belchar
18th December 2008, 12:36 PM
A little background. We produce handheld devices and currently use the MIL-STD-105E for our lot check sampling, but there's some desire to start including some drop testing and waterproofing test in our tests (either at the factory level or at the acceptance level), but I'm hard pressed to find a resource to determine standard sample sizes for this type of testing. Preferably, this would be determined as a subset of our standard lot check sample, but I'm open to any and all info.
Any help?
Kurt
gpainter
18th December 2008, 01:03 PM
On our drop testing we do 3 samples to drop and if pass send and additional to the lab for certification. I assume you mean package testing.
belchar
18th December 2008, 01:25 PM
You use 3 regardless of production size or sample size? It's the handheld device itself that we're testing to ensure that drops don't break it, cause to cease functioning, etc.
Similarly, I'm looking for help with the waterproof testing (ours is an ipx7 rated device).
Kurt
gpainter
18th December 2008, 01:27 PM
A good place to go is the ISTA site. Just do a search and you will get there.
Stijloor
18th December 2008, 01:33 PM
A good place to go is the ISTA site. Just do a search and you will get there.
This one (http://www.ista.org/)?
Stijloor
belchar
18th December 2008, 02:13 PM
Perhaps I've not been clear here. Lord knows that it happens often enough! :)
I'm looking for the sampling information for the device itself, not the package it comes in. Most of the packaging drop tests seem to refer to how well a device is holding up in the package it's shipped in, rather than the device itself (or I'm reading things incorrectly).
Kurt
gholland
30th December 2008, 09:41 PM
Our sample sizes are developed through our quality manual based on Juran and other manuals.
For pass/fail test, the basic equation is [log(confidence level)/log (1-upper limit of population not containing defect)]. This is indepenant of lot size or any population sizes.
For example, if you want 95% confidence that 95% of your 'good' parts are actually good, you would test 59 parts with no failures. (log.95/log .95) = 58.44 rounded to 59. That's a common level for us. For greater confidence in your test or to ensure that a greater percentage of your 'good' parts are actually 'good' you would need to test many more.
For 'critical' parts we test several hundred parts. We're a medical device supplier for context.
:2cents:
