new version of Caliper Accuracy article available

[rgsparber]

I have learned a lot since version 1 came out. For example, the name of this article is now Accuracy and Repeatability, not Accuracy and Precision.

There is now a section that discusses the effects of temperature on the Harbor Freight caliper.

If you are interested, see

http://rick.sparber.org/Articles/PAC/PAC2.pdf

As always, comments are welcome and do have an effect on future versions.

Rick Sparber
[email protected]
web site: http://rick.sparber.org

[tdmidget]

So, Rick I'd be interested to know what qualifies you to pen this article? For the most part it shows that you don't know much about measuring or the instruments to do so.
1. You go on at lengths about possible errors of .001 or less in an instrument that no shop working to a serious level of precision would allow for tolerances of .002 or less, and therefore inconsequential. No slide caliper should be used for measurements closer than this as that is pretty much the limit of precision for the instrument.
2. You cool the instrument, measure with it, then cool the part and measure it. One of the most basic principles of metrology is to have the instrument and the workpiece at the same temperature. All this shows is that metal contracts when cold, no error is shown. This demontrated by the fact that you claim the same "error" each time. And again beyond the capabilities of the instrument. It goes without saying that a steel caliper measuring a steel part at the same temperature has no error since the coefficient of expansion is the same for both.
3. You used as a standard 1-2-3 blocks which are normally .0002 oversize unless you are using Starrett or Herman Schmidt blocks. If you could afford 1-2-3 blocks approaching $300.00 each you probably would not be testing a Horror Freight caliper. In any case 1-2-3 blocks are NOT gage blocks and should never be used as such.
I wonder if you were a manufacturer of anything, calipers, wdgets, or what have you, would want someone "testing" your product with your level of knowledge? Should anyone make a buying decision based on this?

[ckelloug]

tdmiget,

In my read of Rick's article, I see that he did some experiments trying to determine the ultimate accuracy and repeatability of the harbor freight calipers. I'm by no means one of the board's old timers but I know it is of great interest here to know whether inexpensive imported instruments are good enough for home shop use.

I do not know Rick personally but I know he said somewhere that he is a retired electrical engineer. While few here are metrologists, many are quite interested in the ultimate accuracy of instruments. I don't think anybody here would try to use calipers for a critical measurement.

I don't think Rick's point in this was either to advocate the use of calipers beyond reasonable accuracy, or to advocate using instruments not at thermal equilibrium with their surroundings. What he did show is that the accuracy specification for HF calipers is simplified somewhat from the actual performance and that they are not temperature compensated. The points are of academic interest mainly and demonstrate that when used for caliper grade measuring applications that the HF calipers do OK. This mini-paper increases the community's understanding of calipers and provided that silly inferences surely not intended by the author aren't drawn from it, it looks useful.

--Cameron

[speedsport]

OK tdmidget, don't hold back, tell us what you really think.

[rgsparber]

So, Rick I'd be interested to know what qualifies you to pen this article? For the most part it shows that you don't know much about measuring or the instruments to do so.
1. You go on at lengths about possible errors of .001 or less in an instrument that no shop working to a serious level of precision would allow for tolerances of .002 or less, and therefore inconsequential. No slide caliper should be used for measurements closer than this as that is pretty much the limit of precision for the instrument.
2. You cool the instrument, measure with it, then cool the part and measure it. One of the most basic principles of metrology is to have the instrument and the workpiece at the same temperature. All this shows is that metal contracts when cold, no error is shown. This demontrated by the fact that you claim the same "error" each time. And again beyond the capabilities of the instrument. It goes without saying that a steel caliper measuring a steel part at the same temperature has no error since the coefficient of expansion is the same for both.
3. You used as a standard 1-2-3 blocks which are normally .0002 oversize unless you are using Starrett or Herman Schmidt blocks. If you could afford 1-2-3 blocks approaching $300.00 each you probably would not be testing a Horror Freight caliper. In any case 1-2-3 blocks are NOT gage blocks and should never be used as such.
I wonder if you were a manufacturer of anything, calipers, wdgets, or what have you, would want someone "testing" your product with your level of knowledge? Should anyone make a buying decision based on this?

As with everything else on the web, reader beware. If you consider my rantings to be crap, by all means say so in as much detail as possible on the BBS so that others may hear a balanced story. I am a hobbyist that really enjoys finding things out. My conclusions are rarely up to scientific rigor but I do try not to delude myself or mislead others. I strongly believe in the free flow of ideas to flush out the truth. I'm also not afraid to say when I'm wrong.

In another thread I was trying to sort out people that demand accuracy versus those that only need repeatability. It sounds like you are in the accuracy camp. As a simple hobbyist, short term repeatability is almost always good enough for me to get my projects done.

Rick

[tdmidget]

CKelloug read again. If Rick is an electrical engineer, would he evaluate the accuracy of a volt-ohm meter by comparing it to itself? Would he take a meter graduated in ohms and attempt to measure .2 ohms? These very errors were committed in his "tests". You absolutely cannot use an instrument to evaluate itself. When he said that the "error" due to temperature was .0005, was it .0003 or .0007? How would you know with an instrument graduated in .0005 increments? You say it showed it was not "temperature compensated". Let's , for simplicity, say that the reading of 5.9995 on the cold part was actually that, 5.9995. The part shrank .0005 so that would mean that the instrument was dead nuts on size. How can you "temperature compensate" when the two are at different temperatures?
My point is that the whole exercise was meaningless. I am not here to defend Horror Freight. My point is that if you want to pass yourself off as "testing " something then it must be checked by instruments of known accuracy and greater precision than that tested. To freeze the workpiece and then measure it is like cutting an inch off a rope and saying "now there is a one inch error". Of course it was shorter if you froze it. Nothing proved except as I said, someone knows little of measuring and the tools therefor.

[ckelloug]

tdmidget,

In the case of temperature compensation, I had the impression that he was doing a qualitative test to check for temperature compensation.

His hypothesis was that a temperature compensated 6" caliper should read about the same regardless of temperature while a non-compensated caliper should read a bit under when it was cooled and a bit over when it was warmed. Assuming the blocks have the same coefficient of thermal expansion as the caliper (perhaps unwarranted) then if the reading is markedly different for cold caliper room temperature block than it is for cold block room temperature caliper then the caliper isn't compensated. This is quite clever as a qualitative test as long as the coefficients of expansion match between the blocks and the calipers. As a metrology practice, this is not good practice, but it's excellent as a quick research tool into the working of the instrument.

Reviewing the section on accuracy however, I think some improvements could be made. Since the precision of the caliper is only .01mm, accuracy numbers below this scale are rather irrelevant. I think that the linear model of error is probably good to first order as it accounts for linear deformation of the entire scale such as temperature. It may be true however that there are non-linearities due to localized manufacturing defects in the combs inside the caliper.

The stepped error tolerances on the specification are almost certainly worst case values that apply to all instruments coming off the assembly line, not any particular instrument. I would assume that they are more of a convenience to the manufacturer in terms of only needing to check accuracy in a few places and guarantee that the accuracy on the remainder of the bar is less than the checked cases. The stepped tolerances also give the manufacturer wiggle room if some process parameter is off when the caliper is made

In short, we know from a study of taylor series that all sorts of functions can be constructed that pass through the points given in the accuracy data. One must remember of course that the derived error equation is an inequality such that the error is less than or equal to the specified amount for the given range. I would say that from effects like thermal expansion and imperfect manufacture that the first term of the taylor series is the linear one like Rick proposes. Without a bunch of measurements against a wide variety of calibrated gage blocks, there is no way to know what the other terms are.

In short, the accuracy of the caliper doesn't matter that much on small measurements because it is dominated by the repeatability. Since the repeatability is only .0005, no value returned from the caliper by any means can be trusted to be more accurate than this.

