Re:27 Jobs with Unbelievably Tough Interview Questions
by ghaff on Friday February 24, 2017 @04:32PM
(#13728111)
I believe the hoop stress vs. longitudinal stress question is approximately true specifically for thin-walled pressure vessels. However, it is many many years since I would have been able to prove that--probably about the time I took the final exam in that particular course :-)
Especially given modern computer tools, I'm actually not sure how many mechanical engineers would be able to derive that sort of thing in an interview. I wouldn't have been able to but you probably didn't want me designing rockets either.
Well, I have no idea whatsoever how to approach that...but I would not be applying to be a test engineer at a rocket company. This sounds like something that an actual rocket engineer should know, and I would guess that it is even something that would be easy for such an engineer.
> “What is the first thing you’d print with a 3D printer if you had one?”—Linux Systems Administrator I, Rackspace
I'd print whatever test object the printer's instructions say to print to verify that I have set the printer up correctly and that it is working.
> “Please describe an instance where you had to make a decision without all of the necessary information.”—Analytics, athenahealth
The obvious answer is "this interview".
> 27. “How would you find the square root of 1.2?” —Hardware Engineer, Jump Trading
I would guess that some context is missing here. I don't think it means that if you needed the square root of 1.2 for something you were doing, how would you obtain it. I expect it is asking how you would do it in hardware, or in firmware. It's then time to talk about what computational ability the hardware has. For example, if it is going to be firmware, is there a divide instruction on the processor? Do we have floating point or are we using fixed point? How accurate does it need to be? Yadda yadda.
If it is asking how you would obtain it for yourself, then the correct answer is use a calculator, unless you just need a quick approximation, in which case it is to note that 11^2 = 121, so 1.1^2 = 1.21, and so 1.1 is a pretty good approximation to sqrt(1.2). If you need slightly better, but still cannot use a calculator, note that 1.1 is slightly high, and assume the right answer is 1.1-x. That gives (1.1-x)^2 = 1.2, which can be rewritten as x = 0.01/(2.2-x). If x is small, we can neglect it in 2.2-x, suggesting we look at x = 0.01/2.2. Let's approximate 2.2 with 2 to make the mental math easier, giving us x = 0.005. That gives us a new approximation for sqrt(1.2) of 1.1 - 0.005 = 1.095. That is quite close. It's square is 1.199025.
If you want an answer that works well under both the "it's for firmware" interpretation and the "it's for me and I don't have a calculator" interpretation is to use the binomial theorem. For square roots, that says sqrt(1+x) = 1 + x/2 - x^2/8 + x^3/16 - 5x^4/128 + ...
With x = 0.2, that's 1 + 0.1 - 0.005 + 0.0005 - ... Note that if you stop at 3 terms you get the same approximation I got earlier, 1.095. Four terms gives 1.0955. That's a good place to stop because 5 terms gets harder to do mentally because of that stupid 5/128.