Dip test: Rapid cathode activity evaluation of the klystron used in an accelerator

maintenance work is a recurring theme. In particular, life estimation by measuring the cathode activity of the klystron is important for accelerator operation because after long usage of the klystron, the cathode deteriorates. The standard method used is to measure the cathode emission current as a function of heater voltage (Miram plot), but it takes a few hours to obtain a full curve of emission characteristics for a tube and is hard work for large-scale accelerators. In this paper, a dip test method is described in detail as the rapid cathode activation evaluation in the Beijing Electron–Positron Collider (BEPC II) ring klystron and BEPC injector linear accelerator at the Institute of High Energy Physics, China. This method measures the emission dip from the short interval of a ﬁlament power cycle with high voltage remaining on under the high voltage application to the cathode, and evaluates the cathode activity, which relates to the cathode deterioration. In a continuous-wave klystron a 10 s interval for the off–on time, and in pulsed klystron, a 40 s interval is sufﬁcient to perform a dip test. Therefore, there are lots of advantages in using this method due to its rapidity and simplicity.


Introduction
The performance of the radio-frequency (RF) klystron is an important factor for the stable operation of a large-scale accelerator. Various klystron failures can result in interruption of operation. Failures such as the RF window, arcing in the gun, high-voltage (HV) ceramic, and vacuum failure are well known. But apart from these failures, klystron life is limited by the cathode. The electron gun in a klystron employs a thermionic cathode, and supplies the emission beam under the condition of limited space-charge current flow. After long usage, cathode activities are gradually deteriorated and the emission current decreases. In order to recover this decreased current, it is necessary to raise the cathode temperature. To maintain operating conditions, periodic evaluation is necessary. In order to determine the cathode characteristics, it is necessary to measure the current as a function of the cathode temperature (or equivalent values such as filament wattage, voltage, or current) under a fixed applied anode voltage; the measured curve is called a "Miram curve" (plot). One data point for a particular temperature requires about 20 minutes for a cathode with a diameter of several centimetres, to ensure thermal equilibrium. Therefore, a period of two to three hours is required to accomplish one Miram plot, and for a large-scale accelerator system it can take PTEP 2017, 113G02 U. N. Zaib et al. Generally, a klystron is very expensive so if tube life suddenly ends due to emission degradation, it is not easy to replace with a new one without having predicted the proper tube life. In the worst case, accelerator operation has to be stopped. Because of this, it is important to measure a Miram plot at least once per year to confirm the cathode activity level and the filament operating point. However, it is true that it is a hard job to complete this work for all tubes in a large-scale accelerator.
In this paper, we study the dip test for continuous-wave (CW) and pulsed klystrons, a rapid way to evaluate cathode activity proposed by KEK [1]. Due to its simplicity and rapidity, it is possible to perform this test during on-beam operation and to evaluate the cathode activity. The dip test itself is not a new idea [2]; during the klystron's high-voltage operation, filament power is turned off suddenly, and after a short interval it is turned on again. During the power-off time, the emission current goes down due to a decrease in the cathode's temperature, and after the power is turned on the emission current starts increasing. The depth of the decrease in current, called the "dip," depends on the time and the starting temperature of cathode before turn-off. As described in the next section, we can evaluate the cathode activity. The concept of the dip test is shown schematically in Fig. 1. Dip test information essentially includes the same information as derived from a Miram plot. A dip test requires only a few minutes, and then it is possible to determine the cathode activity. On the other hand, a full Miram plot comparison with a dip test result is useful to judge the remaining klystron life. As described in Sect. 3.2, comparison of already accumulated data obtained by successive dip tests is also useful, especially for scandate cathode klystrons [1].
The reason why this kind of simple method is not popular may be due to not knowing about the dip test, or such methods not being favored for turning off the filament power during the operation. Some may be worried about the possible trip of the interlock for the filament power during the operation. Some may worry about the over-voltage due to the mismatching between the line-type pulse modulator in the case when the pulsed klystron (popular in electron linear accelerators, or linacs) is used. However, if one chooses the dip test conditions properly, as described in this paper, one can perform the cathode activity evaluation using the dip test in a short time without any problems. It has been suggested that the dip test is a very powerful method [

