Announcement

Collapse
No announcement yet.

new version of Caliper Accuracy article available

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • tdmidget
    replied
    Thank you JC your explanation is much more eloquent and precise than mine might ever be.
    Evan , in my conversation with the Starrett engineer he made it clear that there is not and never has been such a thing. If there was , would Starrett have parts sitting around for 24 hours before measurement? How , I wonder would a caliper know what the temperature of the workpiece was? Even if it compensated for it's own temperature, (which means , I guess the difference from standard temperature,) how would it compensate for the workpiece temperature? And why would anyone develope such a thing when all you have to do is have everything at the same temperature? Again, the man said that there is not and never has been such a device. He did say that at one time there was a primitive electronic height gage that had a circuit to allow for heat from the electronics. It had no correlation to part or ambient temperature and was not successful.

    Leave a comment:


  • tdmidget
    replied
    Thank you JC your explanation is much more eloquent and precise than Mine might ever be.
    Evan , in my conversation with the Starrett engineer he made it clear that there is not and never has been such a thing. If there was , would Starrett have parts sitting around for 24 hours before measurement? How , I wonder would a caliper know what the temperature of the workpiece was? Even if it compensated for it's own temperature, (which means , I guess the difference from standard temperature,) how would it compensate for the workpiece temperature? And why would anyone develope such a thing when all you have to do is have everything at the same temperature? Again, the man said that there is not and never has been such a device. He did say that at one time there was a primitive electronic height gage that had a circuit to allow for heat from the electronics. It had no correlation to part or ambient temperature and was not successful.

    Leave a comment:


  • kap pullen
    replied
    Originally posted by Evan
    A couple of years ago there was a 72" Starrett Master vernier for sale on E-bay. It had a temperature compensation scale as shown in the photographs. As for the Starrett engineer, he probably wasn't born yet when it was made..
    What is the point of a temperature compensated caliper. All materials expand and contract togather, but at different rates.

    The best guess of size you can make is if part/instrument are the same temp unless calculations have made as respect to a certain situation.

    Even Invar has some thermal movement, It is utalized to point telecopes in deep space with solar powered heating coils.

    If you measure a part that is 100 degrees with a compensated caliper, you will get an exagerated error. The part is in an expanded condition, and the caliper is not, producing more of an error than exhisted with a warm caliper.

    My foreman many years claimed we should measure parts with a caliper
    at 68 Degrees, and the part at 140. This was big stuff measured with a 72 inch Master Starret Caliper.

    The parts hit inspection .03 undersize. The head inspector set him in his place.

    Later we stored the bore gages in pockets in the coolant tanks of cylindrical grinders to mantain reletive temperatures of parts, and gauges.
    This was to hold tenths tolerance.

    Today, some machining centers have water cooled ball screws supposed to mantain size. This will work when the parts do not absorb much heat from the machining, but not on heavy machining operations.

    I, in 38 years have not seen a compensated caliper but there may be one somewhere.

    I could imagine a caliper that can compensate for different materials and temperatures these days could be built, but not before that Starret Engineer was born.

    Kap

    Leave a comment:


  • JCHannum
    replied
    I am still unconvinced as to the existance of a temperature compensation scale on a Starrett Master Vernier caliper. I have seen dozens, and possessed more than a few at one time or another, and none had any such device. It definitely does not appear in any of the catalogs. I do not consider someone's memory of something that may or may not have been correctly described on eBay several years ago as acceptable proof.

    The Starrett Master Vernier caliper 123 Series was introduced in the early 60's. Prior to that they sold only the 122 Series, which had the short, 25 division vernier.

    While the standard of length might be related to some imaginary distance traveled by a beam of light in some unmeasureable amount of time in an unachievable atmosphere, that has nothing to do with taking a measurement.

    All precision measurements are based on standard temperature conditions. A measurement of a part at 120*F will result in a different dimension than the same part measured at 0*F. If the measuring device did not change with temperature, it would indicate two different dimensions. If a part were mis-machined, all that would be needed to bring it into tolerance would be to heat or cool it. If the measuring device were hot and the object cold, or vice-versa, the error would be compounded. This is why all dimensions ultimately must relate to some standard temperature, with the measuring instrument and the part normalized at that temperature.

    There are several different protocols used by the various manufacturers in their digital instruments. There is no reason to believe that they would all behave the same.

    I have no idea who Nazakawa is, perhaps he was selling that caliper on eBay. He is referring to measurement, we are referring here to calibration. The standard method of calibration will rely on the use of the next smallest increment. In the case of metric and decimal measurement, that is a factor of ten.

    Leave a comment:


  • ckelloug
    replied
    td,

    Would you have been satisfied if Rick had used the word deviation rather than error for this experiment? You are correct when you say that it isn't an error by the dictionary definition of error but I see this point more as non-standard usage of the word error to mean deviation than a real error in the paper.

    Evan,

    One of the things that they taught us back in engineering school is that in every digital system there is an analogue system just waiting to escape. For something as simple as a caliper, you can model the entire thing as I did in my earlier post as an analog system with quantization error. Likewise, when you apply frequency domain techniques to a more complicated system, if you need the complete behavior model you do the analog analysis and then multiply by a pulse train to get the digital effect.

    I am not familiar with Nakazawa’s principles of measurement however I would contend based on my above observation that it is just as important to apply them in the analog world as the digital one. It's probably even more important because you are forced to look upon the already complicated analog system through the tiny window afforded you by digital sampling.

    As for temp compensated calipers, A touch of googling gives 0.00000645in/in/degF as the thermal expansion. For a 72 inch caliper with a 40 degree temperature difference from the calibration temp, this is 18 thou. While they do make the 72 inchers at Starrett, this isn't exactly the most common product in Starrett's line. Thermal compensation is probably a pretty minor problem on smaller instruments especially on the 6 inchers. It's a purely academic exercise but you could temperature compensate the HF caliper I posted the data for by modifying the linear coefficient in the modeling equation I posted.

    At any rate, The HF calipers are a pretty good buy when they're on sale and while they may not last forever, in a place like a school lab where the expected life of calipers is only slightly greater than whiteboard pens, they're a bargain. Not as pretty as good Starrett ones though

    --Cameron

    Leave a comment:


  • Evan
    replied
    I have never seen a Starrett Master Vernier caliper with a temperature compensation scale, or any other manufacturer's vernier caliper with a similar device.
    A couple of years ago there was a 72" Starrett Master vernier for sale on E-bay. It had a temperature compensation scale as shown in the photographs. As for the Starrett engineer, he probably wasn't born yet when it was made.

    I also see no mention of calipers being available in anything but tool steel or hardened stainless. For a prohibitive cost, you could conceivably have one made of any material desired, but to what end. Unless the material being measured were of the same material, temperature changes would result in compounding any error. As all measurements are ultimately referenced to standard temperature, the most precision measurements are made under those conditions.
    The standard of length is not referenced to a standard temperature and changes in temperature do not affect the standard. The standard of length for all measuring systems is the distance travelled by light in absolute vacuum in 1/299,792,458 of a second, the metre.

    Calipers, or any other measuring instrument that did not change with temperature would give an accurate measurement of the dimension being measured. Temperature does matter and having a measuring instrument that doesn't change with temperature does not compound the error.

    It is not surprising to see that the HF caliper checks out so closely, a Starrett, Mitutoyo or other will do equally as well I am sure. You would hope that the specifications would represent the worst case, and that the product would be better than that. As manufacturing methods become refined, accuracy and repeatability will increase.
    It's not surprising because they all likely use the same integrated circuits.

    [edit]

    Note also that the "ten times" principle as has been stated here is stated incorrectly. The principle is:

    Nakazawa’s principles of measurement

    First Principle — When measuring the dimensions of an object,
    one must know its temperature.
    Second Principle — Avoid the application of force. (Non-contact
    methods of measurement are preferable to contact methods.)
    Third Principle — The precision of the measuring instrument must
    be five or ten times higher than the expected precision of
    the measured object.

    Fourth Principle — Be aware of the behavior of the instrument. (!)

    http://lmas.berkeley.edu/me220/Lecture-6-07.pdf
    The ten times principle states that the device doing the measuring must be ten times more accurate than the artifact being measured. This is not the same as a calibration standard being used to calibrate a device. This is especially true of digital systems that exhibit entirely different types of errors than analog systems.
    Last edited by Evan; 06-06-2007, 03:58 PM.

    Leave a comment:


  • tdmidget
    replied
    Rick , my problem is that you froze the part, measured it , and called the difference "error". It is not. When you froze the part it became smaller and your new number reflected that. If you ground .0005 off and measured it would that be an error? Of course not. In either case the ppart as measured WAS .0005 smaller.
    Evan, I just got off the Phone with an engineer from Starrett who I'll be nice enough to leave his name out of the post. I asked him to settle this arguement. Has Starrett ever made a caliper with temperature compensation, is it possible? His reply- "What has that guy been smoking?

    Leave a comment:


  • DR
    replied
    A quick random scan of postings in this thread indicates to me the HF calipers may be a good buy based on their accuracy.

    I have a customer who bought a bunch of $18 calipers at the local Schucks/Cragen auto parts store.

    Comparing those to my Mitutoyo, it appears the cheapos are a direct copy. I assume the HF version is the same/similar to the auto parts store version.

    Fast forward a year or so from my customer's purchase. About half their calipers are no longer working, not to mention their's eat batteries.

    So, for the customer were the cheapos a good deal? I don't know......

    Leave a comment:


  • JCHannum
    replied
    I have never seen a Starrett Master Vernier caliper with a temperature compensation scale, or any other manufacturer's vernier caliper with a similar device. None of the catalogs I have from several manufacturers of precision measuring instruments spanning some 70 years or more show such a scale as being available. I would be very interested in seeing one, and how it might be used if it were to exist. Regardless of the length of a caliper, the vernier length and divisions of the beam must remain constant.

    I also see no mention of calipers being available in anything but tool steel or hardened stainless. For a prohibitive cost, you could conceivably have one made of any material desired, but to what end. Unless the material being measured were of the same material, temperature changes would result in compounding any error. As all measurements are ultimately referenced to standard temperature, the most precision measurements are made under those conditions.

    While the electronics might have temperature compensation built in, it is only to maintain the accuracy of the electronics, and has no effect on the dimensional accuracy of the caliper's mechanics. Without such compensation, the electronic calipers would have an even higher potential for inaccuracy.

    In metrology, the rule of ten must be used, at least when using metric or decimal measurement. You cannot determine the accuracy of a linear measuring device unless you compare it to a device of the next level of accuracy.

    While there may be some electronic instruments with built in self checking loops, there are none for measuring instruments that I am aware of. The frequency counter described, is still comparing two different measurements, even though they are contained in the same instrument. For reliable, take it to the bank accuracy, a measuring instrument such as a caliper must be calibrated to known, traceable, standards.

    It is not surprising to see that the HF caliper checks out so closely, a Starrett, Mitutoyo or other will do equally as well I am sure. You would hope that the specifications would represent the worst case, and that the product would be better than that. As manufacturing methods become refined, accuracy and repeatability will increase.

    Leave a comment:


  • oldtiffie
    replied
    Very well done all round

    My sincere thanks and congratulations to all - all - each and every one of you - it is marvelous- and I, being a confirmed sceptic and cynic, am not "blown away" very often at all.

    And thank you Evan, as you've seen, this has been a very "lively" and fascinating thread. Thanks for your measured and considered expert input as well.

    Metrology in all its derivatives and applications has been a fascination to me for a very long time.

    I can remember grinding and lapping and setting-up to these standards. The best/worst impression I got was seeing what I thought to be a very straight, sharp thread cutting tool that I had made when it was put up on the comparator projector (ie x40 and upward). I was mortified when I saw the result. I decided that I could and would do better - which I did and my pride in what I had achieved was surpassed by the Tradesman who set me the task. He felt that he was doing his job as well.

    Anyway.

    This thread is doing a magnificent job of matching the theory and practice/practicable at all levels so that we all have a better appreciation of what "goes on" "up there" and "down on the shop floor".

    Again, many thanks.

    And Rick, do "press on" and gather all the threads together as a lot of people who were "at arms length" now feel that they have "involvement" and "ownership" in it and an interest in seeing it published in due course.

    I think that many who might have appeared at best sceptical are now behind you and supporting you.

    Let's see what the next edition is like.

    Leave a comment:


  • Evan
    replied
    I haven't looked at this thread until this evening. Dang.

    First, tdmidget, there are calipers that provide temperature compensation. The Starrett line of Master Precision verniers in the larger sizes have or at least had a temperature compensation scale on the beam that was/is used to correct the reading.

    Second, some high quality instruments have been made from invar, an alloy with a nearly zero CLE with temperature. No compensation is required.

    Third, many if not most of the ICs in the electronic digital calipers have inbuilt temperature compensation that adjusts the output value based on the temperature of the chip.

