# Posts tagged SPAM

## Improving state prep errors on IBM Quantum systems

It is often taken for granted that qubits are initialized to zero when executing quantum circuits on hardware. However, errors in the reset process mean that this initialization is not perfect leading to state preparation errors. Here we will look at state prep errors and how to improve them on IBM Quantum systems.

Here we will use two lesser known configuration variables available on IBM Quantum systems to do so. First is rep_delay. This is the delay time before the next circuit is executed. Each backend has a different default_rep_delay (although as of the time of writing many are $$250~\mu \rm s$$ or lower) and support tuning this parameter from $$0$$ up to $$500~\mu \rm s$$. Why in the world would I want to vary this parameter? Well if I set rep_delay=0 then I can get much faster throughput as the default delay is nominally longer than the $$T_{1}$$ time of the system. However when doing so you are trusting the implicit reset (called “unconditional reset”) to do its job and reset the qubits to their ground state with high-fidelity. Failure to do so would lead to state-prep errors that propagate through the circuit. This is why the delay is usually longer than $$T_{1}$$; let the enviornment help your reset fidelity. Second is init_qubits that tells the hardware to not perform an unconditional reset. As no reset at all would quickly lead to junk, this setting needs to be done in concert with reset instructions being added to the beginning of your circuits. Perhaps this is better, or maybe a single reset at the beginning and the uncondtional reset. Lets find out.