Quantum Optics Group ETH Zurich A low-noise and scalable FPGA-based analog signal generator for quantum gas experiments. D. Pahl*, L. Pahl*, E. Mustafa*, Z. Liu†, P. Fabritius†, J. Mohan†, P. Clements†, A. Akin†, T. Esslinger† Department of Information Technology and Electrical Engineering, ETH Zurich, Switzerland † Department of Physics, ETH Zurich, Switzerland * 1 Introduction • Each contains all metadata for its channel State-of-the-art quantum gas experiments rely on the precise and dynamical control of experimental parameters like magnetic fields and laser powers to probe quantum systems. Achieving high fidelity for such tasks imposes strict requirements on the stability of surrounding electronics. To address these challenges, we propose a scalable analog signal generator based on a field-programmable gate array (FPGA) development board and a custom PCB hosting a digital-to-analog converter (DAC) with 20-bit precision at 1MSPS. • Each event is specified by the number of samples ( ) and • This pair of data is used by module to control data flow of corresponding sample data from DDR3 belonging to that event 2 System Overview • PC transmits user defined waveforms for each channel via TCP/IP connection • FPGA write waveforms to on-board RAM, parses the data structure, and streams samples as SPI signals at runtime • Breakout boards distribute signals over LVDS Ethernet cables to DAC boards • DAC chips convert digital signals to analog signals and drive devices Figure 1: System overview Event sequence for multiple channels • Arbitrary waveforms can be generated • Each channel is subdivided into events that are defined by sampling frequency and signal length specified in metadata • Events for all channels are concatenated into data packets for transmission to FPGA Event 1 Event 2 Event 3 Event 4 • indicates necessary repetitions for each sample to scale sampling frequency Figure 4: Internal structure of and DDR3 sample readout starts 24 clock cycles (4 bit config + 20 bit sample) • Initially, all modules are triggered simultaSCLK neously to read 20 bits from FIFO and prepend MOSI 4-bit DAC control header • low opens sample readout window, where sample is transmitted bitwise over 24 clock cycles, high triggers DAC to update analog output read MSB 0 1 read MSB-1 0 read LSB 1 1 SYNC min SYNC high: 48ns ACLK Load one word + hold time = 1µs 1 MHz sample freq. Figure 5: Timing diagram Event 5 • 54 MHz FPGA clock () is converted into 27 MHz SPI clock () CH 1 Event 1 Event 7 CH 2 Event 1 CH 3 Event 1 Variation of sampling frequency at arbitrary point in the waveform Asynchronous switching to next waveform across channels, while maintinaing synchronous sample output Signals have 20-bit precision and up to 1 MSPS sampling frequnecy • Using counters, module tracks of Samples for each event • External triggers and shared clock improve synchronization over multiple channels and FPGAs 4 Hardware Non-Isolated Region CH 4 for each sample and Number Isolated Region Power Regualtion Figure 2: Example sequence of events 20-bit DAC Connector time 3 FPGA Figure 3: FPGA Design overview ETH Port LVDS Receiver Power Management and Isolation Unit Power Regualtion Voltage Reference Figure 6: The DAC board • LVDS carried by shielded, twisted pairs in Ethernet cables are noise resistant • Isolation of grounds on PCBs remove ground loops • Credit card sized PCB requires no or very short wires to connect to devices 5 Conclusion • Metadata is written to , waveform sample data is written to DDR3 • DMA modules read DDR3 sample data for their channels, FIFOs ensure interruption-free data streaming • modules read metadata from to independently control the SPI modules We designed an analog signal generator platform which will be able to control a quantum gas experiment with high precision at high speeds. The FPGA-based digital hardware is verified in real-time, and the DAC cards will be verified together with the digital system when they are manufactured. In the future, we plan to extend the system to have more analog output channels at lower sampling rates and high precision data acquisition.