Several posts above, I made reference to exposing a printed cicuit board with a flexible optical fiber.
Years ago, when flat-bed plotters were all the rage, I actually did this; using the pen-up/pen-down solenoid signal to turn an LED off and on. I used the actual PC layout plotting software (Protel) to do the drawing. I selected the "pen width" close to the diameter of the fiber, and got some surprisingly good results. With today's brighter and white LEDs, one can probably do a better and faster job.
Although it is was a little slow, what was neat about this approach, you can get (close to) photo results without involving a photoshop; just software. I absolutely LOVE "soft-tools".
Presently I am also working on another (Laser) based approach that will, if it pans out, reduce the "exposing" time to 10 seconds or less. With 1 mil resolution, a bit costly to manufacture though, but the processing cost should be minimal.
In case anyone is interested, i came across this site that has a nice design for a cnc router/ milling machine. www.cheapcnc.com/cs12-24.html
The idea of using the high density plastic for a frame is a good idea, easy to machine and to work with. As you can see by the construction, it would not be hard to fabricate, and makes for a simple design. If anything, you can get a good idea of the gantry design by the pictures.
And the porter/cable spindle looks like a good pick too, up to 30,000 rpm with a collet holder.
Sorry about the linear vs. chopper confusion. I've been dealign with a similar issue here at work and thought (erroniously) that you meant linear. I really need to disconnect from work more before I post here before I cause real trouble
The laser and fiber optic exposure both sound really interesting. I remeber putting down etch-resist with an HP flatbed plotter some time ago with excellent results. A small UV laser should work wonders for the presensatized boards at small outputs (1-5mW?).
They're manufacturer datasheets, but include the frame sizes you asked for. They're standardized just about everywhere, but these PDFs should make it nice and easy for you:
I know the whole issue of your desired accuracy versus everyone else's is pretty much a done deal, but I have a couple of questions.
1 - In an earlier post, You mentioned building "a 14" square base that was flat to no less than 0.001". Could you give a few more details? What material was used? How thick was the material? How were the pieces fastened together? Since you mentioned using a miter saw, I assume this was open like a picture frame as opposed to a flat, solid plate. How did you measure the flatness of this item?
2 - You talk about accuracy of .0005" (which is 1/2 of 1/1000 of an inch) as if this is a piece of cake to achieve. I do milling on a Bridgeport with a DRO (Digital Readout) and just setting the locks on an axis is often enough to move an axis .0005". I know that the axis has moved because the DRO is good to .0005" and I can watch the number jump. If a machine that weighs over 2000 pounds can so easily move .0005", how do you expect a little tabletop device that weighs under 100 pounds to do better?
I built the machine base from 4" square aluminium extrusions similar to the ones from 8020. It It was a simple, butt-jointed square (like a cheap picture frame)designed for an elevated optical bench apparatus. The pieces were fastened with some glorified right angle brackets I had machined, and tweaked regularly for the better part of a day to get it "just right". The flatness was verified on the granite surface plate at the customer's location (I coldn't get a 0.001" feeler anywhere under it, on either side, btu there was still a small light leak when we looked very carefully), but I checked it while I was building it with an 18" level, a feeler guage, and a24" square 4-layer stackup of 3/4" MDF that was as flat as anything else in my garage. The reason I said it took a weekend was that I spent considerable time aligning my miter saw, tweaking for maximally flat cuts, etc. If I had the pieces precut with square flat edges, I think I could have done it in a few hours total, knowing what I know now.
As for your Bridgeport, I have no doubt that the ways allowed 0.0005" of slop when you tightened the locks. The one ton of weight of the machine keeps it stable and solid when moving it's 100lb spindle assembly around, machining large, hard parts. That's the way it's designed. You're talking about big, heavy moving parts. There's less precision in the movements of thos big clock movements on the towers, and it's not just because one was made in Switzerland. Smaller, lighter components allow finer tolerances when they are moved.
What I'm talking about is a much lighter weight machine, for machining much smaller items from much softer material. Specifically, for the PCB drilling aspect of things, it should be pretty darn easy to keep 0.001" of calibrated accuracy, precision AND repeatibility, and probably 0.0005" if folks take good care in building and maintaing their machines. The sheer issue of scale makes these goals easier to achieve. We're moving a whole lot less mass. We're using a gantry, which has no moving ways or gibs under the workpiece, simplifyign the relative motions. Slop is easily removed from the gantry's motions with ballscrews, recirculating ball nuts, and precision linear bearings (to 0.0005" or better) tho it will cost more money to get that accuracy than with hardware store threaded rod and drawer glides (like a lot of the wood CNC guys use). This won't be a bash-around router for the garage. It will be a piece of fine machinery that likes to sit inside on a benchtop. We too will need a solid base, but because of the moving masses involved, we won't need 2000, or maybe even 200lbs. I think for the goal of making PCBs, the weight of the machine itself will do just fine.
The way that Peter (pjv) is putting this together, I feel strongly that people will have access to not only high precision motion devices, but accurately machined structural components that can easily be assembled to construct a small, precise, lightweight machine capable of sub-mil accuracy. I don't think everyone will achieve that. It will take time, planning, tools, patience, and more patience, but I think it is posible.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
I admire your tenacity in anticipating sub-thou performance. I too hope that that can be achieved, but I continue to believe we should be happy with 2 to 3 thou, and elated with 1 thou. I can only dream of better than that.......I expect temperature variations alone will throw us off some tenths.
Nevertheless, onward !!!!!!!!
Some money making activities got in the way of the design process, but I expect to be back at it early in the week. Please remember this is a volunteer activity, and needs to be accomodated by volunteer hours.
I think we can all relate to the inherantly second-class nature of hobby projects to those providing important things like pizza and mortgagae payments.
I'm tenacious because I know others have done it with a less organized effort. If some hamfisted guy with minimal mechanical engneering exprience can buy things on eBay and assemble a machine that can hit one mil all the time (one guy at work did, and it worked great after we tweaked it for him), with the combined efforts of the real engineers and experienced enthusaists here, it should be a lot easier.
Of course, if we fail misterably, I'll be the first one to eat crow, and I'll like it
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
I guess my confusion arose from the fact that you intend to build a completely different machine than basically everyone else. The thread got started when someone wanted to build a cheap machine for milling and drilling PCBs. In this case, cheap was $500 or less. As soon as you start talking about lead screws and ball nuts driving machinery on precision hardened rails, you're falling off the $500 train pretty fast.
I agree that with the parts you've spec'ed, it's possible to achieve the tolerances you talk about. I disagree that it is easy to build a machine that will hold these tolerances while machining something else. While it might be possible to achieve these tolerances in terms of simply driving the rig around, the minute you add something like a spindle and a cutting bit which have to cut through a layer of copper, then everything just got a lot harder.
As you know, the reason I brought up the weight issue is that when cutting metal (and copper cladding definitely counts as metal), mass is one of the most important attributes needed to achieve accuracy. There's a reason that machine tools are heavy. Basically, if it were this easy to build a metal cutting machine that can hold tolerances to .0005 for less than $500, then everyone and his cousin would already be doing it. That little machine that NewZed had a link to went for something like $270 and wouldn't hold anywhere near the tolerances we're talking about here.
Remember, you are talking about the cuts in the object having an accuracy to .0005, not simply that the rig can move around in a high precision manner without cutting. The cutting bit will be at the end of a spindle and the end will suffer from deflection if the entire machine that holds it in place isn't solid enough.
However, I'm obviously beating a dead horse here, and I apologize for that. It's clear that you have lofty goals and sound far more willing than most to spend the money needed to achieve them. I believe that your stated intention of using precision components will allow you to achieve a considerably higher tolerance than basically everyone else here. This is a good thing, and I wish you success.
I have been reading all the posts but for the most part have kept out of the conversation because I am pretty low on the totem pole of knowledge when it comes to CNC stuff.
Peter mentioned my little Proxxon MF70 mill.· I have installed a Gecko-controlled bipolar on the Y table.· My goal was to punch in 360 degrees on my little keypad and have the table move exactly .110 inches.· I am acheiving that plus or minus 1 mil.· When I reverse directions on the Y table, I lose about .0025; however, the program has a little "nudge" routine built in and I can bump the Y table in increments of .0005 to make up for the loss.· I am monitoring the Y table movement with a dial indicator so I know exactly where I am.· I did not try to attach the motor to the Y table since that would have increased the weight on the table.· I mounted a piece of galvanized, highly polished and waxed, on my workbench and anchored it down.I put another smaller piece under my motor mount which has one side polished and waxed.· This side mates with the larger piece.
I adjusted the height of the motor with strips of making tape and mounting foam until it had no offset when I slid it up to the shaft hole in my handwheel adapter.· Now the motor assembly slides effortlessly as the Y table moves.
I tried to measure the runout on the spindle.· I put the dial indicator pointer on the .125 shank of a .035 drill, just above the point where the shank begins to taper into a drill.· When I rotated the spindle by hand I could detect no runout.· Since I'm sure there must be some runout, I will try this again.
I'm working on motorizing the X table right now.· That should be a bit easier thatn the Y table since the X handwheel does not move.· I also plan to use a timing pulley/belt arrangement so I will not have the problem of alighning the motor shaft with the handwheel adapter.
The mill will handle a single-sided PC board 1.8 x 4.5 inches which is adequate for the moment.· As soon as I figure out an accurate registration procedure I may try double sided.
