1. Investigation of the Resonant Transformer response at HTP (and a little bit of Fast Current Transformer response) Measurements with Krypton and proton beams for measurements at the detection limit and evaluation of new FESA DAQ system H. Bräuning A. Reiter, M. Witthaus, H. Reeg

2. Two beam times: SN March 21st, 2014 & NF April 29th, 2014 Ion species: –Krypton, 300 MeV/u, q=+33, particles/spill: 1E6 – 1.5E9 (3.3E7 – 4.95E10 e)H. Bräuning –Protons, 300 MeV, q=+1, particles/spill: ~3E6 – 1.5E9 A. Reiter RT with modified amplifier: –Peak Detector analog output (existing signal for GSI control system) –Analog output for damped oscillation (new output added to amplifier unit) –Ranges: (1, 10, 100, 1000) nC RT DAQ: –VME system with CERN/GSI timing converter –Struck 16-bit ADC (100 MSa/s) –10 Volt input range (1 ADC ch. = 0.15 mV) FCT with attenuator/amplifier stage (originally designed by H. Reeg, but never installed!) –Amplifier: Miteq AU-1423 (45.2 +/- 0.2 dB; 0.3-500 MHz; -3 dB cutoff at 2 kHz; noise figure =1.10-1.25) –Attenuator: MTS PAS-63/1000-5401, 63 dB (+/-0.3 db) att. in 1 dB steps, 3.5 dB insertion loss; 1.7:1 VSWR) –Coaxial cable from HTP to AP container: RG214, length ~ 75 m FCT DAQ: –VME system with CERN/GSI timing converter –Kicker tank signal as timing reference –Struck 10-bit ADC (1.25 GSa/s for each of the 4 channels) –+/- 1 Volt input range (1 ADC ch. = 1.95 mV) Software and DAQ system by HBr: –FESA class (C++) –GUI (Java) Summary of Beam Times & Experiment Setup

3. Minimum detectable charge limit Evaluation of analogue signal and comparison to peak detector output Developtment and test of new FESA DAQ system Open issue: Comparison with FCT signal This is a difficult task since many parameters can affect the current integral, especially frequency dependence of attenuator, amplifier, coaxial transmission, eddy currents, etc. The same care is needed for the RT electronics chain, although easier. A previous test resulted in a ~7% discrepancy between FCT and RT. The current data at low particle number agree within ~10%, if the measured gain is applied (see slide 32). A more detailed effort is only worthwhile, if the new FAIR transformer with Bergoz FCT and RT is available including a precision calibration circuit (1% accuracy). One possibility might be the re-commissioning of SIS18, if beam is available at HTP. Purpose of Investigation

4. Part I – Krypton data

5. Amplitude of RT analog output (damped oscillation) and peak det. in good agreement March 2014: Krypton, 300 MeV/u, q=+33 High intensity (around 6E8) Q~0.75 nC 1.4E8 Kr (4.6E9 e) Q~0.2 nC 3.8E7 Kr (1.3E9 e)

6. March 2014: Krypton, 300 MeV/u, q=+33 1E6 particles (~3E7 charges): Peak det. unreliable, while RT analog output is relatively stable Q~25 pC 4.7E6 Kr (1.5E8 e) Q~5 pC 9.5E5 Kr (3.1E7 e)

7. Part II – Proton RT data

8. GHTPDT1 in Range 4 (1 nC) for all runs Run 1: 1.5E9FCT signal attenuated, limited use! See slides at end. Run 2: 1E8FCT signal attenuated, not useable! Run 3: 3E7FCT attenuation = 0 dB from now on! Run 4: 1E7 US2DT5 = 2.2 µA Run 5: ~7.2E6US2DT5 von 2.2 µA auf 1.6 µA, scale to Run 4 Run 6: ~4.5E6US2DT5 von 1.6 µA auf 1.0 µA Run 7: ~2.7E6US2DT5 von 1.0 µA auf 0.6 µA Useful numbers:1 nC = 6.25E9 e (or protons) 5000 mV full scale = 1 nC (RT amplifier calibration) 1 mV = 1.25E6 protons 1 ADC ch. = 1.9E5 protons (16 nominal bits, +/-5 Volt) Expected resolution 1% ~ 6.25E7 e (protons) ~ 50 mV ~ 328 ADC ch. April 2014: Proton, 300 MeV, q=+1, Run 7

9. Blue: Peak detector output Red: Raw ADC data Green: Filtered ADC data BW = 250 Hz - 500 kHz, to match raw data for 1.5E9 protons April 2014: Proton, 300 MeV, q=+1, Run 1: N=1.5E9 Analysis: Find 1st peak and its position Calculate peak detector amplitude (PD) as difference base line (mean over 1 period) and output (mean over 3 periods) Calculate analog output (AO) as maximum value of filtered ADC data Base line region PD amp. region 3x t OSC 1x t OSC PDAO 1.5E9 p~0.25 FS~0.24 nC~ 8000 ch. =OK!

10. FFT spectrum of damped oscillation with resonance frequency at 28.9 kHz. Run 1: 1.5E9 p; Converted signal & FFT spectrum

11. AO and PD work reliably Scatter plot of AO vs. PD yields straight line Comparison PD and AO for 50 cycles, N=1.5E9 „Empty cycle“ STD(residuals) ~48 ADC ch. ~0.6 %

12. Cycles with good agreement Cycles with ~15% variation Conclusion: At least one signal is unstable !!! April 2014: Proton, 300 MeV, q=+1, Run 2: N=1E8 Constant PD signal, but AO varies by ~15% Some cycles with good agreement 1E8 p ~ 1.6% FS ~ 530 ch. ~ 16pC = OK!

13. PD jitter of 150 ADC ch. (25% of mean value) for fixed AO signal of 600 ADC ch. 150 ADC ch. ~ 22.5 mV (of 5 Volt) ~ 0.5% ADC full scale Conclusion: Need to check the RAW signal to investigate effect! Comparison PD and AO for 50 cycles AO / ADC ch. PD / ADC ch.STD(residuals) ~47 ADC ch. ~8 %

14. AO: Noise typically +/- 50 ADC ch. with spikes of 100 ch. PD: Smooth, clean, low-noise signal!!! Conclusion: ADC noise very low, noise in AO generated in amplifier or along transmission line! April 2014: Proton, 300 MeV, q=+1, Run 2: N=1E8 PD OK! PD not OK! Small kink the reason?? Noise spike

15. Pulse shape examples in region around 1st maximum ??? When and why does the peak detector circuit fail ???

16. ADC noise within 25 ch. Assume Gaussian distribution: 25 ch.~6 σ 1 σ ~ 4 ch. = 0.6 mV ~ 8E5 e Minimum detectable signal ~ 50 ch. ~1E7 e Question: Can we reduce the amplifier or environmental noise contribution??? April 2014: Proton, 300 MeV, q=+1, Run 3: 3E7 p 3E7 p ~ 0.6% FS ~ 170 ch. ~ 5pC = OK!

17. PD unable to measure at the intensity of 4E7 e (~5 pC) AO yields a reasonable signal within ~100 ch. (or 2E7 e, ~3 pC) Comparison PD and AO for 50 cycles

18. Run 7: FFT spectra of 50 cycles plotted in same spectrum. Peaks appear at regular intervals! Noise contribution at multiples of ~5.14 kHz; beam dependant or EMC problem ? April 2014: Proton, 300 MeV, q=+1, Run 7 Resonance frequency of RT (=LRC circuit) ~29 kHz

