Components and Systems for Communications Research Group

Competition details

Access details

The remoteUPCLab is a remote PA measurement system that everyone is welcome to use! No registration is required. We just ask you not to overload the system.

In order to generate the test signal, upload/download the waveforms data files, and provide a score for each experiment, the following scripts and detailed instructions are provided to the user:

UPC_WebLAB_IMS2025 (User DPD).zip                                                  (MATLAB scripts)

Instructions_DPD_SDC_IMS2025.pdf                                                               (instructions)

NOTE: A Matlab version  R2021b or higher, with SFTP connectivity functions, is required.

The previous compressed file contains a simple example on how to call the basic functions of the remoteUPCLab. The user has only to execute the "Main_DPD_SDC_IMS2025.m" script to check that the MATLAB functions are running OK (control messages are displayed during the waveform Tx/Rx process) and obtain an initial negative score before applying predistortion.

I. Introduction

The goal of the IMS2025 Power Amplifier Linearization through Digital Predistortion (DPD) Student Design Competition (SDC9) is to develop an effective DPD algorithm for linearizing a Load Modulated Balanced Amplifier (LMBA). The LMBA setup requires two RF input signals: one for the balanced port and another for the control port.

Students interested in participating can register here.

Participants will gain web-based access to the remoteUPCLab, which allows them to:

  • Upload their predistorted baseband I/Q signals (for both balanced and control paths) to the Vector Signal Generator (VSG).
  • Retrieve the amplifier’s output and I/Q signal response captured via the Signal and Spectrum Analyzer (SSA).

Details of the measurement setup—used for both competition preparation and the event itself—are available under the Measurement Setup section.

During the on-site competition at IMS2025, participants will use the same remoteUPCLab system. Each team will be allotted a maximum of 30 minutes to tune their DPD algorithm and upload the corresponding predistorted signals. Once the tuning phase ends—or sooner, if the team is ready—a competition test signal will be provided. This signal will include various multi-carrier configurations.

At that point, the team’s DPD algorithm must generate the final predistorted balanced and control signals, which will be uploaded to the signal generator. The jury will then measure and record key performance metrics to determine the team's overall score.

II. Signals

The test signal consists of five 40 MHz bandwidth, 16-QAM modulated, 5G-like waveforms. The DPD algorithm must be robust enough to handle various configurations of this multi-channel signal, ranging from a single active 40 MHz channel to a composite 200 MHz waveform with all five channels active simultaneously.

III. Scoring

At each iteration, the DPD linearizer performance will be scored taking into account five different quality metrics:

  • The average output power (dBm)
  • The PA power efficiency (%)
  • The worst channel ACPR value (dB)
  • The worst channel EVM value (%)
  • The number of coefficients of the DPD function

The out-of-band linearity is measured in terms of the adjacent channel power ratio (ACPR), computed for each band as the difference (in dB) between the integrated in-band power and the highest between the right and left, adjacent, integrated out-of-band powers (within the same bandwidth). The time-domain error is measured as error vector magnitude (EVM) computed as the root mean squared (RMS) value of the error in percentage (%) between the ideal constellation of symbols (being originally generated) and the measured output signals at each band. The number of coefficients of the DPD is represented in real-valued or complex-valued coefficients, accordingly to the DPD function employed by the contestant.

 Scoring trends:

  • ACPR Contribution (Weight: 10): ACPR values below the minimum threshold improve the score, while values above it reduce the score. The minimum ACPR depends on the signal configuration:
      • For 1 or 2 active channels, or the specific case of [0,1,1,1,0]: –45 dB
      • For any configuration with 3 or more active channels: –40 dB
  • EVM Contribution (Weight: 5): EVM values below the minimum threshold improve the score; values above the minimum threshold decrease it.
    • Only considered if the minimum ACPR threshold (–45 dB or –40 dB) is met.
  • Average Output Power (Weight: 10): Output powers above 30 dBm increase the score, while powers below 30 dBm reduce it.
        • Only considered if the minimum ACPR threshold is met.
      • Power Efficiency (Weight: 3):
        • Contributes positively only when the minimum ACPR requirement is satisfied.
      • DPD Model Complexity (Weight: 0.1):
        • Using fewer coefficients than the threshold improves the score; more coefficients reduce it.
        • Complexity thresholds depend on the number of active channels and coefficient type (real or complex):
          • For 1 or 2 channels:
            • Max 50 complex-valued or 200 real-valued coefficients
          • For 3 or more channels:
            • Max 100 complex-valued or 400 real-valued coefficients
        • 1 complex coefficient = 4 real coefficients
        • This contribution is only considered if the minimum ACPR is met.

      where ACPRmin is -45 dB if the signal consists of 2 channels, 1 channel or the [0,1,1,1,0] configuration, otherwise -40 dB. Additionally, ncoeffs,max is 50 if the DPD function operates with complex-valued coefficients, 200 if these coefficients are real and the signal consists of 2 channels or less. Otherwise, for more than 3 channels, ncoeffs,max is 100 if the DPD coefficients are complex-valued or 400 if these are real-valued. Finally, EVMmin =1.5% and Pmin=30 dBm.

