Digital LLRF Test on the Renascence Cryomodule Trent Allison, Rama Bachimanchi, Curt Hovater, John Musson and Tomasz Plawski Introduction The Renascence cryomodule was the first opportunity for testing the newly constructed Digital LLRF FC (Field Control) chassis. During the previous test on the Renascence cryomodule we used the VME style digital LLRF system and results were presented in the JLAB technote [1]. Recent tests took place during June 2009, in the Cryomodule Test Facility (CMTF) and lasted for several days. Numerous measurements were completed, starting with SEL (Self Exciting Loop) mode to recover the cavity gradient then the switch between SEL and GDR (Generator Driven Resonator) or Microphonics Compensation mode and finally stability performance of the entire system. Although originally it was planned to test field and resonance control at gradient ~20 MV/m, the tested cavity gradient was limited to 9 MV/m due to low quench threshold. Cavity loaded Q was ~7.5 x 10^6 thus bandwidth 200 Hz. System Components Figure 1 shows a block diagram of the system. To replicate the master oscillator, an Agilent RF source was used as a 1427 MHz local oscillator and then synchronized to the 70 MHz LLRF system clock. Flexible firmware let the operator chose between a variety of operational modes, gains and filters (dynamic IIR filter). Figure 1: LLRF Test Diagram
SEL mode switch to GDR mode (I&Q) Figure 2 shows the inital switch between SEL and I&Q regulation. Though successful, closer inspection showed that the transition was not smooth. The green line indicates substantial forward power flow during the transition period of 4 ms. Figure 2: SEL to GDR transition The forward power spike occurs if the magnitude and phase are not aligned before the switch. This misalignment causes the I&Q control loops to compensate and thus spike the forward power until the cavity responds. The magnitude can easily be aligned by setting the SEL magnitude set point on the control screen. However, the phase rotates while in SEL mode due to cavity mistuning so it is not possible to align the phase from the slow control screen. A fast firmware algorithm was developed that monitored the rotating phase while in SEL mode and switched to I&Q regulation once the phase rotated into alignment. This algorithm eliminated the forward power spike and allows for a seamless transition. Field Control Tests Amplitude control was tested independently using an Analog Devices (AD8361) amplitude detector (linear in Volts) and an Agilent Vector Signal Analyzer. The ratio of the detected signal spectral density for the frequency bandwidth of 3.1 Hz -5 khz to its DC component gave the rms amplitude noise. Measured rms noise was 1.9x10-4. The microphonics background at the time of the test was around 1.2 deg rms or 2 Hz rms. Figure 3 shows the power spectral density of AC and DC components of the signal from the AD8361 for a closed loop gain of 64. While the amplitude noise was measured for the frequency bandwidth of 3.1 Hz- 5 khz, the phase noise bandwidth was 10 Hz-100 khz due to instrument constrains. Consequently the simple formula amplitude = amplitude x cos(detuning angle) cannot be applied for above situation.
Figure 3: Gradient stability For residual phase noise characterization, the Agilent E5052A Signal Source Analyzer was used. The measurement was made for I&Q regulation with the loop gain of 64. Figure 4 shows cavity phase noise during I&Q closed loop operation. Very low integrated phase noise of 51 mdeg rms indicates a proper operation of the LLRF FC chassis. Figure 4: Cavity field phase noise
Piezo Induced Microphonics To simulate an un-stiffened cavity, the cavity s piezo tuner (PZT) was used to initiate a 23 Hz microphonic detuning. During the test, a PZT amplifier drove a 300V/20 Hz sinusoidal signal which shook the cavity. Without compensation, cavity detuning was 23 Hz/45 deg. After I&Q feedback was turned on detuning dropped down to 0.5 deg rms. Figure 5 shows phase noise of the cavity field for I&Q regulation with the gain of 64. Amplitude AC and DC was measured for the bandwidth of 3.1 Hz 5 khz. The rms of the integrated AC component was 2.9 mv while the DC part was 1.1 V, yielding an amplitude error of 2.6x10-3. While this does not meet specification, it should be noted that the loop back cable was in place during these measurements. This caused an isolation reduction that limited the ability of the control loop to reduce the large amplitude excursions generated by the piezo. On subsequent bench tests (with similar detuning and no loop back cable), the amplitude specification was met. Figure 5: Phase noise of the cavity field for 23 Hz rms microphonics at 20 Hz 109 khz Sideband Spurs We observed spurious sidebands at a frequency of 109.375 khz and level -69 dbc while the system was running open loop. This became an issue in closed loop mode. The gain in the system increased the spur power to ~ 10 dbc, with the potential to rob power from the klystron (~1 kw, power not being applied to acceleration). It was found that the 109.375 khz -69 dbc spur was generated by polling the RF Board Temperature sensor which is located among the probe channel components.. This was eliminated by changing the polling scheme to quickly read the temperature every 75 msec with a long
silence between reads. Further investigation on the bench showed that the spur did not grow to a problematic size if the loop back cable was removed. Loopback cable The present RF board is equipped with a loopback cable that can be used to conduct close loop tests without a cavity (a DSP low pass filter in the firmware is used instead). We found that the loopback cable greatly decreases the isolation between the receiver and the transmitter channels. In a similar manner as the 109 khz sidebands, at gains above 60 it causes other spurious sidebands to develop eventually causing the klystron to waste power at undesired frequencies. Removing the cable reduces the spurious forward power signals to an acceptable level. This feature has been eliminated on future manifestations of the RF board. It should be noted that this problem was not solved until after the Renascence CMTF was completed and was discovered on follow up bench tests investigating for isolation issues. Summary The LLRF test on Resonance was a success. Field control specifications were met given CMTF background microphonics. We were able to verify performance of our 1497 MHz prototype and demonstrate technical feasibility of the LLRF architecture. The conceptual control algorithm was tested and for the most part operated as expected. The flexible digital LLRF controls allowed for modification of the firmware in situ to optimize controls. References 1. H. Dong, A. Hofler, C. Hovater, J. Musson, T. Plawski Digital LLRF Test on Renascence JLAB-TN-06-005