I have been assigned the task of performing calibration interval analysis on two of our company’s products to see if the calibration interval can be extended. A visit to the Metrology Lab provided a copy of RP-1 from which I selected method S2 as being the most appropriate. Preliminary data analysis indicates the current calibration time is at the 95% reliability target, which negates any attempt at extending the calibration interval without further intervention. Rather than take my management bad news, I prefer to take them solutions. I want to lay out a scenario of what needs to change in order to extend the calibration time and let them make the business decision as to whether to implement or not. To accomplish this I decided I need to further analyze individual "variable" failures to get a better visual representation comparing trend, test tolerances, and confidence limits. I can fit a curve through data, calculate standard deviations, and confidence limits, but now I want confidence bands around a curve corresponding to 95% confidence and this confidence band is a function of time.

Enter “Calibration Intervals from Variables Data”

Calibration Intervals from Variables Data.pdf (http://www.isgmax.com/Articles_Papers/Calibration%20Intervals%20from%20Variables%20Data.pdf)

Equation (14) is not working. The model is second order. (14) is returning a one row 2 column “variance” matrix that I can’t seem to pull the elements out of . And if I could get the elements out I have drawn a blank as to how to incorporate these elements into an equation to plot the confidence boundary vs. time.

I am looking for the solution from (14) to a confidence vs. time function OR an alternative set of equations “that work and are understandable” to calculate and plot confidence bands as a function of time.


Thanks

Didn't mean to start my own blog here. After staring at "Uncertainty Growth Estimation in Uncertainty Analyzer" this morning. I got the brainstorm to compromise with MRbar/d2 scaled to enclose 95% of a normal distribution as a graphical solution that may work. I have enough data points. Follow up research this pm on Rbar/d2 revealed "Uncertinty Analysis and Parameter Tolerancing" "Figure 4" "(Deviations from Nominal)" confirms that this may be possible. Will give the math a try next week.

Thanks

Interesting topic. The post got in at the start of the weekend when it's pretty slow here. Hopefully someone will pick up on it by or on Monday.

Also, by all means - Let us know what you come up with as well. We appreciate the information!

Make sure that you test for significance on your equation. Your scatterplot looks an awful lot like a shotgun blast, which does not visually indicate much of a relationship. The equation that you developed may be heavily influenced by a few points.

Thank you for the paper. I do laud the effort to empirically develop methods to improve quality.

As your second post suggests, I assume you have found your answer to your question.

Are you at a point of validating the method? I was greatly intrigued to see the author suggest not using weights. Without sound theoretical justification for the variables chosen and no weights, I wonder if consistency problems will develop in replication.

I would be intrigued to learn from the author why he/ she chose not to use weights in the regression analysis. Plausibly, rational discussion would merit the use of the chosen parameters, but the data may suggest otherwise. Too, it might be interesting to explore some additional predictors of appropriate calibration interval.

Again, thank you for the article. I enjoyed reading it. Please repost with your results, and any correspondence with the author.

scatterplot looks an awful lot like a shotgun blast, which does not visually indicate much of a relationship.

That's what I was thinking, but not being an expert in the field I figure(d) I'd keep my mouth shut.

Yes - It's interesting. I've never seen anyone go this far (precise?) in determining calibration intervals.

Miner: The equation is heavily influenced by a few points. Least squares best fit does this. Its the ouliers that throw the prediction off. Which opens up a little of the next question. How do I deal with outliers. Every article I read has a different perspective. To predict take them out, but leave them in because they are the most intersting cases? Better yet analyze it both ways (which I don't really want to do).

BradM: I have not found the answer to my question, BUT I have an Rbar/d2 variation to try next week. Does least squares best fit return an average? Why not use least difference best fit?

Your observation about weighted vs nonweighted brought a flashback when I read it. I think the author doesn't use weights because he hasn't validated this method yet I also detected some hesitancy about validity of first versus second order fits. Notice my curve is second order, it beat out first order by .001.

Validation! I've never done a calibration interval analysis before so you get the opportunity to watch me struggle up the learning curve. I will do a lot of research and technique selection, crunch 100's of thousands of data points, (there are 313 variables/calibration and 270 As Found calibrations to analyze for the first product) make a lot decisions on what gives the most correct answer (make my best guess). Validation will come with time. As I gain experience the analysis method can and probably will be tweeked. Remember "my" analysis is a combination of RP-1 S2 and Intervals from Variables data. S2 is over glorified validated standard practice, the variable data intervals technique I suspect is not validated, but its theory sounds really good.

