Newzed:
Thanks for the input.
2+ Amps?? Wow, thats a lot of load. And at 165 each... I take it it Mills rather fast.
Would you consider your mill to be of a "professional" grade, vs a "hobbiest" grade?
Graham: THanks for the links. I'm checking them out as we speak.
I can see how this is going to get expensive real fast...
Thanks guys...
BTW: Is there a link or site that actually gives the GCode spec and it's contents? If I'm going to author a GCODE based Mill programm, I'll need to know how to interpet the incomming GCode...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
KK, the Y table is geared dowmn 2:1 because I didn't wnt it to go too fast.· I wasn't sure what speed the cutter could safely travel at using a .015 router - they are pretty fragile.· The X drive is controlled by an old Gecko microstepper that was given to me, so it is quite a bit slower than the Y drive, but I'm working on that.
SuperMill does a pretty good job, but I don't think it could be called "professional grade".· The main problem is the backlash.· The MF-70 has about a 2 to 3 mil backlash and I have compensating routines for that.· The other problem is that on the X table it takes 51.56 pulses to travel 1 mil and on the Y table it takes 20.3 pulses to travel 1 mil.· The Stamp can not handle decimals so on short travels there is always a bit of an error.· I'm hoping the Propeller program, with its floating-point math, will really improve the accuracy.
Regarding gcodes, to create a gcode you must have your etch pattern in a .dxf file, which my Expresspcb can not do.· After you have the .dxf file it must be converted to gcode using something like the ACE converter.
So.........I write my own etching programs.· When I am doing a new board, I draw each etch path, then write down the ending coordinate and whether it is the X or Y direction.· This data is copied into an Excel template which generates a DATA list the Stamp can read.· This DATA list is the pasted into a Stamp conversion program which converts the Excel DATA into a proper etching DATA list that my mill program can read.· Only two directions are generated by the conversion program -
L(eft) and I(n).· Then the mill program reads the DATA list, it compares the coordinate value with the current position.· If the coordinate is greater it travels Left or In a distance equal to coord - posx (or posy).· If the coordinate is less than the current position it travels R(ight) or O(ut) a distance equal to posx (or posy) - coord.· A bit time-consuming sometimes, but the system works very well.
Newzed...
Well, it sounds like you have a system, somewhat long, but working.
While I have been reading this (and the home brew milling) threads, I was thinking about going directly to GCode, but seems problemmatic (Ie no expresspcb support).
And doing GCode is going to take some time, but I do have an alternate idea that will probally work much easier... and better.
Take a screen shot of the board in expresspcb and save it as a 8bit color BMP .
My application will open the BMP and generate some form of code for the Mill to directly run. (Code very simular to your code, if not the same.)
The idea is to get "Human Hands" out of the data stream.
My current approch is to start at 0,0, read across the x, lift the z when coper is needed, drop the z when it's not.
Hit the end of that X line, move to the next y line, and return back to X:=0, lifting and dropping the Z as needed.
Now all this "encoding" could be placed into "data" statements for your mill.
It would need to have a "backlash setting", but that would be mininal due to the reduced +- movment of the axis.
There would be some editing of BMP, adding/ using colors so to 'dimple' the through holes, and other extra things.
Well, this part is just an idea. My buddy and I are very serious at this project.
He's got thoughts on handling backlash, and I'll get a chance to talk to him later on tonight about it, but his words were "I'v got the back lash beat"... more on that later.
What I'll do is suggest his ideas to you two for your "feel" on it (you both know a whole lot more then him or I).
My goal with this project is to be able to produce double sided PBC's within the same spec range as expresspc does... without going out of house or chemicales.
I'm not sure if that feasable or not, but I'll get that information / idea once Joe (my buddy) decides on the hardware purchase (it's his money..)
I like your setup Newzed, but I'm somewhat confused why you don't have all three steppers in the exact same setup, thus advoiding the mill dravel / pulse conversions.
Parsko: I've only just started the research aspect of this. I'v no hardware as of yet...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
KK, consider this.· Excluding any backlash or any other external conditions,one revolution of a handwheel produces .03936 inches of travel.· Let's round that off to .0394.· Suppose you want to travel 10 mils and your stepper controller is operating at half-step.· .010/.0394 = .254
revolutions.· With the controller half-stepping it takes 400 pulses for one evlution, so for .254 revolutions you would want to pulse the stepper
.254*400 = 101.5 pulses.· For a travel of 25 mils, you would pulse the stepper .025/.0394*400 = 253.8 times.· You can't do a fractional pulse so you're always going to drop the decimal, which always introduces a certain amount of error.· On my mill, for the Y travel it is 10.15 pulses per mill; for the X table it is 51.56 pulses per mil.· With the Stamp I get as close as possible with the " */ " operator.· On the Propeller, with the floatmath object, I hope to improve the accuracy.· And there are mechanical considerations - attaching the stepper to the handwheel (or to the leadscrew shaft) so that it is perfect alignment.· Not trying to discourage you by any means -·it took me about four months to get SuperMill to produce an acceptable single-sided PC board, and I'm still working on improvements.· Double-sided is much tougher, even with perfect registration.· I tried to make one, using the shanks of broken drills for registration pins.· It was just barely acceptable.· I haven't tried again since, but I may once I get the Prop hooked up to the mill.· We'll see.
