====== spectrometric chain ====== ===== Introduction ===== The aim of this laboratory is to understand how the spectrometric chain works. First, you will observe the operation of readout electronics blocks like preamplifier and shaper. The next step you will measure the basic parameters like gain or noise of such readout chain. At the we will try to observe pulses from Am61 radio active source. ===== Circuit description ===== ==== Building blocks ==== Schematic diagram of the spectrometric chain is shown below in . {{ :user:kulis:uspec:uspec_front02.png }} The most often used front-end electronics configuration consists of a charge sensitive preamplifier and a CR-(RC)^n the filter known as pseudo-gaussian shaper. The CR-(RC)^n filter is called pseudo-gaussian because the response of such filter to step function becomes exactly gaussian when its order n approaches infinity. The pseudo-gaussian shaper is frequently used because of its simplicity and because it allows to obtain close to optimum S/N ratio [ref-gatti,radeka]. Neglecting initially R_f the preamplifier output voltage can be written as: U_{0}(s) = -\frac{I_{in}(s)}{sC_f} \label{eq:upre} where I_{in}(s) is a Laplace transform of a sensor pulse and assuming that A_{pre} is high enough (reasonable assumption since the gain of operational amplifiers is of the order 10^5-10^6). The shaper output signal can be then written as: U_n(s) = H(s) \cdot U_{0}(s) \label{eq:un} where H(s) is the shaper transfer function and again assuming high enough gain of the operational amplifiers A_i. Writing equations for the first shaper stage: U_1(s) = - \frac{U_0(s)}{R_d + 1/(sC_d)} (R_i||C_i) \label{eq:u_sh1} and for the following shaper stages: U_k(s) = -\frac{U_{k-1}(s)}{R_g}(R_i||C_i) \label{eq:u_shk} and setting equal the differentiating and the integrating time constant \cite{crrc} \tau = R_d C_d = R_i C_i one obtains the shaper transfer function: H(s) = \frac{U_{n}(s)}{U_{0}(s)} = (-1)^{n} \frac{1}{R_g^{n-1} C_i^n R_d} \cdot \frac{s}{(s+1/{\tau})^{n+1}} \label{eq:u_sh1} The transfer function of the whole circuit may be written as: U_{n}(s) = -\frac{I_{in}(s)}{sC_f} \cdot (-1)^{n} \frac{1}{R_g^{n-1} C_i^n R_d} \cdot \frac{s}{(s+1/{\tau})^{n+1}} \label{eq:un_s} Usually, a very good assumption for sensor pulse shape is a dirac delta, i.e. I_{in}(s)=Q_{in} with its integral equal to a total charge Q_{in} deposited in the sensor. This assumption reflects the fact that the charge collection time in the sensor is much shorter than the front-end electronics shaping time. Under this assumption the front-end response in time domain equals: U_{n}(t) = -\frac{Q_{in}}{C_f} \cdot (-1)^{n} \frac{1}{R_g^{n-1} R_d C_i^n} \cdot \frac{t^n e^{-t/\tau}}{n!} \label{eq:un_t} which has the maximum amplitude at time T_{peak}=n\tau which is equal to: U_{n}^{max} = -\frac{Q_{in}}{C_f} \cdot (-1)^{n} \frac{\tau^n}{R_g^{n-1} R_d C_i^n} \cdot \frac{e^{-n}}{n!} \label{eq:un_max} To move from the theoretical considerations to a practical circuit the only thing needed is to add a resistance R_f in parallel to C_f. It is needed for two purposes: first to set the DC level of the preamplifier input and second, to assure that the feedback capacitance C_f gets discharged and so the preamplifier output will not saturate after a number of subsequent pulses from the sensor. The derived above transfer function becomes practically unaffected if the R_f value is sufficiently large (\sim 1 G\Omega in this laboratory circuit). ====Sensor, Preamplifier and Shaper - Noise Analysis==== There are two main sources of noise which deteriorate signal to noise ratio in the proposed spectrometric system: the sensor noise and the front-end electronics noise. The sensor noise comes mainly from a shot noise caused by the sensor leakage current fluctuations I_{leak} and its spectral power density is