pulse

qendra pulse delivers high-fidelity RF waveform control with nanosecond-level precision.
 Developed for trapped-ion and neutral-atom experiments, pulse generates synchronized signals for every stage — from cooling to measurement.

The AMD ZCU216 ready for AOM control

Integrated intensity stabilization can be as simple as pie

16 RF outputs, 16 ADC channels, 16 digital outputs. Ready to pulse

RF outputs

Property Quantity
Channels 16
Skew Length matched in groups of 4, 0.7 ns between earliest and latest group
Sampling rate 1 GSPS
Frequency resolution 32 bits (0.23 Hz)
Amplitude resolution 14 bits (84 dBFS)
Phase resolution 16 bits (96 μrad)
Time resolution 4 ns
Minimum pulse time 40 ns
Digital waveform generation 4 sinusoids per physical channel
Frequency modulation Linear interpolation
Amplitude modulation Cubic polynomials
Phase modulation Zero-order hold
Parameter storage per channel 1 Mib (5957 cubic polynomials or 10922 linear segments)
Sequence storage per channel 640 Kib (20480 instructions or 2560 unique pulses)
Waveform memory per channel octet 1 GSa
AWG channels per channel octet 3 (triggered simultaneously, can be output on any of the 8 channels of the octet)
NCOs per channel 4
Output frequency 0.2 – 450 MHz (AC coupled)
Output power 5 dBm
Phase noise -139 dBc/Hz @ 10 kHz
-148 dBc/Hz @ 1 MHz

The numerically controlled oscillators (NCOs) operate in phase-coherent switching mode by default. The output phase is referenced to the beginning of the pulse sequence.

To meet the demands of neutral atom quantum experiments, the frequency modulator will receive an optional upgrade to support cubic polynomials (tentatively scheduled for Q2 2026). Additionally, the channel mapping will be customizable with multiple logical channels added together to a single physical output or the individual NCOs per channel fanned out to multiple physical channels (e.g. physical channel RF 0 outputs the sum of logical channel CH 0, 1, 2 and 3, while frequency tone 0 and 1 of CH 4 are routed to RF 1 and tone 2 and 3 are routed to RF 2). The multi-channel mapping feature is scheduled for Q2 2026 as well.

Analog inputs

Property Quantity
Channels 16
Skew 1.3 ns between earliest and latest channel
Sampling rate 62.5 MSPS
Amplitude resolution 14 bits (84 dBFS)
Time resolution 16 ns
Input frequency 0 - 8 MHz (DC coupled)
Voltage range ± 0.5 V
Input termination 50 Ohms, differential (isolated ground)
Modes Photodiode monitoring and integrated intensity stabilization

Digital outputs

Property Quantity
Channels 16
Voltage level 5 V
Galvanic isolation Isolated from chassis and from each other in groups of 4
Time resolution 4 ns
Minimum pulse time 12 ns
Sequence storage 1 Mib (32768 instructions or 8192 unique pulses)

Digital inputs

Property Quantity
Channels 3 (including 1 AC-coupled input for < 300 MHz signals)
Voltage level 5 V
Time resolution 4 ns
Modes Time tagging and gated counting
Counter width 18 bits
Maximum window length 4 minutes
Line trigger Included

Sequencer

Advanced sequencing capabilities support loops, branching and functions. A sequence of pulses can be repeated up to 16.7 million times and called at multiple different times of the same experimental sequence. Loops and functions can be nested up to a total of 8 levels. An efficient use of these features makes it possible to efficiently reuse sequence blocks such as sideband cooling or syndrome measurement.

This functionality is complemented with real-time parameter and sequence updates that are popped from a queue in real time. This makes it possible to scan the length and parameters of a pulse within a loop. Furthermore, this can be combined with branching to completely change the sequence structure between loop iterations.

Readout processing

QuENCH follows a CPU-FPGA-hybrid architecture. While the pulse generation and sequencing are offloaded completely to dedicated sequencers on the FPGA, the CPU remains responsible for analysing the readout results. This is possible with atom-based quantum computing because the typical times of a readout operations are much longer than 100 μs such that latencies up to multiple μs are acceptable.

This simplifies the implementation of mathematically non-trivial schemes such as adaptive maximum likelihood readout. Using the branching functionality of the sequencers, the CPU can efficiently steer the further flow of the sequence by updating few instructions in the instruction memory.

This way, latencies of 2 μs were achieved using older hardware based on the AMD Zynq-7020 SoC.