How good is good enough?

[Jeff_Birt]

I really enjoy seeing the fancy test equipment and calibration standards on the EEVBLOG. The tour of the mobile calibration lab was also very interesting. I don't think most folks have a clue what it take to measure anything (mechanical or electrical) and be confident about your measurement (and know how confident you can actually be.) This got me to thinking about how to teach 'young players', as Dave would say, about how to know how good is good enough.

I am a EE but have always worked on electromechanical things, I run the machine shop in the EE department at a university and do everything from machining, to controls, to programming. I run a side business that deals in small CNC equipment too which is relevant to this discussion as to "How good is good enough?"

One thing I see a lot with the students I work with on projects is they have no clue what sort of accuracy/repeatability/precision is required be it electrical or mechanical (I see this with new hobbyist machinist as well). For example, a student designing a brushless motor had trouble understanding that the laminations would all not be exactly the same thickness as there is slight variation across the sheet, they will also not stack exactly perfectly so if you thought you needed 50 laminations you may wind up with 49 or 51 to get the correct stack height. Or, a LiPo battery may not be 'exactly' 3.7V when fully charged, it might be 4.2V and that might just burn up your board.

I guess it comes down to a few different points:

1) What sort of accuracy/precision/repeatability do I really need?
2) How do I specify that?
3) How do I measure that?

What I see very little discussion of is #1, which is basically my topic, 'How good is good enough?'. At university I tend to only see this actively thought about on research projects; it's funny how when you are trying to actually do something physical that you find your perfect theory goes out the window.

As hobbyist we can go to great lengths to make things impractically perfect, as professionals it pays to know just how perfect it needs to be. So how do you as a hobbyist or professional answer that question? How good is good enough? What are your guidelines? How do we teach this to 'young players'?

[MosherIV]

So how do you as a hobbyist or professional answer that question? How good is good enough?

As a professional, it should be defined in a specification.
Who and how the specification is written......well in my professional experience this is either guided/set by the customer or some expert will know the market/product and set the requirements.

How do we teach this to 'young players'?

Not everyone learns in the same way, some are brilliant and will pick it up through teaching or reading books but most need to learn from doing or an example.
So I think in your example of the student having to figure out the motor lamination, the student would have learned there is a difference between the pure theoretical and the practical. I guess I am suggesting more practicals for students.
We have the new batch of interns starting right now and yeah, boy are they hard work

[rstofer]

It sounds a lot like: Design with a micrometer, measure with a yardstick, cut with an axe then grind to fit and paint to match.

The very topic of "good enough" is the underlying question of "Six Sigma" design:

http://www2.warwick.ac.uk/fac/sci/wmg/ftmsc/modules/modulelist/peuss/slides/dfss_lecture_slides_compatibility_mode.pdf

This was a HUGE program at General Electric, everybody, and I mean every single employee, had to take the training and work out some projects.  All of which involved driving costs down.  It is amazing how many processes are institutionalized and can be completely eliminated.  A process that disappears, output being equal, is the ultimate goal.

Every STEM student takes Statistics (... pretty sure).  Few want to use it when the class is over but it is fundamental to designing things that are manufacturable.  Had the students prepared a histogram of the thicknesses of laminations, they would have seen right away how many they might need and how the thicknesses would impact assembly.

[Jeff_Birt]

The very topic of "good enough" is the underlying question of "Six Sigma" design:

When I first started working at the university I was in the Engineering Management Department which also did some manufacturing oriented classes. There are Six Sigma graduate level classes but it is nothing one would learn in a standard engineering tract.   

Who and how the specification is written......well in my professional experience this is either guided/set by the customer or some expert will know the market/product and set the requirements.

It has been my experience that not even the company designing and manufacturing a product will know how to specify things. The last job I had in industry was for an electronics manufacturer. They wanted me to test a new lower cost back up battery so I asked, "How should I test it, what specifications should it meet?" What is the current draw profile we should test against, i.e. do we expect a typical equipment installation will draw x.xxxA with brief spike of y.yyyA? Product engineering had never though of or developed a spec like this and the marketing guy did not care as long as it was cheaper and he could see more of them.

[woodchips]

Interesting questions, but what about resolution to add to the three?

Book that changed my ideas was 'Right First Time' by Frank Price.

This can also apply to working practice, such as 'The Mythical Man Month' by Brookes, 'Peopleware' by De Marco and Lister. If you can't convince those that need convincing that there is a problem, or something needs investigating, then you will get nowhere.

I am clearing out a lifetime's collection of stuff, and in taking it to pieces it amazes me just how poorly power supplies in particular are designed. Not the efficiency, noise etc but wrong size cables, expecting a screw through two pieces of copper to keep tight sort of thing. Then there is cooling, put a fan in it and it will be cooled. Unfortunately not.

[bitseeker]

How good is good enough? What are your guidelines? How do we teach this to 'young players'?

Very good topic, Jeff. Much of the answer comes from experience, whether that's first-hand, handed down, or defined in a specification by someone with experience.

Getting students to start thinking and seeing these issues can begin with what rstofer said:

Had the students prepared a histogram of the thicknesses of laminations, they would have seen right away how many they might need and how the thicknesses would impact assembly.

In other words, have them go out there and gather data on all kinds of things. Then, they'll begin to see what the real world is really like.

[b_force]

To make it even more interesting, a lot of specs (even from top brands) are really sucked out of a big thumb if you really have to rely on it.
That's why real research centers get them calibrated every now and again.

Most people forget the absolute error + extra amount of digits in the result of a meter.
Practically it means that in a lot of cases the last one or even two digits are not very accurate (at all).

Another reason why it's sometimes handy to have just a little bit more

[David Hess]

I am clearing out a lifetime's collection of stuff, and in taking it to pieces it amazes me just how poorly power supplies in particular are designed. Not the efficiency, noise etc but wrong size cables, expecting a screw through two pieces of copper to keep tight sort of thing. Then there is cooling, put a fan in it and it will be cooled. Unfortunately not.

I am always amazed at how well some old integrated power supplies are designed.  They include things like fold-back current limiting, SCR crowbar protection, shutdown with hard start capability, output tracking, reverse output shunts, etc.

[German_EE]

How good is good enough?

If you are repairing a tube receiver it might specify a power voltage of 90V, anything from 85V to 100V will probably be fine but watch the voltage rating on the components.

If you measure the temperature of a room then you're not too bothered about the accuracy of your thermometer. I can see two right now and one reads 32C whilst the other reads 35C. Measuring people however needs greater accuracy, average core temperature is 37C, you're ill if you're 38C and by 40C you might die.

I normally work to 1mm using a ruler or 0.5mm if I want a close fit. A friend of mine works with millimeter microwaves and the cavities he uses are as small as 1.3mm +/- 0.002mm.

In other words, the answer depends on the job that you are doing. I am happy using a cheap DVM to measure the mains supply but for general electronics use I have a 5-digit DVM that is calibrated once a year. Frequency? I'm locked to DCF77 which means an accuracy well below 1Hz but if I'm measuring a 500 MHz signal I don't expect to get it down to the nearest millihertz.