Significant Digits

[wasasparky]

I am an EE and former master electrician.
I am going to take my test to become a master once again.
Have any jurisdictions (or schools? IBEW? etc,) ever taken a look at the accuracy (precision) of their answers?

Two things have made me start this thread:
1. A "recent" thread here about the size of capacitor conductors for motors. (and the precision of the answers)
2. Going through the study material available from the state (MN), I am alarmed at the precision of many of their answers.

For Example: (going from memory, not what is exactly in the guide)
75kVA
208V
Z=1.2%
Available Fault Current answer was 17,348 amps. (Book I think was 17,333)

The 75 kVA I believe would be seen as "nominal"?, so would not necessarily limit the precision of the answer? (been a while since I reviewed significant digit rules)
For sure the 1.2 limits us to TWO significant digits, so the answer should be 17,000 amps...

 

[charlie b]

1. You are right.

2. If you want to pass the test, you need to forget the fact that you are right.

 

[drbond24]

Significant digits? C'mon man, nobody took that any further than college and we only did it then because we HAD to.

I agree with Charlie's assessment.

You're right, but I don't care. My calculator spits out answers with 9 decimal places, and I round that to a number that suits me at the time. This may have to do with my mood, the answer I'm looking for, the time of day, or the amount of lead in my pencil, but it will NEVER EVER have anything to do with the significant digit rules.

Edit to add: The professional engineering exam (which I passed!) was multiple choice. I don't know anything about the test you're going to take, but if it is multiple choice you just need to figure out which circle to darken. The accuracy of your calculations only needs to be good enough to be able to tell which of the choices your answer is closest to.

 

[jrannis]

You will most likey see multiple choice answers.
I wouldn't sweat it too much.

 

[wasasparky]

You will most likey see multiple choice answers.
I wouldn't sweat it too much.

I'm not worried about it at all.
I just wonder if we should pay a little more attention to precision.:-?
Let's show a little restraint.
Come on guys - we have to keep the bar higher than the other trades...:smile:

I don't want it to be to the point where someone gets an answer wrong on a test, but lets at least be reasonably close...

17,333A is just way off the target
17,000A is correct in regard to significant digits, but I wouldn't mind one more
17,300A is reasonable enough...

 

[Cold Fusion]

--- Have any jurisdictions (or schools? IBEW? etc,) ever taken a look at the accuracy (precision) of their answers?
I have not seen any evidence they have - Nor much evidence that the engineering side is paying much attention either

--- I am alarmed at the precision of many of their answers.
Me too. Looks sloppy to me

--- For sure the 1.2 limits us to TWO significant digits, so the answer should be 17,000 amps...
Maybe even a wider range than that.

Consider the normal spec on transformer impedance is +- 7.5%. So the 1.2%Z is anywhere in the range from 1.11% to 1.29%

FLA for 75kva at 208V is 208.179183602A

So available SCC (infinite primary) is from 18,754.8814056A to 16,137.9212095A

The answer really is 17,446.4013076A +- 1,308.48009800:roll::roll:

But, test makers, electricians, engineers have got 12 digit caculators and 208v is exactly 208.000000000V and A 75kva xfm is is exactly ....

dr - I don't agree with you. Any time I see an answer carried out to more significant figures than I know the data is good for - that tells me they either don't understand, or they are too sloppy to pay attention.

But that is an engineering opinion, not a test taking opinion.

cf

 

[Cold Fusion]

I just wonder if we should pay a little more attention to precision.
Let's show a little restraint.
Come on guys - we have to keep the bar higher than the other trades..

I agree. I think we should.

The worst I have ever seen was in the late 70s, working for NRC. I was an engineer (just above junior grade) working with the crew doing calorimetric tests on Naval reactors. We had been hand plotting the data on 8.5x11 graph paper and it is suposed to be a straight line - but the instruments are not perfect so the data is scattered. Then the engineering management would take a straight edge and fight over how they would put the straight line (by eyeball) to fit the data - they could be hours moving the ruler around a few fractions of an inch.

Then we got an HP programmable calculator that would connect to a plotter. We programmed the calculator to do a least squares fit to the data and draw the line.

So, the data gets punched in, hit the calculate button, the calculator blinks for a bit, then the plotter starts moving - little 'x's where the data points are, and then draws a straight line up through the middle. The linear equation coeficients were printed out at the bottom of the graph - to about 9 significant figures. The upper end engineers (including the NRC reps) would nod their collective head, and agree on how accurate the "computer" was.

Nobody ever asked about the programming algorithm. Nobody ever asked about the instruments accuracy.

I checked on the instrument accuracy and did error calculations. I knew the programming was okay - I was part of the crew that did it.:roll:

Turns out that hand plotting the data and eyeballing the ruler in the middle of the data, was just as good an answer as the computer plotted one.

But that didn't worry me as much as the willingnes of the engineering managers to accept the "computer" answer with out any question.

cf

 

[drbond24]

dr - I don't agree with you. Any time I see an answer carried out to more significant figures than I know the data is good for - that tells me they either don't understand, or they are too sloppy to pay attention.

