Evaluation of amplifier parameters
|
Page in progress. |
Evaluation of amplifier parameters 
- is an Engee application designed to analyse power amplifier (PA) characteristics from measured data and to build amplifier models for usage in the Engee simulation environment. It supports signal analysis, adjustment of model parameters, visualisation of results and export of coefficients for later usage.
To open the application, go to the Engee workspace and in the upper left corner under Engee applications 
select Evaluation of amplifier parameters :
Purpose and context of usage
Nonlinear distortion occurs in signal transmission through power amplifiers (PAs), especially in near-saturation modes. This leads to two key effects:
-
Extension of the signal spectrum beyond the bandwidth - violating the spectral mask requirement (ACLR);
-
In-band distortion and degradation of transmission quality - increased power in neighbouring channels (ACPR).
To improve the power efficiency of transmitters, amplifiers are often operated near saturation where nonlinear effects occur. The problem is solved by using Digital Pre-Distortion (DPD). The purpose of this method is to modify (distort) the input signal in advance so that, after passing through a real nonlinear amplifier, the output signal is as close as possible to the ideal linear response. This allows:
-
Increase the efficiency of the amplifier;
-
Reduce nonlinear distortion;
-
Reduce inter-channel interference;
-
Ensure compliance with transmission standards.
The application supports several algorithms for adaptive adjustment of the amplifier model:
-
RLS - recursive least squares method;
-
LMS - classical least squares method;
-
NLMS - normalised LMS;
-
RPEM - recursive error prediction method;
-
Regularised RLS - modification of RLS with robustness to noise.
These methods allow to dynamically adjust the parameters of the model based on input and output signals.
Three model architectures are also supported:
| Architecture | Description |
|---|---|
P (Polynomial). |
Memory-less polynomial model - suitable for simple cases without pronounced hysteresis. |
MP (Memory Polynomial). |
Adds a "memory" effect to the amplifier, improving the accuracy of time-dependent modelling. |
GMP (Generalised Memory Polynomial) |
Extended architecture with additional cross terms - used for complex distortions. |
The choice of algorithm and architecture combination affects the accuracy of the model. The evaluation criterion is usually the NMSE (Normalised Mean Square Error) metric.
Thus, the application Evaluation of amplifier parameters allows modelling and adaptation of digital preamplification systems using both real and synthetic data.
Terms and abbreviations
-
DPD (Digital Pre-Distortion) - Digital Pre-Distortion, a method of compensating for nonlinear distortion in an amplifier.
-
NMSE (Normalised Mean Squared Error) - A measure of model accuracy, measured in decibels.
-
PAPR (Peak-to-Average Power Ratio) - The ratio of a signal’s peak power value to its average value.
-
ACLR (Adjacent Channel Leakage Ratio) - radiation level in the neighbouring frequency channel.
-
ACPR (Adjacent Channel Power Ratio) - ratio of power in the neighbouring channel to the main channel.
-
AM-AM - dependence of output amplitude on input amplitude (amplitude-amplitude characteristic).
-
AM-PM - dependence of output phase on input amplitude (amplitude-phase characteristic).
Application interface
The application interface can be divided into five sections:
-
Yellow section - used to upload files with amplifier samples, download .txt file with KM coefficients;
-
Three blue sections - used to configure the amplifier model;
-
Red section - start the model calculation, control the sampling rate of the spectrum signal for plotting and the plot menu itself.
Next, let’s see how to work with the application using the interface.
Working with the application
Loading input data
To get started, load two files with amplifier signal samples by clicking on the 
button in the yellow section:
-
IQ_in_norm- samples of the signal at the input of the amplifier; -
IQ_out_norm- samples of the signal at the output of the amplifier.
|
The samples files must be named |
The files should be loaded simultaneously by selecting them via Ctrl+ЛКМ in the opened window:
Wait until the files are completely downloaded, the download indicators should change to ✓:
In the process of downloading |
Done (✓) |
|---|---|
|
|
Setting up the amplifier model
Next, set up the model in the blue sections step by step:
-
Select Algorithm and Architecture (section #1) - specify the model type and DPD method (e.g. NLMS). Depending on the algorithm selected, you will need to specify the convergence step or confidence and forgetting factor. For Moore-Penrose algorithm no additional parameters are set.
-
Model parameters (section №2) - for models are set:
-
Depth of memory;
-
Degree of the polynomial;
-
Order of the polynomial.
-
-
Sampling range (section №3) - specifies the range of samples used for model calculation (by default: from
200000to400000).
Start model calculation
After setting up the amplifier model, start its calculation. To do this, click the Start Calculation button. The indicator on the button will show the progress.
-> 
->
After completion, the plotting section will become available.
Plotting graphs
The following types of graphs are available:
-
Signals: input, output, modelled;
-
Signal spectra;
-
AM-AM and AM-PM characteristics;
-
Cumulative distribution function (CCDF).
The sampling frequency can be set manually:
For more convenient work with graphs, the classic Plotly library functionality will help:
-
- download a graph as a PNG. -
- the ability to select an area and zoom in on its contents. -
- tool to move the chart on the coordinate plane. -
- zooms in on the coordinate plane. -
- zooms out the coordinate plane. -
- returns the coordinate plane scale by default. -
- reset coordinate axes.
Examples of graphs
Signal spectrum U input, U output, U mod:
AM-AM amplitude response:
AM-PM amplitude-phase characteristic:
Cumulative probability distribution function CCDF:
To evaluate the performance of the amplifier model, the following are automatically calculated: normalised mean square error NMSE in dB, PAPR peak factor for input, output and simulated signal, e.g.:
Loading coefficients
To save the model coefficients, click the Download file button. The file will be saved in .txt format.