Marc: I don't see a shotgun blast, I see a normal distribution following the plotted curve of the mean. The shotgun blast x axis is "weeks", the y axis is dBm. The trend is -0.5 dBm drift downward in 100 weeks. The test tolerance is +-1dBm. +-1.96 standard deviations is equivalent to the Reliability Target of 95% (95% of As found product will be IN-TOLERANCE) and where it crosses the lower variable test tolerance is the estimate of calibration cycle time. I can eyeball this crossing at 80 weeks, the current calibration cycle is at 52 weeks.

I still want a mathematical technique to get the 1.96 standard deviation curve plotted around the mean, so I can get closer than eyeballing 80 weeks.

I get to do this with 313 variables/calibration. The Pass Fail of a S2 calibration is determined by the variable that goes OUT OF TOLERANCE first. The shot gun plot is one of the Variables that caused a S2 Fail.

I suspect I will find a 1:1 correlation between this variable and TCXO drift. ie the drift in the frequency standard is causing frequency shifts in all the mixers, filters, and IF bandwidths in our product and this is causing the output amplitude to roll off at -.5 dBm / 100 weeks. I will get around to checking for this correlation eventually.

So Marc, the data point in the lower right hand corner caused an entire S2 calibration to fail. Is it a valid data point or an outlier? Do I have a real S2 calibration Fail or can I change it to Pass?

My gut says its a valid data point and this calibration is a confirmed Fail.

Thank you!! What an excellent follow-up to our posts. This is such an interesting topic (at least to me). I'll "mull" over your post, and give you some more of my thoughts. Maybe since we're in the week now some others can chime in.

Please do keep us informed of your testing process.

Yes - It's interesting. I've never seen anyone go this far (precise?) in determining calibration intervals.

Actually, the U.S. Navy at their Engineering Center in Corona has a department that has done only that type of work for 20+ years.....they have a staggering amount of data.....and I believe the Navy also has had a lot of input to RP-1.....

I couldn't download the information tonight, so I couldn't tell if it is taking more of a Monte Carlo approach (which is making a big come back in Metrology it appears), but is appears to have some similarities, based only on the write-up.

Hershal

Just out of curiosity, in what circumstances is such precision in determining calibration intervals necessary?

Found it. My MRbar/d2 alternative to Castrup math works. I believe my solution to growth of standard deviation with time is more validated than that proposed in Castrup’s white paper. d2 for a sample size of two is 1.128379 and since the chart doesn’t go any lower someone once told me to use the 1.128379 for a sample size of one. WRONG! WRONG! WRONG!

I had to go back to basics. MRbar is the average of a one sided normal distribution. To convert MRbar to standard deviation divide by .67449 (d2 for a sample size of one). Then multiply that result by 1.6449 to get the standard deviation for a one tail 95% Confidence.

The first big graph MR (Moving Range - Rbarpoints) has a smoothing function ksmooth, 15 points. That’s the curve weaving up and down. It’s a moving average. A line function returns slope and intercept for the average. The ksmooth and line agree at the two heavy clusters of data. I’m very pleased with this result. Line*1.6449/.67449 is the boundary for 95% confidence as a function of time.

There are 17 points outside the estimated confidence boundary. 17/267 = 6.4%.
A normal distribution would have 5%, but this is real data and the next graph shows some skew on the lower side. Once again I am pleased with the agreement.

Now that I have accomplished this for a normal distribution I realized I can derive factors to convert MRbar to chosen Reliability Targets for non normal distributions.

The last graph is the model selected by residual sum of squares with one tail 95% Reliability Target boundaries. Origin corrected for Type B Expanded Uncertainty.

Validity?

Calibration Intervals from Variables Data: This one variable analysis has 5 fail data points out of 267, 1.87%.

S2 Intervals Analysis: I have looked at 93 calibrations with 2 failed by the above variable, 2.2%.

I now have templates for Predicting Calibration Intervals from S2 and Variables. I have 177 more calibrations and 312 more variables to look at. And am curious as to how well these two methods will agree. Tomorrow: calibration crunching.

I am still finding improvements that can be made to this template. Can you understand the last page? Does it have the pertinent information on it? I think I will add sample size. If you have any suggestions, please comment.

Just out of curiosity, in what circumstances is such precision in determining calibration intervals necessary?

Money.

The military, and the Navy/Marine Corps in particular, have been able to save tens of millions of dollars over the last few decades by continuously studying that.

Hershal

Yes, I understand the money aspect. I was more looking for specific situations where a lot of money would be saved. For example, I wouldn't think a metal stamping house using a CMM typically would need that level of precision to determine calibration frequency for the CMM.

Good question, Marc. This is my take:

First, I would be assuming that you have a calibration program with access to the appropriate historical numbers. Now, say I have six hundred (or more) instruments a year that I have to send out. We all understand the costs associated with sending instruments out. So I have the applicable information to use such a tool.

So, if I have a defensible tool to help me establish a more reliable indicator of calibration intervals, than say six months, one year, etc., it would be feasible for me to utilize such a tool. I have compounded my savings: lowered costs associated with sending instruments out and lowered costs associated with calibration failure. I can not only use this to decrease the frequency, but increase.

Marc: Read the last paragraph.

http://news.minnesota.publicradio.org/features/2003/08/28_zdechlikm_wellstonesettle/

or about VOR meter

http://news.minnesota.publicradio.org/features/2003/03/03_zdechlikm_wellstone/

That VOR signal was calibrated by a technician with some company’s instrument. You get a feel for the money that can be involved over and above the yearly metrology department’s budget. Marc this is getting scary. Your question prompted a search on ““out of tolerance” litigation” and I found the above articles. Both products I have been asked to do this analysis on are capable of calibrating VOR. How about lives instead of money.

Hershal: If there are analysis methodologies similar to S2 or variables analysis available from the Navy, where do I find it? You talked about downloading.

Still request feedback on suitability of Plot03.

Marc: Read the last paragraph....about VOR meter

Point well taken. As a 'former' pilot I am well aware of aspects such as this. However, in the case you cite it doesn't address calibration frequency. I know lots of places that end up using equipment when it is known to be late for calibration, for example, and at times when equipment that is KNOWN to be out of tolerance. The article you linked to says "Federal officials tested that VOR signal and found it was slightly out of tolerance after the crash." It does not address calibration frequency of the instrument used.

How about lives instead of money.

You can't nail me on that. If you're doing aluminum casting of an alternator case and you're off a bit, there's a (probably) statistical probability of failure due to an out of tolerance condition causing loss of life of (I would bet) zero.

Good question, Marc. This is my take....

Yup. I guess what I was getting at is there are situations where such a detailed analysis isn't necessary. In other situations such a level of analysis is not just important, it is critical.

I'm not trying to scare anyone. I bring this up because over the years I have worked with so many different companies and there are so many unique situations. Because so many people come here from all over the world, from big corporations and governments to 4 person mom and pop shops, I wanted to inject a bit of 'common sense' here. A contrast, if you will, so that smaller shops don't take this as a 'Must Do' situation - That everyone should look to their specific company and assess what they need. I had one client, for example, that had 4 measurement devices. One was a scale to weigh pallets for shipping putposes. I have had other clients, like the old Motorola semi-conductor sector, which had thousands. One client made elevator cabs. There were almost no critical measurements (+/- 14 inch was the typical tolerance). In short, I'm throwing some balance in the discussion, I think.

I do want to say this is a most excellent discussion thread. It's the kind that makes this forum excite me. Hershal's participation is especially appreciated because Hershal is one of the rare expert metrologists. Hershal has seen so much, and is so knowledgeable, that I really appreciate his participation.

As an FYI - I 'cut my calibration teeth' in military manufacturing of various avionic and nuclear submarine assemblies (from communication, control and navigation LRU's {line replaceable units} to hermetic bulkhead connectors) at what was then Cincinnati Electronics. The calibration laboratory at CE had very precise standards for obvious reasons. For example, resistance standards were kept in liquid baths in a room where you had to go through three sets of doors just get in. Having said that, it was a different world when I jumped into the 'commercial' world and got involved with companies where tolerances were so much bigger than those required in the environment where I learned about calibration and calibration systems.

I'm not trying to say calibration frequency isn't important. I'm just trying to inject thought about the multitude of scenarios.

Kansas has been in Texas all week witnessing testing.

Marc: To elaborate on BradM post. It’s the cost of quality. It’s a bathtub curve.

For calibration intervals to short, manpower and out of service time create additional cost.

For calibration intervals to long, risk of shipping nonconforming product increases. The lower cost level is Customer Satisfaction and Rework. The upper cost level is litigation both contract and liability. Additional cost, just a different flavor.

