Changelog

All notable changes to this project will be documented in this file. The format is based on Keep a Changelog, and this project adheres to Semantic Versioning. See also our stability guarantees.


Version 2.15.1 – 2024-10-18

Fixed

  • Fix performance regression in lockin mode with small values of df, introduced in v2.15.0

  • Fix wrong output amplitude in SymmetricLockin, introduced in v2.15.0


Version 2.15.0 – 2024-10-15

Added

  • Ability to continuosly stream lockin measurements to the computer. See new stream_pixels() method.

  • Ability to run arbitrary user code in the on-board real-time processor (RPU) in lockin mode. Especially useful for implementing custom feedback schemes.

  • Ability to decrease the sampling rate in the intermediate-frequency (IF) domain in pulsed mode. This allows e.g. to store longer pulses in the output templates, and to acquire longer input waveforms. See downsampling optional argument to the Pulsed class.

Fixed

  • Optmize network usage in lockin mode by changing internal data types.

Changed

  • Drop support for Python 3.8, the minimum-supported Python version is now 3.9


Version 2.14.1 – 2024-08-16

Fixed

  • Specify paramiko as optional dependency and set minimum version to solve authentication issues on some computers.

  • Fix internal command in utils.ssh_restart().

Changed


Version 2.14.0 – 2024-08-12

Added

  • New method Pulsed.setup_flat_matching_pair() creates reference templates for template matching that are flat (i.e. have a constant weight) and can be very long. Much like output templates created via setup_long_drive(), these long flat reference templates will use only one reference-template memory slot.

  • New method Pulsed.setup_flat_pulse() replaces the now-deprecated setup_long_drive(). The new method uses envelope=False by default, consistent with the behavior of setup_template(), and drives both I and Q quadratures by default on mixed mode. See deprecation note on setup_long_drive() for a migration guide.

  • New method Hardware.configure_mixer_base_n() allows to specify strict rational requirements between different NCO frequencies on different ports. It works similarly to configure_mixer(), but you specify a base frequency common to all NCOs and then request that each NCO be exactly an integer multiple of this base frequency, even taking into account small rounding errors in the digital domain.

Changed

  • Method Pulsed.setup_template_matching_pair() will now split reference templates longer than MAX_TEMPLATE_LEN into smaller sub-templates. Previously, the method would raise an exception if a template was too long for a single memory slot. Now the exception is raised if the sub-templates require more memory slots than available.

  • Update internal Linux distribution.

  • Improve error message when the API version installed on the computer is not compatible with the firmware version installed on the hardware.

Fixed

  • Fix output of digital markers/triggers in pulsed mode.

  • Fix bug where, in some cases, the autoconfiguration feature would warn about incompatible DAC configurations and choose a suboptimal configuration.

  • Fix bug in Pulsed.setup_condition() that would cause an error when threshold was negative.

  • Workaround for a bug in upstream code that prevented Hardware.configure_mixer() from using an NCO phase equal to ±π.


Version 2.13.0 – 2024-06-03

Added

  • All 128 reference templates for template matching in pulsed mode can now be used on a single input port. Previously, each input port could access only 8 reference templates. Now the available reference templates are automatically distributed across the input ports based on the user program.

  • New method Pulsed.setup_store() combines the functionality of the now-deprecated set_store_ports() and set_store_duration(): now you won’t have to remember which of the two you should call first.

  • New method Hardware.disable_coeff_override() to disable factory calibration of ADCs.

  • Add support pulsed sequences without any data acquisition (store or match).

Fixed

  • Fix checking of FIR coefficients in Lockin.get_pixels().

  • Improve checking of whether NCO frequencies are compatible with DAC settings.

  • When using Presto with an external reference clock, always check if the frequency lock is good before running the experiment.

Deprecated


Version 2.12.2 – 2024-03-26

Added

Changed

  • Improved internal representation of very long, very sparse pulsed sequences.

Fixed

  • Fixed bug in pulsed experiments with frequency or amplitude sweeps that caused some of the sweep entries to be duplicated or corrupt. The bug was introduced in version 2.12.0.


Version 2.12.1 – 2024-03-14

Fixed

  • Fixed bug that caused Lockin and SymmetricLockin to download the wrong data at the end of a measurement when using summed=True. Measurements with summed=False were not affected by the bug.


