http://www.ghielectronics.com/index.htm has some stuff with examples of interfacing PIC to MMC/SD cards with boards for about $50. Interesting. Still, getting the code to interface in with a gcode interpreter in one system may be difficult.
This means an arm of reliable length with one end located by one of these bearing can swing through a plane describing a pie section. I have in mind a thin wedge of pie. Pi comes in later. The general mechanism I see used to position the pie wedge in it's plane of movement is wire attached to the corners of the arc section. A spring keeps the wire tensioned (maybe with a small block and tackle arrangement so the small spring length changes absorbs a lot of wire movement.
1) Anchor point of spring.
2) Spring
3) Block and tackle
4) Wire to drum (drum driven by stepper, wire held to drum by friction of multiple turns).
5) Wire to corner of wedge.
6) Other corner of wedge has spring to remove backlash.
The upshot of that is that any point on the line drawn between the bearing at the point of the wedge and any place on the arc it's far end can swing through can be precisely located to within the tolerance of the bearing, given the line never changes length with humidy or temperature. Assume there are two such pie sections. There are center lines through the pie pieces these sections can swing through, the center lines go through the bearing and the center of the arc their motions describe.
Now arrange the wedges with their points on the diagonal corners of a square, and the center of the arcs through which their ends away from the bearings move, these two arcs cross with lines tangent to their arcs, these lines are perpendicular to each other. Within the area the wedges overlap, by positioning each wedge through its range of motion and knowing the point of each arc is within the bearing tolerance of one location, then a spindle on the top wedge (which can just be just a bar with the bearing at the far end and the spindle at the other) will draw over any point on the lower wedge that its motion lets it overlap.
To put it another way, since the bearings constrain every point on a line of constant length to be within the tolerance of the bearing of being in the same place, then positioning the swinging end of each with the swinging ends overlapping give you very precise positioning of a point where the two overlap. The tolerance of the bearing is 1/20000 of an inch, and the main downside is how far apart the two bearing have to be keep the angular range of motion convenient.
I am verry focussed on building robots, as this is my greatest hobby. I have started with Lego robots, then got a Boe-Bot, then moved on to building robot controllers myself, using PICs and recently AVRs. I am preparing a robotic kit somewhat·like Boe-Bot, under $100, that I will try to promote and sell in my home country, Romania.
In order to produce chasis parts, wheels and other styff that I need, I want to use a CNC mill or router, at least until I can afford having them laser cut, and for prototyping. Also, I need to make prototyping PCBs.
I have started yesterday morning to read this thread and took me almost 2 full days to read everythig, including G-codes tutorial and take a peak at Basic Stamp sample code you guys are working on. I have to say that I have learned a lot about what making a CNC involves. I realy like and appreciate the effort you all are making to get this done. I hope I can get one before I have to return to Romania, mid February. I do not dispose of much money, so having something that works well, without spending a fortune, will be of most importance. I need at least the electronics, software, lead screws, rails and step motors. Once home, I can build the table and·frame, using the parts and pictures (eventualy plans).
My questions are: do you guys think this project could be completed before mid February? If so, about how much money should I set aside to buy a kit? $500 is still in the range, or it is underestimated?
I will continue to observe your effort, maybe contribute where I feel I can say something of value. While reading the thread, I had a few suggestions to propose, but allways someone did·it before I·got to·the end of the thread. Keep up the good work, I am sure you'll get it done!
My questions are: do you guys think this project could be completed before mid February?
That sounds a bit optimistic.
Gabriel said...
If so, about how much money should I set aside to buy a kit? $500 is still in the range, or it is underestimated?
Gabriel
$500.00 is still the target as far as I know.
It all depends on how well inexpensive material performs.
It may not be up to the task, taking the price upwards.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - PLJack - - -
Perfection in design is not achieved when there is nothing left to add. It is achieved when there is nothing left to take away.
Well, things have progressed somewhat, but subsequently are again at a snail's pace. I had some reasonable results in the last recirculating ball slide bearings; a little noisy which is not a good sign as it would indicate premature wear, and more work is required. At present they are the most important precision-issue part of the whole thing.
As for the scedule; from my perspective I think February is premature. There still is a lot of prototyping to do; but we are making slow progress. I personally have need for a number of these units, modified to act as robot arms for our plastics production machines, so I'm a long way from quitting. Probably many others have become disenchanted. Hopefully their faith will be restored when they see the end results.
As far as price for the mechanics only is concerned, so far I see no change in what was earlier anticipated which as I recall was some $100 base plus $50 to $100 per axis, depending on size.
This link was just posted to the CHIBOTS.org message list: http://www.pdjinc.com/ somebody else's homebrew destop CNC. Looks like a nice implementation for woodworking. At least worth looking at for comparison.
For several years now I have been thinking of scratch building a clone of the 48" x 96" Shopbot. see: http://www.shopbottools.com If I can only find room to put it...
This is the description:
This Desk Master 1"· X-Y-Z travel capacity of 14 x 21 x 4 CNC router is in new condition and small enought to fit on most desk tops measuring 27" by 21" inches. This machine can move at 220· in/min in all three axis and has .0005 inch repeatablity. The screws are stainless steel and loaded self adjusting nylon bearings.· Power is by 3 220 oz. Nema 23 stepper motors are coupled tot he screwds using using 1 to 3 timing gears and belts. Tehe bearings are all self adjusting and and the whole unit is made out of solid aluminum. This unit comes with a Porter Cable router and standard electronics package ready to hook up to your PC (PC and ssoftware not included).
Here's the picture:
He supplies no software. I have emailed a request for suggested CNC software to him.
This presently has 7 bids, up to $350, with 6 days + to go.
So, is this what you have in mind for the community supported CNC milling machine? It looks pretty solid. He claims accuracy of 0.0005"! That's pretty good, isn't it? How would you secure your work piece to something like this?
If you're in the market, and that's the real price, I say go buy it; it looks well made. There is some similarity to the design I'm pursuing, but I suspect he is supporting the gantry on the lead screws; something I believe that will NOT permit his claimed 0.0005 unless they are very large in diameter; more like .005 I'd guess. That said though, I think it's a bargain if it fits your needs.
Sorry to have ignored your post for so long. As I was reading it, things got a little convoluted, and I didn't spend the time to study it in detail.
What you are suggesting certainly is a little unorthodocs, however that itself should not scare anyone, but it certainly is not in line with the path I'm on.
Could you make a drawing or two so we all may have a better understanding, so for those where this will be a fit will be able to make a better assessment.
Basically I think this is a 2 dimensional concept, probably suitable for PC routing, but not light 3 D milling.
I will attempt to embed below a quick bitmap image made from a *.dxf file. It show the rough shape of the intersecting "pie wedge arcs", the bearings are at the points of the wedges. The black shape shows the area swept by a spindle center located at the left end of the upper wedge as the it steps through its full range of motion, then the lower wedge steps once, and the the upper wedge steps back through its range, the lower steps once again, and so on until the full range of motion of each is exhausted.
For milling PCBs of course, a rectangular shape fitting withing that odd shape is all that is useful of the working area. You can see why accuracy is easily had with this approach, but at the expense compactness (nonexistent) and the complex math to make straight lines where the natural range of motion is all arcs. On the other hand, ABEC9P bearings have .5 mil runout, and that accuracy's not bad.
Okay, please try again. If Yahoo uses a dynamic link generation scheme, this one still won't work, so I've included it here as an attachment as well. Thank you for your patience. Tom P. ml, msl, & pfpp
In other words, the intersection of two swinging arms carries the tool-tip, and the controller then uses polar coordinates to control the arms in order to move the tool-tip·in a cartesian coordnate workspace?
How would this be much different from a SCARA stype robot arm, or from a 2-axis "hexapod" (neglecting that it would no longer then be a hexapod [noparse][[/noparse]google for "Stewart platform", or "Stewart-Gough platform"])?
Are you talking of one arm supporting the workpiece (say, the PCB) and the other arm supporting the tool-tip (perhaps the router)?· Then using polar coordinates to implement a cartesian workspace?
I agree with Peter (pjv).· The .0005·seems unrealistic but it looks like a nice machine.
In answer to your question about holding down the work, a friend of mine has a Haas 5' x 10' gantry CNC mill and the whole table is perforated with a plenum below (like a air hockey table).· In between each perforation is a groove so the whole table is a grid of grooves with·the perforations·at the intersections.· The grooves are about 1/8" wide by about 3/16" deep.· To hold down the work, he places foam rubber strips (1/8" x 1/4" cross section) in grooves enclosing an area that lie under the edge of the work piece. He·then positions the work piece on the foam strips and pulls a vacuum on the plenum.· The foam compresses and the work is sucked down flat to the table.· It holds amazingly well and there are no clamps to interfere with the spindle or gantry.
It would take·a bit·of effort to create a vacuum table but it would be ideal for PCBs.
The coordinates are angles only. The PCB is held to one wedge, and the spindle is on the other. As the length of the arms increases relative to the size of the workspace, then moving that arm produces an angular movement that is more purely in one cartesian axis as the the arm length increases. Angular movement of the wedge pointing down produces movement of the PCB in primarily the X axis, and movement of the other moves the spindle primarily in the Y axis. It think it would take trig done in IEEE format floats to not have innacuracies in the math show up in the work, so doing it all in an SX is out of the question. It could be that a table could hold corrections as well. The idea would be to have PC software keep a buffer in an SX filled enough (and a direction flag set) that the SX would just read how many clock ticks 'til it next pulsed a stepper, with enough code to slow things down as buffer under runs and halts approached. As long as the wedges are prevented from moving up and down away from each other (they have to move in their planes) and the materials of construction don't allow the length of the arms to change (with temp and humidity) then within those limits, plus the spindle runout, the machine should be as accurate as the bearings which is 0.5 mils.
Starman, I was thinking mostly of milling printed circuit traces·with this, and routing out the finished boards. However, things I have read so far suggest that the lowest resolution for this is sort of work .001" or .002", if one is doing surface mount ICs with pin spacings less than 0.05". I see from various posts, here and else where, that several people successfully hold things down with carpet tape. Works for me.
Thomm, $7000?!!!!·It has now gone 2 days since I saw this, and it's still $350!! I wonder if he'll continue this auction. I didn't notice if this·had a·reserve price.
Oh well!
kenjj
PS Is there ANY way to shut down that ding dong menu of bouncing emoticons?! I vcan't concentarte!
Tom Perkins, a few questions. What would be the advantage of your design over a conventional 3 axis table? It looks to me that the working area is about 15% of the “footprint”. Would someone write gcode programs using xyz coordinates? Could you draw something with more detail?
The only advantages I see over a 2.5 Axis table (X and Y and Tool Raise/Lower) are that the bearings I am trying to find at reasonable (to me) prices are very accurate at positioning things radially, and since the native mechanical coordinates are angular--nothing is sliding linearly--then the machine's accuracy should be roughly as good as the bearings are. The bearings, if I am understanding their specs correctly, are good for about 1/20,000 of an inch. That's half of a tenth!
The main question seems to be:
1) In Line Skate bearings that are advertised as being ABEC 9 are available at only $4.00 per. But they are not spec'ed to be ABEC 9P, the particular grade I'm interested in. Below is the reply from a skate equipment dealer I got when I asked if they were ABEC 9P.
"Yes- ABEC-9 is ABEC-9 according to the "annual bearing engineers conference" no matter what the brand.
Granted - some brands of ABEC 9 are superior to others, but all are ABEC-9"
It's not that I think he's lying, I just don't get a good feel from this reply that he's knows what I'm asking.
2) The Torrington/Timken group makes ABEC 9P ultra-precision bearings, but they don't seem to let you spec radial play, the spec I'm especially interested in. The part number I've asked them about is: P-MMX-C-9300-K-CR-SUX and I've an open email to them describing what I need and asking for which part I should get.
3) The Pacamor group lets you specify radial play, but instead of implying (with 0.5 tenths radial runout spec), they let you spec a range of radial play, and an allowable range is 1 to 3 tenths. So Pacamor will let me spec precisely what I want to, but the accuracy might only be 1/3333 of an inch. Of course, 1/3333 of an inch will let you mill some nice PCB's, but that's not all I might want to do with it.
Is radial play and runout the same thing? I don't yet know how they are using these terms.
Bottom line, I'm trying to inexpensively get extreme accuracy in 2 axis, at the expense of more complicated pre-processing (of the G-Code in the PC), and a physically large machine. The large size is required because I do want a roughly rectangular work area*, and the sides of the work area that is "native" to the mechanics of the machine are more nearly square the larger the machine gets.
* If I wanted an annular work area, this machine setup would be ideal. If I added a stepped Z axis slide, I could presumably make blisks and blings and homebrew the turbine for a VTOL jumpjet ;^)
The Mill this thread is dedicated to is intended to be capable for other than PCB milling, and is intended to be $100.00 per axis plus $100.00 base. There seems to be some skepticism as to what accuracy can in fact be achieved.
With the cost of say 2 4'x4' pieces of MDF, surplus PC supply for the steppers, surplus stepper, and the bearings, I hope to get at worst 1/3333inch accuracy in 2 axis for more like $60.00 axis plus another $60.00 for incidentals. I suppose with enough spare junk lying around, the all up cost might be as low as the cost of the bearings. The likely accuracy would be about 1/8000 to 1/12000, and if I luck out and get really good bearings and play in other parts are minimal, I don't yet see why I couldn't get 1/15000. On thinking it through further, 1/20000 inches would only be possible if there was no other play anywhere else, which is not likely. On the plus side, the 1/1000 minimum for fine pitch PCB milling and drilling seems to be a sure bet.
Besides the bearings, places where other mechanical tolerances affect accuracy seem to be limited to the play in the spindle and the mechanism that moves it up/down into/out of the work. I don't know if MDF has an unacceptable expansion with heat (unlikely) or humidity (perfectly plausible it will, I don't know).
So that's where I'm at. Gathering info to see if I can trade a large size and mathematical complexity for high accuracy at low cost.
The workspace shape* and size I get with a 3'x3' table seems acceptable, and that lets me use a single 4'x4' sheet of MDF and keep the whole thing smaller. If I can remove the spindle and secure it to an unused part of the table, then the whole thing will go fairly unobtrusively behind a door in a corner.
*D'oh! Scale it linearly and of course the shape doesn't change.
The only advantages I see over a 2.5 Axis table (X and Y and Tool Raise/Lower) are that the bearings I am trying to find at reasonable (to me) prices are very accurate at positioning things radially, and since the native mechanical coordinates are angular--nothing is sliding linearly--then the machine's accuracy should be roughly as good as the bearings are. The bearings, if I am understanding their specs correctly, are good for about 1/20,000 of an inch. That's half of a tenth!
Thank you, Tom Perkins,
ml, msl, & pfpp
Isn't a 1/20,000 of an inch equal to 0.00005 of an inch??? Which is MUCH finer than .05 or 1/2 of a tenth????
Please don't mis-understand my post here, as I am NOT trying to pour cold water on your ideas, but rather point out some over simplifications you are making.
You seem to think that perfect bearings will yield a perfect result. This is not the case; there are other effects at play.
For example:
Even with a perfect bearing how will you fasten the bearing to some holder and ensure that you have "centered" the bearing perfectly to the middle of the race without offset? Any error in THAT translates directlly to an off-center error, and as you swing an arc, that leg will "grow" or "shrink" according to the amount of center offset and the angle swept.
You see, the bearing attaching method is probably some machined post with a thread on it, or a sleeve with a (screw) hole drilled through it. The centricity of this hole or the threaded post must be PERFECTLY machined to the race center, or you will get an offset error. Unless you have VERY good tools, you won't come anywhere near the precisions desired. In fact, I would bet that a hobbyist with typical tools would have difficuly drilling a hole to 10 one thousanths of an inch. On a lathe you could probably center bore one to one thou.
But you speak rather casually about one half of one tenth.....a truly remarkable precision that I believe you have NO HOPE in achieving without very advanced equipment. At our work, we have such tools, and even then we have difficulty (as in it takes a lot of time) to position bore holes to 3 tenths of a thou. You are attempting something 6 times that good and with what kind of tools?........
I think you are focusing and relying too much on the precision of the bearings being the only issue. Sure, its important, but not in relation to all the other things.
Have fun, and let us know how it works out.
Cheers from Maui, Hawaii,
Peter (pjv)
POST EDIT;
Hi Again Tom;
I just went for a stroll on the beach and was reflecting on what I had said regarding centricity of mounting the bearing, and I'm ALL WET; literally.
And please don't take this as a suggestion that I post information without due thought, but in this case I was absolutely wrong, and gave you bad advice; I apologize.
How you mount the bearing has no bearing (pun) on the arms stretching or shrinking. It is the run-out (eccentricity) of the bearing that causes it, and you are proposing near perfect bearings. So you may well have much better luck than I first thought.
That said, there still are other gremlins at play; stretch in the cables due to "run-up/down on the drive cylinder, temperature, etc., and you will need to assess the effect of each on the accuracy/repeatability of your design.
Mostly, be analytical, and don't "gloss-over" or trivialize small dimensions. A tenth of a thou is VERY small for us mere mortals.
Yep, the "runout" spec for ABEC 9P bearings is 1/2 of a tenth, as stated by the PACAMOR.com website, they also give a selectable range of "radial play" tolerances, which sounds more like what I'm looking for. The tightest standard range 1 to 3 tenths, which would at worst be 1/3333 of an inch*. Of course, the runout of a bearing is important for making the spindle, and the die grinder I've got, at 25krpm, will need to be stepped up to 100krpm for milling and drilling at the hole/track pitches I'm interested in.
* Of course if I can preload the system enough while still not stressing the stepper too much, radial play might go away as an issue, and then I'm back (I think) to the 1/20,000 inch runout as a controlling spec.
Tom Perkins
Your design is very interesting. I can see a few advantages if I have it right (not sure of the drive aspect). If the wide end of the wedges were trimmed to a radius and flexible rail were attached, it would become something like a gear segment. Then, a stepper motor with a spur gear could be mounted to drive the gear segment back and forth. Wouldn’t you have an axis movement with only one moving part? (I also was walking on the beach, but it’s 16 degrees here).
As for a more conventional mechanical design , using high quality bearings, rails, and drive components available, along with the ability to make parts to close tolerances , a machine with accuracy of .0005 could be made. It would be hard to do for $500. Most commercial grade “bench top” mills sold at around $3K that claim .0005 accuracy are hardly over engineered to justify the price. Needles to say ,to get this kind of accuracy , components are chosen for performance over cost. For the average guy, to make a machine out of surplus “e-bay treasures” and an “on the fly” design, just getting it to run can be a challenge . Because of vibration and abrupt start/stop movement ,using inexpensive materials will effect the repeatability after a short time . Even “very cheap” has it’s limits.
Yeah I know. If the $4.00 per skate bearings are really ABEC 9P, this is an idea sure worth trying. If bearings that meet the ABEC 9P standard are $400.00, then this idea is not worth pursuing--and even at $40.00, it loses much appeal.
To keep tension constant on the cable, it seems like maybe a continuous loop of cable tensioned by a spring mechanism is a much better idea than just attaching one corner of each arm to a spring, much less a block and tackle arrangement to minimize the inconstant forces of a lengthening spring. Perhaps a servomotor could read the extension of the tensioner, and a PI (Proportional Integral) loop driving the servo could keep the tension of the cable fairly constant from the perspective of the stepper, making the stepper less likely to loose steps.
I need a way to force the inner race against the outer race that cutting and inertia forces can't override, so that will be more springs.
On the plus (more plus, or is that plusser) side, the bearing has the spec'ed runout presumably over 180 degrees of travel, and I'm only using about 15 degrees of that travel...is it plausible that it would only see 1/12th of the runout?
That sounds like it could be a mere 1/240,000 of an inch! Damn! Breathing on it would screw it up more than that!
Yeah, I know that can't be right, I just haven't figured out the specifics of how its wrong yet ;^)
While I retracted my statement about center mounting the bearings, I would point out thet your "block and tackle" would presumably be a home made device, and it IS subject to the centricity I spoke of. Even if you supplied perfect bearings for those pulleys, you would have a bear of a time getting the bearings into the precise center of the pulleys, and the whole arrangement made perfectly round. Any offset in centricity would translate directly into non-linear displacement. So as I tried to indicate before, getting the utmost out of the precision of the bearings (now you are speaking of a rather rediculous two hundred and fortieth of a thou) is not the issue; there are so many other things that are thousands of times worse than those miniscule numbers that you should be focused on.
Again I wish you luck, but I believe you should stop trying to improve the bearing issue (it's been over-done), and deal with the balance of your design.
