Analysis of the NA48 LKr Calorimeter Readout Electronics

The electronics to read out the LKr Calorimeter was developed and produced from 1990 to 1996. The various parts of the electronics have been described in many technical notes and papers, but this is an overall technical description with references available in one place (TWiki or WWW) and it should be useful when analysing the signal processing chain in detail especially for future modifications and adaptations. The different parts of the present LKr signal processing chain are described in separate chapters below.

The NA48 LKr readout electrode capacitance

For an accurate analysis of the LKr readout chain (especially the noise performance) it is important to know the value of the capacitance of the readout electrodes. This can be calculated with help of the Maxwell SV program from ANSOFT .

Electrodes dimensions:

Figure 1 shows the drawing used as input to the Maxwell program. The red strips (readout electrodes) are 18 mm wide, 40 um thick, 127 cm long and has a 2 mm distance to each other. The black strips are at 10 mm distance to the readout strips. The red strips are numbered from left to right a1, a2 ,..., a9. The area marked with in blue is filled with LKr which is assumed to have a dielectric constant = 1.66.

Result from simulations:

The capacitance of the readout strip is found to be 131 pF in total (75 pF + cross talk 2x28 pF). To this there is a calibration capacitor of 22 pF and the input capacitance of the preamplifier (modelled with Spice parameters) and some unknown stray capacitance. The total capacitance value found (153 pF) is within +-5%? It should be noted that for transient analysis of the LKr readout the electrodes should be treated as transmissions lines with time constant of about 5 ns. To simulate this a more complicated model is needed.

Cold Preamplifier

The noise of the whole readout electronics is determined to about 90% by the input transistor of the cold preamplifier. This transistor the JFET IF4500 from the process P_NJ450L of INTERFET has a minimum noise at temperature of 160K and still a very good noise performance at the calorimeter operating temperature of 120K. All other transistors are from the process: P_NJ132L

The final version of the preamplifier is described by C.Cerri. The schematic of the preamplifier is shown below:

The preamplifier is analysed with help of the LTSpice program with parameters obtained in 1991 from S.Rescia of BNL .

Below is shown the LTspice schematic for simulating the preamplifier [ LTSpice_Preamplifier ]:

The output noise spectral density of the preamplfier after the coaxial cable and at input of the transceiver is shown in the Figure below. In order to put some realistic limits on the noise of the electronics the transconductance (gm) of the IF4500 is varied from 25, 55 and 105 mA/V. This give a variation of the noise voltage of of about 25% as shown below. The total noise is plotted in blue, the noise component from the input JFET transistor in red and the 10 kohms feedback resistor RF. At low frequencies the noise of the resistor RF dominates while it is the noise of the JFET IF4500 at high frequencies.

In contrast to the noise, the amplitude and frequency response of the whole electronics chain including the preamplifier is given by passive external components. Therefore the spread in gain and delay between channels are very small (+-1%). This show in the following plot of the amplitude response of the preamplifier:

Transceiver

The transceiver has to amplifiy and amplify to the signal from the preamplifier and send it the readout module the Calorimeter PipeLine Digitizer (CPD) via a twisted pair cable. The tranceiver reduces the coherent noise while having a neglible effect on the thermal noise of the electronics. The preamplifer decay constant of 150 ns is compensated with a method called pole zero compensation see http://en.wikipedia.org/wiki/Pole_splitting. This has the effect that the noise spectrum is amplified with a peak at about 20 MHz. The noise bandwidth is limited by the combined effect of the preamplifer and transceiver but also the cables to about 25 MHz. A first prototype system was tried out in beam test on the prototype calorimeter in 1992. The final version transceiver is described in NA48_97-30.PDF from 1997.

The tranceiver schematic is shown below

An ultra low noise amplifier CLC425.pdf is used as input stage. The output stage consists of a dual wide-band amplifier CLC412.pdf.

Below is shown the schema of LTspice simulation [ LTSpice_Transceiver ] (The preamplifer part is shown above in [ LTSpice_Preamplifier ]

Illustration of the effect of the transceiver on the signal is shown the simulations of the waveforms at the preamplifier, transceiver and shaper input.

The read curve V(6) shows the preamplifier signal shape at the input of the transceiver. The differential signal at the transceiver output V(7)-V(6) is shown in blue and the signal after the twisted cable termination and (20ns) differentiation is shown in green V(out). The effect on the overall delay due to the transconductance of the input JFET of the preamplifier can be seen.

The spectral densitiy of the most important noise sources with (gm) of the IF4500 JFET varied from 25, 55 and 105 mA/V :

The total noise is the curves in blue V(onoise). The noise of the IF4500 is shown in red and the dominating source. All other noise sources have much smaller influence which is illustrated by the curves V(r2), V(r3) and V(rf) from [ LTSpice_Preamplifier ]. Note The "oscillations" in V(r3), due to be non-perfect back termination of the preamplifer.

Shaper

Three different types of shapers were developed to test the basic concepts of the electronics and the LKr calorimeter in beam tests [1], [2], [3], [4], [5]. The analog signal processing in the shaper consists of a 20ns differentiation and Bessel filter. The final version of the shaper is implemented in a custom integrated circuit KRYPTON [6] designed in 1.2 μm BiCMOS technology. The KRYPTON was designed and produced by the company CISS and Austrian Mikrosysteme', Graz, Austria. The Figure 1 below shows the block diagram of the KRYPTON :

It consists of an input line receiver, a 9-pol shaper, HF noise filter and gain switching circuit for augmenting the dynamic range of a 10 bit FADC. The gain is dynamically switched in four ranges from 1, 2.5, 6.1 and 17.6 giving a dynamic range of about 1:15000. There are also analog circuits for trigger summing with programmable gain for use by the neutral trigger. The KRYPTON together with a 10 bit FADC with memory and trigger circuits are implemented on an analog subcard called CPDAS [7]. The CPD CPDAS contains in addition to the shaper, ADC, pipeline memory and trigger circuits. The CPDAS contains two channels and there are 32 cards in the CPD module.

The schematic of the LTSpice_Shaper simulation of the shaper is shown below. (LTSpice_Preamplifier and LTSpice_Transceiver from above is not shown)

All passive R, C and L components are of the SMD 0605 components. The rest of the components are parts of the KRYPTON ASIC.

The Frequency response of the LKr electronics:

The bandwidth of the LKr electronics chain as determined by the shaper is about 8 MHz. The AC frequency response of the total shaper is shown in blue. The curve in red shows the response of the 9-pol Bessel filter with -3 dB high frequency limit at 5.8 MHz. The maximum attenuation of -120 dB. The dip at 75 MHz and increase in the gain above this frequency is due to crosstalk and other parasitic effects. The HF filter after the shaper compensate for this (-3dB at 22 MHz and attenuates more than 20 dB).

The Amplitude response of LKr electronics:

Typical signal shapes and amplitudes at different part of the LKr electronics readout chain for a typical LKr signal with peak 1 uA amplitude (~425 MeV).
a) V(3) = 8mV is the output of the preamplifier before the coaxial cable,,
b) V(tp)-V(tn) = 16 mV the transceiver amplitude after the 16 mV,
c) V(dout) = 12 mV shaper line receiver output
d) V(adc) = 410 mV the shaper output to the ADC.
(Note the small rate effect at 3 μs due to AC coupling internally in the preamplfier in a) and b).

The Pulse response of the LKr electronics:

The variation in pulse shape when the transconductance (gm) of the IF4500 is varied from 25, 55 and 105 mA/V is about 1 % in amplitude, 2ns in in delay and neglible in FWHM.

The spectral_density of the total noise of the LKr electronics:

The integrated noise spectrum of the LKr electronics:

Summary

The simulation results of the noise as function of frequency are shown in the picture in the attachment below:

The table below shows a summary of the contribution of the different noise sources of the LKr readout for the case when the preamplifier JFET IF4500 has a gm of 55 = mA/V.

The conversion factor from mVrms to MeV is about 1.0, (This assumes that 1 GeV shower produces 2.5 uA peak current in the central hit cell of the shower box, while the LTspice simulation shows that 1 uA peak current gives 410 mV peak amplitude of shaped output pulse).

Note; The contributions of the different parts of the electronics to the noise are strongly frequency dependent. If the bandwidth of system is reduced by a factor two the overall noise would be decreased by a factor 3 and the contribution from the preamplifier and shaper would be about the same.

The corresponding total noise value for JFET IF4500 with gm = 105mA/V is 7.7 mVrms and with gm = 25 mA/V 10.0 mVrms, values which are in good agreement with the measured noise performance of the calorimeter.

-- BjornHallgren - 14 Feb 2008