    One thing I haven't seen clearly mentioned here is that the display resolution of the readout is the absolute limiting factor of the device. A digital readout has a plus/minus 1 count on the least significant digit inherent error. That is why the calipers that display to .0005 are rated as having .001 "accuracy".

    It is also important to understand the difference between accuracy and resolution. Accuracy is how close the story is to the truth. Resolution is how many details the story contains.

    It is entirely possible for a measuring instrument to be self referenced for calibration. I have a good quality dual timebase frequency counter that calibrates one time base by measuring the other. This is viable because it is vanishingly unlikely that the two time bases would each develop exactly offsetting errors and so appear to give a correct result when it is not. I have also calibrated it to a very traceable standard, the triple atomic clock stack at the local LORAN C station. The counter was correct to one count.

    The old rule of the standard being at least 10 times more accurate than the device being calibrated is meaningless in the digital world. You can't tell if the display is almost about to change to the next value or not. The Nyquist–Shannon sampling theorem says that the digital sample rate must exceed the data frequency by at minimum two times to avoid aliasing. The inverse is directly applicable to the space domain instead of the time domain and tells us the reference artifact fails to provide any further useful information when it's "truth" exceeds one half the least significant display resolution.

    The accuracy of all measuring instruments is properly defined by their uncertainty, or the error band. In a digital instrument it cannot be better than the display resolution plus/minus one count. This also applies to repeatability since if the display is extremely close to changing on a measurement it may only take a difference of .0000001 in truth to change the display by one count on the next measurement.

    Leave a comment:


  • ckelloug
    replied
    NIST tracable Hard data on HF caliper Accuracy

    Well guys,

    Putting aside my early aggrevation about style of discourse issues, I've actually put my gage blocks to use. The results are as follows.

    For the purposes of metrology, the gage blocks are B89-0 Starrett Webber steel blocks etched ZHH030 Calibrated 1/24/07 with calibration certificate 07-47943A-F from Starrett Webber Gage traceable to NIST via NVLAP partnership under lab code 200038-0. Room temperature was 72 degrees. The gage blocks had acclimated at that temperature for several weeks. The caliper had acclimated for several hours with the box open. Cotton gloves were worn during the handling of the blocks and the caliper to minimize contamination of the blocks and thermal transfer from the hand.

    The first column is the Gage block in use, the second is 5 measured values from the caliper for that block. The caliper was brought to the jaws closed position and zeroed for each measurement. I was blown away on the accuracy of the HF caliper. I tried graphing the raw measurements but the line is so flat that it shows nothing. I graphed the error but this board doesn't let you upload pdf files. I can e-mail it if anyone is interested.

    Finally, I got a fourth order regression fit of the data representing the caliper measurement vs. the absolute length of the item in the jaws which came out to

    M = -0.00017683 + 0.99985 * L + 0.0001182 * L^2 - 0.00011015 * L^3 + 2.3791e-05 * L^4

    Matching terms with the model I published earlier, epsilon is -.00017683, dl/dL is .99985 and d^2l/dL^2 =.0001182. This implies that the scale has about -.0002 offset built in and that it is about 1.5 tenths per inch shorter than ideal . The higher order terms are important as well and they indicate that the error in the scale is a bit nonlinear but that some of the error from low order terms cancels out error from higher order terms in some places. A touch of the error could be measurement technique on my part but it seems at least in part actual error.

    At any rate, here are the results:
    Block Caliper
    .0500 .050
    .0500 .050
    .0500 .0495
    .0500 .050
    .0500 .050
    .1000 .100
    .1000 .100
    .1000 .100
    .1000 .1005
    .1000 .100
    .1001 .100
    .1001 .100
    .1001 .100
    .1001 .100
    .1001 .100
    .1003 .1005
    .1003 .1005
    .1003 .100
    .1003 .1005
    .1003 .100
    .1005 .1005
    .1005 .1005
    .1005 .100
    .1005 .100
    .1005 .1005
    .1006 .1005
    .1006 .1005
    .1006 .1005
    .1006 .100
    .1006 .100
    .1009 .1005
    .1009 .1005
    .1009 .101
    .1009 .101
    .1009 .1005
    .1010 .101
    .1010 .1005
    .1010 .101
    .1010 .101
    .1010 .101
    .1020 .1015
    .1020 .1015
    .1020 .1015
    .1020 .1015
    .1020 .102
    .1030 .103
    .1030 .1025
    .1030 .1025
    .1030 .1025
    .1030 .1015
    .1030 .1025
    .1030 .103
    .5000 .500
    .5000 .4995
    .5000 .500
    .5000 .500
    .5000 .4995
    1.0000 1.000
    1.0000 1.000
    1.0000 .9995
    1.0000 .9995
    1.0000 .9995
    2.0000 2.0005
    2.0000 1.9995
    2.0000 1.9985
    2.0000 1.9995
    2.0000 1.9995
    3.0000 2.9985
    3.0000 3.000
    3.0000 2.998
    3.0000 3.0005
    3.0000 3.000
    4.0000 3.9995
    4.0000 4.000
    4.0000 4.000
    4.0000 4.0005
    4.0000 4.0005
    4.0000 4.0005


    In conclusion, the caliper was never more than .002 off and a bit of that could be technique. My only observation is that the .0005 readout on the caliper is actually worse than useless as it is almost purely noise and the since the caliper seems to read just a touch low, that digit gives a false sense of accuracy. It may be that once I have instrument oil and can wring blocks to measure at 5 and 6 inches that the error will go up more.

    Also hope I didn't get too annoyed on this thread. The comment I made about being right being the main factor in being right was only an indication that I support hard data over soft data even if it proves me wrong. I bailed from the thread initially because I didn't want to hang around to feed the trolls. Now that I know td isn't a troll I'll just say I personally favor polite discourse over the vaguely ad hominem stuff. Most people on this board are so helpful that it's a shame to be mean to any of them. Sorry td if I got a bit upset.

    Regards to all,

    Cameron

    Leave a comment:


  • rgsparber
    replied
    Originally posted by dp
    If you have an inside and an outside measurement standard that are exactly the same size, hysteresis will provide two different measurement readings because you are coming at the size from different directions (compressing the caliper, expanding the caliper). Part of that is slop in the tool, and part of it is the resolution of the digitizer and where, in the digitizer stream, the zero point is taken on what you call a sensor step.

    The hysteresis reported may be different at 1" than at 2.5" as well, owing to it being a simple mechanical device.

    It may be incredibly small, or masked by the 10/1 ration between displayed accuracy and digitizer resolution.
    dp,

    Ah, now I understand. I'll go check it out.

    Thanks,

    Leave a comment:


  • dp
    replied
    Originally posted by rgsparber
    DP,

    I don't recall ever seeing hysteresis in my digital caliper. Can you suggest an experiment that will flush it out?

    A single "sensor step" on these calipers is about 0.05 thou. that is 10 times smaller than the smallest displayed value.

    Rick
    If you have an inside and an outside measurement standard that are exactly the same size, hysteresis will provide two different measurement readings because you are coming at the size from different directions (compressing the caliper, expanding the caliper). Part of that is slop in the tool, and part of it is the resolution of the digitizer and where, in the digitizer stream, the zero point is taken on what you call a sensor step.

    The hysteresis reported may be different at 1" than at 2.5" as well, owing to it being a simple mechanical device.

    It may be incredibly small, or masked by the 10/1 ration between displayed accuracy and digitizer resolution.

    Leave a comment:


  • rgsparber
    replied
    Originally posted by oldtiffie
    I must say I am both very please and relieved that Rick, tdmidget and ckelloug are all (still) in there and getting on with things (and each other).

    I am enjoying the discussions of issues as there really is some "good stuff" in there.

    Also there is little or no "because I said so - end of story" crap either and items posted can be expected to be discussed (dissected even) and a compromise reached.

    This is far too good a thread and the people on it are far too good to be "wasted" as well.

    I've seen a lot of very good and highly relevant points and discussions here.

    I've had to abandon some thoughts and practices I had and modify others as this thread and its discussion progressed.

    This thread seems to be quite balanced.

    Rick, how are you feeling about all this? There should be food for thought as regards methods, validation and relevance in this - nothing drastic perhaps - but a bit of "tweaking"?

    Too all and sundry - many thanks and keep it up.
    As far as I'm concerned, the process worked correctly. Sure there are people that shout and don't listen, but that does not mean they have nothing important to say.

    I can't say I enjoy personal attacks but I certainly DO enjoy learning and sharing what I learned.

    Leave a comment:

Working...
X