Resonator spectroscopy

[eggerdj]

Summary

This PR will replace #90 which is rather old now. The experiment can be run as follows.

spec = ResonatorSpectroscopy(0, frequencies=np.linspace(-10e6, 10e6, 51), backend=backend)
exp_data = spec.run()

Details and comments

This experiment requires special data processing to correct for the fact that the IQ points rotate around in the IQ plane.

TODOs:

• Analysis class
• Get qubit != 0 case working
• Data processing
• Tests

[nkanazawa1989]

Thanks Daniel. This is great start. I added couple of change requests for documentation and simplification. I feel this experiment motivate us to separate processor from analysis so that it can be flexibly configured.

[nkanazawa1989]

What is the reason we cannot reuse superclass method?

[eggerdj]

Because in resonator spectroscopy we are attaching the pulse schedule to the measurement instruction which does not have a parameter. Therefore, we can not do:

circuit.add_calibration("measure", self.physical_qubits, sched_)
assigned_circ = circuit.assign_parameters({freq_param: freq_shift}, inplace=False)

for example, the following simple test case will raise:

import qiskit.pulse as pulse
from qiskit.circuit import Parameter
from qiskit import QuantumCircuit

param = Parameter("amp")

with pulse.build(backend=None) as sched:
    pulse.play(pulse.Drag(160, param, 2,0, 40), pulse.DriveChannel(2))
    
circ = QuantumCircuit(1, 1)
circ.measure(0, 0)
circ.add_calibration("measure", (2, ), sched)
circ.assign_parameters({param: 0.2345566}, inplace=False)

[wshanks]

This looks good to me, but I think all of Naoki's points are good.

I do wonder how the measurement is working here -- is the backend demodulating based on the shifted frequency of the drive or is it always demodulating at the nominal measure frequency.

[wshanks]

Additional comments

[wshanks]

I put a few minor comments in. I think this is very close to done. Besides those comments, there is my previous comment about acquisition duration and delay. Also, perhaps someone from #387 will weigh in.

Besides that, I think this is ready, though I was also hoping to see backend support added for this experiment before it was merged so I could try it out. My understanding is that the backend team has been trying to push out this support all week but has not been able to get the update through the QA pipeline without errors.

[wshanks]

What is the circuit metadata used for? Does it matter that for the resonator spectroscopy the qubit is not really part of the experiment?

[eggerdj]

The qubits entry does not matter. Tests run fine without it. The only crucial part of the metadata is the xval since this is what is used in the curve analysis. I think we can tacitly interpret this as being the qubit for which we are doing readout spec. Alternatively we could override the _add_metadata function in ResonatorSpectroscopy to change this field but this sounds like a bit too much.

[nkanazawa1989]

Looks good to me with minor documentation changes. Please address my final comments and feel free to merge with @wshanks 's approval.

[nkanazawa1989]

You can use #section: examples to write some code. As you mentioned, we cannot set default value to run option since this is provider dependent. Even if we don't assume non-IBM provider for experiments, simulator may crash with invalid options (probably meas level and meas return are okey since they are defined in OpenPulse spec).