22
2 Front-End Electronics
2.3.4 TPC Specific Requirements
The main TPC digital requirements were altered during the final submission phase
of the second prototype due to new simulation results. As there are no major digital
feature changes between the second prototype and the final device and since most
of the digital implementation choices are based on the original requirements, the
original requirements will also be presented here for clarity.
2.3.4.1 Digital Filter
The signals from the GEM detector do not feature a tail like on the MWPC and so
the necessity for a tail cancellation filter is no longer present.
The baseline correction of non-systematic effects might still be useful, this is
where a moving average value is subtracted from the current value to remove perturbations of the baseline. This filter was present in the ALTRO, but could not be
utilized due to sudden swings in the baseline, which caused the calculation to fail
and so also the zero suppression.
2.3.4.2 Other Compression Scenarios
In detectors like the TPC, the data output contains primarily small sample values,
while large values are more rare. The distribution of sample values is approximately
exponential [21]. By employing Huffman coding [22] the data size can be reduced
by assigning shorter codes to frequent values and longer ones for more rare values.
For occupancies above 25%, the Huffman coding will yield better compression factor than zero suppression [18], at lower occupancies the zero suppression is more
efficient, though the Huffman coding is lossless so signals that are close to the noise
will be kept, improving later charge calculations.
2.3.4.3 Readout Ordering
The charge of a track hitting the pad plane might be spread over multiple pads and
so the signals from several pads are needed to find the centre of the charge. As the
pads where a charge is spread over might not all be pads that are connected to one
SAMPA, the calculation needs to be done later where the information from several
SAMPAs is available. The central position and the total charge of each cluster hitting
the pad plane is therefore planned to be calculated in the CRU. By recording only
the position and charge of a cluster, the amount of data that needs to be stored
for analysis is greatly reduced. This would though require that the data from each
channel connected to a pad row arrive together in the CRU. The more out of order
the data arrives, the more buffering is needed to get the data aligned. The possibility
2 Front-End Electronics
2.3.4 TPC Specific Requirements
The main TPC digital requirements were altered during the final submission phase
of the second prototype due to new simulation results. As there are no major digital
feature changes between the second prototype and the final device and since most
of the digital implementation choices are based on the original requirements, the
original requirements will also be presented here for clarity.
2.3.4.1 Digital Filter
The signals from the GEM detector do not feature a tail like on the MWPC and so
the necessity for a tail cancellation filter is no longer present.
The baseline correction of non-systematic effects might still be useful, this is
where a moving average value is subtracted from the current value to remove perturbations of the baseline. This filter was present in the ALTRO, but could not be
utilized due to sudden swings in the baseline, which caused the calculation to fail
and so also the zero suppression.
2.3.4.2 Other Compression Scenarios
In detectors like the TPC, the data output contains primarily small sample values,
while large values are more rare. The distribution of sample values is approximately
exponential [21]. By employing Huffman coding [22] the data size can be reduced
by assigning shorter codes to frequent values and longer ones for more rare values.
For occupancies above 25%, the Huffman coding will yield better compression factor than zero suppression [18], at lower occupancies the zero suppression is more
efficient, though the Huffman coding is lossless so signals that are close to the noise
will be kept, improving later charge calculations.
2.3.4.3 Readout Ordering
The charge of a track hitting the pad plane might be spread over multiple pads and
so the signals from several pads are needed to find the centre of the charge. As the
pads where a charge is spread over might not all be pads that are connected to one
SAMPA, the calculation needs to be done later where the information from several
SAMPAs is available. The central position and the total charge of each cluster hitting
the pad plane is therefore planned to be calculated in the CRU. By recording only
the position and charge of a cluster, the amount of data that needs to be stored
for analysis is greatly reduced. This would though require that the data from each
channel connected to a pad row arrive together in the CRU. The more out of order
the data arrives, the more buffering is needed to get the data aligned. The possibility