Hi, On vr, 13 sep 2019 10:38:23 +0200, Romain Forlot [IoT.bzh] wrote: > Hi everyone, > > I have a question about signed numbers in CAN, but I try to handle sign of a > signal's value whenever it is a signed value. > > I would like to have something that works in any cases obviously but for > what I found, signed value in CAN bus isn't standardized. > > You can have: > > * signal on N-1 bits, with the subtracted bit representing the sign > * either with 1's complement > * or 2's complement. > > You could have some bitfield not encoded within classics bytes but in 3 or > 13 bits by example, so it could complicate the computing of the signed value > depending on the chosen method. A common method is to define a range, like -1000 to 1000 to be encoded into 0..2000 unsigned. In this example, you require 11bits. The advantages of doing so are: 1) it's not so difficult to convert to/from bitfields, i.e. as you found, properly decoding a negative number from an 11bit value is a bit obscure. 2) it leaves explicit room for special/fault codes, above the defined range. Kind regards, Kurt
