SPEAR BTS Toroid Calibration

Transcription

1 SPEAR BTS Toroid Calibration J. Sebek April 3, 2012 Abstract The Booster to SPEAR (BTS) transport line contains several toroids used for measuring the charge that is injected into SPEAR. One of these toroids is used solely for instrumentation. Three others are used to measure charge as part of the SPEAR beam containment system (BCS). This report documents the calibration of these toroids. 1 Introduction The SPEAR storage ring complex consists of a radio-frequency (RF) gun, a linear accelerator (linac), a Booster ring, and the SPEAR storage ring. Transport lines from the gun to the linac (GTL), the linac to Booster (LTB), and the Booster to SPEAR (BTS) connect these parts of the complex. During each 100 ms period when beam is injected into SPEAR, one bucket of charge is accelerated in the Booster, transported to, and injected into SPEAR. One quantity of interest in all of these transport lines is the total charge that goes down the transport line. This charge is sensed by toroids. The beam acts as the primary single turn of the toroid and a winding acts as a secondary that outputs its current. This output current is then processed electronically and recorded. This report documents a calibration of these outputs. One of the toroids is a commercial unit, the Integrating Current Transformer (ICT) from Bergoz Instrumentation. Its signal is amplifed and then measured with a commercial charge to digital conversion module from Caen Electronic Instrumentation. The other three toroids were assembled in house. Their processing electronics was also designed and built in house. The calibration assumes that the Bergoz ICT meets its specifications. We first used laboratory test equipment to calibrate the signal processing used on the ICT signal. We then measured this signal with the Caen module. This calibrated the ICT signal channel. In order to calibrate the other toroid signals we took data acquired during a fill from 0 ma to 350 ma. Data from this fill, spanning a time of more than 6 minutes, was then used to calibrate the other toroid signals against the ICT data. As a cross check we also compared the ICT with the measurements obtained from the Parametric Current Transformer (PCT), also from Bergoz Instrumentation. The PCT measures the total current in SPEAR. We compared the differences of the PCT output every second with the charge that was measured coming down the BTS line. The cross check compares the differences of the PCT are close to, but slightly less than the current measured in the ICT. This ensures that no more beam gets accumulated into SPEAR than passes down the transport line. 1

2 amplitude (mv) 5 typical toroid output with bias removed time (ns) Figure 1: ICT output for typical BTS bunch charge. 2 ICT Calibration 2.1 ICT Signal The signal out of the ICT, for typical fill rates, terminated into 50 is shown in Fig. 1. To obtain a cleaner signal, 1000 consecutive pulses were stored, their offset removed, and their outputs were averaged. This average is shown in Fig. 2. The turns ratio of the ICT transformer is such that the output current of the ICT is calibrated to be 1=10 of the beam current. 2.2 ICT Signal Processing Calibration This signal is then amplified using a commercial radio frequency (RF) amplifier. The output of this amplifier is then measured by the Caen charge to digital converter. In order to calibrate the electronics downstream of the ICT an output signal was programmed on an Agilent 81110A pulse pattern generator. In order 2

3 amplitude (mv) 2 typical toroid output averaged over 1000 pulses with bias removed time (ns) Figure 2: Average of ICT data from 1000 consecutive pulses. 3

4 Setting Delay Width Leading Edge Trailing Edge High Level Low Level Output Value 0 ns 19 ns 10 ns 25 ns 0 mv 620 mv Complement Table 1: Settings of Agilent 81110A pulse pattern generator for electronic calibration. to simulate the toroid pulse as closely as possible the generator was programmed to the settings in Table 1. The generator is designed to produce signals that are significantly larger than the typical ICT pulse. Therefore a 32 db attenuator was placed on the generator output in order to allow the generator to produce clean, controllable signals that were the appropriate amplitude at the amplifier input. 32 db of attenuation is a reduction in signal amplitude of approximately 39:8. The charge in the pulse was calculated by first dividing the voltage by 50 to obtain the pulse current, and then integrating this current to obtain the total charge.figure 3 displays the pulse pattern generator output and a typical toroid pulse. The RF amplifier used in the circuit is capacitively coupled on both its input and output. This coupling distorts the pulse shape; this distortion is evident in Fig. 4. The sharp edges of the pulses are apparent in these waveforms, but the long tails show the discharge time of the circuit capacitors. For this measurement the output of the generator fed the 50 capacitively coupled amplifier input. One channel of the oscilloscope was set to high impedance and sampled the input signal using a tee. The amplifier output was connected to a second channel of the oscilloscope, which was terminated in 50. One can see that the capacitors on both the amplifier input and output add to the distortion. The generator amplitude setting was matched to the amplitude of a typical ICT pulse. In order to calibrate the electronics over amplitudes that cover the expected BTS toroid operating range, data was acquired over two octaves of generator amplitude, with the central value being the one that matched the typical ICT pulse. For each generator amplitude the input charge was calculated by converting the input voltage to current, then integrating that current between the sharp edges of the pulse, each of which is identified by the red x plotted on the waveforem in Fig. 4. Two numbers were calculated for the output current. The first number is obtained by integrating between the sharp edges of the output pulse, each of which is again identified by the red x. The second number integrates the ouput current for 100 ns. This time is the gate width for the charge to digital converter that is set by the control system. The leading edge of the gate arrives 17 ns before the pulse arrives, so the charge to digital converter integrates for 83 ns after the arrival of the pulse. The end point of this integration is identified by the red circle in Fig. 4. Figure 5 shows that the amplifier response is linear over the range of the test. After the amplifier response was characterized, its output was reconnected to the input of the charge to digital converter. The generator was stepped through the same set of amplitude settings and the measured output of the charge to digital converter was recorded in the history buffers, as displayed in Fig. 6. Note that the charge to digital converter is bistable when its input is about 55 pc, the third lowest step of the 4

5 amplitude (mv) 2 test pulse approximation to toroid pulse test pulse typical data time (ns) Figure 3: Pulse pattern generator waveform used for calibration of ICT processing electronics. 5

6 amplitude (mv) 50 V gen = mv Q in = Q out = Q outcaen = pc 0 50 amplifier input (X10) amplifier output time (ns) Figure 4: Typical amplifier input and output signals. The red "x" marks show when the pulse ends and the distortion due to the capacitive coupling occurs. The red circle marks the time when the charge to digital converter stops integrating the charge. 6

7 output charge (pc) output charge (pc) 50 toroid calibration (matched times) Q out = *Q in input charge (pc) 50 toroid calibration (83 ns integration) Q out = *Q in input charge (pc) Figure 5: Calibration of the BTS toroid amplifier. The upper trace integrates the output only over the duration of the pulse. The lower trace integrates over the 83 ns that the charge to digital converter uses. 7

8 charge (pc) 160 history data for calibration pulses time (s) Figure 6: Output of the charge to digital converter as the calibration generator amplitude was stepped through two octaves. calibration measurement. At this value the integrated signal area is at the transition point between two ranges of the converter and it seems to oscillate between the values. For the purposes of fitting this data, we averaged only over the converter values in the higher of the two states. The data was averaged for each step and fit to the equivalent toroid output produced by the generator. This fit is shown in Fig. 7: From this we see that the ICT detection electronics is linear over the two octaves of the calibration. Further, since the offset is relatively small, we do not introduce much of an error by neglecting it. Our calibration shows that the history data under reported the actual toroid signal by approximately 12%. To introduce this calibration into the system we will increase the linear scale factor of the ICT by 1: The current value of this linear factor is 0: and is stored in the process variable BTS:ToroidChargeFast.A. This value will be increased to 0:

9 history data (pc) 160 history data vs toroid output Q history = *Q toroid toroid output (pc) Figure 7: Fit of ICT calibration of history data vs calibration signal. 9

10 3 Beam Containment Toroids Calibration Now that the ICT has been calibrated, we can use it to calibrate the toroids used as average current monitors (ACMs) in the beam containment system (BCS). We took data over a nearly six minute span from the history buffers that recorded a fill from 0 ma to 350 ma. The relevant data is presented in Fig. 8. It includes the ITC output, ACM signals and the fill rate as calculated from the SPEAR parametric current transformer, which measures the stored beam in the machine. There are three ACMs and there are two signals plotted for each ACM. One signal is the digitized value of the averaged charge measured by each ACM electronic module; the other is an averaged value of the integral of each pulse of the amplified toroid signals, which are buffered values of the input signals to the ACM electronics, as calculated by summing over the digitized samples of these pulses. When this data was taken, there existed a calibration of the ACM outputs, but not of the digitized samples of the inputs. The rate and efficiency of this fill was typical, so we assume that we can neglect any loss of charge down the BTS transport line. We will calibrate the ACM data by performing a linear least squares fit of the ACM in question against the ICT. In order to correctly align the data from the history buffers we perform a correlation of the signal in question with the time-reversed signal of the ICT. A signal such as this should have random deviations about its mean value. The correlation between this signal and the time-reversed reference signal should be close to zero when the signals are not properly aligned since each value of the correlation is a sum of products of numbers that are equally likely to be positive or negative. If the numbers are uncorrelated, the individual products will equally likely be positive or negative. However at the time for which the signals are properly aligned the individual products are most probably positive. The result of the correlation of the ICT signal with its time-reversed signal is shown in Fig. 9. In order to amplifiy the relative signal strength at the proper correlation time, we subtracted off the mean of the signal. We see that this method clearly shows the proper time alignment of the signals. Figure 10 shows that the correlation of the ACM signal calculated from the digitized pulsed signal, after it was properly aligned with the ICT signal. A comparison of the similarities of the correlation functions in Figs. 9 and 10 show that the signals are well correlated, as one can also see by examining Fig. 8. Correlation plots for the other ACM signals and the SPEAR fill rate calculated from the parametric current transformer are very simiilar in appearance to Figs. 9 and 10. The least squares fits of the ACM readouts to the (corrected) ICT readings are plotted in Fig. 11. Since the ICT data is assumed to be more accurate than the ACM data, we selected the ICT data as the independent coordinate. The equations from the fits are in the form q ACM = aq ICT + b: We need to invert this equation in order to find the coefficients necessary to modify the current ACM calibration constants in the control system; the coefficients that we require are obtained from q ICT = 1 a q b ACM a : The corrections that could be made to the control system calibration factors are given in Table 2. However since the difference in this calibration is small, within the specifications of the electronics, and the values 10

11 charge (pc) 300 di/dt and toroid signals for fill on feb di/dt*100 acm1 acm2 acm3 acm1(psm) acm2(psm) acm3(psm) bts toroid time (s) Figure 8: Data of the ICT, ACM outputs, and the SPEAR fill rate during a typical fill. ACM a b 1/a -b/a 1 0: : : : : : : : : : : : Table 2: ACM Calibration 11

12 amplitude 8000 matched filter (length = 677) of BTS toroid with itself peak at index Figure 9: A correlation of the AC component of the ICT signal with its time-reversed signal. 12

13 amplitude 2 x 10 4 matched filters (length = 677) of BTS PSM ACMs with toroid ACM1 pk = 339 ACM2 pk = 339 ACM3 pk = index Figure 10: Correlation of the AC component of the ACM signal with the time-reversed ICT reference signal. 13

14 acm charge (pc) acm charge (pc) acm charge (pc) 100 acm 1: Q acm = *Q tor toroid charge (pc) acm 2: Q = *Q acm tor toroid charge (pc) acm 3: Q = *Q acm tor toroid charge (pc) Figure 11: Calibration of ACM readouts based on the calibrated output of the ICT. 14

15 ACM a b 1/a -b/a.a.b 1 2: : : : : : : : : : : : : : : : : : Table 3: ACM Calibration currently presented to the control system are the values displayed on the front panel of the ACM chassis, these changes will not be made to the control system. Figure 12 displays the fit of the output of the digitized signals of the ACM input signal to the calibrated ICT signal. Again, in order to obtain the coefficients that we will enter into the control system, we must invert the linear equations obtained from this fit. The signals to be digitized are rather low in amplitude yet are in a relatively noisy environment. In order to minimize the effect of this noise on the digitized signal, the samples taken before the current pulse are used to calculate a baseline. This baseline is then subtracted from the current pulse. We can change the linear term of the digitized output by scaling the raw signal, but any offset we add to the raw signal will be subtracted out as part of the baseline subtraction. We therefore choose to not put any scaling into the raw signal; instead we correct for the gain and offset after the calculation has been completed. The values of the gain and offset for the raw signal were 10 and 0, respectively; we reset the gain, which for ACM 1 is in process variable BTS-ACM1:Conv.B to 1 and left the offset unchanged. The process variables that we changed are those associated with the calculational record. For ACM 1 the process variables for gain and offset are BTS-ACM1:ChargeFst.A and BTS-ACM1:ChargeFst.B, respectively. The calibration coefficients and the process variable values, taking into account the reduction in gain of the raw signal, are given in Table 3, 4 Comparison with Fill Rate Finally we plot the fit of the fill rate to the calibrated ICT signal. The fill rate was calculated by taking the differences of signals of the SPEAR parametric current transformer. The slope of this curve is the injection efficiency. Our consistency check is that this rate is less than unity. 15

16 acm charge (pc) acm charge (pc) acm charge (pc) 300 psm acm 1: Q acm = *Q tor toroid charge (pc) psm acm 2: Q = *Q acm tor toroid charge (pc) psm acm 3: Q = *Q acm tor toroid charge (pc) Figure 12: Calibration of the ACM readings from the pulsed signal monitoring based on the calibrated ICT signal. 16

17 stored charge (pc) 95 stored charge = *Q tor toroid charge (pc) Figure 13: Fill rate, calculated from the SPEAR parametric current transformer, fit to the calibrated ICT signal 17