1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

Ppk VS Cpk

Discussion in 'Capability - Process, Machine, Gage …' started by Pongsakorn, Feb 19, 2016.

  1. Pongsakorn

    Pongsakorn Active Member

    Joined:
    Aug 10, 2015
    Messages:
    72
    Likes Received:
    16
    Trophy Points:
    7
    Please advise how do I explain to the customer as they need me to use Ppk calculation instead of
    Cpk calculation for ongoing production.
    With the same set of data, the Ppk calculation shows about 1.5 whereas Cpk calculation shows more than 2.00.
    Since the target is 1.67, using Ppk will never meet such target and customer keep asking for the improvement while I have no more idea to improve further.

    Please advise if any reason that I can convince customer to use Cpk calculation instead of Ppk for ongoing production.
     
  2. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    this sounds like a formula / application misunderstanding.
    the goal of 1.67 typically has been applied to short term variation using the 'long term' or total standard deviation. in this application, teh short term variation is assessed by sampling approximately 100 parts all made under the same conditions: hsort ttime period, one set up, one lot of each fo the raw materials, one pice of equipment, one operator, etc. This is usually applied durign pre-production periods when a diversity of material is not available. given a Cpk of 2.00 I am assuming that you are using the formula for 'short term' variation which is the within subgroup standard deviation of any subgroups.

    too often Ppk and Cpk goals are simply set in stone and the only way to 'get around them' is to:
    1. be sure you follow the correct sampling and formula requirements
    2. actually improve your process performance
    3. take this information resulting from 1 and 2 to your suppleir quality rep and have a logical physics based discussion.

    IF you can provide more details about who your Customer is, what product/feature you are asessing, what process creates the feature and how you actually calculated your Ppk/Cpk indices, we can provide better advice. Without this information we can only guess as to what yoru real issue is.
     
  3. Pongsakorn

    Pongsakorn Active Member

    Joined:
    Aug 10, 2015
    Messages:
    72
    Likes Received:
    16
    Trophy Points:
    7
    Hello Bev, really thanks for your reply.
    You are correct, the "Estimator Sigma" (R bar/d2) is used for Cpk calculation and "Summed Square Sigma" is used for Ppk calculation.
    The product is IC Package, the SPC is used for product characteristics "Ball Shear Strength" (destructive test). I am not really able to tell
    customer name.
     
  4. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    261
    Likes Received:
    168
    Trophy Points:
    42
    Location:
    North Carolina
    I laughed until milk came out of my nose.

    Well, coffee actually, it's early.

    Ow.

    :)
     
  5. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    no need to name Customer knowing the industry is enough. can you post the data? (not the horrible minitab reports but the actual data). do you have a distribution of the breaks? at the bond pad, the wire, the package?
    are you sampling the same wire or different wires or a random selection of wires in the package? what is your sample size? how many different set-ups of the wire bonder were used? How many wafer lots were sampled?
    how many assembly lots?
     
    Pongsakorn likes this.
  6. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    some supplier reps are logical; some are not.

    just another reason why these indices are stupid.
     
    Pongsakorn likes this.
  7. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    261
    Likes Received:
    168
    Trophy Points:
    42
    Location:
    North Carolina
    OK. My curiosity is now piqued.

    I am looking at the output of a new process and I want a sense as to how well I will achieve my tolerance. I'd like this sense to be numeric so I can rank them worst to best to go after the "riskier" or less stable process first. Not sure why the capability metrics would be bad ....

    I DO agree they are inverted. They are basically tolerance divided by process noise. So a Cp of 1.33 is 4/3. If I invert that, I get 3/4 = 75%. Or, a +/- 3 sigma distribution would use 75% of my tolerance. To me (at least) that is far easier to visualize than 1.33. A Cp of 2 means my "normal" process variation will use half my tolerance. (Invert 2 and you get 1/2 = 50%)

    But I am unclear as to why they are stupid .... Is there something fundamentally wrong with compare a sample standard deviation to a tolerance?
     
    Pongsakorn likes this.
  8. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    I certainly understand the desire to make something simple and rank it. (after all, sports are very important in our culture)
    I do suggest that you read my "process capability" paper in the resources section...it addresses - from my experience - the issues with the indices.

    briefly tho, the basic ratio of tolerance spread to process spread (as quantified by the SD) was the original point of the Cpk index (L. P. Sullivan "Reducing Variability: A New Approach to Quality”, Quality Progress, July 1984 and “Letters” Quality Progress, April, 1985). It's a neat concept, but since then the whole 'quality profession' has been defrauded by the conflation of categorical defects with a continuous data distribution...
     
    Pongsakorn likes this.
  9. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    Another point about the ranking is that the smaller sample sizes typically used to calculate the SD provide horribly imprecise estimates of the SD. If you use enough data then you could get around this...
     
    Pongsakorn likes this.
  10. Miner

    Miner Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    576
    Likes Received:
    492
    Trophy Points:
    62
    Location:
    Greater Milwaukee USA
    Just a little bit of history for you. Around the same time that Ford came out with the Cp index, GM came out with the Capability Ratio (CR) (could not use the same metric as Ford). The Capability Ratio was the inverse of Cp, so it was exactly as you prefer. Shortly thereafter, Ford added Cpk. GM could not follow suit because this led to "Divide by Zero" issues when the Cpk approached zero. In addition, as Cpk linearly approached zero, the CR would exponentially change. GM dropped the CR after a few years.
     
    Atul Khandekar, Pongsakorn and Bev D like this.
  11. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    261
    Likes Received:
    168
    Trophy Points:
    42
    Location:
    North Carolina
    But ... but ... I'm in supplier quality. If it's not in pretty crayon colors, I can't focus long enough.
     
  12. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    261
    Likes Received:
    168
    Trophy Points:
    42
    Location:
    North Carolina
    No. Because that would make SENSE. I mean, why would one want to use consistent metrics with all the same suppliers and focus on making good parts when we could choose to spend so much effort to not be like the other guy?
     
  13. Miner

    Miner Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    576
    Likes Received:
    492
    Trophy Points:
    62
    Location:
    Greater Milwaukee USA
    What can I say. It was pre-AIAG. No pretty blue manuals.
     
  14. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    261
    Likes Received:
    168
    Trophy Points:
    42
    Location:
    North Carolina
    At BevD: OK. I read it. Not so sure I am ready to throw out Cp/Cpk calculations. Totally agree that by themselves they do not paint the picture. But I'm not sure I have ever accepted JUST the Cp/Cpk numbers from one of my suppliers and called it good. It always comes in a study with a run chart and a histogram or it gets rejected. I still think there is value in saying "Were I to assume normality, how does the process spread compare against the tolerance." I also like to have the raw data.

    Where I see it totally abused is in the receiving inspection department. I have had quality engineers ask them for Cp/Cpk on parts taken randomly from boxes. Which is invalid because you have absolutely zero guarantee you are taking them sequentially as they were made.