Cpk for Solder Paste Height

alonFAI

Involved In Discussions
Hi Guys,

I calculated process capability for solder paste height data.
I used the following ucl and lcl = 8.5 and 5
First, are these acceptable?
Second, the data fails for normality test (although the cp cpk are very good)
any suggestions? am I doing something wrong? Does solder paste height data must be distributed normally?

Thanks again to all!!:applause::bigwave::thanks:
Alon
 

Miner

Forum Moderator
Leader
Admin
Many processes do not follow the normal distribution, so this is not unusual. However, some processes that would otherwise follow a normal distribution may not if the process is unstable or if process streams are mixed. So you need to answer these questions:
  1. Are you mixing different process streams?
  2. Is the process stable (in control)?
  3. How is the solder paste applied? By screening?
 

alonFAI

Involved In Discussions
Miner Hi,

First of all I wanna thank you for taking the time and answer all my questions!!!


  1. Are you mixing different process streams? :caution: What do you mean by mixing different process streams?
  2. Is the process stable (in control)?:caution: I'm not sure the machine is new but in general all the results are between 5 to 8.5 (for height) so I would say yes.
  3. How is the solder paste applied? By screening? :caution:Yes
another question, do you have any idea how should I calculate process capability for the solder paste volume? since the volume is determined by the size of the pad so the upper limit and lower limit cannot be set like in height....:thanks:
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Hi Guys,

I calculated process capability for solder paste height data.
I used the following ucl and lcl = 8.5 and 5
First, are these acceptable?
Second, the data fails for normality test (although the cp cpk are very good)
any suggestions? am I doing something wrong? Does solder paste height data must be distributed normally?

Thanks again to all!!:applause::bigwave::thanks:
Alon

The only data you've provided is the measurement data. What is the specification? No way for us to tell if the paste height is between 5 and 8.5 (units?) is "good" or "bad".

In order to even calculate a process capability, you need to know if the data are in control, you haven't stated if that is true or not. Also, are there any correlations in the data - that could through off normality and the use of SPC. This is also a one - sided distribution - height can't be less than zero, so that in itself may cause a failure of normality.

Definitely hard to tell if anything is "wrong" without seeing the SPC chart with the data and the specification, which I assume is two-sided.
 

Bev D

Heretical Statistician
Leader
Super Moderator
since the solder paste is screened on there will also be non-homogeneity in the data. each board type will have a different screen so the volume will vary from product to product, within a screen or board and from board to board.

A traditional Shewhart chart probably can't be used to demonstrate statistical stability.

I would recommend that the OP provide data for 3-5 pad locations within a board and 3 consecutive boards (ideally from 3 different set-ups of each board) from each of 3 different product types. a multi-vari chart can be used to assess the components of variation, determine homogeneity and stability.

I would also why the OP needs to calculate a Cpk value. is for a Customer requirement?
 

Miner

Forum Moderator
Leader
Admin
Process streams are multiple versions of the same process operation that creates a product characteristic. In your case the deposition of solder paste. Lets say that you have an IC chip with 120 leads. All 120 leads have the same nominal dimensions, but each have their own unique aperture in the solder stencil. Therefore, each have a slightly different average volume and are each a unique process stream. If similar enough, they might safely be combined without changing the distribution, but if different, mixing these 120 process stream could distort the underlying distribution.

In control, refers to whether the volumes, if plotted on a control chart (SPC) are comprised solely of common cause variation. In a stenciling operation, the stencil aperture dimensions are fixed, but the amount of paste applied may vary. Though the stencil thickness is fixed, there are still variables such under-filling, application pressure, solder age, etc. that may affect the thickness.
 
Top Bottom