View Full Version : Checking large plate flatness. HSM workarounds?

08-23-2010, 12:39 PM

08-23-2010, 12:53 PM
See this method I developed.


08-23-2010, 12:57 PM
So I have been wanting to check the flatness of a 32"x45" ground steel plate. Budget runs around $300 max.

How accurate? Optical interferometry type of technique like the home telescope mirror grinder guys use?

How bout a master precision level moved around in a grid pattern combined with lots of surveyor mathematics techniques?

Pretty darn cheap if you already have a precision level and many known thickness shims to align your lathe.

I've seen new import master "precision" levels for like $150. I'm thinking $150 for the level, maybe $100 to contract out to someone like Mr Evan to explain how to do the surveying math and significant figures calculations (I'm not even going to try to explain it, although I could probably hack thru it if I had to) and $50 for headache pills and beer afterwards?

08-23-2010, 12:57 PM

08-23-2010, 12:59 PM
How accurate?

08-23-2010, 01:10 PM
Ideally, I would like to know to .001" accuracy; however, in this case, budget may dictate accuracy moreso than my wishes.

First thing when I google for master precision level is

"Value 639-3618 8''/0.0005'' Per 10'' Master Precision Level" for $154

Your budget and your wishes appear more or less compatible, assuming you'd want to do the "MPL surveying thing".

You "need" a MPL anyway to align your lathe. Easiest way to see if you've twisted the ways is to level it "short way" at headstock and at the tailstock end. Also you can get the headstock and tailstock aligned dead center vertically by sticking in a test bar and getting it level "long way" on the ways and on the test piece.

Folks whom ask why you'd want to level a lathe, forget that two things that are both perfectly level, also happen to be perfectly parallel......

08-23-2010, 01:10 PM

08-23-2010, 01:14 PM

08-23-2010, 01:18 PM

08-23-2010, 01:19 PM
As shown the test doesn't identify the area that isn't flat. That can be checkd by masking the plate using a very thin fence of just about any material to shadow all but the area of interest. The salt test will inform you just what area is being tested.

08-23-2010, 01:25 PM
Ideally, I would like to know to .001" accuracy; however, in this case, budget may dictate accuracy moreso than my wishes.

Wow that's a challenge. sounding like a broken record here....getting something that large flat is not trivial...and that is a VERY large surface. Even the guy who blanchard ground it or however it was prepared would have to have been very good to achieve this flatness - without careful shimming the piece would been sucked down on the chuck and pulled flat, then ground, only to spring back after being released.

Going from a straight edge, which is very narrow to a large surface would be difficult; I'm not seeing how it would work. Scraping is in a way an inspection process and i've not seen a way to scrape (or inspect) a large rectangle for flatness using a narrow straightedge. there may be a way a way to map it if you used a parallel and a master precision level stepped it over, each time shimming to to level then measuring with feeler gauges, but even Starretts best is only good for .0005" over a foot so that won't work.

Just to maintain .001 flatness over something that size will require quite a support system...I'm imaging how much it could move just under its own weight. Can you give some context to its use?

Short of creative optical ideas like Evan's, the only way I know of to check for flatness is comparing to something flat (unless you want to generate three :eek: ). Fortunately I hear big assed old surface plates go cheaply. Even with a plate it would be a big job; I'd guess you'd have to lower the surface plate onto the work separated by three exact height spacers then indicate the distance between the two surfaces.

08-23-2010, 01:31 PM
How exactly would you extend the test with, for example, a 12" level over the full 40+ inches?

That is that "surveying" bit that probably exceeds my ability to explain in English. If you find an understandable surveying book just use their explanation of "how to be a land surveyor" and disregard my attempt to explain.

Basically you'd break the 40+ inch thing into a grid using perhaps a 12 inch level. Then you draw out on the grid intersections how much shims you had to add at which end to make it level. You're kind of trying to make a topographic map.

So call the upper corner perfect zero. So one length out, 12 inches or whatever, to level it you had 0.020 shim on the edge and 0.021 shim 12 inches in. So, at the 12 inch mark its .001 low of "perfect"

Next measure 12 inches in to 24 inches in. This time it took a 0.025 shim on the 12 inch mark and a 0.020 shim on the 24 inch mark. Superficially it looks like 24 inchs in is 0.005 high. But wait, you've got to add the delta from the 12 inch mark, so its actually only 0.004 high. Err, I think. Remember, headache medicine before its done, beer afterwards.

Just do this, in 2d, across the whole thing. Then figure out the average slope across the width and the average slope across the length, and thats how much, on average, out of level the whole plate initially was. So, you proved that lengthwise your plate is 0.001 lower per foot because its resting on some swarf or beer can pull tabs or whatever under one edge. Now go thru and correct all your numbers again.

So 12 inches in you figured it was 0.001 low, but the whole plate on your bench is tipped 0.001 up so add 0.001 giving zero I guess over the first 12 inches its dead nuts perfect flat. You figured the 24 inches in point was 0.004 high but add that 0.001 per foot correction and its actually 0.005 high (oops).

Eventually you end up with a lovely topo map of peaks and valleys. Now you need a calculator or computer, to tell you the average and std deviation and Students-T test and things to tell if it all matters for your application. This is the fuzzy part that requires the headache medication and/or someone whom understands statistics, or at least can carefully explain it.

08-23-2010, 01:45 PM
How are you going to support the plate while measuring it? If your supporting surface isn't PDF (pretty dern flat) the plate is going to flex and give you erroneous measurements.

Dave S.
08-23-2010, 01:49 PM
How thick is this plate. You are going to have to figure a way to lay it flat so it does not bow or twist under it's own weight.

good luck


I type slower than SGW :)

08-23-2010, 02:21 PM
I have used a Starret 36" straight edge to test something 34" long , just using light and shim stock down to 0.0005" - if the light shows through under the straightedge, see how big a shim you can get under there -it's rude and crude but when you ge to this kind of length, it will at least give you some idea. It takes patience, but is quite feasible.

You can reduce your expenditure by using a 36" across the plate, much cheaper than a 48" - and just slide the 36" across the 45" width and again check for light - I'm sure Evan will know for sure, but I read somewhere that the eye can see light coming through a gap of less than 0.0005".

You can buy a set of plastic shim stock from 0.0005" to 0.030" for about $25, and this is also priceless in the home shop anyway, you'll wonder how you ever did without them .

Although very old school, this works fine for 0.001" error, it gets a bit harder if you try going smaller than that. You will also have to take more seriously how this plate is mounted, as unless it is a few inches thick, it will conform slowly to whatever surface it is sitting on -just think of it as very hard rubber :-)

Of course, there is another problem you haven't yet figured -what do you do if there are more errors than you'd like -sometimes it's better not to know :-)

Richard in Los Angeles

08-23-2010, 02:39 PM
What if you put it on the mill table with an indicator in the spindle? You could shim the ends to get an equal reading on both ends and both sides, then see how much the indicator moves between those points. If your mill table sags too much going side to side, maybe do it on a surface grinder or lathe. I have a CNC bed mill that would work fine for that. If you don't have a big enough machine, maybe someone in your area does. It wouldn't take long to make a few passes over the surface with a 1/10 indicator. For the lathe you could put an indicator on a bar sticking out of the spindle. Let the plate ride on the cross slide.

Michael Hall
08-23-2010, 02:44 PM
These come up on Ebay once in a while:


surface plate calibration products, planekator, rahn planekator Federal Spec GGG-P-463c dictates that the proper methods for determining the overall flatness of a granite surface plate. In section 4.5.9, it states, “A…system of diagonals, perimeter and center lines may be calibrated on plates, and especially on small ones, by using a precision calibrated straight edge, adjustable supports and a suitably mounted indicator…”. This is a basic description of the Planekator.

The Planekator kits consist of a granite straight edge, a set of supports (one fixed height and one adjustable), a calibrated .000020” dial indicator with indicator carriage and a storage case. The straight edge is supplied with a correction tape.

When selecting your Planekator, consider the size of the plate(s) you will be inspecting. The straight edge should be at least equal to the full width of the plate and at least equal to 50% of the length of the plate. For example, a 36” Planekator straight edge can be used to calibrate any surface up to 36”x72”.

surface plate calibration products, repeat-o-meter, repeat o meter, rahn repeat-o-meter, surface plate measurement accuracy The Repeat-O-Meter is used to scan visually the work surface of the plate. The range of the indicator readings shall not exceed the values shown in Table II of section 3.3.3 in Federal Specification GGG-P-463c. The Repeat-O-Meter performs the same function as a height gage and gage block, but is more dependable and easier to use.

In a short amount of time, you can check the repeat measurement accuracy of a surface plate by moving the Repeat-O-Meter over the work surface, noting the full indicator range (F.I.R.). In just a few more minutes, by zeroing the R-O-M at the center along a diagonal or centerline and then recording readings at 5” intervals, you can chart the readings from the plate. The difference between the greatest plus and minus readings is the maximum measurement error possible on the plate.


08-23-2010, 02:47 PM

08-23-2010, 03:21 PM
.001 on a bicycle frame??? Your joking, right? A bicycle frame will flex more than they lying down vs upright. Not to mention putting a load on it.

I have seen good quality bikes being built and their fixtures are nowhere near this precise. The choice of material was poor and the support is wrong as well. You could find a used granite slab for the price of what you are willing to spend to check the plate.

This is a case of overspecing. You have one too many zeros here.

But if it is really that big of a deal you can have someone come out and check it for about what you are willing to spend.

08-23-2010, 03:27 PM

08-23-2010, 03:29 PM
For it to sit "flat" on any surface, the plate should be supported at only three points. As already stated, a granite surface plate (even a used one) would have been a more accurate, reliable and probably cheaper solution for a reference surface.

08-23-2010, 03:35 PM
Alright. Thank you all for your replies.

Forrest Addy
08-23-2010, 03:37 PM
My reading of the photo suggets that Blanchard ground plate is smooth and probably parallel within 0.001" but most likely not very flat. Not unless the operator was careful and flipped it several times shimming to eliminate error. The usual Blanchard grind job on a flat plate involves a cleaning up both sides. That plate is thin, probably 1," less the grind stock. I would be amazed if it's flat in its free state within 0.030". Right now the plate is constrained by the stand on which it lays. The plate's overall flatness can be considerably improved by shimming the feet so the plate level all around. Whether there are humps and hollows elsewhere you'll have to check. Here's from something I wrote iome time ago.

I've checked lots of plates for flat using the laser, straight edge, the tight wire, etc. The simplest method by far is by using a plain old precision level. If you have precision machine tools a 10 arc second level comes in mighty handy at the damnedest times. I suggest an 8" long import 0.0005" per 10" level from any of the usual lower end import houses for aout $80. It's not a Starrett 199 but believe it it or not the cheapos are pretty good right out of the box. You should be prepared to adjust it should you get one. They are easily adjusted if a nice clean machine tool table is handy.

When checking plate for flatness you wll be plagued by the effects of heat and deflection. The plate should be thick enough not to deflect significantly under the weight of the level or by its own weight when supported at its Airey points. For a 36" x 48" plate I would think that minimum thickness for acceptable reistance to deflection for loads under 30 lb would be 1 1/2". This is just a guess unsupported by deflection calculations. As for heat, avoid settings where temperature can vary.

Finding the support points. The Federal Standard describing the location for support for granite surface plates is somethng like this: draw a rectange 1/2 the size of the plate's outline centered on the surface. In this case the rectangle will be 18" x 24". Place two supports (screw jacks) under the corners at one end. Find the center of the other end (18") and place a jack under it. There are not the exact Airey points but they are close and more importantly conform to the standard.

If you don't have a rigid surface (a wood bench is NOT rigid) set up directly on the concrete floor. You may wsh to raise the set-up on concrete blocks for the convenience of you knees. Adjust the jacks to level the plate by centering the level between any two supports. Fiddle with the adjustments until you get center bubble. Allow the plate to reach thermal equalibrium for several hours - maybe overnight.

Checking for flatness. Double check the set-up for level and tweak if necessary. Find a mass roughly the same weight as the level. Imagine a line between the single end support and the nearest corner. Place the level on that line at the corned. Note the reading. Place the mass along side the level. Note the reading. Subtract. This is the the worst case deflection of the plate by the level.

Divide up the plate surface like the British flag with margins, diagonals, and centerlines. Make a drawing to note readngs on. Working sytematically take level readings along each line noting the bubble readings with "+" and "-" to represent slope one way or the other along every edge along each diagonal and centerline. Make the first run-through a quick survey for general observation and practice. Check the level of the set-up from time to time.

Once your are practiced make a series of readings for permanent record. Note the readings as above. If your body remains in one position, you have a filament light bulb shining on the plate, or there is some other source of diffuse heat do not be surprised if your readings are slightly inconistent. If you place your hand or lean your forearm on the plate, subsequent readings may be suspect. A thin flat plate krinkles up like a potato chip on a microscopic scale if subjected to localised heat.

Save your readings. They are a map of the plate's errors. If you have to perform work close to the plate's accuracy, knowledge of the error will help you compensate for it.

All the above may sound like a discouraging amount of work but in reality it only takes a couple of hours.

Robin R
08-23-2010, 03:43 PM
You're close to Shars, this should tell you to within 0.0004". http://www.shars.com/products/view/2346/Grade_B_36quot_x_48quot_Black_Granite_Surface_Plat e

08-23-2010, 03:45 PM
Stick a piece of 2" sched 40 pipe in the lathe and turn it round. Take a couple of finish passes of only a few thou to minimize any deflection from tool forces and call it good. Whatever length will fit is good enough. Chop off a few inches from both ends and chamfer the ends slightly. You now have a cylindrical rule that is as straight as your lathe. Taper makes no difference since it isn't going to serve as a square. As long as your lathe is at all reasonable and the tooling is very sharp the pipe should be cut pretty straight. If you have any doubt then make two the same and compare them to each other.

08-23-2010, 03:54 PM
Just as a cursory check, you could put an indicator on a surface or height gage. With a straight edge resting on the surface of the plate, hold it 90 degrees to the surface with an angle block, then indicate across the top of the straight edge. The indicator will show any variation in height between the surface of the plate and the top edge of the straight edge.

loose nut
08-23-2010, 07:56 PM

What is with this , am I missing something.

08-23-2010, 08:12 PM
What is with this , am I missing something.

Seems like Mr. Marks is pretty proud about being able to count from 1 to 10.

Hey, my 3 year old grandson is too...I can identify with that...:rolleyes:

I too missed the point...

08-23-2010, 09:11 PM
???? Wondering the same thing ???

Mr. Arthur Marks, this whole thread doesn't do me as much good for educational purposes without your original question and subsequent follow-up
comments / questions. Why delete them?

Seems like Mr. Marks is pretty proud about being able to count from 1 to 10.

Hey, my 3 year old grandson is too...I can identify with that...:rolleyes:

I too missed the point...