. This simulation setup and data display is identical to the "One Tone", except that now two tones are supplied instead of one. A two tone test signal stresses the device more because of its much higher peak-to-average ratio. The data display from this simulation shows the same information as above and also includes intermodulation distortion.

A different device is used here.

Here also you have to replace the sample device with yours and adjust the biasing as needed. You need to specify the following two variables:

  1. Frequency spacing between the two tones fspacing
  2. Maximum order of intermodulation distortion tones to be included in the simulation Max_IMD_Order.
    The simulation results include similar information as shown above, with the addition of intermodulation distortion.

There is a clear trade-off between PAE and distortion. For this bias point, if you want maximum PAE, you suffer a lot of gain compression and intermodulation distortion.

Tolerating a lower PAE allows much lower gain compression and intermodulation distortion.

WCDMA Signal Simulation

Select WCDMA Loadpull > Constant Power Delivered, Mag/Phase Loadpull to copy WCDMA_LoadPullMagPh_ConstPdel schematic and corresponding data display into your project.

This setup sweeps the load reflection coefficient in a fan-shaped region of the Smith Chart and optimizes the source power level for each load reflection coefficient until the desired power is delivered to the load. The source is a WCDMA signal read in from a dataset, and its amplitude (and thus the available source power) is set by a variable, SFexp, and the gain applied to this signal is 10**(SFexp). The WCDMA signal was generated by connecting a Timed Sink to the output of the signal source in the ADS example examples/WCDMA3G/WCDMA3G_PA_Test_prj/WCDMA3G_PA_UE_ACLR schematic.

The loadpull is performed twice.

  1. In first case, SFexp is set to 0.01. This is assumed to make the input signal small enough that the amplifier is operating linearly. The gain under this condition for each load reflection coefficient is the reference used to compute the gain compression.
  2. In second case, SFexp is optimized until the desired power is delivered to the load.

The data display shows contours of constant PAE, ACLR, bias current, gain, and gain compression. The input reflection coefficient is also shown for a particular load that you specify. This allows you to pick the optimal load that produces the best PAE, ACLR, gain compression, or bias current, or make trade-offs among these specifications.

When using this schematic, there are a number of different things you need to specify, and these are listed in a paragraph on the schematic. First, you would replace the device with your device or amplifier. You have to set the bias voltages or modify the bias network, as needed. However, the data display calculates the DC power consumption assuming current probe Is_low is connected to supply voltage node Vs_low and current probe Is_high is connected to supply voltage node Vs_high. If you delete any of these or re-name them, you will have to modify the equations like Is_highDC=mean(Is_high.i[0]) and the Pdc equation on the schematic.

You have to specify the range of phases and magnitudes of the reflection coefficients. The total simulation time will increase linearly with the product of the numbers of phases and magnitudes simulated. The tradeoff is that you should get better contour lines with more points simulated.

If the device or amplifier is potentially unstable and the region of reflection coefficients that you specify includes the unstable region, the simulation may run into convergence problems. This would be due to the device wanting to oscillate. A solution to this problem is to add stabilizing components at the input, output, or in parallel with the device. You may want to use a simulation setup for this purpose, DesignGuide > Amplifier > S-Parameter Simulations > Feedback Network Optimization to Attain Stability. Another solution is to specify the range of reflection coefficients such that the unstable region is avoided.

You also have to specify the reference impedance, Z0, and the source center frequency, RFfreq.

You have to specify the nominal and allowed range of the signal source gain scale factor exponent, SFexp. It is necessary to adjust this gain to set the available source power, because we are using a voltage source to generate the signal.

It is not obvious what the relationship is between this exponent and the available source power. However, the WCDMA_SrcTest schematic in the same project allows you to sweep this scale factor and calculate the corresponding available source power. In this case, stepping SFexp from -1 to 1 increases the available source power from about -20 dBm to +20 dBm. This does vary with the source impedance you specify.

When setting the range of SFexp values, you want the highest SFexp value to correspond to the maximum available source power you will accept. For example, if you want to deliver 27 dBm to the load and you want the device to provide at least 7 dB of transducer power gain, you would set the maximum value of SFexp to correspond to an available source power of approximately 20 dBm. The optimization will run fastest if SFexp is allowed to vary over a relatively large range, but with the nominal value close to the value needed to give you the desired output power. You may get extremely high gain compression values for some reflection coefficients near the edge of the Smith Chart.