We want to be in the bottom of the tub. …..did I type that?

Follow up on Wellstone crash final report cleared the VOR, citing pilot error. So I’m not so scared any more.

Alternator………would that be the Horton Emergency Vehicles, John Molinari, Bobby Labonte, or just plain Nissan

http://www.ems.ohio.gov/special/NHTSAnotice04.htm

http://www.njatty.com/articles/auto/jmsm03.html

http://www.joegibbsracing.com/season_2005/news_cup/10_oct/051015_bl_race.php

http://www-odi.nhtsa.dot.gov/cars/problems/defect/defectresults.cfm?start=1&SearchType=DrillDown&type=1&year=2003&make=NISSAN&model=MURANO&vehtype=MP&component_id=203&prod_id=201597&PrintVersion=YES

Sorry Marc, couldn’t resist. But you made your point. We don’t care if a few barbecue covers are an inch to long. And a pilot is supposed to be able to recognize a broken VOR.

Back to basics……….

Does anyone out there use any form of variables calibration interval analysis?

Hershal where are you? Still want to know about downloading.

I guess what I was getting at is there are situations where such a detailed analysis isn't necessary. In other situations such a level of analysis is not just important, it is critical.



Sorry, Marc. Did not see your response on this one. It must have been during my "where are my notifications going?" phase.

I agree totally with you :yes: about what I perceive is the sentiments of your follow-up post, and the appreciation for Hersal (and all your professional covers) who take the time daily to share their expertise here.

I think experience and time teaches the professional about calibration frequency. I think even if you are a beginner, after you have instruments calibrated over a few cycles, one can begin to see whether the intervals are established properly (if you're looking).

I got excited (and am still excited) about this thread when there is some life given to actually looking at your calibration program and managing it. So many times, people get something calibrated, file it, and move on. They never look at their certificate, review the work being done, determine if the proper frequency is in place, if they have the right tool, etc. These forums are jam-packed with confused individuals who legitimately ask two fundamental questions: 1) what is my tolerance, and 2) what should my calibration interval be?

Ok, for #1, most of the time we say Mfg. tolerance. Since I'm talking to my cohorts, I ask:" how many of you have confidence in Mfg. specification??" If it's Fluke, I trust it implicitly. If it's XXXXXXXX, I have no confidence, and establish my own. So then there are several instruments in-between. There is no oversight whatsoever as to how companies ascertain their stated accuracies or recommend intervals. I know, that's a broad brush. But like I said, when I look at Fluke's analysis, you know they know what they're doing. Others?? You have to dig (sometimes I have to call) to find if there is any stated tolerance for the equipment even established.

By following established uncertainty analysis procedures, you can determine this. This is why I appreciate Hershal's torch-carrying on ISO17025 and referring people to legtimate, accredited labs that take pride in their work and do it right.

As for #2, in my experience, this one is a little more tricky. Say four of us bought out a calibration department of a corporation. What would be our intervals? Set aside objective analysis for a second. I bet it would be pretty short, right? We're not being deceptive or unethical. We are a business to make money (short interval means more money$$) and we're erring on the customer's safety side by not letting it go too long and being O.O.T. Sounds like a win-win, yes?

I just think it would be neat to have a tool to run some numbers through should I desire. If I could extend some and shorten others, giving me some $$$ saved to demonstrate to management, that has some promise. But to your point, it would probably not be useful to many others who manage a small group of instruments.

But.... that goes to the database thread. Where Access is the greatest thing in the world to me; for people who really work with databases, they prefer much more robust packages.

As for #2, in my experience, this one is a little more tricky. Say four of us bought out a calibration department of a corporation. What would be our intervals? Set aside objective analysis for a second. I bet it would be pretty short, right? We're not being deceptive or unethical. We are a business to make money (short interval means more money$$) and we're erring on the customer's safety side by not letting it go too long and being O.O.T. Sounds like a win-win, yes?

Sorry BradM. Short interval more $$ is not win win. Step back and look at the big picture.

Company A voltmeter has a recommended calibration interval of 1 year. Company B voltmeter has a recommended calibration interval of 2 years. Both meters have the same tolerance and can be used for our application. Guess which one I’m going to buy. So while company A makes short interval dollars they are losing market share to company B.

This is where my calibration interval analysis is going. By identifying the current calibration interval design limitations, we can redesign to improve our product towards longer calibration interval and put the competition out of business. Well at least make them uncomfortable.

By identifying the current calibration interval design limitations, we can redesign to improve our product towards longer calibration interval and put the competition out of business. Well at least make them uncomfortable.

I agree, but then again... Manufacturer claims (even warranty aspects) may not tell the whole story. One would have to take both meters and put them in identical (or very similar) use scenarios and see how well they hold their calibration.

In most situations I would look for what I felt was the best 'built' device or instrument, I would set a 'reasonable' calibration interval and check results each time calibrated. If you get a new meter that is used daily in a very rough environment I wouldn't trust the manufacturers recommendation. I'd look to other multi-meters I already had and start there. Obviously this wouldn't apply in a startup with no history, but those are the exception.

I think for most non-critical (lives do not depend up it) applications which is the case in many companies, measurement equipment calibration cycle time should be looked at in terms of calibration history. My 'rule of thumb': If the instrument keeps coming back in calibration without adjustment, lengthen frequency. If adjustment is necessary but the instrument is within its tolerance, the frequency is probably about right. And, of course, if it comes back having needed adjustment AND was out of tolerance, shorten the cycle. Note that my 'rule of thumb' is general. For example, if a review of the calibration history for the device shows that it was stable until a certain point in time (coming back in calibration without adjustment, or minimal adjustment is necessary but the instrument is within its tolerance), one should be looking at the integrity of the instrument (for example, is it wearing out?).

Now, let's take another scenario I'd like feedback on. A company has 20 digital multi-meters. There are 5 in the calibration laboratory and each is used approximately 5 times a day, 2 are kept by product engineers whose use is not able to be tracked, 13 are used on the line and each is used at 15 minute intervals on each of 2 shifts, and 5 of the 13 are also use on a 3rd shift (same 15 minute interval scenario). Not relying on calibration history, how would one set a calibration cycle for each?

Sorry BradM. Short interval more $$ is not win win. Step back and look at the big picture.

Company A voltmeter has a recommended calibration interval of 1 year. Company B voltmeter has a recommended calibration interval of 2 years. Both meters have the same tolerance and can be used for our application. Guess which one I’m going to buy. So while company A makes short interval dollars they are losing market share to company B.

This is where my calibration interval analysis is going. By identifying the current calibration interval design limitations, we can redesign to improve our product towards longer calibration interval and put the competition out of business. Well at least make them uncomfortable.

Excellent job. And agreed :agree1:

Realize my previous post was Devil's Advocacy, and a bit tongue-in-cheek. Observe my statement about setting objective evalution aside. I was implying that most companies out there (IMO) are approaching it this way. I believe many do not have rational approaches to establishing frequency intervals. Too, more of the population responds positively to price and stated accuracy. Few have the knowledge (or the tools, or the desire) to critically analyze the accuracy, calibration methology, etc. As Marc suggested, there are so many variables to the same instrument, it would be difficult to do. Unless... a robust model can be utilized.

I up came with a system several years ago that is competely arbitrary but has served me well. If an instrument has needed adjustment within two calibration intervals I reduce the inverval by 50%. If it passes through 4 intervals without requiring adjustment I increase it by 50% and then review it after four more calibration intervals to see if I can extend it further.

Not very scientific, but it has worked for me.

I think for most non-critical (lives do not depend up it) applications which is the case in many companies, measurement equipment calibration cycle time should be looked at in terms of calibration history. My 'rule of thumb': If the instrument keeps coming back in calibration without adjustment, lengthen frequency. If adjustment is necessary but the instrument is within its tolerance, the frequency is probably about right. And, of course, if it comes back having needed adjustment AND was out of tolerance, shorten the cycle. Note that my 'rule of thumb' is general. For example, if a review of the calibration history for the device shows that it was stable until a certain point in time (coming back in calibration without adjustment, or minimal adjustment is necessary but the instrument is within its tolerance), one should be looking at the integrity of the instrument (for example, is it wearing out?).

I think that is an excellent system, and would serve most fairly well. Basically you have three decision criteria: 1) Failed calibration, 2) passed calibration (but with adjustment), 3) passed calibration (no adjustment). 1 and 3 are fairly consistent; 2 is my interest. Does the vendor always adjust, never adjust unless OOT, adjust at 50%? If I had a little more of a robust system to take the percentage and estimate a frequency, that would be useful.



