That day, when the Cpk report came out, the entire room fell silent for three seconds, only the rumbling of the machine could be heard.
"Boss, the Cpk for this batch of products is only 1.08, stuck at the edge of the spec. Customers might complain." Xiao Li cautiously placed the report on the table. The boss's brows furrowed instantly into a '川' shape, and the air immediately congealed. I thought to myself, here we go again. This Cpk of 1.08, not good enough, not bad enough, but just unable to pass. The boss glanced at the report and asked, "Are the measuring instruments okay? Is the measurement error too large?" Xiao Li lowered his head and said, "Probably... not? The measuring instruments are regularly calibrated." At this point, I couldn't help but interject, "Are you sure the measuring instrument's resolution is sufficient? Or should we check if that 10:1 golden ratio has been met?"
Where is the problem? Is your measuring instrument "blind"?
Frankly, the 10:1 ratio rule for measuring instrument resolution is essentially reminding us: your measuring instrument cannot be "blind." Imagine if you want to measure the diameter of a hair, but you're using a ruler that can only measure down to centimeters. Then, for you, this ruler is blind. You simply cannot distinguish between 0.01 cm and 0.02 cm, let alone the micrometer level of a hair.
So the key point is that the "smallest distinguishable increment" of your measurement equipment must be much, much smaller than your product's "specification tolerance range." This "much smaller" ratio is generally considered in the industry to be at least 10 times.
In other words, if your product's specification tolerance is 100um, then your measuring instrument's minimum increment must be able to read at least 10um. This way, you can clearly see product variations and avoid "masking" some real quality issues due to a coarse measuring instrument.
How to do it in practice? Divide your tolerance by ten!
To determine if the measuring instrument resolution is sufficient, the method is simple:
- Identify your product's specification tolerance: This is usually your Upper Specification Limit (USL) minus your Lower Specification Limit (LSL). For example, if a product's thickness requirement is 100 ± 5um. Then USL = 105um, LSL = 95um. The tolerance is 105um - 95um = 10um.
- Calculate your "target resolution": Divide your tolerance by 10. For the example above, 10um / 10 = 1um.
- Check your measuring instrument's minimum increment: See what the smallest value your measuring instrument can display is. If your measuring instrument can display down to 0.1um, then it meets the requirement. If your measuring instrument can only display down to 2um, then it is "unqualified."
Frankly, if your measuring instrument's resolution doesn't even meet this basic requirement, any Cpk or DPMO data you see might just be "self-deception." This is like wearing glasses with an insufficient prescription: you see everything blurry but mistakenly think you're seeing clearly.
The most common pitfall: Assuming everything is OK just because the calibration report is fine
I remember once, a new piece of equipment arrived, and the production line colleagues eagerly took it to measure. However, the data measured consistently fluctuated. They showed me the calibration report and said, "Senior colleague, look, the calibration report passed, so this equipment is fine!" I checked, and the calibration report was indeed fine, but the minimum display of that equipment was 1um, while our product's tolerance was only 5um.
According to the 10:1 rule, for a 5um tolerance, the measuring instrument must be able to read at least 0.5um. However, that equipment could only read down to 1um, which means its resolution was simply not enough! It was within the "legal" range, but it couldn't discern details clearly. This is like using a 5x magnifying glass to observe cells; of course, you won't see anything, you need a microscope. Therefore, a calibration report only tells you if the measuring instrument is "accurate," but not if it is "fine enough." Often, it's this seemingly minor oversight that causes your production yield to get stuck and makes you spend more time on repetitive verification.
One thing you can do today
Go back and check your most frequently used measuring instrument: does its "minimum increment" meet "one-tenth of the product tolerance"!