Skip to content

Fix deep matplotlib waveform plot

Luca Scotto Lavina requested to merge fix_records_plot into master

Created by: JelleAalbers

@tunnel reported that plotting large time ranges (> few ms) with st.plot_waveform(..., deep=True) can result in an empty records plot.

index

I suspect the cause is that the records matrix became very large (many GBs), causing rendering goofs.

This modifies st.records_matrix (used by the waveform plot) to create downsampled matrices if the requested time axis exceeds 20_000 samples (configurable through the max_samples argument). This solves the problem:

import straxen
st = straxen.contexts.xenon1t_dali()
run_id = '170204_1710'
df = st.get_array(run_id, "event_info")
event = df[4]
st.plot_waveform(run_id, time_within=event, deep=True)

bla

It also has other benefits:

  • We actually fill a smaller matrix with downsampled waveforms, rather than downsampling a huge matrix. Thus preparing the plot (not just the rendering itself) is much faster and takes far less memory.
  • You can now plot very large time ranges with the matplotlib display -- though the color scale inference could use improvement -- e.g.
st.plot_waveform('180215_1029', seconds_range=(0, 10), deep=True)

bla_long

Merge request reports