Version 2.12.0 – 2024-02-28

Added

  • Autoconfiguration feature: ability to automatically choose DAC configuration (DacMode and DacFSample) based on signal frequencies. See Direct and Mixed mode for more information and how to (optionally) migrate your old code to use the new feature.

  • The plot_sequence() utility now shows operations on IF generators.

  • Ability to trigger signal generation and acquisition in lockin mode on an external trigger, likely generated by Metronomo. See auto_sync optional argument to apply_settings() and get_pixels().

Changed

  • Default parameter values for Pulsed and Lockin now use automatic DAC configuration.

Fixed

  • Fixed ConnectionRefusedError when trying to reconnect to Presto too quickly after closing a previous connection.

  • Fixed bug that cause some very-long sequences in pulsed mode to be rejected due to the repetition period being too short.


Version 2.11.0 – 2023-11-27

Added

  • Historical version information in the documentation of the version module.

  • Option to download measurement data using the lz4 lossless compression. It can improve download times over slow networks. See use_compression keyword argument in Pulsed.get_store_data() and get_template_matching_data().

  • Option to download template-matching results as they become available during an experiment, rather than at the end of the experiment. Can reduce communication overhead for sequences with lots of single-shot readouts. See prefetch_matches keyword argument in Pulsed.run().

  • Ability to configure low-noise mode on an output port: Hardware.set_dac_lownoise().

  • It is now possible to compute real-time histograms on template-matching results on board. See methods Pulsed.setup_histogram(), setup_histogram_2d() and get_histogram_data().

  • Utility to plot the programmed experimental sequence in Pulsed mode. See utils.plot_sequence().

Changed

  • Minimum supported versions: Python 3.8, NumPy 1.22.0.

Fixed

  • Optimize internal data transfers, yielding improved download speed over the network and processing of commands during pulsed sequences.


Version 2.10.0 – 2023-06-27

Note

presto v2.10.0 is the last version supporting Python 3.7, which reached [end of life][pyver] on 2023-06-27. Starting from next version, the presto package will require Python 3.8 or above.

Added


Version 2.9.0 – 2023-03-30

Added

Fixed

  • Improve handling of abrupt interruption of a measurement.

  • Fix error message when rebooting Presto/Vivace via ssh_reboot().


Version 2.8.1 – 2023-02-14

Changed

  • Allow for a list of ports in Pulsed.output_dc_bias() to change DC bias during a pulsed sequence simultaneously on many ports.


Version 2.8.0 – 2023-02-03

Added

Changed

  • Enable positive-only DC-bias ranges in Hardware.set_dc_bias() (range_i settings 0 and 1) for higher resolution on small positive voltages.


Version 2.7.1 – 2023-01-20

Fixed

  • Fixed Python 3.7 compatibility issue in pulsed module.


Version 2.7.0 – 2023-01-18

Added

Changed

  • A DC bias set with Hardware.set_dc_bias() is now persistent across a firmware reload, e.g. when switching between pulsed and lockin mode, unless the hardware is initialized with force_reload=True.

Fixed

  • Documentation for Lockin.set_phase_reset().

  • Fix bug causing spurious Pulsed.next_scale() events that could happen when the repetition period of an experiment was longer that half a millisecond.

  • Fix erroneous error message that would warn of an error when using template matching even though everything was good.


Version 2.6.0 – 2022-12-07

Added

  • Abort a pulsed measurement early if data corruption is detected.

Changed

  • Improve handling of pulsed-mode commands at run time to enable longer two-qubit randomized benchmarking.

  • Update all internal libraries.

  • Speed up reconnecting to Presto if no settings are changed.

Removed

  • Remove experimental feature enable_compression in Pulsed.run(): this feature is now covered by the improved handling of pulse-mode commands.

Fixed


Version 2.5.1 – 2022-11-05

Fixed


Version 2.5.0 – 2022-09-20

Added

  • A pulsed measurement can now be triggered internally, by a SYSREF signal, or (new) by a signal on any of the digital inputs. See method Pulsed.run().

Changed

  • Replace the sliding-window low-pass filter for Lockin introduced in 2.3.0 with a customizable 43-tap FIR filter. For the occasion, the new optional argument fir_coeffs replaces the previous nbox in methods Lockin.get_pixels() and SymmetricLockin.get_pixels().

Fixed

  • Fixed an issue that could result in a DC bias not being set reliably on some units.


Version 2.4.1 – 2022-07-08

Fixed


Version 2.4.0 – 2022-06-23

Added

  • You can now choose whether a long drive is an envelope or a template. See new optional parameter envelope in Pulsed.setup_long_drive(). To maintain backwards compatibility, the default setting is envelope=True. Note that this is different from Pulsed.setup_template() where the default is envelope=False.

  • New interface to asynchronous measurements: Pulsed.run_async() returns a MeasurementHandle that allows interaction with a running measurement.

Changed

  • Methods Lockin.get_pixels() and SymmetricLockin.get_pixels() now return the standard deviation as a complex number, where the real and imaginary parts encode the standard deviation of the real and imaginary parts of the demodulated signals. Taking the absolute value of the complex number results in the total standard deviation that was previously returned.

  • Many methods in Pulsed raise a RuntimeError exception if called after a measurement has already been run(). The previous behavior was not supported, see the warning in the documentation for the Pulsed class.

  • The output_port argument to Pulsed.setup_long_drive() now accepts a list to output the same pulse on multiple ports, much like Pulsed.setup_template() already did.

Fixed

  • Fix inconsistencies in the documentation regarding maximum length of templates.

  • Improve error messages when running out of template slots.


Version 2.3.2 – 2022-06-06

Fixed


Version 2.3.1 – 2022-04-25

Fixed

  • Restore compatibility with Python 3.7+


Version 2.3.0 – 2022-04-21

Added

  • New class SymmetricLockin: a special-purpose version of Lockin with same frequencies on input and output (symmetric). The new class inherits from the previous one, so most of the Lockin functionality is available in SymmetricLockin as well.

  • New features for the lock-in: Presto can now low-pass filter the measured lock-in packets in real time with a sliding-window average, and also compute mean and standard deviation of the measured lock-in packets in real time over chunks with configurable width. Check out the summed, nbox and nsum optional parameters to Lockin.get_pixels().

  • The lock-in can output a trigger signal at the start of each integration window: Lockin.set_trigger_out().

  • utils.sin2() can now create DRAG-corrected pulses.

  • It is now possible to modify the DC bias output during a pulsed measurement sequence: Pulsed.output_dc_bias().

  • Print a warning if the signal on any of the input ports is out of range.

Changed

  • Change the way errors from the hardware are reported to the user. It should help when the API is used through third-party software, such as Labber.

  • Improve handling of NCO frequencies when the same frequency is set on multiple ports. This can avoid long-term drift, see documentation for the tune parameter in Hardware.configure_mixer()

  • Increase length of command buffers: you can now send more pulses closer together! Useful for randomized benchmarking.

Fixed

  • Fix bug in Lockin.get_pixels() where the data in first lock-in packet would often be corrupt.

  • The pulsed mode of operation is now more resilient to aborting measurements with Ctrl+C.


Version 2.2.1 – 2022-02-10

Fixed


Version 2.2.0 – 2022-02-09

Added

  • Method Hardware.set_dc_bias() now handles the DC bias outputs on Presto hardware.

  • Check that all programmed times in a pulsed sequence are nonnegative.

  • Method Pulsed.run() will now check for more inconsistencies and report errors:

    • if the programmed sequence contains overlapping calls to store(). Overlapping stores are not supported and could cause the experimental sequence to terminate abnormally, or not terminate at all.

    • if some select_frequency() or select_scale() event is trying to access indexes beyond what was programmed in the look-up tables with setup_freq_lut() and setup_scale_lut().

  • Method Pulsed.run() can now compress some pulse sequences so that they consume fewer commands. Compression is useful when there’s a long list of small templates output in quick repetition, such as in randomized benchmarking with thousands of Cliffords. To opt in to compression, pass enable_compression=True.

Changed

  • Refactor calibration of ADC and DAC alignment. A calibration routine will be run when a new combination of Converter modes and Converter rates is requested. This will delay the start of a measurement by a few seconds, but only the first time a configuration is requested: the calibration will be stored on Vivace/Presto and will be available for all subsequent measurements without delay!

  • Method Pulsed.get_store_data() returns data in real form when adc_mode=AdcMode.Direct and in complex form when adc_mode=AdcMode.Mixed (when using digital downconversion).

  • Method Pulsed.setup_template_matching_pair() now accept complex-valued arguments when adc_mode=AdcMode.Mixed (when using digital downconversion).

  • Methods Pulsed.setup_long_drive() and Pulsed.setup_template() now accept complex-valued arguments when dac_mode=DacMode.Mixedxx (when using digital upconversion).

