CNC Wood Router how Accurate?

[Bill George]

l realize there may be two or more correct answers for this. One for a commercial shop making furniture or kitchen caninets the other for hobbyists, and perfectionist.
The Forum I am on from time to time has the last two. One insists that if if a 48 inch machine can not hold .001 over 48 inches, its junk. Others think 1/16 inch over 48 is a worthly goal. Frankly I just want the job done and .125 is fine.

I wonder what you guys think? Reminder this is wood.

[Mark Bolton]

There was a thread here a while back regarding machine accuracy over distance as well as machining and measuring some test shapes.

A lot depends on the machine and hardware specs but I'm not sure .001 in 48" is realistic but I would not be able to work with a 1/16th and never an eight anywhere on the machine. Tollerance close to and in the single digits (below .010)

[Gerry Grzadzinski]

The machine will only be as accurate as the leadscrews or gear racks, but even average quality ones should get you within .01" over 4 feet.

It's hard to build a machine less accurate than that.

[Mick Simon]

I think we're considering two different things here - accuracy claimed by the manufacturer and accuracy that's "good enough" for hobbyist work.

Two very different things. If a manufacturer claims to be accurate within .005" over the bed of his machine and it's out .125" (25X the error) there's a problem.

[Bruce Page]

The machine will only be as accurate as the leadscrews or gear racks, but even average quality ones should get you within .01" over 4 feet.

It's hard to build a machine less accurate than that.

There's a big difference between .001 & .010
.001 is as someone said, unrealistic in a wood router. I have worked with machine tools in ISO9000 shops that couldn't hold .001 over 48". You would need a temperature controlled inspection lab to even check it.

[David Falkner]

It depends on what I'm cutting, really. If I am cutting a sign or plaque that's less than two square feet then what I need is repeatability as good as I can get, preferably less than 0.005". That's probably good enough for going back over letters and simple engraving. Our CNC router is 2x4 and the largest piece I've cut is 48" but it was the word 'Grace' in sort of block letters. Ten or fifteen thousandths is close enough for that.

But if I am cutting a bridge for an acoustic guitar and want to mill the slot to 0.126" for a good tight fit of the 0.125" bridge saddle then I need the best accuracy I can get in those few inches. And if I am cutting 0.023" fret slots on a fingerboard that is 18" long (they vary) then I need the best calibrated and most accurate I can get our machine.

And I cut a lot of engineered pieces and parts; pieces that have to fit inside another and slide, parts that lock together, etc., so my requirements are a bit more stringent that most, I would think.

So if you're cutting signs, plaques, profiles that can vary by 1/8" and not matter, etc., then most machines will suffice. But I need accuracy and repeatability often. I am willing to invest the time to get our CNC as accurate as is possible for that 25% of the time when I need the utmost.

David

[Gary Campbell]

Bill...
Lost of info here, most of it good. And all in the same direction, let me add my take to a few of the comments.

Bill: "Reminder this is wood." Right on! Plus or minus.005 if you are lucky and measure within a few seconds of being cut. Double or triple that a few hours later is wood has not reached equilibrium with the shops environment.

Gerry: "It's hard to build a machine less accurate than that." Agree, but for the most part not easy to maintain for the majority of non fulltime operators.

Mick: "accuracy claimed by the manufacturer and accuracy that's "good enough" for hobbyist work." Two completely different items.

CNC mfgr's, and I am speaking of the sub $25K arena here, are in a race to have the best, the highest, the lowest or the fastest specs. Customers buy on specs, they cant help but do that. Unfortunately they usually know as much about them as a 16 yr old buying their first car with a list from Dad for the salesman. That said, most mfgr's post reasonably accurate specs, most customers think they mean something else.

Accuracy, or the ability to hold tolerance under load? Positional accuracy by some mfgrs is simply the step resolution under no load. No account for any backlash in the system. One company offers .002" positional accuracy, but their motor supplier says .003-.006 backlash in their gearmotors. In their (mfgr's) defense, it would be as hard to predict a users accuracy as it would be to predict gas mileage without knowing driving habits or terrain. Yet the number is on every window sticker. TAKE THEM WITH THE SAME GRAIN OF SALT!!

Machine deflection is the key to accuracy. Large parts, small parts, whatever sized parts. Most sub $20K machines will deflect under load far more than one could ever find error in Chinese ballscrews or most rack and pinion setups. These are not milling machines with high tolerance ground, qualified ballscrews. The less the machine costs, the more it deflects. The cost of the bearings , power transmission components and rigidity of the major parts is what determines the price class of these machines.

If one were to take a hardwood plywood, or medium hardwood stock and use a 1/4" bit, 1/4" depth of cut and push it 250ipm @ 15Krpm (.008" chipload) as a constant. (I have) Lets call it "one unit of push". No where near the chipload required on commercial machines, but much higher than many smaller machines can cut anywhere near accurately with.

Accuracy? I say tolerance, or accuracy, under load. Want to know an accuracy number? For steel framed linear rail machines, take the price, divide into 10,000 and multiply that number times .008 (that chipload from above) and you will be very close to the actual deflection for that class of machine under the load I define. If you have steel Vrollers, 1.5 times the deflection, plastic: double it. Supported round rails: add a few thousandths, unsupported: back to double. Add more for extrusion, vs steel, add more for plastic parts, the list goes on and on.

That's why the operator is important. If the chipload is low enough and the machine has been calibrated properly, a good operator can get within a few thousandths of the vector size with a bit of practice. He just cant do it as fast with a $2500 machine as he could with a $25,000 one. Unfortunately most operators try to get their chipload numbers up to match bit mfgr's charts instead of lowering them to match their machines characteristics.

This is the reason shallower depths of cut, higher rpm and slower speeds (much lower chiploads) are recommended on sub $10k machines. But when run at chiploads that are below what the machine frame can resist most will be just as accurate. Think of these machines like you would a musical instrument. It takes practice to be good, and just like any instrument, no talent is purchased with the machine. Like music, only the best operators will get the best results.

I would rather listen to Keith Urban on a $200 guitar than a novice on a $20,000 one.

[Bill George]

I guess I was referring to how accurate over a distance, 4 or 8 feet. What David is doing in a smaller area and I have done about the same with my machine is to be expected. But with a ball screw machine it would be much easier I would think. Mine is R&P and think Davids is also. But you can't get the speed out of ball screws.

Thank you for the replies, I just needed to hear something from professionals.

It was also posted on the "other" forum that Camaster is a 2nd tier machine.

[Mick Simon]

Gary - I love your formula(s) for determining accuracy! Probably the most true-to-life I've seen anywhere. Also agree with your assessment of how these machines get used. One of my students last year bought a Shapeoko XXL for making $$$ guitars. At the time I thought, "Good luck with that." He's a world-class guitar maker. He uses the machine as it's intended to be used and gets superb results.

Bill - The poster in the other forum was not disparaging Camaster in any way. He was making the division between heavy industrial machines - tier 1 - and everything else - tier 2 (BTW, his definition, not mine). Toyota makes a great car, but you don't see Corollas running with the big boys at Indy. At the time, the discussion was about machines that were well under $20K. Camaster makes several models. If you call them wanting to do "serious production" I doubt they'd steer you toward a Stinger I. Cobra Elite - well, there you're starting to blur the line, depending on your idea of serious production.
You can certainly get speeds out of ball screws that well exceed anything that FLA claims for the Saturn 2, especially in the lengths we're talking about. Many tier 1 routers do it day in and day out.

Since I was the OP of the thread you're referring to in the other forum I'll point out that the accuracy of the machine was not my reason for posting. My machine was binding up to the point that the R&P drives were ratcheting because of (IMO) poor build practices used to produce the machine. That, in turn, devolved into a discussion of stated accuracies.

[Jim Becker]

I like that Gary brought up the things he did as it reinforces what is reality: there are many variables and our job as a CNC operator is to do our best to leverage strengths and mitigate weaknesses, all at the same time. It's a dance. I have a decent machine and I know it can cut pretty accurately if I do my job well, including being within a reasonably good tolerance over a ~48" distance. There is going to be a trade-off on speed vs accuracy for sure, but that shouldn't be surprising. I'm still learning this dance each and every job I run. And yea...wood. Wood does what wood does. Sometimes quickly. Sometimes over a long period of time. When you measure matters.

[Gerry Grzadzinski]

But you can't get the speed out of ball screws.

Sure you can. I'm cutting at 1200ipm on a ballscrew machine. And it has a 14' long ballscrew.

It was also posted on the "other" forum that Camaster is a 2nd tier machine.

As Mick pointed out below, I was comparing it to machines costing between $100K and $250K.

Thank you for the replies, I just needed to hear something from professionals.

How do you define "professional"?

[Bill George]

Sure you can. I'm cutting at 1200ipm on a ballscrew machine. And it has a 14' long ballscrew.



As Mick pointed out below, I was comparing it to machines costing between $100K and $250K.




How do you define "professional"?

You must admit most of the posters over "there" are first time users, not like you. Ball Screws vs Rack I guess Gerry I was comparing it to the "other" second tier machines on the market, none seem to have ball screws that are running as fast as the R&P.

[Jim Becker]

My machine has both R&P (X-Y) and Ball Screw (Z) and it can move pretty darn quick...not 1200 ipm, but this specific machine isn't designed for that. 500 ipm cutting with 800 ipm rapids. It's not material on small stuff, however...because you can't accelerate and decelerate in short runs.

[Brad Shipton]

I got a quote from CR Onsrud for a Panelmate machine and during that discussion I was quoted .003" for accuracy. The tech was quick to point out that below that I would need to provide a lot of other information other than "I want to cut some wood parts." They routinely send machines to work in the aerospace industry, so I am quite confident in their specs.

One of the thing to consider is how the entry level companies measure their machine setups for accuracy. Think how you would measure the distance from your left rail to the right rail with an accuracy of .005" or less? Not a simple task. The big companies like Gerry mentions have CMM testing equipment that cost 100's of thousands or more, and the groups most of us deal with do not. You can read about the 3D testing machine that Thermwood uses on their website. Here is a low tech solution(https://www.youtube.com/watch?v=k5kdWaTM_14) (yes, Chinese, but they seem open to post) that shows the setup for the Y Axis . That is a time consuming job for a person that has patience. Done poorly, the rails are not parallel. Keep in mind the surface that rail is mounted to is milled. Many of our entry level machines have rails that are simply screwed to raw square tubing or plate. From the mill, plate and HSS tubing is not perfect (they have tolerance specs). I know both my Y axis rails are out by .015"+- in both planes since I bought a straight edge so I could measure with feeler gauges. While this is small, each little error adds up. I know some will say, just cut some test parts. That works too, but do they cut at the center of the table, the end, how large...? If after I test it some parts are out of spec, how do I fix it? Measuring gives you that answer.

As for accuracy on long parts my rule is I should not be able to detect an error with a typical tape measure. I want to be able to give a part to a customer that he or she can measure in front of me and accept it. For me, that is around 1/32". If they pull out a 5'-0" long set of calipers, I am not their guy.

The biggest difficultly I find with machine inaccuracy is predictability. I would be fine with 1/16" inaccuracy if it was a known constant. The reality is there are areas of your table that produce marginally different results. This first popped its head up for me years back when I was making a stand that needed to fit 1/4" dowels. I ran a routine on one side of the table one day, and for some reason I ran it later on the other side. The fit was not quite the same and I had to edit settings for the tool path. I eventually found the problem, and corrected it as best I could.

At the end of the day, you get what you pay for. If you need <.001"/48" you better be prepared a machine magnitudes heavier than the typical 2,500lb entry level 4'x8' machine. Look at some of the HAAS machines to give you an idea what happens if you need to cut parts for a race car. Yes, that is steel, but the principle is the same.

[Mark Bolton]

While you may not have a set of calipers the length of the longest axis on your machine, a way that we frequently see how accurately our machine is cutting is when cutting large-ish jobs of nested parts. You may have part A cut on the left side of the machine on sheet 3 and part B cut on the right side of the machine out of sheet 9 and so on. When they are all coming off consistently enough that on close inspection during assembly there is zero issues over an 8' part you have a pretty good indicator. We run very tight tolerances on our dado's for instance. The other shops we cut for continually ask to open up these tolerances because they are accustomed to dados cut on the saw that are cut drastically oversize because shimming the dado stack is too time consuming. Why I have no idea, but my guess is out of shear habit and routine they want their dado's sloppy and they push the parts to the interior face of the dado and nail/staple. We on the other hand run our dado's with about a .004 tolerance so on assembly the boxes basically push together with a hand friction fit and stay there while you grab your screw gun or nailer.

Point being, when you have a large pantry side that is 8' long with a .004 tolerance dado at the top, bottom, and a few intermediates, and a mating back, adjacent side, and so on, all coming out of different sheets in the nest, and all from random sides, center, whever on the table, and your lids, decks, and fixed shelves slide in by hand, dead tight without gaps, mallets, or clamps,... I think you can call it good.