qiboteam / qibolab

Quantum hardware module and drivers for Qibo.
https://qibo.science
Apache License 2.0
42 stars 14 forks source link

Active reset #592

Open alecandido opened 1 year ago

alecandido commented 1 year ago

To speed up the execution of independent experiments, the qubits could be reset to the ground state without waiting this to happen naturally after the relaxation time (which is an increasingly worse strategy as much as the relaxation time improves, which is generally a good thing).

@Jacfomg tried to implement this for Zurich Instruments in #506, but it is definitely possible even for Quantum Machines https://www.quantum-machines.co/faq/how-can-i-do-an-active-reset and in general the algorithm is pretty simple, so it should be possible with arbitrary electronics.

I wonder whether this could be implemented in hardware-agnostic way or not (relying on the readout and pi-pulse), but for sure is a useful feature to have.

alecandido commented 3 days ago

Apparently, Qblox supports conditonal pulses across different modules (and clusters) as well through its LINQ protocol, with a guaranteed delay of at most 320ns.

4. Scalable Feedback with LINQ

Our proprietary LINQ protocol allows for low-latency feedback with an arbitrary control flow by distributing measurement outcomes to all modules in less than 320 ns. This all-to-all connectivity allows for applying conditionally generated pulses based on the result of a measurement from any other module. https://docs.qblox.com/en/main/cluster/cluster.html#scalable-feedback-with-linq

It would be interesting to know which are the limitations of the other drivers.

(Just to make it clear, even to my future self, it is mandatory to use different modules, since the acquisition will be performed by the QRM-RF, while control pulses, even on the same qubit, will be created by the QCM or QCM-RF modules - which are necessarily distinct)

This stresses the dependence of this on #917, since we can not provide active reset with a deterministic timing (not within the ns resolution), so it's falling quite beyond what a PulseSequence can currently represent (other than the measurement dependence itself).