This is gonna be messy... Please Help!

[Mike Johnston]

Ok... First of all, thanks in advance for any help I can get... I'm in a crunch and I really have my back against the wall.

A little history... This is a jet channel cover that assembles to a reservoir body for the GM H3 fuel system. We originally had an outside metrology house who knows their stuff... analyze the "profile" of the part and there's something that neither I or my fellow engineers understand.

Our "outside metrology house" is reporting the "T" Values for the profile. They collect the data of a scan of the part and then in Excel report the worst "T" Value conditon multiplied by 2 and report that value as the profile.

My question is this... First of all, (pardon my ignorance) what in detail are the "T" Values? I don't understand how this value is calculated. Any graphical examples would be greatly appreciated.

I can provide more details if needed.... (ie)... Program details and coordinates..

You guys(and gals) are great! I know this is a little vague in question so please let me know if you need more details....

Mike

[joe.sha]

Go to the thread titled "Sheet Metal Axes:T" posted on the 9/14/2006, Peter attaches a really good explanation of "T" values and how to calculate them. Can you attach what your print calls out for?

[Matthew D. Hoedeman]

"T" axis is the (T)otal vector deviation (3-D) of the vector (or surface) point taken as opposed to the X dev or the Y dev or the Z dev, none of which tell you the (T)otal deviation from perfect.

[Mark Farren]

As posted by Matt Hoedeman, this formula for calculating the "T" value..... SQRT(X DEV^2 + Y DEV^2 + Z DEV^2)
(EDIT: Guess you type faster than I Matt!!!)

"T" is the amount of error along the vector from the nominal.

Also included is an attachment, I believe from B&S, with the "T" value calculation explained since according to some, the above method is not accurate, but it came from Matt, so I got faith in it!

[Jeffrey Lovely]

Not sure why they times it by 2

lets say lowest reading id -.2
highest reading + .25
profile is .45
not .50

[Peter Warcholyk]

Multiply by 2 ??

Hello Folks,

First of all I do understand "T" values.

The part that I do not understand concerns the multiplying by 2 the "T" value

Is anyone able to simply and clearly explain as to why this is done, perhaps even with a diagram?

I don't understand the purpose of doing this. Please help.

Many thanks in advance !!

Best Regards,
ZydecoPete

[Matthew D. Hoedeman]

If you have a tolerance of +/-0.75mm, then you have a total, zone of 1.5 centered about the nominal. If your worst reading is +0.65mm, then you would need a minimum of 1.30mm of tolerance zone, centered about the nominal, and this is what they are telling you by doubling the worst reading, the minimum profile tolerance needed for that part.

[Jeffrey Lovely]

T = Profile

B = BEER

[Matthew D. Hoedeman]

By multiplying the largest deviation by 2, it tells you the 'required' profile tolerance required to 'hold' the part you just checked, as long as the tolerance is an equal split between + and -. It will not work if you have a unilateral tolerance or some other split (say +1.0 / -0.5). This is why I ALWAYS just simply use the T with the +/- tolerance, I NEVER report a profile tolerance of a set or scan or group of points, I always dimension them as individual points, XYZ & T, with the T being the value that actually means something. THe XYZ simply tell you WHERE on the part the point is while the T tells you how it checked.

[Mark Farren]

W.M.S. (couldn't get all caps w/o the periods, can I say that word here?? )

[\v/inston]

As posted by Matt Hoedeman, this formula for calculating the "T" value..... SQRT(X DEV^2 + Y DEV^2 + Z DEV^2)
(EDIT: Guess you type faster than I Matt!!!)

"T" is the amount of error along the vector from the nominal.

Also included is an attachment, I believe from B&S, with the "T" value calculation explained since according to some, the above method is not accurate, but it came from Matt, so I got faith in it!

in the illustration shouldn't the nominal point be on the surface instead of below the surface?

[Matthew D. Hoedeman]

As posted by Matt Hoedeman, this formula for calculating the "T" value..... SQRT(X DEV^2 + Y DEV^2 + Z DEV^2)
(EDIT: Guess you type faster than I Matt!!!)

"T" is the amount of error along the vector from the nominal.

Also included is an attachment, I believe from B&S, with the "T" value calculation explained since according to some, the above method is not accurate, but it came from Matt, so I got faith in it!

Well, another point, the SQRT(Xdev^2+Ydev^2+Zdev^2) will really only work correctly in 2 cases:
Case #1) Your machine is so dang tight and accurate, it can actually drive the vector 100% correctly, everytime (yeah, right)
Case #2) You use the handy SNAP POINT feature which 'forces' the actual point back onto the perfect vector line of the point being checked to simulate a machine that can run absolutely 100% (to the split micro) on the vector line of the point.

If you are NOT using SNAP POINT and your machine drifts even a couple microns from the perfect vector line on the approach to the point, then the formula will not 'add up' to what Pcdmis shows. This is a VERY GOOD reason for using SNAP point. ANother reason is, you have a customer who wants XYZ values for every point you check as well as the (T)otal deviation for each point. So, you are repoting XYZT for each point. Now, you just happen to have a point with a sqaure vector, lets say, 0,0,1. Now, unless your machine CAN run absolutely perfect, it will not hit the part at the EXACT XY nominal location to get the Z reading. So, you will have a deviation of 0.025mm in X and -0.018mm in Y and 0.235mm in Z. The (T)otal devation will be 0.237mm if you are NOT using a SNAP point and 0.235mm if you ARE using SNAP point. So, if you are not using SNAP point and your customer caomes in and see a deviation in X and Y for a point that is sqaure in Z and says, "How can you be off in X and Y when it is sqaure to Z, you should only have deviation in the Z axis?" and then YOU have to tell him, "Well, er, the machine can't go to the absolutle exact location and you are seeing the amount it misses by." and since they know nothing about CMM's, they start having a fit about how you and your lab are not capable of doing the job needed and yada, yada, yada, yada, yada and so on and so forth. So, USE the SNAP pointoption. Sure, this example is a little extreme, but when you have to deal with idiot customers (and we all will at some point or another) it is best to find ways to keep this sort of thing from coming up BEFORE it can come up.

My $3000 worth, or whatever the SMA is these days.

[\v/inston]

Well, another point, the SQRT(Xdev^2+Ydev^2+Zdev^2) will really only work correctly in 2 cases:
Case #1) Your machine is so dang tight and accurate, it can actually drive the vector 100% correctly, everytime (yeah, right)
Case #2) You use the handy SNAP POINT feature which 'forces' the actual point back onto the perfect vector line of the point being checked to simulate a machine that can run absolutely 100% (to the split micro) on the vector line of the point.

If you are NOT using SNAP POINT and your machine drifts even a couple microns from the perfect vector line on the approach to the point, then the formula will not 'add up' to what Pcdmis shows. This is a VERY GOOD reason for using SNAP point. ANother reason is, you have a customer who wants XYZ values for every point you check as well as the (T)otal deviation for each point. So, you are repoting XYZT for each point. Now, you just happen to have a point with a sqaure vector, lets say, 0,0,1. Now, unless your machine CAN run absolutely perfect, it will not hit the part at the EXACT XY nominal location to get the Z reading. So, you will have a deviation of 0.025mm in X and -0.018mm in Y and 0.235mm in Z. The (T)otal devation will be 0.237mm if you are NOT using a SNAP point and 0.235mm if you ARE using SNAP point. So, if you are not using SNAP point and your customer caomes in and see a deviation in X and Y for a point that is sqaure in Z and says, "How can you be off in X and Y when it is sqaure to Z, you should only have deviation in the Z axis?" and then YOU have to tell him, "Well, er, the machine can't go to the absolutle exact location and you are seeing the amount it misses by." and since they know nothing about CMM's, they start having a fit about how you and your lab are not capable of doing the job needed and yada, yada, yada, yada, yada and so on and so forth. So, USE the SNAP pointoption. Sure, this example is a little extreme, but when you have to deal with idiot customers (and we all will at some point or another) it is best to find ways to keep this sort of thing from coming up BEFORE it can come up.

My $3000 worth, or whatever the SMA is these days.

i'm glad you posted this, i never really use snap point but will from now on and one nice thing is global update has an option to turn it on and it works so you don't have to go thru every point to turn it on.

matt can you check and see when our sma is due for your tech support it seems like it been more than a year since we paid you.

[Matthew D. Hoedeman]

hahahahahahahahahahahahahahahahahahaha


Actually, I think you are 3 years behind schedule, but I will give you a 50% break for everything over 2 years, so.......