In the post above, I already ditched both the spring and sprung block and tackle, and have decided to to go with a tension loop of cable. Also, referring to the 1/240000 issue, I wrote:
"Yeah, I know that can't be right, I just haven't figured out the specifics of how its wrong yet ;^)"
Comments
I have been lurking on this thread for weeks (it took a while just to read it all ;^) ).
As far as PCB milling and drilling goes, I've had a thought I'd like to bounce off of you.
You want to be very accurate and inexpensive--go large.
Here are the tolerances for ABEC 9 bearings ($4.00 per inline skate bearings).
www.pacamor.com/abectolerances.php
This means an arm of reliable length with one end located by one of these bearing can swing through a plane describing a pie section. I have in mind a thin wedge of pie. Pi comes in later. The general mechanism I see used to position the pie wedge in it's plane of movement is wire attached to the corners of the arc section. A spring keeps the wire tensioned (maybe with a small block and tackle arrangement so the small spring length changes absorbs a lot of wire movement.
1) Anchor point of spring.
2) Spring
3) Block and tackle
4) Wire to drum (drum driven by stepper, wire held to drum by friction of multiple turns).
5) Wire to corner of wedge.
6) Other corner of wedge has spring to remove backlash.
The upshot of that is that any point on the line drawn between the bearing at the point of the wedge and any place on the arc it's far end can swing through can be precisely located to within the tolerance of the bearing, given the line never changes length with humidy or temperature. Assume there are two such pie sections. There are center lines through the pie pieces these sections can swing through, the center lines go through the bearing and the center of the arc their motions describe.
Now arrange the wedges with their points on the diagonal corners of a square, and the center of the arcs through which their ends away from the bearings move, these two arcs cross with lines tangent to their arcs, these lines are perpendicular to each other. Within the area the wedges overlap, by positioning each wedge through its range of motion and knowing the point of each arc is within the bearing tolerance of one location, then a spindle on the top wedge (which can just be just a bar with the bearing at the far end and the spindle at the other) will draw over any point on the lower wedge that its motion lets it overlap.
To put it another way, since the bearings constrain every point on a line of constant length to be within the tolerance of the bearing of being in the same place, then positioning the swinging end of each with the swinging ends overlapping give you very precise positioning of a point where the two overlap. The tolerance of the bearing is 1/20000 of an inch, and the main downside is how far apart the two bearing have to be keep the angular range of motion convenient.
Err...
Whad'ya'think?
Thanks, Tom Perkins
ml, msl, & pfpp
I am verry focussed on building robots, as this is my greatest hobby. I have started with Lego robots, then got a Boe-Bot, then moved on to building robot controllers myself, using PICs and recently AVRs. I am preparing a robotic kit somewhat·like Boe-Bot, under $100, that I will try to promote and sell in my home country, Romania.
In order to produce chasis parts, wheels and other styff that I need, I want to use a CNC mill or router, at least until I can afford having them laser cut, and for prototyping. Also, I need to make prototyping PCBs.
I have started yesterday morning to read this thread and took me almost 2 full days to read everythig, including G-codes tutorial and take a peak at Basic Stamp sample code you guys are working on. I have to say that I have learned a lot about what making a CNC involves. I realy like and appreciate the effort you all are making to get this done. I hope I can get one before I have to return to Romania, mid February. I do not dispose of much money, so having something that works well, without spending a fortune, will be of most importance. I need at least the electronics, software, lead screws, rails and step motors. Once home, I can build the table and·frame, using the parts and pictures (eventualy plans).
My questions are: do you guys think this project could be completed before mid February? If so, about how much money should I set aside to buy a kit? $500 is still in the range, or it is underestimated?
I will continue to observe your effort, maybe contribute where I feel I can say something of value. While reading the thread, I had a few suggestions to propose, but allways someone did·it before I·got to·the end of the thread. Keep up the good work, I am sure you'll get it done!
Gabriel
That sounds a bit optimistic.
$500.00 is still the target as far as I know.
It all depends on how well inexpensive material performs.
It may not be up to the task, taking the price upwards.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - PLJack - - -
Perfection in design is not achieved when there is nothing left to add.
It is achieved when there is nothing left to take away.
Well, things have progressed somewhat, but subsequently are again at a snail's pace. I had some reasonable results in the last recirculating ball slide bearings; a little noisy which is not a good sign as it would indicate premature wear, and more work is required. At present they are the most important precision-issue part of the whole thing.
As for the scedule; from my perspective I think February is premature. There still is a lot of prototyping to do; but we are making slow progress. I personally have need for a number of these units, modified to act as robot arms for our plastics production machines, so I'm a long way from quitting. Probably many others have become disenchanted. Hopefully their faith will be restored when they see the end results.
As far as price for the mechanics only is concerned, so far I see no change in what was earlier anticipated which as I recall was some $100 base plus $50 to $100 per axis, depending on size.
Please keep the faith......
Cheers,
Peter (pjv)
This link was just posted to the CHIBOTS.org message list: http://www.pdjinc.com/ somebody else's homebrew destop CNC. Looks like a nice implementation for woodworking. At least worth looking at for comparison.
For several years now I have been thinking of scratch building a clone of the 48" x 96" Shopbot. see: http://www.shopbottools.com If I can only find room to put it...
Of course then we all need a CNC plasma torch too: http://www.plasmacam.com/
Sorry, have to go. I'm going to be late for my toolaholics annonymous meeting...
