Differences

This shows you the differences between two versions of the page.

--- atmos:citation:research:soda_proecessing_1d_and_2d_hail_2d_hvps1 [2024/05/20 18:16] – klinman
+++ atmos:citation:research:soda_proecessing_1d_and_2d_hail_2d_hvps1 [2024/05/21 14:23] (current) – klinman
@@ Line 1: / Line 1: @@
-On board the T28 aircraft was a Hail Spectrometer and High Volume Particle Spectrometer (HVPS). Both probes output 2D data, and the Hail Spectrometer also output 1D data. The HVPS had an array of 256 diodes, the Hail Spectrometer had 126/64 diodes for 1D/2D data.
+On board the T28 aircraft was a Hail Spectrometer and High Volume Particle Spectrometer (HVPS). Both probes output 2D data, and the Hail Spectrometer also output 1D data. The HVPS had an array of 256 diodes, the Hail Spectrometer had 126/64 diodes for 1D/2D data. If image elements were blocked above a certain threshold a particle(s) would be counted. The steps to use these counts to calculate particle concentration (#/m^4) are detailed below.
-The Hail Spectrometer/HVPS had an array of  image elements .9mm in size. For the 1D outputs, if say a particle were to block four consecutive image elements, then the system would record the top and bottom blocked diode as the size of the particle and output a sized particle based on that. There is no size dimension along the width of the particle or the direction of travel. Once the diodes were no longer blocked, then the probe would “?refresh?” and wait for the next particle to pass through. Because the probe was made for larger particles, it can be assumed that coincident particles are rare. However, in the event there were coincident particles, unless the probe knew to filter them out and ignore them, then an erroneously large particle could be recorded.
+The counts generated from SODA are saved in a 14xn array where n is the number of x second intervals, where x can be specified when using the SODA gui. x is set to 5 seconds as a default in the gui, and is what has been used in the research so far. An example of these counts is shown below,
+{{:atmos:citation:research:count_comparison_header.png?900|}}
+{{:atmos:citation:research:count_comparison.png?900|}}
+The time interval for 241846 - 241851 is shown. Note that the time intervals count up to 5 seconds forward from the start time. The counts from SODA are in the "SODA" row, and you'll notice that they are much larger than the DISP counts which are the 1D counts. Usually they're larger by factors of 5-10. This will be (partially) resolved later on.
+To calculate concentration from counts, the basic forumla is,
-Notes for self
+    Conc (#/m^4) = Counts / (SA * Airspeed * Binsize change)
-  * np.ceil or np.round
-        * for cases where 1 count, but soda does correction of * 1.1 / 5
-              * so the 1 count would be 1 *1.1 /5 = ~.2     does the .2 count as a count since the /5 is a result of the 5 seconds?
-  * hail counts from disp count forward. so 241841 is 41 42 43 44 45
-  * cant say i understand it but in aarons code, the counts are multiplied by the 1.1 correction factor. then the conc is /5 which is the active time.
-        * what doesn't make sense is that the counts are ~5x higher than the disp file, which is wrong because when the concentration is calculated, it takes into account that 5x factor and becomes decently close to the disp calculated conc
-        * what also doesnt make sense is that the concentration from calculating it myself, using that /5 and *1.1 is that the concentration i calculate is consistently ~1.1 times higher than the concentration aaron gets.
-              * could i be looking at the wrong code
-  * ok update to previous bullet point. just not multiplying by 1.1 seems to be the move???
-        * i think im missing where in aarons code he actually makes the concentration file because something is off here
-              * and im still confused about why the counts should be /5 but thats not the step that I see taken in the code
-  * aaron takes conc1d and prints it to the ascii file and does nothing with it except format it
+where
+  * SA is sample area which is (probe arm length * height)
+    * Height is (number of diodes * diode size)
+  * Binsize change is the (Max_Size - Min_Size) for each channel.
+    * For channel 1 for the Hail Spectrometer, this would be 0.50cm - 0.40cm = 0.10cm
+SODA includes extra variables in it's calculations. I can not explain why SODA includes these extra factors, but they are in there, and when included, the resulting concentration is agreeably close to the concentration calculated from the 1D counts. This is why the counts are "(partially) resolved."
+  * Counts are multiplied by a correction factor
+    * This is the Poisson correction factor.
+  * In additions to SA, Airspeed, and Binsize change, SODA also divides by "activetime"
+    * "activetime" is the time interval. SODA defaults to 5 seconds for this interval
-Ok so, notes for making this page.
+The new formula is,
-  * all the counts check out. would be good to show this and include where the time starts
-  * the sample volume checks out. show this too, but not all the soda details. but a comparison of numbers would be good
+    Conc (#/m^4) = (Counts * corr_fac) / (SA * Airspeed * Binsize change * activetime)
-  * explain the differene between sodas counts and the disp counts and  how if you divide by 5 theyre the saame but im not sure why. however, it works out in the concentration with the divide by 5
-    * the counts are multipliued by a correction factor 1.1
+where for the Hail Spectrometer,
-  * with all that considered the concs match well and the math was reproducable using the counts from soda and the numbers, and the conc ultimately lines up decently with the 1D
+  * Probe arm length: 0.99 m (SODA version)
-  * the corr fac is the poisson corr factor
+  * Diode size: 0.009 m
-    * not sure why this is needed tho
+  * Diodes: 64 (SODA version)
-  * the active time correction is for how ever long the chunk is
+  * Airspeed: ~100m/s
-    * not sure why it should be divided by 5? figure it should be a sum so no need for division but oh well it works so whatever
+  * Binsize change: 0.001, 0.001, 0.001, 0.001, 0.0014, 0.0016, 0.002, 0.0026, 0.003, 0.0031, 0.0044, 0.006, 0.0079, 0.01
-  * why the 126/64 diode change?
+  * Corr_fac: ~1.1
+  * activetime: 5
+When you consider the counts being multiplied by the correction factor and divided by the active time, the 2D counts from SODA better match the 1D counts. In the image below, the 2D counts generated by SODA are much closer to the 1D counts from the DISP file. These "corrected counts" agree pretty well with the 1D counts.
+{{:atmos:citation:research:counts_adjustsed_2d.png?900|}}
+With the counts from the first image plugged into the second formula, the resulting concentration is,
+{{:atmos:citation:research:conc_compairon_header.png?900|}}
+{{:atmos:citation:research:soda_conc_comparison.png?900|}}
+The SODA concentrations agree fairly well with the DISP concentrations.
+Notes to self on information to include, mention, explore, etc,
+  * Counts rounded. Ceiling or round up/down? Which is better representation of data?
+  * Time counts forward from time denoted for interval
+  * Idk why the counts are 5x higher. Idk why the correction factor. Idk why the divide by active time. Idk
+  * Sometimes, although I believe this is fixed in my math now, my calculated 2D concentrations are 1.1x higher. But again, I think this isn't an issue anymore.
+  * Sample volumes check out. Airspeed checks out. SODA uses one airspeed for whole time interval. Include a comparison of the numbers
+  * Include a comparison of the different probe specs used in 1D and 2D calculation
   *
+to get rid of sec from airspeed, i like to think of it as counts per time interval. so 100 counts per 5 second interval would be 100/5s which gives the seconds on bottom to cancel out the seconds in the airspeed. 10 counts per 1 second likewise gives 10/1s. and all of this normalizes the concentration to one second, which is why the 5 second soda intervals and 1 second disp intervals produce the same looking numbers