19. The following statements hold for the 1 nC range (6.25E9 protons) 1.5E8 charges (~2.5% FS, 24 pC): Peak detector reliable 1.0E8 charges (~1.6% FS, 15 pC): Peak detector unreliable (~8% resolution) 4E7 & 3.1E7 charges (6 pC & 5 pC): Peak detector misses ~50% of all beam pulses No obvious reason was found what triggers an incorrect peak detector output Peak detector signal very clean (25 ADC channel base line width ~ 4E-4 FS of ADC) Analog output shows noise peaks at 5.14 kHz intervals (systematic noise source) with amplitudes in range +/- 75 ADC channel Struck ADC detection limit (S/N~2) ~50 ADC ch. ~7.5 mV ~1E7 e (1.5E-3 FS of RT) Noise peaks at 5.14 kHz intervals indicate systematic noise source Noise peaks significant up to frequencies above 500 kHz. Investigation of noise sources ongoing!!! Potential of significant improvement of charge determination (lower robust detection limit & better resolution) 18/12/2014: MW has measured unusually high ~15 mV Vpp ripple on output! Every ADC system (or more generally any system providing time-related data sets) should provide a FFT spectrum to user for trouble-shooting purposes (not necessarily as standard function)!!!!! Conclusions

20. Part III – Proton FCT data

21. Run 3 – 3E7 particles 29. April 2014: Proton, 300 MeV, q=+1, Run 3: 3E7 p Attenuator: 0 dB Nominal total gain = 45.2 – 3.5 dB (att. insertion loss) = 41.7 dB (gain=121.6) Measured gain = 43.5 dB (used in this document) Losses in cable are not considered here (~75 m RG214) 50 extractions from SIS18 were stored for offline analysis (for all runs!) (IMPORTANT: The block of cycles is not contiguous! Some cycles are missed, if the DAQ system was still busy to write the output files!) Nominal FCT sensitivity S = 4.2 V/A Gain = 10**( total gain [dB]/20 ) ADC [ch.] * ( ADC_range [V] / No_ADC_ch. [ch.]) Current I [A] = -------------------------------------------------------------------- Gain * sensitivity S [V/A] Particle numbers are calculated from current integral and proton charge (q=1)

22. Basic Formulae and Estimates No. Of particlesN_ion3,00E+07 Charge of ionsq_ion1 Total chargeQ_tot4,81E-03nC pulse length of 1 bunchtau_bunch100ns No. Of triangular bunchesn_bunch4harmonic number Charge per bunchQ_bunch1,20E-03nC Mean current in bunchI_mean1,20E-05A 0,01mA 12,02µA Peak current in bunchI_peak2,40E-05A 0,02mA 24,03µA We start with some basic assumption and calculate the expected peak voltage in units of ADC channels. Assume „typical“ triangular bunch shape for each of the 4 „identical“ bunches! 300 MeV/u protons travel at v = 0.65 c and have a revolution frequency of 0.9 MHz in SIS18 At harmonic number h=4, the RF frequency is f RF = 4x0.9 MHz = 3.6 MHz

23. Sensitivity of FCTS_FCT4,2V/A Mean output voltageU_mean5,05E-05V 0,05mV 50,46µV Peak output voltageU_peak1,01E-04V 0,10mV 100,93µV Amplifier/attemuatorgain_dB41,7dB gain121,62 ADC input rangeV_ADC2V Nominal ADC bitsN_bit10 No. Of ADC channelsN_ADC1024channel Sensitivity Ch./VoltS_ADC512Ch./V Sensitivity V/ch.1,95mV / ch. Mean ADC ouputU_ADC_mean3,1ch. Peak ADC outputU_ADC_peak6,3ch. Basic Formulae and Estimates cont. For 2E9 particles and 16 dB attenuation, one gets U_ADC_peak ~ 66 ch.