The dip test and related issues 2.1. Emission characteristics of a thermal cathode
A dip test is an old technique and was invented long ago, as described in the previous section and in Ref. [2]. Many vacuum tube vendors may use this technique to evaluate cathode activities. Our purpose is to apply this technology to the high-power klystrons used in accelerators in order to measure the cathode activity in a short time and assist maintenance. To begin with, the characteristic of an electron gun using a thermal cathode is described. When a constant voltage is applied across the anode and cathode, the cathode temperature is raised and current flow is set up, determined by the cathode temperature and known as the temperature-limited (TL) flow (J TL ). J TL is a function of the cathode temperature defined by the Richardson-Dushman equation, expressed as [3] J TL = AJ 0 exp 1 kT where J 0 is the Richardson-Dushman current density, V d is the applied voltage, T is the cathode temperature in Kelvin, k is the Boltzmann constant, e is the charge of the electron, and ε 0 is the permeability of the vacuum. If the cathode temperature is further raised, the increase in current becomes smaller and finally the current becomes constant, which is called space-charge-limited flow. This completely constant current is called the full space-charge-limited (FSCL) current and is denoted by J FSCL . J FSCL is determined only by the electrode dimension and the voltage across the cathode and anode, as expressed by the Child-Langmuir equation: where P μ is the perveance and S is the area of the cathode. The characteristics of the emission current in response to the cathode temperature is known as the Miram curve, and is shown in Fig. 1 (left). In the TL operation region, the current increases with temperature. In the FSCL region, at high temperatures the current is independent of temperature. There is an intermediate smooth transition region, called the knee point. In actual cathode operation, Longo introduced the following expression using the operating current density, J OP [3,4]: In an actual observed curve, the knee of the transition (from TL to FSCL) curve predicted by Longo's formula is much broader than that found for most of the cathodes. Vaughan proposed the relation where n can be adjusted to provide the best fit to the data [3,5]. The operating impedance of the klystron is investigated next. Since in the pulse operation of the klystron impedance matching between the pulse modulator and the klystron has an important role to determine the applied voltage on the tube, a line-type pulse modulator is used. In the FSCL region, the klystron's operating impedance is expressed, from Eq. (2), as Therefore, when the applied voltage increases, the impedance decreases as the inverse square root of the applied voltage. In the TL region, the impedance is derived from Eq. (1): In the transition region, using Eq. (4), we obtain

Miram plot and dip test
As described in Sect. 1, stable operation of the klystron requires the proper control of cathode activity, which is effective in extending the klystron life. If this is not properly performed, you find an abrupt decrease in the output RF power due to the emission decrease during on-beam operation of the accelerator. In the extreme case, the klystron's life is ended due to this emission degradation. If the user does not expect this situation, sudden replacement of the tube or shortage of the tube may happen. Therefore, measuring the cathode activity by changing the cathode temperature is recommended at least once per year. This data is called a Miram plot: the emission characteristics as a function of cathode temperature (or equivalent values such as filament wattage, voltage, or current) [3]. The relation between the Miram plot and the dip test is explained in Fig. 1. A Miram plot is shown on the left of Fig. 1: the constant current part is the FSCL region, and the current decreasing part is the TL region; the transient region (the "knee") exists between the two regions. Usually, the operating temperature point is determined by a point some degrees higher than the knee position (for example, 10% higher), or a value specified by the klystron vendor. After long operation, movement of the TL curve toward a higher temperature is observed [3], which is shown as the three lines in Fig. 1 (left). This indicates a reduction in cathode activity; emission is degraded if the temperature remains the same for a long time. The operating point (a) shown on the left of Fig. 1 is not the proper point in the degraded Miram curve, and after confirming this relation, it is necessary to shift it to the higher temperature side. In order to get a full Miram curve, measurement of six to eight points is required; for each measurement, it is necessary to have heat equilibrium on the cathode for about 20 minutes (for a large cathode of diameter about 8 cm)-one Miram curve measurement needs two to three hours. The dip test is shown schematically on the right of Fig. 1 by considering the three cases of the Miram curve having the filament power set to point (a). In a dip test, the filament power is turned off during HV operation, and after a fixed short time the filament power is turned on again. During the filament's power-off condition the cathode temperature goes down, and once the temperature reaches the TL region the emission current decreases rapidly. After the filament is turned on, the current starts to increase as the temperature rises. The amount of the decrease in current is shown as the depth of the current or dip. The current at the bottom of the dip indicates the current at the temperature after the off-on interval. Therefore, if the first setting point (a) is located near to the knee point of the Miram curve, the setting point is considered to be inadequate. In this case a big dip is observed and you move the setting point higher; if the higher setting point is very near to the allowable limit of temperature depending on the cathode (i.e. around 1200 • C), the klystron's life would be nearly ended. The limit is determined by the evaporation rate of barium in the gun, the capability of the filament transformer, or from other causes.  The dip test was introduced to the S-band 50 MW klystron in the KEKB ring injector linac by one of authors (S. Fukuda) [1,7] because this linac used 60 klystrons and it was hard to measure all their Miram curves so frequently. As described in Sect. 3, this dip test required only a minute for the net time of off-on, and 10 to 15 minutes for full recovery. KEK performs this measurement on the maintenance day once per month, and all dip test measurements are performed twice a year. In this paper, we propose to perform the dip test properly during a possible interval in the accelerator operation, and expect readers to understand how the dip test is performed.

