April Through September 2007 CNC Blog Archive
I'm taking a day off from my other activities to do some work on this blog and hopefully get down to the shop.
If you want to run flood coolant, you'll need an enclosure around your mill or the coolant will go everywhere. Here are a few example enclosures that might spark some ideas:
The splash guard on this Tree CNC mill would be a lot easier to build than a full enclosure. It will limit the size of work that can be done, however. The Widgitmaster has built a very similar enclosure for his Bridgeport clone:
The Widgitmaster Enclosure being bolted down to mill table. Note the shiny plate is some work he will dial in. It is located against two pins at the rear.
Locating pins and clamp...
Pins on the bottom locate the enclosure against he mill table...
Here's the enclosure in use...
It seems like it would be straightforward to make one of these and to leave room for a vise as well. It's an ideal setup to key in your vice so it will be properly indicated in when you drop it on. It's basically a fixture plate with plexiglass walls.
Kap Pullen uses a very simimlar enclosure on a mill at his work...
Here's a more traditional enclosure. I love the way the whole front hinges downward for access. Notice how the coolant is channeled into the drip pan at the bottom...
I've always liked the idea of making an enclosure from 80/20 aluminum extrusions:
Here is the Flood Coolant in action...
It drains from the bottom into a tub. Electronics on top are high and dry!
You can see from the pictures why I like the 80/20 route: it goes together with a minimum of fabrication and comes out looking extremely professional. The extrusions aren't cheap, but I really think they're a very attractive way to go.
Speaking of attractive, here is a totally awesome fabricated enclosure for an IH Mill:
Shiny diamond plate sure looks good!
Electronics are right underneath the mill...
Not Machine Tool Oriented, But I Want One!
If you're like me, there is a gaggle of boxes and cables under your desk that are associated with your computer. There are routers, DSL modems, power strips, and every manner of wall wart power supply. It's scary to even go under there and try to plug a new one in or fix the old. Enter the pegboard under-desk organizer:
Such a cool idea!
Slow Times at the CNC Cookbook!
You've no doubt noticed a bit of a hiatus in the postings here. That's because I'm in the process of changing jobs. I'll get back into this thing once the new job is up and running smoothly, but in the meanwhile, things will be a bit sporadic. Bear with me!
Great Idea for a Vibratory Polisher
I couldn't pass this one up because it seems so cheap and easy to try:
Take four pieces of 1in pipe weld them to a flat piece of steel. Make it so a five gallon plastic bucket will slide up and down loosely between the pipes. Then pick up a cheap jitterbug sander from harbor Freight. I got one for $9.99 on sale. put a long piece of threaded rod up through the center of bucket to bolt the top down then epoxy a block of plywood to the bottom of the bucket drill out hole to fit over rod and nut on bottom of bucket. bolt the sander to the block of plywood. You put your parts in the bucket with some crushed walnut hulls You can get 13lbs of them on ebay. for 8 bucks. add a little polishing compound to the media. You attach a air hose to the Jitterbug sander. when you set the whole thing bucket and sander down between the rails of pipe it hits the trigger lever which is on top of the sander. it points down when bolted to the bucket. as long as air goes to it the whole bucket will dance It took me 15mins to build this and guess what it works.
From the Anodizing Yahoo Group. Sounds like a clever solution to me!
Mach 3 Controller Limitations
I recently came across an interesting thread on the Mach 3 boards talking about the limitations of the various controllers. As you may know, there is a relatively new crop of controllers that Mach 3 supports that do the step pulsing in the controller, relieving Mach 3 of the task. Supported controllers include the Gecko GRex, Galil, and the NCPod. In theory, these boards allow much higher performance because they can deliver a smoother pulse train to the step or servo drivers and many more pulses per second. But here is the rub--each outboard controller has a limitation on the number of moves per second it can accept, and all of them accept a lot fewer moves than a parallel port:
Parallel Port: circa 10,000 moves/sec
NCPod: 1,000 moves/sec
Galil: 250 moves/sec
GRex: 50-100 moves/sec
Those are some big differences! Does this mean the outboard controllers are all doomed to poor performance relative to the parallel port? Not necessarily, things are more complex than that. Think of the moves as being a rough estimate of the maximum rate g-code moves can be evaluated. If the g-code is making 0.0001" moves to simulate some sort of smooth 3D curve, you get the following maximum rates:
Parallel Port: circa 10,000 moves/sec = 1"/sec of 0.0001" moves = 60 IPM or 1" of 0.0001" moves
NCPod: 1,000 moves/sec = 0.1"/sec of 0.0001" moves = 6 IPM or 0.1" of 0.0001" moves
Galil: 250 moves/sec = 0.025"/sec of 0.0001" moves = 1.5 IPM or 0.025" of 0.0001" moves
GRex: 50-100 moves/sec = 0.005 - 0.010"/sec = 0.3 - 0.6 IPM or 0.005 to 0.010" of 0.0001" moves
Wow! Those get to be pretty darned slowed speeds at some point. In fact, these are theoreticals, and there are mitigating circumstances of various kinds. For example, the GRex is far more likely to hit its theoretical maximums than the other boards because it has an internal move planner that looks ahead and smooths acceleration. The others have to deal with start/stop jerking, and so won't hit their theoretical maximum. The parallel port's maximum is mitigated by the ability to generate the necessary pulses to move at those speeds, and as Art will tell you, 60 IPM is in the grey zone. Some stepper motors can be tuned up to perform that fast, but many cannot. I know the motors on my CNC MiniRouter cannot--30 IPM is about the best I can do there.
So what's to be done about this? It rears its ugly head most often for 3D profiling operations on a mill. What are the chances you need to profile to 0.0001"? What are the chances your mill is accurate enough that its even possible? You probably don't and can't, but the bad news is your CAM program may be throwing out such moves regardless. I took a look at OneCNC XR2 Mill Advantage, the program I use, and there is a parameter for the tolerance on the 3D profiling. I tried creating a program to profile the rifle stock I designed, which is a demanding 3D task. I set the parameter to 0.0005" tolerance, and posted. The result is a 17,000 line g-code program that OneCNC says would take 3 1/2 hours to run. Incidentally, for wood it wants a feedrate of 4 IPM, which is faster than really small g-code segments can go according to the calculations above, although GRex should be able to hit about 1.5 to 3 IPM with 0.0005" segments. I loaded the g-code into Excel and wrote a quick sheet to determine how many moves were less than 0.0005". Out of the 17,000 odd lines, only 75 made moves less than 0.0005". The total distance moved in this way was 0.023".
Another way to look at this is to look at how much travel can be had with these short segments before the controller starts to get behind. The parallel port cna go 1", NCPod 0.1", Galil 0.025", and GRex 0.005 to 0.010". With the worst case being the GRex, we assume that moves of less than 0.005" can eventually lead to falling behing. Since the buffer is 50 to 100 moves/sec, let's assume 50 moves for GRex, 250 moves for Galil, and 1000 moves for NCPod. On my rifle stock sample, the maximum number of moves that are 0.005" or less is 29. It seems to me that the GRex could tolerate up to 50 such moves before it fell behind, so I should be okay.
What's my conclusion? I'm not going to worry about this particular buggaboo. The rifle stock model I tested is very curvy, yet OneCNC generated very few g-code moves short enough to cause problems with the GRex. Your mileage may vary!
Lathe Touch Off w/ Edgefinder
I knew there was a reason I'd need one of these cool electronic edgefinders after I saw this picture of using one to touch off a CNC converted Lathemaster 9x30:
BTW, this conversion apparently will do 125 ipm! This is the $175 model from J&L Industrial, made by XYZ. Apparently this fellow tried some cheaper ones with poor results.
CNC'ing the "Hula Hula" Steam Engine
I've been working through the drawings for the Hula Hula engine (a Philip Duclos design) and converting them over so I can build the engine using CNC techniques. So far I am just at the Rhino drawing stage, but I'm hoping this can be one of my first projects after completing the conversions on my lathe and mill. Here is a sample drawing:
As you can see, I have slightly redesigned the parts. Some things that were more easy on the manual mill Duclos used can be better done a different way with CNC. These cylinder backplates and associated holes are a prime example. Duclos originally had these pieces with nothing but square cuts and pointed tops. The point on the top would locate the backplate on the Engine Body, and all holes were to be drilled using the backplates as a guide for greater accuracy and match up. With CNC, it's easier to profile the round shapes shown, and the holes can be drilled with enough precision under CNC control that it isn't necessary to match up the pieces.
