- Roof Board
- Chip Development
- Prototype boards
- Glossary and Index
This chip has been developed by Mitch Newcomer at the department of High-Energy
Physics at University of Pennsylvania, Philadelphia USA.
A bonding diagram and a chip-pad description can be found on the
stamp-card postscript page.
The acronym ASDBLR stands for 'Amplifier, shaper, discriminator, base-line restorer', and the prupose of this chip
is to perform the analog front-end of the straw read-out.
The chip is manufactured as a full-custom, analogue, bipolar ASIC with the dimension 6.17mm x 4.78mm.
It provides eight channels with the full functionality of amplifier, shaper, discriminator and base-line restorer.
Figure1 Block diagram of the ASDBLR circuit.
The competing requirements summarised in Table 1 led to the choice of a
largely differential circuit which includes detector tail compensation and a capacitively-coupled, diode
clamped baseline restorer. Separate high (15 fC range) and low (150 fC range) sensitivity discriminators
allow both tracking and TR photon detection on each channel. Selectable ion-tail compensation makes the
circuit compatible with both CF4 and Xe-based gases. Figure 2 shows the SPICE calculated
response of the circuit for a point ionisation at the output of the preamp (solid line) and shaper (dashed line).
The long tail evident in the preamplifier trace is due to the motion of the positive xenon ions in the gas.
A cancellation network in the shaper eliminates most of this extraneous signal allowing fast recovery for good
double pulse resolution.
The peaking time in the shaper has been set to 7.5 ns to optimise signal-to-noise, position
resolution, and double-pulse resolution. Cross-talk between channels has been measured to be
less than 0.2%.
Figure 2 Spice simulation of the output of the preamplifier (solid) and shaping circuit (dashed).
Figure 3 signals after baseline restoration for 2 and 25 fC (scaled) input charges.
The Baseline Restorer
At very high rates, a DC coupled circuit may have an unstable threshold due to pileup caused
by imperfections in the ion tail cancellation, or long integration times in other circuit elements.
There may be several microseconds of uncertain threshold or even persistent triggering due to
large energy depositions that saturate the tail cancellation network and leave the slowly decaying
ionisation current uncompensated. A simple baseline restorer (BLR) has been implemented
to limit these effects. A series capacitor and diode to virtual ground on each of the differential
shaper outputs provides a variable time-constant CR differentiation of the signal. The diode
current is set to 40 µA to provide a 5ns differentiation time constant in quiescent mode. The
exponential behaviour of the diode junction is used to moderate the diode impedance as the signal
is processed. The time constant is increased as the desired lobe of the signal is processed and
dramatically lowered when the signal returns to baseline to reduce the overshoot required to
recharge the capacitors. Figure 3 shows the shape of the signal at the shaper output for both
a minimum-sized 2 fC signal and a larger, average-sized 25 fC signal scaled to the 2 fC signal.
Larger signals have a smaller fraction of overshoot.
The BLR output is connected to a fast discriminator with sufficient hysteresis to guarantee a 4-5
ns minimum output width. The discriminator dead time is about 5 ns. A relatively fast shaping
can deteriorate the energy measurement critical for detecting transition radiation photons
which are only a few keV higher in energy than depositions due to ionising tracks. A photon
conversion presents a sharp, relatively large charge deposition which travels towards both ends
of the straw. With no termination, the signal going toward the far end reflects and returns to the
preamp delayed from the prompt signal by as much as 6 ns. The output signal amplitude depends on
signal shaping time and on the position where the particle crossed the straw. To reduce this effect
and improve the TR performance, the shaping time has been extended to 12 ns before the high-level discriminator.
This extra shaping time allows more efficient integration of the direct and reflected TR photon
signal which will allow for operation at higher effective threshold, increasing the deadband in
charge deposition between ionising tracks and the TR photons.
The output of the two discriminators is encoded into a programmable bi-level output current.
Figure 4 shows the 55 Fe response of the ASDBLR shaper (upper trace) and bi-level discriminator
(lower trace) when attached to a 4mm diameter straw filled with a Xenon-based gas.
The 50mV shaper signal in the left plot is below the high-level threshold and produces a single-level
discriminator output 15ns wide. The 250mV signal on the right is well above the high-level threshold
and produces two levels of discriminator output. The horizontal scale on these plots is 20 ns/div.
|Figure 4 Output of shaper (top) and discriminator (bottom) af ASDBLR (20ns/div). The plot on the left (right) corresponds to a signal below (above) the threshold for the TR discriminator.
The straws are terminated only at the electronics end. In order to avoid multiple reflections of the signal,
a good termination must be achieved at the input of the amplifier. The input stage of the ASDBLR has been
designed so that it provides a good termination and minimises the noise level. The frequency dependence of
the input impedance is shown in Figure 5.
Figure 5 ASDBLR input impedance characteristics.
Measurements with a real straw have shown good termination properties
of the input stage of the preamplifier and no significant reflections have been observed.
Several ASDBLR chips have been measured to determine channel to channel uniformity. The relation between
input signal amplitude (in terms of energy) and discriminator threshold (in terms of mV) is plotted in Figure
7 for sixteen channels. One can see that a 200 mV threshold level corresponds to a 250±25 eV energy
threshold for all 16 channels. For the high level threshold the uniformity is even better.
As mentioned above, the ASDBLR design has been optimised to minimise noise. Noise properties of the chip
have also been measured in a situation with one channel connected to a straw through a short cable (5 cm)
and a Lemo connector. The noise counting rate for different thresholds has been measured (Figure 6).
At a nominal threshold of 200 eV, the counting rate is less than 10 kHz which is well within the specifications
Dedicated studies of the ASDBLR performance at high counting rate have also been performed (up to 24 MHz).
All detailed studies have shown that, although some minor improvements remain to be implemented,
the ASDBLR performances are very close to the desired specifications.
Figure 6 Counting rate as a function of threshold.
|Figure 7 Low-threshold uniformity
||Figure 8 High-threshold uniformity
Text and pictures extracted from Atlas Inner Detector, Technical design Report