The Next Generation of Signal Generators: SHFSG Launch Event – Summary and Q&A

Event Summary

Last Thursday, April 29th, Jan Benhelm, Mark Kasperczyk, and I welcomed a crowd of customers to the interactive online event introducing Zurich Instruments’ newest product, the SHFSG Signal Generator. Our aim was to present multiple perspectives on the launch of this powerful instrument.  Jan provided an overview of Zurich Instruments’ quantum technology roadmap and the role of the second generation of our Quantum Computing Control System (QCCS). I gave an introduction to qubit control, and highlighted the most relevant features and specifications of the SHFSG. It was then Mark’s turn to describe the powerful radio-frequency (RF) engineering technology inside the SHFSG, to explain where and how this instrument makes a difference in a quantum computing experiment, and to demonstrate how the SHFSG generates its signals in practice. Below is a list of some of the topics covered by Mark:

  • Setup of a single-qubit Rabi sequence
  • Generation of a randomized benchmarking sequence
  • Phase control of 2 outputs
  • Instrument trigger with the PQCS over ZSync
  • Performance comparison with an IQ mixer setup
  • First measurement data obtained with a real qubit

If you missed the launch event or would like to watch the video again, please find the complete recording here. The presentation ended with a live Q&A session: the answers provided during the session are summarized below, and video links take you directly to our live replies. Some questions couldn’t be answered during the event due to time constraints, so we’ve also addressed those below. Finally, we encourage you to get in touch with us to ask the questions that haven’t yet been asked!



Is the SHFSG good for other types of qubits besides superconducting and spin qubits?

We certainly see that there is a potential match with applications outside those mentioned. For instance, the frequency range matches with that needed for NV color centers in diamond (many of the statements on the related application page based on the HDAWG are also applicable to the SHFSG), or for high-field NMR. There is no general answer whether the application matches well, therefore we invite you to get in touch with us to discuss your detailed requirements. Answered live at 51:02

Do you have to make compromises when designing a system both for small setups (technology thrust) and large setups (scaling thrust)?

In the scaling thrust, it’s valuable to have the flexibility to tailor the systems according to the specific channel configurations. For the technology thrust, in turn it is important to have the instruments useable stand-alone. We can cover both needs with our approach to hardware and software design, so there are probably no fundamental compromises that we have to make at this stage when building a large system out of components that function as stand-alone instruments. This might change in the future for even larger systems, when the cost per channel becomes more important. Answered live at 52:04

Why is the SHFSG not used for flux pulses, but only for the RF pulses?

We recommend using the HDAWG for generating flux pulses because it offers higher vertical resolution and higher output range than the SHFSG in the relevant frequency range, roughly 0 to 500 MHz. In addition, the HDAWG comes with the real-time precompensation option. That being said, the SHFSG has the essential capability to generate flux pulses, most notably a frequency range down to DC, and may be used to generate flux pulses opportunistically when needed – in full awareness of the mentioned limitations. Answered live at 53:50

You showed a system with several SHFSGs but only a single SHFQA. How realistic is this?

It’s realistic, because the SHFQA Quantum Analyzer supports multiplexed qubit readout, so one SHFQA channel often is combined with multiple SHFSG outputs. A surface-17 setup would for instance require 2 SHFSG8, 1 SHFSG4, but only a single SHFQA2. Answered live at 56:30

What is the waveform loading time and how can waveforms be loaded?

There are multiple ways to define and upload waveforms, e.g. based on CSV files, or using built-in function in the LabOne sequencer language. The pure transfer time to fill the SHFSG waveform memory is actually quite short, of the order of milliseconds. Thanks to the advanced sequencer feature set, e.g. the command table, digital modulation and real-time phase changes, or the playZero instruction, even a small amount of memory is sufficient to represent long an complex signals. Answered live at 58:50

What is the reason for using 6 GHz?

In the measurements comparing SHFSG and IQ mixer SFDR and drift, we chose a carrier frequency of 6 GHz as a representative example in the center of the most often used range of 4 to 8 GHz for controlling superconducting qubits. Across this frequency band, the drift of IQ mixers is equally poor, and the SFDR of the SHFSG is equally good – even better actually at 4 and 8 GHz than at 6 GHz. Therefore, the conclusion drawn from these measurements is valid for the range 4-8 GHz, not only 6 GHz. Answered live at 57:50

What is the SHSFG’s ENOB (effective number of bits) at 1 GHz?

Let me answer by providing the specs of the underlying signal properties that are summarized in the ENOB: spurious level, linearity, and noise. At 1 GHz, the SHFSG performs according to the following specifications:

  • SFDR (excl. harmonics) at 1 GHz: -74 dBc (0 dBm)
  • Worst harmonic component at 1 GHz: -40 dBc (10 dBm)
  • Voltage noise density at 1 GHz: -135 dBm/Hz (10 dBm, offset > 200 kHz)

The ENOB is certainly a helpful summary specification parameter for many application use cases, e.g. telecommunication. For the quantum computing application, in our experience the three underlying parameters have qualitatively different effects in the experiment, which is why we find it useful to have them listed separately.

Answered live at 1:00:50

What’s the SFDR (spurious-free dynamic range) on the SHFSG output channel?

Also here, we refer to the specifications published on the SHFSG website. The SFDR (excluding harmonics) at several frequencies is specifically:

  • -74 dBc (1 GHz, 0 dBm)
  • -66 dBc (4 GHz, 0 dBm)
  • -60 dBc (6 GHz, 0 dBm)
  • -65 dBc (8 GHz, 0 dBm)

Answered live at 1:00:50

Are there plans to increase the max frequency of 8 GHz to 12 GHz? / Are higher frequencies than 8.5 GHz planned?

There are currently no plans to increase the frequency range of the second generation of QCCS instruments (the SHFQA, and the SHFSG). However, using a double balanced mixer with a large intermediate frequency bandwidth combined with filtering, it is possible to reach higher frequencies while keeping the advantages of the double superheterodyne approach.

You mentioned a customer could use the command table to speed up their randomized benchmarking experiments. Can you give any details about this?

There are multiple methods for implementing randomized benchmarking on the HDAWG. Probably the most powerful one makes use of the real-time pseudo-random number generator (PRNG) on the HDAWG. For example, the command table then may contain 30 entries, 1 for each single-qubit Clifford gate. The PRNG may then be used to select random entries of this set, and generate random sequences on the fly instead of uploading them. Keep an eye on our blog, where we’ll shortly publish a post describing this method in more detail. Answered live at 1:04:10

What’s the phase drift between two different channels of SHFSG?

We have performed preliminary measurements, but require further measurements and more statistics to finalize this specification parameter. The drift occurs on the timescale of 1 hour or more, while over shorter timescales (e.g. tens of minutes) significantly higher stability is reached on shorter timescales. In the SHFSG8, neighboring outputs share the same local oscillator reference, which means that their relative phase is not limited by the absolute phase stability of one output.

Are the bandpass filters used in double super heterodyning programmable?

The bandpass filtering after the second frequency conversion is automatically set to one of multiple configurations depending on the center frequency chosen by the user. In that sense, it is only indirectly programmable by the user. The used configuration is chosen in order to maximize SFDR at each point of the instrument’s frequency range.

Does the SHFSG increase gate fidelity? If yes, what is the highest fidelity you achieved?

This is not an easy question to answer. The signal generation is not usually the factor limiting gate fidelity. Nonetheless, there are preliminary single-qubit measurement data obtained with the SHFSG do indeed look very encouraging in the sense that the SHFSG may give access to higher fidelities in practice. But there need to be deeper investigations before reaching a conclusion on this key question. When we have news on this topic, we will certainly communicate it in one of our quarterly QT newsletters. Answered live at 1:03:00

What’s the amplitude nonlinearity in SHFSG?

It is clearly better than that of an IQ upconversion setup. In a qubit measurement, the amplitude nonlinearity is e.g. relevant when calibrating gate pulse amplitudes with a Rabi oscillation measurement. In preliminary measurements, we have seen clearly better data quality obtained with the SHFSG compared with a state-of-the-art IQ mixer setup. This allows for a more accurate gate pulse calibration, especially when needing many calibrated amplitudes (e.g. for quantum simulation).