equal to: i_{d}^{2} \equiv \frac{d \langle i_{d}^{2} \rangle}{df} = 2qI_{leak} \label{eq:s_ileak} The front-end electronics noise is mainly due to the preamplifier and the feedback resistance R_f. In a properly designed system the following shaper stages give small contribution to the total noise (since the signal is already amplified by the preamplifier). {{ :user:kulis:uspec:uspec_noise02.png }} The overall effect of preamplifier noise is described in terms of an equivalent input noise expressed by series (voltage) and parallel (current) noise sources with spectral densities respectively v_{eq}^2 and i_{eq}^2. The feedback resistance R_f is characterized by its thermal noise with spectral density: u_{R_{f}}^{2}\equiv \frac{d \langle u_{R_{f}}^{2} \rangle}{df} = 4kTR_{f} \label{eq:s_rf} The noise diagram of the front-end electronics with the equivalent input noise sources is shown in fig\ref{fig:sch_noise0}. Knowing that for the circuit with transfer function T(s) the input noise spectral power density S_{in}(f) is transformed to the output as: S_{out}(f) = S_{in}(f) | T(f) |^{2} \label{eq:s_transf} where \left| T(f) \right| =\left| T(s = 2 \pi j f) \right|, one can derive the preamplifier output spectral power density. This is easily done using the superposition principle. Since in practical systems R_f is large and the R_f C_f time constant is much larger than the signal duration the contribution from v_{eq}^{2},i_{eq}^{2}, i_{d}^{2} may be calculated neglecting the R_f. To calculate the contribution from u_{R_f}^{2} one assumes that \omega / {R_f C_f} \gg 1 in the signal bandwith. With such approximations the preamplifier output noise density equals: v_{0}^{2} \equiv \frac{d \langle v_{0}^{2} \rangle}{df} = \\ \left| \frac{C_{in} + C_{f}}{C_{f}} \right|^{2} v_{eq}^{2} + \\ \left| \frac{1}{sC_{f}} \right|^{2} (i_{eq}^{2} + i_{d}^{2}+ \\ \frac{u_{R_{f}}^{2}}{{R_{f}}^{2}}) \label{eq:s_pre} and subsequently the shaper(H(s)) output noise density equals: v_{n}^{2} \equiv \frac{d \langle v_{n}^{2} \rangle}{df} = \\ \frac{d \langle v_{0}^{2} \rangle}{df} \left| H(f) \right|^{2} \label{eq:us12a} In order to obtain the voltage noise rms at the output the spectral noise density needs to be integrated in the frequecny domain: v_{n}^{rms} = \sqrt{\int_{0}^{\infty} v_{0}^{2}(f) \left| H(f) \right|^{2} df} \label{eq:us12} For this integration the explicite dependence of v_{eq}^2,i_{eq}^2 on frequency is needed. The main contribution are white (constant) voltage and current noise sources v_{eqw}^2,i_{eqw}^2 . In JFET or CMOS technology also the flicker noise voltage component should be taken into account so one can assume the equivalent spectral noise densities as: v_{eq}^2 = v_{eqw}^2+\frac{v_{1/f}^2}{f} and i_{eq}^2 = i_{eqw}^2. For the proposed pseudo-gaussian shaping (fig.~\ref{fig:sch_front}) the above integral can be calculated analytically as: v_{n}^{rms} = \frac{C_{in}+C_f}{C_{f}} \frac{\tau^n}{R_g^{n-1} C_i^n R_d} \\ \sqrt{\frac{\Gamma(n - \frac{1}{2})}{8 \sqrt{\pi} n!} \\ \left( \frac{v_{eqw}^2}{\tau} + \frac{\tau \cdot (2n - 1)}{(C_{in}+C_f)^{2}} \\ (i_{eqw}^{2} + 2qI_{leak}+ \frac{4kT}{R_{f}}) \right) + \frac{v_{1/f}^2}{2n} } \label{eq:13} where \Gamma is the Gamma function. ====Sensor, Preamplifier and Shaper - Signal to Noise==== Knowing the signal amplitude U_{n}^{max} (eq. \ref{eq:un_max}) and the noise rms value (eq. \ref{eq:13}) the S/N ratio may be easily obtained. This is rarely done. Instead the noise performance is usually expressed in a slightly different way i.e. by means of an equivalent input noise charge (ENC), in the number of electrons. The ENC is calculated dividing v_{n}^{rms} by the amplitude U_{n}^{max} obtained for single electron input charge Q_{in}=q=1.6 \cdot 10^{-19}C and gives in result: ENC = \frac{e^n n!}{q n^n} \\ \sqrt{\frac{\Gamma(n - \frac{1}{2})}{8 \sqrt{\pi} \sqrt{n!} } \\ \left( \frac{v_{eqw}^2}{\tau} (C_{in} + C_{f})^2 + \tau (2n - 1) \\ (i_{eqw}^2 + 2qI_{leak} + \frac{4kT}{R_{f}}) \right)\\ + (C_{in} + C_{f})^2 \frac{v_{1/f}^2}{2n}} \label{eq:enc1} It is seen that in order to minimize the noise one should minimize the sensor leakage curent and use the highest possible R_f value. It is also seen that the voltage white noise decreases with shaping time on the contrary to the current white noise which increases with \tau. For a given filter order n one can minimize the ENC finding an optimum shaping time: \tau_{opt} = (C_{in}+C_f) \sqrt{\frac{v_{eqw}^2}{(2n - 1)(i_{eqw}^2 + 2qI_{leak} + 4kT/R_{f})}} \label{eq:tau_opt} For practical applications it is often more convinient to study the noise performance as the function of T_{peak} instead of \tau. In this case the formula \ref{eq:enc1} will be espressed as: ENC = \frac{e^n n!}{q n^n} \\ \sqrt{\frac{\Gamma(n - \frac{1}{2})}{8 \sqrt{\pi} \sqrt{n!} } \\ \left( \frac{n v_{eqw}^2}{T_{peak}} (C_{in} + C_{f})^2 + T_{peak} \frac{2n - 1}{n} \\ (i_{eqw}^2 + 2qI_{leak} + \frac{4kT}{R_{f}}) \right)\\ + (C_{in} + C_{f})^2 \frac{v_{1/f}^2}{2n}} \label{eq:enc2} Now, instead of optimizing the shaping time \tau the optimization of the peaking time T_{peak} can be done in order to equalize the contribution of voltage and current noise. As before the voltage component decreases with increasing T_{peak} while the current component increases with increasing T_{peak}. The noise dependence on shaping order is more complicated and in general it is better seen with numerical simulations. Qualitatively one can only say that the ratio of voltage to current noise contribution increases with increasing shaping order. To get feeling about the real noise performance it is instructive to plot the ENC as a function of peaking time, for different shaping orders (Fig XX). {{:hw:lab:spectrochain:theory01.png|}} As expected the optimum peaking time exists. Regarding the shaping order it is seen that in general the higher order the better noise performance. Assuming that the further signal processing stages add negligible contribution to the overall system noise one can estimate the expected resolution of the spectrometric system in terms of the number of electrons, on the basis of eq. \ref{eq:enc2}. ==== Practical realization ==== Schematic diagram of the implemented spectrometric chain is shown below on figure## . {{:hw:lab:spectrochain:schemat.png|}} One can see few differences between figures XX and ZZ. At the input of the preamplifier charge adapter is added. Such a circuit allows injecting quasi dirac current pulses. Amount of charge is given by the amplitude of voltage step pulses. Rt resistor is added for line termination. Looking at the resistors R1 and R2 at voltage divider one can write the value of injected charge as: Q_{in}=\frac{R_2}{R_1+R_2} C_{test} V_{in}. All capacitances in the shaper are changeable in a wide range(mode than two orders of magnitude). Such a construction will allow to study the impact of the shaping time value on output signal shape and noise performance of the whole chain. Such a changeable capacitance are build from tens of binary weight capacitances cross-connected by switches. All switches are controlled by microcontroller. On front panel you may read the shaping time value ( \tau=RC ) given in microseconds. To change it you should use the rotary switch (selector). The outputs of subsequent stages are routed via switches to one output connector. Those switches are controlled by the same digital circuit. The active output is displayed in the box OUTPUT. By clicking rotary knob you can select the active variable: shaping time or output. The laboratory module contains also few integrated voltage regulators to generate all needed voltages for both analog and digital blocks. ===== Laboratory Setup ===== During this laboratory we use the following equipment: * front-end electronics circuit * digital sampling oscilloscope (TEKTRONIX TDS3034) * pulse generator (AGILENT 3320) * true RMS meter (HP3400) * power Supply (AGILENT 3631A) * cables * PC with dedicated software The connection of the setup is shown in the figure X. {{:hw:lab:spectrochain:setup.png|}} Ask the leading person to check all connections and do following steps: * set supply voltages to +/- 7V * set channel's one digital scope input impedance to 50 Ohm * set rectangular wave on pulse generator * switch on pulse generator output * set triger to proper channel ===== Measurements ===== ==== Observing the preamplifier output ==== Select 0 as an active output. Try to find output pulse on oscilloscope. Set time base of oscilloscope to 400ns. Does signal look like voltage step? Try to estimate charge gain given by K_u = \frac{V_{out}}{Q_{in}} (having in mind formula "charge adapter") and compare it with value given by formula X ref wzmocnienie pream X. Try to look mode deeply on the pulse head (change time base to 40ns). How would you explain the non zero rising time? Then look at the pulse tail. Try to estimate time constant of this pulse and compare it to "y". ==== Linearity check of pre amplifier ==== Perform measurements of pulse amplitude at the preamplifier output for the set of input test pulse amplitudes. The obtained results write to file ''linearity.dat'' in following format: #vin[mv] vout[v] 500 XXX.Y 1000 XXX.Y 1500 XXX.Y 2000 XXX.Y ... Then run script ''linearity.gnu'' and see results written in the file ''linearity.png''. ==== Observing shaper output ==== Look at the first shaper stage output. Check whether the displayed time is equal to the peaking time (time after which pulse reaches its maximum). Look at the pulses at the outputs of subsequent stages. Find their peaking times and see how they are related to the shaping time constant. Try to set the same peaking times for different shaping orders and compare the pulse shapes. ==== Noise performance ==== Measuring noise performance is two phase process. At the begining one has to measure the output amplitude V_{out} being response to input current pulse carrying charge Q_{in} (use cursors on oscilloscope for this measurement). In the second phase the voltage noise root mean square V_{rms} at the output is measured by the HP3400 true RMS meter. During this measurement input pulse should be disabled. It is important to keep circuit settings (shaping time, observed output) the same for both measurements. The equivalent noise charge (ENC) is given by: ENC[C]=\frac{V_{rms}}{V_{out}} Q_{in} Repeat this measurements for several combinations of shaping time and shaping order. Write down your results to the file ''noise.dat'' in following format: #order time_const[us] v_out[mv] v_rms[mv] 1 0.25 XXX Y.YY 1 0.50 XXX Y.YY 1 0.70 XXX Y.YY 1 1.00 XXX Y.YY ... 2 0.25 XXX Y.YY 2 0.50 XXX Y.YY 2 0.75 XXX Y.YY 2 1.00 XXX Y.YY ... 3 0.25 XXX Y.YY 3 0.50 XXX Y.YY 3 0.75 XXX Y.YY 3 1.00 XXX Y.YY ... To analyse the measured data use script ''noise.gnu''. See the results in the file ''noise.png''. Are you able to show the optimum shaping time for each filter order ? Does it pay off to use higher shaper orders for noise reduction ? ==== Observing pulses from source ==== Uda sie zorganizowac zrodlo ? ==== Discussions topics==== * What is main uncertainty source during gain measurement in such setup? * During this laboratory we were investigating electronic noises. You have to remember that in real experiments You also have to fight with disturbances. Can You give an example of disturbances in real experiment? ===== References ===== E. Gatti i P.F. Manfredi, Processing the Signals from Solid-State Detectors in Elementary-Particle Physics, Revista Del Nuovo Cimento (1986).