CNC Notes

Painful Lessons

gholbrook@digitalintuition.com — Thu, 08 Apr 2010 01:46:44 GMT

I had today to myself in the ye olde shoppe, so I decided to work on some CNC projects. One of the projects that has been stalled for quite a while is the creation of a tool plate for my X3 mill. For this project I purchased a 3/4” thick piece of aluminum. The idea behind the thickness was that I could anchor the edges with short screws so that no part of the fasteners holding the plate to the table would poke out above the top of the plate. Interesting idea, but probably not work the effort. First, it’s really really annoying to try cutting that size chunk of material.

While trimming the material, I broke one bandsaw blade and two 1/8” endmills. First I tried the bandsaw, but I got too eager and broke the blade. It seemed like a good plan to try my mill instead. That worked fine until the endmill made it to a depth of about 1/3”, at which time the cutter was unable to evacuate chips. After two broken cutters, and a failed attempt with a jigsaw, I put a new blade on the bandsaw and took my time cutting through the material. Worked fine.

By the way, VRML extra cool bonus points. Download and install this viewer, then click this link to view the file in 3D. Be sure to push the tilde (~) key to switch to rotation mode. The VRML file is large, so it may take a moment to download.

In Home Digital Machinist I saw an article where the author discussed using a drill bit to remove material. His rationale was that drill bits are much better equipped for removing material. The drawing above shows the toolplate. It is 6”x12”X.75” with a .75” drill pattern throughout and .75”x.25” edges at each end. In the illustration above, the red columns represent drilling to remove material with a 1/2” bit.

<\/param><\/embed><\/object><\/div>";" alt="">

Drilling to remove material worked really well. Unfortunately I got overconfident and choose my cutting parameters poorly. I used a 1/2” endmill to pocket the ends and then took a 1/8” deep cut. This ended up causing a lot of trouble. The spindle tried stalling, and the machine chattered quite a bit. I finally broke down and created a new NC program to use a 1/4” endmill. By that time, the steppers were pretty warm. Eventually I heard the telltale sound of missed steps. I’m not sure exactly what the reason was. The irony is that I had already attached the fittings for feedback encoders on the X and Y axes. Had I chosen to finish that project first, the toolplate would have gone a bit smoother. Once I suspected that I was missing steps, I shut everything down and called it quits for the day.

Lessons learned…drilling is good. Large cuts in aluminum, not good, even with extra material removed. Next time I’ll go with a faster program using a smaller cutter and shallow cuts.

Journey

gholbrook@digitalintuition.com — Fri, 26 Mar 2010 21:04:15 GMT

In the fall of 2009 I had this brilliant idea. I wanted to be able to CNC cut aluminum without damaging my machine. Since my machine at the time was a MaxNC, making cuts in aluminum or mild steel was simply out of the question. It could be done, but it couldn’t be done well. That’s when I got the bright idea to buy an X3 and a conversion kit from CNC Fusion.

When I bought the kit, I had no idea how ballscrews work. I didn’t even know what a ball nut was, and I certainly had no idea how to repack one of them. Seven months later I have learned all of these things and a good deal more.

After taking my mill apart and putting it together about a dozen times, everything finally seems to be in good working order. The machine is completely reassembled, and my workshop is slowly getting organized again. In the picture below you can see my X3, and a small MaxNC on the other side of it. The CNC controller for the X3 is wall mounted behind it.

Today I made my first real attempt at cutting aluminum, thanks to a material donation from my nephew. It’s a simple single pass cut using a 1/8” cutter at a depth of 0.05” and a feed rate of 2”/minute.

If I had a spindle tachometer and axis encoders, I would probably step up the speed a bit. The cut turned out really well. The edges are crisp and the finish is smooth.

Inauguration

gholbrook@digitalintuition.com — Mon, 15 Mar 2010 03:18:25 GMT

Recently our office formed a soccer team. I’ve never played soccer but I joined anyway. So far it’s been a lot of fun, but it started reminding me of the hockey team I played on last summer. We had fun, but scored very few goals. In our first soccer game we didn’t score at all. In our second game we still lost, but we scored twice. To be precise, one of us scored twice. She was the hero of the day, all the more because she isn’t a regular player. I decided that Mo deserved to gloat a bit, but in order to gloat…you need a good conversation starter. Like, a trophy! It would be really great if we had an MVP trophy to pass around. I started designing the trophy. I really wanted “MVP” written around a “ball”. Getting this done turned out to be a bit more work than I expected. TurboCAD Pro can write text on a path, however I only have the “Average Joe” edition. The difference in cost is about $1000, so an upgrade was out of the question. Instead I started with InkScape. Inkscape will create text on a path, so I created a circle and put the text on it. From Inkscape, I exported the text as a DXF and pulled it into TurboCAD. From there I was able to tweak it and finally pull it into CAMBam.

I wanted the base of the trophy to be round, and to have “SET SEG Strikers” written on it. Finally, I wanted to test out my newer mill for this project. The working area on the mill is only about six inches in Y, so I needed to fit the text in a relatively small area. It’s also difficult to get the sizing right, because I really wanted to use a 1/8” endmill. In retrospect, I should have sized the text smaller and used a 1/16” cutter for the circle.

The MVP cutout is the piece that I was most concerned about, however it turned out really well. Apparently my concern was not warranted. Toward the end of the cut it started to chatter loose, so I held it down by hand to keep if from getting too dinged up.

Oddly enough it was the base that ended up giving me some trouble. The letters ran together a bit, which I should have expected…since they cut out exactly the way that they looked on the screen. One thing that amused me is that people who do their own CNC conversions sometimes refer to a “circle test”. This is no “test” at all. If your machine cannot cut a reasonable circle, something is seriously wrong. Still, I was pleasantly surprised to see how nicely the top circle on the pedestal turned out. It wasn’t until the program started cutting out the base of the disc that I grew concerned. It was chattering quite a bit, so I tried to hold it down. That turned out to be a bad idea because I inadvertently started leaning on the table…and I didn’t notice until it missed a couple of steps. It the spindle were running at the proper speed, and I had added the encoders to the motors this wouldn’t have been a problem. I didn’t check the spindle or wire up the encoders yet, though. The motor missed a couple of X axis steps and it shows on the finished product as a bump in the circle.

All in all, I’m pretty happy with the results. I definitely need to get off my duff and add the extra electronics to the mill. The accuracy on all the cuts (in terms of backlash and repeatability) was really great.

Less Lash

gholbrook@digitalintuition.com — Mon, 15 Feb 2010 00:01:00 GMT

I’m waiting for the parts to break in (I may need to polish the ends of the screws where they meet the helical couplers), but the preliminary results are in. After fashioning some parts and putting the X axis on my retrofitted mill back together, the results are quite a bit better than they were initially. I’ve now got double ballnuts on the X axis, along with zero backlash helical couplers.

1/1000 of backlash on the X axis. I don’t actually have the helical coupler screwed in because the initial fit was so tight, so it is possible that it’s causing a wee bit of slip (I doubt it though). 1/1000 is way better than the initial 7.5/1000 that I was getting before though. Later this week I will write a CNC program to run the axis back and forth a couple hundred times…then I’ll recheck the thrust bearings and measure backlash again.

Lash B Gone

gholbrook@digitalintuition.com — Thu, 07 Jan 2010 03:22:49 GMT

After a couple month hiatus, I’ve finally returned to working on my mill. About two months ago I tested the backlash and was sorely disappointed. Both X and Y axes were in relatively poor shape. I got back in touch with the company that I ordered the original parts from and ordered extra ball nuts for the X and Z axes and oversized balls for the Y axis. Along with the balls and nuts, I also ordered HeliCal couplers (zero backlash) and new, higher quality thrust bearings.

So, I got started the other night. Yet again, I found myself carefully repacking ball bearings one at a time. There were actually extra balls in the kit, which is good, since I have a demonstrated ability at losing balls. When I started packing the balls I actually said to myself “I should do this sitting at a table, with my magnetic bowl.” Of course, that isn’t what I did. And I lost a ball as a result.

Having gotten the Y axis repacked with oversized balls, I was pleased to notice a significant reduction in “slop” when I moved the nut around on the screw. Tonight I took on the X axis double ball nuts. When I unpacked the parts I noticed a set screw and wondered what it was for. There turned out to be two set screws, one hiding the other. Too bad it took so long to figure out!

The reason for the set screw became clear once I threaded the second nut onto the screw. There are three sets of threads that need to be synchronized: the ball screw, the left size of the nut adapter, and the right side of the nut adapter. The set screw helps fine tune the threads together, which makes it possible to set the tension to whatever you please. With a little work, I was able to get the double nuts set almost perfectly. They don’t roll quite as easily, but there is absolutely no noticeable movement when moving the screw.

Take a Lashing

gholbrook@digitalintuition.com — Sat, 26 Sep 2009 04:23:10 GMT

First, fun stuff. I cut this on my smaller mill. In the end, the top was pink with black lettering. The best way that I found to paint it was by machining it first, then priming it liberally. After the primer dried, I liberally dabbed black ink into the lettering (at .1” depth) and sponged the excess paint from the top. After the paint dried, I used a roller to roll a couple coats of pink over the top. I wasn’t smart enough to take a picture of the final plaque, but it turned out really well. It’s about 6” wide, and I cut it with a 1/16 endmill. The material is oak, although pine seemed to work just as well.

The big excitement for me this week has been putting those finishing touches on my new mill. Part of what’s left to be done it to check accuracy and make alterations where necessary. This requires a dial indicator. These buggers are interesting, if you’ve never played with one. The travel on them varies with accuracy. If you want more precision, you get less travel. The one I bought was only $25, so it probably isn’t terribly accurate. You can see it in the picture below.

As I started testing accuracy on the Y axis, I noticed that when I programmed 1”, I was actually getting a lot less. The error seemed consistent, though, so I bumped up the number of steps per inch. That worked just fine, except that I got different results for long and short distances. For a bit, I started to wonder if there was something wrong with my cheap dial indicator. When I was nearly about to give up, my friend ~~G-Smack~~ GC happened to call. While we were chatting, he suggested that I should verify that the results were consistent if I traveled 1/10” at a time. The first 1/10 worked fine. The second, third, fourth, and fifth 1/10 increments were each over by the same amount.

Can you spot the problem? Last night, based on the results of distance travel to 9/10”, I thought the machine had very little Y axis backlash. I was wrong, though. It actually has quite a bit (0.0075”). Since I was moving consistently back and forth, the backlash was self-compensating. Once I programmed Mach to compensate for backlash, and reset the steps per inch to the base number, I started getting very consistent travel. The dial indicator showed that travel might be over or under by 1/1000”, but I think that might actually be the indicator. The X axis backlash was even worse, at 0.014”. That, I believe is caused by a loose thrust bearing. When I get paid again, I’ll probably spring for a set of Heli Cal zero backlash couplers. Fortunately, as long as the backlash is consistent, Mach does a great job compensating. If I get really crazy, I might also buy some better ball nuts later on too.

The following video is just a quick demo of using the indicator on the mill. The numbers you see are thousandths of an inch. You can see the needle start on 0, travel 1/4”, and then return solidly back to 0.

<\/param><\/embed><\/object><\/div>";" alt="">

I’m curious what I’ll find when I put the indicator on my MaxNC mill. It uses solid couplers (no backlash) and leadscrews (very, very little backlash vs cheap ballscrews). My guess is that the machine has almost no backlash, but does need to have steps per inch tweaked a bit.

First Chips!

gholbrook@digitalintuition.com — Tue, 22 Sep 2009 10:59:40 GMT

Ok, I’ve been anxious not to stall on this project. Momentum is key! Over the weekend I got all three axis wired up, and I put the electronics in this neat enclosure. Notice the monster heat sink, and the cooling fan.

Last night I finally finished rudimentary software tuning. When I told the Z axis to move 1”, it did so. Since X and Y are configured the same, I could hardly just stop there. All three motors seemed pretty happy, so I decided to try cutting something really simple.

<\/param><\/embed><\/object><\/div>";" alt="">

First, I need to mention that the ball screws and nuts that I installed have a relatively high backlash. If I remember, it’s somewhere on the order of .005”. That’s quite a bit. When I cut out a 2” circle, the actual inside diameter was about 1.9”. If you think about the number of direction changes involved with a single axis while cutting a circle, that’s probably about right. Fortunately this can be calibrated and adjusted in software. Later this week a dial indicator should arrive and I’ll use that to measure and adjust for backlash.

Update: No, it isn’t right. I just took a shower after writing that and realized something really funny. 0.1” seemed like a suspiciously accurate error, so I double checked the gcode that I ran. The script I used to generate my circle uses a roughing pass. When I made the above video, I stopped the program after the roughing pass. The roughing pass used a radius of .825. Add on to that 1/2 of the cutter diameter of 1/4” and you get .95 radius. Double that to get diameter and you get a 1.9” circle. I’ll run the whole program tomorrow and see how it does.

In other news, I noticed that the X axis motor was not holding very well. I could grab the coupler and turn it by hand. Some investigation revealed that VREF was way too low on the board, but it’s also turned up all the way. Z and A also seem to have the same problem. I’ll check into it more tomorrow. All in all, things look pretty good so far!

Axis of Evil, Unperturbed.

gholbrook@digitalintuition.com — Wed, 16 Sep 2009 02:08:51 GMT

This evening I had only a few spare moments, so I raced downstairs and plugged in my soldering iron. Within a minute or so the tip was hot enough to melt solder, and I tinned the ends of the stepper wire that I soldered to the Y axis motor last night. With the motor attached to the table, and the table back on the mill, I was eager to see how it would do!

<\/param><\/embed><\/object><\/div>";" alt="">

So, the motor did just fine. I should be able to run it at 3 amps, but the driver started getting uncomfortably hot. No big deal. I dialed down the power a bit and ran through the paces. In and out fast, then slow. I ran the table all the way to the column, then all the way back out. I slowed as I neared the end of the ball screw that caused me so much grief at the beginning of this project.

My solution to the “problem of the lost bearings” was to drill and tap a hole at the end of the ball screw. I used the hole to affix a washer to the end of the screw. Brilliant, no? Since I was testing the machine I decided to make sure the screw would stop it.

Perhaps you wondered about this too, if you read the description of my solution. The Y axis stepper motor is rated at 425 inch ounces, and the table weighs about 40 lbs. I inched the machine closer to the end of the axis, and then, with no hesitation whatsoever, the machine pushed the retaining screw out of the way and continued barreling toward the end of the shaft.

Brilliant!

Z’s Alive!

gholbrook@digitalintuition.com — Mon, 14 Sep 2009 03:07:28 GMT

In an earlier post, I briefly discussed building the Hobby CNC driver board. In the following photo, you can see the board on my bench while I test it with the power supply and tune in the vref for current control.

By some fluke, the entire kit ‘n kaboodle worked on the very first try! Now, Monstro The Stepper is NOT a fast motor. My initial attempt to run it at a reasonable speed failed quite conclusively. At very low speeds, it does just peachy. While testing, I noted that the stepper driver got very warm. When I came back and did some testing later I used a clothesline clip to fasten a piece of aluminum to the driver. You can see it in the following video. The chip seemed to remain quite cool with this very small amount of heat sink. Just to be safe, I plan on putting large heat sinks on all four drivers.

<\/param><\/embed><\/object><\/div>";" alt="">

Once I recovered from the initial shock of seeing the motor moved, I wasted little time affixing it to the mill. It took a while to solder the hookup wire and get the motor physically attached, however it didn’t have any trouble driving the Z axis without any load. Swell! The motor worked great without a load! Now, the real test would be to attach the head and try again. If I were to guess, I would say the head weighs between 60 and 80 lbs. Getting it wrestled into place is generally not pleasant, but since I was very excited it seemed a lot easier this time. The results can be seen in the following video.

<\/param><\/embed><\/object><\/div>";" alt="">

Now, time for a couple of interesting notes. First, the motor was originally set to a very low speed. Stepper motors have much more power at low speeds. Even at the low original speed, the motor sometimes stalled when coming out of low current idle mode. I slowed it down even more to counter this effect and it worked very well. I’m happy with the results for two reasons. First, the speed is still reasonable. Remember that movement along the Z axis is normally limited to less than an inch so slow isn’t a big deal. Second, I’ve got a gas spring kit that I have not installed yet. If I want, I can install the spring and then increase the stepper speed. I may not do this, however, since heavy is usually better for milling. The spring would actually take away force that the machine can use when plunging.

In the above picture, you can see that the stepper motor is completely covered by the factor electronics enclosure. Also, note the block of wood under the quill. I had a very pleasant and unexpected surprise when I turned off the power to the stepper driver. The head didn’t crash! It would not have been a surprise for the head of the machine to crash once power to the motor was turned off because steppers only have a strong hold while powered. Even when I leaned on the head, it didn’t move. This means that I can use the machine like a drill press when I’m not milling something. A drill press with a 1” capacity! If this works out, I may even put my current drill press in storage. If you’ve seen the Secret Laboratory, you know exactly how much space it doesn’t have.

Made in the USA, sort of.

gholbrook@digitalintuition.com — Sun, 13 Sep 2009 03:25:19 GMT

This evening while Aimee was making fidgets this evening, I took the time to assemble my Hobby CNC Pro 4 Axis control board. This board comes as an olde tyme electronic kit. It’s a circuit board and a bag with about 80 parts in it. The rest is up to the person doing the assembly.

You might be thinking something to the effect of “Let me get this straight, you have dumped an absurd amount of money into this project, but then you skimp on the driver board?” Well, yes, and no. Some of the other drivers\controllers are very nice. Gecko comes to mind. Gecko is also very expensive. Instead of paying $80, I would have paid $700 or so.

There is a really good reason to buy a Hobby CNC board, though. If you happen to burn up a driver chip, you can desolder the chip and replace it with a new one for $10. Because of the user assembled nature of the board, it’s a bit more forgiving. It lacks any onboard intelligence, however I still have the option of upgrading to a Gecko or KFlop controller for about $300 later. In this scenario I would still be able to use the Hobby CNC board as a power driver, still at a fraction of the cost of one of its competitors.

So, perhaps “Made in the USA” is a stretch, but it was assembled in the USA at the very least.