Backend: Zurich Instruments


quantify_scheduler provides a stateless module: zhinst_backend, that abstracts the complexity of setting up experiments using Zurich Instruments hardware. quantify_scheduler uses information on the quantum device and instrument properties to compile a quantify_scheduler.schedules.schedule.Schedule into waveforms and sequencing instructions suitable for execution on Zurich Instruments hardware. More information about compilation can be found in the User Guide.

Using existing programming interfaces provided via zhinst-qcodes and zhinst-toolkit, quantify_scheduler prepares the instruments that are present in the sec-hardware configuration file.

Finally, after configuring and running compile_backend() successfully the instruments are prepared for execution.

The Zurich Instruments backend provides:

  • Synchronized execution of a program in a UHFQA and an HDAWG through the use of Triggers and Markers.

  • Automatic generation of the SeqC Sequencer instructions.

  • Waveform generation and modulation.

  • Memory-efficient Sequencing with the CommandTable.

  • Configuration of the relevant settings on all instruments.

  • Configuration for Triggers and Markers.

Supported Instruments

The Zurich Instruments backend currently supports the HDAWG and the UHFQA instruments. In addition to the Zurich Instruments devices, the hardware backend supports using microwave sources such as the R&S SGS100A.

Basic paradigm and inner workings

The Zurich Instruments back end threats a Schedule as a linear timeline of operations executed on the quantum device and translates this into pulses and acquisitions to be executed on the different input and output channels of an HDAWG and a UHFQA. In this context a single HDAWG is treated as the master device that sends out a single marker pulse for synchronization at the start of each iteration of the schedule that is used to trigger all other devices. After the synchronization trigger is given, all devices execute a compiled program for which the timings of all instructions have been calculated.

The compilation from operations at the quantum-device layer to instructions that the hardware can execute is done in several steps. The zhinst compile_backend() starts from the ScheduleBase.timing_table and maps the operations to channels on the hardware using information specified in the sec-hardware configuration file. Corrections for channel latency, as well as moving operations around to ensure all measurements start at the first sample (0) of a clock cycle are also done at this stage.

Once the starting time and sample of each operation is known, the numerical waveforms that have to be uploaded to the hardware can be generated. The numerical waveforms differ from the idealized waveforms of the device description in that the they include corrections for effects such as mixer-skewness and linear-dynamical distortions (not implemented yet), and intermediate frequency modulation if required.

Both the CompiledSchedule.hardware_timing_table and CompiledSchedule.hardware_waveform_dict are available as properties of the CompiledSchedule.

In the next step, the clock-accurate Seqc instructions for each awg of each device are determined, as well as the settings (nodes) to be configured including the linking of the numerical waveforms to these nodes. All of this information is combined in ZISettingsBuilders and added to the compiled instructions of the CompiledSchedule.


There are several limitation to the paradigm and to the current implementation. Some of these are relatively easy to address while others are more fundamental to the paradigm. Here we give an overview of the known limitations. Note that some of these can be quite specific.

Inherent limitations

There are some inherent limitations to the paradigm of describing the program as a single linear timeline that is started using a single synchronization trigger. These limitations cannot easily be addressed so should be taken into account when thinking about experiments.

  • Because the synchronization of the HDAWG and the UFHQA relies on a trigger on two devices operating at different clock frequencies one cannot guarantee at what sample within the clock domain the slave device gets triggered. The consequence is that although the triggering is stable within an experiment, the exact time difference (in number of samples) between the different devices varies between different experiments. This problem is inherent to the triggering scheme and cannot be easily resolved.

  • The paradigm of a single fixed timeline with a single synchronizing trigger is not compatible with control loop affecting feedback.

Limitations with the current implementation

There are also some practical limitations to the implementation. Keep these in mind when operating the hardware.

  • Real-time modulation is currently not supported, relying on pre-modulated waveforms, it is important to start waveforms at a multiple of the modulation frequency. Sticking to a 10 ns grid, it is recommended to use a modulation frequency of 100 MHz.

