Calculating Tolerance with One Sided Limits - Need a Form to Calculate

[xochi]

My company driven GRR form (excel) requires upper and lower limits for determining the tolerance.

I am currently recieving critical parameters that have only an upper limit.

There does not seem to be a valid internal definition as to how to calculate the tolerance when this happens.

You fine folk seem to have more than a handlfull of good answers, hence my registration today.

Any inputs?

 

[CarolX]

My company driven GRR form (excel) requires upper and lower limits for determining the tolerance.

I am currently recieving critical parameters that have only an upper limit.

There does not seem to be a valid internal definition as to how to calculate the tolerance when this happens.

You fine folk seem to have more than a handlfull of good answers, hence my registration today.

Any inputs?

Can you "redefine" nominal? In other words, if your dimension is 1.00 +.010/-.000, can you redefine to 1.005 +-.005? I do this with my data collection system. And it is much easier on the operators, who are accustomed (sp?) to bi-lateral tolerances.

Hope this helps.

 

[xochi]

In the most recent case (complanarity of a intergrated circuits leads), there is an upper limit of 8 mils. Anything less than that is acceptable, hence there is no lower limit.

Recent suggestion:

Tolerance Width = (abs(Upper limit - Parameter Mean)) X 2

The issue I have is that the Parameter Mean will be different on each of the same gauge. I fear this may be less than acceptable for some auditors.

I am looking for an "auditor friendly" definition.

Please be patient, I am new to this...

(Thanks for the warp speed response CarolX!)

 

[Miner]

In the most recent case (complanarity of a intergrated circuits leads), there is an upper limit of 8 mils. Anything less than that is acceptable, hence there is no lower limit.

Recent suggestion:

Tolerance Width = (abs(Upper limit - Parameter Mean)) X 2

The issue I have is that the Parameter Mean will be different on each of the same gauge. I fear this may be less than acceptable for some auditors.

I am looking for an "auditor friendly" definition.

Please be patient, I am new to this...

(Thanks for the warp speed response CarolX!)

In the case of coplanarity (or parallelism) with a maximum of 8 mils, you have a lower boundary condition of zero that a part could theoretically measure. You could legitimately use 0 - 8 as your tolerance.

If the spec were a minimum, it gets sticky. I have seen the reverse of your equation used, as well as using the opposite 3 or 4 StdDev limit as the opposite spec limit, but nothing really work well.

 

[xochi]

Thanks for the suggestion.

I had thought about that once but had opted to follow instructions. Your explanation makes it seem far more plausable than my vague redition. I have already printed out your explanation (theoretical lower limit) and sent it to my management. I will let you know how it comes out.

Thanks for the input!

 

[KenK - 2009]

Minitab 14.20 uses the following formula for %Tolerance for a one-sided spec:

%Tolerance = 3*sigmaM / ABS(xbar - speclimit) * 100%

where sigmaM is the Gage R&R sigma for measurment variation, ABS() is the absolute value, and xbar is the average of all the part measurements.

The more typical formula for a two-side spec is:

%Tolerance = 6*sigmaM / (USL - LSL) * 100%

 

[xochi]

Thanks!

Greetings one and all,

Due to the variations of answers that I recieved, I was forced to actually research the specifications of the limits in question to determine the best methodology.

To my suprise, there is a lower limit!

My GRR was eazy at that point and all is well.

As a side note, the information gleaned from you all has been passed on to others in my company who are actully suffering from the plight that I thought I was.

Your help was priceless.