Drs4 time corr #236

pawel21 · 2019-11-27T21:40:29Z

Hi,

I have created code to DRS4 time correction.
We have two class TimeCorrectionCalculate to create file with coefficients for time correction curve
of chip DRS4 and PulseTimeCorrection to make time correction on pulse time from extractor.
I added notebook time_calibration.ipynb which show how to create h5py file and how to apply correction to camera calibration run and cosmic events.

@FrancaCassol Please look on my code and tell what do you think?
I think, we should ,,connect" time correction with flatfield calibration

What is the status of gain selection? Because, now code is designed for waveform with both gains.

FrancaCassol · 2019-11-28T07:54:29Z

Hi @pawel21,

I can not test the code before this evening, some questions for the moment:

could you please describe the method and algorithms you use for deriving and applying the corrections?
In the notebook I see you modify not only the width but also the mean value of the time, why?
is your correction dependent from the charge extractor you use?
you speak of h5py, but I guess it is a normal hdf5 file, isn't it? (in this case better you speak of hdf5)

pawel21 · 2019-11-28T11:34:14Z

Hi @pawel21,

I can not test the code before this evening, some questions for the moment:

No problem.

* could you please describe the method and algorithms you use for deriving and applying the corrections?

,,DRS4 chips exhibit an delay of typically 1 ns (up to 4 ns), depending on the absolute location of the signal pulse in the domino ring. The differences can be calibrated with the use of calibration pulses. The computed mean arrival time of a pulse in a channel is a rather complicated function of the time sample position in the ring and differs from one DRS chip to another. For each channel we expand this function into a Fourier series to obtain the correction function." (from Analysis techniques and performance of the Domino Ring Sampler version 4 based readout for the MAGIC telescopes ) To computed mean arrival time of a pulse I am using LocalPeakWindowSum (window_width and window_shift user can set up), then I fit data using Fourier series, I obtain coefficients, which I save to hdf5 file. Then, I am using "correction curve" (which coefficients from previous step) to correct arrival time (pulse time) from extractor in each pixel. For single event time resolution is decreased from 1.6 ns to 0.4 ns.

* In the notebook I see you modify not only the width but also the mean value of the time, why?

The idea of time correction is the following: we got arrival time from calibration pulse (charge, pulse_time = extractor(ev.r1.tel[1].waveform) and then using "correction curve" in class PulseTimeCorrection we got arrival time after correction pulse_corr_array = PulseTimeCorrection.get_corr_pulse(ev, pulse_time). So, my code modify pulse_time

* is your correction dependent from the charge extractor you use?

Yes, this is not big differences. User can set up parameters for LocalPeakWindowSum (I added example in notebook"). It is best to use the same parameters for calibration and for processing data.

* you speak of h5py, but I guess it is a normal hdf5 file, isn't it? (in this case better you speak of hdf5)

Yes, I use normal hdf5 file. Thank you.

FrancaCassol · 2019-11-28T21:15:30Z

Hi @pawel21,

thanks for your explanation. Could you please add the link to the MAGIC document to the doc of the of the classes?

Also, I wonder, is this correction compatible with the sampling time correction of Yuto? Do you know if the correction proposed by Yuto is applied in Magic?

pawel21 · 2019-11-29T09:59:18Z

Hi @pawel21,

thanks for your explanation. Could you please add the link to the MAGIC document to the doc of the of the classes?

Yes, sure

Also, I wonder, is this correction compatible with the sampling time correction of Yuto? Do you know if the correction proposed by Yuto is applied in Magic?

Yes, should be compatible. My method will be applied to time, and "sampling time correction" meted will be applied to the value of the reproduced signal, then they should not bite.
Sampling time correction of Yuko is used to improve the extracted signal. This method is not used in MAGIC.

FrancaCassol

Hi @pawel,
thanks for your reply. I just realised that I didn't submit these more specifics questions

FrancaCassol · 2019-11-28T21:22:36Z

lstchain/calib/camera/time_correction_calculate.py

+                          allow_none=True).tag(config=True)
+
+
+    r1_sample_end = Int(default_value=38,


@pawel21,
why do you need to give the r1_sample_start and r1_sample_end?
This is already defined in the R0 calibrator. At this level you should be able to work with the full r1 and forget r0. Isn't it?

Yes, You are right. Thank you

FrancaCassol · 2019-11-28T21:24:24Z

lstchain/calib/camera/time_correction_calculate.py

+        event : `ctapipe` event-container
+        """
+        if event.r0.tel[self.tel_id].trigger_type == 1:
+            for nr_module in prange(0, n_modules):


why do you use r0 trigger and not r1 trigger?

What is the difference between trigger and r1 trigger? But, I agree use r1 trigger will be more convenient

Hi @pawel21,
there is not difference, at least for the moment. But, after the r0 calibration, I think it is better to work only with r1 elements

FrancaCassol · 2019-11-28T21:30:44Z

lstchain/calib/camera/time_correction_calculate.py

+
+            pixel_ids = event.lst.tel[self.tel_id].svc.pixel_ids
+            charge, pulse_time = self.extractor(event.r1.tel[self.tel_id].waveform[:, :, self.r1_sample_start:self.r1_sample_end])
+            self.calib_pulse_time_jit(charge,


Same question as before, after the LSTR0Corrections r1 samples are 36 and we should be able to use all of them, why do you select a range to be given to the extractor?

I removed. Now, my code use r1 samples. Thank you

FrancaCassol · 2019-11-28T21:35:42Z

lstchain/calib/camera/time_correction_calculate.py

+
+n_gain = 2
+n_channel = 7
+n_modules = 265


Is this compatible with eventually missing modules?

This wasn't compatible. I have modified code to make correction only for working modules ( n_modules_from_event = event.lst.tel[self.tel_id].svc.num_modules line 69 in pulse_time_correction.py)

rlopezcoto · 2019-11-30T15:35:38Z

Thanks @pawel21 for the code and @FrancaCassol for the review. This PR is very necessary to improve signal extraction.

pawel21 added 5 commits November 26, 2019 16:20

added new code to pulse time correction in DRS4

fd7ba39

added jit function to speed up code

506f3d7

clean up code to pulse corr

9e162bb

cleaun up code, add comments

ff6300d

Added comments and cleaned up notebook

8f6bb2a

added set up param to extractor in notebook

8cd73ca

FrancaCassol reviewed Nov 29, 2019

View reviewed changes

change according Franca comments

77c2fd1

FrancaCassol approved these changes Nov 30, 2019

View reviewed changes

rlopezcoto merged commit 4a8732d into cta-observatory:master Nov 30, 2019

pawel21 deleted the drs4_time_corr branch July 10, 2020 11:04

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Drs4 time corr #236

Drs4 time corr #236

pawel21 commented Nov 27, 2019 •

edited

Loading

FrancaCassol commented Nov 28, 2019

pawel21 commented Nov 28, 2019 •

edited

Loading

FrancaCassol commented Nov 28, 2019

pawel21 commented Nov 29, 2019

FrancaCassol left a comment •

edited

Loading

FrancaCassol Nov 28, 2019

pawel21 Nov 29, 2019

FrancaCassol Nov 28, 2019

pawel21 Nov 29, 2019

FrancaCassol Nov 30, 2019

FrancaCassol Nov 28, 2019

pawel21 Nov 29, 2019

FrancaCassol Nov 28, 2019

pawel21 Nov 29, 2019

rlopezcoto commented Nov 30, 2019

		allow_none=True).tag(config=True)


		r1_sample_end = Int(default_value=38,

Drs4 time corr #236

Drs4 time corr #236

Conversation

pawel21 commented Nov 27, 2019 • edited Loading

FrancaCassol commented Nov 28, 2019

pawel21 commented Nov 28, 2019 • edited Loading

FrancaCassol commented Nov 28, 2019

pawel21 commented Nov 29, 2019

FrancaCassol left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rlopezcoto commented Nov 30, 2019

pawel21 commented Nov 27, 2019 •

edited

Loading

pawel21 commented Nov 28, 2019 •

edited

Loading

FrancaCassol left a comment •

edited

Loading