Implement frequency changes for pulse

[eggerdj]

What is the expected enhancement?

PulseChannels create the control fields Re[exp(i 2pi j dt f + phi)d_j]. There should be a Command to change the frequency f of the PulseChannel.

This instruction will be implemented by class SetChannelFrequency(Command). It will have

• one attribute, the new frequency of the channel, i.e. self._frequency,
• the prefix scf,
• and will result in the command_dict {'name': 'scf', 't0': ..., 'ch': ..., 'frequency': ... }

[taalexander]

I think this is an excellent idea! Only recommendation I might have is to call this SetFrequency since pulse commands are implicitly on channels.

Are you planning on implementing @eggerdj?

[eggerdj]

@taalexander Yes I'll implement this. Please assign me.

SetFrequency is also an option and it makes sense given that Commands are applied on channels. My reason for making channel explicit is that for flux-tunable qubits I can imagine a command named SetQubitFrequency. I.e. to avoid any ambiguity I would prefer to have SetChannelFrequency.

[taalexander]

Yes, I can see that being the case. One option would be accept both types into the command in that case. Any thoughts @ajavadia on instruction operand type overloading? I know this is how LLVM does it.

[eggerdj]

I guess accepting both types is an option. I worry that might make Pulse a little confusing for the end user?

[taalexander]

I agree it could, it would help if there was a formal language definition associated with pulse, eg., The equivalent of QASM in which typing in explicit.

[eggerdj]

The WIP PR outlines what this would look like. Changing the name of the command is easy. Let me know if I forgot anything here.

[lcapelluto]

SetQubitFrequency would not be the best approach. The qubit frequency isn't something that can be set directly, but rather something that is controlled via other channels (the flux bias line, in that case). Pulse inherently always operates on channels, and cannot operate on qubits, so I think SetFrequency is still clear.

[eggerdj]

Thinking about this a bit more I am now comfortable with SetFrequency. Changing the qubit frequency for a fixed duration T would require a square pulse (implemented as ConstantPulse(ParametricPulse) or SamplePulse) of duration T on the flux bias line. (If the frequency change is to be until the end of the schedule then a persistent value would probably better). At this point its up to the user to know that his square pulse will change the frequency of the qubit.

[taalexander]

Yes, I agree with this. Frequency should only reference the LO of a given drive channel.