Page 3 of 3
Re: About value inaccuracy
Posted: Fri May 13, 2016 8:08 pm
by KG_is_back
Tulamide, I have one thing to mention. When you save preset with FS preset manager, float values are actually stored as 8digit decimals. This effectively means floats will get quantized to 0.0000001 steps (rounding anything lower than that to zero). It popped up when I tried saving filter coefficients (which are very sensitive to rounding)... It is probably the most serious issue with inaccuracy in flowstone.
Re: About value inaccuracy
Posted: Fri May 13, 2016 9:02 pm
by tulamide
KG_is_back wrote:Tulamide, I have one thing to mention. When you save preset with FS preset manager, float values are actually stored as 8digit decimals. This effectively means floats will get quantized to 0.0000001 steps (rounding anything lower than that to zero). It popped up when I tried saving filter coefficients (which are very sensitive to rounding)... It is probably the most serious issue with inaccuracy in flowstone.
KG many thanks for mentioning that! We discovered such a fine difference as well, as the one I'm working for did the hard works of going through each seperate value and discovered that some of them work correctly when 0.0000001 added (within green). I then made the Ruby test and came to the result that I could simply add 0.0000001 to ALL values to get them to work.
EDIT: I'm talking of 8-bit-integers converted to floats in the range 0-1
Since I thought of a coincidence I didn't want to rely on it and decided to go with your proposals, but if this is a fix behaviour of the preset manager I might save all the time-wasting work of re-coding and just add 0.0000001!