Session aimMARTe tests and commissioning for MLControlGAM and then plasma pulses eventually
Session day2025-05-28Operatorrhenriques
Session IDISTTOK_20250528_1Coordinatorrhenriques

Shot Time Operator Iplasma(kA) n(1018m-3) Duration (ms) Pre/post shot comments
528702025-05-28 17:44:06.512rhenriques000Repeat with increased real-time cycle from 100 us to 150 us | Still many packet lost observed, execution time is still 7.2 us
528692025-05-28 17:37:58.533rhenriques000Repeat to check logger messages | Packet lost happens in large quantity
528682025-05-28 17:33:59.336rhenriques000Repeat with 75 hidden layers | Logger viewer stopped and could not check for lost packets, execution time: 7.2 us
528672025-05-28 17:28:32.045rhenriques000Repeat with increased number of hidden layers (from 25 to 50) to check for lost packets | No packets lost (other than the usual few), MLControlGAM takes about 3.8us to execute (with 25 layers was 2.1)
528662025-05-28 17:23:13.617rhenriques000Repeat with limits set to: Prim -> 40, Vert -> 70, Horz -> -30 A | All 3 clipping limits worked correctly
528652025-05-28 17:18:14.134rhenriques000Repeat with negative output scaler weigths for MLControlGAM to test the opposit clipping limits in ControllerGAM | OK, output is inverted but forgot to set the limits
528642025-05-28 17:10:14.604rhenriques000Repeat with correction of ControllerGAM | All 3 limits OK
528632025-05-28 16:34:29.475rhenriques000Repeat to test current limit clipping in ControllerGAM when using control mode 9 (parsing MLControlGAM signals), testing: negative limit for Prim. and Vert. and + limit for Horz. | P and H OK, V not
528622025-05-28 16:19:25.616rhenriques000Repeat with new MLControlGAM that contains a scaler for the output and with reduced hidden layers: 25 only to check packets lost | No packets lost (other than the usual few), all 3 control modes OK
528612025-05-28 15:18:39.221rhenriques000Repeat with ML control on horizontal field during 100 ms | OK, worked as expected
528602025-05-28 14:59:33.625rhenriques000Test recompiled ISTTOK GAMs from oper/ML folder with MLControlGAM and new MARTeNode for data acquisition and storage | OK, everything looks technically as expected