I've redone the stepper slider assembly for the X table - I'll post a picture of it tomorrow.
Newzed...
Good points... Very good points. I haddn't thought of the actual pulse to step ratios.
I take it half stepping is better off the the qtr stepping as listed?
I'v got some serious math and thinking infront of me...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
Don't forget that the Y table has a 2:1 reduction, so even though I'm half-stepping the stepper, the Y table is getting quarter-stepping.· If you are direct-driving, I would start with qtr-steps.· Go too fast and you'll snap off your cutter.· And you will need a dial indicator - LittleMachineShop.com, $19.95 -· so you can accurately measure table travel when you are setting up your pulse-step routines and your backlash compensation.· And you will need a backup board for your copper.· I use 3/8" x 3" oak strips from Home Depot, cut into 5" lenghts.· The backup is help in place by three strips of golf grip double sided tape - Golfsmith.com, 3/4", $4.19.· I use the same tape to hold my PC board in place - I quit using the two hold-down clamps because·they do not hold the board as flat as the tape.
I just ran Compile/View Info for my Propeller mill program with TV display and for the same program with VGA display.· According to what I see, switching to VGA will save me 3996 longs of memory.· Does that sound reasonable?
although the number of microsteps effects the max system speed it should not effect whether you snap your cutter or not as you should simply reduce the step rate to create a suitable feedrate.
when considering microstepping realize that you can't necessarily make the mill step a single microstep as there is insufficient torque, what it does give you is smoother motion and less noise, the system can start to sound more like it has servos.
I have used it to easily export Gcode from Eagle, then make double sided boards with no problem at all, traces in between pads etc,. It does take a few practice runs. There is a free Eagle demo, with limited pcb size though, but fully functional to try the gcode generator.
You can download Mach3 (Gcode machine output software) free here at :
This will allow you to import some of your gcode, then test it in simulation mode even without a machine hooked up.
I used to use this PCBgcode plugin with Eagle before I burnt out on homespun boards. Now I send the Eagle board file striaght to pcbexpress.com. If you get into Eagle, you can run the same file on your mill or send it out, same files either way, no gerber stuff to deal with. The drag is the cost of Eagle versus the free softwares.
Just currious why you guys are having backlash issues, aren't you using anti backlash nuts?
I am using LIN engineering steppers with encoders, they are quite robust and take a lot of abuse milling aluminum, even stainless sometimes. Another company that I am starting to get some motors in from is US Digital, you guys might like that site as well. Extremely helpful tech guys.
Well, my evening was full of insights.· I spent the better part with a Mill operator (friend of a friend), and plucked his brains.
While his point of view was mostly from the working end of the process, he did have a lot to say about the various mills he's used.
Here are some of his points of "view".· I should preface this with the following.· On the drive to his shop, Joe was very VERY explicit about measuring.· Everything they mill starts in mm, stayes in mm and ends in mm.· THe machines, the tools, the software, EVERYTHING is done in mm.
#1: The lead screw's thread count, and the driving gear train from the stepper all must workout to a basic uint of measurment.
He Explains that on his "best" mill, the lead screw's pitch is quite simple: 1 * 360 = 1 mm of travel.
There's gearing on the stepper to make the stepper's rotation· =· 1 * 360 = 3600 steps.
Now, at this point, stepping is a direct·function of speed, not measurment.
#2: About backlash.· Joe explained to me what backlash was:· The effective slack between the lead screw and the lead nut.
There are quite a few answers·from reduce to·illimate backlash.· Simple ones (backlash nuts & washers) only reduce·backlash to a·stable and non changing measurment.· Elabrate answeres move the backlash measurment below the resolution of the mill.· I asked him to clairfy this statement, and added that each and every mill has it's resolution.· The point at which·changes in measurments·have no effect on the milling process.
An example would be if a Mill had a 1 mm resolution, that means that measurments of .9 and lower have no effect in a stright cut.
His production mill has a measured backlash of .0001 mm, well below the size of the smallest bit used in the machine.· The machine is rated at 0.001 mm resolution. (that's real small, but·neither he or I could tell·where that was in the PCB world.)·
As long as all movments are expressed as truncated numbers above this measurment, backlash is non existant.·
The machine I was shown has·an anti-back lash system on all three axis.· Air over Spring. Something like one of the double stroked hand air pumps (it pumps air in both the up and down directions) but much better, under a constant·pressure of {He said some number and I can't remember it but it like·50 cm^3}· The air cylinder is attached to the lead nut, directly below the lead screw with very heavy duty springs on both directions.
#3: Power.· Joe showed me the controll box for three of the Mills in his shop.· 2 of the three have very distinctive 24V @ 2.5 amp lables.· THe other power supply, the one attached to his production (best) machine as 24V @ 3.0 amps.· Electrally there is no differance between the steppers and power supplys on any of the machines.· THe·difference between the machies is the production has a chuck cradle with 15 bits as the 4th axis, providing for full automatic bit changes within the program run.
..
Ok, so I got a lot of work instore for me, and making it on a budget is going to be fun, but Joe said it's a matter of affording the highest quality parts for the most critacale applications.· Joe suggested I spend the extra money, have·lead screws turned for me, as well as matching lead nuts.· Invest in heilicoil (non back lash gears) for the stepper motors, and most important (according to him), was to take the time to do the setup once (ie when making each axis table / slider), keeping it true.· Motor and spindle selection should come next.· Spindle speed and motor power are the issues when Milling.· THe motor needs to have enough power to keep the spindle speed constant while cutting.· Not nessassarly high speed, but the power to maintain that speed is important.· Lastly, the selection of bits is very important.· His parting words on bit selection was, "Use the right bit for the right job, and dont rush the bit and everything turns out fine."
When we were done,·Joe showed me a board they mill when calibrating the mills (something they do every so often).·
The borad has concentric circles, .001 mm on center, with .001 mm clearance.· I needed a magnafying glass to see the traces!· it was awesome.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
My education was about 80% SAE, 20% Metric. My employer was bought out 4 years ago. I started 5 1/2 years ago, using english (SAE) units, then have since used metric. My concensus is that metric is BY FAR the best way to go, and the US should be darned for listening to the English (though they aren't completely the same with that stinkin "stone" measurement!). Also the reason why we lost that rocket a while back. Base 10, that is all I have to say about metric.
1 micron resolution is probably 100 times more resolution that you would need, if you were only going to make homebrew PCB's. Also, I would recommend that you look into companies like THK (www.thk.com/online_cat/catalog_ball_screw.html), SKF, Danaher, or even MSC/McMaster Carr for lead screws. They are not that expensive. As a rule with that kinda stuff, if it exists out there, it's likely cheaper bought off the shelf, unless you have free machine time... which rules!!!
Bottom line, you should stay focused, cause what Sid has should be more than enough accurate for PCB's.
-Parsko
BTW- Your buddy Joe has some nice stuff! But he probably spent 50-100k, while you will easily get 100 times less accurate, for 1000 times less cost!
Thanks parsko for the link and the comments.
I sent a email off to Joe about cost of the Mills, when he replies I'll post it here.
Any idea how much resolution I'll need for the project?
Sid: The issue about double sided boards, was it the registeration alone or were there other factors?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
If you can make them fit, I would recommend ballscrews on at least the X- and Y-axes. These offer the least backlash and require the least driving torque. The Z-axis is less critical, since the weight of the motor and headstock tend to keep backlash to a minimum. If ballscrews are not possible, at least use a screw with an Acme (square) thread. Machine (V-shaped) threads are fine for fasteners, but not for leadscrews. Differential wear on a machine thread happens more quickly, leaving you with differing amounts of backlash over different parts of the screw -- usually with the worst occurring in the center. (I know, because I used a stepper-retrofitted Sherline mill with 1/4-20 machine screw leads for years, and it was a total pain to keep in adjustment.) If you can keep the amount of backlash constant, you don't need to eliminate all of it if it makes your leads too tight to turn. What's left can be compensated for programmatically by adding extra steps whenever a motor changes directions.
When selecting a stepper motor, pick one with a low voltage winding. It will have the lowest inductance and respond the quickest to your drive pulses. A motor rated at 2A and 5V will probably work well. Make sure to pick a four lead (bipolar) model over a six lead (unipolar) one. These will give you the most torque per volume, since there are no idle windings. Shop around in the surplus catalogs for the best values. There's no reason to pay retail for stepper motors.
Use a 24V, current-regulated dual full-bridge chopper drive for each motor. The high voltage will force current into the motor winding more quickly from zero, giving you more torque and a higher top-end slewing speed. Half-, quarter-, or micro-stepping is a plus, since it helps to reduce resonances that will kill your torque at certain speeds. And if you're doing linear or circular interpolation, it's nearly impossible to avoid speeds at which resonance occurs. (I had to resort to a felt drag washer on the Sherline to cut the resonant chatter.)
Another thing that helps achieve higher slew speeds is ramping. By smoothly transitioning the stepper from a low to a high step rate, you can achieve faster speeds under light loads than you can by starting with your running step rate. Ramping will also help you avoid missed steps if you're not using encoder feedback.
If you direct-drive your leadscrews, use a flexible coupling. Helical is probably the best for precision work, although some composite types may also work well. Without one, you'll grind your bearings to dust in no time.
Anyway, that's my brain dump on retrofitting milling machines. Several years ago, I broke down and bought a used Light Machines CNC mill and scrapped the Sherline I'd used since 1981. I couldn't be happier. But I wouldn't trade the experience I got from doing it myself those many years for anything!
Good luck with your projects!
-Phil
Post Edited (Phil Pilgrim (PhiPi)) : 7/19/2006 8:16:34 PM GMT
I have a serious problem with my new Propeller board - some please help!
I hooked up the Demo board, and it displayed my start menu just fine, and it responds to the keyboard.· After about three hours, I determined that a problem existed with this portion of my program:
if dir =="L" ········· if ((dir == "L") and (dirx ==XRight)) ··········· !outa[noparse][[/noparse]dirx] ··········· pulsePin(stepPin, XBacklash) ········· ········· travel := coord - xpos···· 'travel to get there ········· term.str(string("Going left ")) ········· term.dec(travel) ········· term.str(string(" to ")) ········· term.dec(coord) ········· term.out(13)
Everything is fine up to this point - the screen displays "Going left 100"
The screen never displays "Done" so I know it is hanging up on the floating point math somewhere.· Does anyone have a clue?· Is there anything on the Demo board that would inhibit the floatmath object?
I am using the original version of floatmath, no revisions.
This same program works perfectly on my old Propeller board that I built.
The MF70 has next to no backlash in the leadnut because it is delrin and a tight fit on the screw. The backlash there is comes from the end float of the leadscrew, this can be arrested by adding some thrust bearings for next to no money. The accuracy of the leadscrew in terms of lead error etc is very good, these are turned screws and with a lead of only 1mm even using only half stepping you have a resolution of 0.0025mm or 25microns which sounds cooler. The spindle on the proxxon is very good, the collets are three-slotted and hardened (unlike dremel rubbish) and go up to 1/8th which is common for PCB tooling, the speed of 25krpm is more than adequate for the speeds the machine is capable of.
So although your friend knows his stuff you should try to keep in perspective what you will use the machine for and exactly how good it REALLY needs to be. It is easy to drift off on a feature creeping adventure but try to keep it real.
Also remember that although a ballscrew might be cheap enough it then needs mounting blocks and bearing retainers etc etc
on point #1 the number of steps/mm travel or the number of mm/step are just constants put in to the control program, there is no advantage to having one rotation give a certain number of steps.
on point #2 you can also remove backlash electronically although this is only true for light cutting loads strictly as cutting forces may be able to move the table and nut within its backlash. Backlash compensation just means taking up the slack, something all manual lathe operators will be very familiar with.
Sid, i know you've invested a lot of time in floating point. i started to go that route when doing some math for my sawmill head positioning algorythms, but in the end i decided to multiply everything by 10_000 to avoid the swap.. with 32 bits we have lots of room, (at least i do) and i dont drop back to actual counts/inch ( of course i'm only keeping track to 1/32, .03125 inch ) untill i am totally done with math, but there is extra resolution there that i am not using. it just seems easier
dan
Dan, I'm just about ready to go that route myself.· The floating point works fine on the Prop board I built but doesn't want to work with the Demo board.· The mill can go left, right, in, out, up or down.· The Demo board works fine with up or down but locks up on the other four directions.· I would like to use the Demo because I have the VGA working great, and my board only has TV output.· Right now I'm back to my old board and everything works great.· Going to call Parallax tomorrow and see if they can check this new Demo board out.
Graham:
I can see exactly what your saying. My friend (the one who's interested in the project) and I are looking at everything you and everyone else has said.
Joe spend over 70K on their mill, and it's being 'retro fitted' by the end of the year.
Phil, thanks as well.
I'v learned quite a lot over the past few days about milling, and now it's time for me to put together a project plan and cost.
Everyone, thank you loads for the information and help.
I'v got more to post, but I'm waiting on information.
KK
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
Mike, I'm doing pretty well with the propeller, but I have run into a problem I can not solve.· I have two methods for writing Xpos and Ypos to memory:
PRI WriteWord(location,contents) · WriteByte(location+1,contents >> 8) · WriteByte(location,contents & $FF)
PRI ReadWord(location) · return ReadByte(location+1) << 8 | ReadByte(location)
I have two other methods that call the above:
PRI changeXpos(delta) · xpos += delta · WriteWord(@xpos,xpos)
PRI changeYpos(delta) · ypos += delta · WriteWord(@ypos,ypos)
Everything works when I use the Propeller board I built.· I just received my Propeller Demo board yesterday and was trying to run the same program.· The only way I can get the program to run on the Demo board is by commenting out the two "WriteWord" instructions in the above.· Of course, I am no longer writing the current positions to EEPROM.· Remember, this program works perfectly as written on my old Propeller board.
Any suggestions?· My I2CObject is V1.2 by James Burrows.· My old board has pullup resistors on SDA and SCL, but I do not see these on the schematic of the Demo board.
Maybe it's the I2C setup on the demo board?
If I recall correctly, the demo board does NOT have the pull up's on it.
THeres a I2C object with modifiers for either with or withou the pull ups.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
KK, I spent about two hours yesterday really fine-tuning my backlash compensation.· Ran a test pattern each and there appeared to be absolutely no overshoot or undershoot.· I also made a new slider for the X table stepper.· A photo of the old and the new is attached.· Reduced the slider base contact surface from 16 square inches to 4 square inches.· Works much quieter and smoother now.
So........I decided to tray making a double-sided board again.· Simple layout with a 16-pin DIP, which two of the DIP pins going to isolated pads.· Did the bottom side, then flipped it over, did the top and drilled all the holes.· Turned out quite good - All the holes in the DIP pads were just where they should be on both sides, as were the holes in the two isolated sides.· Now I will have to try something more ambitious.· Laying out the second side is quite a chore -· everything has to be reversed and you have to be very careful laying out the traces and locating all the holes.· The holes are not PTH, so you can't have traces from a DIP on both sides of the board since a DIP has to be soldered on the bottom side.· You can have traces on both sides from a resistor and a cap and such - you just have to remember to solder both sides of the board.· The same with vias.· You can have as many as you want, but you have to stick a bit of wire through the hole and solder both sides of the board.
Sid:
Sweet deak, and congrats on the backlash reduction and the double sided board. Good show. Your explnation of double sided board creation revield to me something I hadn't thought of, soldereing on both sides. It's interesting how the little things like that excape you when you are first thinking of a project.
The new slider looks awesome.
my copy of expresspcb does export to dfx (well, it has the option on the screen and it makes a file). So from there you convert the dfx into gcode using ace. Got that. Now, for the hard part. I think I know the answer, but I'll ask it anyway. GCode assumes some smarts in the controller, like the ability to create arc's and circles. Is this the main reason why you didn't code a gcode interpiter directly?
Anyway, ^5's on the accomplishment.
KK
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
Comments
Thanks for the input.
2+ Amps?? Wow, thats a lot of load. And at 165 each... I take it it Mills rather fast.
Would you consider your mill to be of a "professional" grade, vs a "hobbiest" grade?
Graham: THanks for the links. I'm checking them out as we speak.
I can see how this is going to get expensive real fast...
Thanks guys...
BTW: Is there a link or site that actually gives the GCode spec and it's contents? If I'm going to author a GCODE based Mill programm, I'll need to know how to interpet the incomming GCode...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
KK
·
Mind putting up pictures of your mill setups? I'd like to conquor this myself in the (far far far) future.
-Parsko
SuperMill does a pretty good job, but I don't think it could be called "professional grade".· The main problem is the backlash.· The MF-70 has about a 2 to 3 mil backlash and I have compensating routines for that.· The other problem is that on the X table it takes 51.56 pulses to travel 1 mil and on the Y table it takes 20.3 pulses to travel 1 mil.· The Stamp can not handle decimals so on short travels there is always a bit of an error.· I'm hoping the Propeller program, with its floating-point math, will really improve the accuracy.
Regarding gcodes, to create a gcode you must have your etch pattern in a .dxf file, which my Expresspcb can not do.· After you have the .dxf file it must be converted to gcode using something like the ACE converter.
So.........I write my own etching programs.· When I am doing a new board, I draw each etch path, then write down the ending coordinate and whether it is the X or Y direction.· This data is copied into an Excel template which generates a DATA list the Stamp can read.· This DATA list is the pasted into a Stamp conversion program which converts the Excel DATA into a proper etching DATA list that my mill program can read.· Only two directions are generated by the conversion program -
L(eft) and I(n).· Then the mill program reads the DATA list, it compares the coordinate value with the current position.· If the coordinate is greater it travels Left or In a distance equal to coord - posx (or posy).· If the coordinate is less than the current position it travels R(ight) or O(ut) a distance equal to posx (or posy) - coord.· A bit time-consuming sometimes, but the system works very well.
Sid
Sid
Well, it sounds like you have a system, somewhat long, but working.
While I have been reading this (and the home brew milling) threads, I was thinking about going directly to GCode, but seems problemmatic (Ie no expresspcb support).
And doing GCode is going to take some time, but I do have an alternate idea that will probally work much easier... and better.
Take a screen shot of the board in expresspcb and save it as a 8bit color BMP .
My application will open the BMP and generate some form of code for the Mill to directly run. (Code very simular to your code, if not the same.)
The idea is to get "Human Hands" out of the data stream.
My current approch is to start at 0,0, read across the x, lift the z when coper is needed, drop the z when it's not.
Hit the end of that X line, move to the next y line, and return back to X:=0, lifting and dropping the Z as needed.
Now all this "encoding" could be placed into "data" statements for your mill.
It would need to have a "backlash setting", but that would be mininal due to the reduced +- movment of the axis.
There would be some editing of BMP, adding/ using colors so to 'dimple' the through holes, and other extra things.
Well, this part is just an idea. My buddy and I are very serious at this project.
He's got thoughts on handling backlash, and I'll get a chance to talk to him later on tonight about it, but his words were "I'v got the back lash beat"... more on that later.
What I'll do is suggest his ideas to you two for your "feel" on it (you both know a whole lot more then him or I).
My goal with this project is to be able to produce double sided PBC's within the same spec range as expresspc does... without going out of house or chemicales.
I'm not sure if that feasable or not, but I'll get that information / idea once Joe (my buddy) decides on the hardware purchase (it's his money..)
I like your setup Newzed, but I'm somewhat confused why you don't have all three steppers in the exact same setup, thus advoiding the mill dravel / pulse conversions.
Parsko: I've only just started the research aspect of this. I'v no hardware as of yet...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
KK
·
revolutions.· With the controller half-stepping it takes 400 pulses for one evlution, so for .254 revolutions you would want to pulse the stepper
.254*400 = 101.5 pulses.· For a travel of 25 mils, you would pulse the stepper .025/.0394*400 = 253.8 times.· You can't do a fractional pulse so you're always going to drop the decimal, which always introduces a certain amount of error.· On my mill, for the Y travel it is 10.15 pulses per mill; for the X table it is 51.56 pulses per mil.· With the Stamp I get as close as possible with the " */ " operator.· On the Propeller, with the floatmath object, I hope to improve the accuracy.· And there are mechanical considerations - attaching the stepper to the handwheel (or to the leadscrew shaft) so that it is perfect alignment.· Not trying to discourage you by any means -·it took me about four months to get SuperMill to produce an acceptable single-sided PC board, and I'm still working on improvements.· Double-sided is much tougher, even with perfect registration.· I tried to make one, using the shanks of broken drills for registration pins.· It was just barely acceptable.· I haven't tried again since, but I may once I get the Prop hooked up to the mill.· We'll see.
I've redone the stepper slider assembly for the X table - I'll post a picture of it tomorrow.
Sid
Good points... Very good points. I haddn't thought of the actual pulse to step ratios.
I take it half stepping is better off the the qtr stepping as listed?
I'v got some serious math and thinking infront of me...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
KK
·
Sid
Sid
when considering microstepping realize that you can't necessarily make the mill step a single microstep as there is insufficient torque, what it does give you is smoother motion and less noise, the system can start to sound more like it has servos.
Graham
http://www.allsurplus.net/CNC/
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Mike
quoted earlier
"If I'm going to author a GCODE based Mill programm, I'll need to know how to interpet the incomming GCode..."
If you haven't already looked at this then it may be worth a glance to get some ideas for Gcode generation:
PCBGcode for Eagle at http://groups.yahoo.com/group/pcb-gcode/
I have used it to easily export Gcode from Eagle, then make double sided boards with no problem at all, traces in between pads etc,. It does take a few practice runs. There is a free Eagle demo, with limited pcb size though, but fully functional to try the gcode generator.
You can download Mach3 (Gcode machine output software) free here at :
http://www.machsupport.com/artsoft/index/index.htm
This will allow you to import some of your gcode, then test it in simulation mode even without a machine hooked up.
I used to use this PCBgcode plugin with Eagle before I burnt out on homespun boards. Now I send the Eagle board file striaght to pcbexpress.com. If you get into Eagle, you can run the same file on your mill or send it out, same files either way, no gerber stuff to deal with. The drag is the cost of Eagle versus the free softwares.
Just currious why you guys are having backlash issues, aren't you using anti backlash nuts?
I am using LIN engineering steppers with encoders, they are quite robust and take a lot of abuse milling aluminum, even stainless sometimes. Another company that I am starting to get some motors in from is US Digital, you guys might like that site as well. Extremely helpful tech guys.
While his point of view was mostly from the working end of the process, he did have a lot to say about the various mills he's used.
Here are some of his points of "view".· I should preface this with the following.· On the drive to his shop, Joe was very VERY explicit about measuring.· Everything they mill starts in mm, stayes in mm and ends in mm.· THe machines, the tools, the software, EVERYTHING is done in mm.
#1: The lead screw's thread count, and the driving gear train from the stepper all must workout to a basic uint of measurment.
He Explains that on his "best" mill, the lead screw's pitch is quite simple: 1 * 360 = 1 mm of travel.
There's gearing on the stepper to make the stepper's rotation· =· 1 * 360 = 3600 steps.
Now, at this point, stepping is a direct·function of speed, not measurment.
#2: About backlash.· Joe explained to me what backlash was:· The effective slack between the lead screw and the lead nut.
There are quite a few answers·from reduce to·illimate backlash.· Simple ones (backlash nuts & washers) only reduce·backlash to a·stable and non changing measurment.· Elabrate answeres move the backlash measurment below the resolution of the mill.· I asked him to clairfy this statement, and added that each and every mill has it's resolution.· The point at which·changes in measurments·have no effect on the milling process.
An example would be if a Mill had a 1 mm resolution, that means that measurments of .9 and lower have no effect in a stright cut.
His production mill has a measured backlash of .0001 mm, well below the size of the smallest bit used in the machine.· The machine is rated at 0.001 mm resolution. (that's real small, but·neither he or I could tell·where that was in the PCB world.)·
As long as all movments are expressed as truncated numbers above this measurment, backlash is non existant.·
The machine I was shown has·an anti-back lash system on all three axis.· Air over Spring. Something like one of the double stroked hand air pumps (it pumps air in both the up and down directions) but much better, under a constant·pressure of {He said some number and I can't remember it but it like·50 cm^3}· The air cylinder is attached to the lead nut, directly below the lead screw with very heavy duty springs on both directions.
#3: Power.· Joe showed me the controll box for three of the Mills in his shop.· 2 of the three have very distinctive 24V @ 2.5 amp lables.· THe other power supply, the one attached to his production (best) machine as 24V @ 3.0 amps.· Electrally there is no differance between the steppers and power supplys on any of the machines.· THe·difference between the machies is the production has a chuck cradle with 15 bits as the 4th axis, providing for full automatic bit changes within the program run.
..
Ok, so I got a lot of work instore for me, and making it on a budget is going to be fun, but Joe said it's a matter of affording the highest quality parts for the most critacale applications.· Joe suggested I spend the extra money, have·lead screws turned for me, as well as matching lead nuts.· Invest in heilicoil (non back lash gears) for the stepper motors, and most important (according to him), was to take the time to do the setup once (ie when making each axis table / slider), keeping it true.· Motor and spindle selection should come next.· Spindle speed and motor power are the issues when Milling.· THe motor needs to have enough power to keep the spindle speed constant while cutting.· Not nessassarly high speed, but the power to maintain that speed is important.· Lastly, the selection of bits is very important.· His parting words on bit selection was, "Use the right bit for the right job, and dont rush the bit and everything turns out fine."
When we were done,·Joe showed me a board they mill when calibrating the mills (something they do every so often).·
The borad has concentric circles, .001 mm on center, with .001 mm clearance.· I needed a magnafying glass to see the traces!· it was awesome.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
KK
·
My education was about 80% SAE, 20% Metric. My employer was bought out 4 years ago. I started 5 1/2 years ago, using english (SAE) units, then have since used metric. My concensus is that metric is BY FAR the best way to go, and the US should be darned for listening to the English (though they aren't completely the same with that stinkin "stone" measurement!). Also the reason why we lost that rocket a while back. Base 10, that is all I have to say about metric.
1 micron resolution is probably 100 times more resolution that you would need, if you were only going to make homebrew PCB's. Also, I would recommend that you look into companies like THK (www.thk.com/online_cat/catalog_ball_screw.html), SKF, Danaher, or even MSC/McMaster Carr for lead screws. They are not that expensive. As a rule with that kinda stuff, if it exists out there, it's likely cheaper bought off the shelf, unless you have free machine time... which rules!!!
Bottom line, you should stay focused, cause what Sid has should be more than enough accurate for PCB's.
-Parsko
BTW- Your buddy Joe has some nice stuff! But he probably spent 50-100k, while you will easily get 100 times less accurate, for 1000 times less cost!
Post Edited (parsko) : 7/19/2006 3:50:40 PM GMT
I sent a email off to Joe about cost of the Mills, when he replies I'll post it here.
Any idea how much resolution I'll need for the project?
Sid: The issue about double sided boards, was it the registeration alone or were there other factors?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
Sid
When selecting a stepper motor, pick one with a low voltage winding. It will have the lowest inductance and respond the quickest to your drive pulses. A motor rated at 2A and 5V will probably work well. Make sure to pick a four lead (bipolar) model over a six lead (unipolar) one. These will give you the most torque per volume, since there are no idle windings. Shop around in the surplus catalogs for the best values. There's no reason to pay retail for stepper motors.
Use a 24V, current-regulated dual full-bridge chopper drive for each motor. The high voltage will force current into the motor winding more quickly from zero, giving you more torque and a higher top-end slewing speed. Half-, quarter-, or micro-stepping is a plus, since it helps to reduce resonances that will kill your torque at certain speeds. And if you're doing linear or circular interpolation, it's nearly impossible to avoid speeds at which resonance occurs. (I had to resort to a felt drag washer on the Sherline to cut the resonant chatter.)
Another thing that helps achieve higher slew speeds is ramping. By smoothly transitioning the stepper from a low to a high step rate, you can achieve faster speeds under light loads than you can by starting with your running step rate. Ramping will also help you avoid missed steps if you're not using encoder feedback.
If you direct-drive your leadscrews, use a flexible coupling. Helical is probably the best for precision work, although some composite types may also work well. Without one, you'll grind your bearings to dust in no time.
Anyway, that's my brain dump on retrofitting milling machines. Several years ago, I broke down and bought a used Light Machines CNC mill and scrapped the Sherline I'd used since 1981. I couldn't be happier. But I wouldn't trade the experience I got from doing it myself those many years for anything!
Good luck with your projects!
-Phil
Post Edited (Phil Pilgrim (PhiPi)) : 7/19/2006 8:16:34 PM GMT
I hooked up the Demo board, and it displayed my start menu just fine, and it responds to the keyboard.· After about three hours, I determined that a problem existed with this portion of my program:
if dir =="L"
········· if ((dir == "L") and (dirx ==XRight))
··········· !outa[noparse][[/noparse]dirx]
··········· pulsePin(stepPin, XBacklash)
·········
········· travel := coord - xpos···· 'travel to get there
········· term.str(string("Going left "))
········· term.dec(travel)
········· term.str(string(" to "))
········· term.dec(coord)
········· term.out(13)
Everything is fine up to this point - the screen displays "Going left 100"
Here is the balance of that sequence:
········· posit := travel
········· travel := fp.ffloat(travel)
········· pulsetrain := fp.fround(fp.fmul(travel, XPulsesPerMil))
········· stepPin := stpx
········· directionPin := dirx
········· stepDirection := XLeft
········· pulsePin (stpx, pulsetrain)
········· changeXpos(posit)
········· term.str(string("Done",13))
The screen never displays "Done" so I know it is hanging up on the floating point math somewhere.· Does anyone have a clue?· Is there anything on the Demo board that would inhibit the floatmath object?
I am using the original version of floatmath, no revisions.
This same program works perfectly on my old Propeller board that I built.
Sid
···
The MF70 has next to no backlash in the leadnut because it is delrin and a tight fit on the screw. The backlash there is comes from the end float of the leadscrew, this can be arrested by adding some thrust bearings for next to no money. The accuracy of the leadscrew in terms of lead error etc is very good, these are turned screws and with a lead of only 1mm even using only half stepping you have a resolution of 0.0025mm or 25microns which sounds cooler. The spindle on the proxxon is very good, the collets are three-slotted and hardened (unlike dremel rubbish) and go up to 1/8th which is common for PCB tooling, the speed of 25krpm is more than adequate for the speeds the machine is capable of.
So although your friend knows his stuff you should try to keep in perspective what you will use the machine for and exactly how good it REALLY needs to be. It is easy to drift off on a feature creeping adventure but try to keep it real.
Also remember that although a ballscrew might be cheap enough it then needs mounting blocks and bearing retainers etc etc
on point #1 the number of steps/mm travel or the number of mm/step are just constants put in to the control program, there is no advantage to having one rotation give a certain number of steps.
on point #2 you can also remove backlash electronically although this is only true for light cutting loads strictly as cutting forces may be able to move the table and nut within its backlash. Backlash compensation just means taking up the slack, something all manual lathe operators will be very familiar with.
Graham
dan
Sid
I can see exactly what your saying. My friend (the one who's interested in the project) and I are looking at everything you and everyone else has said.
Joe spend over 70K on their mill, and it's being 'retro fitted' by the end of the year.
Phil, thanks as well.
I'v learned quite a lot over the past few days about milling, and now it's time for me to put together a project plan and cost.
Everyone, thank you loads for the information and help.
I'v got more to post, but I'm waiting on information.
KK
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
PRI WriteWord(location,contents)
· WriteByte(location+1,contents >> 8)
· WriteByte(location,contents & $FF)
PRI ReadWord(location)
· return ReadByte(location+1) << 8 | ReadByte(location)
I have two other methods that call the above:
PRI changeXpos(delta)
· xpos += delta
· WriteWord(@xpos,xpos)
PRI changeYpos(delta)
· ypos += delta
· WriteWord(@ypos,ypos)
Everything works when I use the Propeller board I built.· I just received my Propeller Demo board yesterday and was trying to run the same program.· The only way I can get the program to run on the Demo board is by commenting out the two "WriteWord" instructions in the above.· Of course, I am no longer writing the current positions to EEPROM.· Remember, this program works perfectly as written on my old Propeller board.
Any suggestions?· My I2CObject is V1.2 by James Burrows.· My old board has pullup resistors on SDA and SCL, but I do not see these on the schematic of the Demo board.
Sid
·
If I recall correctly, the demo board does NOT have the pull up's on it.
THeres a I2C object with modifiers for either with or withou the pull ups.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
Sid
http://forums.parallax.com/showthread.php?p=596482
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
Does anyone know where I can download a copy of I2cObject V1.3?
Sid
It at the end of this thread (the first thread ref's this thread, http://forums.parallax.com/forums/default.aspx?f=25&m=135261 )
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·
So........I decided to tray making a double-sided board again.· Simple layout with a 16-pin DIP, which two of the DIP pins going to isolated pads.· Did the bottom side, then flipped it over, did the top and drilled all the holes.· Turned out quite good - All the holes in the DIP pads were just where they should be on both sides, as were the holes in the two isolated sides.· Now I will have to try something more ambitious.· Laying out the second side is quite a chore -· everything has to be reversed and you have to be very careful laying out the traces and locating all the holes.· The holes are not PTH, so you can't have traces from a DIP on both sides of the board since a DIP has to be soldered on the bottom side.· You can have traces on both sides from a resistor and a cap and such - you just have to remember to solder both sides of the board.· The same with vias.· You can have as many as you want, but you have to stick a bit of wire through the hole and solder both sides of the board.
Sid
Sweet deak, and congrats on the backlash reduction and the double sided board. Good show. Your explnation of double sided board creation revield to me something I hadn't thought of, soldereing on both sides. It's interesting how the little things like that excape you when you are first thinking of a project.
The new slider looks awesome.
my copy of expresspcb does export to dfx (well, it has the option on the screen and it makes a file). So from there you convert the dfx into gcode using ace. Got that. Now, for the hard part. I think I know the answer, but I'll ask it anyway. GCode assumes some smarts in the controller, like the ability to create arc's and circles. Is this the main reason why you didn't code a gcode interpiter directly?
Anyway, ^5's on the accomplishment.
KK
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
·