Randomized benchmarking¶

This notebook shows a compact randomized benchmarking workflow: prepare a good single-qubit gate set, run standard RB, and then estimate specific gate errors with interleaved RB.

1. Create an Experiment¶

In [ ]:

import numpy as np

import qubex as qx

exp = qx.Experiment(
    system_id="YOUR_SYSTEM_ID",
    muxes=[0],
    # config_dir="/path/to/qubex-config/config",
    # params_dir="/path/to/qubex-config/params/YOUR_SYSTEM_ID",
)

import numpy as np import qubex as qx exp = qx.Experiment( system_id="YOUR_SYSTEM_ID", muxes=[0], # config_dir="/path/to/qubex-config/config", # params_dir="/path/to/qubex-config/params/YOUR_SYSTEM_ID", )

2. Connect to the setup¶

Connect before you run the prerequisite calibrations.

In [ ]:

exp.connect()

exp.connect()

3. Check the waveform¶

A waveform check is a quick sanity check before you spend time on benchmarking.

In [ ]:

waveform_result = exp.check_waveform()

waveform_result = exp.check_waveform()

4. Measure a baseline Rabi oscillation¶

Use the baseline Rabi result to confirm that the drive amplitude is in a usable range.

In [ ]:

rabi_result = exp.obtain_rabi_params()

rabi_result = exp.obtain_rabi_params()

5. Calibrate the DRAG half-pi pulse¶

Run the DRAG half-pi calibration first so the primitive gate set is in good shape.

In [ ]:

drag_hpi_result = exp.calibrate_drag_hpi_pulse()

drag_hpi_result = exp.calibrate_drag_hpi_pulse()

6. Repeat the DRAG half-pi pulse¶

Use a repeated-sequence check to confirm that the calibrated half-pi pulse accumulates as expected.

In [ ]:

repeat_drag_hpi = exp.repeat_sequence(exp.drag_hpi_pulse)

repeat_drag_hpi = exp.repeat_sequence(exp.drag_hpi_pulse)

7. Calibrate the DRAG pi pulse¶

Next, calibrate the DRAG pi pulse used in the benchmarking gate set.

In [ ]:

drag_pi_result = exp.calibrate_drag_pi_pulse()

drag_pi_result = exp.calibrate_drag_pi_pulse()

8. Repeat the DRAG pi pulse¶

Check the calibrated pi pulse with the same repeated-sequence workflow.

In [ ]:

repeat_drag_pi = exp.repeat_sequence(exp.drag_pi_pulse)

repeat_drag_pi = exp.repeat_sequence(exp.drag_pi_pulse)

9. Save the calibration note¶

Save the calibrated pulses before you move on to the benchmarking measurements.

In [ ]:

exp.calib_note.save()

exp.calib_note.save()

10. Inspect the calibrated pulses¶

Plot the DRAG half-pi and pi pulses for the target qubit so you can inspect the cached waveforms directly.

In [ ]:

Q0 = exp.qubit_labels[0]

print("target qubit:", Q0)

Q0 = exp.qubit_labels[0] print("target qubit:", Q0)

In [ ]:

drag_hpi = exp.drag_hpi_pulse[Q0]
drag_hpi.plot()

drag_pi = exp.drag_pi_pulse[Q0]
drag_pi.plot()

drag_hpi = exp.drag_hpi_pulse[Q0] drag_hpi.plot() drag_pi = exp.drag_pi_pulse[Q0] drag_pi.plot()

11. Run standard randomized benchmarking¶

Use the calibrated single-qubit gate set to estimate the average Clifford error.

In [ ]:

rb_result = exp.randomized_benchmarking(
    Q0,
    n_cliffords_range=np.arange(0, 1001, 100),
    n_trials=30,
    save_image=True,
)

rb_result = exp.randomized_benchmarking( Q0, n_cliffords_range=np.arange(0, 1001, 100), n_trials=30, save_image=True, )

12. Run interleaved RB for X90¶

Interleave X90 with the RB sequence to estimate the contribution of that gate.

In [ ]:

irb_x90 = exp.interleaved_randomized_benchmarking(
    Q0,
    interleaved_clifford="X90",
    n_cliffords_range=range(0, 1001, 100),
    n_trials=30,
    save_image=True,
)

irb_x90 = exp.interleaved_randomized_benchmarking( Q0, interleaved_clifford="X90", n_cliffords_range=range(0, 1001, 100), n_trials=30, save_image=True, )

13. Run interleaved RB for X180¶

Repeat the interleaved analysis for X180.

In [ ]:

irb_x180 = exp.interleaved_randomized_benchmarking(
    Q0,
    interleaved_clifford="X180",
    n_cliffords_range=range(0, 1001, 100),
    n_trials=30,
    save_image=True,
)

irb_x180 = exp.interleaved_randomized_benchmarking( Q0, interleaved_clifford="X180", n_cliffords_range=range(0, 1001, 100), n_trials=30, save_image=True, )