24. A block of 10.000 samples is saved for each extraction, with respect to the trigger timing of the extraction kicker that is fed to the DAQ system! Some data are faulty! Fit baseline in two regions, one to the left (1600 – 2400) ns and one to the right of beam signal (4160 – 4960) ns. Subtract fit from raw data to yield „corrected FCT signal“. Integrate beam pulse in region (2800 – 4000) ns Raw ADC signal agrees in magnitude with estimate! Data Analysis I – Baseline Correction

25. The corrected FCT signal is scaled to units of ampere via sensitivity of S = 4.2 V/A and ADC sensitivity of 1.95 mV/ch (nominal ADC values). The FFT of the corrected FCT signal is calculated and displayed up to 30 MHz. The spectrum shows the total revolution frequency at 3.6 MHz of the bunch train, higher order components, and a peak at frequencies < 1 MHz. Data Analysis II – Conversion to physical unit

26. The „empty“ data set shows a clean baseline of ~8 channels total width Data Analysis – Conversion to physical unit „Empty cycle“ or weak beam The total baseline width is ~8 channels (15.6 mV) The standard deviation can be estimated to be 1.5 ch. (0.15% FS) = 3 mV

27. Data Analysis – Conversion to physical unit „Empty cycle“ or weak beam

28. Nice comparison between mean signal of all data (green) and single signal trace (blue) Zero-value in trending plot : Kicker not detected ! 2 empty cycles at number 35 and 36 Signal peak ~10-11 channels. Peaks are clearly detected. Run 3 – Trend of FCT integral Histogram

29. Zero-value in trending plot: Kicker not detected! Weak signal of <8 channels, but peaks are clearly detected. Run 4 – Trend of FCT integral

30. Zero-value in trending plot: Kicker not detected! Extremely weak signal of a few ~5 ADC channels. Peaks are now „thinner“ compared to Run 5. Run 6 – Trend of FCT integral

31. Zero-value in trending plot : Kicker not detected! Extremely weak signal of a few ~5 ADC channels. Large fluctuation in FCT integral. Similar signal strength to Run 6. Very nice resolution for mean value down to the few micro-Ampere range!!! Run 7 – Trend of FCT integral

32. Run 3 – Trend of FCT and RT charges Resonant TransformerFast Current Transformer Calculated particle numbers do agree to within 10%, if measured gain is used! Cable losses in RG214 can shift FCT charge by ~10-15%: Attenuation(3.6 MHz, 75 m) = 0,975 dB or factor ~0.89. Cycles are not identical due to gaps in DAQ system outputs and asynchronous (manual) start of 2 different DAQ systems for FCT and RT

33. Run 1 – Trend of FCT and RT particle numbers DAQ systems did not save every single extraction! Asynchronous exports. FCT and RT charges agree, only if the time stamp is identical!

34. Nice comparison between mean signal (green) and single signal trace (blue) FFT spectrum shows larger „main peaks“ at multiples of 3.6 MHz (h=4) and smaller „satellite peaks“ at multiples of 0.9 MHz (h=1) Run 1 – 1.5E9 particles, 16 dB attenuation The attenuator was set to 16 dB (peak value ~ 65 ADC ch.) by accident! The FCT data export was started „as is“ as a test in Run1. The signal was not checked online! For 2E9 protons the peak current in the bunch is about 1.6 mA.

35. Trendline through satellite peaks Satellite peak frequencies as function of harmonic number. The gradient of 0.9 MHz corresponds to the revolution frequency of a single bunch at v = 0.65 c Gaps correspond to positions of the „main peaks“ (that obscure „satellites“)

36. Baseline of Struck ADC very clean. Width of baseline ~8 ch. (of 1024) or 0.15%. At a total number of 2E7 protons bunches are clearly detectable. Minimum detectable proton number in a single peak is about 5E6 for a triangular pulse shape of 100 ns total width. There is a significant discrepancy (~20-30 %) between particle numbers derived by FCT and RT when the available nominal calibration values are used. This discrepancy is reduced to ~10%, if the measured gain is applied. AR will re-check his code for bugs, only if needed and if all calibrations and gain/attenuation values are available! See next item. Necessary investigation: –The RT calibration should be re-checked locally at HTP and in the AP Container –The FCT amp/att unit incl. Transmission line to AP Container should be measured at 3.6 MHz (and higher harmonics) sine waves and some pulsed signals. Inject since wave at amp. input and attenuator input (2 separate measurements). Measurements with pulsed signals inject signal at amp. input. Conclusions