If I may repeat a statement I made at the beginning of this project, the little MF70 seems like an excellent starting point, especially for the little guys like me.
I just rechecked the runout on my MF70 spindle.· I performed the test 5 times, each time approaching the spindle at a different angle.· There was no runout error detected, and I can easily se a deflection of .0001.· I am not saying there is no runout - I am saying that it is less than 1 ten-thousandth.
I think you're overlooking what I was saying in my last post. I don't expect folks can build the aluminium milling machine I want with half-mil tolerances for $500. I do however think folks can make the PCB mill that accurate for around $500. The ballnuts and linear bearings for a 12x12x3" machine can be had on eBay for under $200 if one shops carefully. Folks do it all the time. ABEC-7 and -9 spindle bearings can be had at high end skate shops for a paltry sum, compared to the normal bearing distributors. When cutting FR4 and 1oz copper with a carbide bit spinning in excess fo 10k rpm, and if one is willing to go slowly (ie, not 6in/s of linear cutting rate), the tolerances can be tight AND the machine can be lightweight.
Big machines are big and heavy because they need to work on big things with lots of force. If we limit the work envelope, the machine can get smaller, lighter, and cheaper, because we're working on small things with very little force. Go slow, work on relatively soft materials in a small area, and precision gets easier. The tests Sid has done with his little mill show that the precision (save the backlash on his leadscrews) can be had in small, lightweight machines. What Peter (pjv) is proposing in terms of the material and motion equipment will make it achievable.
I'm not dismissing your point of view. I held it for a long time myself, having learned machining on an ancient and gigantic 3 ton Bridgeport in a shop in college. It wasn't until I saw the 3D dental mill at my dentist's office that I started to see what was possible if one only wants to mill small things that are relatively soft. I started poking around and found out how cheap THK linear bearings can be acquired if you shop around the surplus markets, and how easy it is to make a high precision spindle with aluminum and steel rod, a lathe, and some skate bearings. I have an acquaintance that has a gantry mill that he's gotten the accuracy down to 0.0003 over it's 24" maximum dimension, and he mills hard brass and mild steel! He paid a lot more for prime heavy duty parts because he doesn't like the idea of surplus, but with the lighter duty of the PCB mill, we have more options, and can get more done for less money. The $500 is predicated on people doing a lot of DIY work to build the mill, and folks with aptitudes beyond changing a tire should hit a mil or two with ease. Those with more motivtion and time can do better.
I honestly believe we can do this.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
I would like to mention that we have an LPKF milling machine at work that I make a fair number of PCBs on. I recently purchased one of the toner transfer systems from pulsar. I have to say that I will most likely only be using the LPKF for drilling holes and singulating the boards from the panel. The pulsar system is fast, easy and cheap. The milling system is slow, LOUD and expensive. The pulsar website says that you can make a PCB in 15 minutes. It took me about 20 minutes for my 1st single sided board. It best part is if I made a panel of 9 boards it would have STILL taken only 20 minutes. The LPKF would have taken hours literally.
I know that there are some that will insist that our "beast" be able to mill a pcb, but I tell you having tried both, I'll stick to etching. Sure it's cool to see the mill cutting out your pcb. But you have to adjust the bit depth constantly you have to change bits constantly. The machine has a noisy vaccum that we must wear earphones when working in the same room. The one downside of the etching system is...You guessed it...Chemical disposal [noparse]:([/noparse]
Just my opinion, take it for what it's worth...But I do recommend the pulsar system www.pulsar.gs
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
I share your sentiments, but there are times when the machine can sit there and grind away without much further baby-sitting. Depth I expect to be OK, but bits, well.....
There are numerous rectangular shaped areas I need for my RF work, and these ought not take too long, whereas breaking out the etcher every time and heating the solution.....
Furthermore, there are MANY other uses for such a positioning machine; the mind boggles.
Some more design has been done, but time is real precious just now, so please be patient. I think something good is coming....
Speaking of other uses. Something that the LPKF says it can do, but we have never tried is "dispensing solder paste".
I think we would have to buy some kind of extra equipment, but the software supports it. That would be way cool.
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
Right Bean, and right along with that a glue dot dispenser for SMT components. The list just goes on.....
I imagine in due course we will be fabricating all kinds of attachments. The way the Z axis is being designed, it allows for quick-change tooling with just the proverbial "flick of the wrist".
PLJack; I'm not hearing of any advancement on the software front. This could be huge effort.
Terry; Will the LPKF equipment support tracing the outline of any Gerber file? Is there any layout package out there that supports edge tracing directly?
I like the Pulsar product. it works great for me. What I like also is their Green Film. Putting it over the toner fills in pits and hardens the etch mask. I can etch the board with 2 ounces of ferric chloride and a small foam paint brush. A 4 inch square board took me four minutes to etch. No need for etching tanks etc.
I also bought at $49 laminator for precision work. (I'm moving to surface mount devices and my traces are getting small.) For through hole components a laundry iron worked fine for me.
Given the persistent doubts about the accuracy of our machine, should we focus more on cutting board outlines and holes than milling traces for the first round? I know that I'd be pretty happy if I could get a machine to mill out my panelized PCBs for me and drill the holes I need drilled. Internal routes would be easy, as would oddball shapes. It would make the software effort a little more managable too (grab and parse the gerber layers for the outline and the .drd and .drl files for PCB drills).
Thoughts?
My lofty goals are still mine, but the more I think about it, the easier it will be to build them on the shoulders of simpler processes and systems.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
Post Edited (Dave Paton) : 7/6/2005 1:31:35 PM GMT
I think Drilling and routing would be a logical stepping stone to a full milling software package.
Another comment...I would like the software to handle DXF file also. I (and many others) use CAD software for board layout, and there is no cheap software to convert to gerber formats. I have a turbo pascal program that I wrote to read DXF files if that helps.
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
PLJack; I'm not hearing of any advancement on the software front. This could be huge effort.
Peter (pjv)
Fear not. I'm still researching.
I have written many Comm applications so I'm not concerned about that aspect.
Two items:
1) As you have said the applications for this "mill" are endless. So a straight CNC app
is probably not the way to go. That greatly effects my software design. I need to keep
that in mind.
2) Software can not be designed around vagueness. I need more specifications.
We need to settle on a PC to IC protocol. I Could probably implement most of the
"G-Code" specifications but is that the best approach for a multi-purpose mill.
How will the SX implement my commands.
Should we stop calling this project a Mill?
There are many questions like this I need to consider before I start coding.
This is normal to software development. There is a lot of conceptual time on the
front end before coding begins.
Also to work the software we need a development environment, such as an SX chip with code to test.
And to finish the software someone needs to build a version of this mill.
I will restate here that I am committed to all of the above.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - 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.
I was tackling the challenge of mounting the pcb on the CNC carriage for drilling and getting it perfectly aligned, trying to get holes that line up perfectly with the artwork.
...
So a simple mathematical approach came to mind.
...
This is working accurately, even if I stick the board down sideways,
upside down or at a diagonal. The interactive board calibration
procedure takes about 20 seconds.
So what is the consensus here.· Does the project want tidbits like this referenced?· If so, does the project just want a link to the information, or should the relevant information be quoted into this thread?·
Continuing with my MF70 project, everything is working good, but there is still that .002 backlash.· So....I determined that a 7 degree rotation moved the table exactly .002.· I modified my program so that when the program detects a reversal in travel direction, it automatically inserts a routine to perform the 7 degree rotation.· I have to make a lot more tests but right now it's looking pretty good.· Which all goes to show that with a program change or two you can compensate for almost any consistent error.
Do we have the movement hardware defined ? If so let me know and I'll start on the SX code.
Also how are we going to "zero" the position ?
As for commands, I assume something like "M+100,-25,50" where "M" means move, +100 is the "X" adjustment in steps, -25 is the "Y" in steps, and 50 is the speed. The X and Y motors would be steped proportionally to create a straight line to the destination position.
How kind of "Z" axis command are required ?
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
··· G1··· ==> move ··· X100· ==> X-axis·100 units ··· Y-25· ==> Y-axis·-25 units ··· F50 · ==> resultant vector's·velocity is 50·units ··· Z-3·· ==>·Z-axis·-3 units
The G-Codes would not take many more bytes, using the axis name as the delimiter instead of the comma, and the numbers are labeled.
Just ignore whitespace and comments (parenthesis delimted).
Then the SX motion "processor" could simply be·feed a "part program" from something as simple as·typing into Hyperterm for very simple work/testing to the many CAM packages in the middle, to the project's own PC front-end at the high-end.
Bean and Daniel, I think you are talking about two different things.
I really dont think the SX will have enough RAM to decode G-Codes.
G-Code translation is the job of the PC software.
Daniel is talking about the PC side. Bean is talking about the SX side.
I feel G-Codes should be just one format the PC software should support.
Whatever the format I will translate it to be compatible with Beans SX code.
Bean, Design your code to be the most efficient way for an SX to manipulate motors and perform
serial communication.
I will follow your lead with the software.
I have one request for the SX code. Given that this software will be running on several
OS's and all kinds of hardware I feel the SX and PC software should be in constant communication.
To send a command to the SX and assume that it was carried out is hopeful at best.
When a command is sent to the Sx the PC software should wait for the SX to confirm successful
completion before the next command is sent.
This will remove any dependance on hardware/software timing issues.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - 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.
Jack,
You hit the nail on the head. The SX should be given only the most rudimentary commands.
This will not only make my work easier [noparse];)[/noparse] but will make the hardware more flexible.
I assume we ARE going to be using steppers ?
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
Comments
For those interested......
Several posts above, I made reference to exposing a printed cicuit board with a flexible optical fiber.
Years ago, when flat-bed plotters were all the rage, I actually did this; using the pen-up/pen-down solenoid signal to turn an LED off and on. I used the actual PC layout plotting software (Protel) to do the drawing. I selected the "pen width" close to the diameter of the fiber, and got some surprisingly good results. With today's brighter and white LEDs, one can probably do a better and faster job.
Although it is was a little slow, what was neat about this approach, you can get (close to) photo results without involving a photoshop; just software. I absolutely LOVE "soft-tools".
Presently I am also working on another (Laser) based approach that will, if it pans out, reduce the "exposing" time to 10 seconds or less. With 1 mil resolution, a bit costly to manufacture though, but the processing cost should be minimal.
Any interest in these ideas??????
Cheers,
Peter (pjv)
The idea of using the high density plastic for a frame is a good idea, easy to machine and to work with. As you can see by the construction, it would not be hard to fabricate, and makes for a simple design. If anything, you can get a good idea of the gantry design by the pictures.
And the porter/cable spindle looks like a good pick too, up to 30,000 rpm with a collet holder.
kelvin
Sorry about the linear vs. chopper confusion. I've been dealign with a similar issue here at work and thought (erroniously) that you meant linear. I really need to disconnect from work more before I post here before I cause real trouble
The laser and fiber optic exposure both sound really interesting. I remeber putting down etch-resist with an HP flatbed plotter some time ago with excellent results. A small UV laser should work wonders for the presensatized boards at small outputs (1-5mW?).
They're manufacturer datasheets, but include the frame sizes you asked for. They're standardized just about everywhere, but these PDFs should make it nice and easy for you:
http://www.pennmotion.com/pdf/Size23.pdf
http://www.pennmotion.com/pdf/Size34.pdf
http://www.pennmotion.com/pdf/Size42.pdf
Shafts are 0.25", 0.5", and 0.625" nominal diameter, respectively.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
I know the whole issue of your desired accuracy versus everyone else's is pretty much a done deal, but I have a couple of questions.
1 - In an earlier post, You mentioned building "a 14" square base that was flat to no less than 0.001". Could you give a few more details? What material was used? How thick was the material? How were the pieces fastened together? Since you mentioned using a miter saw, I assume this was open like a picture frame as opposed to a flat, solid plate. How did you measure the flatness of this item?
2 - You talk about accuracy of .0005" (which is 1/2 of 1/1000 of an inch) as if this is a piece of cake to achieve. I do milling on a Bridgeport with a DRO (Digital Readout) and just setting the locks on an axis is often enough to move an axis .0005". I know that the axis has moved because the DRO is good to .0005" and I can watch the number jump. If a machine that weighs over 2000 pounds can so easily move .0005", how do you expect a little tabletop device that weighs under 100 pounds to do better?
Thanks, PeterM
I built the machine base from 4" square aluminium extrusions similar to the ones from 8020. It It was a simple, butt-jointed square (like a cheap picture frame)designed for an elevated optical bench apparatus. The pieces were fastened with some glorified right angle brackets I had machined, and tweaked regularly for the better part of a day to get it "just right". The flatness was verified on the granite surface plate at the customer's location (I coldn't get a 0.001" feeler anywhere under it, on either side, btu there was still a small light leak when we looked very carefully), but I checked it while I was building it with an 18" level, a feeler guage, and a24" square 4-layer stackup of 3/4" MDF that was as flat as anything else in my garage. The reason I said it took a weekend was that I spent considerable time aligning my miter saw, tweaking for maximally flat cuts, etc. If I had the pieces precut with square flat edges, I think I could have done it in a few hours total, knowing what I know now.
As for your Bridgeport, I have no doubt that the ways allowed 0.0005" of slop when you tightened the locks. The one ton of weight of the machine keeps it stable and solid when moving it's 100lb spindle assembly around, machining large, hard parts. That's the way it's designed. You're talking about big, heavy moving parts. There's less precision in the movements of thos big clock movements on the towers, and it's not just because one was made in Switzerland. Smaller, lighter components allow finer tolerances when they are moved.
