|
|
cdc.png (257,118 bytes) |
|
|
For 4) I experimented a little further. It is a compromise between speed and map coverage. Currently I'm at something like: ( (searchmode='T') or ((field_diameter>4 degrees) and ((area290=1) or (area290=290))) or (sqr( delta_ra*cos_telescope_dec) + sqr(dec2-telescope_dec)<sqr(field_diameter * factorX) ) ) {check on areas 1 and 290 fixes some coverage problems with celestial poles in the corner} So only for the two pole areas 1 and 290. The main filtering is earlier done. Furthermore it works better for me if I calculate the distance by using the rule of Pythagoras. In my program I use a zoom factor rather field_diamater, so I can't test above condition. factorX should be little larger then one. Han |
|
|
Thank you Han, I apply the change and for the filter I use: ( (searchmode='T') or ((field_diameter>4*pi/180) and ((area290=1) or (area290=290))) or (sqr( delta_ra*cos_telescope_dec) + sqr(dec2-telescope_dec)<sqr(field_diameter * 1.5) ) ) This look good. https://github.com/pchev/skychart/commit/6cb1edfb6d55b00be7d663466ae57ac7b9353952 |
|
|
Good morning Patrick, I'm still not fully happy with fix of problem 4). The basis of the code was written more then 20 years ago. It seems no problem to calculate the distance of each star individually and more elegant. The plot time is not really influenced using the processing power of modern CPU's. I will experiment today a little further and come back on this . Han |
|
|
This is not a big problem for CdC to get some extra stars to reject because this is the first thing the program do after converting to screen coordinates, so there is no try to plot this stars. |
|
|
Okay I have implemented a more rigorous method which works well with the same processing speed. The rigorous method is implemented in a new procedure ang_sep_fast where cos, sin and arccos values are read from arrays rather then being calculated. High accuracy is not required. This gives the best performance. For short distances the angular seperation is calculated using a simpler method but using the star declination rather then screen center declination. Also the cos array "cos_value" is used for speed. Just check the the factors *1.5/2 in line 3233 and in 3238. If it is the diameter, the factor should be 1.5/2. If it is the radius the factor is 1.5. Too small factor will remove stars in the corners of the window. delta_ra:=abs(ra2-telescope_ra); if delta_ra>pi then delta_ra:=pi*2-delta_ra; if delta_ra>pi/3 then begin {far away in ra} ang_sep_fast(telescope_ra,telescope_dec,ra2,dec2, sep );{calculate seperation using rigorous method but with approximate cos and sin and cos values from tables} in_sight:=sep<(field_diameter * 1.5/2);{star in sight?} end else begin {nearby in ra} sep:=(sqr( delta_ra*cos_value[round(dec2*180/pi)] ) + sqr(dec2-telescope_dec));{calculate seperation using approximate method for shorter distances} in_sight:=sep<sqr(field_diameter * 1.5/2); {{star in sight? sqr(field_diameter) is faster then sqrt(seperation)} end; See attached updated u_290.pas. I haven't compiled it but it should work since I copied the parts from my code. |
|
|
Happy New Year Han! The new method work very well. Good idea to use table for speed, old trick are sometime still useful. The only change I do to make the compilation to work is to add the local variable in_sight:boolean to the function readdatabase290. I also remove a multiplier of 1.5 to the fov I have in CdC, now it work without that. https://github.com/pchev/skychart/commit/43a300f8a647a6dfe702ab8e890a87f76137120a |
- Anonymous