      On the day of the competition at the IMS venue in San Francisco, the final test will be run three times with different signal configurations (unknown to contestants until the test). Your final SCORE will be the average of these three runs with varying channel configurations.

      IV. Matlab Functions Description

      The overall procedure can be summarized following these steps:    

      1-Signal generation

       [TX,PARAM]=IMS2025_generate_signal(config_chan)

       This function provides i) a multi-channel signal composed of 40 MHz bandwidth, 16-QAM modulated, 5G-like waveforms; and ii) A struct array containing the signal generation and acquisition settings (PARAM). The generated signal always has the same mean power but different PAPR.

       

      Inputs:

      - config_chan - A 5-element binary array indicating the status of each 40-MHz carrier (1 for active and 0 for inactive) (e.g., the figure above corresponds to [1,1,1,1,1]). If the input is left blank, i.e., [TX,PARAM]=IMS2025_generate_signal(), a random configuration of active and inactive channels is used for testing.

      Outputs:

      - TX - Structure array containing, among other fields, the generated multi-channel signal.

      - PARAM - is a structure array containing relevant information on the baseband signal processing, for example: OFDM signals bandwidth, frequency bands, length of the data vectors, the sampling frequency.

      1.1-Signal back-off

      - Use code line: uBB=TX.uBB*GainBB

      Line description: This line uses the gain GainBB ranging from 0 to 1 to set the amplitude of the baseband generated signal and to control the input power. The absolute value of the IQ signals to be sent to the remoteUPCLab must not exceed 1; if the value exceeds 1, the program will automatically clip the signals to 1 or generate and error message. All signals are generated with the same average power but different PAPR, so users can select lower gain values if desired to work with higher back-off levels.

      2-LMBA Shaping Function to Generate balance and control input signals

       [TX.xBB_Bal,TX.xBB_CSP]=IMS2025_LMBA_shaping_function(xBB,PARAM,Delay)

      This function generates the balanced TX.xBB_Bal and the control TX.xBB_CSP signals from the predistorted baseband signal xBB. The shaping function used to generate the balanced and control signals is based on a frequency domain linear shaping (FLS) [Li23]. The frequency offset and slope parameters of the FSL are already optimized and cannot be changed by the user. However, the user must find and set the optimal time delay (in samples) Delay between the balanced and control output signals.

      Inputs:

      - xBB - is the baseband IQ predistorted signal.

      - PARAM - is a structure array containing relevant information on the baseband signal processing, for example: OFDM signals bandwidth, frequency bands, length of the data vectors, the sampling frequency.

      - Delay - is the delay (in samples) between the two LMBA inputs (balance and control signals). The optimal value is not provided and must be determined by the user.

      Outputs:                                                                  

      - TX.xBB_Bal - is the balanced input of the LMBA.

      - TX.xBB_CSP - is the control input of the LMBA.

       3-Waveform UL-PA excitation-DL

      [RX]=TX_RX_Remote_UPC_WebLAB_IMS2025(TX)

      This function is used to upload the TX structure, containing the balanced and control input signals, to the remoteUPCLab FTP Matlab server (in the control PC) for further generation with the VSG. The SSA will then capture the Rx waveform at the output of the PA after down conversion and the FTP Matlab server will send this waveform back to the user through the RX structure which also includes relevant power metrics given by the measurement setup and relevant hardware settings information.

      - Notes: The absolute value of the TX waveforms TX.xBB_Bal and TX.xBB_CSP cannot exceed 1.

      4-Scoring

      [SCORE]=IMS2025_meas_score(RX,TX,PARAM,iteration)

       Function description: This function provides the SCORE.

      Inputs:

      - RX - Structure array containing, among other fields, the measured output of the PA.

      - TX - Structure array containing, among other fields, the generated multi-channel signal.

      - PARAM - Structure array containing information on the most relevant remoteUPCLAb parameters.

      - iteration - is the current iteration of the DPD algorithm

      Outputs:

      - SCORE - is the numerical value calculated as defined in section III.

      Access details

      Our ambition is to have the remoteUPCLab permanently available. However, please respect that we may need to shut it down temporarily for maintenance.

      REFERENCES

      [Qua18] R. Quaglia and S. Cripps, "A Load Modulated Balanced Amplifier for Telecom Applications," in IEEE Transactions on Microwave Theory and Techniques, vol. 66, no. 3, pp. 1328-1338, March 2018, doi: 10.1109/TMTT.2017.2766066.

      [Li23] W. Li, G. Montoro, W. Thompson, K. Chuang and P. L. Gilabert, "Digital Shaping and Linearization of a Dual-Input Load-Modulated Balanced Amplifier," 2023 International Workshop on Integrated Nonlinear Microwave and Millimetre-Wave Circuits (INMMIC), Aveiro, Portugal, 2023, pp. 1-3, doi: 10.1109/INMMIC57329.2023.10321805.

      [Gui22] E. Guillena, W. Li, G. Montoro, R. Quaglia and P. L. Gilabert, "Reconfigurable DPD Based on ANNs for Wideband Load Modulated Balanced Amplifiers Under Dynamic Operation From 1.8 to 2.4 GHz," in IEEE Transactions on Microwave Theory and Techniques, vol. 70, no. 1, pp. 453-465, Jan. 2022. doi: 10.1109/TMTT.2021.3091672.