Re-sizing a gage block

[cecilstrange]

My old Webber set is missing its .100 wear blocks and, the .100 block, the .10005 block, and the .1001 block. I bought a Chinese set that was missing a lot of blocks for 99 cents. It did not have any of the blocks I needed but it had some which were close. I lapped the oversized Chinese blocks with successive sizes of diamond paste on a piece of milled cast iron. Darn if they didn't come out as close as I can measure with a calibrated mike, surface plate, and dial indicator with one micron resolution.

By the way, the rest of the Chinese blocks are off up to .0003". Some of the Webbers are off by .0002". Could that be growth from age? I am being careful to keep everything at 68 degrees F and very clean.

[The real Leigh]

Originally Posted by cecilstrange

Some of the Webbers are off by .0002".

Errors of that magnitude would lead me to question the accuracy of your measurement equipment or technique.

Where did you find a dial indicator with 1 micron resolution? I've never heard of one. In that range, the reading would change if you breath on it.

I have a Hamilton dial indicator that's calibrated in 10 millionths increments. I can use it with a 2" steel gauge as a thermometer in the office.

- Leigh

[PeteM]

Originally Posted by cecilstrange

I lapped the oversized Chinese blocks with successive sizes of diamond paste on a piece of milled cast iron.

Might also want to check how flat and parallel they are -- hand lapping to gage block tolerances is a tricky job. First thing to check is if they wring. Second thing is to get a quartz flat and monochromatic light and check the fringes. Let us know if you check the flatness and parallelism.

[bryan_machine]

Indicator?

Dorsey metrology:
http://www.dorseymetrology.com/produ...ndicators.html

Speced as 20 millionths or .5 micron, a couple of repair websites report that they really are.

And yes, combined with anything that expands much, they would in fact function as a thermometer.

But how does an indicator tell you a gage block is off? What reference do you compare it to? (Another gage block.... OK....)

[Forrest Addy]

You can make comparative measurements using a comparator stand and an LDVT gage head and a gage amp. I have a Federal that registers within 0.000005" and higher by 5 ranges. These aren't cheap or easy to find but by persistant haunting of eBay and other suction sites such treasures can be found.

Here's one costing an effing fortune:

http://cgi.ebay.com/Starrett-Electro...item5627fb0e2e

I bought several in the past in the $150 range.

Anyway Cecil has the smarts to lap gage blocks to hgh accuracy but it doesn't sound like he has the gear to support the sizing and calibration . If we're getting fussy, a good calibrated set of gage blocks need to be on hand if the finish size on the re-laped block is worth taking to the bank.

[cecilstrange]

The indicator is a Mahr Millimess. You're right that hysteresis has to be accounted for. I do it with light taps from a pencil. The indicator is rigidly mounted to a granite comparator stand, with a light preload, so I can just slide blocks underneath for comparative measurements. This is really just a double check on measurements from the mike, but it also makes it easier to explore for high and low spots on the block. My measurements repeat across three quality mikes using friction stops. My comparisons are to a set of Hoke blocks from the 60's which mike out as perfect as I can see. The mikes are also calibrated to the Hoke blocks.

The lapped surfaces wring just fine. I do have an optical flat, but I am just learning how to use it and do not feel confident about making any pronouncements yet.

I was amazed myself that rubbing in a simple figure eight pattern on a piece of milled cast iron gave me enough control. I did have to vary finger pressure to restore parallelism while I progressed. Of course I was going slow and progressed from 320 grit diamond to 3000 grit in eight steps. I couldn't detect any material was removed in the last four steps but the luster improved dramatically.

[The real Leigh]

Originally Posted by cecilstrange

This is really just a double check on measurements from the mike, but it also makes it easier to explore for high and low spots on the block. My measurements repeat across three quality mikes using friction stops.

You're mikes aren't accurate enough for these measurements, by a couple of orders of magnitude.

Depending on accuracy class, gage blocks are typically within ±10 millionths of an inch of nominal dimension. Even a tenth-reading mike is more than an order of magnitude less accurate.

Metrology SOP dictates that the measuring equipment being used must be at least an order of magnitude MORE accurate than the tolerance of the parameter being measured.

That requires a limit of error in your measuring system of ±1 millionth of an inch for ±10 millionths tolerance.

As Forrest mentioned, measurements to this level of precision require electronic equipment.

- Leigh

[The real Leigh]

Originally Posted by bryan_machine

Indicator? Dorsey metrology:
http://www.dorseymetrology.com/produ...ndicators.html

Yeah, that's 20 millionths per division.

My Hamilton is 10 millionths per division.

But they're both way less accurate than would be required for gage block calibration.

- Leigh

[cecilstrange]

There is no argument here about calibration standards, and I never claimed I was calibrating anything. I only reported I made a gage block thinner while keeping it parallel and to target size AS CLOSE AS I COULD MEASURE.

All of my measurements were comparative. If I can repeatably detect that one gage block is thicker than another, and they are marked .1004" and .1005", is my comparison meaningless because I am incapable of calibrating the vernier scale on a micrometer or the motion of a one micron indicator?

What is most interesting to me is detection that some, but not many, of my Webber blocks are oversize by up to .002". I could believe the deduction is faulty because of measurement error, but that is hard since measurements of another set of blocks does not show errors from nominal size.

[whitis]

Originally Posted by The real Leigh

Metrology SOP dictates that the measuring equipment being used must be at least an order of magnitude MORE accurate than the tolerance of the parameter being measured.

I am not a metrology expert but this tends to get overstated a bit, often more so than here.

The pirates code is more like guidelines, than actual rules. :-)

If you have 10X accuracy, you are allowed to ignore error accumulation in the chain back to international standards, possibly not calculate the accuracy of simple measurements to painstaking levels of detail, not do buckets of measurements, etc. Life is much simpler. I forget the exact distinction but it involved being allowed to ignore some forms of error accumulation to an extent that was questionable.

But if this was an iron clad rule, we would never have those 10X standards to begin with. Of course, there is a certain amount of arbitrary fiat in setting master standards. The inch is based on the meter. The meter is based on a the distance light travels in a certain fraction of a second (time being our most accurate measurement). A number chosen, of course, based on yesterday's (for sufficient values of yesterday ) definition of the meter which was based on a multiple the wavelength of light from a particular atomic transition, a number chosen based on the day before yesterday's definition of meter. Back in 1790, the meter was also based on time - the distance of travel of a pendulum with a half period of one second.

On the flip side, you can follow the 10X rule and still end up with results that are crap.

Yes, he needs more accurate instruments to get those blocks back to original claimed tolerances. However, he could calibrate them to a lower accuracy but not necessarily 10X lower than measurement equipment. And it might be a worthwhile learning exercise, even it the end product leaves something to be desired.

Also, accuracy of measurement equipment isn't an issue when the equipment is used in relative mode rather than absolute mode.

One approach might be to effectively subtract gage block sizes to get the desired result by using two surfaces. For the missing .1005 block:
.1005 + .1006 = .1004+.1007
Errors:
tolerance on .1006 block
tolerance on .1004 block
tolerance on .1007 block
tolerance of surface plate
precision/repeatability (not accuracy) of indicator, as used
temperature induced error
wringing film error (1 film included in each block size, film is pretty small)

Even 1 degree C temperature error can give something like 12microinch error. Non-precision temperature measurements frequently aren't any better than 1 degree C and sometimes not that good. However, much of the temperature error will cancel out, since steel expands at the same rate as steel. Change in temperature over time and lack of thermal equilibrium is another can of worms (I have done optical measurements precise enough that it took days to stabilize). Then there is body heat, light sources, etc. which may heat things unevenly.

Use the measured values of the various blocks from the calibration certificate as your starting point, not the nominal values. Actual error may be less than calculated. Gage block set accuracy is measured relative to nominal sizes; the tolerance on the measured values is likely tighter - or at least it would be if you were using a set that didn't have some known out of tolerance blocks suggesting the whole set is a dud. Ordinarily, they would have compared the block to a master (with probably 10X accuracy) using a comparator (with probably 10X accuracy), so the measured sizes may be 5X more accurate than the claimed accuracy of the set. By doing his own comparisons, he may end up with a calculated tolerance of about 5X wider than the specs for the block set but a block that is just barely within the claimed accuracy due to the two factors of 5 cancelling out. But you don't know for sure. And since there are other descrepancies in the set...

Also, adding the tolerances on the 3 gage blocks used in the proceedure tends to exagerate the actual error (but not worst case).

For digital indicators, you need to add 1 count if you zero each time you compare and 2 if you don't (inch/metric units have higher internal precision than displayed, so zeroing is more accurate than subracting readings - you may be able to take advantage of the extra precision using SPC output and time averaging).

The stacking method provides a way of double checking the gage block errors reported by the micrometer

Different areas on the surface plate can be used. Many measurements need to be taken across the surface. Hand scraping techniques (bluing, etc.) can be used - add the tolerance on the reference surface to your list of errors. Optical flats, if available.

Wringing film is typically 0.4 microinch but some can be over 1 microinch while others less than 0 microinches (not a typo! 2007 NIST Gage Block Handbook pg 134).

LVDT comparators are used for calculating high grade sets, you may not need such precise equipment for "recalibrating" a shop grade set.

NIST handbook also describes some other techniques of using multiple blocks for comparison (pg 131).

The variation due to age fluctuation (not wear) should be something like 0.000002" on a 0.1" block, not 0.002". Were these blocks mislabeled? Calibration all wrong? Were they not properly stress relieved? Is there an oil crust? Fakes? Not original? imported and not checked? Measured at body temperature? Drastically mishandled?

[The real Leigh]

Originally Posted by whitis

On the flip side, you can follow the 10X rule and still end up with results that are crap.

Isn't that the truth?

I worked as a metrology technician in an NIST-traceable cal lab for several years, and the accuracy guidelines were emphasized daily.

Originally Posted by whitis

The variation due to age fluctuation (not wear) should be something like 0.000002" on a 0.1" block, not 0.002". Were these blocks mislabeled? Calibration all wrong? Were they not properly stress relieved? Is there an oil crust? Fakes? Not original? imported and not checked? Measured at body temperature? Drastically mishandled?

You forgot... Used as a drift punch?

I can't think of any type of abuse that would cause elongation, other than using one as an anvil.

Very good analysis.

- Leigh

p.s. I corrected a typo in your previous post.

[cecilstrange]

Whitis, thank you for your helpful perspective. A couple of things from my post did not translate perfectly and I hesitate to mention them because the differences in no way detract from the logic or precision of your analysis.

I am not missing a .1005" block; it is a .10005" block that is missing. I think I will leave it missing, or buy one, as a constant reminder of the limits to the information value of ordinary workshop measurements. Next time I work on this I will use your suggestion to check measurements by stacking blocks and subtracting. I've already triple checked using a surface gage, test indicator, different surface plate, and a Cadillac height gage. Dependable or not, repeatable indications to the tenth using different equipment generates confidence, but I know I would cross into the realm of the fool hearty if I started to think my gear could split a tenth.

Also, I described a .0002" variation in some blocks in my old Webber set, not .002". Thanks for quantifying the significance of changes due to age. Obviously that is not the problem, nor do I know what is.

"Mislabeled, Not original, Fakes, Imported and not checked?" -- Perhaps but it seems unlikely. Two of the larger blocks are labeled Webber, LSS Co, the rest are labeled Webber, Cleveland. The smaller blocks are labeled just Webber, but sometimes the brand is at the top edge and sometimes in the middle. The box has the original metal nameplate. Obviously this set has some replacements, but I wouldn't suspect counterfeit.

"Calibration all wrong?" -- This wouldn't account for blocks labeled .001" apart being oversized from the labeled size by a tenth or two.

"Were they not properly stress relieved?" -- I have no idea. The vast majority of the blocks measure as labeled. As with any reputable brand that is generally trusted, I guess defective manufacture is possible.

"Is there an oil crust?" I've cleaned them with tolulene, acetone and alcohol. I've used leather chamois, cotton shop rags, fingernail and paper towel for abrasion. There is some remaining staining so perhaps there is a build-up I did not suspect. But the offending blocks do not appear more stained than the ones that check out to size. I'll try some lye to soften whatever is left before I resort to an abrasive polish.

"Measured at body temperature?" I know not to inject them with heat from my body or proximity to local sources. I handle them with a spring type clothes pin.

"Drastically mishandled?" -- This is anybody's guess. The set is obviously very old and the swarf in the box indicates a lot of shop exposure. The gauging surfaces wring and look unblemished, except for the staining and the butcher mark on the side of the four inch block from collision with a grinding wheel.

This is an educational exercise but with a practical purpose. I use my Hoke blocks with my layout equipment. My grinder is in a different room and I dug out this old set of blocks to have handy for sine bar work on the grinder. Attempting to replace missing blocks was mainly a curiosity explored when I had time to fill, but the wear blocks I've fabricated will get used. The blocks graduated in tenths may never have their oil film wiped off again.

[whitis]

Originally Posted by The real Leigh

I can't think of any type of abuse that would cause elongation, other than using one as an anvil.

Yeah, when I said drastic, I was thinking of stuff along those lines. Impact, using it to shim up a lathe the size of an aircraft carrier, etc. But also, extreme temperature changes (more extreme than those used in stress relief, at either end of the scale). Immersing in liquid nitrogen or heating really hot. It isn't a good idea to use your gage blocks to demonstrate coefficient of thermal expansion, especially not if you go to extremes :-) Or to use them to hold parts in alignment in the brazing furnace.

Take the plastic window from a envelope and expose it to autoclave type temperatures. It will shrink down to a very small, but thick, chip. Which is probably what it's original shape was before it was rolled to a thin sheet.
Metal often gets rolled, too. Metal isn't plastic, of course, but it does sometimes act like it, it is just too subtle about it for most people to notice. Except nitinol memory wire. If the gage block was cut from metal sideways, then when the stress was inadvertently relieved it might want to elongate in two directions and shrink on the other. I.E., there might be a method to its madness, as opposed to the usual behavior of rolling into a pretzel.

I cringed when the guy in one of the machining videos said that a piece of metal would always be "right on the money" 6 months later. Of course, the same guy WELDED an indicator holder/transfer stand to a precision ground ring of material to check for squareness on the surface plate. Ok, you still have to calibrate the indicator each time with a precision square (which he doesn't show), but a stunt like that deserves a disclaimer.

p.s. I corrected a typo in your previous post.

Only one?

[whitis]

Originally Posted by cecilstrange

I am not missing a .1005" block; it is a .10005" block that is missing. I think I will leave it missing, or buy one, as a constant reminder of the limits to the information value of ordinary workshop measurements. Next time I work on this I will use your suggestion to check measurements by stacking blocks and subtracting. I've already triple checked using a surface gage, test indicator, different surface plate, and a Cadillac height gage. Dependable or not, repeatable indications to the tenth using different equipment generates confidence, but I know I would cross into the realm of the fool hearty if I started to think my gear could split a tenth.


Also, I described a .0002" variation in some blocks in my old Webber set, not .002". Thanks for quantifying the significance of changes due to age. Obviously that is not the problem, nor do I know what is.

Sorry, didn't remember the numbers right.

"Mislabeled, Not original, Fakes, Imported and not checked?" -- Perhaps but it seems unlikely. Two of the larger blocks are labeled Webber, LSS Co, the rest are labeled Webber, Cleveland. The smaller blocks are labeled just Webber, but sometimes the brand is at the top edge and sometimes in the middle. The box has the original metal nameplate. Obviously this set has some replacements, but I wouldn't suspect counterfeit.

"Calibration all wrong?" -- This wouldn't account for blocks labeled .001" apart being oversized from the labeled size by a tenth or two.

Well, you don't want to underestimate the ways to screw up calibration. When it works, it works; when it doesn't, anything can happen. Look at Hubble. Two tests. Passed one, failed the other. They decided to believe the one it happened to pass, because it was more accurate, even though it was the same test they used in making the part - of course it passed. And that was on a multi-billion dollar project. One bad bit on an A/D converter on the LVDT probe system could cause havoc. You would think there would be a completely separate QA/QC phase.

But these blocks with the mix and match sound like they had a checkered history so we don't know what happened after they left the manufacturer.

I've cleaned them with tolulene, acetone and alcohol. I've used leather chamois, cotton shop rags, fingernail and paper towel for abrasion. There is some remaining staining so perhaps there is a build-up I did not suspect. But the offending blocks do not appear more stained than the ones that check out to size. I'll try some lye to soften whatever is left before I resort to an abrasive polish.

"Measured at body temperature?" I know not to inject them with heat from my body or proximity to local sources. I handle them with a spring type clothes pin.

Yeah, I figured you knew enough to avoid body heat, etc.

I wouldn't get too drastic (lye, abrasive polish). Yeah, steel likes things a little on the alkaline side but lye might be pushing it. Dilute might be ok, though. Abrasive polish could also cause trouble, though it sounds like you can handle it.

Could your stains be rust? That could puff up the surface.

This is an educational exercise but with a practical purpose.

Made for an interesting thread.

[The real Leigh]

Originally Posted by whitis

Only one?

Yeah. I'm not an editor. I only make corrections when it might affect the meaning.

In the case at hand, you had misspelled "our" as "out".

- Leigh

[bryan_machine]

re: The Real Leigh:

"Quote:
Originally Posted by bryan_machine
Indicator? Dorsey metrology:
http://www.dorseymetrology.com/produ...ndicators.html

Yeah, that's 20 millionths per division.

My Hamilton is 10 millionths per division.

But they're both way less accurate than would be required for gage block calibration."

