The first recipients of the award are Niels Bultink and Adriaan Rol, both from the Delft University of Technology. Under the leadership of Leo DiCarlo, the head of the Quantum Transport group, they supported Zurich Instruments during the development of the UHFQA Quantum Analyzer and the HDAWG Arbitrary Waveform Generator.

Niels and Adriaan received the prize from Sadik Hafizovic, CEO of Zurich Instruments, during a lab visit at the TU Delft. On the picture: Leo DiCarlo, Sadik Hafizovic, Adriaan Rol, and Niels Bultink (from top left)

The very first HDAWG prototype outside Zurich Instruments's R&D lab was delivered to you in the summer of 2017. At that time, the device was still under development and did not always behave as expected. What did the role of an early adopter of the HDAWG entail?

Adriaan: I would say that my role consisted mostly of a combination of debugging and suggesting features that would improve the usability of the HDAWG. We were also closely involved in the development of some key features such as the real-time predistortion filters.

When we first got the HDAWG it was not yet in the shape it is in now. My first task was to hook it up to our setup. This may seem rather simple, but a lot of the basics were still missing, for instance there was no driver for our Python framework. Even when everything was done right, there were often bugs such as parameters that would return nonsensical values when read out, or output voltages that did not correspond to what was indicated by the instrument. After some discussions with ZI's software team, a feature was added to the ZI Python API that allows us to auto-generate these drivers based on the API itself, including the documentation for each parameter. This is very useful when new features are added as we have effectively eliminated the driver writing.

Tell us about your worst experiences with the prototype?

Adriaan: My worst experience has to be what we internally call the "staircase test" but a particular mode dependent rounding error in waveforms is a close second. In our experiments, we use the HDAWG by triggering specific predefined waveforms using a codeword trigger send using a digital input/output (DIO) signal. To test the synchronization of this DIO protocol, we would provide triggers to generate a staircase pattern of waveforms. We quickly found that the timings of this protocol need to be calibrated to ensure the right waveforms were played. The problem was that not only was the initial calibration protocol not very reliable, requiring a lot of restarts and other hacks to get it to work, it was also possible for the staircase to look fine but only glitch once every few minutes. The last part especially made it very frustrating to use the HDAWG in experiments. Since then, a new DIO calibration routine has been implemented that addresses this problem.

What was your experience of the collaboration with our R&D team?

Adriaan: I am very positive about our collaboration with the R&D team, both for immediate support as well as their expertise in developing new features. Besides our bi-weekly Skype meetings we can call in any moment and receive support using TeamViewer. It also helps that we know most of the R&D team having met them several times in the last couple of years.

What were the biggest measurement challenges that you faced and can now be solved with the HDAWG?

Adriaan: The biggest problem that the HDAWG solves for us is that of real-time distortion corrections. When we use the HDAWG to generate flux-pulses to perform two-qubit gates using transmon qubits, the waveforms are typically distorted on their way to the qubits. The traditional way of correcting this is by applying a predistortion filter to the waveforms being played. The problem with this is that these filters are history dependent, and as a consequence all pulses in a program need to be combined into one very long waveform that contains the predistortion correction for all the waveforms. Having these very long waveforms is undesirable for several reasons; besides the memory limitations and loading times, requiring a single very long waveform is incompatible with a flexible control scheme that relies on using codewords to trigger individual pulses. More fundamentally, requiring a very long waveform makes it almost impossible to perform real-time feedback.

By applying the pre-distortion corrections in real-time all these problems disappear, significantly reducing the complexity of these experiments.

Can you give us an idea about your experimental setup where you use the HDAWG?

Adriaan: We perform experiments on superconducting transmon qubits. Single qubit gates are performed using microwave pulses generated by an HDAWG, these pulses are then routed to the right qubit using a Vector Switch Matrix that allows us to use the same primitive pulses for same frequency qubits. Two-qubit gates are performed using flux pulses generated by another HDAWG unit. Readout is performed using multiple UHFQAs. All of these instruments are controlled using a central controller that provides codeword based triggers that determine what operation is performed at what point in time.

Where do you see the biggest value in working with the UHFQA?

Niels: The UHFQA is one of the first commercial solutions for multiplexed qubit readout. Besides, the concept of combining pulse generation and data acquisition into one instrument has great advantages in terms of synchronization.

What was your own experience of the collaboration with our R&D team?

Niels: The relation with the ZI R&D team has always been very pleasurable. I remember one of the first meetings in Santa Fe, New Mexico three years ago where I met part of the R&D team. One of your guys had just had a serious bike accident, and was still walking on crouches, while I had just dislocated my shoulder when skiing. The both of us surely were the last two remaining in the bar to discuss technical matters and of course beyond that. I guess developing instruments together is as much understanding each other’s strengths, as it is being able to compensate for each other’s weakness.

What did the role of an early adopter of the UHFQA entail?

Niels: Beyond the many face-to-face meetings, the role of early adopter has meant a countless number of Skype meetings to debug the instrument (and sometimes other parts of the setup). Although this often meant experimental delays on our side and unforeseen time investments on ZI’s side, communication was always strong and pleasurable.

Can you give us an idea about your experimental setup where you use the UHFQA?

Niels: We’ve used up to three UHFQA units in a single setup to readout 17 superconducting qubits. We interface with the UHFQA with our all-digital FPGA-based controller from which we can order it to readout specific qubits. Measurement results are in turn sent back to the controller that then closes a fast feedback loop. In the near future it will also run an error decoder in real time for Surface Code error correction.

How did the UHFQA help you to publish your scientific results faster?

Niels: Developing and integrating new hardware is not necessarily the fastest way to boost your scientific output in the short run. It has however helped us to reduce the complexity that is exposed to the experimentalists. This is precisely the most important step to scale up your quantum computer in the long run.

***
The first results of the experiments Niels mentions in the interview are already published. Read more:
C. Bultink, B. Tarasinski, N. Haandbaek, S. Poletto, N. Haider, D. Michalak, A. Bruno, and L. DiCarlo, "General method for extracting the quantum efficiency of dispersive qubit readout in circuit QED" arXiv:1711.05336, November 2017