pedestal leveling

[tomeichlersmith]

To Do

• code cleanup and documentation, move task to own file with header
• add another iteration to level even more? going to leave further dev of the algorithm for the future, this is just a first draft and it already works great
• replace copy of DecodeAndBuffer with one in main, move calculation of medians to utility

Functionality Test

I ran this new task after a hard reset to check how it does. The pftool log of what I did is 180-commands.log
I then ran

$ python ana/pedestal/violinplot.py data/pedestal-2025-07-15-180927.csv -o default.png
$ python ana/pedestal/violinplot.py data/pedestal-2025-07-15-181017.csv -o leveled.png

to produce the following. The calib channels do not move (as expected, we don't bother leveling them).

[tomeichlersmith]

This does change the median (slightly) while significantly tightening the distribution of pedestal values across all channels.

[tomeichlersmith]

After my code cleanup, I re-ran the checks

# on zcu
./pftool -z -s 180-check.pftool | tee 180-check.log
# on local computer
python ana/pedestal/violinplot.py data/pedestal-2025-07-17-default.csv -o default.png
python ana/pedestal/violinplot.py data/pedestal-2025-07-17-leveled.csv -o leveld.png
# other histogram in jupyter notebook

180-check.log

[tomeichlersmith]

@erik-wallin @taylorjcolaizzi I am requesting your reviews to see how this functions on the HGCROCs you have. I want to make sure that it runs successfully. If you see results similar to mine that is an extra win.

I've been testing it by doing a roc.hardreset in order to see the default-parameter pedestals and then running tasks.level_pedestals in order to see the leveled pedestals. Note: after a roc.hardreset you will need to daq.setup.standard.
The 180-check.log shows exactly what I did.

# on zcu
cd pflib
git switch 180-pedestal-leveling
cd build
make -j

[taylorjcolaizzi]

@erik-wallin @taylorjcolaizzi I am requesting your reviews to see how this functions on the HGCROCs you have. I want to make sure that it runs successfully. If you see results similar to mine that is an extra win.

I've been testing it by doing a roc.hardreset in order to see the default-parameter pedestals and then running tasks.level_pedestals in order to see the leveled pedestals. Note: after a roc.hardreset you will need to daq.setup.standard. The 180-check.log shows exactly what I did.

# on zcu
cd pflib
git switch 180-pedestal-leveling
cd build
make -j

Sure thing! I followed your steps and share my output here. I'm attaching a zip file of my yaml file, since Github won't let me upload .yaml here.

level-pedestals-roc-0-settings_20250717_183223.yaml.zip

One thing I noticed is that CH_42.TRIM_INV = 2147483647, which might be a little high.

[taylorjcolaizzi]

And here's a pedestal run I took afterward!
pedestals_leveled.csv
Hopefully they are level :)

[OwenCole04]

Ran this on the ROC at UVA; This is what I got from a pedestal run after:

[tomeichlersmith]

Thank you both! I now have two more things to look into.

• fix ultra-high TRIM_INV issue. I'm guessing this is because the algo is accidentally finding a negative TRIM_INV which is then being converted to a super-high value when it is static cast into a unsigned int.
• fix upper limit of TRIM_INV. two of the channels that are far below the link median (70 and 34) have a TRIM_INV bigger than 63 mean which is then cropped to only 6 bits and thus are not shifted close to the median at all.

[taylorjcolaizzi]

[like] Colaizzi, Taylor James (zxk4gs) reacted to your message:

…

________________________________ From: Tom Eichlersmith ***@***.***> Sent: Thursday, July 17, 2025 7:49:24 PM To: LDMX-Software/pflib ***@***.***> Cc: Colaizzi, Taylor James (zxk4gs) ***@***.***>; Mention ***@***.***> Subject: Re: [LDMX-Software/pflib] pedestal leveling (PR #182) [https://avatars.githubusercontent.com/u/31970302?s=20&v=4]tomeichlersmith left a comment (LDMX-Software/pflib#182)<#182 (comment)> Thank you both! I now have two more things to look into. * fix ultra-high TRIM_INV issue. I'm guessing this is because the algo is accidentally finding a negative TRIM_INV which is then being converted to a super-high value when it is static cast into a unsigned int. * fix upper limit of TRIM_INV. two of the channels that are far below the link median (70 and 34) have a TRIM_INV bigger than 63 mean which is then cropped to only 6 bits and thus are not shifted close to the median at all. — Reply to this email directly, view it on GitHub<#182 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BS6CCXVS75437PCXRZ4F4UT3I744JAVCNFSM6AAAAACBSXHRESVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTAOBVGI4DAMBSG4>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[tomeichlersmith]

Alright, I added some checks in both the compiler and in the level pedestals algorithm.

@taylorjcolaizzi @OwenCole04 @verenamh Please give it a try again, I would appreciate the results in terms of the violin plot that this branch can make.

# on zcu
git switch 180-pedestal-leveling
git pull
cd build
make
./pftool -z
# roc.hardreset
# daq.setup.standard
# daq.pedestal -> call this default.csv
# tasks.level_pedestals
# daq.pedestal -> call this leveled.csv

# on local computer to make plots
cd pflib
git fetch
git switch 180-pedestal-leveling
python ana/pedestal/violinplot.py default.csv leveled.csv
# upload violinplot.png created

[taylorjcolaizzi]

Nice work! Looks quite nice! The leveling script printed an error for me during runtime, saying that channels 42 and 70 are so far below target that TRIM_INV had to be set to its maximum.

[verenamh]

I got the same error as Taylor, just for channel 0 and 2! Here is the plot I got: