Changed the way how the baseline rms is computed. Created new dtype raw_records which is identical to the previous records dtype.

Added a new field to the record dytpe called rms.

Author: Wenz

strax/dtypes.py

@@ -6,11 +6,11 @@
 """
 import numpy as np
 
-__all__ = ('interval_dtype record_dtype hit_dtype peak_dtype '
+__all__ = ('interval_dtype record_dtype raw_record_dtype hit_dtype peak_dtype '
            'DIGITAL_SUM_WAVEFORM_CHANNEL DEFAULT_RECORD_LENGTH').split()
 
 DIGITAL_SUM_WAVEFORM_CHANNEL = -1
-DEFAULT_RECORD_LENGTH = 110
+DEFAULT_RECORD_LENGTH = 600
 
 
 # Base dtype for interval-like objects (pulse, peak, hit)
@@ -29,6 +29,7 @@
     # and int32 for per-channel waveforms (area in ADC x samples)
 ]
 
+
 def raw_record_dtype(samples_per_record=DEFAULT_RECORD_LENGTH):
     """Data type for a waveform raw_record.
 
@@ -60,9 +61,9 @@ def record_dtype(samples_per_record=DEFAULT_RECORD_LENGTH):
     Length can be shorter than the number of samples in data,
     this indicates a record with zero-padding at the end.
     """
-    return interval_dtype + raw_record_dtype(samples_per_record) + [
+    return raw_record_dtype(samples_per_record) + [
         (('Baseline RMS in ADC counts. data = baseline - data_orig',
-          'baseline'), np.float32)
+          'rms'), np.float32)
     ]
 
 

strax/processing/pulse_processing.py

@@ -7,7 +7,7 @@
 
 import strax
 export, __all__ = strax.exporter()
-__all__ += ['NO_RECORD_LINK']
+__all__ += ['NO_RECORD_LINK', 'baseline_rms']
 
 # Constant for use in record_links, to indicate there is no prev/next record
 NO_RECORD_LINK = -1
@@ -31,26 +31,58 @@ def baseline(records, baseline_samples=40):
     # We only care about the channels in this set of records; a single .max()
     # is worth avoiding the hassle of passing n_channels around
     last_bl_in = np.zeros(records['channel'].max() + 1, dtype=np.int16)
-    last_rms_in = np.zeros(records['channel'].max() + 1, dtype=np.int16)
 
     for d_i, d in enumerate(records):
 
         # Compute the baseline if we're the first record of the pulse,
         # otherwise take the last baseline we've seen in the channel
         if d.record_i == 0:
             bl = last_bl_in[d.channel] = d.data[:baseline_samples].mean()
-            rms = _baseline_rms(d['data'], bl, baseline_samples)
         else:
             bl = last_bl_in[d.channel]
-            rms = last_rms_in[d['channel']]
 
         # Subtract baseline from all data samples in the record
         # (any additional zeros should be kept at zero)
         last = min(samples_per_record,
                    d.pulse_length - d.record_i * samples_per_record)
         d.data[:last] = int(bl) - d.data[:last]
         d.baseline = bl
-        d['rms'] = rms
+
+
+@export
+@numba.njit(cache=True, nogil=True)
+def baseline_rms(records, nsampels=40):
+    """
+    Function which estimates the baseline rms within a certain number of samples.
+
+    The rms value is estimated for all samples with adc counts <= 0.
+
+    Args:
+        records (np.array): Array of the data_kind raw_records or records.
+
+    Keyword Args:
+        nsampels (int): First n samples on which the rms is estimated.
+
+    Note:
+
+
+    Returns:
+        np.array: array of the length of records containing the rms values.
+    """
+    # Init result and temp_rms storage:
+    res = np.zeros(len(records))
+    last_rms_in = np.zeros(records['channel'].max() + 1, dtype=np.float32)
+
+    for ind, r in enumerate(records):
+        if r['record_i'] == 0:
+            rms = _baseline_rms(r['data'], r['baseline']%1, nsampels)
+            last_rms_in[r['channel']] = rms
+            res[ind] = rms
+        else:
+            # if higher fragment take previous rms value:
+            res[ind] = last_rms_in[r['channel']]
+
+    return res
 
 
 @numba.njit(cache=True, nogil=True)
@@ -67,7 +99,7 @@ def _baseline_rms(d, b, n_samples=40):
     Keyword Args:
         n_samples (int): First n samples on which the rms is estimated.
     """
-    d = b - d
+    d = b + d
     n = 0
     rms = 0
     for s in d[:n_samples]: