Status of The European XFEL Transverse Intra Bunch Train Feedback System

The European XFEL (E-XFEL) will have a transverse intra bunch train feedback system (IBFB) that is capable of correcting the beam position of individual bunches in the ~650us long bunch train, with a minimal bunch spacing of 222ns. The IBFB measures the beam positions with high-resolution cavity BPMs, and corrects the position of each bunch via stripline kicker magnets driven by class AB solid-state RF power amplifiers. The production of the IBFB BPM pickups is finished, and a pre-series version of the low-latency BPM electronics, including firmware and software, has been successfully tested with beam. After successful production and tests of prototypes, the series production of IBFB kicker magnets and RF power amplifiers is in progress. The IBFB feedback electronics hardware development is mainly finished, while firmware and software development is still ongoing. This report summarizes the latest design status and test results of the different IBFB system components. IBFB SYSTEM OVERVIEW Figure 1 shows the layout of the IBFB. The core of the system is located just upstream of the E-XFEL beam distribution kicker system and downstream of the collimation area. Four cavity BPMs (CBPMs) downstream of the IBFB (“downstream BPMs”) are used to implement a fast feedback loop, where two vertical and two horizontal stripline kickers can apply individual kicks to each bunch in order to correct the beam trajectory at the downstream BPMs to the desired position. A feedback loop latency of ~1μs is expected to be sufficient to damp all relevant perturbations. Figure 1: IBFB System. The necessary kick amplitudes are calculated by an FPGA board that receives the beam position data from the CBPM electronics via fast fiber optic links. The FPGA board applies the kicks via two digital-to-analog converter (DAC) mezzanines with four 16-bit 500MSPS DACs each. In order to apply corrective kicks to each bunch, the DAC mezzanines generate suitable output waveforms that are amplified by pulsed solid-state RF power amplifiers driving the stripline kicker magnets. Each stripline kicker has two amplifiers for driving its two opposite strips in push-pull mode, i.e. with opposite voltages. Four CBPMs upstream of the kickers (“upstream BPMs”) are used by the IBFB to predict the beam position at the downstream BPMs from the upstream BPM readings and DAC set values. This enables the IBFB e.g. to detect failures or drifts of the RF power amplifiers, variations of the beam energy, or to check and adjust the IBFB timing. The IBFB also receives the data of a dispersive CBPM in the collimator section (for beam energy measurement and kicker scaling factor adjustment) as well as data from all undulator CBPMs via digital multi-gigabit fiber optic links. In order to reduce the amount of cables, the undulator CBPMs of each of the three initial undulators (SASE1, SASE2, SASE3) are connected in a daisy chain, where only the first and last CBPM electronics of each chain is connected to the IBFB core system via singlemode fiber optic cables up to 1km length. When the first bunches of the E-XFEL bunch train (with up to 650μs train length and down to 222ns bunch spacing) arrive at the IBFB, it first corrects the trajectory using only downstream BPM data. As soon as the first undulator CBPM data is received by the IBFB, it fine-tunes the beam trajectory (if necessary), such that the following bunches reach the desired beam position in the undulators. Due to the long distance from IBFB to undulator CBPMs, the resulting latency of this correction is 4 to 10μs, depending on undulator and BPM location. However, the beam trajectory perturbations that occur between IBFB and undulators are expected to be either low-frequent (e.g. quadrupole magnet vibrations) or predictable, therefore it is sufficient to do this fine-tuning of the undulator beam trajectory once at the beginning of the bunch train, and then with a low correction bandwidth for the remaining part of the bunch train, in combination with the above mentioned fast (low-latency) feedback loop based on BPMs near the IBFB. In addition to this feedback-based correction, the IBFB will also perform an ___________________________________________ * This work was partially funded by the Swiss State Secretariat for Education, Research and Innovation SERI. TUPB064 Proceedings of IBIC2015, Melbourne, Australia ISBN 978-3-95450-176-2 492 C op yr ig ht © 20 15 C C -B Y3. 0 an d by th e re sp ec tiv e au th or s BPMs and Beam Stability adaptive feed-forward correction of the beam positions, by predicting perturbations that are reproducible or varying slowly from bunch train to bunch train, such that the feedback loop only has to correct the unpredictable perturbations. BEAM POSITION MONITORS The performance requirements to the IBFB BPMs are identical to the requirements of the undulator BPMs [2], i.e. <1μm resolution and weekly drift at 100-1000pC and ±0.5mm range, with the exception that the latency of the BPM electronics must be so small that an overall feedback loop latency of ~1μs can be achieved. It should be noted that the electronics of all E-XFEL cavity BPMs will have the required low latency, thus allowing to use also cavity BPMs in undulators and transfer lines for intra bunch train corrections. Figure 2: E-XFEL cavity BPM electronics (MBU = Modular BPM electronics), with two cavity BPM RF front-ends (RFFEs, top) and FPGA carrier board with two ADC mezzanines (bottom) [2]. The high-resolution BPMs in E-XFEL have two types of 3.3GHz dual-resonator cavity BPM pickups: One type with 10mm aperture used in the undulator intersections, and one with 40.5mm aperture used in the warm beam transfer lines, including the CBPMs of the IBFB core system. All these CBPMs have basically the same electronics, firmware and software. However, since the 40.5mm CBPMs have a position cavity sensitivity (in units of V/mm/nC) and Q factor comparable to the 10mm CBPMs, their position cavity signal may be up to four times higher for the same beam charge when the beam is close to the aperture limit. Therefore the non-IBFB CBPMs have additional attenuators at their RF front-end (RFFE) inputs, where the necessary attenuation to protect the RFFE from overvoltage is still low enough to reach their specified single-bunch position RMS noise of <10μm RMS for a measurement range of ±10mm. The 40.5mm BPMs used by the IBFB core system for the ultra-fast feedback loop require a resolution <1μm RMS for a desired range of ±1mm. Therefore their RFFEs do not have additional attenuators, but a special input protection circuit that clips the input signals when they reach a certain voltage. The IBFB CBPMs also determine the bunch charge range where the IBFB can be used. At very low bunch charge, the noise of the BPMs scales inversely with the bunch charge, and so does the noise modulated by the IBFB onto the beam (for fixed settings of the feedback parameters). Although the E-XFEL IBFB and undulator cavity BPMs have been specified to reach <1μm RMS noise only between 100pC and 1000pC, PSI designed the BPM electronics to reach this performance also for much lower charge down to about 20pC (see Figure 3). Moreover, the noise modulated by the IBFB onto the beam can be reduced if necessary by changing the feedback algorithm parameters to reduce the feedback loop bandwidth. However, this will of course also lower the frequency up to which random perturbations can be corrected, and reduce the correction efficiency at lower frequencies. Therefore the IBFB will allow to adjust the feedback algorithm settings to the bunch charge, either manually or automatically by measuring the bunch charge with the IBFB BPMs and adjusting the feedback algorithm parameters accordingly. Figure 3: Position resolution (RMS noise) of E-XFEL cavity BPM, measured by correlating data of several BPMs, for different bunch charges [3].