Flatness (non electronic) anyone here know ? or where to go ?

[Simon]

I'm having a discussion with a colleague (Quality inspectors). We are trying to define flatness tolerancing.

Most tolerances are stated as a +/- figure or there will be a specific + and - size allowed. but in the case of flatness (NOT parallelism) it is just a number. so is this number the maximum you can measure from any two furthest points or is it the maximum that the surface can deflect to either way from a given point. As the point of "reference" is never given but the whole surface is refered to I'm guessing it is the maximum deviation between any two extreme points ?

googling has not turned up a reliable explanation. Can anyone otherwise recommend an 2engineering2 type forum ?

[Rerouter]

i would imagine flatness would be the tolerance between any 2 points on the surface of the object,

here is a link i feel is relevent http://www.emachineshop.com/machine-shop/Flatness/page591.html

and if not that, then i can only think in the simplest way, get a hard straight piece of metal, and lie it across your surface, and make sure the gap  (vertical gap between the 2 points) never exceeds your tolerance,

[Simon]

and if not that, then i can only think in the simplest way, get a hard straight piece of metal, and lie it across your surface, and make sure the gap  (vertical gap between the 2 points) never exceeds your tolerance,

yes that is sort of what i do, but I need to know that i'm talking the same language as the supplier and customer. i had a load of goods returned that are marked as 3mm out of flatness, I measured them corner to corner and got 6mm, so i think they are incorrectly assuming it is +/-3mm - thats fine by me and gives me more lee ways between the manufacturer and final customer.

[Simon]

here is a link i feel is relevent http://www.emachineshop.com/machine-shop/Flatness/page591.html

Ah yes i did find that, i think it tells me i am right 

[Rerouter]

also, with the 6mm issue, where you measuring with a flat edge, or from the base to top of each corner, if with a edge, then if it was 3mm out at the middle, it would appear 6mm out at the other corner  while only being 3mm out

(sorry if that sounds silly, but i myself have made worse mistakes, and just want to confirm)

[Simon]

the item in question is a large radiator. so i put 4 machined blocks on a surface table and place the radiator on top so that there is a block on each corner. I then rock it to see which way it is out. press down one corner and measure how much it lifts at the opposite corner, so i have a measurement that is the maximum between the two most out points.

what our customer seems to have done is to have halved that thinking that the flatness figure is a +/- figure, so when they measure 6mm they decided it was +/- 3mm - I do not think this is right but as i said it does work in my favour because if they have put 2 on the drawing and think this is +/-2 then that actually gives me 4mm by my method of measuring.

[Rerouter]

yeah, i think you could get away with that due to the technicality of the measurement with no fixed point of reference, (makes things sound better)

to compare to an electrical analogy its like saying +-10V without specifying ground

[EEVblog]

Maybe try:
http://en.wikipedia.org/wiki/Surface_plate
which leads to:
http://www.tru-stone.com/pages/faq.asp

3) How is surface plate flatness defined and specified?
A) Flatness can be considered as all points on the surface being contained within two parallel planes, the base plane and the roof plane. The measurement of distance between the planes is the overall flatness of the surface. This flatness measurement commonly carries a tolerance and may include a grade designation.
The flatness tolerances for three standard grades are defined in the federal specification as determined by the following formula:

Laboratory Grade AA = (40 + diagonal squared/25) x .000001" (unilateral)
Inspection Grade A = Laboratory Grade AA x 2
Tool Room Grade B = Laboratory Grade AA x 4.

For standard sized surface plates, we guarantee flatness tolerances which exceed the requirements of this specification. In addition to flatness, Federal Specification GGG-P-463c address topics including: repeat measurement accuracy, material properties of surface plate granites, surface finish, support point location, stiffness, acceptable methods of inspection, installation of threaded inserts, etc.

So yes, it's basically the difference between the peaks.

Dave.

[Simon]

@ rerouter

precisely, the spec is just a number and is for a 2 dimensional surface. My argument to my colleague is that if it is a +/- dimension then I need a point on the 2 dimensional surface to be specified as a datum to measure from but all the info around says that flatness does not have a datum it's sort of the surface to itself so supports my theory that is is simply the biggest deviation to can measure between two points

[robrenz]

Flatness as used in ANSI standard dimensioning and tolerancing is:
Any point on the surface be within two paralllel planes separated by the tolerance value.

Your setup on the surface plate and the measurement you took is correct.

[HLA-27b]

In terms of machine tools there are two terms associated with flatness. One is the surface roughness measured in geometric average Ra. This of course applies to any surface including curved ones.

The real flatness is a bit weird because it changes with scale. Something which is flat at one scale may not be flat on smaller or larger scales. When there is a known scale and a known datum plane, the flatness may be quantified as deviation from the datum plane in say millimeters.

Another approach is to mate the part to be measured to a reference flat with engineering blue on it. The bluing transfers to the measured part at the highest points only, thus giving an indication of the surface flatness. By scraping the high spots it is possible to achieve an extremely flat average albeit with highish Ra. The quantification is simply the count of contact points per square inch. The count is also expected to be uniform across the entire surface.

It is possible to use scraping to generate flat surfaces without any external reference simply by averaging three surfaces to one another. Scraping is a manual skill rarely possessed by average engineers or machinists. Those who poses it are as rare as electronic engineers able to make their own ICs. 

Another technique to asses flatness is the use of optical flat to generate interference rings between the flat and the measured surface. The density and the shape of the interference rings is a function of the flatness ant the angle between the surfaces. This is generally limited to the scale of hand held objects. The precision can be sub wavelength.

In land surveying the instrument of choice is the optical level. The flatness or more commonly lack thereof can be indicated as deviation from level in meters (or other units). The commonly achievable accuracy is +-1mm over a kilometer. With more involved technique it is possible to achieve +-0.001mm or better over shorter distances.

[robrenz]

It is possible to use scraping to generate flat surfaces without any external reference simply by averaging three surfaces to one another. Scraping is a manual skill rarely possessed by average engineers or machinists. Those who poses it are as rare as electronic engineers able to make their own ICs. 

HAL-42b,  nice overview of flatness.  As you say it depends on the scale you are working with.  32 years ago I wanted to learn how to hand scrape surfaces. So I bought (3) .5 meter square cast iron surface plates and proceded to use the whitworth 3 plate method to establish the flat plane.  After 6 months of scraping during break time and lunch at work I achieved 15 microinch flatness as checked with an autocollimator. It is a invaluable skill that has served me well through the years.

[HLA-27b]

HAL-42b,  nice overview of flatness.  As you say it depends on the scale you are working with.  32 years ago I wanted to learn how to hand scrape surfaces. So I bought (3) .5 meter square cast iron surface plates and proceded to use the whitworth 3 plate method to establish the flat plane.  After 6 months of scraping during break time and lunch at work I achieved 15 microinch flatness as checked with an autocollimator. It was a invaluable skill that has served me well through the years.

Something which wil never ever be available cheaply

[FreeThinker]

It is possible to use scraping to generate flat surfaces without any external reference simply by averaging three surfaces to one another. Scraping is a manual skill rarely possessed by average engineers or machinists. Those who poses it are as rare as electronic engineers able to make their own ICs. 

HAL-42b,  nice overview of flatness.  As you say it depends on the scale you are working with.  32 years ago I wanted to learn how to hand scrape surfaces. So I bought (3) .5 meter square cast iron surface plates and proceded to use the whitworth 3 plate method to establish the flat plane.  After 6 months of scraping during break time and lunch at work I achieved 15 microinch flatness as checked with an autocollimator. It is a invaluable skill that has served me well through the years.

Yikes! The Whitworth 3 plate method .. Shudder! Brings back memorys of my apprenticeship days and my 1st Year in the training school. Well done on the finish, I can only take my hat of to you and wonder Why? (I'm Electrically biased so forgive me Lol)

[Simon]

well our machine shop supervisor decided he was going to show me how to grind a surface table (it had been left to rust) but time never allowed and he lost interest. My main job is to just measure stuff. The surface table I need for these units is a battered old table that has very poor flatness as surface tables go but fine for my measurements that are down to 1/10 mm

[robrenz]

I have had a Dave Jones level of Passion about machining and precision since I was 12.  I knew that mastering hand scraping was a necessity for the type of work I wanted to do.  Here are some pics of one of the two I still have. High quality scraping is having approximately 40 bearing points per inch.  These plates are well over 100 points per inch. At this level the bearing points are burnished to where I can read the fringes of an Optical flat on the surface plate.

[HLA-27b]

100 points per inch is not something achievable with engineering blue I presume, so you must have been reading the famous book from Moore Nanotechnology 

Also don't hesitate to show what an "assumed flat" surface looks like when mated to a real hand generated flat. Young whipper snappers have to learn somehow.

[robrenz]

100 points per inch is not something achievable with engineering blue I presume, so you must have been reading the famous book from Moore Nanotechnology 

Also don't hesitate to show what an "assumed flat" surface looks like when mated to a real hand generated flat. Young whipper snappers have to learn somehow.

Actually it is achievable with very precise application of very dry bluing. You cant use the runny stuff out of the tube. And you need an equally fine tuned red lead.  I have done micro scraping at well over 150 points per inch for some instruments, picture below. The first few pages of The foundations of mechanical accuracy by Moore was what inspired me to do the 3 plates. That is an awesome, indispensable book and I still have my copy.  The last I checked it was selling for upwards of US$350.00. I have been to Moore many times and was responsible for the selection, purchase and operation of a Moore #3 universal measuring machine for my last employer many years ago.

[HLA-27b]

I have been to Moore many times and was responsible for the selection, purchase and operation of a Moore #3 universal measuring machine for my last employer many years ago.

I'm green with envy

A machining shop I used to order parts from had a SIP with optical readouts sitting in the corner,obviously acquired from a bankrupt predecessor. The lumberjack at the helm had figured out how to use it for basic milling tasks. His comment: "Its quite accurate for such an old mill (!!!) but I wish boss would sell it for something more functional like a knee type or sutin"