I'm unsure how to go about what I need. I've looked at the library code and the documentation and am not even sure what I want is possible, though it should be. In most of our code we use the si system and have defined composite dimensions for most of what is unique to our application. However, there is one dimension where there is no standard unit in the si system: flow coefficient. The flow coefficient is Q/sqrt(dP), a volumetric flow/sqrt(pressure). We have derived and defined a unit that would be the si standard for this quantity if there was one and things work fine. I have a runtime unit system that then converts between user values that make sense, to our internal si version that does not but works. However, now I want to write some alternative functions that use the actual units that engineers really use: Cv and Kv. I would like these alternative functions to then convert into this si unit we've created but the interface should distinguish between Cv, Kv, and my made up unit. The first part is easy: make a system that only contains this funky dimension and then use it to make quantities that are not in the si system. The second part is what I can't find. The conversion templates in conversion.hpp seem to specifically expect base units which at least the si part is not and my conversions should use my funky si unit as their basis. So the question I have is how can I write conversions such that I can do this: quantitysi::flow_coefficient fc(42 * Kv); and this: void f(quantityKv::coefficient); void f(quantityCv::coefficient); void f(quantitysi::coefficient); Where Kv and Cv are base units in their own systems. The problem is sort of like degrees and radians, which I was going to use as a basis when I first was looking at this issue, except that in that case the unit is a base unit for a base dimension and in my case I'm working with a derived dimension within the si system. The only way I can think of to do it is to make a system with the components within the derived dimension and make conversion factors for those. The problem is that I am not sure such systems can be coherent and it would be a significant effort above and beyond the direct conversion factors I already know. -- http://crazyeddiecpp.blogspot.com/