Now, let's take another scenario I'd like feedback on. A company has 20 digital multi-meters. There are 5 in the calibration laboratory and each is used approximately 5 times a day, 2 are kept by product engineers whose use is not able to be tracked, 13 are used on the line and each is used at 15 minute intervals on each of 2 shifts, and 5 of the 13 are also use on a 3rd shift (same 15 minute interval scenario). Not relying on calibration history, how would one set a calibration cycle for each?

Good one. Why would you not want to at least consider calibration history? Is it not available? I’m saying whether you go off your rule of thumb, or the most sophisticated system available, any realistic forecast should start with historical performance (IMO).

ALGORITHMIC METHODS
Other methods utilize simple to complex decision algorithms to adjust calibration intervals in response to in-tolerance or out-of-tolerance conditions observed during calibration. Typically, these approaches consist of instructions to lengthen or shorten calibration intervals in response to current or recent observations. Because of their nature, these methods are labeled algorithmic methods. Algorithmic methods have achieved wide acceptance due to their simplicity and low cost of implementation. However, most algorithmic methods suffer from several drawbacks.

The following list is fairly representative:

1. With most algorithmic methods, interval changes are in response to small numbers (usually one or two) of observed in-tolerance or out-of-tolerance conditions. It can be easily shown that any given in-tolerance or out-of tolerance condition is a random occurrence. Adjusting an interval in response to small numbers of calibration results is, accordingly, equivalent to attempting to control a process by adjusting to random fluctuations. Such practices are inherently futile.

2. Algorithmic methods make no attempt to model underlying uncertainty growth mechanisms. Consequently, if an interval change is required, the appropriate magnitude of the change cannot be readily determined.

3. Algorithmic methods cannot be readily tailored to prescribed reliability targets that are commensurate with quality objectives. The level of reliability attainable with a given algorithmic method can be discovered only by trial and error or by simulation.

4. If an interval is attained that is consistent with a desired level of reliability, the results of the next calibration or next few calibrations will likely cause a change away from the correct interval. To see that this is so, consider cases where reliability targets are high, e.g., 90%. For a 90% target, if the interval is correct for an item, there is a 0.9 probability that it will be observed in-tolerance at any given calibration. Likewise, there is a 0.81 probability that it will be observed in-tolerance at two successive calibrations. With most algorithmic methods, such observations will cause an adjustment away from the item’s current interval. Thus, algorithmic methods tend to cause a change away from a correct interval in response to events that are highly probable if the interval is correct.

5. With algorithmic methods, although a correct interval cannot be maintained, a time-averaged steady-state measurement reliability can be achieved. The typical time required ranges from fifteen to sixty years.

6. With algorithmic methods, interval changes are ordinarily computed manually by calibrating technicians, rather than established via automated methods. Accordingly, operating costs can be high.
Quote from Dr. Castrup

In metrology world it is the user of the product that has responsibility to establish calibration interval based on their usage. Which is in conflict with the fact the manufacture of the product has the largest database of information available to establish calibration intervals.

I think the dilemma is “instrument” vs. “instruments”. If I have one instrument and want to adjust cal interval the only method that will work is an algorithmic method. “John Nabors” has a good one, but once the average is found the cycle of 4 good cals and two OOT cals is a Reliability of 66.66%. Which means it is returned out of cal 33.33% of the time. There just isn’t enough data to work with.

I have a large group of instruments and if you refer back to plot03 last graph you will note the average is 125 weeks the dogs are 69 weeks and the gems are 180 weeks. I don’t currently know if this is random with increasing uncertainty over time or if there really are gem instruments that will go for 180 weeks. If it’s random “John Nabors” algorithmic will be constantly searching a 69 to 180 week window. If there are really gems in the group and “Jim Nabors” owns it algorithmic may save some money.

The variables calibration interval analysis does work, but it’s not as good as RP-1 S2. Set to 99% confidence bands you can get an estimated threshold as to when a variable will go out of tolerance. It suffers from the same resolution issue that occurs with the calculation of Cpk. If the resolution of the measurement is such that it can’t estimate a standard deviation then its prediction is questionable. It also suffers greatly from effects of outliers which can affect the prediction by as much as four months. The residual sum of squares is used to select the best order fit and in many cases they just look wrong, so many liberties were taken with outliers and order of fit. It also suffers from effects of non normal distributions. But it does a good job at filtering MTE random fails to get at the underlying trend over time, and you can make a spreadsheet listing variables with weeks to threshold of OOT, which does provide a feel for which variables will go OOT beyond the current calibration interval.

Its report time and the next question: In a calibration interval analysis should broken equipment returned for repair by a customer be considered as a failed incoming calibration? There seems to be some confusion between reliability of broken equipment and reliability of calibration OOT. I tend to think of them as separate issues, is there an industry standard?

Dear Distinguished people of the forum,

I am a make shift calibration engineer(from the instrument service and repair dept) and i am still green to all these calibration terms,definations and etc.
I was redesignated to the RF/Microwave calibration team and find myself facing the ardous test of being being audited by the 17025 accreditation body.
We are scheduled to be audited by the 17025 body this July.
We are accreditated in the AC/DC measurement and now, we are trying for the RF/Microwave measurement as well.
It is an extension to our capabilities...does that mean that i can use the same quality manual and laboratory manual fo the RF calibration as well?
I am trying to get used to our standards in the RF/Microwave lab but i am blur when it comes to the MU calculation and quality manuals that go along with it.
I require some inputs on coming out with the MU calculations for power measurements..levels in dBm and in ppm as well.
If you have any materials pertaining any RF/Microwave MU budgets, do let me know?
Appreciate the help.

Rgds
Ruben

Depending on the frequency and what type of conduit you use, there could be many factors.....some of the common ones are:

Uncertainty on your equipment, as reported by the accredited lab that cal'd them

VSWR (Voltage Standing Wave Ratio), or how much reflection you have

RH (Relative Humidity), below 20% you may have influences from static charges in the ambient environment

Surface loss of the conduit

There will be other factors also.....but these are common ones.

Hope this helps.

Hershal

I like the work stated. You run the risk with dimensional contact measuring devices of missing special causes - such as dropped gages - if you run all the way to the period that approaches the tolerance within (1-a) confidence level. If the period is 5 years, and you drop the gage in 30 days, you have 4 years and 11 months of suspect product. Nasty risk. The linear regression is a good approach for dimensional contact measuring devices. It works because the fundamental variation should be wear - a uniform distribution. The measurement error about that wear rate will be a normal distribution - but if the calibration standard and technique is appropriate, it should be statistically insignificant to the wear rate for the technique to be accurate.

You run the risk with dimensional contact measuring devices of missing special causes - such as dropped gages - if you run all the way to the period that approaches the tolerance within (1-a) confidence level. If the period is 5 years, and you drop the gage in 30 days, you have 4 years and 11 months of suspect product. Nasty risk.

If your company adjusts calibration intervals to reduce risk caused by dropped gages it means that you are always producing suspect product, because you can not predict when the gage is dropped. No risk here, you’re 100% guaranteed to always ship suspect product, sometimes more sometimes less. Put everything on a 1 day cal interval to avoid the risk and if it’s dropped an hour after calibration you’ve still shipped 23 hours of suspect product.

The calibration interval is for “typical equipment usage”.
It is ultimately the “user’s” responsibility to know if their instrument is out of tolerance.

As a manufacturer we sell instruments to customers, we are not responsible for customer dropped product and run no risk. Same for inside the company, production floor personnel have the responsibility to know their instrument is out of tolerance, not Metrology. Metrology runs no risk.

I could fix your implied scenario by adding a second final inspection to the process performed with new personnel and a separate instrument. But this fix does not address the possibility that there would then be two teams playing catch over the noon hour with the gages. Or sabotage in which case both gages will be abused in less than 30 days. This fix is called Tampering - taking action based on the belief that a common cause is a special cause. The tendancy to take action, often leads to action without reason which causes more problems than it fixes. Dr. Deming stated that most variation (97% plus) was common cause variation not due to special causes. Tampering can also be considered a form of variation.

Our “typical equipment usage” does not include dropped. One might want to consider: incompetent employee, insufficient training, inappropriate procedures, absence of clearly defined standing operating procedures, inexperience, and sabotage. All are classified as “Common Cause” not special cause (wikipedia).

I am very pleased to be able to say. I have witnessed our production personnel delivering gages to Metrology to have them “checked”, because they were dropped.

Managing common cause risk is a management function not a calibration interval function.

I am glad you agree the risk exists and needs to be managed. I never said it had to be managed by the calibration period - but I will say if it is not properly managed, then your calibration period will be useless - no matter how long it is. :cool:

My misinterpretation. Agreed. :agree1: