Add postselection to readout characterization

[sorewachigauyo]

We have been getting poor readout on our 5Q device for a while. On checking further, I noticed that there is a bit of ground state shots in the excited state clusteroid.

By adding a pre-measurement and filtering based on that, the readout fidelity improves by quite a bit. (Results not shown because plotting of the postselected data hasn't been implemented yet)

TODO:

• Tidy initial dirty implementation
• Streamline characterization protocol
• Plotting of pre-and-postselected data
• Tests?

[codecov]

Codecov Report

❌ Patch coverage is 70.37037% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 94.29%. Comparing base (883e5c5) to head (23488e6).

Files with missing lines	Patch %	Lines
...ocal/protocols/readout/readout_characterization.py	68.00%	8 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1548      +/-   ##
==========================================
- Coverage   94.35%   94.29%   -0.06%     
==========================================
  Files         136      136              
  Lines       10638    10653      +15     
==========================================
+ Hits        10037    10045       +8     
- Misses        601      608       +7     

Flag	Coverage Δ
unittests	94.29% <70.37%> (-0.06%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/qibocal/protocols/utils.py	89.59% <100.00%> (+0.25%)	⬆️
...ocal/protocols/readout/readout_characterization.py	93.04% <68.00%> (-6.96%)	⬇️

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[alecandido]

But then, if you are dropping states which are excited at the beginning of the experiment, are those states generated by a too-short relaxation? Or is it a residual thermal population? It seems quite a large difference...

In any case, I'm fine with the concept of this PR, but I was considering that we may want to implement the initialization measurement at a more general level, in Qibocal or Qibolab (preferentially the first one), such that it can be toggled for most (or all) experiments, in a similar fashion to the reset (relaxation time).

[sorewachigauyo]

But then, if you are dropping states which are excited at the beginning of the experiment, are those states generated by a too-short relaxation? Or is it a residual thermal population? It seems quite a large difference...

For this chip, the relaxation time is 200us when the T1 is about 20-30us. Currently our suspicion is on the thermalization.

In any case, I'm fine with the concept of this PR, but I was considering that we may want to implement the initialization measurement at a more general level, in Qibocal or Qibolab (preferentially the first one), such that it can be toggled for most (or all) experiments, in a similar fashion to the reset (relaxation time).

Sure, I can work on that in Qibocal.

[alecandido]

For this chip, the relaxation time is 200us when the T1 is about 20-30us. Currently our suspicion is on the thermalization.

A bit unexpected, but I understand (I do not really see any other option).

Sure, I can work on that in Qibocal.

No worries: right now there is no way to do that without touching each and every protocol. I will write an issue for that, and it may align with a proposed simplification for the protocols' acquisition definition.