DSP of RF PA and DPD

Summary of simulation

• Summary_PA vs DPD.md
• Summary_DPD coefs. applied vs Waveform.md
• Summary_DPD coefs. applied vs PA.md

Main: PA_DPD_main

• Initialization

flag_orx_ripple = 0   
flag_orx_impair = 1   
flag_pa_awgn = 0    
flag_pa_combine = 0   
flag_pa_product = '20W'   
flag_pa_Nstages = 1   

• Power assignment

pwrdBm_In_target = 3.5

• AWGN

pwrNoise_dBmHz = -174

• Signal settings

condSig.aclr = 1;   
condSig.aclr_fs = fs;   
condSig.aclr_bwInband = bwInband;   
condSig.aclr_foffset = foffset;   
condSig.ccdf = 1;   
condSig.ccdf_Nsamps = numel(x);   
condSig.pwr = 1;    
condSig.pwr_fs = fs;    
condSig.pwr_bwInband = bwInband;    

• PA Model: Memoryless/am2pm/noise ...

% pa model
pa.model = "Cubic polynomial"; % model of am/am and am/pm   
pa.gain_dB = 32; % Input gain_dB   
pa.am2pm = []; % Input am/pm   
pa.oip3_dBm = []; % Input oip3    
pa.opsat_W = 20;   
pa.opsat_dBm = 10*log10(1000*pa.opsat_W);   
pa.TOISpecification = 'OPsat';   

% pa ampm
pa_ampm_pwr = [10 15 20 25 30 35 38 40]   
pa_ampm_shifter = [0 1 2 4.5 9 15 19.5 19.5]   
pa_ampm_pwrIn = pa_ampm_pwr - pa.gain_dB     

% pa noise
pwrNoisedBmHz = -150

% pa combination and phaseShift
pa.pa_cmb_Mstages = 2;
pa.pa_cmb_Pshifts = [0, 0]

• PA output
  • generate pa model class

paCls =

powerAmp with properties:

               Method: 'paModel'
             pa_model: "Cubic polynomial"
           pa_gain_dB: 32
         pa_am2pm_deg: 0
  pa_TOISpecification: "OPsat"
          pa_iip3_dBm: Inf
          pa_oip3_dBm: Inf
         pa_ip1dB_dBm: Inf
         pa_op1dB_dBm: Inf
         pa_ipsat_dBm: Inf
         pa_opsat_dBm: 43.0102999566398
pa_PowerLowerLimit_In: -30
pa_PowerUpperLimit_In: Inf
        pa_Table_cell: {[0]}
           pa_opsat_W: 20
            am2pm_deg: {[8×2 double]}
          noise_dBmHz: -150
              flat_dB: 0
                    x: []
                    y: []
               pwrdBm: []
                   fs: 1966080000
              Nstages: 1
             lna_NFdB: NaN
             flag_lna: 'pa'
                  plt: [1×1 struct]
       pa_cmb_Mstages: 0
       pa_cmb_Pshifts: 0

• plot settings

paCls.plt.fnum = 1;
paCls.plt.flag = 'aclr';
paCls.plt.fs = fs;
paCls.plt.bwInband = bwInband;
paCls.plt.foffset = foffset;
paCls.plt.Nfft = Nfft;
paCls.plt.legend = '20W, power combine = 0'

• output

• Plot: AM/AM, AM/PM

plt_ampm.Rohm = 1;
plt_ampm.unit = 'dBm';
plt_ampm.xlim = [0, 40]
plt_ampm.type = 'amgain';

• CCDF

• ACLR

• Learning parameters, input: dpm: dpd parameters

DPM.order_poly = 9;
DPM.depth_memory = 7;
DPM.Niterations = 30;
DPM.flag_even_order_poly = 1;
DPM.flag_conj = 0;   % Conjugate branch. Currently only set up for MP (lag = 0)
DPM.flag_dc_term = 0; % Adds an additional term for DC
DPM.flag_LS_exclude_zero_second = 0;
DPM.modelFit = 'WIN'; % 'GMP'/'HAM'/'WIN'
DPM.evm = bwInband;
DPM.learning_arc = 'DLA'; % better!
DPM.fnum = 0721;
DPM.flag_Multicarrier = '2C';

• ORX parameters

DPM.ORX_RippledB = 0;
DPM.ORX_SNRdB = 80;
DPM.ORX_SNRdB = [];

• Plot parameters

plt.fs = fs;
plt.bwInband = bwInband;
plt.offset = foffset;
plt.dispTitle = [];
plt.legend = paCls.plt.legend
plt.title = 'DPD, ACLR'

• DPD learning

  • generate dpd class

  DPD_g with properties:

                 order_poly: 9
               depth_memory: 7
             order_poly_lag: 0
           depth_memory_lag: 0
                  depth_lag: 0
                Niterations: 30
       flag_even_order_poly: 1
                  flag_conj: 0
               flag_dc_term: 0
               learning_arc: 'DLA'
              learning_rate: []
            learning_method: []
  flag_LS_exclude_zero_second: 0
                   modelFit: 'WIN'
          flag_Multicarrier: '2C'
                 ORX_FlatdB: 0
                  ORX_SNRdB: []
                    ORX_LPF: []
                        CFR: []
               paprdB_limit: []
                       fnum: []
                  coeffs_PM: [63×1 double]
                    Ncoeffs: 63
          coeffs_PM_history: []
                     paprdB: []
        sigOut_u_predistort: []
             sigOut_y_paout: []
                        evm: [-249140000 … ]
                  ACLRdBcal: []
                     NMSEdB: []
               ACLRdBmargin: -1
               IpwrdBerrCal: []
            flag_Ipwr_check: 1
            flag_aclr_check: {[70]}
        DPDexpansionDBLimit: 5
         DPDexpanIterations: []
  

  • learning

  dpd =

  DPD_g with properties:
  
                 order_poly: 9
               depth_memory: 7
             order_poly_lag: 0
           depth_memory_lag: 0
                  depth_lag: 0
                Niterations: 30
       flag_even_order_poly: 1
                  flag_conj: 0
               flag_dc_term: 0
               learning_arc: 'DLA'
              learning_rate: []
            learning_method: []
  flag_LS_exclude_zero_second: 0
                   modelFit: 'WIN'
          flag_Multicarrier: '2C'
                 ORX_FlatdB: 0
                  ORX_SNRdB: []
                    ORX_LPF: []
                        CFR: []
               paprdB_limit: []
                       fnum: []
                  coeffs_PM: [63×1 double]
                    Ncoeffs: 63
          coeffs_PM_history: [63×30 double]
                     paprdB: [7.35 7.89 8.13 … ]
        sigOut_u_predistort: [1966080×1 double]
             sigOut_y_paout: [1966080×1 double]
                        evm: 0.943088130181803
                  ACLRdBcal: 50.28
                     NMSEdB: -40.53
               ACLRdBmargin: -19.72
               IpwrdBerrCal: -0.25
            flag_Ipwr_check: 1
            flag_aclr_check: {[70]}
        DPDexpansionDBLimit: 5
         DPDexpanIterations: []
  

• DPD result: ACLR

• DPD result: CCDF

• DPD result: AM/AM, AM/PM