Lost Tolerance?

[erik_f]

It seems my bot is not producing parts vary accurately...some of the parts are dead on and other can be off as much as .05" maybe more...also my circles come out with this too...the problem is that the same circle will have multiple diameters when measured and I measued one that was out of round by .064"...any one have any ideas? Software maybe?

Erik F.

[artisan]

Erik, I always stress looking for repeatability of your errors when this sort of thing happens....and it happens to most all of us sooner or later. Generally speaking, if you can repeat your errors, it's usually a software issue. If errors are random, they are usually mechanical.
You mentioned different sized circles. Try a test cut in some foam or your spoilboard. Cut a circle, then move the bot over 6 inches and cut it again (just scratching the surface so you can measure it). Are they the same? Increase your dimensions x2 and cut another....are the dimensions exactly double (meaning the error doubled exactly .128). If this all checks out it's most likely software and the Bot is just following orders. Post your software type and a little more of your process and someone on here will help figure it out...D

[erik_f]

Thanks Darrell,
I'm using BobCad to design with and Fanuc 6M as the post processor to produce the G-code...then converting to .SBP...I will try and run another file tonight on some scrap. Up to this point it has been quite a struggle to get the shopbot to proform like I had expected when I bought it...I guess I'm just paying my dues...

Erik

[artisan]

Your Bot will impress you I think...when you find the way to communicate with it, but, it's only as good as the commands you give it. I'm familiar with the Fanuc....it should be fine. Brady Watson and some of the other guys on here use BobCad and will be of more help there than I am. Once you run a couple of test cuts you'll know more and someone will spot your problem and help you fix it....D

[erik_f]

Thanks again...what is your experiance with the accuracy of the bot...as well as repeatablity? I know that the spec is +/- .015...but my understanding was that was worst case...on a general basis...I was really hoping to do much better than that. I'm not making aircraft parts or anything...but the whole idea behind the getting the bot was it would be faster and more accurate than me...it can produce complex shapes much faster than me, but if they aren't accurate...whats the point? Thanks for letting me rant a little.
Erik

[artisan]

No problem Erik. What ARE you producing? I do lot's of Signs and 3D sculptures. I own a theming company which produces unusual one-of-a-kind pieces for theme parks, malls, restaurants etc, so a CNC Shopbot is ideal for me. My personal experience on repeatability is in the .005 range. This is not constant however. I am constantly resurfacing spoilboards and checking my "ways" for wear. If I have some really critical work, I spend as much as half a day setting up the table before I start. There are too many variables like the weather, which can affect your work once you get into the really tight tolerances. I seldom accept this type of work anymore... though I just engraved some aluminum motor heads for a speed boat. I choose instead to use the Bot for what it's best at....creating complex shapes that can be accurately machined in 2 or 3 dimensions. I have saws that will cut straight lines all day, but the Bot can actually translate my computer designs into complex parts that fit perfectly over and over....if that's the way I drew them. I look forward to seeing your results posted. Good Luck....D

[erik_f]

I'm producing high end speaker systems...I'm using the bot to produce some designs that have been in my head for about 10 years, but would have taken that long to produce if I did them by hand! I did compare the piece I told you about to the other pieces for the last cut and the two and the similar pieces mic out to with in .003" of each other...even though they are the wrong size...the part I don't understand is there are 4 pieces with the same outside specs the only thing that is changes are the holes in them for the speakers...so there are 2 sets of rectangles. So my guess is since the like rectangles are coming out almost exactly the same the shopbot is doing its job right...maybe the software is wacky...I am going to run the same file again on some scrap to see if the piece come out exactly wrong again.
Thanks,
Erik
3944

[jay]

Erik, welcome to the club. I wish I could offer some specific help, the learning curve for each of us is different.

When I first set up my Bot it did things I did not expect it to do. A number of the files I had prepared in preparation to receiving my bot cut a little differently then expected. I was very disapointed and wondered if I had made a lousy purchase. But in the end the machine was only doing what I had told it to do. I have not had any machine errors as yet, they have all been operator errors.

To say there is a learning curve is somewhat of an understatement.

[erik_f]

Thanks for all the help guys...I did cut a second batch of the same file...and the mistakes are the same...from old batch to new the parts are with in .006"...I could live with numbers like that...now if I can only get the parts to come out that tight AND the right size...we'll be cooking with canola oil! Must give a call to BobCAD tommorrow...Thanks again.

Erik

[gerald_d]

On tolerancing problems, I would look only at mechanical features, and definitely not at the software. The normal culprits are:

Out of square setups
Unit values
Loose gears
Loose motor tension
Loose V-rollers on z-axis
Bit diameter
Router bearing run-out
Bit deflection
Blunt bit
Direction of cut
Loose router mounting
Etc.

Most of these factors will give repeatable results.

[Brady Watson]

Lookin' good, Erik!

You may want to check out glassman's custom forum on Yahoo groups. Lots of good ideas on enclosure construction.

I like BobCad for cranking out DXF files. I haven't done much toolpathing with it since I use PartWizard for 2D stuff and Rhino/MillWizard for 3D stuff. When I did do toolpathing with BC, I ran into similar issues and found it to require more time and thinking on my part than I wanted to invest...and also don't like wasting material on failed toolpaths. Call me spoiled, but I like using the DelCam stuff because it is reliable and I just have to fill in the required fields and it does the rest.

-Brady

[bill.young]

Hey Erik,

I think Gerald's on the right track. You might try something simple like cutting a couple of squares with the CR command (one clockwise and one counter-clockwise) and see how they turn out. If they're off, is it the same amount in the X and Y axis...if not look for something that's loose in the axis that's off? If it's the same amount in both axis, try cutting the rectangles twice the size...is the error doubled? If so maybe it's a unit value issue? And maybe try cutting them in different parts of the table...if there's a variation then maybe your rails aren't quite straight

Are they both square but one oversize and the other undersize? There is always some of this because of bit deflection and router runout, but if it's excessive look for bit or router problems.

Just a few things to look at,
Bill

[erik_f]

I will try the tests and check what you guys said, but I did run the same file twice on scrap last night and all the pieces came out exactly like the first run...and from first run to second run they were with in .003-.006". I don't know if that tells you anything...I don't think I'll get a chance to do any major testing until this afternoon...thanks for the help everyone.

Erik

[gerald_d]

Just out of curiosity, what measuring tools are you using?

[erik_f]

I am using a digital caliper (6")...I really think it has something to do with the way BobCad and the bot are talking. More over how BobCad is writing code...and the bots controler sees the directions. There are many settings in the NC side of bobcad that I don't really know what they mean yet...I haven't looked yet, does anyone know where I can find specs for the bots controller and what kind of directions it needs? Not sure if the question makes sense...but I am a total newbie to all of this...Thanks everyone.

[artisan]

Erik, this is easy to check. As Bill suggested earlier, use the CR command....and I suggest the CC command as well. Be sure and run each command in clockwise and counter clockwise directions. Measure your circles and squares. You can then determine if you have (A) a software or post processor issue and/or (B) deflection...which will simply mean a different bit or cutting strategy. If they are dead on...it's Not your Shopbot...D

[artisan]

PS...in the future, this type of thread probably belongs under "Troubleshooting" or "Cutting Techniques"....D

[erik_f]

I did run a couple of squares and circles just from the control software the and in opposite cutting directions. The circles still seem to measure different diameters when measured across at different points...Grant suggested that the table is out of square...I did not have time to run a large rectangle test and measure the diaganals...but I did recheck by clamping a board of MDF to the X rail and using a square to see how close I am on the Y...from both sides of the table...the square as dead on...I will run the rectangle test tonight. Thanks

Erik

[erik_f]

Made a 30" x 50" rectangle and measured corners to corners and to my eyes its dead on square...checked the Y pinion gear...its tight.

[bob buttons (Unregistered Guest)]

I have been having the same problem for a long time now and just learned to compensate for it since there wasnt an answer for me out there. my machine is square and the cr/cc commands run true but when i convert dxf's from my software and run them my sizes are unpredictable at times. sure bit deflection is a problem but just how much?? i have tried to watch the xyz positioning while jobs run to see if the shopbot is being told to go to the wrong location but i havent noticed that. it seems fine. yet my sizes are still off. im not sure what the answer is?
is a dxf a dxf?

[grandpaspastime]

I had a similar problem and was not x or y but some slop in z bearings. also found that the router was running to slow and forward speed a little too fast. corrected the bearing tightness and set vatriable speed router up a couple of notches and slowed feed rates down a little. seemed to fix the off.

[artisan]

Re: DXF Files. Gerald posted a piece on DXF and DWG files you can search in the archives that will shed some light. There are essentially 5 major flavors....R12, R13, R14, R2000 and R2004, with R14 being the most commonly encountered in my experience. They all translate similarly, but NOT exactly the same. Usually by updating your software you can solve the problem, however there are "translators" available if you have problems. I've never had a problem with the Shopbot DXF converter, but I really don't use it much anymore. Perhaps Shopbot could shed some light on the built in converter and whether it is backwards compatible with all the formats....D

[gerald_d]

All dxf's give accurate results. They do not introduce a scale or tolerance inaccuracy. The so-called "bad" dxf's are those that your software will refuse to open. But, if your software can open it, it produces accurate lines.

If your dxf has a line 30.000 inches long, and if your SB's unit value is set correctly, the SB's motor & gear & rack will travel 30.000 inches.....always.

Edited to add: You can open a dxf file with a text file editor, and you can see the coordinates that are ascribed to the start and end of lines. Those numbers always correspond to the length of line that you drew.

[erik_f]

Well, I've now checked the pinion gear, slop in the z-axis, the table for square...everything seems to be in order...BobCad has looked over the code and they say it has no flaws. Perhaps the control software has been corrupted? At first I thought it was that the larger parts were coming out correctly, but I have just taked a look at them and some are correct and some are not...by at least 1/16th. They are still suggesting bit deflection...but that would mean that I can cut in one direction and get good results and if I cut in the other direction the machine is unreliable...besides, I'm using a 1/2" bit, with a 1/2" shank...cutting .75" (or 19.05mm for Gerald...I know how much he likes it when we Yankees use mm
) at 1 inch per second in two passes! I really don't think that is my problem...I'm going to make sure everything is in the same direction and cut it again anyway...but doesn't make sense to me if the machine only is accurate in one direction.

Erik

[bob buttons (Unregistered Guest)]

please keep us posted i would love to know the answer here. 1/2 bit 2 passes at that speed. deflection should be minimal. .01" would be a lot with those parameters. if you get a chance heres the real test. put a pen in your shopbot then measure the pcs. deflection will not be an issue then. if still bad what is going on!!?? im hoping for some dxf conversion problem cause ive run out of ideas on my end.

[erik_f]

Bob, do you mean to draw the parts with a pen...like a sharpie? Both shopbot and bobcad have looked at my files and they both say there is nothing wrong with them...both the dxf, tap, and sbp files. My hope is that it could be something corrupted in the control software that may go away with a reinstall, but I have no real reason to beleave t his will work. What kind of computer are you using? I'm using an old e-machine...I know this may sound crazy, but it could be a problem with the processor and this specifc version of the control software... I've heard of some cad programs not working on some PC's with certain processors...so I'm going to swap in another PC ,I know its a long reach, but I'm going to try this as well, but I also don't think it will change anything...but at this point I am mystified...and something needs to happen soon I am nearing the end of my rope. I have only run 4 projects on this machine and all of them have come out wrong is some shape or form...I had to send them out anyway...and one customer is never coming back...but he was the biggest...how can I promote a CNC service if I can't produce simple boxes better than what I could do with hand tools? The worst part of it all is that I doubt myself...and wonder if there is some small simple detail that I have missed and that is why everything is not working right...but in the end...I cannot find it and am sure I followed the manual to the letter and double and triple checked everything. I suppose my frustration is showing through, but I think to myself...either the other folks that have put these together are way smarter than me...or there is just something plain wrong with something somewhere in the shopbot, and I'm not talking about how I bolted it on...

Erik

[gerald_d]

Erik, I would say that you are totally wasting your time looking at software and computers. If you are getting out-of-shape circles, then your problem can only be something loose, or something deflecting too much.

Too much deflection is caused by too much force.

Blunt bits, too high a move speed, are the main things that increase the cutting force. (a 1/2" bit feeding at 1" per second is quite a substantial load.)

Unplug your router and push the cutting bit from side to side with your hand. In x and y directions. If you see any deflection there, that is the cause of your circles being out of shape.

[gary_n]

Erik,
I think Bob's point with the sharpie is that if it draws the image OK, then that would seem to eliminate any software problem, but suggest it may be have something to do with the bit and cutting speed. If it draws the image the same as it cuts it, then there may be a problem somewhere in the machine.

[artisan]

Erik, I am leaning towards Gerald's view. If your Bot won't cut identical squares and circles from within the Shopbot software, then you have a problem somewhere in the table. Perhaps something as simple as how your router is attached. If I understand you correctly, you ran the CC and CR commands in opposite directions and got different results? This would suggest that it is NOT a software issue, but a mechanical one. If you are cutting only deep enough to scratch the surface, then it's really hard to imagine that deflection is a problem. Instead, it really sounds like your table is out of square, or something is loose....D

[gerald_d]

And remember to report your problem directly to the ShopBot people. They do not necessarily monitor this Forum.

[erik_f]

It would seem the problem almost has to be mechanical at this point...I guess I will just have to take a long weekend and break everything down. I used a sharpie to make 4 points on the table and measured the diagonals to my eyes it was dead on...I checked the pinions gears, but and going to totally remove the steppers and double check them that way this time to be extra sure they are on the flats of the stepper shaft. While the table is on and idle...I have pulled on the bit and there seems to very little if any play that I can detect...if I pull on the y car will alot of force it will move a little, but I'm sure I'm putting more than 80lbs of force onto it...I weight 220lbs...and the same for the x-car...Question...in your guys experiance...with a machine in proper running order...if you were to cut a bunch of 5" squares...how close would YOU expect them to be? Just trying to get an idea of the experianced botter's out come. Thanks for all the help guys...I know it seems like I'm griping, but...this is ALL new to me and I'm on unfamiliar ground.

Erik

[gerald_d]

We regularly cut perspex(plexiglass?) rectangles for a guy who wants 0.2mm accuracy (0.008") and we give it to him by tweaking the settings. i.e. we cut one, measure the error, compensate for the error with a new setting, and than run the batch of say 50 pieces - all of them within 0.2mm of each other.

I must add that we cut critical jobs with supersharp 6mm bits moving no faster than 1.5" per second, and plunging a max of 3/8" per pass - the load on our SB is probably less than the typical SB'er.

We also don't do clean-up passes. Clean-up passes only load the bit on one side, causing deflection and an unexpected final dimension. I mention this here because you be trying to cut the same blank circle smaller and smaller while you are doing your tests, and this could be giving you the odd results. Your bit must always be cutting a full-width "groove" if you want a repeatable test.

[gerald_d]

Related to my post regarding clean-up cuts above, let's say you want green circles out of yellow blanks:


3945

If your blank is so small (left) so that the bit breaks out of the sides, there will be a change in the amount of deflection at those 4 spots, and the resultant circle will not be perfectly round.

A bigger blank (right) is one way of solving this problem.

[erik_f]

Thanks Gerald...The pieces are cut from a solid sheet and they do not exit the board and then re-enter it to make the cut. The holes are cut for speaker drivers...the smaller speaker drivers I measure with the digital caliper before designing...then I add a little extra to make sure they slide in with ease...they are also recessed so they are flush mounted on the board...but the process I use to do this is...I make two files...one for the mortis for the flush mount and one for the hole for the actual speaker driver to fit into...the mortis is cut first into the wood, then the second cut file is run with tabs...so there is always wood on both sides of the bit when I'm cutting these files...I also am not running any sort of clean up as the MDF is cut so cleanly there is no need...I get the tabs off with a trim router bit and router table. I just rechecked the table for square after looking at all the pinion gears...everything on that end is in order...what else could be loose? If there was bearing run out on the router...I can try running these files as .5" per second in 3 passes...I will have to bring my sleeping bag, but this would take almost all the load off.

Erik

[erik_f]

For Bob Buttons,
Got an e-mail that someone had the pinion gear of the Y car was rubbing on the Unistrut of the X car rails...they said this was causing him to have variations similar to ours...I'm going to check mine and see if it is.

Erik

[gerald_d]

When you do a full circle, does the finishing point end up exactly at the starting point?

[K Simmerer (Unregistered Guest)]

Double check your distance between the X rails. If this is off the rails will ride at varying depths in the v grooves causing all kinds of accuracy problems. This certainly sounds like a mechanical problem. I am still curious to hear the results of the cut circle and cut rectangle commands CC, CR.

Keith

[gerald_d]

Also check for a straight, parallel & correct distance apart set of rails for the y-axis as well. The y-axis is more sensitive to the problem, that Keith mentions, than the x-axis because there is less weight pushing the v-rollers into place. A spacer rod, cut to fit between the rails is useful to check with at various points.

[erik_f]

Yes, Its seems that the circles do not line up with where they started from...it is very hard to see, but from my estimation that is what is happening at least with the circles. As far as the CC, and CR commands...the squares were not perfect nor were the circles...and I ran them in both directions...I can give a better account of the size and variation of the clockwise and counter clock wise numbers later...but the circles did seem to measure out of round...When I checked for square yesterday...I chucked a pen into the collet again and put piece of cardboard in the middle of the table and proceeded to make points for a 20 x 32" rectangle...corner to corner was dead on(I don't have any way to measure this but with a tape...but the point was made with a fine tip pen)! This is partly why I'm confused...and why if it is not something software...points to machine flex or runnout of some sort...but a 16th of an inch is alot even for run or bit deflection wouldn't you say? The router and collet only have 10 hours on them. So another question...if cutting clockwise and getting the parts coming out nearly perfect, the cutting counter-clockwise, shouldn't the parts come out equally as well and any flex or run out would just shift the entire position the piece , but the size should remain the same...the only thing I can account for is the direction of the cutter spin some how effecting the pull or push into or away from the wood...but I would think that as long as everything is tight this wouldn't really matter unless edge quality was an issue, but in this case it is not...am I totally wrong...am is my thinking sound?

Erik

[artisan]

I'm reading conflicting accounts...I think. First you've stated that with the CC and CR tests "the squares were not perfect nor were the circles". Then you state that "a 20 x 32" rectangle...corner to corner was dead on". Please do this for us....Chuck up your fine point pen once more and move your router to the center of the table. Using the CR command, type in a 12 x 12 box and run it. From the same starting point, now run the 12 x 12 in counterclockwise fashion.

Is there an offset? What is it? Where is it?

Repeat this test with a 12 inch circle. Clockwise, then counter from the same start point.

Is there an offset? What is it? Where is it?

Increase the the size of the square or circle to 4 ft and run again in both directions if there was an offset and let us know this measurement as well please....D

[gerald_d]

Erik, looking from the top down onto the workpiece, with a clockwise turning bit, deflection makes the cutter move to the left of the direction of travel. Therefore, a clockwise cut circle (or any shape)will be bigger than an anti-clockwise cut circle (or any shape). It is not a shift of position, it is a change of size - the deflection direction depends on the direction of cut.

1/16" errors are totally unacceptable.

[erik_f]

Darrell,
I actually cut into wood with the CC, CR commands, and for squaring the table...I only made points with the pen...I didn't actually drag it to make lines, only corner points and measured those...I'm at my day job now...so all testing has to be done at night or on the weekend. How much of a difference in size would one expect from a clockwise compared to a counter clockwise cut? Assuming of course the machine is in good running order.

Erik

[artisan]

None

[erik_f]

Ok, so there should be no change is size...two posts before Gerald says there would be...but I'm not trying to start a flame war...here, just trying to get an understanding of something I may be doing to cause problems and even when things are right with the bot to be able to get the most accurate cuts in the future.

[gerald_d]

My post was on the theory of deflection, and to show you that deflection doesn't simply cause a change in position. It was in response to this "So another question...if cutting clockwise and getting the parts coming out nearly perfect, the cutting counter-clockwise, shouldn't the parts come out equally as well and any flex or run out would just shift the entire position the piece , but the size should remain the same"

I would also not expect to see a visible difference between the two directions (without fancy measuring equipment) if the machine is properly set up and deflection has been eliminated as far as possible.

[erik_f]

Ok, so I had a chance to run a test...it would seem as if the problem is located in the Y-axis as the circles that are the most wrong when the Y-axis is doing most of the work as well as the squares being only out of spec on the sides that the Y-axis is cutting. The sides when the x-axis is doing most of the work are within thousands...oh and...I made sure everything was being cut in the same direction except for one piece...these two piece even though one is cut count clockwise and the other is cut clockwise match up with no visual differance...Thanks for all the info all...any ideas now? You guy have been alot help...I think we're almost there!

Erik

[artisan]

Yes, I meant with zero deflection. A bit moving through oak at 3 ips would certainly deflect more than a bit moving through foam at .1 ips. These are calculations you will have to make with experience....there is no magic bullet. In the end Erik, once you are sure you have set your tool up properly....and again, I and others are happy to try and help you achieve this, you're going to have run your own tests and calibrate depending on the job at hand. The Bot is reliable....just not perfect....D

[chamcook]

I wonder if the hold down rollers that hit your Y-rails side on could be loose?
As for whether a circle should measure the same regardless of direction of travel I would say it depends on how accurately you measure. If you measure closely enough there will certainly be some deflection. This will be the combination of flex of various parts, play in bearings, play in the spindle,flex of the bit itself.

Here is an example of what we are experiencing on our machine. Last week I was cutting the inside rabbet of a series of cherry window frames and then trimming the outsides of the windows to fit on the frames with an even 1/8" all around. The total error over the 4 cuts (inside of the frame both sides and outside of the window both sides) was around 1/16 . ie 1/64 or .015625"per cut. I was somewhat surprised that it was this much. All the cuts were climb cuts because we were dealing with radiused corners and arched tops and Cherry splinters very readily when you cut against the grain. The climb cut tends to cut a part larger when going around the outside. The error is repeatable and has been corrected by offsetting the edge of the door slightly.
I would guess even the really high end machines have to allow for some amount of deflection. No machine will be 100% tight and stiff. And of course, some of the deflection comes from the bit itself bending. Try cutting hardwood with a long 1/4" bit with a 1/4" shank and you can actually see the bit deflect if you move into the stock from the side. This is what Gerald was pointing out with his pictures of circle cutting.
It is because of deflection that you must have an even load on the bit to get a really accurate shape. When cutting these doors with radiused corners I am carefull to pocket away the waste on the corners before taking a full cut around the perimeterof the door. If I shaved .0625 off of the side and then cut across solid wood when rounding the corner it would surely effect the fit.
It is only when you start fitting together parts with close tolerances that this stuff becomes an issue again. For majority of things I have done on my machine 1/64" (approx 15 thousandths) is not an issue.

David

[Ryan Patterson]

This may be to simple but are the bolts on the y car tight enough. The side of the Y car without the motor might not stay in the same location as the motor side. (deflection in the Y car) I Know some have added a second motor to the Y car for this reason.

[erik_f]

Yes these are tight...the problem seems to be one of 3 things...I will run the tests later today after I sleep for a little while! Even though I have checked the pinion twice and it seems tight...I am going to remove the set screws and reaply locktite to them...maybe somehow the set screws are not firmly seated on the flats of the stepper and after changing position jams the set screw so it seems tight...but when changing position again unjams it and jams it on the other side...give a error in the cut...also I noticed...that under a pretty good amount of pressure...the stepper swing arm wants to push down...so maybe I don't have enough tension on the tensioner spring...I am going to work both of these tommorrow and then if this still doesn't work, then the last place I can look is the unit values...Grant at shopbot told me they had one other case like this, that the unit values somehow...even though right on paper...were not working in real life...but I guess this problem with the unit values would be a real freak thing...but there is a compensation for it under Units that is there for the old cable driven SB's that was there to account for the strech in the cables...Well it is a good feeling to think I've got the problem isolated...even if I don't know yet exactly what is causing the problem yet...at least I know where to look now. You guys have been really great...I'll keep you up to speed...What happened to Bob Buttons? I would think he would be most interested in what happens here...could be the fix for you errors to.
Erik F.

[gerald_d]

Erik, I don't think it is the pinions, because you will then see a step mark, or the machine doesn't come back to the startpoint.

I don't think it is your unit values either, because when you plot big items with a pen, they measure out correctly.

Did you understand the issue with the V-rollers riding up the rails if the distance between the rails is incorrect?

[erik_f]

I actually had problems with the Z-when I first got the machine together...it took me almost a whole day just to get the Y car riding right(riding up on the rails was giving me goofy Z depth problems)...I'm aware of that problem and am sure that the y car is moving smoothly on the rails and the hold downs are very tight and pulling down. To tell you the truth, and this may seem like a foolish oversite on my part...after ploting the points of the large rectangle...I only measured it the long way...which would have been the distance made by the X...yesterday afternoon I remembered that I only measured in one direction and decided to measure the other direction after I started to suspect the problem was soley in the Y axis and sure enough, for at least that plot, the marks were off by almost an 1/8" in the Yr...thinking about it more...the parts that are coming out short were pretty close to 1/16th off and they are 9" in the Y...the rectangle was 20" in the Y...could be just a chance happening, but it seems like the error was double when the distance was doubled...that is what leads me to believe Unit values...but I'm going to recheck mechanicals before I go messing with UVs. The doubling of the error could just be chance. I will know more this afternoon.

Erik

[erik_f]

Ok...so I'm sure I will get yelled at...but I think the problem is fixed as good as it is going to get...at least for now...it was really only a problem of tweaking...and wasn't just one thing...but more a matter of a few things that seemed ok, but after readjusting...made the differance. The first was I put more tension on the Y axis swing arm and gave the swing arm bolt a little bit more tension as well, I ran another file and the problem was still there, but less, so I decided to take the vac system off so I could get the Y stepper and router as close together as possible...and this brought the bot in even a little tighter...I think I will investigate the second y axis stepper as I don't see why the y should be any less accurate than the x axis...I guess it really did all have to do with flexing in the y-car...the router is now as close to the x car as possible without scraping it while it moves side to side...now I just have to figure out a way to put the vac system back on...I think I'm going to have to make a new one as the shopbot verion will not fit with the router that close. I'm still going to see were I can tweak this thing a little more and hopefully be able to get the last degree of accuracy out of it. Thanks everyone...you all have been great...I'll keep you up to date on any other changes.

[artisan]

Though I'm surprised at the severity of some of your adjustments, I'm happy to hear you are getting it tweaked. Maybe one of the best things about the Shopbot is that there is very little that you cannot adjust or tweak on your own.

I don't think anyone will yell at you here....unless you're doing something unsafe maybe. Sometimes....I think some of the posts here are done out of laziness and rely a bit too much on the generosity of some of the talented people on here. Your problem was valid and I really hope you're on your way. One of the benefits of thrashing out another's problems is the learning that comes from it for all of us and especially me. I was entirely new to CNC 3 years ago and now my business revolves around it. I could not have done it without many searches through this forum for answers at all hours of the day and night....D

[gerald_d]

@ Erik (and @ Darrel)

[stevem]

Darrell,

I may be able to help you with the dust collection. My bot also has the Z axis very close to the Y stepper. I am currently designing a high efficiency dust chute that attaches behind the Z stepper. Will post a pic of the prototype soon.

[erik_f]

Circles still seem to be coming out wrong! Supposed to be 4.12"...never hits 4.12" closest is 4.108" and gets worse from there, measuring other cross points it gets as bad as 4.072". That is a change of .036" in diameter in the same circle. This is .048" off spec...my bits are 1/2" and mic out at .501". The squares seem to be coming out better, but still a bit short in the Y by about .020" off of spec. I'm lost at this point.

Erik

[gerald_d]

If you have asked for a 4.12 circle and your results lie between 4.108 and 4.072 that means you are on average .03 undersize. Therefore, ask for a 4.12 +.03 = 4.15 circle and you are going to get a result of 4.12 with deviations of .018 bigger at places and .018 smaller at places. This is pretty close to the .015 advertised for the SB without any router mounted.

Methinks that this is going to be as good as you will get it. Remember to add .03 for that job.

[erik_f]

I suppose, but I thought that the .015" was including router run out. Kind of dissapointing, but I guess if this is as good as it gets, this is as good as it gets.

Erik

[erik_f]

I don't want to come across as a jerk...but have finished running some more parts and am still getting out of spec parts small by as much as .05"...this is way out of spec...even if I was hitting worst case on the bot of having +.015 and -.015 on the same piece. Positional accuracy is supposed to be .005" and cutting accuracy is supposed to be .015"...from everything I've read in the manual and talking to the guys at shopbot that includes the router. Is it really this hard to get the machine to cut better than a Craftsman table saw? I guess I shouldn't write anymore tonight because I'm pretty freakin heated...and there seems to be no end in sight.

[gerald_d]

Talk to the SB people if your system doesn't meet the advertised spec.

[artisan]

Erik.....I believe your machine WILL give accurate, repeatable cuts in the .005 area if it's setup correctly and there are no defects. You can have confidence in that. You must be frustrated at this point, however, and I believe the Shopbot staff is your best resolution now. Where are you located? Can you attend a jamboree to see another Bot in action? Also....feel free to send me a file and I will run it to see if my results are comparable to yours....D

[erik_f]

I'm going to call them again today...thanks

[erik_f]

We are still working on some ideas...but if it was bearing runnout...the local shop that rebuilds the PC router said that I wouldn't notice the runnout by hand...and that it is possible for the runnout to be only in one direction..ie: runnout in the y moves but not in the x...another question is...wouldn't runnout happen in both directions as fars making things too big AND too small? Or is is possible that the runnout would be causing things to be too small only? I know this is getting boring...but I'm looking for anything at this point.

Erik

[erik_f]

I think I fixed the problem...wasn't in the router...didn't really think it was...again...a simple thing...I will of course have to run some more tests, but it seems that the hold down bearing were tight enough that it must have been causeing major flex in the Y car, but not enough to make it miss steps...on a hunch I took off the hold down bearings and it seems to be much better on the y-axis...I ran the same file and did get some variation in the some of the parts, but that could be due to the fact that I as running no hold downs...but all the rest were spot on...I am going to readjust the holddown bearing to a very light setting as it would seem it works pretty well without them.

Erik

[gerald_d]

Check if your so-called "hold down" bearings are not actually behaving as "lift off" bearings. (Remember that I tried to discuss the v-rollers riding correctly on the rails with you two weeks ago?) In our case we took the "hold downs" off permanently because they were more trouble than they were worth - but we generally take lighter cuts than most of you guys.

[erik_f]

I remember...and if it makes you feel any better I do feel like a dumb ass...but it would seem they are nearly impossible to get the hold down bearings to ride right, as the hold downs were only pulling down, but it seem the variations in the rail cause binding is some areas and freer movement in others, for what ever reason the binding was not enough to loose steps, but some kind of flex or maybe lift off...and it also, at least in my experiance, useless to try and true the Y rail by way of the Uni-strut...I think there may be new plans for an updated y car...but in the mean time I think I may have to try and think of a simple easy way to get hold down on the y car as I did notice that after testing with cuts .1" deep then going through 3/4" MDF 3/8" a pass I did lose some accuracy, or maybe try it without the holddowns for a while and just cut slower with a higher spindle speed...So really...thanks everyone again...and the guys at shopbot...

Erik

[gerald_d]

The tension spring oy the y-motor is already holding down the y-car on the one side. Suggest you test this by grabbing hold of the router nose and pushing/pulling to see if you can lift the v-rollers on that side of the y-car (we can't). If you are happy with the amount of hold down on the motor side, then I suggest building a spring loaded idler arm to roll under the y-rail of the opposite side. A normal flat-top ball bearing would make a perfect roller, plus a spring, flat bar, two bolts and a couple of nuts. . . . . .

[erik_f]

I had thought of something similar, but was thinking of useing a small wheel from an inline skate for the sake of smoothness.

[gerald_d]

Anything will be smoother than the gear wheel rolling on the rack on the other side. The v-rollers on top give the smooth ride to the router.

[erik_f]

I did notice under heavier cutting that I wasn't as good as under light cutting, and I guess this may be an obvious answer, but, do you think the variations are coming from the Y car vibration and movement from lack of hold down? The variations are not as bad as when the hold downs were in place, but I did notice a fair amount of vibration in the y car under heavy cutting...I think I'm answering my own questions here, but I just want to make sure so I can decide the best way to go from here.

Erik

Erik

[pacoisa@cooptel.qc.ca]

So, is accuracy with SB is "only" a matter of slow and carefull machining?!


Erik; say more about Y carr update...

[gerald_d]

Maybe you got the vibrations because your y-car was only riding on the hold down rollers for some sections and that a main v-roller (or more) had lifted clear of the rails? This flex that you talk of in the y-car might actually be the flex in the hold down roller brackets, because they are supporting the full weight of the car?

You would get this sort of problem if your y-gantry rails had a twist in them. ie. if you inadvertantly twisted the rails when sandwiching the unistrut channel to them. We couldn't sandwich ours straight enough, so we welded the sandwich together and re-machined them perfectly straight.

[erik_f]

I don't have the means to machine the rail...and don't know if I would feel comfortable making such heavy mods...I guess there has to be some sort of middle ground between where the holddown bearing were set and where they were too loose before...I may now be able to tweak them better as I think I have a better understanding of what the machine needs after going through it the last couple of weeks. I cannot give you any info on the new Y-car...as I don't really have any info on it...only a short conversation from Grant at shopbot...but from what I understand it will do away with the Z brackets for the hold downs and may be made of a single piece.

Erik

[pacoisa@cooptel.qc.ca]

THANK YOU VERY MUCH FOR SHARING INFORMATIONS WITH ME Erick!!!!!!! I APPRECIATE A L-O-T!!!!! REALY!!!! Hope you fix your "prob" so you could enjoy CNC more... 8-)
What about...someone else... suggestion about spring loading "carrs"; this way they'll fallow right and load some on a too light a "carr"...
But it seem your not the only one disliking this setup...seek; their is other threads about it...

Paco

What about my other question...about accuracy and low speed machining? C'mon "experienced" ShopBotters!!! 8-)

[srwtlc]

I too have been frustrated by the hold downs (old and new). I just setup a new machine for a High School in the area yesterday and spent too much time trying to get them adjusted. even the slightest amount of pressure would push/pull one or the other of the vertical rollers away from the track. I never did get it adjusted to my satisfaction. I told them to exercise the machine for a while and I'd come back and readjust it. The hold down situation does really need to be done another way. On my machine I'd like to have another rack and pinion on the other side. Not only to eliminate the need for the hold down bearings (because of spring tension on both sides then), but to stabilize that side too. One other nice way might be to have a rack on that side also, with a spring loaded pinion/idler. Maybe there's even a motion dampening bearing that could be used.

[stevem]

Try this http://www.talkshopbot.com/forum/messages/7/3031.html?1077239080

[gerald_d]

Steve, your method assumes that the two ground edges of each angle iron are perfectly parallel to each other. On our early model PRT, the second edge of the angle iron was not even ground.

Scott, you don't need a second rack & pinion solely to get hold down - you can run any old roller directly under the angle iron. Heck, I might even go so far as to fit one to our SB and put a photo here just to show how really simple it can be - but we actually like the fact that our y-car jumps off the rails when dive into the table by mistake, or hit a clamp, etc. It is like a safety valve now.

[stevem]

Gerald,

My method assumes nothing. It has the same amount of adjustability as the original brackets, but without the frustration. In addition, the two connecting square tubes stiffen the Y carriage to some degree.

[gerald_d]

Steve, the original brackets are more flexible, therefore they can flex to conform to a less than perfectly parallel rail. Plus they are height adjustable to ensure that they are pulling down and not pushing up. All very flexible, but can theoretically be set so that they have a "spring-loaded" pull down.

Your solution seems to be too stiff and unforgiving. ie. your hold down rollers might be doing the whole job of the main rollers, and your main rollers may be free-wheeling. Unless you have a height adjusting system . . . . . . in which case the set-up becomes more finicky than the flexible arms supplied at the moment.

Remember that I am saying all this as an armchair critic since we don't have any hold downs at all.

[erik_f]

More fun in Shopbot land...ok...the machine is put back together and running, what seems to be smoothly...the question is...if I measure one piece and it is .014" shorter than the next one...would that be +/-.007"? Or is it an impossible question to answer.

Erik F.

[gerald_d]

You would have to measure say 100 pieces and find the bell graph to see the distribution of results. The statisticians will drive you nuts on this one.

[bob buttons (Unregistered Guest)]

this is an easy one i think. your plus or minus is simply the amount you are off of the size you are trying for in your cad design. not the amount each piece is off from each other. hopefully the size you are shooting for is in the middle of your cut results. as Gerald said this one will drive you nuts because you are not getting what you want and cant zero in on it. too many factors effect each piece you cut esp the direction each piece is cut. if they are the same thats a whole different situation. if you are floating +/- .007 from your cad number that is pretty good.

[erik_f]

I'm -.017 on one piece from my CAD. I measured one of the 1/2" bits...they measure .501"...so I guess that would be .002" right there on an outside cut...which would bring the machine in to .015"...perfectionism is a real pain in the ass! Still want to do better than that...but...I guess I'll have to learn more tricks as I go.

Erik

[Paul Amos (Unregistered Guest)]

Eric

I ignore the actual bit size, what I tend to do is run a groove with the cutter then measure the width of the groove and use this... it also takes into account bearing runout


Paul

[erik_f]

I'll try it...thanks


Erik