BreakPoint
"Golden" units
I wish to rant.
It is my strongly-held opinion that using "Golden" units is a blatant admission that we don't know what we're measuring and, therefore, can't calibrate our equipment back to a NIST certified instrument (or a group thereof). The fall-back tactic in these cases seems to be to select a "Golden" unit from a pool of working units, slap a calibration sticker on it, run it through the test equipment periodically and compare its performance with the original data. If it continues to pass, the equipment is deemed to still be "in calibration". I hate this.
I believe that the community can offer some insight that will soften my position and help me to feel better about the "engineering" involved in this small part of my world. Please help me by offering some comments that validate this approach to production measurement.
Thanks. I feel better for getting that off of my chest.
@johnsold wrote:
[...] I could see an argument for using an known-good unit as a quick operational check to make sure the test apparatus had not been damaged,
Sure, we all have a drawer full of parts to make quick sanity-checks. Known bad parts are very useful, too. Otherwise we get accused of writing code like this:
[...] particularly if a defective DUT could propagate damage to the tester.
Test technician: "Hey, the test adapter is smoking! It must be bad. Let's test this part on the other adapter. Hey! This one smoked, too. What are the odds of both adapters going bad at the same time."
... later that morning
Test engineer: "groan..."
I worked years ago for a cal house that depended upon standards that were calibrated by the mfgr. As part of applying for A2LA accreditation, we had to provide traceability and uncertainty for many of our measurements that were made by an RLC meter.
We used four standards provided by the mfgr to cal the RLC meter. These were sent to them each year for cal. How did they cal these?
The answer from the mfgr was that they compared our 4 resistor set against three sets of standards they owned, which were in turn compared to a NIST traceable DC resistance standard. But all of RLC measurements were AC, and the 4 resistors were not "standard" values that NIST could not calibrate. Needless to say I was unhappy in trying to document the traceability, so our metrologist was responsible for documenting the calibration. We were lucky that the original designer of the meter was local and still available to provide design and operational expertise of the Digibridge.
You can read the history of it here:
http://www.ietlabs.com/notes/digibridge
http://www.ietlabs.com/pdf/application_notes/OntheCalibrationofDigibridgesandtheVerificationHPH.pdf
http://www.ietlabs.com/pdf/application_notes/OntheFrequencyResponseofStandardResistorsHPH8-94.pdf
A 'gold unit' with sufficiently documented operational details and an scheduled check against a base traceable value (RLC, time, phys property) may be the only way to validate your system.
Henry P. Hall was the designer of the Didgibridge. He is a VERY SMART as well as practical man.
http://www.ietlabs.com/pdf/application_notes/12-Henry%20Hall%20Digibridge%20Patent-Digibridge.pdf
"Gold" only has value if you believe it does. 
@SnowMule wrote:
I use "golden" units to satisfy manufacturing that the equipment is working right. That's fine. I also use them the other way around to show manufacturing that the ATE is fine and their units really are bad.
Calibration stickers don't mean much to them. ...until it expires and QA slaps a huge "DO NOT USE" sticker on their equipment.
When we've got the line running rework/refurbished units, they'll occasionally get a string of failures in a row... is there something wrong with the equipment (Bent pin? Broken wire somewhere?) or is it actually doing what its supposed to do?
A "gold" unit is a good way to keep them happy and quickly prove that the equipment is working and doing what it should. The proper calibration keeps the Quality, Design, Manufacturing and Test engineers happy. It's a part of V&V that doesn't involve any hand-waving.
@Phillip Brooks wrote:
I worked years ago for a cal house that depended upon standards that were calibrated by the mfgr. As part of applying for A2LA accreditation, we had to provide traceability and uncertainty for many of our measurements that were made by an RLC meter.
We used four standards provided by the mfgr to cal the RLC meter. These were sent to them each year for cal. How did they cal these?
The answer from the mfgr was that they compared our 4 resistor set against three sets of standards they owned, which were in turn compared to a NIST traceable DC resistance standard. But all of RLC measurements were AC, and the 4 resistors were not "standard" values that NIST could not calibrate. Needless to say I was unhappy in trying to document the traceability, so our metrologist was responsible for documenting the calibration. We were lucky that the original designer of the meter was local and still available to provide design and operational expertise of the Digibridge. But, with his assistance you were able to validate your process?
You can read the history of it here:
http://www.ietlabs.com/notes/digibridge
http://www.ietlabs.com/pdf/application_notes/OntheCalibrationofDigibridgesandtheVerificationHPH.pdf
http://www.ietlabs.com/pdf/application_notes/OntheFrequencyResponseofStandardResistorsHPH8-94.pdf
A 'gold unit' with sufficiently documented operational details and an scheduled check against a base traceable value (RLC, time, phys property) may be the only way to validate your system. Perhaps I wasn't clear; I meant an uncalibrated gold unit. It's gold just because we say it is, not because we can compare its value with anything traceable. Good point, though.
Henry P. Hall was the designer of the Didgibridge. He is a VERY SMART as well as practical man.
http://www.ietlabs.com/pdf/application_notes/12-Henry%20Hall%20Digibridge%20Patent-Digibridge.pdf
"Gold"A US Federal Reserve Note only has value if you believe it does.
