Engineers, Quality Control, and thinking people - a dilemma

[oznerol]

Simple scenario:

Customer gives specs for a product.
One particular spec is for a voltage.
Spec states "the voltage's acceptable values are -1.0V, -1.1V, -1.2V"

We sent the product to QC and got back a reading of -0.97

Does it pass or fail?

Head of engineering says pass. Head of quality says fail.

Head engineer states the significant digit argument. The spec said -1.0, so in that case -0.97 = -1.0, so it is within spec and passes.
Head of quality says no, we measure to 10x precision of the spec, if it falls below that then it is a fail.

Product needs to ship soon or get delayed, so it's very heated.

What do you guys think?

 

[FoBoT]

shouldn't the original specs have included tolerances?

i blame the project manager for getting vague requirements, it is always the project managers fault

 

[TheNinja]

It obviously fails. The specs say the voltage must be -1.0v or lower. Clearly -0.97 is not "lower" than -1.0V. Although I can see the engineer's side in which 0.97 is close enough to 1.0 that it will work fine. According to the specs, it technically fails.

 

[NatePo717]

What are the tolerances on your Voltages? If it allows for +/- 0.03V your golden.

 

[MotionMan]

Do the calculations on a Pentium in Excel and it should pass with no problem!



MotionMan

 

[torpid]

I voted fail, but the real solution is to ask the customer.

 

[Indolent]

Technically, it fails. But the logical response to me would be to ask the customer?

 

[TBone48]

I vote fail. If your QC dept. has to certify that the product falls within a stated specification they have to fail it if it doesn't. That doesn't mean they won't be overruled, but the certification would have to detail the decision. My department will approve product that doesn't meet specification if we get approval from the customer beforehand. Then we document it on our certification.

 

[Train]

Well they didnt say 1V they said 1.0V, a factor of 10 times more accuracy is required. 3% off is usually the fine line between Good and Acceptable in most quality assurance departments, but that totally depends on the industry and type of product.

But in this case they also specified "and lower" sorta giving you your room to work with on one end, not the other. Should have targetted for -1.15 or so if you knew it wouldn't hit exactly on. The real mistake here was aiming for 1V (assuming thats what you guys did) when its clear you should ahve aimed lower.

Fail.

Howerver in the "real world" if its really gonna hurt, and the client wont notice, ship it now, and hurry up and fix it in the mean time without telling them. Hopefully by the time they notice (if they ever do) you can be like, "umm no, these last 10 samples are all within spec.

I've pulled this one off in the software world a dozen times. Ship with known bug (known to us, no one else) to meet a deadline, then fixed it before any customers found it, or if one did find it, blame it on some bizarre pc configuration then silently install the patch.

 

[hypn0tik]

Significant figures!

They required a spec -1.0V. In this instance, they only need accuracy to within a tenth of a volt. If they needed a higher accuracy, they would have specified 1.00V or 1.000V. Clearly they didn't. In this case, -1.0 means that the voltage could be anywhere from -0.95V to -1.04V. Since -0.97V falls within the acceptable range, it's good to go.

 

[Train]

Originally posted by: hypn0tik
Significant figures!

They required a spec -1.0V. In this instance, they only need accuracy to within a tenth of a volt. If they needed a higher accuracy, they would have specified 1.00V or 1.000V. Clearly they didn't. In this case, -1.0 means that the voltage could be anywhere from -0.95V to -1.04V. Since -0.97V falls within the acceptable range, it's good to go.

If they had only specified -1.0V, I would agree with you. Howerver They said "And Lower" pretty much making it one sided. The significant figure is meaningless since he also said -1.3V was acceptable.

 

[torpid]

Significant digit rounding may not be applicable. Depending on the device, a voltage of -0.97 may not be acceptable / usable even though it is rounded to -1.0.

 

[oznerol]

Alright, so both sides have been stated so I can explain the reasoning.

Head engineer states the significant digit argument. The spec said -1.0, so in that case -0.97 = -1.0, so it is within spec and passes.
Head of quality says no, we measure to 10x precision of the spec, if it falls below that then it is a fail.

Both have a valid point.

The "and lower" part is not exactly correct I am finding. I believe it is -1.0, -1.1, -1.2. I have not seen the exact spec since I am not working on the project. It is quite lively here though, as neither side wants to budge and someone needs to make a decision.

 

[flxnimprtmscl]

Originally posted by: hypn0tik
Significant figures!

They required a spec -1.0V. In this instance, they only need accuracy to within a tenth of a volt. If they needed a higher accuracy, they would have specified 1.00V or 1.000V. Clearly they didn't. In this case, -1.0 means that the voltage could be anywhere from -0.95V to -1.04V. Since -0.97V falls within the acceptable range, it's good to go.

Good point. I say it passes.

 

[Train]

Originally posted by: ducci
Alright, so both sides have been stated so I can explain the reasoning.

Head engineer states the significant digit argument. The spec said -1.0, so in that case -0.97 = -1.0, so it is within spec and passes.
Head of quality says no, we measure to 10x precision of the spec, if it falls below that then it is a fail.

Both have a valid point.

The "and lower" part is not exactly correct I am finding. I believe it is -1.0, -1.1, -1.2. I have not seen the exact spec since I am not working on the project. It is quite lively here though, as neither side wants to budge and someone needs to make a decision.

Theres only one way to solve this then. Engineers vs QA in the octagon!

w00t! you better have the vid on youtube by tomorow.

 

[hypn0tik]

Originally posted by: Train

Originally posted by: hypn0tik
Significant figures!

They required a spec -1.0V. In this instance, they only need accuracy to within a tenth of a volt. If they needed a higher accuracy, they would have specified 1.00V or 1.000V. Clearly they didn't. In this case, -1.0 means that the voltage could be anywhere from -0.95V to -1.04V. Since -0.97V falls within the acceptable range, it's good to go.

If they had only specified -1.0V, I would agree with you. Howerver They said "And Lower" pretty much making it one sided. The significant figure is meaningless since he also said -1.3V was acceptable.

The "and lower" has no bearing on the fact that they specified to an accuracy of 1/10th of a volt.

It's really an elementary scientific principle. You can't specify your result more accurately than what you start out with. In this case, -0.97 is more accurate than -1.0. You should express your result with the same accuracy. In this case, -0.97 -> -1.0, which is indeed the required specification.


On the other hand, it is extremely poor engineering practice to design for something right at the boundary. Their engineers should have designed for something LESS than -1.0V (-1.2 for example) so that they'd have a tolerance of 20%. Regardless, this is not the issue we are looking at.

 

[hypn0tik]

Originally posted by: ducci
Alright, so both sides have been stated so I can explain the reasoning.

Head engineer states the significant digit argument. The spec said -1.0, so in that case -0.97 = -1.0, so it is within spec and passes.
Head of quality says no, we measure to 10x precision of the spec, if it falls below that then it is a fail.

Both have a valid point.

The "and lower" part is not exactly correct I am finding. I believe it is -1.0, -1.1, -1.2. I have not seen the exact spec since I am not working on the project. It is quite lively here though, as neither side wants to budge and someone needs to make a decision.

It is company policy to measure 10x the precision of the spec?

 

[Specop 007]

Originally posted by: hypn0tik

Originally posted by: Train

Originally posted by: hypn0tik
Significant figures!

They required a spec -1.0V. In this instance, they only need accuracy to within a tenth of a volt. If they needed a higher accuracy, they would have specified 1.00V or 1.000V. Clearly they didn't. In this case, -1.0 means that the voltage could be anywhere from -0.95V to -1.04V. Since -0.97V falls within the acceptable range, it's good to go.

If they had only specified -1.0V, I would agree with you. Howerver They said "And Lower" pretty much making it one sided. The significant figure is meaningless since he also said -1.3V was acceptable.

The "and lower" has no bearing on the fact that they specified to an accuracy of 1/10th of a volt.

It's really an elementary scientific principle. You can't specify your result more accurately than what you start out with. In this case, -0.97 is more accurate than -1.0. You should express your result with the same accuracy. In this case, -0.97 -> -1.0, which is indeed the required specification.


On the other hand, it is extremely poor engineering practice to design for something right at the boundary. Their engineers should have designed for something LESS than -1.0V (-1.2 for example) so that they'd have a tolerance of 20%. Regardless, this is not the issue we are looking at.

What company do you work for.
I want to make sure I never do business with a company with shady employees selling out of spec stuff.

 

[torpid]

Originally posted by: hypn0tik
The "and lower" has no bearing on the fact that they specified to an accuracy of 1/10th of a volt.

It's really an elementary scientific principle. You can't specify your result more accurately than what you start out with. In this case, -0.97 is more accurate than -1.0. You should express your result with the same accuracy. In this case, -0.97 -> -1.0, which is indeed the required specification.

This is based on the possibly flawed assumption that the client specified -1.0 because it only cared about two significant digits and that there is an implicit understanding between the client and the OP's company that -1.0 does not imply a minimum limit / threshold beyond which the product will not be acceptable. That is an inference that may not be accurate in this case. The only way to gain clarity is to consult with the client (or for the OP to elaborate).

 

[torpid]

Originally posted by: hypn0tik

Originally posted by: ducci
Alright, so both sides have been stated so I can explain the reasoning.

Head engineer states the significant digit argument. The spec said -1.0, so in that case -0.97 = -1.0, so it is within spec and passes.
Head of quality says no, we measure to 10x precision of the spec, if it falls below that then it is a fail.

Both have a valid point.

The "and lower" part is not exactly correct I am finding. I believe it is -1.0, -1.1, -1.2. I have not seen the exact spec since I am not working on the project. It is quite lively here though, as neither side wants to budge and someone needs to make a decision.

It is company policy to measure 10x the precision of the spec?

That is a standard QC practice.

 

[hypn0tik]

Originally posted by: Specop 007

Originally posted by: hypn0tik

Originally posted by: Train

Originally posted by: hypn0tik
Significant figures!

They required a spec -1.0V. In this instance, they only need accuracy to within a tenth of a volt. If they needed a higher accuracy, they would have specified 1.00V or 1.000V. Clearly they didn't. In this case, -1.0 means that the voltage could be anywhere from -0.95V to -1.04V. Since -0.97V falls within the acceptable range, it's good to go.

If they had only specified -1.0V, I would agree with you. Howerver They said "And Lower" pretty much making it one sided. The significant figure is meaningless since he also said -1.3V was acceptable.

The "and lower" has no bearing on the fact that they specified to an accuracy of 1/10th of a volt.

It's really an elementary scientific principle. You can't specify your result more accurately than what you start out with. In this case, -0.97 is more accurate than -1.0. You should express your result with the same accuracy. In this case, -0.97 -> -1.0, which is indeed the required specification.


On the other hand, it is extremely poor engineering practice to design for something right at the boundary. Their engineers should have designed for something LESS than -1.0V (-1.2 for example) so that they'd have a tolerance of 20%. Regardless, this is not the issue we are looking at.

What company do you work for.
I want to make sure I never do business with a company with shady employees selling out of spec stuff.

Mind if I ask what you do for a living?

You're obviously not an Engineer based on your comments, so you probably believe that all the numbers you see on products you buy don't have any tolerances associated with them.

Fortunately for you though, I'm in grad. school (just started my M.A.Sc in EE). You don't have to worry about doing business that employs me for a couple years.

 

[hypn0tik]

Originally posted by: torpid

Originally posted by: hypn0tik

Originally posted by: ducci
Alright, so both sides have been stated so I can explain the reasoning.

Head engineer states the significant digit argument. The spec said -1.0, so in that case -0.97 = -1.0, so it is within spec and passes.
Head of quality says no, we measure to 10x precision of the spec, if it falls below that then it is a fail.

Both have a valid point.

The "and lower" part is not exactly correct I am finding. I believe it is -1.0, -1.1, -1.2. I have not seen the exact spec since I am not working on the project. It is quite lively here though, as neither side wants to budge and someone needs to make a decision.

It is company policy to measure 10x the precision of the spec?

That is a standard QC practice.

But is it company policy?

Standard engineering practice is to express your results with the same precision as the problem statement. Pretty much any person in science will tell you that.

 

[torpid]

Originally posted by: hypn0tik
Mind if I ask what you do for a living?

You're obviously not an Engineer based on your comments, so you probably believe that all the numbers you see on products you buy don't have any tolerances associated with them.

Fortunately for you though, I'm in grad. school (just started my M.A.Sc in EE). You don't have to worry about doing business that employs me for a couple years.

Where I work, and every place I've worked in the past, it would be frowned upon pretty heavily (or worse) if you made a guess about the client's needs when a simple, 10 second phone call could provide a clear answer.

 

[hypn0tik]

Originally posted by: torpid

Originally posted by: hypn0tik
Mind if I ask what you do for a living?

You're obviously not an Engineer based on your comments, so you probably believe that all the numbers you see on products you buy don't have any tolerances associated with them.

Fortunately for you though, I'm in grad. school (just started my M.A.Sc in EE). You don't have to worry about doing business that employs me for a couple years.

Where I work, and every place I've worked in the past, it would be frowned upon pretty heavily (or worse) if you made a guess about the client's needs when a simple, 10 second phone call could provide a clear answer.

Agreed.

However, I was under the assumption that they didn't call the customer (and didn't plan on it) to get an answer from them. If they did call the customer, there would be no reason to debate, lol.

 

[pnho]

Based on the customer spec it will pass due to significant figures and would be ok to ship out to them. Although it fails based on your internal spec, the main issue would be cost consideration because if the product spec is very critical then the customer should have requested a tighter spec value. So if its justifiable to send it out, the cost issue is revelant, and your company's reputation won't be affected...then send it out.