iSixSigma

What is considered a high standard deviation

Six Sigma – iSixSigma Forums Old Forums General What is considered a high standard deviation

Viewing 13 posts - 1 through 13 (of 13 total)
  • Author
    Posts
  • #39589

    BB Australia
    Participant

    Hi all,
    When looking at standard deviation, what is the cut off point where we agree that there is a high degree of variation rather than a low degree of variation? Is is 10, 20.5 etc..
    Thank-you for your time!
     

    0
    #120731

    Anonymous
    Guest

    Good day …
    I typically use the ratio of sigma to target. The reason is because as the target becomes large, the ratio of sigma to target becomes smaller. Using this approach 10% is typical, but it very much depends on the process.
    Cheers,
    Andy

    0
    #120735

    New
    Participant

    I dont agree to this … it totally depends on your process how much variation your process can tolerate

    0
    #120736

    Anonymous
    Guest

    Anyone can specify unreasonble tolerances where a variation of  0% of target is not good enough …. 
    Take some technologies:
    Photolithography target = 0.25 microns, 1 sigma = 0.25 (10%)
    Grinding Form < 10 microns, 1 sigma = 1 micron (10%)
    Diffusion oxide thickness target = 50 Angstroms, 1 sigma = 5 Angstroms (10%)
    I could give many examples from many different types of industry – you only have to provide one!
     

    0
    #120739

    Anonymous
    Guest

    I found some typos – here is the correct response to your objection.
    ———————-
    Anyone can specify unreasonble tolerances where a variation of  10% of target is not good enough …. 
    Take some technologies:
    Photolithography target = 0.25 microns, 1 sigma = 0.025 (10%)
    Grinding Form < 10 microns, 1 sigma = 1 micron (10%)
    Diffusion oxide thickness target = 50 Angstroms, 1 sigma = 5 Angstroms (10%)
    I could give many examples from many different types of industry – you only have to provide one!
    ———————
    Awaiting for your rebuttle …

    0
    #120744

    Thai
    Participant

    Consider the equation to calculate Cp: Tol / (6*stdev).  If you want a six sigma process (Cp=2), then your stdev needs to be 1/12th of the total tolerance or 8.3%.  For a 3 Sigma Process (the point where many people begin to consider a process to be capable, the stdev needs to be 1/6th the tolerance or 16.7%.  All this considered, the 10% value Andy U talked about is a good round number to target.
    Kirk

    0
    #120752

    Utah123
    Member

    The key point to remember is variation must be looked at within the context of tolerance.  If within control limits and specification limits, your resources are perhaps better spend elsewhere.  If looking at a single factor, I would not consider the magnitude of the variation outside the context of the process capability.
    If looking at multiple factors in an effort to determine which is more actionable, COV allows you to rank the factors relative to the porportion of variation.  Good luck.
    Coefficient of variation
    A measure of relative variability, usually calculated only when all values are positive, as in weight, sales, or exam data. The coefficient of variation expresses the standard deviation of the data as a percentage of the mean.

    0
    #120757

    Ken Feldman
    Participant

    I can agree with this and in the past I have posted that a COV of about 10% is what I use as a rule of thumb.  No one has come up with a good rationale for doing anything significantly different.

    0
    #120761

    Savage
    Participant

    I propose we open this up for discussion and hold a biased, one sided debate here in Phoenix.

    0
    #120763

    Ken Feldman
    Participant

    I will be glad to write a Beige Paper supporting my position which I will never submit.  And you can count on me not showing up anyway.  I learned from the Master :-).  But, get Reigle to work on it anyway, it will keep him out of trouble.

    0
    #120797

    vidyut
    Member

    You will agree that no number is high or low, unless there is a reference.
    The customer or competition will decide whether a standard deviation value is high or low.
    What is high for one could be low for another.

    0
    #120802

    Anonymous
    Guest

    I will not agree – provide an example of a technolgy or compare some equipment so that we know you’re not an armchair engineer.

    0
    #120887

    Kumar
    Participant

    You will agree that no number is high or low, unless there is a reference.
    The customer or competition will decide whether a standard deviation value is high or low.
    What is high for one could be low for another.
    To give an example/ case that I have seen –
    Zener diodes -of say 7.5 V, tolerance =+/- 5% ,  or total of 750 mV 
    Now what would be good sigma for this product.?
    How about 75 mV.? Looks good ??
    How about 50 mV ?  Looks better??
    How about 10mV ? Is still better???
    But may not be the best….
    I have seen manufacturers and users for all these values / products.
    Wherever they compete, the low sigma wins – provided customer is aware of the difference/ knows the benefits..
     

    0
Viewing 13 posts - 1 through 13 (of 13 total)

The forum ‘General’ is closed to new topics and replies.