# Cluster

In this section we introduce how to configure the cluster with Qblox backend, and which options are available in quantify. For information about the lower-level functionalities (qblox-instruments level) of the cluster see Clusters. If you are not familiar with how to compile hardware configuration in general, see Compiling to Hardware.

To use the Qblox backend, "quantify_scheduler.backends.qblox_backend.hardware_compile" has to be used as a "backend" in the hardware mapping configuration.

## General hardware mapping structure, example

We start by looking at an example config for a single cluster. The hardware configuration specifies which outputs are used, clock frequency properties, gains and attenuations among other properties. The general structure is that the cluster has multiple modules, and each module can use multiple portclocks.

 1mapping_config = {
2    "backend": "quantify_scheduler.backends.qblox_backend.hardware_compile",
3    "cluster0": {
4        "instrument_type": "Cluster",
5        "ref": "internal",
6        "cluster0_module1": {
7            "instrument_type": "QCM",
8            "complex_output_0": {
9                "lo_name": "lo0",
10                "portclock_configs": [
11                    {
12                        "clock": "q4.01",
13                        "interm_freq": 200000000.0,
14                        "mixer_amp_ratio": 0.9999,
15                        "mixer_phase_error_deg": -4.2,
16                        "port": "q4:mw",
17                    },
18                ]
19            },
20        },
21        "cluster0_module2": {
22            "instrument_type": "QCM_RF",
23            "complex_output_0": {
24                "portclock_configs": [
25                    {
26                        "clock": "q5.01",
27                        "interm_freq": 50000000.0,
28                        "port": "q5:mw"
29                    }
30                ]
31            },
32        },
33    },
34    "lo0": {"instrument_type": "LocalOscillator", "frequency": None, "power": 20},
35}


Notice the "quantify_scheduler.backends.qblox_backend.hardware_compile" backend is used. In the example, we notice that the cluster is specified using an instrument with "instrument_type": "Cluster". In the backend, the cluster instrument functions as a collection of modules. The modules themselves can be configured with portclock_configs.

Also notice, that not only a cluster, but a local oscillator can also be configured with Qblox. Currently the only instrument types that can be at the top level are:

• "Cluster"

• "LocalOscillator"

• pulsars

## Cluster configuration

The cluster configuration must be at top level, and its "instrument_type" must be "Cluster". The name of the cluster (the key of the structure, "cluster0" in the example) can be chosen freely.

It has only one required key "ref", which can be "internal" or "external". This sets the reference source, which is a 10 MHz clock source.

To add a new module mapping to the cluster, add a new key with a valid "instrument_type".

### Write sequencer program to files

It is possible to optionally set "sequence_to_file" key to True or False. If it’s not set Quantify will behave the same way as if it was set to True. If it is True, a file will be created for each sequencer with the program that’s uploaded to the sequencer with the filename <data_dir>/schedules/<year><month><day>-<hour><minute><seconds>-<milliseconds>-<random>_<port>_<clock>.json in a JSON format, where <random> is 6 random characters in the range 0-9, a-f.

It is possible to overwrite this parameter to "True" in each module configuration for each module.

{
"backend": "quantify_scheduler.backends.qblox_backend.hardware_compile",
"cluster0": {
"instrument_type": "Cluster",
"ref": "internal",
"sequence_to_file": True,
"module0": {...},
"module1": {...},
...
}
}


## Module configuration

For each module configuration the key must be "<cluster_name>_module<n>", where <n> is the module number in the cluster. "instrument_type" is mandatory, and can be one of

• "QCM",

• "QRM",

• "QCM_RF",

• "QRM_RF".

Apart from the "instrument_type", the only possible key in the module configuration for a cluster are the inputs/outputs. The possible inputs/outputs are

• for "QCM": "complex_output_{0,1}", "real_output_{0,1,2,3}",

• for "QRM": "complex_{output,input}_0", "real_{output,input}_{0,1}".

• for "QCM_RF": "complex_output_{0,1}",

• for "QRM_RF": "complex_{output,input}_0".

Note, for RF hardware, if an output is unused, it will be turned off. (This is to ensure that unused local oscillators do not interfere with used outputs.)

### Real mode

To use real mode, the output/input name must start with "real_". When using real outputs, the backend automatically maps the signals to the correct output paths. We note that for real outputs, it is not allowed to use any pulses that have an imaginary component i.e. only real valued pulses are allowed. If you were to use a complex pulse, the backend will produce an error, e.g. square and ramp pulses are allowed but DRAG pulses not.

Warning

When using real mode, we highly recommend using the cluster in combination with the InstrumentCoordinator as the outputs need to be configured correctly in order for this to function.

"qcm0": {
"instrument_type": "QCM",
"ref": "internal",
"real_output_0": {
"portclock_configs": [
{
"port": "q0:mw",
"clock": "q0.01",
}
]
},
"real_output_1": {
"portclock_configs": [
{
"port": "q1:mw",
"clock": "q1.01",
}
]
},
"real_output_2": {
"portclock_configs": [
{
"port": "q2:mw",
"clock": "q2.01",
}
]
}
},


### Marker configuration

Currently it is not possible to set marker configuration by the user on the modules. For all module types, the markers are set to all high at the start of the schedule.

### Mixer corrections

The backend also supports setting the parameters that are used by the hardware to correct for mixer imperfections in real-time.

We configure this by adding "dc_mixer_offset_I" and/or "dc_mixer_offset_Q" to outputs, like the following example.

"complex_output_0": {
"dc_mixer_offset_I": -0.054,
"dc_mixer_offset_Q": -0.034,
...
}


And you can also add "mixer_amp_ratio" and "mixer_phase_error_deg" to a specific portclock in order to set the amplitude and phase correction to correct for imperfect rejection of the unwanted sideband. See the following example.

"complex_output_0": {
...
"portclock_configs": [
{
"port": <port>,
"clock": <clock>,
"mixer_amp_ratio": 0.9997,
"mixer_phase_error_deg": -4.0,
...
}
]
}


### Gain and attenuation

For QRM, QRM-RF and QCM-RF modules you can set the gain and attenuation parameters in dB.

#### Gain configuration

• The parameters "input_gain_I/0" and input_gain_Q/1 for QRM correspond to the qcodes parameters in0_gain and in1_gain respectively.

Note, these parameters only affect the QRM modules. For complex inputs you have to use "input_gain_I" and "input_gain_Q", and for real inputs "input_gain_0" and "input_gain_1".

...
"cluster0_module1": {
"instrument_type": "QRM",
"complex_input_0": {
"input_gain_I": 2,
"input_gain_Q": 3,
...
},
},
"cluster0_module2": {
"instrument_type": "QRM",
"real_input_0": {
"input_gain_0": 2,
...
},
"real_input_1": {
"input_gain_1": 3,
...
},
},


#### Attenuation configuration

• The parameter "complex_output_*"."output_att" and "complex_input_0.input_att" for QRM-RF correspond to the qcodes parameters out0_att and in0_att respectively.

• The parameter "complex_output_*"."output_att" for QCM-RF correspond to the qcodes parameters out0_att and out1_att.

Note, that these parameters only affect RF modules.

...
"cluster0_module1": {
"instrument_type": "QRM_RF",
"complex_output_0": {
"output_att": 12,
...
},
"complex_input_0": {
"input_att": 10,
...
}
},
"cluster0_module2": {
"instrument_type": "QCM_RF",
"complex_output_0": {
"output_att": 4,
...
},
"complex_output_1": {
"output_att": 6,
...
},
},


See Qblox Instruments: QCM-QRM documentation for allowed values.

### Maximum AWG output voltage

Note

This subsection on max_awg_output_voltage is still under construction.

### Clock settings

The aim of quantify-scheduler is to only specify the final RF frequency when the signal arrives at the chip, rather than any parameters related to I/Q modulation. However, you still need to provide some parameters for the up/downconversion.

The backend assumes that upconversion happens according to the relation

$f_{RF} = f_{IF} + f_{LO}$

You can specify $$f_{RF}$$ in multiple ways. You can specify it when you add a ClockResource with freq argument to your Schedule, or when you specify the BasicTransmonElement.clock_freqs.

Note

If you use gate level operations, you have to follow strict rules for the naming of the clock resource, for each kind of operation:

• "<transmon name>.01" for Rxy operation (and its derived operations),

• "<transmon name>.ro" for any measure operation,

• "<transmon name>.12" for the $$|1\rangle \rightarrow |2\rangle$$ transition.

Then,

• for baseband modules, you can optionally specify a local oscillator by its name using the "lo_name" key. If you specify it, the "frequency" key in the local oscillator specification (see the example below) specifies $$f_{LO}$$ of this local oscillator. Otherwise, $$f_{LO} = 0$$ and $$f_{RF} = f_{IF}$$. $$f_{RF} = f_{IF}$$ can also be set in the hardware mapping explicitly with the "interm_freq" key in the portclock configuration.

• For RF modules, you can specify $$f_{IF}$$ inside each portclock configuration in the hardware mapping for each portclock with the "interm_freq" key, and/or you can specify the local oscillator for each output with the "lo_freq", because they have internal local oscillators. Note, if you specify both, the relationship between these frequencies should hold, otherwise you get an error message. It’s important to note, that fast frequency sweeps only work when $$f_{LO}$$ is fixed, and $$f_{IF}$$ is unspecified. Because of this, it is generally advised to specify $$f_{LO}$$ only.

In the following example for the baseband modules "complex_output_0"’s $$f_{IF}$$ is the same as the "q0.01" clock resource’s frequency, and "complex_output_1"’s $$f_{IF}$$ is calculated using the frequency of "lo1" and "q1.01" For the RF modules, "complex_output_0"’s $$f_{IF}$$ is calculated using the provided "lo_freq" and the frequency of "q2.01", and for "complex_output_1", it’s $$f_{LO}$$ is calculated using the provided "interm_freq" and the frequency of "q3.01".

mapping_config = {
"backend": "quantify_scheduler.backends.qblox_backend.hardware_compile",
"cluster0": {
"instrument_type": "Cluster",
"ref": "internal",
"cluster0_module0": {
"instrument_type": "QCM",
"complex_output_0": {
"portclock_configs": [
{
"clock": "q0.01",
"port": "q0:mw"
}
]
},
"complex_output_1": {
"lo_name": "lo1",
"portclock_configs": [
{
"clock": "q1.01",
"port": "q1:mw"
}
]
},
},
"cluster0_module1": {
"instrument_type": "QCM_RF",
"complex_output_0": {
"lo_freq": 7e9,
"portclock_configs": [
{
"clock": "q2.01",
"port": "q2:mw"
}
]
},
"complex_output_1": {
"portclock_configs": [
{
"clock": "q3.01",
"interm_freq": 50000000.0,
"port": "q3:mw"
}
]
},
},
},
"lo1": {"instrument_type": "LocalOscillator", "frequency": 5e9, "power": 20},
}

test_sched = Schedule("test_sched")
test_sched.add_resource(ClockResource(name="q0.01", freq=8e9))
test_sched.add_resource(ClockResource(name="q1.01", freq=9e9))
test_sched.add_resource(ClockResource(name="q2.01", freq=8e9))
test_sched.add_resource(ClockResource(name="q3.01", freq=9e9))

test_sched.add(SquarePulse(amp=1, duration=1e-6, port="q0:mw", clock="q0.01"))
test_sched.add(SquarePulse(amp=0.25, duration=1e-6, port="q1:mw", clock="q1.01"))
test_sched.add(SquarePulse(amp=0.25, duration=1e-6, port="q2:mw", clock="q2.01"))
test_sched.add(SquarePulse(amp=0.25, duration=1e-6, port="q3:mw", clock="q3.01"))
test_sched = determine_absolute_timing(test_sched)
hardware_compile(test_sched, mapping_config)


#### Downconverter

Note

This section is only relevant for users with custom qblox downconverter hardware.

Some users may have a custom Qblox downconverter module operating at 4.4 GHz. In order to use it with this backend, we should specify a "downconverter_freq" entry in the outputs that are connected to this module, as exemplified below. The result is that the downconversion stage will be taken into account when calculating the IF or LO frequency (whichever was undefined) during compilation, such that the signal reaching the target port is at the desired clock frequency.

For baseband modules, downconversion will not happen if "mix_lo" is not True and there is no external LO specified. ("mix_lo" is True by default.) For RF modules, "mix_lo" has no effect.

 1mapping_config = {
2    "backend": "quantify_scheduler.backends.qblox_backend.hardware_compile",
3    "cluster0": {
4      "cluster0_module0": {
5          "instrument_type": "QCM",
6          "ref": "internal",
7          "complex_output_0": {
8              "downconverter_freq": 9000000000,
9              "mix_lo": True,
10              "portclock_configs": [
11                  {
12                      "port": "q0:mw",
13                      "clock": "q0.01",
14                      "interm_freq": 50000000.0
15                  }
16              ]
17          }
18       }
19    },
20    "cluster0_module1": {
21          "instrument_type": "QCM_RF",
22          "ref": "internal",
23          "complex_output_0": {
24              "downconverter_freq": 9000000000,
25              "portclock_configs": [
26                  {
27                      "port": "q0:mw",
28                      "clock": "q0.01",
29                      "interm_freq": 50000000.0
30                  }
31              ]
32          }
33       }
34    }
35}
36hardware_compile(test_sched, mapping_config)


### Portclock configuration

Each module can have at most 6 portclocks defined, and the name for each "port" and "clock" combination must be unique. Each of these portclocks is associated with one sequencer in the Qblox hardware.

Note

If you use gate level operations, you have to follow strict rules for each kind of operation on which port name you can use (what’s the naming convention for each port resource).

• "<device element name>:mw" for Rxy operation (and its derived operations),

• "<device element name>:res" for any measure operation,

• "<device element name>:fl" for the flux port.

The only required keys are the "port" and "clock" which are needed to be defined. The following parameters are available.

• "interm_freq" defines the $$f_{IF}$$, see Clock settings,

• "mixer_amp_ratio" by default 1.0, must be between 0.5 and 2.0, see Mixer corrections,

• "mixer_phase_error_deg" by default 0.0, must be between -45 and 45, Mixer corrections,

• "ttl_acq_threshold",

• "init_offset_awg_path_0" by default 0.0, must be between -1.0 and 1.0,

• "init_offset_awg_path_1" by default 0.0, must be between -1.0 and 1.0,

• "init_gain_awg_path_0" by default 1.0, must be between -1.0 and 1.0,

• "init_gain_awg_path_1" by default 1.0, must be between -1.0 and 1.0,

• "qasm_hook_func", see QASM hook,

• "instruction_generated_pulses_enabled", see Instruction generated pulses.

Note

We note that it is a requirement of the backend that each combination of a port and a clock is unique, i.e. it is possible to use the same port or clock multiple times in the hardware config but the combination of a port with a certain clock can only occur once.

#### QASM hook

It is possible to inject custom qasm instructions for each portclock (sequencer) after the compiler inserts the footer and the stop instruction in the generated qasm program. See the following example to insert a NOP (no operation) at the end of the program.

def _func_for_hook_test(qasm: QASMProgram):
qasm.instructions.insert(
0, QASMProgram.get_instruction_as_list(q1asm_instructions.NOP)
)

hw_config = {
"backend": "quantify_scheduler.backends.qblox_backend.hardware_compile",
"cluster0_module1": {
"instrument_type": "QCM_RF",
"ref": "internal",
"complex_output_0": {
"downconverter_freq": 9000000000,
"portclock_configs": [
{
"port": "q0:mw",
"clock": "q0.01",
"qasm_hook_func": _func_for_hook_test,
}
]
}
}
}


#### Instruction generated pulses

The Qblox backend contains some intelligence that allows it to generate certain specific waveforms from the pulse library using a more complicated series of sequencer instructions, which helps conserve waveform memory. Though in order to keep the backend fully transparent, all such advanced capabilities are disabled by default.

In order to enable the advanced capabilities we need to add line "instruction_generated_pulses_enabled": True to the port-clock configuration.

hw_config = {
"backend": "quantify_scheduler.backends.qblox_backend.hardware_compile",
"cluster0_module1": {
"instrument_type": "QCM_RF",
"ref": "internal",
"complex_output_0": {
"downconverter_freq": 9000000000,
"portclock_configs": [
{
"port": "q0:mw",
"clock": "q0.01",
"instruction_generated_pulses_enabled": True,
}
]
}
}
}


Currently, this has the following effects:

• Long square pulses get broken up into separate pulses with durations <= 1 us, which allows the modules to play square pulses longer than the waveform memory normally allows.

• Staircase pulses are generated using offset instructions instead of using waveform memory

## Local Oscillator configuration

Local oscillator instrument can be added and then used for baseband modules. You can then reference the local oscillator instrument at the output with "lo_name".

The three mandatory parameters are the "instrument_type" (which should be "LocalOscillator"), and "frequency" in Hz or None, and "power".

It is also possible to add "generic_icc_name" as an optional parameter, but only "generic" is supported currently with the Qblox backend.

"backend": "quantify_scheduler.backends.qblox_backend.hardware_compile",
"cluster0": {
"instrument_type": "Cluster",
"ref": "internal",
"cluster0_module0": {
"instrument_type": "QCM",
"complex_output_1": {
"lo_name": "lo1",
"portclock_configs": [
{
"clock": "q1.01",
"port": "q1:mw"
}
]
},
},
},
"lo1": {"instrument_type": "LocalOscillator", "frequency": 5e9, "power": 20},


## Latency corrections

Latency corrections is a dict containing the delays for each port-clock combination. It is possible to specify them under the key "latency_corrections" in the hardware config, at the top-level. See the following example.

"latency_corrections": {
"q4:mw-q4.01": 8e-9,
"q5:mw-q5.01": 4e-9
}


Each correction is in nanoseconds. For each specified port-clock, the program start will be delayed by this amount of time. Note, the delay still has to be a multiple of the grid time.

## Distortion corrections

Distortion corrections apply a function on the pulses which are in the schedule. Note, that this will not be applied to outputs generated by modifying the offset and gain/attenuation. The "distortion_corrections" is an optional key in the hardware config, at the top-level. See the following example.

"distortion_corrections": {
"q0:fl-cl0.baseband": {
"filter_func": "scipy.signal.lfilter",
"input_var_name": "x",
"kwargs": {
"b": [0.0, 0.5, 1.0],
"a": [1]
},
"clipping_values": [-2.5, 2.5]
}
}


If "distortion_corrections" are set, then "filter_func", "input_var_name" and "kwargs" are required. If "clipping_values" are set, its value must be a list with exactly 2 floats.

Clipping values are the boundaries to which the corrected pulses will be clipped, upon exceeding, these are optional to supply.

The "filter_func" is a python function that we apply with "kwargs" arguments. The waveform to be modified will be passed to this function in the argument name specified by "input_var_name". The waveform will be passed as a np.ndarray.