3.2 Digital Implementation
61
Fig. 3.12 Moving average principle
The glitch filtering implemented for the GdSP has been kept. The glitch filter,
when enabled, prevents the activation of the post- and pre-samples when a single
sample is outside the threshold. This makes it possible to keep the thresholds closer
to the noise, which will provide more precision for the pulse exclusion.
3.2.3.4 Filter Changes and Improvements
As reported in [9], the ALICE TPC had previously experienced issues with the BC2
filter of the ALTRO, an example of this is shown in Fig. 3.13. In certain cases where
the baseline is shifting down and up again, like seen at around time-bin 650, and a
pulse is riding on top of the baseline change, the slope change might be steeper than
what the filter can handle and the input signal settles at a level that is above the upper
threshold for updating the baseline. Consequently, the baseline will not be updated
and all the values on the output of the filter will be larger than zero.
The zero suppression compression that follows the filtering stages has the task
of suppressing all values below a certain threshold value before storage, to save
bandwidth and buffer space. The threshold level is usually set at zero, or at a few
counts above, and everything below this level is suppressed and run length encoded.
When all the values are above zero, all the samples are considered as one large
cluster and every sample of the event is stored in the buffer. This quickly fills up all
the available space, as the output bandwidth is not normally designed to handle the
equivalent of a raw data readout.
The ALTRO has no way of resetting the filter, except by resetting the complete
device, which means that the configuration registers needs to be reprogrammed as
well. Like in the S-ALTRO, the SAMPA device also incorporates a soft reset functionality that resets the device, but keeps the configuration registers intact. To use
the soft reset to recover from a stuck baseline, the upstream device would need to
monitor the buffer overflow flag in the header of the data packets. As this is not an
61
Fig. 3.12 Moving average principle
The glitch filtering implemented for the GdSP has been kept. The glitch filter,
when enabled, prevents the activation of the post- and pre-samples when a single
sample is outside the threshold. This makes it possible to keep the thresholds closer
to the noise, which will provide more precision for the pulse exclusion.
3.2.3.4 Filter Changes and Improvements
As reported in [9], the ALICE TPC had previously experienced issues with the BC2
filter of the ALTRO, an example of this is shown in Fig. 3.13. In certain cases where
the baseline is shifting down and up again, like seen at around time-bin 650, and a
pulse is riding on top of the baseline change, the slope change might be steeper than
what the filter can handle and the input signal settles at a level that is above the upper
threshold for updating the baseline. Consequently, the baseline will not be updated
and all the values on the output of the filter will be larger than zero.
The zero suppression compression that follows the filtering stages has the task
of suppressing all values below a certain threshold value before storage, to save
bandwidth and buffer space. The threshold level is usually set at zero, or at a few
counts above, and everything below this level is suppressed and run length encoded.
When all the values are above zero, all the samples are considered as one large
cluster and every sample of the event is stored in the buffer. This quickly fills up all
the available space, as the output bandwidth is not normally designed to handle the
equivalent of a raw data readout.
The ALTRO has no way of resetting the filter, except by resetting the complete
device, which means that the configuration registers needs to be reprogrammed as
well. Like in the S-ALTRO, the SAMPA device also incorporates a soft reset functionality that resets the device, but keeps the configuration registers intact. To use
the soft reset to recover from a stuck baseline, the upstream device would need to
monitor the buffer overflow flag in the header of the data packets. As this is not an