Rick B.
http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=6018508186&ssPageName=ADME:B:SS:US:1
This is the description:
This Desk Master 1"· X-Y-Z travel capacity of 14 x 21 x 4 CNC router is in new condition and small enought to fit on most desk tops measuring 27" by 21" inches. This machine can move at 220· in/min in all three axis and has .0005 inch repeatablity. The screws are stainless steel and loaded self adjusting nylon bearings.· Power is by 3 220 oz. Nema 23 stepper motors are coupled tot he screwds using using 1 to 3 timing gears and belts. Tehe bearings are all self adjusting and and the whole unit is made out of solid aluminum. This unit comes with a Porter Cable router and standard electronics package ready to hook up to your PC (PC and ssoftware not included).
Here's the picture:
He supplies no software. I have emailed a request for suggested CNC software to him.
This presently has 7 bids, up to $350, with 6 days + to go.
So, is this what you have in mind for the community supported CNC milling machine? It looks pretty solid. He claims accuracy of 0.0005"! That's pretty good, isn't it? How would you secure your work piece to something like this?
Happy holidays!
kenjj
If you're in the market, and that's the real price, I say go buy it; it looks well made. There is some similarity to the design I'm pursuing, but I suspect he is supporting the gantry on the lead screws; something I believe that will NOT permit his claimed 0.0005 unless they are very large in diameter; more like .005 I'd guess. That said though, I think it's a bargain if it fits your needs.
Cheers from Maui, Hawaii,
Peter (pjv)
Sorry to have ignored your post for so long. As I was reading it, things got a little convoluted, and I didn't spend the time to study it in detail.
What you are suggesting certainly is a little unorthodocs, however that itself should not scare anyone, but it certainly is not in line with the path I'm on.
Could you make a drawing or two so we all may have a better understanding, so for those where this will be a fit will be able to make a better assessment.
Basically I think this is a 2 dimensional concept, probably suitable for PC routing, but not light 3 D milling.
Cheers from Maui, Hawaii,
Peter (pjv)
I will attempt to embed below a quick bitmap image made from a *.dxf file. It show the rough shape of the intersecting "pie wedge arcs", the bearings are at the points of the wedges. The black shape shows the area swept by a spindle center located at the left end of the upper wedge as the it steps through its full range of motion, then the lower wedge steps once, and the the upper wedge steps back through its range, the lower steps once again, and so on until the full range of motion of each is exhausted.
For milling PCBs of course, a rectangular shape fitting withing that odd shape is all that is useful of the working area. You can see why accuracy is easily had with this approach, but at the expense compactness (nonexistent) and the complex math to make straight lines where the natural range of motion is all arcs. On the other hand, ABEC9P bearings have .5 mil runout, and that accuracy's not bad.
Thank you, Tom Perkins,
ml, msl, & pfpp
Post Edited (tperkins) : 12/2/2005 4:17:14 PM GMT
I get a no-show on the bit-map.
Cheers,
Peter (pjv)
Same here. I think your link to your Yahoo photo album might be broken.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
How would this be much different from a SCARA stype robot arm, or from a 2-axis "hexapod" (neglecting that it would no longer then be a hexapod [noparse][[/noparse]google for "Stewart platform", or "Stewart-Gough platform"])?
Daniel
I just reread your post.
Are you talking of one arm supporting the workpiece (say, the PCB) and the other arm supporting the tool-tip (perhaps the router)?· Then using polar coordinates to implement a cartesian workspace?
Daniel
I agree with Peter (pjv).· The .0005·seems unrealistic but it looks like a nice machine.
In answer to your question about holding down the work, a friend of mine has a Haas 5' x 10' gantry CNC mill and the whole table is perforated with a plenum below (like a air hockey table).· In between each perforation is a groove so the whole table is a grid of grooves with·the perforations·at the intersections.· The grooves are about 1/8" wide by about 3/16" deep.· To hold down the work, he places foam rubber strips (1/8" x 1/4" cross section) in grooves enclosing an area that lie under the edge of the work piece. He·then positions the work piece on the foam strips and pulls a vacuum on the plenum.· The foam compresses and the work is sucked down flat to the table.· It holds amazingly well and there are no clamps to interfere with the spindle or gantry.
It would take·a bit·of effort to create a vacuum table but it would be ideal for PCBs.
Chris I.
The coordinates are angles only. The PCB is held to one wedge, and the spindle is on the other. As the length of the arms increases relative to the size of the workspace, then moving that arm produces an angular movement that is more purely in one cartesian axis as the the arm length increases. Angular movement of the wedge pointing down produces movement of the PCB in primarily the X axis, and movement of the other moves the spindle primarily in the Y axis. It think it would take trig done in IEEE format floats to not have innacuracies in the math show up in the work, so doing it all in an SX is out of the question. It could be that a table could hold corrections as well. The idea would be to have PC software keep a buffer in an SX filled enough (and a direction flag set) that the SX would just read how many clock ticks 'til it next pulsed a stepper, with enough code to slow things down as buffer under runs and halts approached. As long as the wedges are prevented from moving up and down away from each other (they have to move in their planes) and the materials of construction don't allow the length of the arms to change (with temp and humidity) then within those limits, plus the spindle runout, the machine should be as accurate as the bearings which is 0.5 mils.
Thanks, Tom Perkins,
ml, msl, & pfpp
Thomm, $7000?!!!!·It has now gone 2 days since I saw this, and it's still $350!! I wonder if he'll continue this auction. I didn't notice if this·had a·reserve price.
Oh well!
kenjj
PS Is there ANY way to shut down that ding dong menu of bouncing emoticons?! I vcan't concentarte!
The only advantages I see over a 2.5 Axis table (X and Y and Tool Raise/Lower) are that the bearings I am trying to find at reasonable (to me) prices are very accurate at positioning things radially, and since the native mechanical coordinates are angular--nothing is sliding linearly--then the machine's accuracy should be roughly as good as the bearings are. The bearings, if I am understanding their specs correctly, are good for about 1/20,000 of an inch. That's half of a tenth!
The main question seems to be:
1) In Line Skate bearings that are advertised as being ABEC 9 are available at only $4.00 per. But they are not spec'ed to be ABEC 9P, the particular grade I'm interested in. Below is the reply from a skate equipment dealer I got when I asked if they were ABEC 9P.
"Yes- ABEC-9 is ABEC-9 according to the "annual bearing engineers conference" no matter what the brand.
Granted - some brands of ABEC 9 are superior to others, but all are ABEC-9"
It's not that I think he's lying, I just don't get a good feel from this reply that he's knows what I'm asking.
2) The Torrington/Timken group makes ABEC 9P ultra-precision bearings, but they don't seem to let you spec radial play, the spec I'm especially interested in. The part number I've asked them about is: P-MMX-C-9300-K-CR-SUX and I've an open email to them describing what I need and asking for which part I should get.
3) The Pacamor group lets you specify radial play, but instead of implying (with 0.5 tenths radial runout spec), they let you spec a range of radial play, and an allowable range is 1 to 3 tenths. So Pacamor will let me spec precisely what I want to, but the accuracy might only be 1/3333 of an inch. Of course, 1/3333 of an inch will let you mill some nice PCB's, but that's not all I might want to do with it.
Is radial play and runout the same thing? I don't yet know how they are using these terms.
Bottom line, I'm trying to inexpensively get extreme accuracy in 2 axis, at the expense of more complicated pre-processing (of the G-Code in the PC), and a physically large machine. The large size is required because I do want a roughly rectangular work area*, and the sides of the work area that is "native" to the mechanics of the machine are more nearly square the larger the machine gets.
* If I wanted an annular work area, this machine setup would be ideal. If I added a stepped Z axis slide, I could presumably make blisks and blings and homebrew the turbine for a VTOL jumpjet ;^)
The Mill this thread is dedicated to is intended to be capable for other than PCB milling, and is intended to be $100.00 per axis plus $100.00 base. There seems to be some skepticism as to what accuracy can in fact be achieved.
With the cost of say 2 4'x4' pieces of MDF, surplus PC supply for the steppers, surplus stepper, and the bearings, I hope to get at worst 1/3333inch accuracy in 2 axis for more like $60.00 axis plus another $60.00 for incidentals. I suppose with enough spare junk lying around, the all up cost might be as low as the cost of the bearings. The likely accuracy would be about 1/8000 to 1/12000, and if I luck out and get really good bearings and play in other parts are minimal, I don't yet see why I couldn't get 1/15000. On thinking it through further, 1/20000 inches would only be possible if there was no other play anywhere else, which is not likely. On the plus side, the 1/1000 minimum for fine pitch PCB milling and drilling seems to be a sure bet.
Besides the bearings, places where other mechanical tolerances affect accuracy seem to be limited to the play in the spindle and the mechanism that moves it up/down into/out of the work. I don't know if MDF has an unacceptable expansion with heat (unlikely) or humidity (perfectly plausible it will, I don't know).
So that's where I'm at. Gathering info to see if I can trade a large size and mathematical complexity for high accuracy at low cost.
The workspace shape* and size I get with a 3'x3' table seems acceptable, and that lets me use a single 4'x4' sheet of MDF and keep the whole thing smaller. If I can remove the spindle and secure it to an unused part of the table, then the whole thing will go fairly unobtrusively behind a door in a corner.
*D'oh! Scale it linearly and of course the shape doesn't change.
Thank you for your interest.
I will keep posting any progress in this thread.
Thank you, Tom Perkins,
ml, msl, & pfpp
Post Edited (tperkins) : 12/3/2005 4:40:55 PM GMT
Isn't a 1/20,000 of an inch equal to 0.00005 of an inch??? Which is MUCH finer than .05 or 1/2 of a tenth????
WOW
Bob N9LVU
Please don't mis-understand my post here, as I am NOT trying to pour cold water on your ideas, but rather point out some over simplifications you are making.
You seem to think that perfect bearings will yield a perfect result. This is not the case; there are other effects at play.
For example:
Even with a perfect bearing how will you fasten the bearing to some holder and ensure that you have "centered" the bearing perfectly to the middle of the race without offset? Any error in THAT translates directlly to an off-center error, and as you swing an arc, that leg will "grow" or "shrink" according to the amount of center offset and the angle swept.
You see, the bearing attaching method is probably some machined post with a thread on it, or a sleeve with a (screw) hole drilled through it. The centricity of this hole or the threaded post must be PERFECTLY machined to the race center, or you will get an offset error. Unless you have VERY good tools, you won't come anywhere near the precisions desired. In fact, I would bet that a hobbyist with typical tools would have difficuly drilling a hole to 10 one thousanths of an inch. On a lathe you could probably center bore one to one thou.
But you speak rather casually about one half of one tenth.....a truly remarkable precision that I believe you have NO HOPE in achieving without very advanced equipment. At our work, we have such tools, and even then we have difficulty (as in it takes a lot of time) to position bore holes to 3 tenths of a thou. You are attempting something 6 times that good and with what kind of tools?........
I think you are focusing and relying too much on the precision of the bearings being the only issue. Sure, its important, but not in relation to all the other things.
Have fun, and let us know how it works out.
Cheers from Maui, Hawaii,
Peter (pjv)
POST EDIT;
Hi Again Tom;
I just went for a stroll on the beach and was reflecting on what I had said regarding centricity of mounting the bearing, and I'm ALL WET; literally.
And please don't take this as a suggestion that I post information without due thought, but in this case I was absolutely wrong, and gave you bad advice; I apologize.
How you mount the bearing has no bearing (pun) on the arms stretching or shrinking. It is the run-out (eccentricity) of the bearing that causes it, and you are proposing near perfect bearings. So you may well have much better luck than I first thought.
That said, there still are other gremlins at play; stretch in the cables due to "run-up/down on the drive cylinder, temperature, etc., and you will need to assess the effect of each on the accuracy/repeatability of your design.
Mostly, be analytical, and don't "gloss-over" or trivialize small dimensions. A tenth of a thou is VERY small for us mere mortals.
Good luck from Maui, Hawaii,
Peter (pjv)
Post Edited (pjv) : 12/3/2005 8:24:54 PM GMT
Yep, the "runout" spec for ABEC 9P bearings is 1/2 of a tenth, as stated by the PACAMOR.com website, they also give a selectable range of "radial play" tolerances, which sounds more like what I'm looking for. The tightest standard range 1 to 3 tenths, which would at worst be 1/3333 of an inch*. Of course, the runout of a bearing is important for making the spindle, and the die grinder I've got, at 25krpm, will need to be stepped up to 100krpm for milling and drilling at the hole/track pitches I'm interested in.
* Of course if I can preload the system enough while still not stressing the stepper too much, radial play might go away as an issue, and then I'm back (I think) to the 1/20,000 inch runout as a controlling spec.
Is anyone here an expert on ball bearings?
Thanks, Tom Perkins
Your design is very interesting. I can see a few advantages if I have it right (not sure of the drive aspect). If the wide end of the wedges were trimmed to a radius and flexible rail were attached, it would become something like a gear segment. Then, a stepper motor with a spur gear could be mounted to drive the gear segment back and forth. Wouldn’t you have an axis movement with only one moving part? (I also was walking on the beach, but it’s 16 degrees here).
"and you are proposing near perfect bearings."
Yeah I know. If the $4.00 per skate bearings are really ABEC 9P, this is an idea sure worth trying. If bearings that meet the ABEC 9P standard are $400.00, then this idea is not worth pursuing--and even at $40.00, it loses much appeal.
To keep tension constant on the cable, it seems like maybe a continuous loop of cable tensioned by a spring mechanism is a much better idea than just attaching one corner of each arm to a spring, much less a block and tackle arrangement to minimize the inconstant forces of a lengthening spring. Perhaps a servomotor could read the extension of the tensioner, and a PI (Proportional Integral) loop driving the servo could keep the tension of the cable fairly constant from the perspective of the stepper, making the stepper less likely to loose steps.
I need a way to force the inner race against the outer race that cutting and inertia forces can't override, so that will be more springs.
On the plus (more plus, or is that plusser) side, the bearing has the spec'ed runout presumably over 180 degrees of travel, and I'm only using about 15 degrees of that travel...is it plausible that it would only see 1/12th of the runout?
That sounds like it could be a mere 1/240,000 of an inch! Damn! Breathing on it would screw it up more than that!
Yeah, I know that can't be right, I just haven't figured out the specifics of how its wrong yet ;^)
Thank you all for your interest and comments.
Yours, Tom Perkins,
ml, msl, & pfpp
While I retracted my statement about center mounting the bearings, I would point out thet your "block and tackle" would presumably be a home made device, and it IS subject to the centricity I spoke of. Even if you supplied perfect bearings for those pulleys, you would have a bear of a time getting the bearings into the precise center of the pulleys, and the whole arrangement made perfectly round. Any offset in centricity would translate directly into non-linear displacement. So as I tried to indicate before, getting the utmost out of the precision of the bearings (now you are speaking of a rather rediculous two hundred and fortieth of a thou) is not the issue; there are so many other things that are thousands of times worse than those miniscule numbers that you should be focused on.
Again I wish you luck, but I believe you should stop trying to improve the bearing issue (it's been over-done), and deal with the balance of your design.
Cheers from Maui, Hawaii,
Peter (pjv)
In the post above, I already ditched both the spring and sprung block and tackle, and have decided to to go with a tension loop of cable. Also, referring to the 1/240000 issue, I wrote:
"Yeah, I know that can't be right, I just haven't figured out the specifics of how its wrong yet ;^)"
I was poking fun at myself.
Yours, Tom Perkins, ml, msl, & pfpp