Getting Value error while executing _generalized_psi function

[mukherjeearup]

Hi,
Many thanks for answering my previous question, it helps. :)
Now I am trying to run Gnpy code (specifically Non linear parts like GN, GGN, SRS) in jypyter notebook. GN, SRS ran fine but I am getting one error while executing GGN. To be specific, "_generalized_psi" function is failing in below line:

----> 8 f1_array = arange(pump_carrier.frequency - (pump_carrier.baud_rate * (1 + pump_carrier.roll_off) / 2),
pump_carrier.frequency + (pump_carrier.baud_rate * (1 + pump_carrier.roll_off) / 2),
f_pump_resolution)

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

@AndreaDAmico Can you please help to fix this error?

[AndreaDAmico]

Hi,
That error is raised within the numpy method arange when at least one of the three inputs is an array instead of being a scalar. You should find out which of them is an array and why. This behavior has to depend on your specific usage (maybe a wrong format in the input files) as it is not the expected behavior (that function is covered within the GNPy tests).

[mukherjeearup]

Sure @ojnas , I am attaching my GGN ipynb file. Attaching the results from Gnpy app as well for same inputs.
Please note that I have verified SRS power profile (which is used as one of the inputs for GGN) with GnPy code and it is exactly matching.
Desktop.zip

Inputs to GGN model:
Start freq = 191.35
end freq = 196.1
spacing = 50 GHz
Per channel launch power into fiber = 0 dBm (same for all 96 channels)
Connector in = 0
connector out = 0
span type = SSMF
span length = 80 km
atten coeff = 0.2 dB/km
No raman pump, no lumped losses
n2 = 2.6e-20
effective_area = 83e-12
ref_wavelength = 1550e-9
c = 299792458
all channels baud rate = 32 GBd
roll off = 0.15
DEFAULT_SLOT_WIDTH_STEP = 12.5e9
dispersion_tolerance=1
phase_shift_tolerance=0.1
k_ref = 5
beta2_ref = 21.3e-27
delta_f_ref = 50e9
rs_ref = 32e9
dispersion = 1.67e-5

Can anyone please check the reason for NLI OSNR diff? @AndreaDAmico @ojnas

[mukherjeearup]

Hi @ojnas , @AndreaDAmico did you get a chance to look at my code?

[ojnas]

Hi @mukherjeearup, the main issue with your code is that you're setting beta2 = abs(beta2). Another thing is that you set cuts to all carriers, while I believe you run gnpy with default sim_params.json, which specifies a limited number of "computed_channels". A third thing is that you set slot_width to 37.5 GHz, while by default in GNPy it would be equal to spacing (but I think this probably has negligible impact).

[mukherjeearup]

Many thanks @ojnas . Now I ran my code for less number of cuts as mentioned in sim_params.json and NLI values are now closer :)

[ojnas]

Note that you can simply remove the "computed_channels" attribute from sim_params.json if you want cuts to be all carriers. If I do that, and make the other two changes in your code that I mentioned above, I get identical NLI with your code and with gnpy.app.

[ojnas]

Hi @mukherjeearup,
the results with gnpy.app should not be different. You can also verify by running the gnpy-transmission-example script included with gnpy, which should give the same results as gnpy.app, with the same inputs. One thing I noticed is that the results you attached were for 2 dBm launch power, while you wrote 1 dBm, but maybe that was just a typo.