Tubing Bender and "Unbender"
How about this beautifully made tubing bender for model engines?
And the matching "unbender" for straightening tubing:
Making Square Holes With Round Pegs
I've mentioned in the past that I like to allocate a certain amount of shop time to experimentation. I find that if I do all my learning on projects, I get too fixated on finishing the project, and may shortcut the learning process. When experimenting, everything is scrap, and the only outcome is knowledge. A little of both practical projects and experimentation goes a long way towards making the whole greater than the sum of its parts.
Today, I decided to try to make a square hole in a piece of steel. A friend was asking if I knew where to get sockets with square holes for some unusual bolts he had. I suggested a couple of possibilities, but then stepped up and said, "We can make one too out of a donor socket and some scrap metal." I decided I'd better get going with an experiment before he shows up locking for his socket.
My plan was to do the old trick where you cut a square slot, and then "glue" two of those together to make a square hole. In this case, I took a 3/8" end mill, slotted the end of a piece I had squared earlier, sawed then end off, cut that part in two, and then Tig welded the result together. After grinding down the welds, I found the technique to be eminently workable:
Square Hole: Mission Accomplished!
A couple of things I learned or that should be noted:
- The left and right sides are not exactly square--they seem to curve to the left. This is because I cut the slot in a single pass the width of the cutter. You want to use a cutter narrower than the slot so that each side gets a nice pass all to itself without cutting forces deflecting things and curving the walls. This is no biggie, I knew that even as I was making the cuts. This was just a test, but I want you to be sure you understand.
- Z depth control is critical to squareness. I have some techniques I've written about to get me within 0.001" in the past. Make sure you can do the same if you want your hole to come out square.
- I started with a block I had squared. Good idea!
- I needed to make a few practice passes with the Tig, but instead I just dove in, so the bead is pretty nasty. However, I ground off the bubbles and it is okay. I wouldn't want to run a 5HP motor shaft with it, but it will work to turn square headed bolts! Next time I'll lay a couple beads on some scrap before diving in.
- I tried a slitting saw to cut the slotted piece in two. This was my first time. They work great! I looked up some feeds and speeds and found they want to go real slow. In fact, I found the feed speed was too slow for power feeding, so I just fed manually. As long as you take it easy, I found they cut extremely well and leave a very clean accurate slot behind. I'll be trying one of these again sometime soon!
So all in all, it was a pretty happy 1 hour segment in the shop to learn a couple of things. I'll hand my pal the test piece and see whether he wants to go ahead making up a custom "square head" socket.
Schaublin Rear Cutoff Tool
A number of commercial lathes I've run across have an option for a rear cutoff tool, usually pneumatic or servo operated. I really liked the manual lever operated tool on this Schaublin lathe:
Something like that would not be too hard to build and seems like it would be very handy!
Another Gantry Crane
That color would work in my shop, eh? This fellow recommends 2 dollies on the crane as coming in handy a lot...
An Insanely Nice Southbend Lathe Restoration
I'm not too sure I could bring myself to cut any chips on this newly restored Southbend lathe if it were mine:
It belongs to a lady named Paula, who says she uses a removeable metal pan to catch chips so the wooden top is protected. But you ain't seen nothing yet! See the Rivett below.
Museum Quality Rivett Lathe
The Southbend above was beautiful, but the Rivett is even more so. They're a more precision machine than the Southbend design. Just look at all the hand scraped surfaces:
Super Titanium Penlight
Came across this really neat AA flashlight on the PM boards:
Nicely made design for an AA flashlight...