What I'm talking about is a much lighter weight machine, for machining much smaller items from much softer material. Specifically, for the PCB drilling aspect of things, it should be pretty darn easy to keep 0.001" of calibrated accuracy, precision AND repeatibility, and probably 0.0005" if folks take good care in building and maintaing their machines. The sheer issue of scale makes these goals easier to achieve. We're moving a whole lot less mass. We're using a gantry, which has no moving ways or gibs under the workpiece, simplifyign the relative motions. Slop is easily removed from the gantry's motions with ballscrews, recirculating ball nuts, and precision linear bearings (to 0.0005" or better) tho it will cost more money to get that accuracy than with hardware store threaded rod and drawer glides (like a lot of the wood CNC guys use). This won't be a bash-around router for the garage. It will be a piece of fine machinery that likes to sit inside on a benchtop. We too will need a solid base, but because of the moving masses involved, we won't need 2000, or maybe even 200lbs. I think for the goal of making PCBs, the weight of the machine itself will do just fine.
The way that Peter (pjv) is putting this together, I feel strongly that people will have access to not only high precision motion devices, but accurately machined structural components that can easily be assembled to construct a small, precise, lightweight machine capable of sub-mil accuracy. I don't think everyone will achieve that. It will take time, planning, tools, patience, and more patience, but I think it is posible.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
I admire your tenacity in anticipating sub-thou performance. I too hope that that can be achieved, but I continue to believe we should be happy with 2 to 3 thou, and elated with 1 thou. I can only dream of better than that.......I expect temperature variations alone will throw us off some tenths.
Nevertheless, onward !!!!!!!!
Some money making activities got in the way of the design process, but I expect to be back at it early in the week. Please remember this is a volunteer activity, and needs to be accomodated by volunteer hours.
Cheers,
Peter (pjv)
I can relate to that.
Do what you can Peter. We all appreciate your effort.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Jack
I think we can all relate to the inherantly second-class nature of hobby projects to those providing important things like pizza and mortgagae payments.
I'm tenacious because I know others have done it with a less organized effort. If some hamfisted guy with minimal mechanical engneering exprience can buy things on eBay and assemble a machine that can hit one mil all the time (one guy at work did, and it worked great after we tweaked it for him), with the combined efforts of the real engineers and experienced enthusaists here, it should be a lot easier.
Of course, if we fail misterably, I'll be the first one to eat crow, and I'll like it
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
I guess my confusion arose from the fact that you intend to build a completely different machine than basically everyone else. The thread got started when someone wanted to build a cheap machine for milling and drilling PCBs. In this case, cheap was $500 or less. As soon as you start talking about lead screws and ball nuts driving machinery on precision hardened rails, you're falling off the $500 train pretty fast.
I agree that with the parts you've spec'ed, it's possible to achieve the tolerances you talk about. I disagree that it is easy to build a machine that will hold these tolerances while machining something else. While it might be possible to achieve these tolerances in terms of simply driving the rig around, the minute you add something like a spindle and a cutting bit which have to cut through a layer of copper, then everything just got a lot harder.
As you know, the reason I brought up the weight issue is that when cutting metal (and copper cladding definitely counts as metal), mass is one of the most important attributes needed to achieve accuracy. There's a reason that machine tools are heavy. Basically, if it were this easy to build a metal cutting machine that can hold tolerances to .0005 for less than $500, then everyone and his cousin would already be doing it. That little machine that NewZed had a link to went for something like $270 and wouldn't hold anywhere near the tolerances we're talking about here.
Remember, you are talking about the cuts in the object having an accuracy to .0005, not simply that the rig can move around in a high precision manner without cutting. The cutting bit will be at the end of a spindle and the end will suffer from deflection if the entire machine that holds it in place isn't solid enough.
However, I'm obviously beating a dead horse here, and I apologize for that. It's clear that you have lofty goals and sound far more willing than most to spend the money needed to achieve them. I believe that your stated intention of using precision components will allow you to achieve a considerably higher tolerance than basically everyone else here. This is a good thing, and I wish you success.
Thanks, PeterM
Peter mentioned my little Proxxon MF70 mill.· I have installed a Gecko-controlled bipolar on the Y table.· My goal was to punch in 360 degrees on my little keypad and have the table move exactly .110 inches.· I am acheiving that plus or minus 1 mil.· When I reverse directions on the Y table, I lose about .0025; however, the program has a little "nudge" routine built in and I can bump the Y table in increments of .0005 to make up for the loss.· I am monitoring the Y table movement with a dial indicator so I know exactly where I am.· I did not try to attach the motor to the Y table since that would have increased the weight on the table.· I mounted a piece of galvanized, highly polished and waxed, on my workbench and anchored it down.I put another smaller piece under my motor mount which has one side polished and waxed.· This side mates with the larger piece.
I adjusted the height of the motor with strips of making tape and mounting foam until it had no offset when I slid it up to the shaft hole in my handwheel adapter.· Now the motor assembly slides effortlessly as the Y table moves.
I tried to measure the runout on the spindle.· I put the dial indicator pointer on the .125 shank of a .035 drill, just above the point where the shank begins to taper into a drill.· When I rotated the spindle by hand I could detect no runout.· Since I'm sure there must be some runout, I will try this again.
I'm working on motorizing the X table right now.· That should be a bit easier thatn the Y table since the X handwheel does not move.· I also plan to use a timing pulley/belt arrangement so I will not have the problem of alighning the motor shaft with the handwheel adapter.
The mill will handle a single-sided PC board 1.8 x 4.5 inches which is adequate for the moment.· As soon as I figure out an accurate registration procedure I may try double sided.
If I may repeat a statement I made at the beginning of this project, the little MF70 seems like an excellent starting point, especially for the little guys like me.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Sid Weaver
Do you have a Stamp Tester yet?
http://hometown.aol.com/newzed/index.html
·
Not too hot at typing, either.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Sid Weaver
Do you have a Stamp Tester yet?
http://hometown.aol.com/newzed/index.html
·
I·shall now return to the shadows.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Sid Weaver
Do you have a Stamp Tester yet?
http://hometown.aol.com/newzed/index.html
·
I think you're overlooking what I was saying in my last post. I don't expect folks can build the aluminium milling machine I want with half-mil tolerances for $500. I do however think folks can make the PCB mill that accurate for around $500. The ballnuts and linear bearings for a 12x12x3" machine can be had on eBay for under $200 if one shops carefully. Folks do it all the time. ABEC-7 and -9 spindle bearings can be had at high end skate shops for a paltry sum, compared to the normal bearing distributors. When cutting FR4 and 1oz copper with a carbide bit spinning in excess fo 10k rpm, and if one is willing to go slowly (ie, not 6in/s of linear cutting rate), the tolerances can be tight AND the machine can be lightweight.
Big machines are big and heavy because they need to work on big things with lots of force. If we limit the work envelope, the machine can get smaller, lighter, and cheaper, because we're working on small things with very little force. Go slow, work on relatively soft materials in a small area, and precision gets easier. The tests Sid has done with his little mill show that the precision (save the backlash on his leadscrews) can be had in small, lightweight machines. What Peter (pjv) is proposing in terms of the material and motion equipment will make it achievable.
I'm not dismissing your point of view. I held it for a long time myself, having learned machining on an ancient and gigantic 3 ton Bridgeport in a shop in college. It wasn't until I saw the 3D dental mill at my dentist's office that I started to see what was possible if one only wants to mill small things that are relatively soft. I started poking around and found out how cheap THK linear bearings can be acquired if you shop around the surplus markets, and how easy it is to make a high precision spindle with aluminum and steel rod, a lathe, and some skate bearings. I have an acquaintance that has a gantry mill that he's gotten the accuracy down to 0.0003 over it's 24" maximum dimension, and he mills hard brass and mild steel! He paid a lot more for prime heavy duty parts because he doesn't like the idea of surplus, but with the lighter duty of the PCB mill, we have more options, and can get more done for less money. The $500 is predicated on people doing a lot of DIY work to build the mill, and folks with aptitudes beyond changing a tire should hit a mil or two with ease. Those with more motivtion and time can do better.
I honestly believe we can do this.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
Happy building!
Thanks, PeterM
I know that there are some that will insist that our "beast" be able to mill a pcb, but I tell you having tried both, I'll stick to etching. Sure it's cool to see the mill cutting out your pcb. But you have to adjust the bit depth constantly you have to change bits constantly. The machine has a noisy vaccum that we must wear earphones when working in the same room. The one downside of the etching system is...You guessed it...Chemical disposal [noparse]:([/noparse]
Just my opinion, take it for what it's worth...But I do recommend the pulsar system www.pulsar.gs
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
http://www.parallax.com/detail.asp?product_id=30012
Product web site: www.sxvm.com
"What's the difference between ignorance and apathy ?"
"I don't know, and I don't care."
·
I share your sentiments, but there are times when the machine can sit there and grind away without much further baby-sitting. Depth I expect to be OK, but bits, well.....
There are numerous rectangular shaped areas I need for my RF work, and these ought not take too long, whereas breaking out the etcher every time and heating the solution.....
Furthermore, there are MANY other uses for such a positioning machine; the mind boggles.
Some more design has been done, but time is real precious just now, so please be patient. I think something good is coming....
Cheers,
Peter (pjv)
I think we would have to buy some kind of extra equipment, but the software supports it. That would be way cool.
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
http://www.parallax.com/detail.asp?product_id=30012
Product web site: www.sxvm.com
"What's the difference between ignorance and apathy ?"
"I don't know, and I don't care."
·
I imagine in due course we will be fabricating all kinds of attachments. The way the Z axis is being designed, it allows for quick-change tooling with just the proverbial "flick of the wrist".
PLJack; I'm not hearing of any advancement on the software front. This could be huge effort.
Terry; Will the LPKF equipment support tracing the outline of any Gerber file? Is there any layout package out there that supports edge tracing directly?
Cheers,
Peter (pjv)
I like the Pulsar product. it works great for me. What I like also is their Green Film. Putting it over the toner fills in pits and hardens the etch mask. I can etch the board with 2 ounces of ferric chloride and a small foam paint brush. A 4 inch square board took me four minutes to etch. No need for etching tanks etc.
I also bought at $49 laminator for precision work. (I'm moving to surface mount devices and my traces are getting small.) For through hole components a laundry iron worked fine for me.
Thoughts?
My lofty goals are still mine, but the more I think about it, the easier it will be to build them on the shoulders of simpler processes and systems.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
Post Edited (Dave Paton) : 7/6/2005 1:31:35 PM GMT
Another comment...I would like the software to handle DXF file also. I (and many others) use CAD software for board layout, and there is no cheap software to convert to gerber formats. I have a turbo pascal program that I wrote to read DXF files if that helps.
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
http://www.parallax.com/detail.asp?product_id=30012
Product web site: www.sxvm.com
"What's the difference between ignorance and apathy ?"
"I don't know, and I don't care."
·
Fear not. I'm still researching.
I have written many Comm applications so I'm not concerned about that aspect.
Two items:
1) As you have said the applications for this "mill" are endless. So a straight CNC app
is probably not the way to go. That greatly effects my software design. I need to keep
that in mind.
2) Software can not be designed around vagueness. I need more specifications.
We need to settle on a PC to IC protocol. I Could probably implement most of the
"G-Code" specifications but is that the best approach for a multi-purpose mill.
How will the SX implement my commands.
Should we stop calling this project a Mill?
There are many questions like this I need to consider before I start coding.
This is normal to software development. There is a lot of conceptual time on the
front end before coding begins.
Also to work the software we need a development environment, such as an SX chip with code to test.
And to finish the software someone needs to build a version of this mill.
I will restate here that I am committed to all of the above.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - 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.
So what is the consensus here.· Does the project want tidbits like this referenced?· If so, does the project just want a link to the information, or should the relevant information be quoted into this thread?·
Daniel
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Sid Weaver
Do you have a Stamp Tester yet?
http://hometown.aol.com/newzed/index.html
·
Also how are we going to "zero" the position ?
As for commands, I assume something like "M+100,-25,50" where "M" means move, +100 is the "X" adjustment in steps, -25 is the "Y" in steps, and 50 is the speed. The X and Y motors would be steped proportionally to create a straight line to the destination position.
How kind of "Z" axis command are required ?
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
http://www.parallax.com/detail.asp?product_id=30012
Product web site: www.sxvm.com
"What's the difference between ignorance and apathy ?"
"I don't know, and I don't care."
·
··· G1X100Y-25F50··· (move to pad)
··· G1Z-3F10······· (drill hole)
··· G1Z3F50········ (retract drill)
where
··· G1··· ==> move
··· X100· ==> X-axis·100 units
··· Y-25· ==> Y-axis·-25 units
··· F50 · ==> resultant vector's·velocity is 50·units
··· Z-3·· ==>·Z-axis·-3 units
The G-Codes would not take many more bytes, using the axis name as the delimiter instead of the comma, and the numbers are labeled.
Just ignore whitespace and comments (parenthesis delimted).
Then the SX motion "processor" could simply be·feed a "part program" from something as simple as·typing into Hyperterm for very simple work/testing to the many CAM packages in the middle, to the project's own PC front-end at the high-end.
Daniel
·
I really dont think the SX will have enough RAM to decode G-Codes.
G-Code translation is the job of the PC software.
Daniel is talking about the PC side. Bean is talking about the SX side.
I feel G-Codes should be just one format the PC software should support.
Whatever the format I will translate it to be compatible with Beans SX code.
Bean, Design your code to be the most efficient way for an SX to manipulate motors and perform
serial communication.
I will follow your lead with the software.
I have one request for the SX code. Given that this software will be running on several
OS's and all kinds of hardware I feel the SX and PC software should be in constant communication.
To send a command to the SX and assume that it was carried out is hopeful at best.
When a command is sent to the Sx the PC software should wait for the SX to confirm successful
completion before the next command is sent.
This will remove any dependance on hardware/software timing issues.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - 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.
You hit the nail on the head. The SX should be given only the most rudimentary commands.
This will not only make my work easier [noparse];)[/noparse] but will make the hardware more flexible.
I assume we ARE going to be using steppers ?
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
http://www.parallax.com/detail.asp?product_id=30012
Product web site: www.sxvm.com
"What's the difference between ignorance and apathy ?"
"I don't know, and I don't care."
·
More research. Yah!!!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - 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.