How to obtain the output signal of the final cascaded system？

[Flashwxy]

After running run_dpd, only the predistorted signal was obtained. How can it be made to pass through PA to obtain the final linearized output?

[gaochangw]

Thanks for your question. We can update run_dpd.py to implement that. Before that, you can also easily add a few lines of code yourself to pass the dpd_out to the PA model again after line 82 in run_dpd.py.

[Flashwxy]

Thank you for your answer. My approach is basically the same as yours, which is to input the predistorted signal into the already trained PA model. However, the ACLR of this method differs significantly from that of the best model obtained by train_dpd. Could you provide a specific code?
TEST_ACLR_AVG -38.93 dBc -44.76701682 dBc

PA_out.py

[yizhuo990]

Which dataset are you referring to?

[Flashwxy]

DPA_200MHz

[yizhuo990]

The train_dpd results save the best performance on the validation dataset. Please check if these two are consistent.

[Flashwxy]

I am currently using the predistorted signal obtained from run_dpd as the input, calling the model trained by train_pa to obtain the final output signal, and then writing a code to calculate the ACLR myself. The result obtained is quite different from that of train_dpd.

[yizhuo990]

The test signal consists of three frames (if I remember correctly, but it's not a single-frame dataset) and 20 MHz × 10 channels for each frame. Therefore, we need to split the signal correctly and pick the right channel to calculate the ACLR. Could you please verify that you are using the same method to calculate the ACLR?

[Flashwxy]

The above-mentioned methods were indeed not taken into account. Could you please provide some specific guidance? Also, could you please clarify exactly how to call the trained PA model to obtain the final output? I'm worried there might be an error here.

[Flashwxy]

Could you please clarify exactly how to call the trained PA model to obtain the final output? I really don't quite understand.