  • All operations need to start at an integer number of samples. Because of the choice of sampling rates of 2.4 GSps (~0.416 ns) and 1.8 GSps (~0.555 ns) it is useful to stick to a 10 ns grid for HDAWG (microwave and flux) pulses and a 40 ns grid for UHFQA (readout) pulses.

  • Different instructions on the same “awg” cannot start in the same clock cycle. This implies that the readout acquisition delay cannot be 0 (but it can be 40 ns or - 40ns).

  • All measurements are triggered simultaneously using the StartQA(QA_INT_ALL, true) instruction. This implies it is not possible to read out only a specific qubit/channel.

  • All measurements have to start at the same sample within a clock cycle because one can only define a single integration weight per channel. To guarantee this, all operations are shifted around a bit (the measurement fixpoint correction). As a consequence, the reset/initialization operation can sometimes be a bit longer than specified in the schedule.

  • Because the timing between two devices needs to align over longer schedules, it is important that the clock-rates are accurate. To ensure phase stability, use a 10 MHz shared reference and operate the hardware in external reference mode.

  • Only a single HDAWG supported as the primary device, other HDAWGs need to be configured as secondary devices.

  • Multiplexed readout is currently not supported. One can only read out a single channel. (#191)

Hardware configuration


This section will move to the documentation of the hardware configuration file themselves. See issue #222.

The zhinst_backend allows Zurich Instruments to be configured individually or collectively by enabling master/slave configurations via Triggers and Markers.

Instruments can be configured by adding them to the hardware configuration file. The configuration file contains parameters about the Instruments and properties required to map quantify_scheduler.operations.operation.Operations, which act on qubits, onto physical properties of the instrument.

The Zurich Instruments hardware configuration file is divided in four main sections.

1. The backend property defines the python method which will be executed by qcompile() in order to compile the backend.

  1. The local_oscillators property is a list of dicts which describe the available local oscillators in the hardware setup. An example entry are as follows:

 2  "backend": "quantify_scheduler.backends.zhinst_backend.compile_backend",
 3  "local_oscillators": [
 4    {
 5      "unique_name": "mw_qubit_ch1",
 6      "instrument_name": "mw_qubit",
 7      "frequency":
 8          {
 9              "ch_1.frequency": null
10          },
11      "power":
12          {
13              "power": 13
14          },
15      "phase":
16          {
17              "ch_1.phase": 90
18          }
19    }
20  ]
  • In the example, the particular local_oscillator is given a unique_name which is then used in the devices to couple the channel of the device to that local oscillator.

  • The instrument_name is the QCoDes Instrument name for the local oscillator object.

  • The frequency property maps the frequency parameter which is used to set the frequency of the local oscillator. If set to null, then the local oscillator frequency is automatically calculated from the relation, LO frequency = RF frequency - Intermodulation frequency.

  • The power property is an optional key which maps the power parameter used to set the power of the local oscillator. If the key is not provided, no value will be set. Note that the units are based on the instruments used (i.e. if the QCoDeS instrument sets the power in dbm, then the power value should be in dbm).

  • The phase property is an optional key that maps the phase parameter used to set the phase of the local oscillator signal. If the key is not provided, no value will be set. Note that the units are based on the instruments used (i.e. if the QCoDeS instrument sets the phase in radians, then the phase value should be in radians).

  1. The latency_corrections property specifies a delay on a port-clock combination which is implemented by incrementing the abs_time of all operations applied to the port-clock combination. The delay is used to manually adjust the timing to correct for e.g., delays due to different cable lengths.

 2  "backend": "quantify_scheduler.backends.zhinst_backend.compile_backend",
 3  "latency_corrections":
 4      {
 5          "q0:mw-q0.01": 95e-9,
 6          "q1:mw-q1.01": 95e-9
 7          "": -95e-9,
 8          "": -95e-9,
10      }

In this example, the user has specified latencies at each port-clock combination in the hardware config. The relative latencies to the minimum of the latencies are then calculated and applied to the signals for the specific port-clock combination. For this case, pulses with port=q0:res, and clock=q0.r0 i.e. “” will have no corrections, whilst, pulses with port=q0:mw, and clock=q0.01 i.e. “q0:mw-q0.01” will have a corrected delay of 190e-9 s added to the pulses.

4. The devices property is an array of Device. A Device describes the type of Zurich Instruments and the physical setup.

2  "backend": "quantify_scheduler.backends.zhinst_backend.compile_backend",
3  "devices": [
5  ]

The entries in the devices section of the configuration file are strictly mapped according to the Device and Output domain models.

  • In order for the backend to find the QCodes Instrument it is required that the Device’s name must be equal to the name given to the QCodes Instrument during instantiation with an ic prepend.

    • Example: If the hdawg QCodes Instrument name is “hdawg_dev8831” then the Device’s name is “ic_hdawg_dev8831”

  • The type property defines the instrument’s model. The DeviceType is parsed from the string as well as the number of channels.

    • Example: “HDAWG8”

  • The ref property describes if the instrument uses Markers (int), Triggers (ext) or none.

    • int Enables sending Marker

    • ext Enables waiting for Marker

    • none Ignores waiting for Marker

  • The channelgrouping property sets the HDAWG channel grouping value and impacts the amount of HDAWG channels per AWG must be used.

 2  "backend": "quantify_scheduler.backends.zhinst_backend.compile_backend",
 3  "devices": [
 4    {
 5      "name": "hdawg0",
 6      "ref": "int",
 7      "channelgrouping": 0,
 8      "channel_0": {
 9        ...
10      },
11    },
12  ]
17instrument = zhinst.qcodes.HDAWG(name='hdawg0', serial='dev1234', ...)
  • The channel_{0..3} properties of the hardware configuration are mapped to the Output domain model. A single channel represents a complex output, consisting of two physical I/O channels on the Instrument.

  • The port and clock properties map the Operations to physical and frequency space.

  • The mode property specifies the channel mode: real or complex.

  • The modulation property specifies if the uploaded waveforms are modulated. The backend supports:

    • Premodulation “premod”

    • No modulation “none”

  • The interm_freq property specifies the inter-modulation frequency.

  • The markers property specifies which markers to trigger on each sequencer iteration. The values are used as input for the setTrigger sequencer instruction.

  • The triggers property specifies for a sequencer which digital trigger to wait for. The first value of the triggers array is used as input parameter for the waitDigTrigger sequencer instruction.

class Output(port, clock, mode, modulation, local_oscillator, clock_frequency=None, gain1=0, gain2=0, trigger=None, markers=<factory>, mixer_corrections=None)[source]

The definition class for zhinst channel properties.

This class maps to the zhinst backend JSON “channel” properties defined in the hardware mapping.

  • port (str) – The port resource.

  • clock (str) – The Clock resource.

  • clock_frequency (Optional[float]) – The frequency for the clock resource (AKA RF/signal frequency).

  • mode (quantify_scheduler.enums.SignalModeType) – The output mode type.

  • modulation (quantify_scheduler.backends.types.common.Modulation) – The modulation settings.

  • local_oscillator (str) – The LocalOscillator name.

  • gain1 (int) – The output1 IQ modulation gain. Accepted value between -1 and + 1. (default = 1.0)

  • gain2 (int) – The output2 IQ modulation gain. Accepted value between -1 and + 1. (default = 1.0)

  • trigger (Optional[int]) – The ZI Instrument input trigger. (default = None) Setting this will declare the device secondary

  • markers (List[Union[str, int]]) – The ZI Instrument output triggers. (default = [])

  • mixer_corrections (Optional[quantify_scheduler.backends.types.common.MixerCorrections]) – The output mixer corrections.