I. INTRODUCTION

Acceleration structures of high gradient and high repetition rate, and the control and measurement in high-power tests of high-gradient acceleration structures, are important in developing compact linacs [1-3]. The RF systems of linac-testing facilities at CLIC, Spring-8, SNS etc. benefit from the use of automatic control. These controllers are basically implemented in hardware to achieve signal acquisition, and the C, C++, JAVA or other software are used in data analysis [4]. But still, it is hard to find a breakdown location in real time. This paper is aimed at implementing the collected data waveforms in field programmable gate arrays to spot a specific breakdown location in real time. This saves time in solving problems of hardware malfunction of a linac. The conditioning process can be terminated soon, with increased safety and accuracy in conditioning the high-gradient accelerating structure.

II. HARDWARE

The PXI bus framework including the ICS-572B board and the RF front with LLRF (low level radio frequency) is shown in Fig. 1 [5-7]. It is a new virtual instrument released by National Instruments (NI) in 1997, and is suitable for testing acceleration structure [8]. The ICS-572B board is equipped with two ADC (AD6645) sampling frequencies of up to 105 MHz, which can simultaneously sample the input and reflected signals of the acceleration structure. It contains the high-performance of Xilinx Virtex-II FPGA. The amount of logic gates is up to six million, being in full compliance with the design requirements. It includes also a PCI bus interface in specifications meeting PCI2.2 at speed of 66 MHz/64-bit. Thus, the data communication between the underlying hardware and upper software is well achieved. According to the project requirements, two PXI-5122 data acquisition boards in 100 MHz sampling rate are added.

Fig. 1.Full-screen View

(Color online) The framework of the PXI.

As shown in Fig. 2, the test platform for high-power S-band accelerating structures is mainly composed of a 45 MW klystron and a SLAC accelerating structure [9, 10]. On the ICS-572B, the DAC output signal is of 25.6 MHz. The frequency of downstream RF front-end is 2997.924 MHz, i.e. the amplifier input. The maximum output power of the state amplifier, which provides input signal to the klystron, is about 1000 W. The output and reflected signals of the klystron, together with the input and output signals of the accelerating structure, and the ICT signal, are displayed and stored in the ICS-572B and PXI-5122 boards [11]. The user interface of the experimental platform is LABVIEW, from which we can set experimental data and display the waveforms. Once the reflected power increases suddenly and exceeds the threshold, the output of DAC module will be reduced to zero to protect the klystron from damage. This kind of protection function is also included in the user interface. In addition, the system can automatically record the reflected signal waveform when breakdown happens.

Fig. 2.Full-screen View

(Color online) Block diagram of the hardware of the testing platform.

III. SOFTWARE

The software includes three core modules: auto conditioning, interlock protection and breakdown locating. As shown in Fig. 3, I and Q value of the input and reflected signals can be obtained after down-conversion [12], AD sampling and FIR filter. Then, their amplitude can be obtained by using the CORDIC algorithm [13, 14].

Fig. 3.Full-screen View

(Color online) Block diagram of the software of the test platform.

B. Interlock protection module

This module detects the power of reflected signal and the vacuum in real time. When the number of breakdown in the accelerating structure exceeds the threshold and the vacuum degrades, the input power of the DAC shall be cut off immediately to protect the system from damage. Figure 4 is the execution procedure of the algorithm [16].

Fig. 4.Full-screen View

(Color online) Arithmetic diagram of the interlock protection.

C. The location of breakdown module

The algorithm is based on the difference between the falling edges of the incident and reflected signals to locate the breakdown position [17]. When a breakdown occurs, the subsequent RF signal can hardly pass through the acceleration structure back and forth. The system recognizes the falling edge of the incident and reflected signals inside the FPGA [18], and calculates the time difference t by means of the counter.

The group velocity equation in the S-band accelerating structure is written as Eq. (1):

$v_g=(\omega /Q)\left[l-\left(1-e^{-2\tau }\right)z\right]/(1-e^{-2\tau })\mathrm{,}$ (1)

where, l and Q are total length and quality factor of the acceleration tube, respectively; ω is the frequency; z is the displacement distance from the origin; and τ is the attenuation constant. Let A = ωl/[Q(1 - e^-2τ)] and B = ω/Q, Eq. (1) can be converted to Eq. (2):

$z=At/(2+Bt)\mathrm{.}$ (2)

In conditioning an S-band acceleration structure, with given values of z and Q, the group velocity can be calculated easily. Then, using Eq. (2), a lookup table can be made to correspond the relationship between the time when the microwave arrives at each cell and the distance z. So the actual position of breakdown in the accelerating structure can be located against the lookup table, with known value of time t. Figure 5 is the principle diagram of the location of breakdown.

Fig. 5.Full-screen View

(Color online) Principle diagram of locating the breakdown position.

IV. THE EXPERIMENT OF CONDITIONING S-BAND ACCELERATION STRUCTURE

The S-band acceleration structure was tested on the platform. In practical operation, the accelerating gradient is supposed to reach 18 MV/m with about 10 Hz petition rate. Before the experiment, it was desirable to provide the klystron with 340 W, 3 μs wide RF excitation signal according to design requirements, and to increase the voltage of the modulator to 5 kV. The width of output pulse of the klystron was set to 0.5 μs by adjusting the output of solid-state amplifier. The voltage was increased by 0.5 kV every 10 minutes. The vacuum interlock point was set to 1×10^-5 Pa and the recovery point was 5×10^-6 Pa. The output and reflected of the klystron, together with the input and output of the accelerating structure were detected by the ICS-572B and PXI-5122.

If the power was conditioned to 20 MW, the modulator output should be decreased to zero. In the experiment, the power could be conditioned to 20 MW at RF pulse width of 0.5, 1.0, 1.5 and 3.0 μs. Figure 6 shows the location of breakdown of these cases.

Fig. 6.Full-screen View

(Color online) The distribution of location of breakdowns in different cases.

Owing to limited hardware conditions, precision of the system is about four cells length, because the sampling frequency (105 MHz) of the ICS-572B board is not enough to locate the breakdown position to one cell. As we know, the effect of attenuation of the reflected signal along the structure may introduce a bias in the analysis. But in most cases, we can confirm the occurrence of breakdown when amplitude of the reflected signal exceeds the threshold value in spite of the existence of the attenuation.

The breakdown rates were monitored at 20, 30 and 36 MW output power. The results are given in Table 1. The breakdown rate is higher at output power of 36 MW. According to the breakdown distribution in Fig. 7, most of the breakdowns occurred in cells number 80–90. So we checked up those cells, and found Cell 89 was impaired.

Fig. 7.Full-screen View

(Color online) The distribution of breakdowns in breakdown rate testing.

V. CONCLUSION

The testing platform realized the detection of the breakdown position by means of a real-time, fast and accurate field programmable gate array (FPGA). The system not only locates the position of breakdown, but also displays the results in real time. If the breakdowns occur continuously in the same position, it is easy to find the problems and analyze the issues immediately, being of help in protecting the accelerating structures from damage.