[atomgonuclear]

I recently started working on a CNC machine for a computer interfacing class and I was wondering what I should expect as far as accuracy for a homebuilt project. I used smooth drawer slides that are supposedly "high precision" and I went to the local steel mill and got some aluminum for the frame. I'm using a print mech and rails from an old printer. I kept everything as perfectly perpendicular and parallel as I could. I beleive im within 1/32th of an inch accuracy. When im finished is there a program or some type of way to determine how accurate it is and how I can improve it??

If anybody has any help or would like to share their homebuilt's accuracy and how they measured please post.

I will post some early pics later.

[NC Cams]

What do you mean by accurate??? What are you looking for???

In our shop, we have to work to tolerances of 0.0001" or better. Thus 1/32" is over 300 times too much tolerance. If you're doing wood working for framing construction, this is plenty close enough.

You can probably figure within 1/32" as mentioned in your post PLUS whatever other errors creep up in your tolerance stacks ( IE: slop due to clearance, backlash and/or any lead error in your drive system(s) etc). Basically, it all adds up....

Until, or unless you machine and or tram everything into place and use precise ball screw drives, you probably can't expect much better.

[swarfmacdaddy]

Someone posted a g-code file for a ruler on here awhile back. That might be a good place to start for checking how accurate you machine will be. Sorry, but i leave it to you to find said file. Enjoy!

[mkaake]

What do you mean by accurate??? What are you looking for???

In our shop, we have to work to tolerances of 0.0001" or better. Thus 1/32" is over 300 times too much tolerance. If you're doing wood working for framing construction, this is plenty close enough.

You can probably figure within 1/32" as mentioned in your post PLUS whatever other errors creep up in your tolerance stacks ( IE: slop due to clearance, backlash and/or any lead error in your drive system(s) etc). Basically, it all adds up....

Until, or unless you machine and or tram everything into place and use precise ball screw drives, you probably can't expect much better.

your shop tries to keep accurate to 2.5 microns? or less?

Sorry, but you'd be fortunate to keep your tooling to a 3 micron tolerance (assuming you're checking it optically), much less machining.

[NC Cams]

We're not checking it optically. We measure to 0.000001" with electronic encoders/counters and grind to 0.0001 tolerances on cam profiles regularly and this is done via Heidenhein encoders.

Granted, we may be +/- 0.001" in absolute size but we got to extremes to control the 2nd derivative of motion curve. To do this, you MUST control the rate of change of acceleration to 0.0001"s. Takes well maintainted grinder and strict grinding procedures to do this.

Our cams have won NASCAR 'Cup races many times over the past 5 years - we like to think that we know what we are doing.

We also grind cams for a number of automotive OEM's (prototypes) as well as a world reknown automotive engineering firm in Detroit as well as OEM suppliers. No customers have espresse dissatisfaction with our work or have challenged our accuracy.

The 0.0001' flats mentioned previously were measured on our cam measuring machine (it measures to 0.000001" electronically, feel it is demonstrably accurate to 0.0001)

[mxtras]

Mkaake -

For those accustomed to precision machining, .001" is a mile.

It is quite easy to machine within .0004". In terms of precision grinding, this is a mile.


Atomgonuclear -

As far as home built precision - there is no reason that you can not attain pretty good precision with a home-built machine but it is not likely, not common and it is a lot more expensive than you would likely believe. Even achieving .003" accuracy, which would be considered pretty good for a home built machine in my opinion, requires a solid structure, fine machining and near perfect alignment of precision motion elements. Drawer slides are not precision elements in terms of machine tools but sure do beat the heck out of doing things by hand or with no machine at all.

Keep us up to date.

Scott

[Geof]

....It is quite easy to machine within .0004"....Scott

this is a question a bit similar to the one posted a while back about what is a tenth; 0.1 or 0.0001.

When you say within .0004" do you mean: SIZE +/-0.0004" or SIZE +/-0.0002"?

Which is the commonly accepted meaning among the cognoscenti?

[Tinmuk]

When a machinist says "a tenth", he means .0001, when a carpenter says "a tenth" he means .1.

We hold +/-.0001 on our Citizen L20 swiss all day long in 303, 440, 316, etc. some jobs we have +.0000-.0001. On one assembly we did for Raytheon we had a shaft/bearing/housing assembly with 60 millionths total runout.

Every one of ours bearings made in the last 18 years has not varied more than .0001 total in ID or OD. we currently have over 600,000 in stock. Yes, it is possible to hold these kind of tolerances with the right machines and the right process.

[mxtras]

this is a question a bit similar to the one posted a while back about what is a tenth; 0.1 or 0.0001.

When you say within .0004" do you mean: SIZE +/-0.0004" or SIZE +/-0.0002"?

Which is the commonly accepted meaning among the cognoscenti?

I think it depends on the industry you are in. To me it means size or feature location and/or profile shape and/or relationship within .0004" (0.01mm) maximum deviation from the design.

Scott

[Mcgyver]

We hold +/-.0001 on our Citizen L20 swiss all day long in 303, 440, 316, etc. some jobs we have +.0000-.0001. On one assembly we did for Raytheon we had a shaft/bearing/housing assembly with 60 millionths total runout.
.

shoot me if i've slipped a decimal, but if your tolerance is better than a tenth, what's so special about 60 millionths? 100 millionths = 1 tenth, right? 60 millionths would pretty much be your standard deviation

just being a smartass, anyone who works to millionths has my respect

[Geof]

....just being a smartass, anyone who works to millionths has my respect

Considering you do not put an upper or lower limit to how many millionths I would say you are spreading your respect pretty widely.

[NC Cams]

In our case, we do onesy and twosy pieces - thus, we must have 100% of our parts made right. Thus, this issue of accuracy of parts per million is a sheer falacy to/for us. We inspect 100% before we ship and RARELY scrap out a part - a worker who does is made to realize that his screw up IS/WAS costly. Result:they take pride in NOT screwing up.

We've had the QS9000 and the other questions re: std deviation, this coefficient or that posed to us. When we explain that our procedures don't fit the ppm scenario, and inspect 100% to verify quality, we've never had an issue with the Q/C depts.

Besides, their argument falls appart when you hit them with the "how do you do a size distribution study when you are working with a statistical sample of one???" inquiry.

We don't do aerospace work but NASCAR Nextel stuff is close - failures are QUITE visible and not tolerated. If you are shooting for a +/-0.0001" total tolerance spread, you creep up on the size. Simple thermal expansion can make the difference in a part being in or out of spec.

Thus, if you "hit' size at +0.000060" (6o millionths) of nominal, you stop grinding, read size, lett cool/thermally stabilize and read again, redo/regrind if required.

You also have to condider what your part is being used with. Thus, if you are making an aluminum widget that fits into a steel whatsits, the RELATIVE size at operating temp is more critical than ABSOLUTE size. YOu also have to know if the part is going to be thrown together or custom fit by the installer.

All the stuff we do is custom fit by user. We still shoot for/hit the nominal size when we grind.

When Mr. Honda was supposedly told by a Q/C guy that he was wasting money trying to hold better quality than 2-3ppm, his reply was, "Yes, that is true but the 2 or 3 customers got the bad parts out of the million who bought my product had a 100% defect rated which is unacceptable in any Q/C program...."

I'm sure the discussion will rage on.

BTW, ball bearings are made in the millions and they control internal bearing clearances within a few microns day in and day out.

It is amazing but if you have to hold a particular tolerance and the customer or market demands will pay you to do it, you find a way - or else go out of business because if you don't, someone else will.

[Mcgyver]

Considering you do not put an upper or lower limit to how many millionths I would say you are spreading your respect pretty widely.

I don't care how big the part is but if the resolution of measurement is in millionths that does it for me

, the RELATIVE size at operating temp is more critical than ABSOLUTE size. .

yup, witness the model engineers who have made marvelous working IC engines without the benefit of even a mic - equiped with treadle lathes and calipers

[mkaake]

I stand corrected

I was under the impression that you had been talking about a home built machine...

In any event, I work for an auto manufacturer, and have been looking at the tolerances that we keep in production - 10 mics is about what we normally hold (that's pretty easy on precision equipment, I know). But what really got me was watching a tool setup a few weeks ago where the x axis moved 3 mics +/- if you accidentally touched the machine during the setup procedure...

Anyhew, didn't mean to stir up any pots, just thought you were talking about home machines. And yeah, it's very hard to run statistics with a sample size of 1 or 2 parts

[NC Cams]

The auto industry is a HUGE driver for precision stuff at dirt cheap prices.

I interviewed at an noted aerospace hydraulic component supplier. He looked down on me since I worked in auto industry previously - his stuff was so hotsy totsy - at least in his mind.

When I mentioned the precision fits of a hyd lifter (at the time in millionths), he said you couldn't make a part like that for under some ungodly sum of money ($25 or $30 or some other absurd price). He liked to you know what when he learned that we SOLD the parts wholesale for $1.25 each at the time and made money doing it.

He also couldn't fathom how we could run valve springs at the stresses and travels we did - his stuff was marginal at a fraction of what they were doing at the time.

At that point I learned that aircraft inspects quality into their parts - the auto people (even with the so-called lame quality they are alleged to have today) build quality in.

ONce you get the setup and have equipment that will hold the setup and tolerances, it is pretty easy to make jewelry. You'd be amazed at the construction and assembly of ball bearings. True poetry and a symphony of motion and Q/C, day in and day out at hundreds of pieces per hour.

YOu can get home machines to be deadly accurate if you spend the time and money. Heck we got a 10 uear old Bridgeport to darn near match the accuracy of a $1.25 million CNC grinder while master milling with TLC and some bucks spent wisely.

[mkaake]

Oh, you should watch a crank being ground on a three spindle machine - one for the crank, and two following with massive wheels to grind the pins... awesome.