 |
| Stephen Brooks 2014-08-27 18:48:29 | The 1Xc2 lattice is the same as 1Xc2_nosample but with network-wide sample files turned on, so a direct comparison there. It is also similar to 4Xc2 but with the linac only containing one block rather than four. The 8Xc3 lattice changes the parameterisation of the RF phase again. I found that letting it vary over two cycles generally optimised better than only letting it vary over one (perhaps because getting stuck at the ends of the range is unnatural since it's a circular variable). So I'm trying three cycles (and using 8 blocks, as gave the highest linac yield under v4.45 previously). |
| Stephen Brooks 2014-09-04 15:40:09 | Something interesting is already happening if you compare 1Xc2 with 1Xc2_nosample. Here is the same optimisation done with and without sample files, and the yields are already similar. The question is can 1Xc2 with samplefiles beat the _nosample in a shorter time? |
