As my original thread was quickly becoming a Project Log rather than a seeking help thread I decided to move it over here.
Here's my project (quoted from my original thread here):
Originally Posted by djem
Hi, I've been looking into this diy CNC stuff now for about a year and I feel ready to start building my first one. I like to think that I have collected enough information and stayed interested long enough that this will be something that I'll actually finish some day (hopefully in less that a year).
Having some spare time today I decided to sit down and make some quick modeling of what I've got in my head. I want this router to be small enough to transport in a normal size car (since I don't have a permanent home atm). I also want this to be a cheap build as I don't expect my first attempt to be a keeper. For these reasons I decided to go for a working area of about 300x200x70mm and use a dremel and not some fancy brushless system. I decided to go for the slide system shown at bildyourcnc.com (unless you know of something better for the same price).
I intend to make all the electronics involved myself using microcontrollers. This will be very good exercise for me as I am an electronics engineering student atm.
So, what do you think? Do you see anything that could be done better?
Do note that these are sketches, not work plans.
Today I received a small stepping motor from China that I'm going to use when developing the controlling circuits. It's only 0.4A so I should be able to run this directly from a ULN2803 darlington array. It's got the standard 1.8deg step angle and has 6 wires so I should be able to use this exactly as a bigger motor (I think..), the only thing I need to step up to a bigger motor would be some bigger transistors and a current limiting circuit.
I won't get to do much today though, my atmega chip isn't due to arrive until tomorrow. I will try to make all the software able to fit a 20Mhz atmega88, but I did order atmega16 MCUs too in case I want some more features and I run out of cycles or flash memory. All this isn't quite clear in my head yet as I don't have much knowledge of how fast the interrupts has to be with all the kind of step modes there is.
Much to learn I have.
The postman arrived with my atmega88 chip and a AVRISP programmer today, joy. I connected up my stepper to the mcu and started programming some short lines to try to move the motor. That was no problem. I managed to do full and half stepping without problems in unipolar mode. I was not able to communicate with the computer though as I haven't got a LPT port on my computer. I guess I have to get an old computer dedicated for that as this one doesn't have PCI slots to spare. I did not try bipolar nor microstepping today, that will be something I'll have to do tomorrow. The project is going smoothly for now.
Building the moving parts of the machine has unfortunately been delayed until week 42, but I'll do as much of the electronics as I can manage until then.
I have to wait a little before doing bipolar stepping. It turns out that I did not have all the parts I needed for making an H-bridge (missing the diodes). I have however tried making the stepper do microstepps. That turns out to be a lot harder to do than I had originally anticipated. Those small avr chips has some complicated timers built in to them and it took me a whole day to get a hang of them (have to use timers to do PWM).
As the AVR chips don't have more than 6 PWM channels and a unipolar motor need 4, I though I'd play with the Idea of software based PWMs. I would like to control all 3 stepper motors from the same chip. Doing 256 steps per step I reached a limit for the chip at about 100rpm. This is not theoretical, but It's the result from a test I did with a short loop that gives the PWM loop signal to move to the next step. In a real life application this signal would be given from a input pin. This is quite disappointing results. From what I understand most steppers do 3000rpm at max, thats about 30 times faster that I achieved. Lowering the steps from 256 to 1/8th steps would give me the speed I need, but with very little to spare. Doing three motors then would not be possible. This was with the on board oscillator running at 8MHz, going for an outboard oscillator would give me a maximum of 20MHz. Thats 2.5 times the speed. Still not enough for 3 motors. So it seams I either have to optimize my code a lot, or I have to trash the Idea of using only one driver for all motors and do one driver per motor instead. One driver per motor would be much easier to do also, and I could optimize my code a lot more than I would be able to when having to work with all the variables included in doing all motors at once.
So the conclusion is that I will go for one driver per motor (one mcu per motor in other words). I will try to make it full speed capable at 1/16 micro stepping. And to work on 16MHz oscillator (Then I can use the cheaper chips at about $2 a piece). And full hardware PWM control.
So far my estimated price per driver in components are ~$10 (6 Amps capable). This might end up as a low cost solution after all. (not counting all the reusable equipment witch makes it about $150 to make one, and $10 more for every additional driver I'll make)
So far I can see that I'm reaching my goal. One of the main reasons for doing the electronics myself instead of buying it was to learn. Three days ago I had never programmed a AVR MCU using C (did it once using basic back in school 6 years ago) and today I can do timers . One week ago I didn't know what an H bridge was, still haven't build one, but I've successfully simulated one using Proteus. And lastly, never touched a stepper motor until I got the the one from china earlier this week. So far I'm happy with my project. I'm learning a lot.
Update:
Miner setback. I managed to blow out my voltage regulator, thats bad. Somehow blowing out the regulator resulted in the regulator spewing out the full voltage from my supply (12V) right into the circuit. Only luck (and possibly some builtin failsafes) saved my programmer, just hoping it didn't fry the microcontroller, won't be able to test it until next week. When I get the power up and running again I'll incorporate a fuse so this doesn't happen again.
On the good side I managed to do 1/4 steps using hardware PWM and the results was uplifting. I was almost finished with a 1/4 microstepping driver that would work with parallel port signal. At least I think it would work, I didn't get to try it and I fried the regulator while trying to work out why my interrupt wouldn't fire.
Here's the unfinished code if anyone want to see it. It's not optimized in any way, but I don't think I'll do more with the code after I get the interrupt working. I'm not going to do unipolar and this code only works on unipolar.
Hi - You have taken on a lot in pursuit of knowledge - that is good.
IMHO, pcb routing is just about the most challenging thing people do with a small DIY router. The reason, is that copper is kind of a pain to mill, and if the end mill tip starts vibrating at all, it can lift the copper foil right off of the board.
I would be tempted to consider moving the entire gantry down so that the Z motion is only about 1 - 2 cm to start with. You can still leave it so that the whole thing can be moved up, but just consider to make this very small to start with.
I would also consider a slight design change, especially since you are only loading small parts. Consider instead of long, moving uprights on each side, make the entire fixed sides of the X axis taller, so that the Y gantry rides right on the top of the X axis. I probably am not explaining this well, so I will try to find a picture.
Imagine a box that is the X and Y size you want, with 300mm high "sides". The table is the bottom of the box, and the gantry rides on the top of the fixed sides instead of on "risers". It is a lot stiffer design, just a bit harder to load if you have large sheets of material, which yours will not.
edit - ok, I found an example - www.mechmate.com I am not saying to build a mechmate or to mimick its rather complex gantry assembly, but take a look at how the gantry runs on the raised sides. I believe this is a very stiff concept, and perhaps easier to transport than the existing drawing you have. (nice animations btw - how did you do it? )
Thanks for the info harryn. I didn't think that PCB would be so difficult, I have never tried milling PCB so it's good to hear from someone who has. Luckily I haven't start building the machine yet so there's time to do a design change. I can see how the design you are suggesting should be a lot more stable, but it looks like it might be just a little bit more challenging too. I will definitively take a crack at making a model in Inventor. I have also though about doing moving table too, but I though I could sacrifice the stability for the size. Your idea could probably make it even smaller.
I made the Z motion bigger than needed because I thought I could use it as an all purpose machine, but I'm starting to think I need to build this one only for PCB and make an other one for multi purpose. I will try to do a stress analysis on the models when I've got both too see how big the difference is. Unfortunately I can't use the data from the stress analysis for anything other that to prove witch design is best as Inventor can't simulate MDF or even multiple materials at once. Had my design been in steel or aluminum it wouldn't be a problem, but I don't have the machines nor the knowledge too machine any of them with the precision needed.
It's a good time to start drawing now as my new voltage regulator haven't arrived yet. The new regulator claimes it self as indestructible, and I ordered 10 of them so I shouldn't be without a power source again anytime soon.
There are a lot of different opinions about materials choices, some have to do with performance, others with cost, others with longer term stability.
MDF
- popular for being cheap and actually is quite accurately made
- good for dampening vibration
- very heavy
- not at all rugged and is easily damaged
- I don't use it, but others have - honestly - don't use it.
Plywood
- medium priced
- easy to make into parts
- warps more than I expected, but might be ok for a smaller machine
- My second choice for building things, but I use it a lot in my machine anyway
Aluminum
- I was surprised how much easier Al is to work with than I thought it would be.
- In many ways, you can think of it as a really hard wood - quite different from steel IMHO.
- Can be machined really with the same power tools you use for wood, just clamp them really well and use carbide tipped tools.
- If you use 6mm thick Al, it is really quite stiff and is thick enough to tap, which is quite handy
- This would be my suggestion for your build
Steel
- Quite good, but I would not bother, as it is a little too hard to make things from if you don't have steel machining equipment.
As far as PCB milling - I have not done it, but I did job some out. There actually is a section of the forum on pcb milling - just keep scrolling down.
I don't think I would build 2 machines, it just might be that you do a "setup" for short work like pcbs and a different "setup" for somewhat taller items. One machine should be fine - just make sure you make things really stiff. 6mm thick parts should be your minium, and don't hesitate to make something even 12mm thick - you would be surprised how much things flex under load.
Perhaps some C channel or I beam shape Al would work for you as the main frame.
Since you need it portable anyway, don't hesitate to setup your system right on the floor or perhaps sitting on 25mm thick MDF board to have a flat surface. Floors are very rigid and don' t move around much.
I did some modeling, but I'm having great difficulties making everything small. It kind of seams like this design doesn't combine very well with the size of the ball bearing slides compared to the small area it has to move on. At this point I can't get it to work without making the design more complex than I'd like on a first build. You wouldn't happen to have a picture of something smaller with that design?
In the meanwhile I will try to redesign it with moving table and lower Z axis. I'm kind of in a hurry right now as the final design have to be ready by friday.
Thanks harryn for more info. I won't do aluminum this time around since I don't want to mix in too many unknown elements, also price is of an essence as I'm only a student and have to live off a student load. With no scrapyards close by aluminum would be an expensive material. If this project turns out to be useless for PCB milling I'll just sell it off to some RC freak and do the old fashion etching, it's much more the fun of building that I do this than the wish for a PCB mill.
Didn't notice the PCB section until now. Oh well, it doesn't have a build log sub category anyway so me talking to myself might be better off here anyways.
Did a small change to the model to see how it looks like with moving table, It's a lot bigger, but might work, just have to figure out a few more things before making a good model (this one is just my way of sketching).
Last edited by djem; 10-06-2009 at 05:24 AM.
Reason: wrong picture
Hi - fixed gantry designs are really stiff and IMHO, make for an excellent pcb router approach. A fixed gantry is what I am attempting to build as well, just a larger model. As you have noticed, size can be a challenge as you need to deal with the movement of the table. By the time you are finished with the size things really are, it might not make that much difference.
Just as another point of reference to the design I was trying to describe before, here is a thread on one under construction.
The advantage of this approach over a conventional moving gantry (your first picture) is less flex for similar materials and components. It is also potentially smaller than a similar capability fixed gantry.
The downside is that you are limited to loading from the ends. On a machine like yours, it is a non issue. For large machines, if you try to load from the side, you get covered in grease. In order to keep it really stiff, the bearings that carry the gantry must be widely spaced - not more than 1/2 the distance of height they are spread.
If you are going to stay with a wood based design, I would probably stay with plywood then, or perhaps just some dimension lumber (I don't know what they are called there - here we call them 2x4s, 2x6s, and 2x8s (rough size in inches) If you build the frame sides from some dry wood cross sections of approx 40 x 200 or 300mm, that should work well. Perhaps reinforce the corners with some angle aluminum.
By the way - is the software you are using for those drawings easy to use?
Thanks
HarryN
I use Autodesk Inventor 2009
I don't know how easy it is to learn on your own as I learned it in school, but I find it very easy to use and better than most other software I know of. The downside is the price of the software. As I'm still a student I get a free educational version directly from Autodesk, isn't that nice? It's also very easy to get on torrent pages, but that might not be legal (Even if you know you could never afford the software it's not nice to steel it. Autodesk wouldn't know the difference, but the software developer you would buy a cheaper CAD software from would. Just a though, not a statement).
LinkBack URL
About LinkBacks
Originally Posted by djem 
. One week ago I didn't know what an H bridge was, still haven't build one, but I've successfully simulated one using Proteus. And lastly, never touched a stepper motor until I got the the one from china earlier this week. So far I'm happy with my project. I'm learning a lot.
