Taiwan Experiment on Neutrino

  Read in   |   Print Print

"Time resolution of NI FPGA hardware totally fits our research. Timed-loop function, which can strictly execute premeditated set periods combined with a 50 MHz ADC adapter, can easily differentiate time differences between cosmic rays passing through two detectors."

- Feng-Kai Lin, Institute of Physics, Academia Sinica

The Challenge:
Developing a complete, complex algorithm under the nanosecond scale to precisely measure relevant physical quantities when they hit a detector, which is easily affected by ambient gamma rays and high-energy cosmic rays rather than neutrinos.

The Solution:
Using the PXI Express platform and FPGA modules to deploy algorithms like random trigger and reset inhibit, creating a high-speed operation that enables us to take 50 analog signals parallel and finish complicated operations within 20 nanoseconds and with a dead time of less than 0.5 percent.

Author(s):
Feng-Kai Lin - Institute of Physics, Academia Sinica
Henry Tsz King Wong - Institute of Physics, Academia Sinica

Introduction

The main challenge of this project is to develop a stable DAQ system under a high trigger rate, random trigger source, and long-term operation. Meanwhile, the data-lost rate must be low enough to reduce statistical error in the numerical analysis procedure, and keep flexibility for new algorithm or real-time pulse-shape discrimination in the future.

Over the past few decades, the acquisition system commonly used by high-energy physics academia can barely handle high trigger rate because of low-speed buses (VME) and too many suppliers that provide different operation environments. Plus, the mainstream software language is C, relevant libraries are very complicated, and professional experts must handle source code. Different research groups own different hardware and they cannot easily share the programs to control them. Fortunately, the European Organization for Nuclear Research (CERN) has taken the lead by using LabVIEW software. Although still not widely familiar in academia, this move has encouraged researchers around the world to change their DAQ hardware configurations.

TEXONO was established in 1997 as the first high-energy physics experiment team in Taiwan. The team’s preliminary goal is to precisely measure neutrino mass, which is notoriously difficult to do because of extremely low cross section to normal material. Therefore, we needed to put our experiment site close to a special place that can generate high neutrino flux. The core of a nuclear power plant releases huge amounts of neutrino through beta decay, which makes it an ideal site for this research. The main detector of this experiment is a high-purity germanium detector, which is known for super low background noise and super high energy resolution. A sodium iodine scintillation detector surrounds the germanium detector as a veto of ambient radiation. We place these two detectors in a 1 m shielding made of copper, lead, and polymer to block gamma and neutrons from outside.

Figure 1. High-purity germanium with the front end surrounded by copper brick to block ambient radiation. Right hand side is dewar flask for lowering germanium crystal temperature. Figure 2. Sodium Iodine detector brazen part is brass, which coats the detector. Upper part is photomultiplier tube and the cables are signal cable and high voltage cable.

Figure 3. Shielding, with the white part as polymer for blocking neutrons.

Plastic scintillators surround the shielding and two to four photomultiplier tubes attach to it to detect the most possible trigger source—cosmic rays. There are 16 plastic scintillators and 40 photomultiplier tubes.

Figure 4. Cosmic ray detector in which the upper parts connected with cables are photomultiplier tubes and the lower parts are plastic scintillators.

The whole site is 28 meters from the #1 reactor core, Kuo-Sheng nuclear power station. There are five output signals from main detector, two from sodium iodine detector, and one from each photomultiplier tube. Combined with other auxiliary signals, there are a total of 50 signals that need online calculation and recording for each trigger with a rate of at least 40 times per second. Because of different timing properties of the detectors, there are 60 MHz, 200 MHz, and 2 GHz digitizers on site (two for each one). The FPGA judges the trigger discrimination. If the condition is a positive edge trigger and is continuously 140 nanoseconds higher than threshold, it is considered a physical event and the FPGA sends several TTL signals to activate digitizers.

Figure 5. The DAQ system in the control room. The uppermost white modules in the crate are amplifiers and an extension of the main detector that cannot be replaced. The modules in the crate below are analog modules for auxiliary purposes. The FPGA replaces their functions.

CAEN, which is known in this field, provided the previous DAQ system. The modules communicate with the controller through VME buses, a familiar technology in the world of high-energy physics. But its low bandwidth goes against data transfer under high trigger rates, and the onboard memory of the digitizer is not big enough to handle sudden trigger rate bursts. The long history of the VME bus system makes it difficult to control hardware from different suppliers.

We have nearly 50 detectors in our experiment structure. They have different trigger conditions because of different signal types. Very few of them have obvious signals at one time and others are inconspicuous. A more common condition (99.9 percent of events) is that all signals from detectors are unobvious, some of them just slightly higher than threshold. We needed a real-time algorithm to determine if this small signal (a particle-deposited small amount of energy to detectors) is a real physical event or a fake signal due to a disturbance of electric noise in the cable.

Figure 6. Signals from the main detector should be a Gaussian-like pulse. It is hard to
determine if a small signal, which is comparable with noise, is triggered by noise or triggered by a real physical event in which a particle only deposited tiny amount of energy to the detector.

Figure 8. The spectra of the experimental site shows raw spectrum is all data without any treatment.
Green shows events without cosmic ray signal (not induced by cosmic ray, which is neutron-rich). Red spectrum shows signals without Sodium Iodine fired, which is a gamma-rich event. Blue spectrum is a combination of residual background plus neutrino events. Noise edge is around 350 eV, much lower than analog modules (450–500 eV).

In traditional ways, scientists used many analog modules to copy signals, judge if signals pass threshold or not, and correlate any timing between two signals (anti-Compton and more). We can use these analog modules for big signals. However, our research project focuses on extremely low-energy events, which means that only signals compatible with thermal noise interest us. The more analog modules the original signals go through, the higher the noise component grows.

This noise component can easily overwhelm a preliminary detector signal, which is already very small. The FPGA modules provided by NI have 20 nanosecond, 14-bit time and voltage resolution to monitor signals all the time and immediately execute complicated algorithms (random trigger, reset inhibit, TDC, and more). Previously, we could only perform this using lots of analog modules to determine if we need to digitize and store an event to hard disk. We have eliminated all the vulnerable, expensive, heavy, calorifacient analog modules. Moreover, the FPGA turns the original analog signal of a detector to a digital signal right after it is created. Even if we copy it or perform a logic operation to it, the preliminary pulse shape is not affected. This is very helpful to projects in pursuit of extremely small signals like TEXONO. The digitizer has enough onboard memory space to provide a big buffer to accommodate sudden trigger bursts so that the program need not handle stacked data by sacrificing systematic resources. We use parallel functions in LabVIEW that can separate data acquisition (producer loop) and data storage (consumer loop) to utilize all hardware resources completely.

Figure 7. Traditional analog modules are shown.

LabVIEW Software Advantages:

  1. Simple User Interface: Except for the above description of parallel operation, users can easily input many parameters with the LabVIEW user interface. Users can easily change them during the running period, which they cannot do with traditional text-based languages.
  2. Intuitive Output Information: Showing a trigger pulse shape on screen by some LabVIEW built-in functions is good for remote monitoring and real-time modification. For example, an unstable power supply can affect sophisticated instruments. Traditional languages can only provide numbers, making it hard to quantize signal stability at any moment. This is not a problem with LabVIEW.

 NI Hardware Advantages:

  1. Closely Follows Industrial Trends: The hardware provides various chassis for different bandwidths of PXI Express buses for painless integration.
  2. FPGA Simplifies Experiment Configuration: Users can quickly address and modify complex and real-time logic operation. Another benefit of excluding traditional analog modules is that we can greatly reduce the number of cables, have less crosstalk, and save on maintenance costs.
  3. Easy-to-Extend FPGA: Another merit of the NI FPGA is that users can extend easily by simply adding another FPGA module. The I/O boxes can have sixteen more channels to receive detector signals. It’s flexible for adding output detectors in the future.
  4. Easy to Compile: More output channels means a more complicated algorithm. In FPGA configuration, a more complicated algorithm means complexity of program rather than outer (analog) hardware. It saves laboratory space and refrigeration budget. Besides, trial-and-error is inevitable for cutting-edge research. One-key operation compile greatly shortens the process of algorithm verification.
  5. Precise and High-Time Resolution: Time resolution of NI FPGA hardware totally fits our research. Timed-loop function, which can strictly execute premeditated set periods combined with a 50 MHz ADC adapter, can easily differentiate time differences between cosmic rays passing through two detectors.
  6. Enough Onboard Memory: We need sufficient memory to handle trigger rate burst. NI provides a total solution through the Standard Service Plan (SSP). The NI professional team in Taiwan has helped us solve many delicate problems we faced when building a perfect DAQ system. For example, if the controller doesn’t read data stored in a circular buffer of the digitizer when one “circle” is complete, the error message shows up so that the FPGA stops monitoring the signal and the controller must focus on digesting data in onboard memory. We were troubled by this for a long time, but the problem was solved by the NI service team. Moreover, my knowledge of FPGA is from an FPGA course from NI. I’m not from the electrical engineering field, so it is difficult for me to read a textbook and the information on the Internet is incomplete. The NI FPGA course helped me understand this tool in a comprehensive way.

The ultimate goal of this project is to get all the signals that pass the threshold and FPGA logic selection as far as possible. Execution speed is our main consideration. However, we found that the speed of LabVIEW executed on the Windows OS is not stable. We plan to use a real-time system in the future to eliminate all the uncontrollable background programs and gain more system resources.

Author Information:
Feng-Kai Lin
Institute of Physics, Academia Sinica
Taiwan
fklin@phys.sinica.edu.tw

Bookmark and Share


Explore the NI Developer Community

Discover and collaborate on the latest example code and tutorials with a worldwide community of engineers and scientists.

‌Check‌ out‌ the‌ NI‌ Community


Who is National Instruments?

National Instruments provides a graphical system design platform for test, control, and embedded design applications that is transforming the way engineers and scientists design, prototype, and deploy systems.

‌Learn‌ more‌ about‌ NI