<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Measurement Setup | Jose Alejandro Galaviz-Aguilar</title><link>https://galaviz-rf.com/tag/measurement-setup/</link><atom:link href="https://galaviz-rf.com/tag/measurement-setup/index.xml" rel="self" type="application/rss+xml"/><description>Measurement Setup</description><generator>Wowchemy (https://wowchemy.com)</generator><language>en-us</language><lastBuildDate>Wed, 20 May 2026 00:00:00 +0000</lastBuildDate><image><url>https://galaviz-rf.com/media/icon_huc0a21cd13d3b330e570311c0697204cf_39767_512x512_fill_lanczos_center_3.png</url><title>Measurement Setup</title><link>https://galaviz-rf.com/tag/measurement-setup/</link></image><item><title>A Low-Cost FPGA/SDR Testbed for Power Amplifier Characterization</title><link>https://galaviz-rf.com/post/measurement-setup-lowcost/</link><pubDate>Wed, 20 May 2026 00:00:00 +0000</pubDate><guid>https://galaviz-rf.com/post/measurement-setup-lowcost/</guid><description>&lt;style>
.article-style p,
.article-style li,
.article-container p,
.article-container li {
text-align: justify;
}
&lt;/style>
&lt;h2 id="overview">Overview&lt;/h2>
&lt;p>This post summarizes the measurement testbed described in our paper &lt;em>&amp;ldquo;A Comparison of Surrogate Behavioral Models for Power Amplifier Linearization under High Sparse Data&amp;rdquo;&lt;/em> (&lt;em>Sensors&lt;/em>, 2022, 22(19), 7461). The goal of the setup is to capture real RF measurements from a power amplifier (PA) under test, so that behavioral models — and the digital predistortion (DPD) that linearizes the PA — can be trained and validated against hardware rather than simulation alone.&lt;/p>
&lt;p>Instead of relying on expensive dedicated RF instrumentation, the testbed is built around a &lt;strong>software-defined radio (SDR)&lt;/strong> running on an FPGA system-on-chip. This keeps the platform affordable and reproducible while still supporting wideband signals and the acquisition of crest factor reduction (CFR), PAPR, and ACPR measurements at different instantaneous bandwidths.&lt;/p>
&lt;h2 id="testbed-architecture">Testbed Architecture&lt;/h2>
&lt;p>The testbed is organized in two stages:&lt;/p>
&lt;ul>
&lt;li>&lt;strong>Digital stage (FPGA/SDR).&lt;/strong> A Cyclone V FPGA SoC kit hosts the SDR. An embedded Linux running on a Microblaze/ARM (Zynq) architecture uses the Analog Devices &lt;code>libiio&lt;/code> library to provide a full-duplex link between the host PC (MATLAB) and the FPGA. Baseband waveforms are streamed from MATLAB, stored in the FPGA&amp;rsquo;s DDR memory, and exchanged over a standard &lt;strong>Ethernet&lt;/strong> connection. An AXI interconnect provides the low-latency, high-throughput data path on-chip.&lt;/li>
&lt;li>&lt;strong>Analog/RF stage.&lt;/strong> An &lt;strong>AD9361 RF transceiver&lt;/strong> up-converts the baseband signal with a quadrature modulator driven by an internal local oscillator set to &lt;strong>2.45 GHz&lt;/strong>, generating the stimulus that drives the PA. A feedback path routes the PA output through a power splitter to a spectrum analyzer for frequency-domain measurements.&lt;/li>
&lt;/ul>
&lt;p>The transceiver uses direct conversion with 2×2 I/Q channels, a 320 MSPS DAC and a 640 MSPS ADC, and a 128-tap interpolation FIR filter with an adjustable internal sampling rate.&lt;/p>
&lt;h2 id="device-under-test-and-instrumentation">Device Under Test and Instrumentation&lt;/h2>
&lt;ul>
&lt;li>&lt;strong>Power amplifier (DUT):&lt;/strong> Mini-Circuits &lt;strong>ZX60-V63+&lt;/strong>, a 50 Ω coaxial gain block biased at 5 V, with roughly 20 dB of gain at 2.45 GHz (measured at 18.79 dB for the LTE-10 MHz signal and 18.45 dB for LTE-15 MHz).&lt;/li>
&lt;li>&lt;strong>SDR / transceiver:&lt;/strong> Analog Devices AD9361 Agile Transceiver (ARRadio / AD-FMCOMMS3-EBZ card).&lt;/li>
&lt;li>&lt;strong>FPGA:&lt;/strong> Altera Cyclone V SoC kit.&lt;/li>
&lt;li>&lt;strong>Power supply:&lt;/strong> GW INSTEK GPS-3303.&lt;/li>
&lt;li>&lt;strong>Spectrum analyzer:&lt;/strong> GW INSTEK GSP-830.&lt;/li>
&lt;li>&lt;strong>Host:&lt;/strong> PC running MATLAB.&lt;/li>
&lt;/ul>
&lt;p>
&lt;figure >
&lt;div class="d-flex justify-content-center">
&lt;div class="w-100" >&lt;img alt="Low-cost FPGA/SDR measurement testbed" srcset="
/post/measurement-setup-lowcost/testbed_hu93991c26fccddd8b1ebbba943eb04960_1205391_96a728eb4f53d28dd89358173297022c.webp 400w,
/post/measurement-setup-lowcost/testbed_hu93991c26fccddd8b1ebbba943eb04960_1205391_66455411e144ab3b5dd446b5202ef61d.webp 760w,
/post/measurement-setup-lowcost/testbed_hu93991c26fccddd8b1ebbba943eb04960_1205391_1200x1200_fit_q75_h2_lanczos_3.webp 1200w"
src="https://galaviz-rf.com/post/measurement-setup-lowcost/testbed_hu93991c26fccddd8b1ebbba943eb04960_1205391_96a728eb4f53d28dd89358173297022c.webp"
width="760"
height="428"
loading="lazy" data-zoomable />&lt;/div>
&lt;/div>&lt;/figure>
&lt;/p>
&lt;h2 id="measurement-workflow">Measurement Workflow&lt;/h2>
&lt;p>Test signals are single-carrier &lt;strong>64-QAM multiplexed over OFDM&lt;/strong>, following LTE at &lt;strong>10 MHz and 15 MHz&lt;/strong> bandwidths. Before each run, the system is calibrated so that a properly synchronized signal reaches the PA input, with the calibration power set for both peak and average levels at the PA output.&lt;/p>
&lt;p>Captured, real-valued I/Q data are post-processed in MATLAB using the LTE Toolbox and RF Blockset models to compute the key figures of merit: &lt;strong>NMSE, PAPR, ACPR, and CCDF&lt;/strong>. These measurements feed the training and testing of the surrogate behavioral models compared in the paper.&lt;/p>
&lt;h2 id="why-a-low-cost-setup-matters">Why a Low-Cost Setup Matters&lt;/h2>
&lt;p>By combining an FPGA-based SDR, an inexpensive commercial PA, and budget-tier bench instruments, this testbed delivers a complete, reproducible PA characterization and DPD-validation flow at a fraction of the cost of a traditional instrument rack. It lowers the barrier for research and teaching in RF linearization, where access to high-end vector signal generators and analyzers is often the limiting factor.&lt;/p></description></item></channel></rss>