Wire Tension Testing Device used in the construction of the CLEO Central Drift Chamber

Transcription

1 Wire Tension Testing Device used in the construction of the CLEO Central Drift Chamber STEVEN E. CSORNA AND SZABOLCS MÁRKA Dept. of Physics and Astronomy Vanderbilt University Nashville, TN 79 USA Abstract: - We describe an instrument [] designed and built to measure wire tension in the CLEO central drift chamber []. Sense-field wire pairs are excited by means of high voltage pulses and we measure the electrical signal that develops when one of the oscillating wires is held at constant high voltage. The advantage of using this instrument is that only one side of the drift chamber needs to be accessible, no magnetic field is required and the tensions of both wires are measured simultaneously. Key-Words: - Drift Chamber, CLEO, Wire, Tension, HEP, Particle, Physics, FFT, Sag Introduction We describe a device for exciting wire pairs of different linear mass densities for the purpose of measuring their natural frequency and decay constant. The measurement makes possible the calculation of the individual wire tensions by use of the formula for the natural frequency of vibration of a wire rigidly held at the two ends: ( ) f = L T λ where λ is the wire mass per unit length, T is the wire tension and L is the wire length. This application was developed for use in the construction of the new.m long CLEO III drift chamber (DR). CLEO III is the solenoidal detector used to measure events resulting from collisions of GeV/c electrons and positrons produced by the Cornell Electron Storage Ring. The DR design (incorporating approximately, wires) calls for constant cell dimensions, and therefore requires that the sag due to gravitation be the same for both sense and field wires. The gravitational sag is given by the formula: ( ) δ y = g λ L 8 T In order to keep the sag the same for sense ( µm diameter tungsten) and field wires ( µm diameter Aluminum) the tensions of the two kind of wires has to be in the ratio of the linear mass densities and therefore they will have the same natural frequencies. Independently of the method of detection, a measurement of the natural frequencies alone is not sufficient to either determine the presence of two vibrating wires or to identify which frequency is associated with which wire (if more than one frequency is observed). We have observed that the sense wires and field wires have substantially different damping. The amplitude of vibration of a typical field wire decreases exponentially with a decay constant of.7 sec -, a factor. smaller than the decay constant of a sense wire. This observation leads to a reliable method for determining the number of

2 vibrating wires and for associating each with the appropriate frequency. The technique that we utilize is summarized as follows. A chosen field-sense wire pair is pulsed several times with a shark fin shaped (distorted square wave) high voltage pulse of approximately 8 V at a frequency near the anticipated natural frequency. The pulsing shakes the wire pair, and as soon as it is over, the computer fixes one of the wires (usually the sense wire) at high voltage. The other wire (usually a field wire connected to a sensor head probe containing a preamplifier) is connected to ground by a resistor. The changing capacitance between the oscillating wires induces a current that is amplified and detected. The signal consists of the superposition of two exponentially damped sinusoidal currents. After amplification, this current is digitized by an ADC (8 samples over. sec.) and Fast Fourier Transformed (FFT). The power spectrum of the FFT, being the sum of two interfering Lorentzian amplitudes, will display peaks at the natural frequencies of the vibrating wires. The width of each peak in the power spectrum tags which wire has that frequency: if it is broad it is a sense wire, if it is narrow it is a field wire. In the case that both wires vibrate at the same frequency, a fit to the spectrum reveals the contribution from each type of wire. If only one wire vibrates, the fit shows that the contribution from the other one is negligible. Our calculation to determine the effect of the coupling between the wires shows that a correction of -.% (-.%) needs to be applied to the measured frequency of the Aluminum (Tungsten) wire; similarly, a correction of +.% (-.%) applies to the measured decay constants. Due to the fact that the aluminum (field) wires creep, they are strung with higher tension during the construction and reach the required tension later. Therefore, the natural frequency of the two types of wire is different right after the stringing process when our method is applied to test the wires. However, if the two resonant frequencies accidentally coincide, there are other ways of breaking the degeneracy. It has been observed that due to the different thermal expansion coefficient of Aluminum and Tungsten, a one degree change in temperature (Celsius) will separate the peaks by.8 Hertz. Thus, heating the Aluminum field wire by only a few degrees (without warming the chamber body) separates the peaks sufficiently to make reasonable measurements. We have tested this scheme and find that it works; however, we will not go into details in this paper. Operationally, the user connects the high voltage and the sensing probe to the wire pair of interest, the measurement proceeds under the control of a PC. The high voltage is pulsed, the signal is acquired, the FFT is calculated and peaks are sought. If only one peak is found, the high voltage pulse frequency is adjusted to the frequency of the single peak and the procedure is repeated. If two peaks are found, the optimal frequency for getting both wires to oscillate maximally is calculated and applied. The results of this second pass, including a calculation of the position of the peaks by means of a weighted mean method, are displayed to the user, who can make a decision about accepting the wires or rejecting either of them. Additional information from an 8 parameter (amplitude, frequency, width and phase for each wire) MINUIT fit of the frequency spectrum is available upon user request. This technique was developed and tested at Vanderbilt on a small -wire prototype chamber and later applied to the axial layers of the large DR prototype at Cornell. In the next sections we discuss the individual components of the tension measuring device, the software that provides a user-friendly interface, controls data acquisition, and performs data analysis. We conclude by analyzing the data obtained by measurements done on the DR prototype. Technical description The schematic of the hardware is shown in Figure. The high voltage (HV) pulsing and the signal read-out is controlled by an IBM PC via a multifunction data acquisition (DAC) card. To avoid Hz noise the sensor head containing the preamplifier is directly connected onto the pin that secures the field wire. The sensor head is very small and light weight in order to decrease

3 the mechanical stress on the pin that holds the wires in place. A very low input current and low Wires HV Pulser Unit Preamplifier Pulser Logic Isolational amplifier and Automatic gain controll Data Acquisition Card IBM PC/AT Controll software Figure. Block diagram of the hardware noise JFET operating amplifier is used as a preamplifier. In order to separate the HV and the PC an isolation amplifier was used. This ensured that the ill effects of an accidental HV short couldn t propagate into the PC. A variable gain amplifier is used to adjust the amplitude of the signal to the fixed range of the -bit successive approximation A/D converter. The gain is controlled by the output of a bit DAC. The two functions of the HV pulser unit (HVPU) are to isolate the computer from the high voltage and to provide the necessary high voltage pulses to the wire. The HV unit is able to pulse the wires, to fix the HV in its high state on the wire during signal read-out and to fix the HV in its low state on the wire after the read-out is finished. In the low HV state the voltage on the wire drops to about V which is enough to saturate the pre-amp in the case of shorted wires. This allows the system to check for shorted wires before any pulsing. The HV current is limited to µa to eliminate any risk of injury to the user. The optically isolated HVPU is controlled by a simple logical circuit, which combines the periodic and the two 'fixing' signals provided by the computer. The stand-alone control program contains three main logical units: data collection, data preparation and fitting. After checking for possible shorts between wires, the program excites the pair of wires at the expected frequency to find the correct amplification. After setting the amplification, the wires are excited again and a. second sample is taken. A FFT is applied to the data and the software searches for the optimal shaking frequency. The optimal frequency ensures that both the broad and narrow width peaks will be present in the frequency spectrum with nearly equal amplitude. After setting the appropriate amplification, the wires are pulsed again with the optimized frequency and a. second data stream is recorded. The FFT spectra of the full data set, the first and last /8 second are computed in the next stage. The FFT of the first /8 second contains mostly the signal from the wire of broad width and the last /8 second's FFT contains only the long lived wire's fingerprint. On the basis of the FFT of the full data set an automatic peak finder algorithm locates the Amplitude [V] Amplitude Signal of Al and W wires Time [s] FFT transform and MINUIT FIT Frequency [Hz] frequency of the two peaks. The FFT spectra, peak frequencies, time domain signal, pulsing frequency and the amplification is displayed. Under most circumstances this data has proven FFT MINUIT FIT Figure. The signal in the time domain, its FFT and MINUIT fit.

4 sufficient for the operator to decide whether the wires are acceptable or not. In principle, it can happen that the two peaks are closer than the width of the signals, making it necessary to fit the FFT spectrum in order to verify the presence of two peaks of different widths. Therefore the program triggers a MINUIT fit. The fit results (amplitudes, peak widths, frequencies, phases, log file and fitted histogram) are transferred back to the PC. All the relevant information (shown in Figure ) including the results of the MINUIT fit are displayed on the operator's CRT. This gives enough information to make a decision about the wires. plots are due to a number of factors, including the variations that occur during the stringing process (temperature variations, nonuniform tensioning, variations in the properties of materials, etc.), the inherent spread in the signal and the measurement error. In the DR geometry, each cell consists of a sense wire surrounded by 8 field wires, with neighboring cells sharing three field wires. Measurements were performed in such a way that no field wire measurements were duplicated, however, each sense wire was actually measured to times, each time in conjunction with a different field wire. A better measurement of the Resolution of sense wire frequency measurement Measurements done on the DR prototype We have measured field wire tensions and sense wire tensions in the axial section of the DR prototype. The ability of the device to separate the sense from the field wires is shown in Figure, a scatter plot of the measured width versus the resonant frequency. Decay constant [Hz] 7 Number of events Damping vs. wire frequency Resonance frequency [Hz] Projection on frequency axis RMS =. Hz Mean = 9. Hz Resonance frequency [Hz] RMS =. Hz Mean =. Hz Figure. The measurements displayed in a scatter plot of width versus frequency. Projection on Decay constant axis 7 Number of events The data clearly indicate a clump centered at frequency of 9. Hz and decay constant.7 Hz (field wires) and another more diffuse clump centered at frequency of. Hz and decay constant.8 Hz (sense wires). Note that the widths of the peaks observed in the projected Decay constant [Hz] RMS =. Hz Mean =.8 Hz RMS =. Hz Mean =.7 Hz Number of events Deviation from mean [Hz] sense wire frequency is arrived at by averaging the results of the various measurements involving the same sense wire. Frequencies calculated in this way were used for the sense wires in Fig.. We estimate our effective measurement error for sense wires by calculating the deviations of the individual measurements from their average. The histogram for all the sense wire data indicates a resolution of approximately.7 Hz Conclusion RMS =.8 Hz Figure. Resolution of the frequency measurement We have developed an instrument capable of measuring wire tensions in the DR environment, thus providing a capability to separate badly strung wires from good ones. Tests on the DR prototype indicate that the instrument's resolution in frequency is of order.7 Hz. The technique does not require a magnetic field, requires access to only one side of the chamber and measures the

5 tensions of two wires at a time. We have shown that the quality of the data and the analysis technique allow the operator to handle all the anticipated problems that can arise, including the presence of shorted wires and badly strung wires.. Acknowledgement We would like to thank Lászlo Baksay, Gyula Zilizi and the DRIII group for their help and support. This work was supported by the U.S National Science Foundation. References: [] S. Csorna and Sz. Márka, Wire Tension Testing Device for DR, CBX 9-. [] S. Csorna et al., Construction of the CLEO III drift chamber, NIM, A9(998), - 9.