Incorrect eVOA settings with Raman usage?
cgkelly opened this issue · 4 comments
Using release 2.3.1.
I have a trial 5 span network (Roadm-ILAs-ROADM), using Raman fiber, without/with Raman pumps defined.
Without the pumps, the amp output eVOAs were set to 2-3 dB, and the OSNR (sig BW) was ~ 18 dB, SNR NLI ~24 dB on average.
With the Raman pumps, the EDFA gains dropped as expected, but the output eVOA values also dropped (0 to 1 dB setpoints).
As a result, the OSNR improved to ~21 dB, but the SNR NLI dropped to around the same level, resulting in a net SNR of ~18 dB, only slightly better (~0.5 dB) than without the pumps.
The behavior I expected was lower launch powers for the Raman case.
This occured with my defined EDFAs, and to make sure it wasn't a quirk of my amps, I then tried it with the default Juniper HG amp (as the gain range was about right to cover both scenarios). However, I modified the output power limit from 21 dB to 23 dB in order to "enhance" the nonlinearities if the eVOA loss was too low.
In both cases, I modified the target tilts manually, as auto tilt adjustment did not work all that well for these runs.
GNPY topo files for eVOA issue.zip
@cgkelly, as I mentioned in #398, GNPy auto-design is completely unaware of Raman gain and you probably shouldn't rely on it to set delta_p (or gain) and out_voa when you have a RamanFiber with pumps.
In this specific case there is also a difference between the topologies with and without pumps: for the Raman fibers with pumps the con_out is almost 3 dB higher (I guess to model insertion loss of pump combiners and other Raman amp components), which has two consequences:
-
The input power to the Edfas expected by auto-design is lower (since Raman gain is not taken into account) so the output power is limited by max gain and not max power (you can test this by increasing gain_flatmax).
-
The higher exit span loss means auto-design actually tries to set a higher target output power, which means lower out_voa.
So you get higher NLI both because of the higher power into the fiber and because of the Raman gain. In a real case, like you say, you would probably set a lower Edfa output power into a Raman amplified span but this is not currently supported by auto-design so you have to specify it manually.
Does the lack of Raman awareness all impact the amp tilt settings? Currently, with power mode set to off, and manual tilt and eVOA settings, I get full amp Pout at my ILAs and ROADM ingress amp (but not the ROADM egress amp, to achieve this I had to set a high gain target). However, the ROADM egress amp gain has the correct (set) tilt, but the ILA amplifiers have less than 1/2 of the set tilts. (I have not compared this with the non-Raman fiber case).
First of all, there is no "auto tilt adjustment" implemented in GNPy. You always have to set the EDFA tilt manually. So there should be no direct impact from using Raman amplification.