3.2 Digital Implementation
65
THRESHOLD
FLAG BIT
MINSEQ = 2
(a)Glitch filtering with minimum samples
above threshold of 2. Samples in solid
black are treated as if they were below
the threshold.
PRES = 0
POSTS = 2
FLAG BIT
THRESHOLD
(b)Merging of close clusters. Samples in red
are included to make one complete cluster.
Fig. 3.16 Zero suppression filtering and merging
the signal overcomes the threshold and a sequence of samples (post-samples) after
the signal returns below the threshold are also recorded (Fig. 3.15b). The number of
pre-samples and the number of post-samples can vary independently in the range
between 0 and 3 for pre-samples and 0 to 7 for post-samples.
To reduce the impulsive noise sensitivity, a glitch filter checks for a consecutive number of samples above threshold, confirming the existence of a real pulse
(Fig. 3.16a). The minimum sequence of samples above the threshold (glitch filter)
which defines a pulse can vary from 1 to 3, not including any pre- or post-samples.
The pulse thus identified and isolated must be tagged with a time stamp, in order
to be synchronized with the trigger decision for validation. Otherwise, the timing
information would be lost by the removal of a variable number of samples between
accepted pulses. This requires the addition of a time data to the set of sample data.
Besides that, in a data format where the addition of flag bits is not allowed, a further
word is needed to distinguish the sample data from the time data. This extra word
represents the number of words in the set. Since for each new set of data we have two
extra words, the merging of two consecutive sets, which are closer than 3 samples,
is performed (Fig. 3.16b).
In case of the cluster sum mode, the data in one cluster is summed together into a
20-bit word and the requirement for the minimum distance between clusters is then
changed to 3.
3.2.4.2 Lossy data compression
Due to the removal of a variable number of samples between accepted clusters in
the lossy encodings (zero suppression, cluster sum), the timing information would
be lost in the process. This requires the addition of a time-data to each accepted
set of samples. Since 1023 is the maximum length of the data stream that can be
processed by the SAMPA chip, the time information can be encoded in a 10-bit
65
THRESHOLD
FLAG BIT
MINSEQ = 2
(a)Glitch filtering with minimum samples
above threshold of 2. Samples in solid
black are treated as if they were below
the threshold.
PRES = 0
POSTS = 2
FLAG BIT
THRESHOLD
(b)Merging of close clusters. Samples in red
are included to make one complete cluster.
Fig. 3.16 Zero suppression filtering and merging
the signal overcomes the threshold and a sequence of samples (post-samples) after
the signal returns below the threshold are also recorded (Fig. 3.15b). The number of
pre-samples and the number of post-samples can vary independently in the range
between 0 and 3 for pre-samples and 0 to 7 for post-samples.
To reduce the impulsive noise sensitivity, a glitch filter checks for a consecutive number of samples above threshold, confirming the existence of a real pulse
(Fig. 3.16a). The minimum sequence of samples above the threshold (glitch filter)
which defines a pulse can vary from 1 to 3, not including any pre- or post-samples.
The pulse thus identified and isolated must be tagged with a time stamp, in order
to be synchronized with the trigger decision for validation. Otherwise, the timing
information would be lost by the removal of a variable number of samples between
accepted pulses. This requires the addition of a time data to the set of sample data.
Besides that, in a data format where the addition of flag bits is not allowed, a further
word is needed to distinguish the sample data from the time data. This extra word
represents the number of words in the set. Since for each new set of data we have two
extra words, the merging of two consecutive sets, which are closer than 3 samples,
is performed (Fig. 3.16b).
In case of the cluster sum mode, the data in one cluster is summed together into a
20-bit word and the requirement for the minimum distance between clusters is then
changed to 3.
3.2.4.2 Lossy data compression
Due to the removal of a variable number of samples between accepted clusters in
the lossy encodings (zero suppression, cluster sum), the timing information would
be lost in the process. This requires the addition of a time-data to each accepted
set of samples. Since 1023 is the maximum length of the data stream that can be
processed by the SAMPA chip, the time information can be encoded in a 10-bit