Fixed


Version 2.1.0 – 2021-12-01

  • Ability to choose frequency of external reference clock: see the optional argument ext_ref_clk to Pulsed and Lockin

  • Improvement to calibration routine during update.

pulsed module

  • Fix a bug that caused the experiment to terminate too early or to never terminate when running long sequences with lots of averages.

  • A single sequence (without accounting for repetitions and averages) can now run for ≈6.5 days, up from ≈8.6 seconds.

Pulsed class

  • Fix printing of estimated time left in a measurement when running from within some Python IDEs, e.g. Spyder.

  • Set tolerance for event times to 1% of get_clk_T(), rather than fixed at 1 ps.

  • Method run():

    • Better handling of startup time for a sequence: fixes an inconsistency with requirements on the period argument.

    • Raise an error if the user runs a sequence without setting up the scale look-up table on an output port and group used by some template, or if the frequency/phase look-up table was non set up when some templates are used as envelopes.

  • The parameter output_port to setup_template() can now be set to a list to more conveniently output the same pulse on multiple output ports.

  • Improve error messages when using not supported combinations of adc_mode, adc_fsample, dac_mode and dac_fsample

  • Fix error when running with dry_run set to True.

LongDrive class

  • New method get_duration() for compatibility with Template class.

lockin module

Lockin class

  • Fix bug in tune_perfect()

  • Improve error messages when using not supported combinations of adc_mode, adc_fsample, dac_mode and dac_fsample


Version 2.0.0 – 2021-05-21

This is the first major-version change since the previous Vivace API. This new API is compatible with both Vivace and Presto microwave platforms, and follows the same overall programming interface. There are, however, many differences that break backwards compatibility. See below for a summary of such changes.

hardware module

New module to control configurations common to all modes of operation. Accessible as the hardware attribute of any instance of Pulsed and Lockin.

pulsed module

Pulsed class

  • The class constructor takes 4 new optional arguments: adc_mode, adc_fsample, dac_mode and dac_fsample. These configure the Converter modes and Converter rates for the input analog-to-digital (ADC) and output digital-to-analog (DAC) converters. With these options it is possible to enable digital up- and downconversion using the built-in digital IQ mixers.

  • The methods next_frequency(), next_scale(), select_frequency() and select_scale() take a new optional argument group. If not specified, both output groups on output_ports are affected.

  • The method run() (previously perform_measurement) does not return the result of store() events: use get_store_data() after running the experiment. To get the result of match() events, use get_template_matching_data() (same as before).

  • Method setup_freq_lut():

    • the old carrier argument (1 or 2) is replaced by the new group argument (0 or 1)

    • new argument phases_q must be specified when using the built-in digital IQ mixers for digital upconversion

    • the optional argument repeat_count is removed

  • Method setup_long_drive():

    • the old carrier argument (1 or 2) is replaced by the new group argument (0 or 1)

    • the optional argument use_scale is removed: all output templates are subject to the output scaler for the appropriate port and group

  • Method setup_scale_lut():

    • new mandatory argument group (0 or 1)

    • the optional argument repeat_count is removed

  • Method setup_template():

    • new mandatory argument group (0 or 1)

    • new argument template_q must be specified when using the built-in digital IQ mixers for digital upconversion

    • the optional argument use_scale is removed: all output templates are subject to the output scaler for the appropriate port and group

    • the optional argument envelope is now a boolean. To specify which signal generator to use when envelope=True, use the group argument (0 or 1)

  • New methods:

  • Deprecated methods:

    • perform_measurement: use run() instead

  • Some methods are moved to the new Hardware class, which is available as the hardware attribute to any instance of Pulsed:

LongDrive class

  • Renamed methods:

    • update_flat_duration: now set_flat_duration()

    • update_total_duration: now set_total_duration()

  • New methods:

    • get_flat_duration()

    • get_total_duration()

Other

utils module

  • New function rotate_opt()

  • The function untwist_downconversion() now follows a different convention. The new convention is consistent with the built-in digital IQ mixers. It might or might not be consistent with the convention used by an external analog IQ mixer.

version module

  • New attributes:

    • version_api: the only one of interest to the user. It is also exported as the __version__ attribute to the presto module

    • version_conductor

    • version_fpga

    • version_rpu

  • Removed functions:

    • get_variant_firmware

    • get_version_all

    • get_version_api

    • get_version_clock

    • get_version_firmware

    • get_version_server