it's normal. it'll only be changing from the data u put in. it wont change again after it saves.
it's just the way the data recalculates when it get's saved back in.
That is discretization error, aka quantization error...
the binary representation of a real number (i.e. a number with a fractional part) allows only specific discrete values;
for example, straight binary fixed point allows the following values (each bit position represents the next power of 2 down, i.e. half the previous bit):
0.1b = 0.5
0.01b = 0.25
0.001b = 0.125
so you would represent 0.75 exactly as 0.110b, and 0.375 exactly as 0.011b, and 0.625 exactly as 0.101b;
but in 3 bits, 0.70 and 0.80 cannot be exactly represented, the closest representation for either of these would be 0.75 i.e. 0.110b.
So when you reopen the tunetool this is what you're seeing (looks like it's fixed point and not floating point because the same bit positions change).
The sub-millisecond values are out by about 2%, the millisecond values are out by 0.2% or less (as the millisecond value gets larger).
You won't notice any difference in 2% error on the sub-millisecond time adder.
Last edited by joecar; April 10th, 2017 at 04:41 AM.
Gotcha! That makes sense. Just wanted to make sure we weren't going to see problems there. Thanks for the quick reply!