There is a 3rd perfectly good explanation for it which I already mentioned in my previous post. I understand the subject matter quite well and I am not sloppy at work (had to add that clarifier....my garage is very sloppy at the moment :wink: ). When it comes to significant digits, I don't care. 17,000 is approximately equal to 17,300 which is approximately equal to 17,446.4013076 A. They are all the same number to me. They all tell me exactly the same thing; the answer to my question is about 17k. That works on tests and in real world engineering. If you are cutting something so close that a 2% swing in the current will throw your design into disarray, you were cutting it too close.

Besides, I am far more likely to have too few significant digits than too many. As I said above, 17k is it for me. Unless the choices were a) 17,301 b)17,302 c)17,303 and d)17,304 I don't need to get any more exact than that.

 

[Cold Fusion]

--- I understand the subject matter quite well and I am not sloppy at work ---

I was (am) pretty sure you did and pretty sure you weren't.

--- When it comes to significant digits, I don't care. 17,000 is approximately equal to 17,300 which is approximately equal to 17,446.4013076 A. They are all the same number to me. ---

First, for this case, your're right. They all tell you the same thing.

The parts I am not agreeing with are:

--- Significant digits? C'mon man, nobody took that any further than college and we only did it then because we HAD to. ---

---You're right, but I don't care. :grin::wink: My calculator spits out answers with 9 decimal places, and I round that to a number that suits me at the time. This may have to do with my mood, the answer I'm looking for, the time of day, or the amount of lead in my pencil, but it will NEVER EVER have anything to do with the significant digit rules. ---

That is not true for most of the work I do. I'm not taking tests, and I do as little design work as I can. So mostly, while I am building it, commissioning it, or fixing it after it doesn't work, I am checking the engineering to see if there are issues. And 4 digit answers involving 2 digit data tells me to look closer. Or design decisions based on 4 digit answers based on 2 digit data, doesn't sound real great either.

--- The professional engineering exam (which I passed!) was multiple choice. I don't know anything about the test you're going to take, but if it is multiple choice you just need to figure out which circle to darken. The accuracy of your calculations only needs to be good enough to be able to tell which of the choices your answer is closest to. ---

I didn't know the PE exam changed to all multiple choice. In 1990, it was half mc and half state assumptions, show your work.

That is indifferent. The important part is you passed your PE test and got your ticket. That's an accomplishment:smile:

cf

 

[charlie b]

Then we got an HP programmable calculator that would connect to a plotter.

That part of the story sent chills down my spine! It brought back an old memory. Were you and I on the same ship at the same time, and was I your boss (i.e., the Reactor Control Assistant)? My crew did a computerized analysis in just the same manner. We almost failed the inspection because of this "great idea".

It turned out that the HP would stop and display values from time to time, as it was programmed to do. The techs would dutifully write those numbers on the standardized form, and that constituted our test report. The problem is that the HP would remember more significant digits than it printed out. So a person performing an audit by only looking at the hand written report (without benefit of using the HP calculator's printout), and who manually entered the printed data into his own calculator, would not get the exact same answers from one line to the next line.

Fortunately (for my career), the error introduced by the difference between the HP's memory and the numbers we wrote down from looking at its screen turned out to be "conservative." If you have to be wrong, it is better to be wrong in one direction than wrong in the other direction.

 

[Besoeker]

I am an EE and former master electrician.
I am going to take my test to become a master once again.
Have any jurisdictions (or schools? IBEW? etc,) ever taken a look at the accuracy (precision) of their answers?

Two things have made me start this thread:
1. A "recent" thread here about the size of capacitor conductors for motors. (and the precision of the answers)
2. Going through the study material available from the state (MN), I am alarmed at the precision of many of their answers.

For Example: (going from memory, not what is exactly in the guide)
75kVA
208V
Z=1.2%
Available Fault Current answer was 17,348 amps. (Book I think was 17,333)

The 75 kVA I believe would be seen as "nominal"?, so would not necessarily limit the precision of the answer? (been a while since I reviewed significant digit rules)
For sure the 1.2 limits us to TWO significant digits, so the answer should be 17,000 amps...

I'd have said 17.5 kA in round figures (with an improbable Z = 1.2%) but I take your point.
Move the 1.2% less than 0.05% either way and the spread is approximately 1.4506665 kA.
:wink:

 

[drbond24]

That is not true for most of the work I do. I'm not taking tests, and I do as little design work as I can. So mostly, while I am building it, commissioning it, or fixing it after it doesn't work, I am checking the engineering to see if there are issues. And 4 digit answers involving 2 digit data tells me to look closer. Or design decisions based on 4 digit answers based on 2 digit data, doesn't sound real great either.

Oooohhh, ok. I see where you're coming from now. That makes sense.

The PE changed to multiple choice just a handful of years ago. I think they got tired of all the challeges and subjectivity involved with hand written answers. It could be argued that multiple choice is easier, but at least there is no question about whether or not your answered correctly.

 

[Pierre C Belarge]

Have any jurisdictions (or schools? IBEW? etc,) ever taken a look at the accuracy (precision) of their answers?

The tests are generally written to the code book and not field practices.
Take a look at 220.5 Calculations.

 

[wasasparky]

The tests are generally written to the code book and not field practices.
Take a look at 220.5 Calculations.

That just deals with fractions of an amp.
Answers with 5 digits, as a result of data with two digits is my concern.
(A level of precision is represented that far exceeds the precision of the input...)

Now we all know how they got from 17,333.3333333333333 to 17,333:smile: