Digitisation and tracking steering files for the new ILD_l5_v09 model

[jfklama]

BEGINRELEASENOTES

• Digitisation including the hits from CLIC-like outer tracker and changes for the FTD, allowing for further use of the Conformal Tracking
• Conformal tracking with input collections from the new ILD_l5_v09 model, parameters setup as in CLIC config.

ENDRELEASENOTES

This is still work in progress, requires e.g. some further parameter tuning, but I was suggested to share these files (which of course makes a lot of sense).

[tmadlener]

Hi @jfklama, thanks for this. Quick question, in order to run the silicon tracking instead of the usual TPC, what changes are necessary in MarlinSTdReco.xml? Is replacing the TrackingDigi and TrackingReco groups with SiliconTrackingDigi and ConformalTrackingReco enough (at least from a technical point of view)? Or are more steps necessary?

[jfklama]

Hi @tmadlener, right, good point. I've just recently figured out how to set up ParticleFlow and HighLevelReco groups to run them with this model, so the corresponding steering files also need changes. I will push them together with some more recent changes I made to the Conformal Tracking setup. However, I haven't tested their performance yet. In any case, I would not merge it until I finish tests on the ttbar events.

Apart from that, and what you have mentioned, no changes are needed in MarlinStdReco.xml.

[tmadlener]

Very nice. In this case, I will mark this as a draft, and let you remove that again, once you think it is ready for review / to be merged. In case there are any questions in the meantime, or some interim reviews you want to have, simply tag me in a comment, and I will have a look.

[jfklama]

@tmadlener,

Now it corresponds to the results for ttbar events I have presented at the last meeting and from my side this is ready. Please just note that I tested only tracking. For HighLevelReco and PandoraPFA I have checked the inputs/outputs and some other settings such that it does not crash, but proper tests would require probably much more time and effort.

[tmadlener]

Hi @jfklama thanks for updating.

I have checked the inputs/outputs and some other settings such that it does not crash

Do I understand correctly, that you have run the full reconstruction with these changes in place? Or did you simply adapt the steering file to have the correct collection names?

[jfklama]

Do I understand correctly, that you have run the full reconstruction with these changes in place? Or did you simply adapt the steering file to have the correct collection names?

Yes, I have run the full reconstruction chain - I checked that it does not crash and all the respective output collections are saved, but I just did not inspect the "quality" of their content.

[tmadlener]

Excellent, thanks for checking that as well. Then I think we can merge this as it is for now. Getting "proper physics performance" is probably a whole task in itself, and will take quite a bit of time as you already said.

Quick question about how involved it is in the end to run everything with silicon tracking: Presumably, you simply have to use the ILD_l5_v09 compact file for ddsim and for reconstruction. Additionally in the reconstruction you have to replace a few of the groups in the execute section. Is there more to do, or does that cover everything to at least get something running? Could you add a very brief section to the README in the StandardConfig/production folder that simply states these steps, so that people don't have to go look into the PR? A few bullet points would be enough, it doesn't have to be a "copy these commands and everything works" solution!

[jfklama]

What you described is perfectly sufficient to run simulation and reconstruction, I've added very short section to README.

Maybe one more comment: as you can see, this could be even more simplified if we name digi and tracking groups the same as by default (then you just need to include proper steering files). I thought it could be somehow confusing maybe, so I set different names, but I can change it if you think it would be better that way.