QT3Scope update to display rolling mean line in figure and acquired values as text, and fix animation-related warning. #117
Conversation
…`oscilloscope.py`.
src/applications/oscilloscope.py
Outdated
| def get_new_rolling_mean_val(self): | ||
| """ Calculates new rolling mean value. """ | ||
| values_of_interest = list(self.ydata)[len(self.ydata) - 2 * self.half_window_size: | ||
| len(self.ydata)] | ||
| return np.mean(values_of_interest) | ||
|
|
||
| def get_new_rolling_stdev_val(self): | ||
| """ Calculates new rolling standard deviation. """ | ||
| values_of_interest = list(self.ydata)[len(self.ydata) - 2 * self.half_window_size: | ||
| len(self.ydata)] | ||
| return np.std(values_of_interest) |
There was a problem hiding this comment.
@vasilisniaouris I think this is more efficient (~2x as fast according to stack overflow -- although perhaps we won't see that much speedup since our deque is smaller than the one used in the measurement that I read)
import itertools
values_of_interest = list(itertools.islice(self.ydata, start, stop))
Also, if we added 'more_itertools' to the project, we could use islice_extended which allows for negative values
values_of_interest = list(more_itertools.islice_extended(self.ydata, -2*self.half_window_size, 0))
There was a problem hiding this comment.
oh that looks good! I will implement it
src/applications/oscilloscope.py
Outdated
| if self.show_rolling_mean: | ||
| new_rm_value = self.update_rolling_mean() | ||
| self.rolling_mean_line.set_ydata(self.rolling_mean) | ||
|
|
||
| if self.show_reading_text: | ||
| text = self._get_text_value(y) | ||
| if self.show_rolling_mean: | ||
| text = f'{text}\n {self._get_text_value(new_rm_value)}' | ||
| measured_stdev = self.get_new_rolling_stdev_val() | ||
| expected_stdev = np.sqrt(new_rm_value) | ||
| text = f'{text}\n {self._get_text_value(measured_stdev)} ({self._get_text_value(expected_stdev)})' | ||
| self.current_value_text.set_text(text) |
There was a problem hiding this comment.
So, instead of creating a list in memory for rolling mean and updating it every time, can we just use the scipy.uniform_filter1d here directly on self.ydata at each update to generate a full rolling mean? Or is that incredibly slow and why you chose not to use that?
There was a problem hiding this comment.
I am not sure how slow it would be, I haven't tested it. At the time of writing, I thought it would be best to not call the filter, since it is a convolution function (if I remember correctly) and I believe they can get quite expensive.
| rolling_mean[:self.half_window_size] = np.nan | ||
| rolling_mean[len(self.ydata) - self.half_window_size:] = np.nan | ||
| return rolling_mean | ||
|
|
There was a problem hiding this comment.
See message above asking if we need to store a self.rolling_mean list in memory and perform this popleft / append step every time. It makes sense if using scipy to compute the full rolling mean is slow.
src/applications/oscilloscope.py
Outdated
| # TODO: Put all axis preferences to .mplstyle and load through terminal or gui. | ||
| self.ax.set_ylabel('Reading (counts / sec)') | ||
| self.ax.ticklabel_format(style='sci', scilimits=(-3, 4), axis='y') | ||
| self.ax.set_xlabel('Acquisition Number') |
There was a problem hiding this comment.
It would be nice if we could format the x-axis so it displays values of time..... Any other labels we could use here? "Time bin"? or "Time (arb. units)" until somebody does the work to convert the values to actual time values.
There was a problem hiding this comment.
I am planning on implementing a different method on collecting data with the new device interface class, which will always include time stamps. For now, maybe we can multiply the index number with the acquisition period?
There was a problem hiding this comment.
looking into it, I realized that the ScopeFigure does not know anything about the update interval, so we should probably leave it that way for now. Updated it to "Time bin (arb. units)"
|
Just an update:
|
|
Hey @vasilisniaouris -- I totally forgot about this PR. I'm wondering if we can cherry pick some of the code in here after we've finished the interface. It will probably be difficult to merge them together at this point, unfortunately. |
Hello all. This is my attempt to tackle issues 54 and 93. I have tested this code on my own setup in the QDL lab. I am attaching two example pictures of how it looks like with random data and with real data.