DPMO calc and Overall Manufacturing Index (OMI) calculations

To clarify on "Opportunities"-

With 1000 components, the opportunities are far greater than 1000. You have 1 opportunity per component for a component defect i.e. bent lead, damaged part, tested bad, etc. You have a second opportunity per component for placement defects i.e. missing, misaligned, reversed, tombstone, etc. You have multiple opportunities per component for each termination (solder joint, etc.)

The sum total of all opportunites is the basis for calculating PPM.

example-

1000 components 999 placements (the PCB is not counted for a placement) 9001 terminations

11000 opportunities for the assembly

Steve is right in saying that examining the defect type and doing something positive about it is the real goal.

By using the PPM method and categorizing the defect types you will quickly see where you should spend your efforts in trouble shooting the process.

I have some assembles that have most defects in the terminations - I then trouble shoot the paste print, and wave solder processes.

Others are more heavily placement defects - which focuses on Pick and Place, through hole placement, 2nd ops assemblers errors.

If you have a good Incoming Inspection, component defects are usually the last thing to worry about.

I only use First Pass Yields in percentage for Test, but all defects found in test are also included into the daily PPM calculation per assembly.

One last comment - OMI calculations are good for examining the quality of a single assembly produced at multiple sites. An OEM would use this calculation to examine various suppliers. Since my company is a CM, I don't bother with OMI, that's for my customers to implement if they are concerned about multiple manufacturing sites.

reply »

IPC7912 deals with the end result, ie an OEM getting boards from a contract manufacturer. I have a friend who is sitting on these committees. He tells me there is a new standard due out for actually doing in-process DPMO.

From a process standpoint, DPMO (or ppm) is much more sensitive to variattions in the process. It also allows you to compare your boards. FPY does not take into account board complexities. For example: I have two different boards. The first has 2500 oportunities for error. It is a fairly complex board. FPY is about 80%. PPM is less than 150. My second board is very simple, 57 oppotunities. It has a 95% FPY, but a ppm level of 700. Hmmm which do I work on first? Looking at FPY would lead you to look at the more complex board, while PPM says check the simpler board. I normally find that the board with the higher ppm is the one to attack first.

Brian

reply »

Dave et al,

I just print out and read a copy of the aforementioned recommended article. Mr. Mangin gives a clear picture of why DPMO is to be considered a more pure data than First Pass Yield (FPY).

However, I must note that his interpretation of how to formulate DPMO is not entirely in line with the IPC 7912. It may be that his article is outdated in that respect and it may be that his current opinion would be changed to more closely follow IPC.

I have also solicited and received a draft of IPC 9261 for In-Process DPMO. It is a straight forward and precise data collection and calculation method that should produce the desired results. It does require a dedication to detail and is a complex system that strict discipline for data collection must be followed to achieve meaningful results.

What is still missing from all information I have perused is the target values and ranges, Best In Class, for DPMO. It could be premature to expect this data but I would still like to hear from the community at large what their experiences may be and collectively we may come up with recommendations that all can follow to achieve the goal of Zero (or significantly reduced) defects.

reply »