While #1558 should provide the most stable experiment to fit the full conditional phase and leakage landscape for the SNZ pulse, the problem is that it is also going to be quite expensive.
Estimate
E.g. for a limited number of pixels, we estimate that the protocol may take ~1h on Qblox, because of the many executions required due to the 3 nested sweepers.
This back-of-the-envelope calculation could be based on experiments dominated by a $300~\mu s$ reset through relaxation, considering 50 amplitude points, 20 values for the shape ($B/A$ ratio), and 10 phases for the conditional phase fit. If we add ~7s of reset time between different executions for the shape, this would be:
(
3e-4 * # relaxation
1e3 * # shots
50 * # amplitudes
10 + # phases
7 # experiment overhead (cluster reset)
) * 20 # shapes
This amounts to 50 minutes of experiments, and couple of minutes just going into the reset.
Proposal
Instead, we could compute the relative phase with just two measurements, which will be the X and Y coordinates (expectation values) of the point rotated by the gate.
We will measure the same phase, with the backdraw of being more noise-prone, resulting in a less accurate measurement per-pixel.
However, this would be a speed-up of a factor 5 over the 10 phases considered above. And still a boost of 2-3, even if we assumed we could fit a sine with 4-6 points strategically distributed (which, most likely, it means evenly spaced).
While #1558 should provide the most stable experiment to fit the full conditional phase and leakage landscape for the SNZ pulse, the problem is that it is also going to be quite expensive.
Estimate
E.g. for a limited number of pixels, we estimate that the protocol may take ~1h on Qblox, because of the many executions required due to the 3 nested sweepers.$300~\mu s$ reset through relaxation, considering 50 amplitude points, 20 values for the shape ($B/A$ ratio), and 10 phases for the conditional phase fit. If we add ~7s of reset time between different executions for the shape, this would be:
This back-of-the-envelope calculation could be based on experiments dominated by a
This amounts to 50 minutes of experiments, and couple of minutes just going into the reset.
Proposal
Instead, we could compute the relative phase with just two measurements, which will be the X and Y coordinates (expectation values) of the point rotated by the gate.
We will measure the same phase, with the backdraw of being more noise-prone, resulting in a less accurate measurement per-pixel.
However, this would be a speed-up of a factor 5 over the 10 phases considered above. And still a boost of 2-3, even if we assumed we could fit a sine with 4-6 points strategically distributed (which, most likely, it means evenly spaced).