How accurately can you machine?

I do work for the REMAP charity. Some years ago we had a big project (a very special wheelchair) that required several people for it's implementation. One of our members, a professed AutoCad expert, decided to do the drawings for us. He produced dimensions accurate to 0.0001" (there was no way we could work to that level). Some way through the actual build, one of the parts didn't fit - it was 1.0000" too big (but otherwise accurate to 1/10 thou). He had committed the cardinal sin of changing a dimension by hand, instead of redrawing the component so that AutoCad would dimension it.

Sorry. As a new member I hadn't realised precision and accuracy was/is such a hot potato (or a can of worms, if you prefer meat to veg!).

I simply judge it the same as a target shooter would. Precision is measured by the size of the

grouping and accuracy refers to how close that group is to the bullseye. Nothing mystical or contentious enough to get hot under the collar about!

I hope you saw my smiley... look up past discussions on the subject and judge for yourself

Neil

Oh it is, i was there and it got pretty ugly at one point. I wont delve into it in case i act as the tinder so to speak.

Michael W

When making things accurate, several factors come to mind. Does it actually need it ?. Is the material actually stable enough to achieve those sizes? Can you measure it ? Did you compensate for the temperature of the part at the time it was being made. This really applies to Ali parts that are bigger than 100 mm diameter. I can remember ceramic gauge blocks coming onto the market. The down side was the expansion difference at different temperatures. So if the tools were not 20 C, it would show an incorrect reading, especially when checking the larger mics, or when someone was checking a milled pocket on a plate in the summer time. So the steel gauge blocks are still used and they sold the ceramic ones. I see that Mitutoyo now sell a sub micron micrometer these days for the very series people. But they will be superceded with laser based measuring equipment before too long or non contact measuring. Simple shapes are a lot easier to make to 0.01mm than a complex shape. If the shape is complex, a lot of cnc lathes are not capable of making a part to 0.01mm accuracy. Usually as they do not have the measuring gear to measure the part, or they do not have the measuring equipment to measure accurately the radius of the cutting tools to properly set the geometry offsets for the programming.

Neil

The purpose of precision and associated tolerances is to make spare parts interchangeable and to ensure the tightness of fit is suitable for the assembly.

Colin

I would have to dispute that about contact measuring, i worked for a precision manufacturer and we found that in practice, the contact measuring machines were the final word on everything, because non contact will only measure what the image projects and any burrs or particles in the way of the image can alter the reading quite drastically. What they say can be accurate to on paper is not what you necessarily get in reality. You're never going to be able to beat the fact that a probe that is physically present on a given surface will always be more reliable than a probe which is not physically present on the surface and is reading an image of the surface. 

Having said that, if you're working on things which arent too fussy you can save yourself alot of time by non contact measuring. Thats the alure of it anyway.

As an inspector i was told to take the shadow graph readings with a pinch of salt because it's a shadow that you're measuring.  

Michael W

Edited By Michael Walters on 07/06/2016 14:08:40

In our world repeatability and reproducibility are much more important than accuracy per se. Even if your micrometer is 10 thou out you'll have no problem in machining a nominal 1 1/8" bore and 1 1/8" piston to match so long as you only use that micrometer. So telescopic gauges for the bore then! If your outside micrometer is + 10 thou and your inside one - 10 thou the potential problems are obvious! For us accuracy of measurement is primarily to verify repeatability rather than producing things to tenths tolerances. Even a low end micrometer is going to be pretty good at showing up the sort of small variations we are likely to encounter. Big variations mean you have a whole n'other set of problems to deal with!

Classic example of small variation accuracy is the tenths thou vernier on a decent micrometer. The overall constructional quality essential to meeting the measurement specification over the full range of the instrument and between any pair of random sizes within that range is such that for small ranges, maybe a couple or three turns of the thimble, tenths thous can be meaningfully measured. Useful for assessing small variations between parts but no good for proper measurements because there is no way of knowing what the error on the nominal dimension is. Obviously it should be less than a thou but that still leaves several tenths to play with.

In practice one you get below ± 1 thou you have to start thinking about temperature and a myriad of other conditions if you need accuracy relative to standards. If its just your own personal shop then keeping conditions and the way you work pretty much the same part to part variations will be small. If you do need real accuracy direct comparison off a set of gauge blocks will kill most of the potential errors. If your micrometer measures a 0.750" block at 0.761" and you make the part to 0.761" on the micrometer it will be pretty darn close to 0.750". Who cares why the micrometer reads different.

