CNC Cookbook: Blog: Feb 09 through Jul 09

First There Was "Mini-Me". Now, Say Hello to "Mini-Haas". Guldberg's KX-3 Enclosure and Slant Bed Lathe Projects

Great build thread by a Danish man named Guldberg over on CNCZone. It's a CNC'd Sieg KX-3 with an extremely nice enclosure:

Back from Uncle's machine shop. Construction is welded plate. Linear ways. Surfaces were trued on a big CNC horizontal mill.

3-Jaw clamped to mill to machine the locking plates for the lathe's tool turret...

Here is the turret. Air cylinder unlocks the turret by spreading the teeth. A stepper rotates the next tool into position. Belleville's apply pressure to lock the teeth...

Some of the projects you see out there are just amazing. Guldberg is definitely doing first rate work and having a lot of fun at it. He's got me thinking about an enclosure for my IH mill now. Sure would be nice to keep the chips up off the floor!

There is a massive and typically inflammatory thread about indexable tooling and small machines over on the HSM board. If you can get through all the posturing (maybe just look at the pictures?), there is some good information there. I summarized it in a post as follows:

The catalogs and the posters here attribute a lot more to those 4 letters in an insert designation than I can read into that spec which seems a lot more related to making sure inserts are interchangeable with their holders than anything else. For all the pages here, there are really very few points being made that matter for making chips:

1. Positive rake requires less cutting force and is generally better for lighter machines. In fact, positive rakes are taking over from negative even for heavier machines in many cases because the geometry cuts better. You can see that reading through the PM board to see what those guys use/recommend. Negative rake is principally useful for durability, but as the positives get better at interrupted and other "difficult" conditions, why bother with negative?

2. The meaning of the various letters in an insert designation is pretty prosaic. Some things we can determine from it but most we can't. We don't know the rake unless we factor in the toolholder and the top surface of the insert. Those two are actually not called out very well by ANSI. Therefore, we have to understand our toolholders and the meaning of positive rake and visualize what will happen with a particular insert. Those "sharp" inserts are clearly very much going to have positive rake. Many other inserts it isn't so clear.

3. Since you can't really tell from just the 4 ANSI letters what's going on, you'd better have one or more of the following in hand before buying the insert (unless you just want to experiment):

- Full ID on the insert so you can go consult the manufacturer's catalogs. This is often hard on eBay.

- A big picture of the insert and enough practical knowledge (more than enough in this thread) to guess how it will cut.

- A solid recommendation for the insert from someone doing similar work on a similar machine.

- Help from a rep picking out your inserts. Clearly YMMV depending on how good the rep is. This is why peeps like the "Exkenna" guy over on PM so much, or Frank Mari. Their advice has prooved out.

Just as an example, not long ago the CCGT inserts were the ticket for smaller lathes. Peeps thought the "G" meant "Sharp", but it was only the tolerance. It wasn't long before manufacturers were selling "G"'s that weren't sharp because there was demand to pay the higher price. Then I started seeing CCMT's that were Sharp. The whole ANSI business ceased being a useful determiner of anything other than whether the insert would fit my toolholders. Hence the 4 criteria above.

4. The idea of a lathe "too lightweight" for carbide is an interesting one. A small Southbend will clearly handle carbide. My Lathemaster 9x30 does too. The ubiquitous 9x20 is noticeably less rigid than these, but with the common mods, seems like it would work. What then is "too lightweight"? Unimats? 7x14's? For the hobbyist, carbide is a matter of personal taste more than anything. Do you want to spend your time trying to understand the minutiae of insert selection (feels like stamp collecting sometimes), or grinding HSS tools? Either takes away from making chips, but I like my carbides though I also do some HSS.

Last point: Be careful if you have an indexable tooling fetish as I do not to accumulate tools that use too many different insert types, or inserts of exotic design. It's just too hard to keep up with it all. Hence all my mill tooling uses APKT (or equivalent) and all my lathe tooling except the boring bars and parting tool uses CCMT.

Apparently cheap toolholders look like such a deal until you're paying $10 and up per insert on a facemill with 7 inserts. Ouch!

I managed to get a couple of projects completed over the long weekend. First was a tooling rack to make it easy to organize my tooling:

I been playing with Turner's Cubes for a little while now. This one is my first CNC'd Turner's Cube:

Powered drawbars are extremely handy things. They make tool changes a snap. I can vouch personally for how great the one I made from an impact wrench is. There are basically two versions. The first, and most common, is like my impact wrench version. It literally simulates the normal operation of an R8 drawbar, and just automates the tightening and loosening along with a little "gentle persuasion" ala the mallet tap that comes in the form of the impact wrenches hammering as well as downward pressure from the air cylinder.

The second version is less commonly seen, and attempts to function more like powered drawbars in commercial VMC's. This one simply uses a stack of Bellville washers to tension the drawbar, and an air cylinder to release that tension for a tool change. The tool is held in an R8 collet that stays quasi-permanently attached to the drawbar. When it is pulled into the taper, it tightens. When pushed out, it releases. This system is a little simpler that the impact wrench in its basic form, and gives a faster release cycle. Hoss built one for his X2 mill:

As you can see from the video, it works really well and is very slick. A lot of folks, having seen this style from Hoss, decided to build them for larger mills. This is where the problems start. The X2 doesn't use a lot of horsepower. Larger mills like Bridgeports and RF-42's like my Industrial Hobbies, can have considerably more horsepower. Let's say up to about 3 HP. With larger tooling and more aggressive cutting, there have been problems with the tooling pulling out of the collet. The hobby crowd has basically decided this is just a matter of applying more drawbar force, but there is more at work here than meets the eye.

Consider that the numerous folks who have manufactured R8 powered drawbar systems would probably love to have a purely pneumatic system so they can reduce their costs, but for the most part, such systems are unavailable. Did they all just fail to build one with strong enough air cylinder and linkage. Not likely.

If we take a look at one of the few systems that did see the light of day, the Mach1 system, there are some interesting learnings to be had. They say their system only uses a 600lb die spring, for example. That's hardly any force. There are reports of tools slipping even with 2500 lbs of pull force. The builder, Scott (Poppabear), comments in that thread about his ATC project for the Tormach milling machine. He tried 2500lbs, and could not take a 1" axial cut with a 3/8" cutter without it pulling out. Ultimately Scott decided the project was not feasible.

Is it really just a matter of more force on the drawbar? What if Scott had used 4000lb? Color me skeptical. Drawbar tensions on machines designed for ATC's, such as CAT40, are typically under 4000lb. BTW, these huge tensions need to be applied in such a way that the force is not transferred to your precision spindle bearings with potentially disasterous results. Ray L. did a great job on that with his design.

Why is the R8 taper so problematic? In looking over the Mach1, I figured out the secret: there are really two issues to consider. First is locking the tool holder to the mill. That's going to be a function of the surface area of the R8 taper and the pull force. I am not too surprised that 600 lbs suffices for that force, even with quite a lot of "work" being done by the spindle.

But there is a second force, and the way it works is hugely counterproductive to the first. That second force is the squeeze on a collet to hold the tool. It doesn't exist with solid tool holders, and it is the reason the Mach1 system uses a special R8 collet closer. Note that Mach1 can also use solid R8 tooling. The special collet closer serves 3 purposes:

1. It's threaded cap compresses the collet on the tool with a lot more than 600lbs of force. I have read somewhere that 5C collets use 1500lbs of force, BTW.

2. It creates a reference surface to preserve the Z repeatability against the spindle nose. This gives the Mach1 system the equivalent of the Tormach Tooling System. Note that solid holders are repeatable already to a few tenths. I've measured that myself before.

3. This is really the big secret that tells why the Mach1 system works with only 600lbs of retention. That special collet holder creates a new, more precise, and more rigid R8 male taper. You can see this clearly from the patent illustration:

Patent illustration showing how the collet closer creates a new R8 male taper: #26

Why is this new male taper so important? Because the deformation of the collet as it locks down on the tool really interferes with its ability to make good contact with the R8 taper. If you think of bluing tapers, there is no way in heck that there is much precision in that interface. So now the drawbar force must not only provide sufficient clamping, but it must also combat the reduced surface area and hence friction of the collet in the taper. The Mach1 system avoids all of that.

Folks get started on these air-cylinder only systems because they seem simpler than an impact wrench system. But they're really not unless you're prepared to live with a huge amount of drawbar tension, and even then I wonder how well they are going to work with a facemill or a large silver and deming bit. People keep saying that this has been tried over and over, and it has. The drawbar manufacturers would love a simpler cheaper mechanism, if only one would work. Yet they keep shipping impact wrench based systems for R8, or special patented tricks like Mach1.

What I will tell you is that a rookie machinist can build an impact wrench system in an afternoon and it won't suffer from any of these problems. It can be completely automated for use in an ATC if desired. It's simpler and cheaper. Your biggest challenge for the ATC is that you'll be using solid R8 holders which don't have a standard interface for the ATC carousel. That's no big deal. You'll need to fab some collars for the tooling that serves that purpose. Meanwhile, you will be saving a fortune on TTS holders and you'll have a more reliable and rigid system to boot. Don't take my word for it. Look at what industry does, and look at how many have tried and failed to produce an air cylinder-only system. Note that this only applies to larger mills. Let's say mills with more than 1HP.