I'd model the system like this: Let the length of a perfect gage block measured by the caliper be L. The caliper can be modeled as providing you with measured length M=epsilon + L*dl/dL + L^2/2 * d^2l/dL^2+... +X rounded to the nearest .0005. where dl/dL is the ratio of noise free caliper length/gage length , epsilon is an arbitrary offset error which may be zero and X is a gaussian random variable with mean 0 and 2*sigma=.0005 and the ellipsis is higher order terms of the taylor series relating the hypothetical noise free measurement l to the real measurement L. I don't see a physical reason for the stepped accuracy in an actual caliper.

The important thing to remember is that the above model is what the caliper is returning to you, not length directly. A quick analysis of this model shows that very near zero movement, most of the value of M is the repeatability. As the values of L get larger, the ratio between the noise free measurement and the actual gage begins to dominate the value of X. Because X is a random variable however, the value for M will never have a smaller standard deviation than X.

If this isn't incoherent yet, it will be shortly. Off to bed.

[oldtiffie]

A superb response and support of a member under unjustified siege

tdmidget,

In the case of temperature compensation, I had the impression that he was doing a qualitative test to check for temperature compensation.

His hypothesis was that a temperature compensated 6" caliper should read about the same regardless of temperature while a non-compensated caliper should read a bit under when it was cooled and a bit over when it was warmed. Assuming the blocks have the same coefficient of thermal expansion as the caliper (perhaps unwarranted) then if the reading is markedly different for cold caliper room temperature block than it is for cold block room temperature caliper then the caliper isn't compensated. This is quite clever as a qualitative test as long as the coefficients of expansion match between the blocks and the calipers. As a metrology practice, this is not good practice, but it's excellent as a quick research tool into the working of the instrument.
........................................
..........................................
.........................................


If this isn't incoherent yet, it will be shortly. Off to bed.

Thank you ckelloug for such a comprehensive report and reply as well as support of Rick Barber who was under unwarranted siege.

I for one thoroughly enjoyed the substance of your report as it was very enlightening on many fronts - particularly as it applied to the "on the job in the real world" of using calipers - in this case digital calipers.

You must have put in a lot of your very high level expertise and your own time into this.

Thank you.

Your closing remark about going to bed could/should be well undertaken by some on this thread - provided that it is preceded by a long cold shower.

[IOWOLF]

Has anyone heard of room temperature?

At Vickers we worked in a climate controlled building,and yes we worked in 0.0001 (tenths).

I never could figure out the guys who did Instrument calibrations in a air cond. (65*) room then give them to the guys in the shop where temps reach 100* or more.

[tdmidget]

A "temperature compensated caliper"? I am a journeyman machinist and I never heard of such a thing. I have looked in the Starrett and B&S catalogues and they show no such thing. Why, when the potential error is beyond the resolution of the instrument would anyone want or make such? How about give those of us who do this for living a part number in case we need one? I don't believe that it exists.

[ckelloug]

Thanks to oldtiffie and an Offer of help to Rick

oldtiffie,

Thanks for the support. I've enjoyed your kind, insightful and level headed posts all over this forum. I'm sticking up for Rick because of a bad experience I had being flamed on the Polymer Concrete Frame thread on CNCZone despite having put in what probably amounts to a man week of time performing calculations and research. My response was very similar to Rick's response.

I suspect Rick's interest in this subject has to do with using these calipers in the Shumatech DRO.

My authentic Bridgy and chinese Birmingham lathe should be arriving Wednesday and I don't have power to the shop yet so I can't say I don't have too much to do. I do however happen to have a brand new set of Starrett Webber gage blocks grade B89 0 with calibration certificates from January of this year that I got on e-bay. I also have a kitchen table maintained at 72 Farenheit.

Rick,

If you can loan me the HF caliper you used in your paper, I'll do a calibration study of it and you can tweak your error model. PM me if this is an offer you can't refuse.

--Cameron

[lazlo]

If you can loan me the HF caliper you used in your paper, I'll do a calibration study of it and you can tweak your error model.

They're on sale this weekend for $9. Not even worth the shipping -- just go pick one up!

[ckelloug]

Lazlo,

Thanks. I haven't been tracking Harbor Freight lately and so I figured it might be a more expensive item. I'm awash in cheap Starrett verniers from e-bay so lord knows I don't need another caliper but why not. . .

tdmidget,

Remember, the primary reason that this is of interest is that the calipers are partially disassembled and used as part of the Shumatech homebrew DRO system. There may easily be no commerically made temperature compensated calipers made but that's not the point. The point is that these particular calipers are not temperature compensated and if this feature is desired in the Shumatech system then someone must add it. It would be quite easy to add digital temperature compensation to the Shumatch DRO with a simple IC temperature sensor.

Since these calipers are already being used to their absolute error limits as DRO scales, folks are interested in minimizing all of the error sources, especially on the longer models. Temperature compensating a 24 inch caliper cum DRO scale probably only corrects for one ULP (Unit in the Last Place) of error but since engineering time is free on the Shumatech system and the BOM change would cost only a few bucks, why deliberately avoid temperature compensation if you can do it.


I would like to know however what Starrett means by stabilized when they stabilized master bar on their 24 inch verniers if you can offer any insight.

For pure use of calipers in the shop, Rick's post doesn't make sense. Taken in the context of understanding the component for the purpose of engineering a device with it, it makes perfect sense that one would want to know the answers to the questions he asked. I'm not sure I agree with all of his answers but I do think he is asking the right questions.

At any rate, it's good to have another Journeyman machinist on the thread. People here do unconventional things here however so we must all be prepared to think out of the box. Many here including me also want advice about how to think in the box and you're more qualified to give that than I am.

P.S. In response to your comment about using a DVM to test itself, I used to build microchip testers and it was standard practice to test one part of the system against another part to provide a quick test as to whether anything was out of calibration. This of course was not sufficient to put it back into calibration but hundreds of thousands of hours of field data showed it was an excellent check to determine when it was time to connect up to the calibrator and recalibrate. Think 512 channel 100 Mhz ocilliscope with 512 signal generators running simultaneously and synchronized to picoseconds. . .

Regards to all,
Cameron

[Bill Pace]

Midget, Obviously you are a new member...one would hope that you lurked for some period of time to try and get a 'feel' for the forum before you joined. By the tone of the replies (read: flaming) on this post, that sure doesnt seem to be the case.

Rick is a relativly new member also, but he has certainly blended in with the group very well with an energetic and inquisitive nature to prod and poke and question on many subjects that have gathered many of of us readers in to follow his findings. He goes to great lengths to disclaim any scientificly controlled method to his quest for knowledge... just curiosity and the best way he can think of to do it, and then presenting to this group for more insight as to how to refine it, and, as he gathers this additional info he will go back and revise the article ---- just as in the beginning of this post.... "New Version of....."

It almost seems that you havent been following Ricks postings,--- from the start, he's presented his musings as just sharing his findings tinkering around in his workshop, not some definative do-all, end-all.

You might slack up on the flaming, this guy dont deserve it.

[tdmidget]

Bill There has been no mention of using this as a DRO until now. Still it does not matter, the accuracy is unchanged. The tests such as measuring a cold part which is now smaller and claiming an "error" due to the size change is NOT valid. As for temperature compensation, you already have it. If everything is at the same temperture and the caliper beam and the part are both steel, there is no error as the coefficient of expansion is the same.
When Starrett ( and others), say that a beam is "stabilized" that means stress relieved, most likely these days by cryogenic means. A warped beam would introduce error both in measuring and reading the vernier scale, maybe even impossible to read if enough curve made more than one graduation on the vernier line up. Shouldn't be a problem when the Horror Freight caliper cost $15.00 or less and is capable of doing the job. The jury may still be out on durability but I have one and if it screws up or I have a bad day I can throw it against the wall and get more 15 bucks of stress relief of my own.
As for it's use in a DRO, look at Mitutoyo's "quill mate". Same as their caliper, same accuracy and precision. Apparently Mitutoyo has already invented this wheel.