Actually, you asked about 1 micron reading dial indicators, you can get the dorsey marked that way as well as millionths. Of course no such device is adequate for certifying gauge blocks, at least without very elaborate statisical analysis and big error bars. It is a happy and perhaps essential fact that interferometers preceed gauge blocks (by a few years.)

But given a good surface plate, and a collection of measurements, a .5 micron or 0.000 020" indicator should allow one, with high confidence, to say that an old weber block is actually different from some certified block, by enough to not trust the weber. Which all just says the scale of the errors suggests that cecilstrange has found a real issue with some of his blocks.

I personally wonder what happens if you put a gauge block in the heat treating oven, and then immediately plunge it into liquid nitrogen. Sounds like a fine experiment for a youtube video

It also seems that a gauge block and fine micrometer would be an excellent way to make a steam punk thermometer

Seriously, a known good gauge block (with current lab cert) and a known good and very sensitive dial indicator, would give a good check to local temperature control, no? Of course, in my shop where the furnace blasts a column of hot air randomly into the middle, this would mostly confirm a problem I already know about

[The real Leigh]

Originally Posted by bryan_machine

Actually, you asked about 1 micron reading dial indicators, you can get the dorsey marked that way as well as millionths.

That was an error on my part.

I don't work in metric, so when I saw "microns" I was thinking millionths of an inch.

Originally Posted by bryan_machine

But given a good surface plate, and a collection of measurements, a .5 micron or 0.000 020" indicator should allow one, with high confidence, to say that an old weber block is actually different from some certified block, by enough to not trust the weber. Which all just says the scale of the errors suggests that cecilstrange has found a real issue with some of his blocks.

Yes, I'm inclined to agree.

That would suggest some really serious abuse of the Webber blocks, but such is not impossible. I've seen "machinists" use stacks of gage blocks to support the outboard end of a work hold down clamp.

These are the same folks who use micrometers as C-clamps.

Originally Posted by bryan_machine

Seriously, a known good gauge block (with current lab cert) and a known good and very sensitive dial indicator, would give a good check to local temperature control, no?

Yep. I believe I mentioned a similar configuration early in this thread.

My earlier comments were just meant to point out that accurate precision measurements require analysis of the instruments and the methodology.

For example, the tightest-resolution Dorsey dial indicator is spec'd at ±20 millionths accuracy and ±10 millionths repeatability. When used as a comparator, you can only guarantee the reading to ±10 millionths if both readings are identical (and they may not "really" be identical). If they differ, the accuracy spec must also be considered.

Anyhow, interesting thread.

- Leigh

[Screwmachine]

Does anyone know how gage blocks are lapped at the factory? For some reason it seems like an ever so slightly convex lap would be needed, as there will almost invariably be more influence on the outer edges than in the middle .

[The real Leigh]

Originally Posted by Screwmachine

Does anyone know how gage blocks are lapped at the factory?

I saw an article once on the lapping technique. They use a machine with the blocks sitting on a rotating circular lap. It was designed to keep the blocks dead straight while turning them axially to provide a uniform surface. Quite a complex machine.

I expect Google could find some info.

Originally Posted by Screwmachine

For some reason it seems like an ever so slightly convex lap would be needed...

Won't work. For them to wring they must be dead flat.

Cal certificates for gage blocks typically show five measurements, one in the center, and the others at some distance out, but not at the corners. I believe there's a standard for the exact location.

I just pulled a Starrett cert for one of my sets. It has five columns, one labeled Reference Point ("Ref Pt") and the others "Deviation at Auxiliary Points", numbered 1 through 4.

For example, measurements for the .125 block show a Ref Pt of +1 and the other four as +2, +1, +1, and +1 (all values in microinches).


- Leigh

[Screwmachine]

Originally Posted by The real Leigh

Won't work. For them to wring they must be dead flat.

- Leigh

I guess what I was saying was that if a block was manipulated on a perfectly flat lap manually, the drag of the lap would impart a tilting force to the gage favoring the leading edge, which would almost certainly effect the flatness. A slightly domed lap- and I mean really slightly- would counteract that.

What brought it to mind is a lap used to refinish burnishers used in old school watchmaking. These were used to finish pivots in a special lathe, and were straight pieces of flat steel with a 'tooth' imparted by the lap. The commercial laps seen in the old catalogs were convex, to insure that the burnishers don't become convex (obviously some skill was required).

The machine you describe I've also seen in industrial watchmaking shops, with a large rotating lap and a system to rotate the work on the lap as it turns. I'm sure the ones used for gages are quite a bit more sophisticated.