In the book Modern Vibrations Primer. Peter Moretti says “In general, problems should be solved in whatever units the input data are given; converting units back and forth unnecessarily is not acceptable practice.”
However, it is often neater and involves fewer calculations if problems are converted to SI units, solved, and the answers converted back if necessary. I say this because often the various constants of nature are not readily available in alternate unit systems, and because many equations have constants that are 1 in SI units but other values one must dig up in other unit systems. Moreover, many constants are readily available in older and unfavored unit systems only in older references that use obsoleted or less accurate values.
So, do I misunderstand his recommendation? Or is his recommendation a poor one? Or is there some precedent for preferring
I’d say that in general you should minimize the necessary calculations because each step potentially introduces error. If you have all the data necessary in the original system of units, the additional steps required to convert to SI and then back are just errors waiting to happen. If you lack a few constants in the original units, it might be safer to convert the SI values you have to the other units rather than converting the problem data to SI. If you have any reason to doubt the value of your data, whether it’s an old reference or some other reason, then find another source. In the end, you have to use your own judgement about the most efficient and accurate method.
So his recommendation is a good one, but his use of “in general” and “unnecessarily” is intended to give you the latitude to use your own judgement.
Of course, the best practice is for the input data to be in SI to begin with. If you’re measuring lengths, use your metric tape measure. If you’re measuring mass, put kilograms on the other pan of your balance, and so on.
For working in the system of the input data, you also have the complication that some constants are defined differently in different unit systems, such that one simply can’t convert the constants directly. This is especially common in electromagnetism… What does 1 equal in N*m[sup]2[/sup]C[sup]-2[/sup]? It doesn’t, since the former is dimensionless and the latter is not. And yet, in Gaussian units, Coulomb’s constant is 1, while in SI units, Coulomb’s constant has units of Nm[sup]2[/sup]*C[sup]-2[/sup]. This is possible because what’s referred to as charge in Gaussian units and SI units actually refer to two different things.
Speeking as a person who has not done technical training in English units and used them very little IRL, I would still suggest using SI units, because if you do all your equation work using the units, the accordance of the units is often an excellent way to make sure your algbra is all right.
Plus, imho, the risk of unit conversion a the beginning & end is more manageable than the need for additional constants in the equations that the user may or may not remember, be familiar with, or have available.
Well, I’m not a scientist, but that sure sounds stupid to me. Where are you going to get something like the ideal gas constant in pounds and cubic feet? You could convert that into your imperial units, but I’d suspect you’d introduce as much or more error there. Since conversions are readily available with far more digits than you’ll need, you can avoid conversion error without difficulty.
When I was in undergrad, all the basic thermodynamic texts came in two versions, one with English units and one with SI. They each had steam tables, constants, etc. in the appropriate units. There are plenty of references out there for any units you want, and if you’re used to working in those units you probably have those references handy instead of having to convert from an SI reference. Back when I was a thermo geek, we had a (homemade) reference listing all appropriate data in units like yule log and fortnight, but that’s another story.
BTW, my English-unit copy of Van Wylen and Sonntag lists the universal gas constant as 1545 ft-lbf/lb mol R. If I had English-unit data, I’d certainly use that instead of converting to SI.
I was curious if my experiences with English units in Naval nuclear power were unusual or the norm, and, after the poll, it appears that in thermodynamics it is quite standard to do everything in BTUs and the like – as micco said, all of the steam tables and such are written in those units.
I was always taught to do any kind of nasty physics or chemistry math in abstract and only plug values in at the end. In that case, it’s pretty easy to just put all the units down, include your conversion factors for known constants, and then cancel out units to get the answer. Then, clearly, your answer would be in the units given.