The relationship between 'tx_power_dbm' in the spectrum.json file and the GSNR of each channel in the final results.

[litengxiao]

I casually used a simple link with two Transceiver, two Roadm, amplified by a few EDFA segments in between, and then used Spectrum to set the transmit power for each channel, but the result was the same gsnr each time.
"spectrum": [
{
"f_min": 191350000000000.0,
"f_max": 191350000000000.0,
"baud_rate": 32000000000.0,
"slot_width": 50000000000.0,
"roll_off": 0.15,
"tx_osnr": 40,
"tx_power_dbm": -4.5
},
{
"f_min": 191400000000000.0,
"f_max": 191400000000000.0,
"baud_rate": 32000000000.0,
"slot_width": 50000000000.0,
"roll_off": 0.15,
"tx_osnr": 40,
"tx_power_dbm": -5.2
},
The settings are similar to this, and the GSNR of the result does not change
Ch. # Channel frequency (THz) Channel power (dBm) OSNR ASE (signal bw, dB) SNR NLI (signal bw, dB) GSNR (signal bw, dB)
1 191.35000 -20.04 21.44 26.91 20.35
2 191.40000 -20.04 21.44 26.34 20.22
3 191.45000 -20.04 21.43 26.09 20.16
4 191.50000 -20.04 21.43 25.93 20.11
5 191.55000 -20.04 21.43 25.82 20.08
6 191.60000 -20.04 21.43 25.73 20.06
7 191.65000 -20.04 21.43 25.65 20.04
8 191.70000 -20.04 21.43 25.59 20.02
I'm a newbie to gnpy and don't know why

[EstherLerouzic]

From what you show, I think that the --spectrum option has not been used to input the proper spectrum (only 2 channels from what I see of your spectrum definition), so that the propagation uses the spectrum defined in the SI section of the library instead. The proper command should be :

gnpy-transmission-example my_network.json --spectrum <FILE.json>

Did you use the --spectrum option ?

Please note that GNPy works well for "full" load. We do not recommand propagation for only a small number of channels in the spectrum as results may suffer from inaccuracy.

[litengxiao]

Yes, I used the --spectrum option. The snippet above only shows the configuration for two channels, but here is my complete spectrum.json file. After running with the spectrum option in power‑mode, setting different tx_power_dbm values for each channel doesn’t seem to have any effect; the power_dbm value in the SI settings appears to flatten out any differences. As a result, I can’t observe the effect of changing tx_power_dbm each time. Do you have any good methods to diagnose or work around this? Thank you very much for your help!
spectrum.json

[EstherLerouzic]

Your simulation probably has ROADMs, and ROADMs equalize power at their outputs (using target_pch_out_db for example https://gnpy.readthedocs.io/en/master/json.html#arbitrary-channel-definition). This may explain why the tx_power_dbm does not have much effect. It will have an effect only if you have defined a max-loss for your ROADM impairment. then if the input power is too low, the ROADM can not properly equalize power. If you would like to define different power at the fiber span input, you need to use delta_pdb instead: then the ROADM will consider these delta for the equalization. Does this solve your case ?

[litengxiao]

Ok, thanks for the reply I'm going to give it a try

[litengxiao]

Yes, I tried to remove the ROADM in the link first, and it did change due to the power equalization, thank you for your answer, very, very much