Figure: Prior to starting on-screen editing

Figure: While performing on-screen editing

During the optimization, this variable is adjusted within the limits until the power that you want is delivered to the load.

You specify the desired power to be delivered to the load in OptimGoal1.

In this case, we want the power delivered to be within 0.25 dB of 27 dBm. You may also specify different load and source impedances at the harmonic frequencies and (for the source) at the fundamental frequency.

There is a bank of raised cosine filters connected to the output node. These are used to compute the upper and lower adjacent and second adjacent channel leakage ratios. If you modify this setup to simulate a signal corresponding to a different standard (non-WCDMA, for example) then you most likely will need to modify the filter parameters and the channel frequency limits, which are specified as an offset from the carrier center frequency, RFfreq.

As mentioned above the simulation time is directly proportional to the total number of different load reflection coefficients. It also depends directly on the number of symbols simulated at each load. When initially exploring the Smith Chart to find an approximate optimal load, it might be useful to run the simulation with a relatively small number of symbols. Later, after determining a smaller, optimal region of the Smith Chart, you might want to increase the number of symbols to get more accurate results.

If using a data file as the source, setting the simulation time step tstep equal to the time step in the data file is good, to minimize effects that may arise due to interpolation.

The Envelope analysis includes a SweepOffset. This simulates but does not keep the first 12 symbols in the simulation during which the input signal amplitude is ramping on. If you want to include this turn-on ramp data in the post-processing computations, just set SweepOffset=0.

Because the simulation includes an optimization, you launch the simulation by clicking on the optimization icon. If instead you just launch the simulation by hitting the F7 key or selecting Simulate > Simulate, an optimization will not be run and the data display will not display the simulation results.

After running the optimization, this WCDMA_LoadPullMagPh_ConstPdel data display shows the results.

To see the contours effectively, you may need to change the CurrentStep, PAE_step, Gain_step, GainCompStep, and ACLR_step variables. These set the step sizes between the contours. The bias supply current calculations only include the current in the probe Is_high. If you change the name of the current probe, you will need to edit the BiasCurrent equation on the Equations page.

The upper Smith Chart shows contours of constant gain and gain compression. The lower left Smith Chart shows contours of constant bias current, power-added efficiency (PAE), and lower adjacent channel ACLR, as well as the simulated load reflection coefficients and the corresponding input reflection coefficients. The lower right Smith Chart shows the same data on a Smith Chart with a different reference impedance.

In the red boxes on the left side are data that correspond to a particular optimal condition such as minimum bias current, maximum PAE, minimum gain compression, or minimum ACLR. However, you have to make sure that the desired power delivered was actually achieved. In some cases with load reflection coefficients very close to the edge of the Smith Chart, the desired power will not be achieved.

The results show there is a trade-off between power-added efficiency and distortion. You can get slightly better PAE if you are willing to tolerate higher ACLR levels.

Note that the power delivered specification is not satisfied. You could re-run the simulation, allowing SFexp to vary over a larger range, or you could move marker m1 to a load near this one and see if the power delivered specification is satisfied. Moving marker m1 allows you to select any of the simulated load reflection coefficients. The corresponding data appears in a separate box, which allows you to see potential tradeoffs as you move around the Smith Chart.

The gain compression is still excessive, however.

The ACLR is somewhat sensitive to the load impedance.

The WCDMA Loadpull > Constant Power Delivered, Circular Region Loadpull menu pick copies into your project a schematic and data display nearly identical to the WCDMA_LoadPullMagPh_ConstPdel ones above, except that it sweeps a circular region of the Smith Chart instead of a fan-shaped region.

Time Taken by Simulation

For 25 different load reflection coefficients, numSymbols=128, and the optimization type set to Gradient with MaxIters=5, this simulation required about 4 minutes. For 49 different load reflection coefficients, about 6.75 minutes were required.

Please help us improve
Please help us improve
Was this topic helpful? Yes No