As I got looking at how my R8 tooling was being scattered around the various crowded flat surfaces in the shop, it became obvious I needed a better way. So I designed a new tooling rack and set about building it from some 1/2" "poly" plastic I got from US Plastics. I really like how the stuff machines. I didn't quite get finished, but here are some piccys of progress underway:

The rack will sit atop and to the rear of the rolling cart that goes with my mill...

Couldn't resist a bit of a trial, now could I? While all the holes are filled, many of these are seldom used. I expect this will lead to the purchase of more tool holders. I hate when that happens!

I intend to number the holes, and use the numbers to fill in the tool table in Mach3. This way I can swap out a tool and Z will already be calibrated since the R8 holders are very repeatable. Between just grabbing the right number and using my impact wrench-driven drawbar, it'll be pretty danged fast. Eventually I'll build a true automatic toolchanger (ATC), but for now, this will be a big step forward!

I've written quite a bit about how to find the right carbide inserts for small lathes, but not much about indexable milling tools and inserts for small mills. Recently I bought a new Iscar face mill that uses APKT-style inserts:

It's a little 2" face mill, which is about right for the size mill I have. I had been using a Lovejoy facemill, but it uses these SPEX inserts that I've only ever seen used with Lovejoy tooling, so they're expensive. I also have a little Iscar Helimill 5/8" diameter indexable endmill that I really love. It uses APKT inserts and I had heard good things about these inserts in a lot of places such as the PM boards. So, I went looking for a deal on eBay and eventually bought another Iscar mill.

Along the way, I discovered there is an insert type called APET that is a super sharp aluminum cutting insert that fits any APKT cutter. Cool!

The APET is specifically designed for aluminum and has a sharper edge. I haven't had a chance to try the new face mill yet. I'm waiting on an R8 shanked arbor for it, as well as the right project. Full details when I get to try making some chips. If it performs nearly as well as the little indexable end mill, it'll be great.

For curiousity's sake, here is a tool shape recommended for a fly cutter on HSM:

The shape is not unlike the APKT. Perhaps these inserts would make for good fly cutting!

For all those who are skeptical about those crazy toolpaths where the cutter never turns a corner and so can go much faster (see my post below on cutter engagement for more), check out this CNCZone video of a router cutting through steel at up to 120IPM:

Normal these little routers barely have the rigidity for aluminum, let alone steel. Look at it go!

I often collect photos of things I know I'll be doing some day. I call these pages "Idea Notebooks". Here is the collection available today:

Electronics Boxes: How to build a neatly wired enclosure for your CNC electronics.

I squared my mill column a few weekends ago as part of an accurizing process I'm going through on the mill, but I only just took the pix off the camera and processed them today. I used a cylindrical square to measure how far off I was, leveled the table, and then shimmed the mill column to take care of the remaining error until I was just a few tenths off. The details are on my Mill Tips and Techniques page, but here is a teaser picture:

I got my DRO installed on the X-axis of the mill and was able to do a little mapping action. Full details on my Mill Tuneup page, but here are some teaser pix:

The left axis shows the actual move of each commanded 0.5000" move as measured by the DRO. If the ballscrew were perfectly accurate, the graph would be a straight line centered on 0.5000".

You can see the righthand 40% of the ballscrew is qutie a bit more accurate than the left, although the first maybe 10% on the left is quite good too. Nevertheless, the whole screw moves to well under a thousandth of accuracy. You can also see that the errors are not cumulative, but are more periodic. The total error in 24 inches of motion was 5.6 thousandths and the screws are advertised as having less than 3 thou per 12", so this screw is within spec.

Mach 3 has the ability to take a map like that and correct for these errors. I haven't tried that yet, but it would be an entertaining experiment!

I've gotten interested in understanding more about CNC toolpaths lately, and one of the most interesting topics is cutter engagement. Cutter engagement is the fraction of your cutter that is actually doing any cutting. It turns out that this can change quite a lot as your cutter travels through most toolpaths. In particular, it gets markedly worse in corners. This diagram will illustrate:

Cutter Engagement: Blue = Material left behind, Purple = Material being removed by toolpath, Red = Cutters in two stages of engagement

The cutter is moving right to left through a corner as the arrow shows. I've captured the cutter engagement at two positions in red. Note that when moving along a straight wall the cutter has a 90 degree engagement, but when it is buried in the corner the engagement is 180 degrees. That means the cutter suddenly has to work twice as hard when it hits the corner. I've shown a radial depth of cut of 1/2 the cutter diameter, but the same principle applies (albeit the angles will be less) with less extreme depths of cut. Any time we go through a corner like this, our cutter engagement increases.

What does that mean for your speed of machining? Well, to put it simply, something has to give. Normally we run the same feedrate throughout the entire toolpath. Yet the cutter works twice as hard in the corners. So that means we either run a feedrate that is slow enough to do the corners well, and we shortchange the long straights, or we run a feedrate that is fast enough for the straights, but it is way too fast for the corners. In the latter case we get chatter, lousy tool wear, or worse a broken cutter.

Most of the time, we therefore opt for the former. Most all of the recommendations we get from the cutter manufacturer for feeds and speeds assume we will run a constant feedrate and go crashing into corners, so they're conservative relative to the straight line performance the cutter could deliver.

"Ah ha," says the clever machinist. We just need to vary the feedrate based on the engagement angle and we can optimize for faster machining. Yep, that will absolutely work. In fact, it would be pretty straightforward to hand tweak the feedrate on the corners for simple toolpaths. It's tedious work this hand tweaking, but you'll definitely speed up the program. Lots of CAM programs have an option to vary the feedrate automatically as well. How much can we tweak the feedrate? Well, from the illustration, a right angle (90 degree) corner has twice the engagement, so in theory, we can run that corner at the cutter's recommended feeds and speeds, and double the feedrate for the straightline. In practice I would not be so bold, probably opting for more like a 50-75% increase in feedrate on the straight lines. You'll have to try it out and be aware that some cutter breakage is likely while you sort out what works for your particular combination.

Varying the feedrate works, but that is considered Old School these days for a variety of reasons. One of the more obvious is that you'll be able to visible see the different feedrates in the surface finish.

If you don't want to vary feedrate, the latest thinking is that you need to create toolpaths that don't turn corners. What? How is that possible? What if I need a square corner in the pocket I'm cutting?

Don't get me wrong, eventually the cutter will have to follow that corner, but we can do everything in our power to avoid it except where we absolutely must, and then we can do so very gingerly. Lots of approaches have been tried and they're computed by CAM programs in lots of different ways, but in general, the produce a series of arc-like cuts instead of straight line cuts. Imagine something like this:

Imagine the tool following these circular paths as it converges into the lower left corner. A little clean up pass will be need along the edge of the bound to pick off the triangular waste pieces between the arcs, but in general, we've cut a corner with fairly constant corner engagement.

This can be coded up by hand for simple situations. Cutting a rectangular pocket, for example, or a rectangular slot. The results can be pretty amazing. Check out this slot cutting program Geof from CNCZone wrote:

You can see the tool moving in circles as it slices through the slot. The cutter is a 1/2" five flute running at 6000 rpm, 1.24" deep cutting a 3/4" wide slot through 1" hot rolled with a radial depth of cut (stepover) of 0.025". My favorite speed and feed calculator, MEPro, would have suggested 2498 rpm and about 35 IPM, and Geof is able to run 2.5x the spindle speed and over 4x the feedrate!

There is more going on here than meets the eye. I won't bore you with the math, but I have a spreadsheet that calculates the cutter engagement given the diameter of the circle (3/4" for Geof's program), cutter diameter (0.500") and depth of cut (0.025"). In this case, Geof is getting about 23% engagement. It looks like there is a lot of wasted motion on that cutter on the backside of the circle where it isn't cutting, but this motion serves a useful purpose. The cutter is only engaged 23% when cutting. But, it is not even 23% engaged for the whole of the circle. The numbers aren't exactly right, but pretty close if we assume we get the 23% engagement on the front cutting half. That means we have a duty cycle of 1/2 times 23% or an effective 12% engagement. It isn't 12% from the standpoint of cutting force, that's why I refer to it as a duty cycle. Rather, it is 12% from the standpoint of cooling the cutter. So we spend 88% of each loop cutting air to cool the cutter and only 12% effectively cutting.

That cooling is less important with aluminum, but vitally important when cutting steel, which is what Geof is doing.

The latest CAM programs all have toolpaths that are designed to work this way for arbitrarily shaped pockets and in 3D. Looking at how much more performance Geof got on his simple slot cutting, you can imagine where such CAM programs can radically reduce your machining time.

I came across this slick little engraver's vise on the Candlepower Forums (dedicated to building custom flashlights):

I like the use of dowel pins to hold irregular parts. In fact, this is why they're called "peg vises." Available from jewelry supply houses for under $20. Also available from Harbor Freight.

One of the most accurate ways to measure a bore is with a dial bore gage, but these can be really expensive and seldom used tools. I loved this gadget by gbritnell I saw on HMEM to adapt a DTI to the task:

Check out the Videos Page for a couple of new vids as I made a bracket to hold the reader head for the X-axis DRO on my mill. I'm only installing the DRO temporarily. It can read to 0.0002" and I want to use it to wring the last bit of precision out of the mill.

I've been eyeballing ger21's "Aqua" screen set for quite a little while. I even loaded it on my home office PC to play with so I could see whether I'd like it well enough to start using it. There are a lot of things I like better about the screen set than the default 1024.set that comes with Mach3:

However, it was missing a feature or two I really coveted. Most importantly, it was missing the ability to cycle through a list of jog increments. So I set about customizing it. Read more on my page about the project.

One of the issues facing every CNC mill user is telling the CNC software where the end of each tool is. They vary by length of tool. Worse, sometimes they can vary each time a tool is inserted in the machine, even when it is the same tool. If you're plagued with the latter, you have to remeasure tool length every time the tool is changed. What a pain!

Most "professional" CNC machines use spindle tapers that eliminate this problem for you. A CAT40 goes into the machine the same way each and every time no matter what. Your worst challenge is having a chip get caught between spindle and toolholder, or perhaps having the wrong drawbar tension. Unfortunately, the R8 taper has a reputation for not being so accurate.

One answer to the R8 problem is to use a tooling system that indexes off the spindle nose for repeatability. Companies like Tormach sell these tooling systems, and sometimes people make their own equivalents as well. These systems have a round shank that goes into an R8 collet. They have a collar with a shoulder that provides a positive stop against the spindle nose. Release the pressure on the R8 collet, shove a toolholder up until the collar makes contact with the spindle nose, tighten the collet again, and you have repeatability in the Z-axis. You can record the tool offset for that toolholder and it'll be the same the next time you plug the tool in. Much time measuring tool heights is saved as a result!

That's all fine and well when using R8 collets, but I've never especially liked them anyway. I prefer endmill holders and collet chucks. These have a solid R8 shank, so I have a hard time believing their repeatability is significantly less than a system like Tormach's. It's easy to see why collets don't repeat-- they're pulled up into the taper a variable amount until they lock down on whatever shank they're gripping. Differences in shank diameter lead to different Z-lengths. I see this all the time when I use my 5C collets on my lathe.

But a solid endmill holder or collet chuck has no "give". Can't you tighten it up in the taper and expect the same result each time?

Of course I had to try the experiment. So I ran downstairs to the shop, grabbed my HF butterfly impact wrench (I just use it handheld on the CNC mill), and my Z-axis toolsetter:

Checking tool Z-axis height. Mill is covered in chips because I was just making parts...

I established a baseline, and then I replaced the tool 5 times by removing it completely from the spindle, reinserting it, and rechecking the tool height versus the baseline. Repeatability was +/- 0.0005".

Now I have a hard time understanding why people pay so much for the Tormach Tooling System when regular R8 shank tooling seems to repeat so well. If there is more going on here than meets the eye, send me a note. I'd love to learn more.

I got a note from a regular reader who indicates the TTS is all about automatic tool change and being able to change tools a lot faster. Since I have both a powered drawbar I built and a handheld butterfly impact wrench for my second mill, I'm not seeing that particular advantage. It's a matter of seconds with one of those wrenches, and if you build the powered drawbar, it's trivial to set it up to work with air solenoids so it can be completely CNC controlled.

Others are concerned about how the holders can sit in the carousel of an automatic tool changer. One can use the indexing ring on the TTS as the place where the ATC carousel grips the tool. That's a better argument, but you can't guarantee that with stock tooling. You need a little shoulder on the ring for best results. The problem is the diameter and shape of the tooling below the ring may interfere if things aren't properly laid out. Tormach sells holders intended for toolchanger applications that have an additional groove. Getting the tools to sit at a defined height in a carousel is one I have some ideas about. More on that later when I finally get to building the carousel!

There is a great thread on CNCZone about creating hypocycloidal speed reducers. The goal of the thread is to create a very low backlash drive for a rotary table 4th axis on CNC machines. I don't know how successful they are with the backlash, but the designs are fascinating, and several gearboxes have been made:

These are the kinds of projects it would have been impossible for me to imagine someone being able to do in their home workshop. It's just an amazing bit of work, and one that could only be attempted with CNC.

Someone on the thread suggests these hypocycloid gear systems would make a lovely clock if encased in clear plastic. I agree!

I was down in my shop last night fooling around with a part and trying to line up my tool height relative to the workpiece top. I was getting lots of Z-axis faults. Doh!

My first thought was servo tuning, but that didn't make sense. I spent quite a bit of time tuning up the servos earlier and they should still be good. The faults were coming only on very short distance moves (I was jog stepping 0.001" to get an indicator to touch off on the bottom of an endmill). I keep my gibbs pretty tight, and so faults on small moves sounds like sticktion.

I reached over and gave my single shot oiler a few pumps until I saw it come out on the ways. I jogged all 3 axes for some pretty good travel (need to make up a short g-code to do that while I'm pumping and put a button on Mach3 for it), lined up the tool height properly (with no faults), and then did a short dry run on the part to make sure I wouldn't fault again. Man the axes even sounded a lot better with a little way oil.

Nice to have built a one shot oiler for the mill, but I guess I need to make it automatic and electric though!

Lots of folks have asked for video of my IH CNC mill. I've tried to take some video on probably 6 or 8 different occasions. It almost always comes out unwatchable. I have a terrible time hand-holding the camera steady for long enough at points interesting enough in the process. Sometimes my camera also focuses on the wrong thing. What I need is one of those cool Flip HD video cameras mounted on a tripod. Meanwhile, I did finally get this barely watchable video:

Surfacing some 6061 with a 5/8" Helimill indexable endmill. 1600 rpm spindle, 10 IPM, 0.100" depth of cut. Ran very smooth and delivered good surface finish...

I've added a new Videos Page (accessible from the navigation palette top left on every page) where I'll put all my videos. They're also uploaded to YouTube if you want to subscribe to them. So far there is just the one, but I'll try to take more of my projects as people really seem to like videos.

I recently came across the idea of a tramming "key" to be installed in the jaws of your milling vise. The idea is due to John Stevenson and looks like this:

Insert the U-shaped key in your vise jaws, tighten the jaws, press the key against the top T-slot edge, and tight down the vise. Nothing could be faster or simpler!

I like the way this enclosure is set up for a small mill. Looks like full width of the table for the door, and that is acrylic. The sides are for the table travel and are metal. There is an access panel on the ends...

Good view of the interior of the enclosure and other details like the gas springs that counterbalance the head or the nice metal way covers...

A lot of what makes these kind of machines look sexy is just sheet metal. Harbor Freight (and others) sell this one. Wait for one of their 20% off discounts and you've got a steal on a turkey little CNC to get started on. If I add up the costs on my IH CNC mill conversion, I probably spent $2k more than you could buy this turnkey cnc mill for. My mill is more capable, but it was a lot more work before I could cut any chips.

This one is made by CNC Automation, and I came across it on eBay. This would sure help control the chips and coolant flying around:

The rear is rubber. I assume that so there are no costly collisions with the mill head or rear column...

The splash guard just bolts down to the table's T-slots. It also bolts to the front slot (where the limit stops go)...

I wound up buying one after thinking about it. My mill is slinging chips all over my work area. On eBay they're $279. On their web site they're $500. It's designed for a 9x42 table, and a 12" Y-axis travel. My IH mill has a 39.5" x 9.5" table and 12.5" Y travel. I figure that's close enough.

Pretty good thread over on the 'Zone about engraving. It starts at bottom of page here. I just thought I'd add that my own CAD program, Rhino3D, has a font function that makes it easy to take any Windows font and convert it to a curve or solid:

I'm going to have to give that a try some time soon. I captured the supplier links for engraving suppliers and added them to my supplier links page as well. Just search that page for "engraving".

The components. The bracket that keeps the ballnuts from rotating relative to one another replaces the old clamp for the ball return circuits...

The arrows show where the cutting force is attempting to deflect the cutter. The takeaway is that when the accuracy of the wall's location is critical, conventional cutting yields a better result. It deflects the cutter in a direction that is less directly vectored towards or away from the wall. OTOH, it is well known that surface finish is better when climb milling.

Sunday night I got a chance to throw some parts into my vibrating polisher for a little deburring action. Here is the before:

It took about 5 hours and did a nice job taking away the tooling marks and leaving a nice satin finish. Full details on the new Vibratory Deburring and Polishing page.

You need a pair of ballnuts with preloading between them to really get backlash down to a few tenths. My IH mill CNC kit came with these for the X and Y axes, and used Rockford parts. The trouble for hobby conversions is that they're pretty expensive. A preloaded ballnut pair is circa $150 for 0.631" diameter ballscrews. Put these on X and Y and you're looking at $300. OTOH, single ballnuts are available for $22.85 from the same source. Four singles would cost $80, less than 1/3 the cost. Evidently there is considerable value in making up our own preload arrangement!

The issue when doing so is to suitably place some Belleville washers between the two nuts to force them apart with sufficient preload to do the job, and to prevent the ballnuts from rotating relative to one another as any rotation can reduce the initial preload.

Those are the square Rockford ballscrews in 0.631" size. You can see there is a collar threaded onto the lefthand ballnut's mounting threads. It holds the Bellville's which push against an inside lip on the left and the body of the ballnut on the right. So far so good. The bracket on top is attached via 2 of the holes that hold the ball bearing return tubes in place on the ball nut. It keeps the ballnuts from rotating relative to one another. Just one problem unless I'm missing something: the bolts are also keeping the ballnuts a fixed distance apart which prevents the preload from working its magic. The nuts have to be able to "float" along the axis of the ballscrew.

Here is an alternative way to machine that bracket so the lefthand ballnut can move:

A slot allows motion. Ideally we'd use a shoulder bolt on that lefthand side so that we can tight down the bolt and there is a nicely machined shoulder that rides in the slot. I'm still not thrilled with the thin plate, but this design would at least allow the nuts to move along the axis relative to one another without rotating as is desired.

One could also envision a design that uses a dowel pin to slide in and out of a hole in a bracket mounted in the same bolt hole, or even a design that is integrated with the ballnut mount. For example, here is a ballnut mount integral sketch:

In this design, an outrigger from the ballnut mount (green) provides a sliding track for the rear ballnut (ballnuts are red). The Bellville preload assembly (gray) is threaded onto the rear ballnut and bears against the front ballnut.

And here is yet another approach originated by Country Bubba and then followed by Pete from TN on CNCZone:

There is a cylindrical housing for the Belleville pack, and a socket head cap screw goes through one of the holes around the rim to lock the ballnut from rotating once the preload is set. Very simple and elegant design.

What's going on here is amazing. He's built his own toolchanger, powered drawbar, flood cooling system with enclosure, tooling plate on the table, repeatable Z tool holders (like Tormach's tooling system), and a whole bunch of other goodies. Amazing to watch it go through its paces. It's a full on VMC built on a hobby budget from a little Sieg X2 imported mill.

I was exchanging some notes with a professional machinist friend of mine (one of those long-suffering mentors who has devoted too much of their valuable personal time educting me on the obvious things I can't figure out for myself, thank you Peter!) and he wanted to know if I was going to tumble deburr my first CNC parts. From his perspective, they looked like perfect candidates.

I have a little vibratory polisher that I tried out once on some brass for polishing, not deburring. I wasn't impressed with the results, but I do know that this is how it's done in industry and I have wanted to learn more about it. Peter pointed out what kind of media his shop would use, and suggested perhaps my problem earlier had been impatience. He says smaller machines can take a lot longer to do the job and that I might've needed 12 hours or even a little more. I certainly hadn't run it that long, so thought I'd give another try on these aluminum parts. The other difference is that this is deburring and that had been an attempt at actual polishing in lieu of hand buffing.

The recommended media were plastic wedges from C&M Topline. He recommended the smaller wedge shape on this page for my application. Then I started in with my stupid questions, starting with, "Gee, that's a 50lb bag, I'll never use all that media, is there another source?" Enter Grainger and this recommendation. Not exactly the same stuff, but available in 6lb quantity instead of 50 lb.

Of course I had to get some to try! So I went through the checkout, and ordered 12 lbs to start. Imagine my surprise when they gave me free shipping. I didn't have any codes like you'd use with Enco, it just did it. The overall online catalog experience was great too.

I finally got started making parts with the newly CNC'd Industrial Hobbies mill. It was a lot of fun last weekend, and eventually the following parts emerged:

My first CNC parts. The one on the left has an 0.010" finish pass, and the one on the right is just roughed with a 0.050" depth of cut. The parts were profiled with a 3/16" 2 Flute end mill. This photo represents about 3X magnification over actual size. Full details on how I made them are on the Comber Rotary page...

These are bearing blocks for Elmer's Comber Rotary Steam Engine. They're for another HMEM Team Build I am participating in. These are trial runs and not finished parts, although the one on the left was intended to be. Unfortunately, when I measured it, there were a number of dimensional errors amounting to several thousandths in a variety of directions. After wracking my brain quite a lot, I finally mic'd the diameter of the 3/16" end mill. It came out as 0.1837" which is considerably different than the 0.1875" that was expected. That would account for a lot of error! Now I need to adjust either the g-code or Mach-3's tool wear offset to account for that difference and run a new part. I'll check it again, and if that doesn't bring tolerances to acceptible levels I'll keep looking for more things to fix.

It consists of a piece of threaded rod, a micrometer dial, and a pretty typical clamping solution to the way. There is a set screw that locks the stop by bearing on a groove cut in the threaded rod. The edge of the block serves as an elegant but easily read micrometer indicator pointer.

The groove that rides on the lathe ways was cut by placing the block on V-blocks so it rested at an angle and then using an endmill...

Click here to see another fellow make up a bunch of these for his shop classe's South Bend lathes.

That power surge was really painful, but I finally got the mill running again on all 3 axes late last weekend. Turns out I had blown the Smoothstepper and 2 out of the 3 Geckodrives.

Having gotten the 2 drives and Smoothstepper replaced, the worst part was just figuring out what was wrong. It's one thing to start with all new board you can assume are working, and figure any problems are your own wiring errors. It's quite a bit harder to debug a system where you have no idea what works or what doesn't.

Having gotten the axes nominally going, my next task was to tune each axis. Servos have to be tuned. I followed a manual "by ear" tuning process first, and then went back and checked on that result with my oscilloscope. Full details are on my servo tuning page, but here are a few photos I snapped for your enjoyment:

Oscilloscope is connected, but the axis isn't moving yet. I also haven't set everything up or you wouldn't see that trace without a moving axis!

I use the circle pocket wizard's g-code for servo tuning. Set a small diameter circle and a relatively high feed rate and you'll get lots of direction reversals to use for tuning...

I got my mill X and Y axes up to 50 in/sec/sec or 0.13g's acceleration with the o-scope. Without it, I could only get to maybe 40'ish by ear. Z has the heavy mill head, so about half this much acceleration is available...

The X-axis right after o-scope tuning. Full clockwise current, nearly full gain, a little bit less damping. Your tuning settings will definitely be something different!

I did not get a chance yet to see what kind of rapids are possible. My tuning was focused on acceleration as it is a more difficult (and many advise more useful) performance characteristic to optimize. I was pretty happy with the results, but I intend to "detune" (back off slightly) in order to provide a margin for error.

I first started messing with these little butterfly impact wrenches from Harbor Freight when I built a powered drawbar for my mill from one. It worked so well I bought another one of the little wrenches to use for other things. Here is what they look like:

That's the wrench on the right. The air cylinder on the left was the other component of the powered drawbar I made...

One day I was laboriously tapping a bunch of holes by hand and I spied the wrench hanging there. "Isn't there some way to use it for tapping?" I wondered. Low and behold, I came across the following little gizmo from Enco not long after:

I've seen others use cordless drills, but what I like about this wrench is the way it fits in your hand, is easily adjusted for torque via the regulator, and can be reversed with the one touch paddle switch. Tapping sure goes fast with one of these, and I've yet to break a tap. I keep the torque relatively light and the wrench just stalls out before anything too terrible can happen.

Next thing I'm going to do is build a parallelogram linkage to make a tapping arm similar to what John Stevenson shows over on the HSM board:

Disassembled view of the torque limited tapping chuck that came with John's tapping head. The torque is limited because 3 ball bearings in little pockets mate with the tapping head to drive the tap chuck. The balls are held in the pockets by the belville washers. Apply enough torque and the force from the washers is overcome, the balls pop up out of the pockets and nothing more happens. John is concerned the the impact "hammer" action is bad for the taps, but I've had no problems. This little impact wrench doesn't have a lot of guts to screw things up except perhaps on a very small tap. He does suggest it would be possible to disassemble the impact wrench and stop the hammering, and I may look into that at some point.

He's using a small 1.5mm endmill to cut a hex shaped hole in the workpiece rather than broaching. The little Rotozip laminate router is conveniently sized for the application.

These are models of a scratch-built CNC lathe that recently changed hands from S_J_H to rubes as portrayed on HSM:

Gorgeous renderings, eh? These were done by Autodesk's Inventor 2009. I must say, the rendering built into Rhino3D (my CAD program) doesn't do nearly this well. They have a separate rendering program, but I haven't wanted to spend the money. I wish they'd incorporate some nicer rendering in the base product. Clearly their competition has.

Full details are on the CNC conversion home page, but I got first chips cut last weekend. I had expected to be cutting more ambitious chips this weekend, but a couple of mishaps have stalled progress until next weekend. First, I managed to blow up the X-axis servo drive through some over zealous tuning. Then I lost the Smoothstepper to (I think) a storm in the area that must've zapped it with a power surge. I've already got a new Gecko, but the Smoothstepper blew just this weekend, so I won't be able to get a replacement until some time next week.

The mill was working well and I was ready to have some real fun with it this weekend, but it was not to be. Next weekend!

Zoho's joints had to be pretty close to a human being's to create such lifelike poses!

Zoho stands 43 cm tall, weighs 6 kg, has 920 parts, 101 of which are found in each hand. Of the 920 parts, 85 are mobile. Zoho was constructed of bronze and stainless steel.

I had an electrician come by, so I now have 220 for the mill (and a big compressor and a few extra outlets for other things). The next logical step is to mount the spindle head, but i'm stuck until I get some parts I ordered from McMaster-Carr. I need the square head bolts needed to secure the spindle to the Z-axis. They got misplaced somehow from the box of parts that came with the mill. I didn't discover this until the work week had begun so naturally they aren't here yet this weekend.

So, I was casting about for something else to do on the mill in the meanwhile, and I decide to calibrate the X and Y axis. This is not too hard to do and makes a big difference for the accuracy of the mill. Here is a little video that Hoss just published that tells how he went about it:

A fellow was asking recently how to design a model engine so the cylinders could be exactly opposite one another without the con rods interfering. Normally the cylinders on opposing banks of a "V" engine are slightly off so that the con rods can ride side by side on the crank.

Close up of the con rods. Interesting how there is a fork and then one smaller con rod in the middle...

I made an adapter for the servo shaft and got the Z-axis running today. Minor tuning was needed, but the Z runs pretty smoothly.

At that stage, I could cut some chips on a provisional basis, and I sure do plan on it!