24
2 Front-End Electronics
At the submission of the second prototype of the SAMPA a test mode was available
that could serialize all the data from the 32 channels when running with a serial speed
that is 32 times the sampling speed. The data is output on 10 serial links where the
10-bit samples from each channel are sent consecutively in one sampling period.
Running the SAMPA at the originally envisioned sampling speed of 10 MHz
would though produce 16 Gbps of raw data from the five devices, whereas the two
GBTx devices can only forward 6.4 Gbps in the normal operation mode with FEC,
or 8.96 Gbps in the Wide Bus Mode where the FEC is turned off [24]. It is thus
necessary to halve the sampling frequency to 5 MHz to avoid doubling the number
of GBTx devices and optical links. An uplink speed of only 8 Gbps is then required.
Simulations have proved that operating at 5 MHz does not degrade the detector performance [23].
As one of the SAMPAs will be connected to two GBTx, which might connect to
different CRUs, there also needs to be an option to serialize half the channels on five
links.
2.3.6 MCH Specific Requirements
With a drift length of only 2.5 mm in the detector, there are no issues with pileup
from multiple events for the MCH like there is for the TPC. It also means the signal
from an event will only be a single pulse of a few samples for the pads that are hit.
This means the data amount per device will be very small, which requires another
readout architecture than the TPC to keep the amount of GBTx devices and CRUs
low.
2.3.6.1 Daisy Chaining
The MCH hit rate should be low enough so the data traffic from one chip can be
routed through only one serial output link with a bandwidth of 40 Mbps. One of the
two chips on the front-end card sends its data-readout stream to the other chip on the
same front-end card, which then merges the data stream of its neighbour to its own
data stream. In this way, each front-end card has only one single serial output link
of 80 Mbps. This effectively halves the number of GBTx and optical links needed,
significantly reducing the cost of the installation.
2.3.6.2 Data Compression Using Zero Suppression and Cluster
Summing
Two devices of 32 channels each operating at 10 MHz produce in total 6.4 Gb/s
of raw data. To reduce the data amount below the serial output link bandwidth of
80 Mb/s an efficient compression must be employed. Due to the bandwidth limit,
2 Front-End Electronics
At the submission of the second prototype of the SAMPA a test mode was available
that could serialize all the data from the 32 channels when running with a serial speed
that is 32 times the sampling speed. The data is output on 10 serial links where the
10-bit samples from each channel are sent consecutively in one sampling period.
Running the SAMPA at the originally envisioned sampling speed of 10 MHz
would though produce 16 Gbps of raw data from the five devices, whereas the two
GBTx devices can only forward 6.4 Gbps in the normal operation mode with FEC,
or 8.96 Gbps in the Wide Bus Mode where the FEC is turned off [24]. It is thus
necessary to halve the sampling frequency to 5 MHz to avoid doubling the number
of GBTx devices and optical links. An uplink speed of only 8 Gbps is then required.
Simulations have proved that operating at 5 MHz does not degrade the detector performance [23].
As one of the SAMPAs will be connected to two GBTx, which might connect to
different CRUs, there also needs to be an option to serialize half the channels on five
links.
2.3.6 MCH Specific Requirements
With a drift length of only 2.5 mm in the detector, there are no issues with pileup
from multiple events for the MCH like there is for the TPC. It also means the signal
from an event will only be a single pulse of a few samples for the pads that are hit.
This means the data amount per device will be very small, which requires another
readout architecture than the TPC to keep the amount of GBTx devices and CRUs
low.
2.3.6.1 Daisy Chaining
The MCH hit rate should be low enough so the data traffic from one chip can be
routed through only one serial output link with a bandwidth of 40 Mbps. One of the
two chips on the front-end card sends its data-readout stream to the other chip on the
same front-end card, which then merges the data stream of its neighbour to its own
data stream. In this way, each front-end card has only one single serial output link
of 80 Mbps. This effectively halves the number of GBTx and optical links needed,
significantly reducing the cost of the installation.
2.3.6.2 Data Compression Using Zero Suppression and Cluster
Summing
Two devices of 32 channels each operating at 10 MHz produce in total 6.4 Gb/s
of raw data. To reduce the data amount below the serial output link bandwidth of
80 Mb/s an efficient compression must be employed. Due to the bandwidth limit,