But when joe says, "It's yet to damage a meter", it isn't?
Got it.
OK, what if we took joe's actual results:
Pass
Pass
Pass
Pass
Pass
Fail
I believe there is some confusion that I am partly to blame for. If you go back and read my comments about not damaging a meter yet, these refer to the new homemade ESD gun, not the piezo ignitor. While both devices put out a fairly fast transient, if you followed along when I tested the 181A and later was working on the new gun, you are aware that the gun can supply much higher current levels than the piezo.
The data that you present I assume came from my spreadsheet and you just missed on meter in the beginning. This data is for the piezo ignitor, not the new gun. To avoid this confusion, I have added a new column to the spreadsheet for the new gun.
I had planned to stop using the piezo ignitor but people have requested that I leave it in. I think this stems partly from my comments about how no meter should ever be damaged by it and that we have seen so many UNI-Ts fail this test. It's pretty much what I consider to be the least stressful test I have ran on the meters.
OK, what if we took joe's actual results:
[snip]
What would that be called?
What predictions could you make about an unknown, randomly chosen meter using that information?
Joe's result (including his documentation about his test setup and the actual meter models) would definitely be data. Using this to predict another random meter is tricky. If we assume that the group Joe tested is representative for the group we pick the random meter from, I would say that the meter has an expected pass chance of 38/43th. However, if Joe only ever tested Fluke meters which are the definition of perfection (), and now started tested $5 meters, that expectation may be way off.
You might be able to improve the estimate if you split the results by manufacturer or price class. How many of the meters that Joe tested with the current procedure were around the $20 mark? And even then, you can only figure out an expected pass chance. A meter that is 90% likely to pass a test can still fail.
If Joe is not comfortable in extrapolating the data to an untested meter, then I would not extrapolate the data either.
I agree. Even if we are all talking about the piezo and not the new gun the sample size is just too small to say much about all meters available. Even if I had ran more brands and models, just from the little testing I have done, I would be VERY hesitant to extrapolate how new meters would perform.
Considering that 3 or the 5 meters that failed the piezo test were UNI-Ts, I fully agree that splitting the results would improve the estimate. Still my confidence factor would be poor.
However, if Joe only ever tested Fluke meters which are the definition of perfection (), and now started tested $5 meters, that expectation may be way off.
Really? The Fluke 87V failed a test that this meter's $15 sibling passed.
FWIW, here are the PCBs of the two meters in question (AN8002 and AN8008).
Me? I've got $10 that says the AN8008 will pass the sparker test.
PS: Is that a Rubycon capacitor?
I agree that the 87V is really a flyer (outside the norm) from what I have seen with Fluke. Similar to the UNI-T 139c that actually fairs well in my tests compared with the vast majority of UNI-T DVMs I have looked at. I wonder if the new gun would damage one.. Anyway, this is exactly why even within a brand, I would be hesitant to comment about how an untested meter would perform.
If you now understand that there are two different ESD tests, one with the piezo grill ignitor and one with the new gun, what are you considering the "sparker test"?
ESD is a fun one to design for. There's a fair difference between the two PCBs you show.