A great advantage of analogue instruments is that you can see small shifts directly on the scale. Reading a micrometer to say 1.125 plus two and a half line thicknesses is perfectly valid. Your judgement as to what one and a half line thicknesses is will probably be different to anyone elses but you know the difference between two and a half and a half. You also know what the difference means in terms of sliding fit. Probably have a good idea as to how much extra you need to tweak the final cut on the lathe to get that change too. Digital won't do that. In my view digital is only reliable with a sprung probe for constant force on whatever you are measuring. Instrument bodies are surprisingly flexi at the thou' level, especially cheap calipers.

Your normal working accuracy depends on the machine you are using. With care and thought you can generally coerce better than normal performance out of even a fairly ropey or well worn machine. But thats for special occasions only. What it does when you just walk up and use it sets the norm. The first machine I had with any pretensions to reasoanble performance was a Pools Special, generally had to start thinking to get much below ± 10 thou. The two SouthBend 9" that followed were probably ± 5 thou walk up, ± 1 thou think about it and possibly better if you tried real hard. The SouthBend Heavy 10 toolroom that followed was in the ± 2 thou or better class for walk up work but the lightweight slides meant that the lathe didn't really match the capabilites of the superb plain bearing headstock. Rather light cuts being the order of the day for top performance but working to a few tenths thou wasn't unreasonably difficult. These days I've moved out of the normal Home Shop Guy equipment level with a Smart & Brown 1024 VSL as my main machine. Tenths thous are within reach without undue effort and if a walk up job goes outside ±1 thou I've either been careless or too darn lazy to put a sharp tool up. All for about half as much again as a good Myford!

Clive.

Edited By Clive Foster on 07/06/2016 14:24:02

Hi Clive,

In principle I agree with you. However (there's always a 'however' ), when you say

"If your micrometer measures a 0.750" block at 0.761" and you make the part to 0.761" on the micrometer it will be pretty darn close to 0.750". Who cares why the micrometer reads different."

the problem comes down to the drawings, which naturally have everything dimensioned with the correct value. Unless you are going to mark-up your own set of drawings with dimensions (values) that would be indicated by your micrometer, and verify (preferably before starting the project) that the error is constant over the range of values that you will need to measure, then life is going to be very tedious setting up blocks all the time to determine what measured values you actually need to work to.

It gets worse if the build depends on a stack-up of components. And even worse again if you happen to have another (dodgy, or accurate) micrometer that you may happen to use alternately during the day. And even worse again if you need to use two different types of dodgy instruments (your bore and cylinder example).

All of which is just the tip of the iceberg. But I'm sure you are perfectly aware of that.

A good read is the NPL 'Good Practice Guide No 80 (Fundamental Good Practice in Dimensional Metrology). Good Practice Guide No. 80
Roger

Edited By Roger Head on 08/06/2016 02:40:01

Edited By Roger Head on 08/06/2016 02:47:24

"Some years ago we had a big project (a very special wheelchair) that required several people for it's implementation. One of our members, a professed AutoCad expert, decided to do the drawings for us. He produced dimensions accurate to 0.0001" (there was no way we could work to that level)"

Another mis-understood word, along the lines of precision and accurate is "professional".

A professional is a member of a profession or any person who earns their living from a specified professional activity but this does not always mean that they are an expert in what they are doing. The professed AutoCad expert above may know how to create drawings in Autocad but could have no practical experience or qualifications in engineering which appears to be the case with his dimensions. CAD is only a tool for making drawings.

The designer/draftsman (sorry draftperson to be non gender specific) should dimension a component to achieve an acceptable fit or finish for a particular part. The machinist is expected to achieve that which is shown on the drawing and specifying a tighter fit or surface finish than that which is required could be costly.

Paul

Big difference between a professional and one who professes, but.

Many years ago, an old engineer told me "If you can't make it accurate, make it adjustable."

Edited By Dinosaur Engineer on 08/06/2016 10:55:01

Edited By Dinosaur Engineer on 08/06/2016 10:55:44

Calibration of high accuracy gauge blocks is done using laser interferometry with uncertainties in the tens of nanometres. Rather odd given that it's a non-contact measurement.

Andrew

This is probably the worlds most accurate hand held micrometer,

http://ecatalog.mitutoyo.com/MDH-Micrometer-High-Accuracy-Sub-Micron-Digimatic-Micrometer-C1816.aspx

This is the new series of measuring technologies that will make the mechanical stuff obsolete like digital calipers have taken over from vernier calipers, but maybe more so. Interferometry I think is the next level of accurate measuring. Not priced for the hobby market or the small engineering companies yet, but prices will come down I am sure.

http://www.keyence.com/products/measure/micrometer/

Neil

I suspect we are talking about the difference between using them in a lab and using them in a workshop environment.

Neil

What happens if your probe happens to also make contact with one of these burrs or particles rather than a clean surface?

Tim asked:

So, it would help if you could let me know how expert you think you are, and how accurate would your output be, remembering that you would be working with a home or hobby machine, and not with stuff worth more than a three-bed semi in Alderley Edge.

• I'm a member of the amateur bodge team who is trying to improve
• I usually go for a good fit rather than absolute accuracy or precision.
• Only the most expensive of my digital calipers came with a documented claim, it's +/- 0.02mm
• The DRO on my mill sometimes jumps 0.02 or even 0.03mm
• My 4 digital calipers all produce similar results up to 30mm. Over longer lengths they start to disagree, worst case about 0.06mm
• None of my measuring equipment has ever been properly calibrated
• I'm not as careful about removing dirt as I could be
• I have precision parallels but no gauge blocks
• Depending on time of year, the temperature in my workshop varies between 4 and 28C. I never compensate for that.
• The unskilled readings I take vary noticeably with hand pressure and angle. I get better repeatability from my micrometer but the results still wobble about a bit.
• When I make to a specific size against some sort of fixed gauge, there will be cursing and quite a few rejects.
• I don't do much with angles apart from using an inexpensive rotary table with a vernier dial. I successfully made a short taper for a drill chuck once, I'm sure more by luck than judgement.

It would be interesting to test just how effective people really are at taking measurements at one of the exhibitions. People could show up with their favourite measuring stick, be ushered alone into a tent containing various objects, and asked to submit their results under a psuedonym. The winner would be the guy with the lowest error rate. All we need is the NPL to provide some test objects and MEW to provide a gigantic cash prize...

Cheers,

Dave

An interesting debate, everyone - and it stayed mainly on topic. One or two answers:

The reason for the question was this - I suggested the use of an old roller-bearing race as a spacer in a machining job, and I was told that surely it was imperative to check the bearing carefully as it might not be parallel (or of a given dimension). My answer was that I thought all rolling bearings are made to high dimensional standards - higher than you can get with a home lathe or mill, in any event, and so you could have confidence that no problems would arise on that score. I said that the ordinary (non-time-served) operator could normally work to a thou as a routine minimum, but anything better than half that took time and special treatment. Those of you who answered the question with rigour confirmed this.

It might help to know that the non-contact methods include the use of interference patterns (Newton's Rings is a classic example) and what you are counting is wavelengths of light (or UV, etc). This requires a decently flat surface, of course, and you could reasonably expect any 'burrs or particles' to be sorted before entering the lab.

Thanks again

Regards, Tim

Edited By Tim Stevens on 08/06/2016 16:41:03

Since I made a 0.5mm pitch cross slide screw and nut with the 0 to 100 dial from the 100 thou pitched screw, I can easily turn diameters to 0.005mm in diameter for small parts under 30mm diameter. The advantage is that the amount on the mic is the same as the cross slide dial. On lengths I can easily get to less than 0.01mm for lengths. To get lengths less than 0.005mm does take a lot more effort that is for sure and then knowing the work piece temperature can be important. For the head inserts for engines making the lengths is easy to 0.01mm and the diameters to within 0.005 is easy for me. Pistons are the hardest part that require making something to Ø 0.001mm is very difficult but can be done. But the pistons are made to fit a specific liner and not just made to a specific size to a tolerance of 0.001mm which is a very different thing. Precision milling to better than 0.01mm does take quite a concerted effort that is for sure and most lower end mills I have seen for the home workshop are not capable of better than 0.02mm precision work.

Neil