March 2011

Things to do to improve the next orrery:
1) Buy better wood.

2) No biscuits when gluing up the wood.

3) Let the machine do the Saturn gear (with smaller bit).

4) Let the machine do the base.

5) Correct the pocket holes on the top.

6) Correct the CAD models of the planet holders and let the machine drill the holes.

7) Let the machine drill the crank plate.

8) Put on finish, then stick on the dial.

9) Try to make the support rods the same wood as the base and top.

10) Make CAD models of the Saturn lantern parts and let the machine do that.

11) Better wire for the planet supports – needs to be stiffer. Lantern gear wire is fine.

12) Didn’t need to polish the shafts – just needed to clean off the price tag gunk.

13) More of an angle on the peg ends.

14) The shaft holes in base and top shouldn’t be through holes – should be blind – will change the shaft length a little.

15) Actually sand it well like you’re supposed to.

16) Check the position of the shaft holes on the base and top – somehow the first one ended up off by a tiny bit which made the gears not level until I stuck a shim on top of the crank plate.

17) Replace crank handle to one that rotates in its socket – makes it easier to turn.

I thought that I needed a nice photo of the orrery.  The ones I had taken were OK but I wanted something more.  Still, it’s a big pain to get an appropriate backdrop, set up lights, etc. but then I realized: when I bought Pinnacle Studio, the box came with a sheet of green screen material to use with the Chroma-Key feature. Well, that’s just what the doctor ordered.

I draped that over a bench in the living room one evening when Mel and Erin were out shopping for prom shoes and then gathered several reading lamps from all over and placed them appropriately.  I put the orrery onto the green screen and snapped away.

I then did some photoshoppage (GIMP actually).  There is a command where you can simply replace one color with another and so when you have a green screen like that you just take out the green and replace it with something much darker to give all the emphasis to the orrery.

Naturally, this did not go perfectly – nothing ever does.  I learned something that all photographers and movie makers already know: that uneven lighting and/or wrinkles and folds in your green screen will cause the resulting image to have not one shade of green but many.  True, they’re all green and I guess the reason that the “green screen” is the outlandish green color that it is is that you almost never encounter that shade of green in normal life.  So in the end, I had to do some extra work to capture all the shades of green that were present but it worked out well anyway.

I did go back and smooth out some wrinkles and even up my lighting to make the color replacement a bit easier.  It also turns out that the orrery itself reflected some green in its surface so that if you replace all the green you end up replacing some of the pixels in the orrery image itself.  In the end I just had to be satisfied with that and not many people notice it.  The image looks way better than the ones where you could see living room stuff in the background and perfection would have taken much longer and probably required more sophisticated equipment. 

Life’s all about balance – you have to decide when something’s good enough.  If you want perfection, you won’t actually get very much done.

My photo looks good.

I made a short video of my orrery.

A better portrait by gregwest98
A better portrait, a photo by gregwest98 on Flickr.

I took a better photo of my orrery tonight. I did something fancy in that I remembered that when I bought Pinnacle software, it came with a sheet of green screen material. I draped this over a bench and photographed it; then used Photoshop to remove the green color. Most of it anyway. I left it sort of gray rather than leaving it totally black. This is a much better photo.

I had not intended to publish any more blog entries on this but it’s probably worth noting a couple of things. 

First, the plans included a dial that you glue to the top that has months, days, and zodiac signs.  This looks pretty cool so after a lot of fiddling to make it print across four sheets of paper on my printer (surprisingly hard to do – for some reason the CAD program thinks the entire drawing covers 12 standard sheets of paper), I put in some cool-looking parchment patterned paper and printed it out in sections.  I cut these out and used some Type 77 spray adhesive to stick them on.  I think that if I had had the appropriate font installed, it would have come out with a vaguely Arabic look but it still looks pretty cool to me with a standard blocky font.

I then set out to put a decent coat of finish on it.  I got out whatever I had which was a can of minwax wipe-on poly.  I had previously tested this on a scrap printout to make sure that the poly did not dissolve the printing from the inkjet.  I wiped on the first coat.

Then I was annoyed to find out that the poly seemed to dissolve the spray adhesive and parts of the dial were lifting up.  After the poly dried, I thought perhaps that shellac might work better but things only got worse.  After things dried really well, the paper flattened back out at which point I took a toothpick and put some wood glue under the edges and stacked quarters, coasters, and whatever flat things I could find on the dial to hold it down.  I guess putting a finish over the dial is not in the cards.  From that point on, I just put finish on the wood.  It looks pretty good now.

Another issue is the gear trains.  This is my fault for being cautious.  The drive gear train rotates as a unit and all gears are glued to each other.  When I initially did this I was very sparing with the glue and furthermore, I only applied glue to one surface which is generally a no-no with wood glue – at least in my experience.  So twice now I’ve had the drive gear train come unglued at one of the spacer joints while cranking it.  In each case I disassemble the thing and reglued the joint with an appropriate amount of glue and with glue on both pieces.  Here’s hoping that this will hold although I’ve only repaired two – there are about 10 more similar joints. 

Also, my planet mounts are pretty lame.  As I mentioned before, I drilled these out by hand rather than letting the CNC machine do it and they just don’t fit like they should.  I have to be careful to make sure that none of these planet holders actually touch any of the others or the friction between them will cause the motion to be off. They drag on each other and usually one gets ‘coupled’ to the other.  If I was certain that I’d never have to take the thing apart again, I would squirt some thin CA glue around the shaft and lock them all into place but for now I am considering just remaking these parts and trying again.

So I can safely say that all the parts that were cut by the CNC machine have worked well; my problems have come in wherever I did some hand work.  It’s not that my hand work is sloppy although that is true in one or two places; it’s that I did something without enough forethought and testing.  In retrospect, this is not surprising. Patience is a virtue; one that I am still trying to nurture. If I had gone back to my CAD models and made them match the drawings I had and if I had waited for my friend to get back to town so that we could use the machine, some of these things would have not happened. 

But then again, there’s value in doing something by hand in order to get a full appreciation for what the machine is doing for you.  When I was in graduate school, the machinist who made our experimental apparatus would occasionally make me do some grunt work “to get my mind right” which is kind of the same thing.  So the effort isn’t wasted.  But my orrery still needs to be fixed.

This whole build is starting to remind me of having a newborn child.  The “design and build” process was super fun but now the daily maintenance has begun and that’s kind of a chore.  I hope I can get the thing running reliably and go into a crank-only mode soon.

This is a list of my orrery build entries.  You can click on the tag “orrery” and get them all but they come up in reverse order.  So, in the proper order:

The first introduction.

An intro to the design.

Gear design.

Thoughts on the planets.

How I made the planets.

Making prototype gears on the shopbot.


The first CNC session.

Hand-cutting the base.

The second long CNC session.



The Saturn Gear.

Completing Saturn.

Sticking on patterns.

Dry Fit.

The Lantern Gear.

The Last Gear – Mercury.


Mounting The Planets.

The Finished Product.

What I Learned About CNC.

I’ve learned a lot from the orrery project.  I’ve learned a lot about how gears are made.  But mostly, I’ve learned how things are different when building things with a CNC mill as opposed to the typical woodworker way.  I guess this parallels the changes that were brought about by power tools like table saws and routers when woodworkers had only been using hand tools before that.

This is part of a generic process that I’ll call “fabrication”.  These days there are two types of computer controlled machines that will fab things for you.  There are the so-called “3D Printers” that squirt out some sort of material in the shape you want.  The simplest of these use a string of ABS plastic that the extruder heats up and squirts out onto a substrate. There are also some types that harden a material with a laser and thus build up your ‘thing’ one line at a time.  Then there are the machines that take a solid block of material and cut away whatever you don’t want.  That’s what I was using.

I guess the first thing is that the terms “CNC mill” and “CNC router” mean almost the same thing.  The first comes from the metalworking world and the second from woodworking.

Anyway, when doing things the CNC way, you tend to try to think of ways to keep all your parts flat since creating anything with curves in three dimensions will complicate your design process by at least a factor of 10 and sometimes it can complicate things at the machine as well (some machines only work in 2 dimensions).  By sticking to flat parts, you can just draw things on the typical CAD screen and then the machine will cut along those lines you drew (after you tell the machine how thick your stock is – more on that later.)

You also have to give some thought to the router bits themselves.  You can use typical woodworkers bits but there are specialized sources that most CNC folks use because of a couple of issues.  In the CNC world, they refer to them as “end mills” (as opposed to “ball mills” which woodworkers would call a round-nosed bit).  If you’re cutting plywood, you tend to prefer a ‘downcut’ spiral bit because it will keep the veneer from lifting up and creating splintered surfaces.  But it will also force the sawdust down into the cut where it will build up if the resulting channel is pretty narrow (and doesn’t go all the way through) and cause heat to build up.  This can, in principle, start a fire and I’m told you can see smoking occasionally. I haven’t but it definitely smells like burning wood sometimes.  So ‘upcut’ bits are preferred for cutting narrow channels or anything that is not a cut all the way through the stock.  You can get the best of both worlds with what is called a ‘compression’ bit which has both upcut and downcut flutes; one set on the end, the other set near the shank. I can’t imagine how these are produced but I suppose they are made by CNC.  They look like something from an M.C. Escher print.

The same bit (or “mill”) does everything: cutting and drilling.  Drilling is interesting: frequently you tell the machine to “peck drill” a hole which is where the router lowers, drills a fraction of an inch deep, then raises back up to clear the chips; repeating this process until the desired depth is achieved.  As this goes on, the machine appears to be pecking at the stock.  If the required hole is larger than the bit then it will spiral out from the center until your hole has the required depth and diameter.

You’ve really got to think hard to come up with any sort of joint that you can make with a CNC machine that looks nice like the traditional woodworking joints.  It’s hard to be original and so at the moment, CNC milling seems to work best for cutting out parts from flat stock.  A clever person (or a person with access to expensive software) can use it to carve pictures into any reasonably soft surface so I imagine in the future, clever woodworking joints will be developed.  I’ve seen things that look something like dovetails that seem to work well.  As for my friend, he’s written his own software to design and then carve out guitar necks.

When you’re cutting a part out of a piece of wood, you have to make sure you have the software cut on the correct side of your line.  Similarly, if you have something that needs to be cut out that will end up as scrap, you can just have the machine mill away all of it but that takes lots more time.  I’ve heard this referred to as “eating the scrap” but I don’t know if that’s a generic term or not. It’s easier to just leave tabs there to hold the cutoff piece.  That way, the waste pieces are held in place until you come back and cut the tabs with a knife.

Coming from ordinary woodworking, it was hard for me to think ahead and take advantage of everything.  You need to have all your dimensions handy before you start instead of the old “cutting to fit” routine that I used to use.  In other words, it requires you to completely design everything down to the last dimension before you start cutting. You also need to forget the fractional measurement system and switch to decimal.  It was interesting to start a woodworking project with a calculator and a caliper.  You can also do some cool things that were not easy before, for example, my project called for half inch thick plywood but I only had quarter inch material.  No matter, although I could have gone out and bought some, we simply cut twice as many parts and glued them together.  It was plywood so it didn’t look any different after it had been laminated together that it otherwise would have.  We included some alignment holes in the parts so that I could stick in a 1/8 inch dowel to keep things aligned while the glue set up.  This dowel was just cut off and sanded flush – it was barely noticeable.  That’s not the sort of thing you do in traditional woodworking.  If you’re willing to change bits (which is no small task by the way since you have to recalibrate the machine after a bit change) you can switch to an engraving bit (v-groove bit) and engrave the parts with labels.  That would be handy if you were producing large numbers of parts. We didn’t do that because we were still prototyping but that leads us to another issue.

You have to tell the machine before you start exactly how large your stock is (so that you or the software can place your parts on it) and where your stock is (so that the machine will cut where you think it should).  This latter is something you do after you clamp your stock to the bed (by any of several means).  First, you use the arrow keys (in our case) to move the router around until you are at the lower left corner of the stock which is probably where you defined your zero point to be in your original drawings.  (A very Cartesian way of looking at things. It doesn’t have to be this way – you just want to align the part you’re making with the stock you have.)  Then you place a metal plate on the stock and tell the machine to find it by lowering the router and bit until it just touches the metal.  It does this twice just to make sure things stay consistent.  In this way it knows where the surface of your stock it (less the thickness of that metal plate which is well known from the factory.)  At this point you can start cutting.  Noobs will usually make the mistake of designing a part and then not noticing where the origin of their system is and cut a part out of the middle of a sheet of stock.  This can screw up a lot of material depending on how extensive your toolpath is.  If we had wanted to change bits (for example to label the parts with a v-groove bit, we’d have to have gone through the calibration process again.  If you’re using a laser cutter then presumably you would not have this problem.)

Dust collection is an issue.  If you’re doing a lot of cutting, the router can be in continuous operation for an hour or more so a lot of dust can get created.  Dust collection is built in to most of these things and that’s fortunate.  Perhaps it’s even necessary depending on your allergy situation.  I suppose you could just stand there with a shopvac hose; it can get pretty boring when the machine is doing its thing for an hour.  You might feel that you may as well be doing something.  Still, I prefer the automated dust collection.  I was taking photographs during my milling session.

Holding down your stock is another issue.  The bed of most of these machines is made of sacrificial MDF which you replace when it gets too chewed up.   So you have some options in how you stick down the pieces you’re milling but screwing your stock to the bed is the most obvious option.  You just have to make sure that no screws are anywhere that the bit will be cutting.  You can also use holddown clamps as long as the router doesn’t get too close to them and knock them off.  A little forethought will keep that from happening.  People who use CNC mills for a lot of production will commonly use a vacuum clamp to hold down the stock.  This complicates the process in that you need a vacuum system first and a shopvac will not do, nor will a dust collector.  You need something dedicated but once you have your fixture built, it becomes a simple matter to put your stock onto the vacuum fixture and turn on the vacuum.  This holds things down admirably and there are no obstructions to the router when you’re cutting.

This is all just what I learned from playing around with this for a few days.  There’s much more to learn but the biggest steps are: designing your parts with a CAD program, importing them into your CAM software (which usually comes with the machine) to convert your parts into a toolpath (usually saved in a language called ‘g-code’), and the actual machine control software.  To get from the CAD to the CAM phase, you need to save your files in a common format; “DXF” is the most common and this may limit the CAD software you use.   Google Sketchup for example will not save in this format unless you purchase the professional version or purchase one of several add-ons that will do the same thing.

Once you go through the process, you begin to see that if you can draw it on the computer, the machine will cut it out for you and it will do so perfectly and as many times as you want.  With all this capability it’s surprising to me that what most CNC router owners (Shopbot specifically) end up doing is making carved wooden signs.  You can literally make anything; and a few hardy souls do just that.  My friend is using his to carve guitar parts (necks, all the internal braces, fretboard, etc).

There are also other ways to have flat parts cut which might work better in some cases.  There are laser cutters that work much the same way.  Probably the water jet cutter has been around the longest and is advertised most often as the way to get many different materials cut.  Places that do this are all over the country and often advertise themselves as being willing to do only one article.  The possibilities are endless it seems.

It’s fun. And it can change the way you see the building process.

Next Page »