Emission characteristics of BI M-type and scandate cathodes
In this section, the Miram plot behavior of two typical accelerator klystron cathodes is described. J FSCL for the different cathodes show different features on the Miram curve. One cathode, called the M-type barium-impregnated (BI) cathode, has been improved by coating a metal film on the surface to reduce the work function. This is widely employed for various microwave tubes, especially for CW and long-pulsed klystrons. Another type is the scandate cathode, developed by Philips, whose emitting surface is composed of a layer of tungsten mixed with Sc 2 O 3 (∼5% by weight) [3,8]. This was employed in the S-band pulsed klystron, 5045, by SLAC [9], and it was later employed in many pulsed tubes. This latter type of cathode has a very low work function, while J FSCL is not constant in the space-charge-limited region; this results in different responses when we perform a dip test. In Fig. 2, Miram plots for the two types of cathodes are shown.

Dip tests and results
New dip tests were performed to confirm the procedure in the BEPC ring klystron hall and the BEPC injector linac at IHEP, China [10]. These two test results showed the different features depending on CW or pulsed operation and the type of cathode, as described above.

Dip test for a klystron using an M-type BI cathode driven by a constant-voltage power supply
The BEPC ring has been using the Thales Klystron, TH2163 A-1, with a frequency of 499. power supply adopting a pulse step modulation (PSM) system. This klystron employs an M-type BI cathode that shows almost constant J FSCL in the space-charge-limited region as shown in the theoretical formula (2) and in Fig. 2. Measured Miram curves for the klystrons used are shown in Fig. 3 (left), and the sequence of taking the Miram plot data is shown in Fig. 3 (right). It took two and a half hours to obtain the blue coloured points of Fig. 3 (left). In order to get a stable I b (emission current), it takes about 20 min for each setting point. In the dip test, we tried to look for an optimized off-on time from 50 s to 10 s by changing the filament setting (temperature) as shown in Fig. 4. The red line shows the klystron current and dips when the filament power is turned off, and the blue line indicates the filament voltage. Then we compare the corresponding dips and we can conclude that a 10 s off-on time is enough to judge whether the setting point of cathode operation is adequate or not. This corresponds to a roughly 10 W lower filament wattage change. Later, after 1-2 minutes, the 6/10 cathode recovered nearly to the initial state. In Fig. 5, the dip test results are summarized. Figure 5 (A) shows the Miram curve of Fig. 3 (left) and the three setting points indicated by arrows to measure the dip with a 10 s off-on time. Figure 5 (B) shows the measured dips for different setting points; we got very clear dip results corresponding to the position of the initial setting points. From this result, we can set the most appropriate initial operating point near to the position (a), or slightly higher, as in Fig. 5, showing a small dip. If the dip becomes bigger after a short period of operation, it indicates that the knee point is close to the setting point due to degradation. In Fig. 5 (C) there is replot of the Miram curves that predicts the future tendency of the cathode characteristics. The pink and green Miram plots show the degradation of cathode activity by 7% each with respect to the blue Miram plot. If the setting point is kept the same then the emission current decreases as pointed out with the pink and green dots. Therefore, the setting point should move to the points indicated by the pink and green arrows. With successive dip tests after a fixed period, the dip results change as shown in Fig. 5 (B). If the knee position reaches the high-temperature region and exceeds a certain temperature, Ba evaporation may occur which may shorten the cathode life; this is undesirable. When taking this data, cathode filament information and data on the klystron emission current were recorded by the EPICS system. Usually, in order to protect the klystron in the case of a filament power supply failure, an interlock is prepared. In the case of the BEPC ring, by setting the operation mode selection, the interlock is overridden. If one uses any other interlock system, it is necessary to disable this interlock during the dip test interval. From this test, we can conclude that a 10 s turn-off time is enough for the dip test to evaluate the cathode activity. Because of the flatness of J FSCL it is clear that any observation of a dip means to reach the knee point, and the dip depth is directly related to the correctness of the operation point. An off-on time of 10 s is valid for many CW klystrons of a few hundred kW, which employ a cathode with a diameter of 4 to 5 cm.  SLAC employed scandate cathodes in the well-known pulsed klystron 5045, many vendors followed this. The Toshiba/Mitsubishi S-band pulsed klystrons used in KEK and BEPC employ such a scandate cathode from Spectra-mat Corp [11]. Concerning with the 5045 cathode, there was an interesting report describing the beginning and ending emission characteristics for a klystron with a "good" and "bad" cathode. Clear data of Miram plots for the good and bad cathodes were shown [12], and it indicated that the initial selection of the cathode is important.

