PDA

View Full Version : While on the subject of levela and machine levelling..

The Artful Bodger
01-04-2012, 02:35 PM
While we are on the subject of machine levelling..and the levels to do the job. Just how does the stated sensitivity of the level relate to graduations on the vial?

I have a Czech level with a test sheet that shows 0.17 per 1000mm. Even to my eyes it is easy enough to visually interpolate to about 1/10th of a graduation which seems to me to be very useful sensitivity.

So, is the stated sensitivity per graduation or what is visually perceptable?

Tony
01-04-2012, 03:05 PM
I'm sure someone will mention the correct ISO/DIN/ANSI/AME standard
for levelling levels and it'll be 0.001 x 1000 but one thing I can tell you
(and I didn't mean to be sarcastic with that)..

I got two levels the "box" kind both 10" on each side one good to
0.02mm/m and the other 0.03mm/m ... (two for 70 euro.. is that a gloat?)

anyway the 0.03mm I can work with.. meaning I can get it into reading
range with 0.001" shims, etc..

the 0.02mm level however has rocket engines on the bubble.. if i so
much as stare at it too long the bubble will fly to the other side.

-Tony

01-04-2012, 06:07 PM
The vial was probably checked with a sine plate and the bubble ends sighted with coincidence optics of some kind. I don't know for sure; I'm speculating. But if I had a level vial factoy where I made vials of refined and calibrated sensitivity for an elite clientele, I'd want a means of inspection superior to plain eyeball and reverse to repeat zero. I'd want a means to precisely quantify vial accuracy graduation to graduation slope representation.

Setting up to quantify a vial's graduation accuracy requires only a stable flat surface and a long sine plate. An arc second is 1 in 1,286,000 slope exactly (60 x 60 x 360). I got curious and quick-checked my Starrrett 199 levels to determine deviation from advertized sensitivity. 0.0005" in 10" is claimed. By shifting Jo blocks in and out and calculating I'm guessing with a little confidence the actual sensitivity is a bit less - closer to 0.0006" in 10" than 0.0005 but without repeated trials tossing out the high and low results and averaging the others, a second trial another day, seeking second party cofirmation, etc. In other words, seeking as refined and accurate data as I can manage pushing the envelope of my apparatus.

This kind of refinement is for them that like to get fussy. Tenth arc seconds and parsing slopes to parts per million has a real application in astronomical telescopes, particle accelerators, NIST investigations, etc but not anything in machine tools where slopes of 1 in 100,000 is adequate for the most refined manufacturing process and 1 in 20,000 is adequae for all but the very finest of hope shop equipment.

The Artful Bodger
01-05-2012, 05:48 PM
Thanks for the comments, but can anyone enlighten me, does the stated sensitivity relate to vial graduations? Thanks.

oldtiffie
01-05-2012, 07:11 PM
Yes - or it should.

The graduations on the vial should be such that the best or quoted accuracy is one unit - ie from one mark to the next.

I hope this helps.

John Garner
01-05-2012, 08:32 PM
Artful --

The sensitivity of precision level vials is most commonly stated as the tilt necessary to move the bubble within the vial a distance equal to one graduation interval, or, much more rarely, as the radius of the vial's lengthwise curvature.

US and older British precision vials are generally graduated at 1/10 inch intervals, European, Asian, and some newer British vials are graduated in 2 millimeter intervals. At least two studies (that I know of) demonstrate that the human eye can reliably subdivide 1/10 inch graduations into 5 parts, 2 millimeter graduations into 4 parts . . . in other words, the eye effectively resolves bubble position to 0.02 inch or the nearly-equal 1/2 millimeter.

Because 1/10 inch graduations are very nearly 1 1/4 as far apart as 2 millimeter graduations, a 1/10 inch per graduation vial of a given tilt sensitivity per division will have a longitudinal radius of curvature that is very nearly 1 1/4 times as long as a vial of the same tilt-per-division sensitivity graduated at 2 millimeter increments.

Precision level vials are generally calibrated by their maker, using a "level trier" that would, in machine-shop lingo, probably be termed a "micrometer sine bar". The usual test protocol is to set the vial in the level trier's V-block vial carrier, the 1) adjust the micrometer until the right edge of the vial is tangent to the center-most right-side-of-vial graduation, and note the micrometer reading. Then, 2) adjust the micrometer until the right edge of the vial is tangent to the outer-most right-side-of-vial graduation, and note the micrometer reading. From there, its a matter of calculating the angle between the two bubble-in-vial positions and dividing by the number of graduations the bubble moved.

The thus-determined tilt-per-division is then generally scribed into the glass of the vial, near one end, although less-sensitive vials may have the sensitivity value silk-screened onto the vial. I speculate that the coarser vials don't warrant individual testing on the level trier.

Forrest --

Movies have been made about Challenging the Master, and they describe a bit of what I'm feeling right now . . . .

An arc second is 1 of 1,296,000 equal parts of a circle, but a slope of 1 arc second, expressed as Rise : Run, is very nearly 1 : 206,265.

And as for the Starrett Model 199Z . . . Starrett somewhat contradicts themselves in specifying the sensitivity of the 199Z at 0.0005 inch per foot, which they state is equivalent to 10 arc seconds per 1/10 inch vial division. Arithmetic shows that a Rise of 0.0005 inch over a Run of 1 foot is really equivalent to a tilt angle of 8.6 arc seconds (rounded to the nearest 1/10 arcsecond).

Having performed both in-situ vial calibrations and vial replacements on a fair number of Starrett 199Z levels, I've found their usual sensitivity to be either 8 or 9 arc seconds per division, with more-sensitive and less-sensitive vials being much rarer.

John

The Artful Bodger
01-05-2012, 08:50 PM
Tiffie, John, thanks, thats exactly the information I was looking for.:)

oldtiffie
01-06-2012, 12:47 AM
AB.

Having settled the issue of the vial marks and sensitivity of the level the next issue is to determine if the bubble is centred on a horizontal flat surface. A surface plate will be ideal.

It is first necessary to use the level - irrespective of its setting or centreing - to set the flat surface to the horizontal plane. That will act as a referencee plane for the level.

The next is to use an iterative process to adjust and test the vial to get the bubble centred in the vial when the level is turned 180 degrees on the reference surface.

It seems and quite often is finicky and frustrating but once mastered takes very little time.

The level should be checked and if required (re)adjusted to a reference flat horizontal plane before using the level as a test instrument - such as levelling a lathe.

danlb
01-06-2012, 02:03 AM
For the purposes of leveling a lathe, wouldn't you get sufficient accuracy by using a liquid level of the kind used on foundations and such?

For those who have not seen one... from a web site.

Water seeks a level automatically. One can make a water level by attaching garden hose fittings to two (2) foot lengths of clear plastic tubing, a male and female fitting on opposite ends.

Hold the ends next to the things that you need level and you have an accurate reference point.

It seems that if you use a water level against both ends of the lathe that you can easily discern differences better than .010 inch over 4 foot using no other tools. How much more accurate does it need to be?

Dan

philbur
01-06-2012, 03:04 AM
I think the normally accepted value is 1/4 of a graduation. To reliably get to a 1/10th of a graduation I think you may need a vernier?

Possibly the way to consider it is: If 10 (suitable) people estimated a particular reading to 1/10th of a graduation would they ALL come up with the same value. If the answer is maybe not then it's not reliable at that resolution?

Phil:)

Even to my eyes it is easy enough to visually interpolate to about 1/10th of a graduation which seems to me to be very useful sensitivity.

The Artful Bodger
01-06-2012, 04:49 AM
Tiffie, Dan, Phil...thanks for the comments.

For my application, which is testing my machine is not twisted, I am not so concerned with the accuracy of my level. In fact I assume it is not accurate and therefore always use it aligned the same way, if practical. However it may be useful in the future to know how to test and if necessary adjust for level so thanks for that Tiffie. Meanwhile, I think my level has a (pitch?) plug covering where the vial adjustment might be.

Dan, I recall my efforts at our previous house to level the lawn using a water level as you describe. It was a very frustrating process and I can only assume that it was air dissolved in the water that was causing me gyp. Nonetheless, a water level may be practical, I just dont know, but I suspect a simple plumb bob could be used to check for twisting of the lathe ways, simply clamp a vertical bar on the tool post, longer is better depending on the overhead space. Hang a plumb bob from the top end and align the point of the bob with a mark on the cross slide. Wind the carriage to the other end and any gross error should be obvious (I think?).

Phil, I used to read meteorlogical instruments as my daily job, several times a day, and we were shown how to interpolate to 1/10th. It is not too difficult, first judge if it is more or less than half then more or less than half of that. The hardest part of reading the level on a lathe is getting one's eyes in a position to avoid parallax error, IMHO.

philbur
01-06-2012, 05:31 AM
But that only takes you to a 1/4?

Phil:)

....and we were shown how to interpolate to 1/10th. It is not too difficult, first judge if it is more or less than half then more or less than half of that.

Davo J
01-06-2012, 05:45 AM
Hi John,
I picked up a Rabone level on ebay years ago and didn't know what resolution it was, so I sat it on top of a 1.2 meter long carpenters level, then leveled it up with packers 1 meter apart. After that I started placing shims under one end until it moved one devision on the Rabone level.

Dave

J Tiers
01-06-2012, 08:55 AM
Nonetheless, a water level may be practical, I just dont know, but I suspect a simple plumb bob could be used to check for twisting of the lathe ways, simply clamp a vertical bar on the tool post, longer is better depending on the overhead space. Hang a plumb bob from the top end and align the point of the bob with a mark on the cross slide. Wind the carriage to the other end and any gross error should be obvious (I think?).

Phil, I used to read meteorlogical instruments as my daily job, several times a day, and we were shown how to interpolate to 1/10th. It is not too difficult, first judge if it is more or less than half then more or less than half of that. The hardest part of reading the level on a lathe is getting one's eyes in a position to avoid parallax error, IMHO.

In principle, of course a plumb bob would be usable..... the problem with it is in seeing the divisions. Friction issues and repeatability *should* be a non-issue with a plumb bob. Wind might be a problem, though.

Back to "seeing the divisions"..... that's the problem with almost ANY measuring device..... once you get past the issue of the device being actually made carefully enough to be accurate*.

For small changes, you would need a microscope to see the movement of the bubble in a carpenter's level, even if the vial was made perfectly accurate so that the bubble moved exactly correctly. Same for the movement of the "bob" for any sensible length of support for a plumb bob.

And that is precisely what a "sensitive" level fixes....... actually it fixes BOTH issues....

First, by having a high "sensitivity", it magnifies the differences so that they are easy to see. The sensitivity spreads out the divisions to make them visible. if the "perfect" carpenter level measured 10 arc seconds, it would be very difficult to spot the movement, the division lines would be tiny and very closely spaced.

Then also, the spread-out divisions make the vial easier to make.... errors in the fractional ten-thousandths are easier to fix than errors in the fractional millionths (or less). The latter might be needed to grind and put a scale on the "perfect" carpenter level vial. The division marks on the sensitive level can be put on fairly crudely and still not affect the accuracy by very much

* I once posted here a comparison of two machinists rules/scales. One was a B&S (IIRC), one was a different brand.... They did not agree, being at least one division line width "out" of agreement over just a few inches, amounting to nearly 20 thou as I recall.

Oddly, the B&S was the one that was wrong...... and clearly the inherent accuracy of the B&S piece was no better than one line width. So much for setting spring calipers to it with a magnifier.....

The Artful Bodger
01-07-2012, 01:45 AM
But that only takes you to a 1/4?

Phil:)

A bit more than 1/4 is .3, a bit less is .2, a bit more than 0 is .1, and so on...

The Artful Bodger
01-07-2012, 01:54 AM
JT, you are right of course, you cannot make Grandpa's old pocket watch more accurate by examining the face with a magnifyer.

However, some of the things we do are more a matter of comparison rather that absolute readings so it does not really matter how accurate our level is provided it is sensitive enough to show a twist in the lathe bed.

Some instruments have a mirror behind the needle to help avoid parallax error and although I have never tried it there may be a way of using that or a similar principle to get more accuracy/sensitivity with a plumb bob.

For sure a sensitive level is the best tool for the job that most of us could aspire to but even a modest one like mine costs enough to make one stop and think.

01-07-2012, 05:10 AM
Forrest --

Movies have been made about Challenging the Master, and they describe a bit of what I'm feeling right now . . . .

An arc second is 1 of 1,296,000 equal parts of a circle, but a slope of 1 arc second, expressed as Rise : Run, is very nearly 1 : 206,265.

And as for the Starrett Model 199Z . . . Starrett somewhat contradicts themselves in specifying the sensitivity of the 199Z at 0.0005 inch per foot, which they state is equivalent to 10 arc seconds per 1/10 inch vial division. Arithmetic shows that a Rise of 0.0005 inch over a Run of 1 foot is really equivalent to a tilt angle of 8.6 arc seconds (rounded to the nearest 1/10 arcsecond).

Having performed both in-situ vial calibrations and vial replacements on a fair number of Starrett 199Z levels, I've found their usual sensitivity to be either 8 or 9 arc seconds per division, with more-sensitive and less-sensitive vials being much rarer.

John

Agree to all but the last paragraph where I take a small exception. My data was rough, taken in haste and was declared as such.

My shot at quantifying the slope indicated by bubble movement of a single vial graduation of my particular level was as stated in my earlier post: "...closer to 0.0006" in 10" than 0.0005... I said a "bit closer to... are the operative words: Since you're splitting hairs I shall too. Call is maybe 0.00056" (putting a shakey figure a bit closer) in 10". That works out to 11.55 seconds. You got 8 to 9 arc seconds on a different level probably taking greater care than I so you probably got better data. My check of my level was hurried so its entirely possible I may have blown the slope a tenth or two. Had I conducted my check with more care I'd cheerfuly defend my data but I'm not that confident. I'd certainly never put a calibration sticker on the level I checked with a quick and dirty slope test of a single graduation.

I need to rant: Starrett needs to clear up the slope per 10", slope per ft, angle subtended by a single graduation. It's confusing for people working to the tool's limit of accuracy - there's a 17% conflict in Starret's claim.

Another point is Starrett makes a big deal over the reference face being hand scraped but they don't discuss the 0.0005" deliberate concavity in 15". Most people would assume it to be scraped flat. If the level was offset over parallels for some reason the surface being leveled could be in error but a small amout. These are fiddling little things but dammit, the Starrett propaganda states :"Master Precision Level" Why can't Starrett address these issues which may be important to those few needing the last iota of accuracy from the tool.

01-07-2012, 05:42 AM
Thanks for the comments, but can anyone enlighten me, does the stated sensitivity relate to vial graduations? Thanks.

Doesn't it say right on your level "1 graduation = xxxx" followed by an angle expressed in arc seconds or a slope? My Starrett 199's does and so do my Polish levels and both block levels.

philbur
01-07-2012, 07:34 AM
http://www.leveldevelopments.com/sensitivity-explained.htm

J Tiers
01-07-2012, 10:44 AM
Another point is Starrett makes a big deal over the reference face being hand scraped but they don't discuss the 0.0005" deliberate concavity in 15". Most people would assume it to be scraped flat. If the level was offset over parallels for some reason the surface being leveled could be in error but a small amout.

Seems that is a Federal purchasing spec, and was discussed here..... the upshot being that it was not necessarily 0.0005" concave, but some unknown amount that was not held to any tight spec for total depth of concavity, smoothness of concavity, etc.

Lufkin only put "feet" at each end of theirs.

The Artful Bodger
01-07-2012, 03:02 PM
Doesn't it say right on your level "1 graduation = xxxx" followed by an angle expressed in arc seconds or a slope? My Starrett 199's does and so do my Polish levels and both block levels.

Perhaps it does on the calibration sheet but I dont read Czesky.:(

01-07-2012, 03:44 PM
Yeah, Jerry, that's true but My beef with Starrett is withholding of cricial information in their advertizing.

There are differences between 10 arc seconds, 0.0005.10", and 0.0005" per ft all if thich have been used deparately as sensitivity fijures.

BTW I looked for the Fed Spec on precision levels. I thought I had it but I don't and my Google skills came up dry morning. You wouldn't have a number for it.....

John Garner
01-09-2012, 03:25 PM
Forrest (and Company) --

After re-reading my earlier posting, I can see how you understood my last two paragraphs to question your findings about the sensitivity of the Starrett 199Z vial; not only was that not my intent, I was actually trying to say that your conclusion, of 0.0006 inch per foot, is in the range I've come to expect for the 199Z. I do apologize for the obtuse transitioning, and I'll try to make my change-of-context breaks more evident in the future.

On to another subject . . . my printed copy of the now-obsolete U.S. "FEDERAL SPECIFICATION LEVELS AND PLUMBS, AND LEVELS", GGG-L-211C, as amended January 30, 1970, in Paragraph 3.17, "Type IX, precision (master) levels", defines only one class, Class 1, thusly: "Class 1, single face levels shall consist essentially of a stock, one adjustable level vial, one fixed cross test vial, a vial holder, a cover, and a case."

Paragraph 3.17.1.1 says "Stocks. Level of this class shall have a single-faced stock of steel, or close-grain cas iron, well seasoned, and of the design to preclude deflection or warping due to temperature change. The base shall be fine precision scraped."

Paragraph 3.17.1.2 says, in part: "Vials. The main level vial shall be a lime or heat-resistant boro-silicate glass tube and shall be ground to an inside radius of 2,070-inches, minimum, equivalent to a vial sensitivity of 10 seconds, maximum. There shall be no less than three auxiliary lines, permanently marked, spaced 1/10-inch apart on both sides of both bubble defining lines. The angular values of these lines, in either degrees or inches per foot, shall be permanently marked on the level."

The maximum permissible concavity of the scraped surface varies, being 0.0002 inch for levels with stocks 12, 15, or 18 inches long, 0.0003 inch for levels with stocks 27 inches long. Convex surfaces are forbidden.

John

oldtiffie
01-09-2012, 11:24 PM
Forrest (and Company) --

After re-reading my earlier posting, I can see how you understood my last two paragraphs to question your findings about the sensitivity of the Starrett 199Z vial; not only was that not my intent, I was actually trying to say that your conclusion, of 0.0006 inch per foot, is in the range I've come to expect for the 199Z. I do apologize for the obtuse transitioning, and I'll try to make my change-of-context breaks more evident in the future.

On to another subject . . . my printed copy of the now-obsolete U.S. "FEDERAL SPECIFICATION LEVELS AND PLUMBS, AND LEVELS", GGG-L-211C, as amended January 30, 1970, in Paragraph 3.17, "Type IX, precision (master) levels", defines only one class, Class 1, thusly: "Class 1, single face levels shall consist essentially of a stock, one adjustable level vial, one fixed cross test vial, a vial holder, a cover, and a case."

Paragraph 3.17.1.1 says "Stocks. Level of this class shall have a single-faced stock of steel, or close-grain cas iron, well seasoned, and of the design to preclude deflection or warping due to temperature change. The base shall be fine precision scraped."

Paragraph 3.17.1.2 says, in part: "Vials. The main level vial shall be a lime or heat-resistant boro-silicate glass tube and shall be ground to an inside radius of 2,070-inches, minimum, equivalent to a vial sensitivity of 10 seconds, maximum. There shall be no less than three auxiliary lines, permanently marked, spaced 1/10-inch apart on both sides of both bubble defining lines. The angular values of these lines, in either degrees or inches per foot, shall be permanently marked on the level."

The maximum permissible concavity of the scraped surface varies, being 0.0002 inch for levels with stocks 12, 15, or 18 inches long, 0.0003 inch for levels with stocks 27 inches long. Convex surfaces are forbidden.

John

A good well stated and researched post and very much appreciated post John - many thanks.

The matter of whether a Starret precision level base is flat or not or if not whether it may be concave or convex and if so the what extent is neatly covered in the final paragraph of John post:

The maximum permissible concavity of the scraped surface varies, being 0.0002 inch for levels with stocks 12, 15, or 18 inches long, 0.0003 inch for levels with stocks 27 inches long. Convex surfaces are forbidden.

So it may not be convex but may have limits of flat to a concave of 0.0003".

It does NOT say that the concave (if there is one) must be a true curve hence it would be unwise to assume that it was/is. The concave could be an irregular curve or a set of positive and/or negative flats - with the 0>0.0003" limits.

If that is the case I would assume that the level was only accurate (and "truely parallel) to the bubble when centred if the level was only bearing or supported on its ends. It would not be safe or prudent to assume that if the level was supported at/on two points each equidistant from the ends of the level that the bubble would be centred.

If it were me, I'd be inclined to scrape the level base flat (even on a new level) using a good (Class 1 or AA) surface plate which is flat to quite high order of accuracy.

For those that may be interested in the standards that John quoted (plus others) see: