Accuracy

[mikejohn]

Today I reset my z axis with a lot of trammeling, and it was a long way out.
In the end I got it pretty accurate, about .5mm over 120 cm diameter.
Then I did a number of tests including cutting two grooves 6mm deep into MDF with a brand spanking new 5mm bit.
The first groove I cut in 2 passes each 3mm deep, the second in a single 6mm pass.
I checked the groove width with the tip of my calipers,which has a precise 5mm width.
The single pass fitted very very snug, perfect even.
The one with two passes had a little play, not enough to insert a 0.1mm feeler guage in, but almost enough.
Now to my question.
In the ShopBot specs for my machine, the latest PRT96, they quote a positioning accuracy of +/- 0.127mm (my conversion from their imperial), and a cutting accuracy of +/- 0.381mm.
My 'error' in the two grooves is well inside even the first of these.
Finally my question.
Does this mean that if I continually cut the same groove, many times, it could end up 0.762mm wider than planned?
And do I need also to add the positioning error to this?
In fact, is it theoretically possible to cut just twice, and end up with +.127 and +.381 (.508mm or 0.02") error?
In all this I am assuming that there is no physical error, that in fact the accuracy of the way the Shopbot is controlled might lead to these type of errors?
Although this question is a theoretical one, and not something I have noticed whilst operating the shopbot, nevertheless I am interested in what all these specification figures mean.
My tramelling today gave me an error, at the bit, well inside the ShopBot specs.
I recognise that the set up of the ShopBot, and any bit run out can be even greater than this.
However, I am interested in the true theoretical repeatable accuracy of a perfectly set up ShopBot.

.........Mike

[gerald_d]

Are you talking of the same "errors" as in the current dovetail thread (http://www.talkshopbot.com/forum/messages/2/10866.html)?

[mikejohn]

Gerald
No, but it does have a bearing on that thread.
Morris Dovey-tail says And no one is likely to specify a gap as large as 0.01" unless they plan to grout/spackle the joint.
My question above tries to ask what the specification accuracy of the ShopBot is.
I presume that if the file asks the shopbot to go to 100,50 it sends this information to the shopbot, but ends up at 99.83,50.18 because of the way the drivers or steppers work.

And I'm assuming there is some mechanical innacuracy with the pinions etc even in a theoretically perfectly set up 'Bot.
Practically the question is, if I did have a .01" glue gap. might this not simply dissapear within the normal shopbot working tollerances?

..........Mike

[richards]

Mike,

I think you only have a few 'unadjustable' errors that you have to worry about.

1. 1/2-step width. The stepper motors make finite moves, which are non-divisible; therefore, if your axis's destination ends up between steps, you will be, at most, 1/2-step width away from your desired position.

2. Mechanical looseness or slop. This includes such things as: V-rails that are not perfectly straight or that have tooling/cutting marks. V-rollers that are not perfectly round or have loose/worn bearings. Worn router or spindle bearings. Worn or out-of-tolerance router/spindle collets. Worn or out-of-round tooling, etc.

3. Temperature. The temperature in my shop has varied from 105-degrees F. in July to 4-degrees F. in mid-December. The metal table has to expand/contract based on the temperature. Even bringing the shop to more livable temperatures before cutting does not guarantee that the steel table has reached room temperature.

4. Material. I've had the cutter wander when cutting hardwood, especially when the feed speed is at its maximum and the tool path cuts close to a knot.

[stevem]

5. The pinion to rack ratio does not divide evenly into increments of 1”, thus causing some rounding off.

[mikejohn]

Mike
I recognise your points 2,3 and 4.
It's your 1 and Steves 5 that I am wondering about
If I am having accuracy problems, how much is outside of my control?

..............Mike

[mikejohn]

edited double post

[gerald_d]

Theoretically, a rack & pinion system has no "linearity" errors. However, manufacturing tolerances and subsequent wear & tear will make the rack & pinion less than accurate, but it will not introduce a coarse accumulative error.

[richards]

Mike,

Stepper resolution is a factor, but I believe a minor factor with absolute accuracy on a CNC router used to cut wood. The main reason that I'm saying that, is because, for particle board, I use a chip load of 0.025-inch. Even if the machine had absolute accuracy of 0.001-inch, particle board 'chunks' as it's cut so my absolute accuracy probably has a variance of +/- 0.025. (Put a straight edge across a freshly cut piece of melamine coated particle board, and you'll see what I mean. There are bumps and dips in the particle board, but the melamine is almost always perfectly smooth.)

Now, back to stepper resolution. If we assume that a basic stepper motor has full steps of 1.8-degree, or 200 steps per revolution, and that the Shopbot has a pinion gear with a pitch diameter of 1-inch, then for every step the stepper motor makes, the axis will move 0.0157 inches (3.14 inches / 200). If the stepper motor half-steps, then each step will move the axis 0.00785 inches (3.14 inches / 400). If the stepper has a 3:1 gearbox, and uses full-steps, then the movement per step will be 0.00523 (3.14 inches / 600). Other pitch diameters, gear ratios, steps-per-revolution, etc. will determine the minimum distance an axis will move per step. To find out how your machine is configured, type the UV command or look at the VU fill-in sheet. You'll be shown the number of steps per inch (steps per millimeter?) that your machine moves.

[dirk]

While trying to determine a drive for my new machine, I wrote a quick little Excel spreadsheet. It will show the effects of different size motors,gearing, cutting force and step resolution, etc.
I'll try to attach a copy.


Stepper Calculator

3242 (20.0 k)

Dirk

[richards]

Dirk,

Great spreadsheet!

[Brady Watson]

PRTs are 1/4 stepping. Alphas are micro-stepping (1/10). A PRT will have approximately 800 steps per rev (4 X 200 steps) and the Alphas have 2000 steps per rev.

Micro-stepping is a lot smoother when compared to it's 1/4 stepping variant...BUT real positional accuracy is the same. Any stepper is physically limited to 1/4 step accuracy due to the way a stepper's rotor is designed. Areas in between a 1/4 step position don't have a good magnetic lock, and as a result can drift anywhere in that (let's say) 12 to 3 o'clock position. Even with an encoder on the stepper, there is no gaurantee that you will land on a spot in between 1/4 steps.

As Mike R. points out, by increasing the gear reduction, you can minimize the effect that this would have on accuracy...at the expense of speed. If you had 3.6 or 7.2:1 gearboxes on an alpha...that would add some serious accuracy to the setup, ruling out other mechanical gremlins.

-Brady

[dirk]

Brady
According to your expanation above it sounds like your saying the owners of PRTs with gearboxes have better "real accuracy" then the Alpha because the Alpha's not geared. Is this what you mean?
Dirk

[richards]

Brady,

Gear reduction gives greater accuracy ONLY if there is no backlash. In my former life as a process control computer designer/programer, I ALWAYS used belt drives instead of gear boxes.

Gearing a motor increases torque and decreases the length of each step - depending on the gear reduction. Both factors are big pluses IF there is sufficient speed to run each axis as desired and IF there is no backlash.

I've been playing with 1/10 step drivers and motors for several months and still have not reached any final conclusions. Every test that I've run agrees with the White Paper released by Maris at Gecko, but in real life, I'm still playing with the toys.

As far as a general purpose CNC router goes, I haven't found anything that beats the Shopbot Alpha for speed and for precision - at an affordable price. When I add up just the hardware costs to build my own machine with ball screws, tight tolerance rails, servo or steppers with feed back, the price is several times the cost of an Alpha direct from the factory. Which means to me that sometimes it's better to tweak the Alpha as tight as possible and then get out the sandpaper to smooth the cuts when necessary.

[Brady Watson]

Dirk,
To be honest, I really don't know. The Alpha motors seem to be very unique in the sense that they have a VERY wide power band for a stepper. This could be partly due to the fact that they run @ 170v. I believe that the alphas also run different sized pinions than a PRT (smaller).

Mike,
That's why I said, "...that would add some serious accuracy to the setup, ruling out other mechanical gremlins" I should have clarified the 'gremlin' part!

Contrary to what belt manufacturers will have you believe about their 'zero-backlash' belts...they DO stretch and DO exhibit backlash in the real world. The ShopBot gearboxes on the PRT are the more expensive variety of gearboxes from Oriental with taper-hobbed gears and very low backlash. I am still impressed by my PRT motors & am sure glad that ShopBot didn't cheese out on the motors. I am not saying that one configuration is better than another. If belts were truly zero backlash, I would put money on ShopBot using them on their tools, over the expensive gearbox motors. I do agree that your machine is only as tight/accurate as your greatest degree of slop/backlash.

I suppose the big question all of us are asking ourselves is, just how much resolution (accuracy) do we really need on a CNC router? I could see us balking at .005" or even .0015" on a metal mill...but what are people cutting that need that degree of resolution? I would venture to say that my PRT's cutting accuracy (when run at the 'right' speed) is WAY better than the advertised .015" number. Repeatability of my PRT is excellent...granted I have make a few refinements to tighten up the machine, but all of the running gear is untouched, except for new pinions on all 3 axes replaced this year.

-Brady

[mikejohn]

Back to the theory.
If the published tolerances are plus or minus .015" then an "error" of .03 is possible, and still be within specs.
I still don't know if you also add the positional error, .02" (twice times published .01") to the .03 to get .05".
Take Mr. Doveys dovetail joint glue gap, where he feels .01" excessive. Is it not possible that you can cut tails and pins, and they will not even fit but still be within ShopBot tolerances?
This thread was started to investigate the theoretical accuracy, following ShopBots published data. I recognise its relevance (or lack of it) in day to day cutting.
I just want to know how they arrive at these tolerances, and what practical effect they have.


..........Mike

[mrdovey]

Brady...

How much accuracy do we really need? Good question - and the answer (IMO) depends on the work we're trying to do.

For American kitchen base cabinet boxes we don't need much. For euro-cabinets a bit more might be needed. For most signs we don't need much. For furniture-grade joinery like dovetails the requirement goes up several notches, because cut edges need to mate cleanly and not show glue lines.

I've found cutting situations where conventional/climb cutting choices at different parts of a joint seemed to produce better or worse fits - and I've had situations where, after trying to rout something on the 'Bot, I switched to a hand-held router and tinkered to get a (measured) depth of cut somewhere between 0.2500" and 0.2505" - because that was what was needed to make a joint meet requirements.

Like you, I'm seeing accuracy on my PRT that's much better than +/-0.015" - some of the stuff I've cut has measured closer to +/- 0.0015", but I only measure when I feel tolerances in the work are critical and I'd be hesitant to advertize that I could routinely produce work that met that tolerance.

If I had a factory-built CNC milling machine that couldn't do better than +/-0.0005" I think I'd be pretty unhappy with my purchase decision.

Mike...

On the theoretical side I'd guess that you're absolutely correct; but keep in mind that:

"In theory, there is no difference between 'theoretical' and 'real world' - but in the real world there often is".

There's a bit of wry humor that describes the process of "calculating to seventeen significant figures, measuring with a yard [meter] stick, cutting with an axe, and adjusting (as necessary) to fit."

In the dovetail example, the practical solution to adjusting the fit appears to be making a choice between: (a) climb-cutting both tails and sockets for tightest fit, (b) climb-cutting one side of the joint and conventional-cutting the other side for medium fit, or (c) conventional cutting both sides for loosest fit.

Sometimes, it's just a matter of figuring out how to do the "adjust to fit" part.


...Morris

[dvanr]

Brady,

Your comment,

"Micro-stepping is a lot smoother when compared to it's 1/4 stepping variant...BUT real positional accuracy is the same. Any stepper is physically limited to 1/4 step accuracy due to the way a stepper's rotor is designed. Areas in between a 1/4 step position don't have a good magnetic lock, and as a result can drift anywhere in that (let's say) 12 to 3 o'clock position. Even with an encoder on the stepper, there is no gaurantee that you will land on a spot in between 1/4 steps."

is interesting. Would this be a possible cause of rough cuts on curves or circles when the Bot microsteps around a curve or circle and is unable to hold (or drifts) at 1/10th microsteps?

( Sorry Mike for the thread drift )

[mikejohn]

Dick
No apology needed ( I am less ambivalent of Ozzies after last years success
)
The thread drift often gives better information than the questioner intended.
Maybe ShopBot will pop up with a definitive answer for me.
............Mike

[Brady Watson]

Dick,
Actually, I should not have used the analogy of '12 to 3' because it paints the picture that a 1/4 step is a 90° rotation of the shaft...not true. On a 200 step per rev stepper, there are 200 'grooves' around the rotor and magnet-backed grooves in the case of the motor. Depending on the position of the rotor, this determines if there is full torque (each 1/4 step) or a position of less torque due to the magnetic/electrical flux holding the motor's position. Think of it as a sort of 'electrical backlash'.

As far as troubleshooting your rough cuts on curves, there are a few things that I have encountered that contribute to this...although I do believe that (on an alpha) there would be a benefit to gear reduction to decrease low-speed cogging on the 1:1 motors. (Not that I have witnessed low speed cogging...but other 1:1 steppers would be cogging pretty badly with this type of load)

The machine really needs a good going over to check that ALL of the v-roller bearings are seated properly in all parts of their travel. Wiggle every part of the gantry & Z assembly (with steppers on) and check that EVERYTHING is tight. Even a little looseness will transfer vibration to the cut. The new alpha/PRT gantry is leaps and bounds better than the older style one that came on my PRT. A tight gantry can be a double-edged sword as it brought a lot of things to light on my machine. The old gantry soaked up a LOT of vibration. After I welded it, it immediately started to see ridging on the edges of my parts, even on stright cuts. This turned out to be my pinions. It only cost me $60 to replace all 4 to fresh ones and the difference was remarkable. (This is after 3+ yrs on my PRT doing a LOT of 3D work...back and forth rastering. If you have an alpha and use the machine frequently, inspect your gears for signs of wear. The increase in speed, and in turn force on these gears is pretty substantial. Depending on what you are cutting and how well you greased the tracks, it might be a safe bet to say that you'll want to swap these out with new ones in the 1st year...or at least inspect them for wear. The other big thing that adds vibration to the cut is a mushy table. The alpha-style reinforcement plates are a MUST even on a PRT. Try air-cutting a 24" circle with your hand near the end of the c-channel, feeling for Y-direction movement. This is *very* minor on an alpha...and of gargantuan proportions on a PRT with no reinforcement plates. I will be adding these to my PRT soon! I have a 5X16' PRT...and it does shake...and those shakes DO transfer to your cut. So check to make sure (by air cutting or while working) that the table isn't moving etc. An older PRT would also benefit from a new set of XY motor mounting plates from ShopBot. They are MUCH stuffer than the single piece of angle holding them right now and will increase your pinion life. ALSO...you will get better circle cutting performance out of PW/Pro if you use the Arcs post processor and limit the tolerance to .001" or more.

-Brady

[jeff_guinn]

Dirk & Mike R,
Thanks for the spreadsheet. I entered 7.2 gear ratio,1.8 step angle,18 teeth pinions & the results are exactly twice the recommended unit values. Also, the recommended unit values for my ball screws are 2000. The screw has a .2 pitch & motor has no gearbox with a 1.8 step angle which I caculated at 1000 steps per inch. I'm totally lost & could use a good explanation.
Thanks,
Jeff

[gerald_d]

Jeff, is your system perhaps half-stepping? deleted

This (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=28&post=2721#POST2721) old post by Ted is essential reading. There have been numerous combinations and permutations, even within the same model number, or even for different axes on the same machine.

[evan]

Brady,
PW/Pro? Is that the same as Parts Wizard Ver.2?

Evan

[Brady Watson]

Jeff,
A '.2 pitch' ballscrew is essentially moves 0.2" per revolution of the motor (200 full steps) You can also think of it as a 5 turns per inch screw, which essentially functions as a 5:1 gear reduction.

So...If you are using a .2 pitch screw in conjunction with a 7.2:1 geared stepper, then you would simply multiply the ratios together to get your final drive ratio:

5 X 7.2 = 36:1

This means that the stepper motor itself must turn 36 times in order to turn the output shaft (the screw in this case) 1 full revolution. Since there is no free lunch per se, you will get gobs of torque at the expense of speed. For instance if you were using a 7.2:1 geared PRT motor with a .2 pitch screw, (using factory PRT voltage supply) you would have a max speed of about 1 inch per second...BUT you would have a theoretical max of about 3,000 Oz-in of torque. This would be a perfect setup for a high-resolution indexer that was swinging a heavy load, but not a good choice for a Z-axis as it just wouldn't be able to keep up with the X & Y movements.

I am confused...what is your application? Is this a PR or cable tool? Why the screw?

Your unit values are the number of steps (or patial steps) needed to move the tool one inch. If you are 1/2 stepping (as Gerald points out as a possibility) then your unit values would be somewhere around 400 (200 steps per rev * 2)....1/4 stepping machines would have a unit value twice that of a 1/2 stepping PRT up around 800 or so. The size of the pinion on each axis also plays a role with your unit values. It is typical to have 25 tooth pinions on the XY axes and a 20 tooth on the Z. Stock values for a PRT with 3.6:1 boxes on the XY is about 733 steps per inch and 916 steps per inch on the Z. A 7.2:1 Z axis would have exactly 2X the number of steps per inch (so more like 1832 or so on the Z).

Confused yet?


-Brady

[Brady Watson]

Evan,
Yes. You have the option to use the regular ShopBot Inch post or Arcs_Inch in both programs. The Arcs post does circular moves in arcs...the inch post does them in very small straight movements. You can choose either post in PW2, Insignia and ArtCAM Pro. I leave it on arcs_inch all the time since it doesn't use arcs for straight moves.

-Brady

[dirk]

Jeff
It sounds like your talking about a PR machine with ball screws on the Z axis. The PR uses 1/2 steps so that is why the unit value is 2000. 200 steps x 2(halfstepping) = 400 x 5(turns per inch)=2000. Half stepping would also make unit values for other Axis 2 x different. If thats not the case post or email your recomended unit values and let me check spreadsheet for error.
Dirk

[dirk]

Jeff
Another note is the 7.2 gear ratio. Are you sure it's not a 3.6, as far as I know the 7.2 is used only on the Z and 3.6 ratios are used on other axis.
Dirk

[dirk]

I posted the question as to position accuracy on Gecko's Forum. I got the following response from Marris of Gecko Drives:


I disagree. Static microstep accuracy depends on good motor linearity
and a drive that generates accurate sine and cosine phase currents.
Unloaded accuracy routinely measures as good as +/- 0.018 degrees over
the span of a single step and +/- 3% (0.054 degrees) of cyclic error
for a full revolution.

These kind of results can only be obtained from "square" step
motors, "round" step motors are far less linear.

Load adversely affects positioning accuracy. Motor torque is
proportional to the sine of the error angle (1 full step equals 90
degrees). An error angle of 1/10 step (9 degrees) results at load
torque equal to 15.6% of holding torque. That is the limit to which
1/10th step accuracy can be maintained; at a full holding torque load
the error is necessarily 1 full step.

Dynamically things get a little more dicey. All step motors develop and
unloaded lag of 1 full step when they are running on the inverse torque
region of their speed-torque curve. This is because inductive reactance
is limiting current and current lags voltage by 90 degrees (1 full
step) in this region. The motor will be located 1.8 degrees behind the
expected location while unloaded and will stall when it lags 3.6
degrees due to applied load.

Conclusion: A 1/10th step accuracy can be relied on if the motor load
is less than 15.6% of its holding torque. This is a usable value
because a properly designed open loop step motor application should be
biased at no more than 30 to 50% of the motor's holding torque. Put
differently, a 600 in-oz torque motor develops 94 in-oz at a 1/10th
step error. This is 184 lbs of thrust when applied to a 5 TPI leadscrew.

Mariss

[Brady Watson]

"A 1/10th step accuracy can be relied on if the motor load is less than 15.6% of its holding torque." put a bit more eloquently than me saying, "Areas in between a 1/4 step position don't have a good magnetic lock, and as a result can drift..."

Thanks for posting Mariss' response...I forgot about the '15% rule'.

-Brady

[jeff_guinn]

Dirk,
Thank you for clearly answering my question about my ball screw drive.I never realized that the PR used 1/2 stepping. Yes,it is a PR model with 2 Y & 2 X steppers which are in fact geared 7.2 with a recommended unit value of 254.6479 which was exactly twice the values in your spreadsheet.
Gerald-Thanks for the response-You shouldn't have deleted it!
Brady- Thanks. As I originally stated, the ball screw motor is direct drive-no gearbox. I stayed with the ball screw because of its' accuracy,durability & compactness. As a matter of fact,I just completed building & testing a new ball screw Z axis, for another application, using a larger stepper(PK268 series).
Thanks,
Jeff

[dirk]

Jeff
I checked the steps per inch calculation on the spreadsheet and I couldn’t find any errors. Maybe someone else could run a calculation to confirm the results. If I remember right The PR handled unit values for the Z axis differently than x and Y so that may be the difference.
Dirk

[mikejohn]

Now I am amongst the worst 'thread drifters' on the Forum, and as I said to Dick above, thread drift often gives a lot more information than the original question intended, which has certainly happened here.
Now all above is explaining how and why the tolerances evolve, but no one has yet answered my original question. If I accept ShopBots specifications "Does this mean that if I continually cut the same groove, many times, it could end up 0.762mm wider than planned?
And do I need also to add the positioning error to this?
In fact, is it theoretically possible to cut just twice, and end up with +.127 and +.381 (.508mm or 0.02") error?"

I will repeat that it is the theoretical possibility I am asking about.

(and if any one has "The Dummies Guide to All above" I would love a copy
)

.................Mike

[rhfurniture]

My guess is that if you issue the same instruction twice the result is electronically identical.
I havn't tested it though.

R.

[gerald_d]

Jeff, I deleted it because I realised that I didn't know what you meant by "recommended" or what model you had, as I hadn't really been following the thread. It was just luck that I said something that helped.

Why wasn't I following this thread.....because I have enough real problems without looking for hypothetical ones!

[mikejohn]

Gerald
Don't you see that , if you are chasing down an accuracy problem, unless you know what can happen in a 'perfect' set up as far as inaccuracy goes, you may in fact not be chasing a 1mm inaccuracy, but in fact a .1mm inaccuracy?
It all comes down to knowing when you should stop tweaking.
I do not understand why no one can answer my question?

rh
That is my point.
If I say
J3, 10,20,-5
MY,50
then send the bot to 20 different positions around the table with M3 or J3 commands, then go
J3, 10,20,5
J3, 10,20,-5
MY,50

can the second MY move be inaccurate in any of the 3 axis by the maximum published tolerances of ShopBot?
If so, do I add position and cutting tolerances together?
As it is published + and -, is the possible 'error' double the combination of both, in the worst possible case?
............Mike

[gerald_d]

Mike, if your budget doesn't balance at the end of the month, do you start analysing the accuracy of your calculator?

I live under the assumption that the PC's accuracy of calculation is a couple of orders better than the flexing and lashing blue beast with joined racks on double-sided tape, grubscrew-held pinions, turnbuckles, bedsprings, soft rails, etc. Why argue about little numbers your screen......because they are so numerical? Don't let the digits lull you into a sense of accuracy. You and I can measure exactly the same groove with the same digital caliper and get different readings - my thumb may push harder than yours.


Having said that, I take the published accuracy figures with a huge dose of salt. Too many factors influencing the accuracy are not defined, and you are not going to get the SB company to tighten up on their definition because that will close convenient loopholes.

[mikejohn]

I guess what it comes down to is this.
Inaccuracy occurs for a number of reasons.
Often one inaccuracy may be reduced by an opposing inaccuracy.
And somewhere within this, dwells the theoretical accuracy, which is, in fact, unmeasurable and indeterminable amongst the general inaccuracy.
So it doesn't matter any way.

I think I will forget the NASA contract



...................Mike

[rhfurniture]

My guess is that provided you repeated the step before the one you want to measure to take out any mechanical slop variations, it would be damn near, provided you havn't lost steps.

PS: Absolute exists only in Black holes.

R.

[mikejohn]

I'm thinking of using Black hole dust as a means of hold-down.
Just a bit worried where the ShopBot might end up if I hit it with the bit!

...............Mike

[rhfurniture]

Actually if I remember right from when I used to browse cnc specs, some makers quoted repeatability factors (shopbot don't) which was usually some stupidly small number, say 1/100th of the accuracy figures.
Of course cutting pull will make a monkey of it all anyway. (sorry monkeys, no offence meant)

PS: Absolute also polutes the mind (unless you are a mathematician)

R.

[jsfrost]

Sorry Mike, I can't answer your original question either.

When it comes to accuracy, there is often the sssumption that all inaccuracys should be summed to find the worst case accuracy. In an absolute sense this is true, but in a practical sense it typically is not. So long as the individual errors are not correlated, probability comes into play. I could win the lottery if I was dumb enough to buy a ticket, and all the errors on my SB could be additive. But in the real world neither is likely. As Mike said, "Often one inaccuracy may be reduced by an opposing inaccuracy." I might substitute usually for often.

Never the less, it's valuable to understand as many sources of inaccuracy as possible.

Jim

[Brady Watson]

In theory...if you want to be theoretical, if you are not running a spindle, you are already giving up at least .004-.009" via runout from the router. The longer the bit...the greater the error. Let's also point out that it is possible for your material to swell or shrink from the time you left your warm dry shop at the end of the day, to the time you come back in the morning to a cold damp workplace.

Any number of factors will eat away at your bottom-line accuracy. I find it helpful to delve in and deeply focus on the theory of why my machine hasn't cut the way that it was supposed to (rare occourance these days), and then, just as important, zoom back out and take into consideration variables beyond my control, and those that we just have to live with.

I think that if a noobie were to view this thread, they would think that these machines were as sloppy as a damp flounder...not the case. So the question seated in reality would be, What kind of errors am I getting? and How much accuracy do I really need for the types of materials that I cut? (all the while bearing in mind expansion and contration of wood materials and metal machine components)

Mike,
You say that your cut is off by xx.xx mm...Did you mic the bit? I have learned that I have to mic router bits...end mills are spot on. Some of them can be off as much as a 1/64" in advertised diameter.

I took the liberty of posting this from the SB Manual:

ShopBot Users Guide Page -- 100 & 101

TOLERANCE BUILD-UP
Numerous small problems of tune and adjustment, over time, can contribute to
"tolerance build up" and start showing up in the quality of your cutting or
machining. Here’s a list of the things to keep an eye on:

1. Your step resolution is .00136 and there’s not much you can do about that.
Keep in mind that this number represents the absolute best that your tool could do
for accuracy, and it means that every step it takes is a digital move of this length.
This digital stepping can show up on diagonals and curves.

2. Check the collet - it should be tight, clean and in good shape. High mileage
collets belong in the trash. Good quality spring collets advertise a life of <700
hours. As the collet goes bad, it no longer holds the bit correctly, but it will cause
the router to vibrate from asymmetry.

3. Check you bit. Dull and out of balance bits are the most frequent cause of bad
cuts from chatter and uneven movement. Improperly sized bits can also increase
deflection. The type of bit you choose can also have a good or bad effect on edge
quality. Your tool has a lot of power, but you don’t want heavy forces put on the
bit because it will flex the bit and your tool and put a serious strain on the cutting
process.

4. Check the run-out (side to side) and end play (up and down) in your router
bearings with a dial indicator. You will find most new off the shelf routers with a
range of .004”-.006” in both directions. Unless you spring for a spindle, you will
need to live with this run-out. Be aware that as the bearings wear, this run-out will
get worse. When the run-out gets bad, replace worn bearings with new ones.

5. Make sure your part is held down properly. $100,000 CNC’s that hold very
high tolerances still have a ¼” error when their part slips ¼”. Constantly evaluate
how you are holding your parts. Consider a special jig or vacuum box. A vacuum
table will not only hold the part horizontally, but will prevent it from vibrating up
and down.

6. Check the Z axis: Grab the Z-axis at the base of the router (please unplug) and
apply pressure along the X and Y-axis separately with the stepper motors powered
up.
*Each bearing on your Z-axis should be tight and riding two surfaces for the
length of the rail.
*The Z axis has eccentric adjustment nuts on each of the four bearings and
will need adjustment as the Z-plate wears on the ground edges. Rotate each nut
slightly, counter-clockwize to tighten the Z car up.

7. Now check for up and down play that can come from backlash on the rack and
pinions.
*The stepper motor pivots into the rack and can be tightened to remove any
backlash.

8. Next, we can check for any play in the X- and Y-axis carriages that might
indicate a loose set-screw and pinion that is rotating slightly on the motor shaft.
With the motor on and the tool not moving, push the carriages back and forth.
They should be locked in position.
*If a pinion moves with the carriage and the shaft remains still, then tighten
your shaft set screws.

9. Rack and Pinion Angle – It is very important that the pinion gears fit into the
rack at the correct angle. The motor should be perfectly perpendicular to the rack
and approximately horizontal when sighted down the track.

10. Spring Tension: Too much tension on the motor springs can add resistance to
the stepper motor and take away some of the motors power. Not enough tension
can create backlash between the rack and pinion gear or allow the gear to hop out
under load.
*To adjust the spring tension, pull the pinion gear up firmly into the rack and
take the slack out of the turnbuckle, then rotate the turnbuckle 3 full turns.
*Lubricate the Rack with bearing grease.

11. Hold down bearings: All four hold down bearings on the Y axis should be
tight enough to just turn in with your fingers. This means they won't spin easily,
but not be so tight that you can't spin them at all. They should also be adjusted to
be slightly preferentially touching the bottom side of the bearing – and thus
pulling down a bit. This will prevent the Y-carriage from rotating and lifting up.
*Tip-If you are using a single Z-axis, you should mount the router as close to
the Y-motor as you can. This will shorten the lever arm between the drive point
(the pinion) and the cutting point (the router bit) and thus provide the most rigid
support.

[gerald_d]

"if you are not running a spindle, you are already giving up at least .004-.009" via runout from the router." Here we go again.....

Brady, where do you get this factoid that everything other than a spindle has a pre-defined runout of .004-.009 inches? This type of statement destroys your credibility.

[Brady Watson]

Gerald,
I'll let YOU do the research to answer your own question: here (http://search.yahoo.com/search?p=router%2Bshaft%2Brunout&fr=FP-tab-web-t&toggle=1&cop=&ei=UTF-8)

No 'here we don't go again'. It is a matter of a finer piece of hand built & tested equipment. Just like my Harbor Freight crappy digital caliper is NOT on par with a Mitutoyo or Starrett digital caliper. If you would like to group yourself in with those disillusioned that believe that they are the same thing, well, you go right ahead.

-Brady

[jeff_guinn]

Dirk,
There is nothing wrong with your spread sheet.Here is some info I finally was able to locate in my Usersguide-
Unit Values for Different Gear Ratios (Rack & Pinion Tools Only)
A Rack & Pinion ShopBot can come configured with a variety of gear ratios. This flexibility allows users trade-offs between speed on the one hand, and power and resolution on the other. To configure a ShopBot for a particular gear ratio, the correct Unit Value for the X and Y Axes must be entered using the [VU; 'V'alues 'U'nit] Command. The correct Unit Values for various gearing combinations can be found in the following table. For your information the Unit Value is a number that is 1/4 the number of motor-steps/inch. (Note that if you have a cable-drive ShopBot, Unit values for your specific tool are determined with the calibration Command [UC]).





Unit Values for Gear Head Motors:

Gear Type
18 tooth pinion
20 tooth pinion
22 tooth pinion

Standard 3.6 Gear Ratio
127.3240
114.5916
104.1742

Hi-Res 7.2 Gear Ratio
254.6479
229.1831
208.3483

This 1/4 value must have something to do with the way the controller works.My steps per inch = 1018.5916 indicating 1/2 stepping.
Thanks,
Jeff

[gerald_d]

Brady, I did my homework when you started spreading this factoid. We have 2 Metabo die-grinders used as routers and 2 Makita routers. All four of them have less than 0.001" runout, measured with a Mitutoyo dial gauge. It is patently not true to state that ALL routers have as much runout as you claim.

[Brady Watson]

Gerald,
I never stated that ALL routers had X amount of runout. I am simply pointing out that it COULD have that much runout & that it would be a place to look when tracking down accuracy issues. I DO know that all of the Porter Cable routers that I have seen on other Bots were very hot to the touch and had noticable wiggle to the shaft whether new or old. Spindles I have setup do not. I am not sure how well the Metabo or Makita are engineered, but it sounds like a real feat of mass production to exhibit only .001" RO. My spindle was hand-built and it has, according to the hand-filled sheet, .0002" RO. Did the router or die-grinder some with one of these sheets?

If we are discussing accuracy to the degree of micro-stepping, pinion lash and micro-adjustments, it is only appropriate to examine the center stage of the entire machine; the cutting tool. This means whoever thinks that their router COULD have runout, then put a dial indicator on it yourself and see. These things are mass-produced and the castings, stampings and bearing sets that go into a router have +/- tolerances as well. You must have blocked out the 1st 8 words of my post that prefaced my comment with: "In theory...if you want to be theoretical,..."

Gerald, if you disagree with something that I say, there is an appropriate way to address that, which does not involve name calling and words about my reputation. That sort of thing does nothing for others who view these topics and only cheapens the words you type, IMHO. If my facts are incorrect, perhaps the best way would be to include your findings in a factual way online and then send me an e-mail if you have something else to add not related to technical facts...this includes taking things out of context as you have done above.

-Brady

[bleeth]

Perspective check: There is no doubt in my mind, without measuring, that a Metabo made tool is a smoother and more accurate tool than other similar pieces. Makita, on the other hand, has a known history of being quite variable in their quality between one model run and the next. My 3HP Mak plunge router, which gets very occaisonal use and almost none since I bought my bot, deteriorated at a very unacceptable rate with regards to runout. That being said, we all know that most router equipped bots are set up with PC's, and frankly, they ain't 1/2 the tool they used to be and never were designed for machineing precision in a cnc. That being said, the weakest link I can identify in the inaccuracy chain comes a whole lot more from the gantry than the router.
My problem isn't now and never has been a wobble of .001 or .005, but rather a 24 X 85 inch panel being out of square by a good 1/16" after I squared the y to paper thin guageing. Hence my previous and ongoing interest in gantry stiffining.

Dave

[mikejohn]

Brady
Yes, I did mic the bit.
It came out at 5mm.
The single cut groove came out also at 5mm. At least I was unable to measure any significant difference.
It was when I made two cuts that I managed to insert a .005mm feeler gauge between caliper end and edge of groove.
This is what inspired my original question.
Should I be chasing down some error of set up (although with .05mm I doubt I would). or because .05mm is within published specs, can this be occurring on a perfect machine? The theoretical question.
I guess what I am asking is, when is it no longer reasonable to try and correct inaccuracies with a router on a PRT96 (new version)?
My own answer is, when experience tells you, you have got is as good as you are going to.
As far as the router / spindle debate, I seem to remember this Forum agreed ages ago a spindle was superior to a router, but a router was suitable for many peoples uses, and lots cheaper.
What I understood Gerald was saying, although it is not for me to put words in his mouth, is that you provide a wealth og great information to this Forum, as indeed he does, but sometimes your enthusiasm for one particular 'method, appears to blind you to alternative possibilities.
At least, that is how it sometimes comes across.
Also, be certain that what you write says what you mean exactly, and doesn't say something else.
If a message is misunderstood, the fault always lies with the sender, never the receiver.

............Mike

[bleeth]

Mike:
"If a message is misunderstood, the fault always lies with the sender, never the receiver."

I guess that's why I find myself telling guys out on the floor several times a week something that goes along the lines of "No, your other right hand".

Dave

[mikejohn]

Dave
It's a standard O & M lesson.
If you asked your people what you said, and they said "right hand" then the message has been received correctly.
The fact they don't carry out the instruction is another thing entirely.

On one occasion in my flying days, working in very congested airspace, I told the controller I was turning right onto a particular compass heading in 3 minutes.
3 minutes later I instruct the pilot( I was a navigator) to turn left onto the compass heading, which he duly did.
A voice from the ground simply said "Oh! that right!"
This exchange (within the London control zone) must have been heard by many pilots.
Embarrasing, or what?

...........Mike

[gerald_d]

Brady, here are a couple of places where you have trashed the choice of tools of folk like me. I have called you on it once before, and you retracted then.

"Even if your machine was perfect, expect it to be out .008 from the start with a standard router." link (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=7&post=23129#POST23129)

"runout compared to PC @ .008" when NEW" link (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=27&post=21728#POST21728)

"a PC has .008" RO right out of the box when new." link (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=29&post=25223#POST25223)

"a factory-fresh PC router has .008" of runout" link (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=312&post=26892#POST26892) Retracted

And then in this thread we again see a statement like "if you are not running a spindle, you are already giving up at least .004-.009" via runout from the router."

If you want to be a tall tree standing out above the forest, you need solid branches to withstand the wind. This is a public forum and if you want credibility you need to be sure that your statements of "fact" can withstand public scrutiny. This (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=29&post=22555#POST22555) little exchange has not been forgotten either.

[Brady Watson]

Gerald,
Let me just say that I am flattered that you would spend that much energy to make your point.

The day Komo gives up on spindles and goes to Porter Cable routers, will be the day I put up my 5HP. If routers were 'just as good' as spindles, then spindles would no longer exist in the marketplace, except for suckers like me who 'wasted' their money on one. Regardless of what router/spindle/die grinder you use, it makes your ShopBot no more or less valid than the next guy's. Intent is everything. I never come from the standpoint of belittling someone because they have a router. If you believe otherwise, well you'll have to take responsibility for your own beliefs.

Let's get back on topic. The thread is about accuracy. Check your spindle & or router runout as a potential source of slop in the machine.

-Brady

[paco]

3243

[slendon]

Whilst not wishing to fan the flames, it is quite clear Brady that you did state(in your Jan 6 11.44 post) that by using a router you ARE (not could be) loosing .004/.009 in runout.

No amount of 'spin' will change the meaning of what you wrote...

By the way, I appreciate, as I am sure do many others, the posts made by both Brady and Gerald.

The rigour with which all our posts are scrutinised by others is an important element for the on going value of this forum.

Thanks again for all your posts to date.

Steve

[Brady Watson]

OK. I'm wrong, you're right.

Now let's move on.

-Brady

My Jan 6 11.44 post is prefaced by the words 'in theory' and 'theoretical'. No spinning necessary. The definition of the word 'theoretical' is not synonymous with the word 'absolute', and affords a certain amount of latitude when explaining something. This is why we have the word, 'theory' in our language. Omitting relevant words that modify the meaning of a post when quoting, is like tabloids when they don't tell the whole story.

[gerald_d]

Mike, my apologies for my contribution towards the fracas in your thread.

I think we need a standard .sbp "test file" that each of us can run on our own machines, where we can then measure and compare real results with each other and against factory specs. For example, how about a file that simply cuts four 2 inch wheels nested in a tight pattern on a 5"x5" square piece of 3/4" MDF. The cutting is only 5/8" deep max, all with a 1/2" cutter. Simple stuff - but, the file must cut two wheels as conventional and two wheels as climb-cut. Also, the file must contain different speeds (light and heavy cutter loads). And we need helical entries for those wanting to keep plunge loads light.

Should we have such a "standard" testpiece, then we can take a digital caliper and compare the roundness and diameters of the wheels, also on the diagonals. All the deviations measured off the theoretical 2" diameter would be attributable to "accuracy".

I think that we could have some fun designing such a test.sbp file and running it all over the world and then comparing results. We could have a standard table of measurements to fill in and our resident programmers could give us a little utility that reads our results table and then tells us things out our flex, squareness, backlash, cut width (bit diam/runout).

A test.sbp would need to be in an inch and a millimeter version. We need to agree on the measuring instrument (probably a 6" digital caliper) that would give consistent results for everyone.

Anyway, I just wanted to plant this seed of thought for the guys wanting something to do while their shops are too cold to work in........design a simple test-piece that anybody can run run anywhere. (similar to the test page of a printer?)

[mikejohn]

What do you mean,cold?
I would tell you what it was like outside, but all the windows are frozen over, and I can not see out. Rumour has it, it's around -17ºC (7ºF) outside!
Run that plan about moving to Africa by me again.

[Support]

Thread closed while admin tries to sort out useful info from the rest.