QCoDeS [UNRELEASED DRAFT] (2026-11-09)

Breaking Changes:

  • Several instrument driver parameters now return numbers or booleans instead of the raw string from the instrument, or an integer for an on/off state. Code that compares the returned value to a string, or to 0/1 by identity, needs updating:

    • Keithley2450.line_frequency and Keithley2450Sense.count now return int, and Keithley2450Sense.nplc and Keithley2450Source.delay return float.

    • Keithley7510Sense.user_delay now returns float, and Keithley7510Sense.average_count, Keithley7510DigitizeSense.acq_rate and Keithley7510DigitizeSense.count return int.

    • SR86xBuffer.count_capture_kilobytes now returns int.

    • RigolDG1062Burst.ncycles now returns int, and RigolDG1062Burst.phase, RigolDG1062Burst.time_delay and RigolDG1062Channel.duty_cycle return float.

    • RigolDS1074Z.trigger_level and TektronixTPS2012.trigger_level now return float.

    • TektronixDPOHorizontal.record_length now returns int instead of float, as the record length is a number of samples.

    • AgilentE8267C.pulse_width now returns float and its unit is s instead of ns, matching the value sent to and returned by the instrument.

    • RohdeSchwarzRTO1000ScopeChannel.impedance and KeysightE4980A.range now return float.

    • RohdeSchwarzZNBChannel.bandwidth now returns float, e.g. 10000.0 instead of 10000.

    • The channel state and direct_output and the instrument DC_output of TektronixAWG5014 now return bool. They accept True/False, 1/0 and "on"/"off". TektronixAWG5014.sequence_pos now returns int.

    • KeysightB1520A.ac_dc_volt_monitor now returns bool and also accepts 1/0 and "on"/"off". (#8658)

  • The field units of an AMIModel430 can no longer be changed while it is part of an AMIModel4303D, and doing so now raises a RuntimeError. The set point and field limit of the 3D driver are in the field units of the magnets, so changing them silently changed their meaning by a factor of 10. Set the field units before creating the 3D driver, or close the 3D driver first.

    The coordinate vectors of AMIModel4303D (cartesian, spherical, cylindrical and their *_measured versions) now return a named tuple (CartesianVector, SphericalVector or CylindricalVector from qcodes.math_utils) instead of a list, so the components can also be accessed by name, e.g. cartesian_measured().z. Code that compares the result to a list, or modifies it, needs updating. Setting these parameters still accepts lists, tuples and numpy arrays. (#8679)

  • The trigger_count parameter of Keithley6500, which sets the number of times a scan is repeated (ROUT:SCAN:COUN:SCAN), now accepts the full range documented in the DMM6500 Reference Manual: 0 to 100,000,000, where 0 repeats the scan until it is aborted. It previously only accepted 1 to 9999 or the strings "inf", "default", "minimum" and "maximum", which were copied from the Keithley 2000 and are not documented for this command. These strings are now rejected; use 0 instead of "inf". Whole-number floats are accepted and sent as integers, and the parameter is annotated as Parameter[int, Self]. (#8694)

  • The auto_zero_once parameter of Keithley7510Sense and the trigger_in_ext_clear parameter of Keithley7510 have been replaced by methods of the same name. Both were command-only parameters, so calling them without an argument, e.g. dmm.sense.auto_zero_once(), only returned the cached value and did not send anything to the instrument. Calling them like this now sends the command. Code that passed a value to them, e.g. dmm.trigger_in_ext_clear(None), must drop the argument, and they are no longer part of the instrument’s parameters or snapshot. (#8694)

Improved:

  • Fixed a rare error on Python 3.12 and 3.13 where closing an instrument could fail with RuntimeError: dictionary changed size during iteration if the garbage collector cleaned up another instrument at the same time. The instrument then stayed registered, so its name could not be reused. (#8680)

  • The “Creating Instrument Drivers” example notebook now explains how to annotate a parameter with the type of its value, e.g. Parameter[float, Self], and all of its example drivers are annotated this way. This also fixed bugs in the example code: the Ithaco example passed units instead of unit, the Keithley 2600 example lacked its model attribute and _display_settext method, and the Alazar example used outdated TraceParameter arguments. (#8694)

  • Dataset preparation now writes startup metadata and the snapshot in one transaction, inserting the snapshot after the other run metadata updates to avoid repeatedly rewriting large snapshots. The complete snapshot is committed before background writers and pending subscribers are started.

Improved Drivers:

  • Add convenience methods to Tektronix DPO7200xx (single, download_waveforms, get_timebase) and AWG70000A (upload_seqx) drivers for streamlined waveform acquisition and sequence upload workflows. (#7865)

  • Most instrument driver parameters are now annotated with the type of their value, e.g. Parameter[float, Self], so static type checkers can infer the type returned by get and accepted by set. As part of this, the mock instruments no longer pass the default parameter_class=Parameter explicitly, and Keysight34980A.module is typed as a mapping from slot number to switch matrix submodule.

    Fixed the following driver bugs found along the way:

    • RigolDS4000.enable_auto_scale always read back as True, since bool("0") is True, and was set with True/False instead of 1/0.

    • RohdeSchwarzRTO1000.trigger_mode queried the trigger source instead of the trigger mode.

    • Keithley6500.digits sent a set command with a stray ?.

    • MessageBuilder.pch_query of the Keysight B1500 sent PCH instead of PCH?.

    • LakeshoreModel325Heater.output_range accepted the high (25 W) range on loop 2, which is only supported on loop 1.

    • LakeshoreModel325Sensor.curve_index now accepts 0 (no curve), which the instrument also reports when the selected curve does not match the input, and rejects non-integer values. Whole-number floats such as 23.0 are still accepted and sent as integers.

    • The LakeshoreModel325 curve list now includes user curve 35, so upload_curve(35, ...) no longer raises an IndexError.

    • LakeshoreModel325Curve.get_data now reads all 200 points of a curve instead of 199.

    • LakeshoreModel325Curve.set_data now writes curve points with 6 significant digits, the resolution of the instrument, instead of rounding them to 3 decimals.

    • LakeshoreModel325Sensor.curve now raises a RuntimeError when no curve is selected for the sensor, instead of returning a curve 0, which the instrument does not have.

    • RohdeSchwarzZNBChannel.bandwidth failed to read a non-integer bandwidth such as 1.5 Hz, which the instrument supports.

    • The current_compliance and power_compliance of the KeysightB1517A sweep could not be set back to None, which leaves the compliance out of the WV command. Since a power compliance can only be sent together with a current compliance, clearing the current compliance also clears the power compliance.

    • Keysight34980A.scan_slots can now be called again; it skips slots that already have a module instead of failing to register a duplicate submodule.

    • Keithley7510DigitizeSense.aperture was labelled in us although the instrument uses seconds; its unit is now s.

    • averages, points and active_trace of the Keysight PNA (KeysightPNABase) drivers now reject non-integer values. Previously averages failed to format a float such as 2.5 and the others sent it to the instrument. Whole-number floats such as 2.0 are still accepted and sent as integers. (#8658)

  • CopperMountainM5xxx.averages, AimTTiChannel.curr_range, RigolDG1062Burst.ncycles, Keithley7510Sense.average_count and the dwell, pause, curve_number and excitation_range_number of LakeshoreModel372Channel now reject non-integer values. Previously these were sent to the instrument, or for CopperMountainM5xxx.averages silently truncated. Whole-number floats such as 2.0 are still accepted and sent as integers. Also removed the incorrect unit of AimTTiChannel.curr_range, which is a range number, fixed the label of LakeshoreModel372Channel.temperature_coefficient, and removed incorrect units from RigolDS1074Z.trigger_mode, waveform_npoints, waveform_yorigin and waveform_yref. (#8678)

  • The field parameters of AMIModel4303D, such as cartesian, x and field_measured, now use the field units of the individual magnets, tesla or kilogauss, instead of always reporting tesla. A warning is logged if the magnets use different field units. Also removed the placeholder unit tuple? of AMIModel4303D.spherical and cylindrical, whose components have different units, which are now described in their docstrings. Closing an AMIModel4303D now detaches its magnets, so set_field, zero and changing the units of a magnet that is still open no longer fail. (#8679)

  • The parameters of the Basel SP983, SP983a and SP983c, Ithaco 1211, Keithley 2450, DMM7510, S46 and 2600 drivers that were not yet typed are now annotated with the type of their value, e.g. Parameter[float, Self], so static type checkers can infer the type returned by get and accepted by set. (#8694)

New:

  • Added the named tuples CartesianVector (x, y, z), SphericalVector (field, theta, phi) and CylindricalVector (rho, phi, z) to qcodes.math_utils, for the components of a field vector in a given coordinate system. (#8679)

  • Added recording of parameter access: inside a with qcodes.parameters.record_parameter_accesses() as accesses: block, every get and set of a parameter made by the current thread (or asyncio task) is appended to accesses as a qcodes.parameters.ParameterAccess: the names of the parameter and its instrument, the value and raw value, when the access started, how long it took and the error it raised, if any. A set that ramps in steps is recorded once per step. Accesses are appended to a new list, or to a mutable sequence that is passed in, whose append must only store them. qcodes.parameters.log_parameter_accesses logs the accesses to the qcodes.parameters.parameter_access logger instead. Recording only stores data: it is not a way to react to an access, and a record holds names rather than the parameter or instrument. This makes it possible to see all instrument communication made by a piece of code without changing it. (#8684)

Under the hood:

  • The test suite now runs faster. Tests using the simulated Keysight B220x and B1500 instruments no longer first try to connect to a real instrument, and a number of unneeded sleeps and waits have been removed or shortened. (#8725)

  • The documentation builds faster: the InstrumentGroup example notebook ramps its gate in fewer steps, and the qcodes.sphinx_extensions.parse_parameter_attr Sphinx extension now parses the source of each class only once instead of once for every attribute documented on it. (#8742)