Dip test for a klystron using a scandate cathode driven by a line-type pulse modulator
The emission current from the scandate cathode in the FSCL region is not constant as shown in Fig. 2, and it phenomenologically corresponds to a small value of n in the Longo/Vaughan formula (4). These two features make the results of the dip test different from the ones described in the previous section. Of course, for the case of the pulsed klystron using the usual (M-type) BI cathode, the results are almost the same as described in the previous section. In the case of a scandate cathode, the dip test always gives a certain dip, but the dip depth apparently gets smaller if the setting point is far away from the knee point in the Miram curve. Well-informed judgment can be aided by the accumulated dip test data from the past, or by comparison between the Miram curve and the dip test. An important point which must be considered is that the applied voltage gets increased when the dip test is conducted. In many line-type pulse modulators, klystrons are operated under the condition of matching between the impedance of the pulse-forming network (PFN), Z PFN , and the corresponding load impedance of the klystron, Z FSCL L /N 2 (where N is the step-up ratio of pulse transformer), i.e. Z FSCL L /N 2 = Z PFN , which is defined by Eq. (5). During the dip test, the impedance of the tube is expressed by Eq. (5) at first, and changed to Eq. (7) or (6) depending on the situation. If this change induces a positive mismatch and a higher applied voltage is applied to the tube, there is the possibility of tube failure [13]. In Fig. 6, the klystron voltage and current as a function of filament voltage are shown in the case when a line-type pulse modulator is used (IHEP, KEK). Due to the scandate cathode, J FSCL is not constant, and the applied voltage increases in the TL region gradually. In order to avoid this excess applied voltage, the applied voltage is lowered by 10% before the dip test, with a suitable off-on time for the dip test in the BEPC and KEKB linacs. In both tests, the klystrons are the same ones, manufactured by Toshiba, E3780, which employ a scandate cathode of diameter 8.5 cm. A new dip test was conducted in the BEPC linac to investigate the dependence of the initial setting point, and to determine the optimum off-on time in the same manner as the CW klystron of the BEPC ring. In Fig. 7   voltage. These measurements are performed with different setting points of the filament voltage from V f = 19 V to V f = 15 V. We conclude that the most likely off-on time was 40 s, and it was shorter than the 60 s determined at KEK [1]. This indicates a difference of roughly 35 W. In Fig. 7 (C), dip data for the 40 s off-on time and different setting points, as indicated by the arrows in Fig. 7 (B), are summarized. Our operating point is V f = 17.2 V (W f = 296 W), and in this case the dip is 1 A. The acceptable range as judged from the dip is about 2.5 A, in comparison with the Miram curve in Fig. 7 (B). In this dip test, the klystron current and voltage waveforms were extracted from the waveform of the current transformer (pulse current waveform) and the capacitive divider (pulse voltage waveform) observed by a digital oscilloscope for which digital data was taken by personal computer. We used the Keysight oscilloscope, Dso-x 3034A. Every 0.1 s data was recorded, and to avoid noise or unwanted disturbance, we used an averaging acquisition mode of 64, which not only reduced noise but also increased the vertical resolution [14]. In the PC, direct data extraction to Excel is possible, or software such as Matlab is also available. If this data is accumulated in the PC or a central computer and checked against old data, it is obvious to judge whether a filament setting point is good or not, or the cathode activity is deteriorating or not. In this case also the interlock on the filament power supply under HV application is required to be temporarily disabled. It is easy to judge the klystron cathode activity from the accumulation of the past data comparison. shows two examples of data from KEK: a case of dips with small change (indicated by the blue circle: no problem) and a case of dips increasing with operation time (indicated by the red circle: the cathode is deteriorating and near to the replacement of the tube). The reason for the difference shown in Fig. 8 is different cathode activity in the two cases, therefore it is important to confirm the dip trend regularly. These methods are useful and applicable not only at KEK/IHEP but also at other institutes where klystrons are being used. From the KEK experience, it was confirmed that this dip test method reveals no harmful effects on klystron performance, and this method will be introduced in many institutes for the quick evaluation of cathode activity.

Conclusions
In this paper, a dip test is described, which enables us to evaluate cathode activity rapidly even during a small interval in accelerator operation. In the case of a CW klystron used in a synchrotron radiation source and colliding machine, the dip test requires only 10 s and full recovery requires five minutes; hence, very rapid evaluation is possible. In the case of a pulsed klystron with an output power of several tens of MW, the dip test requires 40-60 s or less after lowering the applied voltage by 10%, and full recovery requires less than 15 minutes. If this rapid method of evaluating cathode activity is performed, the current cathode activity can be understood. If the test indicates the tube is near the end of its life, further Miram curve measurement to assess the remaining life is performed once per year and this data can be used for a good procurement plan to maintain klystron activity.