Obsessed With Precision? (Moore Angle Plate)
I guess you could say I'm obsessed with precision. At least I'm obsessed with having the ability to measure and align precisely. A lot of the shortcomings of cheap machine tools can be at least partially overcome if you can measure precisely. For example, adding a precision DRO to a tool can greatly increase the precision of the work you can do on the machine.
The science of measurement is called Metrology, and its a complicated business. I keep my eye open for items on eBay that give me a chance to substantially improve my chances for precision. One such that came along recently was this Moore Angle Plate:
The picture doesn't do it justice, because this thing is 36" long and its two surfaces are hand scraped:
Hand scraped precision surface...
The name "Moore" is synonymous with ultraprecision in the machine tool world, so I reckon getting my hands on this piece for $140 shipped was a real bargain! I now have a very precise standard that I can use a a straightedge or for checking 90 degree angles. If I wanted to make a set of box ways, I could use this angle plate as a standard for scraping them in. The reality is it will probably sit in its box and see only occassional use, but that's okay. I feel like I've significantly extended my shop's capacity for precision.
I've made a couple of other recent eBay purchases in the precision department. First up is a pair of Browne & Sharp ground parallels. Doesn't sound to impressive? Well, these parallels are 1" x 2" x 12", so they're really BIG! These are ground to 0.0002". I figure they may be handy for a precision setup of some kind some day. For example, to spread the clamping force if I am milling a big plate I could put these on the edges atop the plate and clamp on them.
I've also got a cylindrical square on the way that looks pretty nice. They're just the thing for checking spindles on mills and that sort of thing.
Lastly, I've been on the lookout for Dial Indicators. I mean the plunger style, not swinging indicators, which are called Dial Test Indicators. I've made do for quite some time with a really cheezy Central Tool indicator, and I wanted something nicer with decent travel. I scored a brand new Starrett with a snazzy 2" dial for about $20 on eBay. The difference in the feel of the action between the Starrett and my cheap indicator are like night and day. The Starret is silky smooth!
Having some extra indicators will let me keep them set up in their fixtures for faster use.
A Shopmade Follow Rest and Box Cutters
Turning a small diameter piece that's long and skinny is difficult. The piece is going to want to bend and flex and generally make it difficult. Imagine trying to work this piece, for example:
That long skinny piece would give me fits!
Steady rests and follower rests were made to pick up where the support a tailstock can lend is not enough. Most lathes come with these gadgets, and most folks don't use them often enough--I know I certainly don't. Part of my problem is I have the usual cheap Asian brass tipped units, and I just can't feature that brass riding very happily along the workpiece. It has been in the back of my mind to add some ball bearings. I've got a ton of cheap skate bearings floating around the shop that should be perfect for the job.
The fellow that did the dollar bill piece above came up with a solution when turning that piece:
Frank Ford, whom I've mentioned here before (be sure to visit his awesome site!) has a mini-follower built right into a QCTP toolholder:
Here's another version, the one I think Frank said he modeled on, that you may be able to purchase from King if you don't have time to make one:
The only thing I don't neccessarily like about the QCTP mounted versions is your tool has to be square onto the workpiece. For some operations, that isn't the case, so I think mounting one to the carriage as in the first example might be a little more versatile.
Sine Jaws for the Kurt Vise
After cogitating on the idea of putting dowel pins in vise jaws (of which more below on how a fellow did that for another purpose), I did a Rhino drawing to figure out the placement of pins needed to provide a set of angle stops in 5 degree increments from 5 to 45 degrees:
This would be a dead simple thing to do with CNC, but I went ahead and listed DRO coordinates for those who want to play with it on a manual mill. A rotary table would let you do the work very easily as well. The idea is to drill and ream holes for 1/8" precision dowel pins. As you can see these holes are divided between a 2" and a 2.5" radius so they don't land too close together. One could either make a little bar with two pins to use as a stop to align the workpiece at the desired angle, or just use two dowel pins in the appropriate holes. I would bore the opposing jaw with holes slightly oversized so the dowel pin ends go in when tightening the vise on thinner workpieces.
I got the idea to use dowel pins from Jim Sehr whom I mention below for this set of jaws:
Headstock Adjustment Bolt
Birmingham lathes have a bolt to adjust the headstock so it cuts true to the ways with no taper:
Nifty Soft Jaw Tricks
Soft jaws for your Kurt vise are fantastically useful gizmos and easy to make. But there are a few clever things I saw recently while prowling the web:
Use slots instead of holes and the soft jaws are quick release...
A series of holes for precision dowel pins and you get:
An easy way to hold round parts for drilling. Normally I would make do with a V-block in the vise...
And here is one of the most interesting tricks. This is a mini sine bar. Depending on which holes the two pivot pins use, you get angles from 0 to 90 degrees. Exotic, but useful! Available on eBay...
Posted Some New Workshop Pix
There've been some updates since I first posted about my workshop! Check the workshop page for the newest photos:
QCTP Indicator Holders
Someday in the not too distant future I plan to give my lathe a tune up. I want to check spindle runout, adjust the preload on the spindle bearings, check the headstock and tailstock alignments, and generally give it a little TLC. I want it running as accurately as possible before I tackle making the end blocks to hold the ballscrews when I convert the lathe over to ballscrews. Angular contact bearings need some pretty close tolerances for installation. I'll also be turning the ends of the ballscrews, so it pays to have it all shipshape.
One of the things that I've been seeing for a long time and thinking I need to build is a QCTP holder with an indicator in it. I recently saw another one and thought I'd do a little roundup article here so I've got the details all in one place.
Needs no Dovetail Cutter
If you don't have a dovetail cutter for making QCTP holders (I made one, it isn't hard!), you might consider this fellow's approach of just doing it by milling the dovetails as separate parts:
The QCTP Indicator Holder...
Using a Tilting Vise Fixture to Mill the Dovetails...
The Components. Note How the Dovetails Are Bolted to the Holder...
"Flapper" For Irregular Shapes
Marv Klotz gave us the "Flapper" design for dialing in irregular shapes or square stock in the 4-jaw:
This one just uses a magnet to attach itself...
5Bears Indicator Holder
5Bears (the Swede) modified an unused QCTP (looks like a knurler) for this purpose:
CNCCookbook "Instant" Indicator Holder
As I was writing this, I was staring at an indicator holder that fits onto a height gage I got off eBay. These replace the carbide scriber and can be used to increase sensitivity and accuracy of the height gage. eBay seller discount_machine (I think that's Shars) has them for $8.95:
If you want one (I ordered a second after seeing how useful they can be), do an eBay search for "HEIGHT GAGE INDICATOR". They only have them on "Buy it now" in their store, so you may have to look carefully.
I took this little gadget together with the QCTP knurler holder (everyone has one and they aren't that hot if you get a scissors knurler, so its great to reuse it) and put them together to get this:
It wouldn't take much to rework the mounting bar so it was just like 5-Bears holder.
Frank Ford's Holder
Frank has not one, but two versions, although the second is just an improvement he made to the first:
The Mark I...
Mark II: Now with a blade so you don't care if it's on center!BTW, Frank Ford's "Frets" site is filled with wonderful tips and projects. Do check it out if you haven't found it already!
Someone is Finally Putting Linear Rails on an Asian Mill!
I got the idea to try this when I came across a Tormach mill for sale on eBay that had bad column ways. I bid on, but did not get the mill. This fellow has actually started to make the mod and its looking quite interesting:
The router is being used to mill a nice flat spot on the side of the column using the original dovetail ways and saddle as a guide. This looks pretty cool!
The linear rail looks very beefy as well. Can't wait to see how this turns out. This fellow is also doing a belt drive conversion on the mill head.
Need More Lathe Precision? Add Some Indicators to the Axes
It's tough to be more precise than you can measure (although some lathes will be less precise than they can measure!), so maybe you should improve your lathe's capabilities in that department. This fellow builds his own super performance model engines for R/C boat racing and added tenths indicators to both axes of his Emco Maier Compact 11 lathe:
Tenths indicators on both lathe axes...
You could achieve the same result with a very sensitive DRO on each axis too.
Door Hinge Thingey
I dunno what it is, but the thought was that it would look nice as a door hinge on a hot rod maybe:
Stunning Paintball Pistol
Came across this completely awesome custom paintball marker made by DocsMachine:
Electronic solenoid-actuated paintball gun (heavily-modded WGP Autococker) with infrared break-beam ball detection, programmable firing modes, and operated off 4,500 psi compressed air. (Regulated, of course, down to around 180 psi operating pressure.)
The workmanship and technology here is just amazing!
Things People Like to Make
It's cute, and I might need something for the kids...
Quarter Scale Tractor Pulling!?!!
My Internet Service Provider had a serious hardware failure so the site was down and out for two days. They never did get everything properly restored as far as I can see, but I think I was able to get everything back. Drop me a note if you encounter any odd behavior. Thanks for your patience!
Super Precision Machine Accuracy Checking: Ballbars and Circle Diamond Tests
Tree machine tools used to ship a circle-diamond test part with every CNC mill to prove that the machine performed to spec. The part looked something like this:
It's an interesting looking test, and one I had never seen much written about. I recently did find a brief article describing the test that I found interesting. Low and behold the circle-diamond test is an official government test that machines used to have to pass for aerospace work. The test is called "NAS 979", where NAS is an acronym for "National Aerospace Standard". Interesting!
The NAS 979 test is designed to measure:
It's supposed to be a pretty complete test of a milling machine, so sometime I may try to get together some g-code for it. As originally concieved, it is supposed to be cut from a 14" x 14" x 2" block of aluminum, so I may try something a little smaller!
Now let's fast forward to the invention that has replaced the circle-diamond test, something called a "ballbar":
Ball Bar: Precision linear measurement held between 2 spheres...
The ballbar was invented in the mid-80's at Lawrence Livermore Labs and consists of a precision linear transducer held between two spheres--one in the spindle and one on the table. The transducer tells the ballbar computer software whether the two spheres are moving closer or futher apart as the machine moves through a series of circles around the fixed sphere on the table. A ballbar can determine the following:
Control Loop Errors
Axis Slide Way Errors
Pretty nifty device, eh? Shops use the ballbar as a way of tracking their machine performance, anticipating when maintenance may be required, diagnosing machine problems, and providing documentation to customers that their machines are in working order.
How About a Really Nice Monarch 10EE?
There's something about the look of these lathes that just can't be beat. Some other lathes, like the Hardinges are also nice, but the Monarch 10EE is machine poetry at its finest. Here is a particularly nice example, recently restored for a total cost of $12,000 (phew!):
How about those monster leveling "spacers" underneath?
I don't know that I'll ever own one of these beauties as I am pretty firmly committed to CNC. It would be an awful shame to convert one to CNC too as far as I'm concerned.
On the Matter of Cheap vs Expensive Angular Contact Bearings for Ballscrews and Spindles
Without ball bearings of various types, machine tools would be impossible. Their most critical applications involve ballscrew mounting and spindles. Unfortunately, these very same critical applications often call out for very expensive bearings that are out of reach for hobby class machine work. I have a confession to make: I harbor a deep resentment for those expensive "machine tool quality" angular contact bearings.
It may be an unreasonable resentment from some perspectives. NCCams over on CNCZone will tell you all day long that you get what you pay for and you have to buy the most expensive bearings you can't afford, but my resentment leads me to wonder whether it is all really necessary. Yes, if I'm building a vertical machining center with micron accuracy that's capable of 600 ipm rapids, I'm sure they're necessary. NCCams has built a precision machine used to make camshafts for NASCAR winning seems, surely a very exacting application, and one that needed great bearings. But do I really need those costly bearings to do an 6K rpm spindle for a hobby mill? Do I need them for a ballscrew bearing block on a machine I hope will be repeatable to a thousandth? I feel the resentment is reasonable for the hobby machines. Someone needs to speak for them!
There are tantalizing clues about this conundrum that I run across from time to time, and it always perks up my interest. Some examples:
- I am told by various sources that garden variety bearings of today are every bit as accurate as the machine tool quality bearings of the 40's and 50's.
- Many Asian-built machines such as the Tormach mills do not use ABEC7 bearings, they get by on lesser grades.
- See my blog post below "When the wrong bearings" wherein I explore the use of multiple deep groove bearings to achieve levels of stiffness comparable to angular contact bearings. I can't see why 4 of these dirt cheap bearings couldn't be made to perform like $400-600 worth of expensive AC bearings. This article is also copied on the belt drive page.
- See my notes on the belt drive page about how to go about hand fitting unmatched bearing pairs to be preloaded duplex pairs.
- I constantly see examples where machinists are able to get superior performance from worn out or inferior machinery because they know the right tricks. Why can't that be true here too?
All of this will sound like a lot of sour grapes belly aching on costs and snake oil selling to those professionals who think nothing of just specing the expensive bearings designed to do the job. They may be right, but I have a sneaking suspicion there is more to it than this. Why might the professionals use expensive bearings if cheaper ones would do? I can think of several reasons.
First, look at it from the standpoint of manufacturing repeatability. Machine tools have to be warranted to certain performance levels despite variations in their construction and component parts. Tightening up the specs on the components makes it less likely the tolerances will stack up poorly and a machine will leave the line that isn't up to specs. The science of Six Sigmas and quality control will tell you it is cheaper to set up the manufacturing process to avoid these mistakes in the first place rather than find them after the machines are built and have to rework the out of spec machines. So the overall cost to mass produce a machine may be less while the individual cost of a single machine may be more.
Second, consider the warranty aspects and especially durability and wear considerations. An expensive machine tool sold for production business use will be expected to run tirelessly around the clock day after day in order to justify its cost. A hobby or light use machine need not be so durable in order to fulfill its purpose. Also, for many manufacturers, the warranty cost of a failure is very expensive to cure. The cost to an individual to fix their hobby machine themselves may be so much lower they're much more willing to risk it. I read somewhere on CNCZone of a fellow running a CNC router shop that uses hardware store routers and buys crates of bearings for them. He says it costs about $2.50 to replace each bearing and he gets 100 hours of continuous routing from a bearing. To him it is worth it to keep on replacing cheap bearings. To Haas, who sold you a very much more expensive gantry router and a warranty that causes them to have to send someone out to fix your bearings if they break, it isn't worth it.
Third, commercial machines have a radically different performance envelope than hobby machines. We kid ourselves we can do what they do, but we can't. We may be able to get a part made that is very close, but it will take us much longer to do it. We don't run nearly the rapids, we often run steppers instead of servos, cutting loads are probably much less, we are babying our cutters and our machines, while the pros are cranking out 110% on the spindle load meters and creating so many chips our little home shops would be buried in no time if we tried it (not to mention all the other problems!). What we really care about most in these hobby class machines is accuracy and repeatability, and not all that much of that. Maybe someday the cutting speeds and efficiencies will matter to us, but for now, we'd just like to get the parts made reliably to a thousandth or so. I submit that this is a far simpler requirement than what most of these high end expensive bearings are being designed to deliver, and that we can therefore get by on less. The level of accuracy and performance needed to cut precision cams to be used in zillion dollar NASCAR racing may be a touch more than what we really need to build tabletop steam engines.
Lastly, there is a labor intensiveness factor that matters more to the manufacturer and less to the hobbyist (or to the Asian manufacturer for that matter!). If I am a hobbyist, I can take the time needed to take two relatively unmatched angular contact bearings and grind them for a desired preload. I can then hand fit them to the shafts and bores they'll live in, lapping, honing, or using whatever means is necessary to achieve a good result. If an experiment of this kind is marginal or fails, I can always try again and perhaps do better the second time. Most of the investment I have in is just time. OTOH, if I am a manufacturer, unless my labor is extremely cheap, I want no part of that process. I will spend quite a bit of premium to buy a matched pair of preloaded AC bearings off the shelf so I don't have to mess with grinding them and trial and error. A really fancy set of these bearings is maybe $800. It takes surprisingly little shop time for someone to run up $800 of labor, and they may screw it up! If I just buy the bearings, they're guaranteed and someone else takes the risk for the screw ups. Hence I just buy the bearings if I'm a manufacturer. Things were not always this way. Bridgeport used to hand fit spindle bearings, for example. Not because it made a better mill, but because it was cheaper than buying more expensive bearings at the labor rates in those days.
Along comes another member with the same machine who bought a cheap pair of unmatched AC bearings off eBay for $35. He got after them with a surface grinder to grind the inner races to create a preload condition, installed them on his machine, and measured 0.0008" play while still being able to turn the ballscrew relatively freely (too much preload will make the screw stiff and there will be a tradeoff between less play and too much stiffness). Backlash of 0.0008" would be fine for hobbyists wanting to machine with 0.001" accuracy. The price is right and a small amount of labor delivered this happy ending. Would a manufacturer do it this way? They probably would in China, but not from one of the "big name" machine builders they wouldn't!
Computer Upgrade Time
I dread upgrading because everything stops working for a little while. However, this is a good time for me to find the hours needed, it's been 3 years so machines have gotten a lot faster, and AMD recently put some big price cuts on their CPU's. That being said, I just ordered up a new set of "guts" (I like to keep reusing my custom Borg Cube case) from my favorite dealer, NewEgg.com.
Here's what the new machine will have going for it:
AMD Athlon64 X2 6000+ 3.0GHz
This is a dual core CPU running at a much higher clock speed than the single core chip I have today, so it should be about twice as fast!
AMD and Intel periodically trade places for the speed champ, with Intel usually offering a slightly faster chip for a lot more money. At the moment they seem neck and neck with AMD still a tad cheaper.
ASUS M2N32-SLI DLX WIFI AM2
I've been partial to Asus boards for a while now. They're extremely well made and stable. This one isn't cheap, but it is crammed to the gills with features and will make a nice platform I can upgrade at least once with better video and CPU.
ASUS EAX1950PRO/HTDP/256M R
Not the absolute fastest board available, but it is the fastest board under $200 at the moment. It will be dramatically faster than the old GeForce card I have today!
1Gx2 Corsair TWIN2X2048-6400C4D R
There's faster memory for a premium, but this is 2 gigs of memory that is plenty fast enough as I don't intend to overclock this system.
Western Digital 10K rpm 150GB 16M SATA WD1500ADFD
I love these Western Digital 10K rpm drives--awesome speed! My current machine runs 2 of them in a RAID configuration for even better speed. The RAID formats are not necessarily compatible between motherboards, so I bought this disk to use to back up the other two so I can then bring them up on the new mobo. It will take some backing and forthing, but I gotta have that RAID speed. When I'm done, this disk will remain to use for nightly backups.
These goodies, plus the power supply, DVD-R/W, and other bits and pieces in my current machine should give me quite an upgrade in performance. That will free up my current machine's guts to trickle down and improve my kid's machines. Isn't it funny how it always costs right around $1000 to do the PC upgrade you'd like to do?
I plan to buy another machine before too long to use for my CNC mill conversion. I think a Shuttle PC will fit inside the NEMA enclosure quite nicely along with the rest of the electronics. Stay tuned, I won't tackle that until the main machine has been upgraded!
Nifty Shop Photos: Truing a Lathe Chuck, Steady Rest, 4th Axis Gear Cutting...
Here's some nifty photo postings I recently came across by Jim Hubbel:
Truing a Lathe Chuck. Note the washer to keep the jaws loaded while grinding them true, as well as the sparks flying!
Heck of a Steady Rest! Big ball bearing unit...
Parts for the Steady Rest...
I need to make a tailstock for my Phase II table just like this one...
A Fellow GRex Rack Mounter!
I came across this fellow's thread on CNCZone and felt a kindred spirit. His rack system is pretty similar to my CNC lathe's rack chassis:
There are several features of his case that I really like. First, It was definitely simpler to just mount the Geckos to the wall rather than use the heatsink I did. Second, I like bringing out the fuses with pilots to the front panel for all axes. I think adding a disable switch to each axis would be nice as well, though it wouldn't get used an awful lot.
The GRex is so powerful and compact, you can fit a lot into one of these small cases!
Network Connectivity to the Shop
Networks and printers usually make for a frustrating day working on computers. Getting network connectivity down to my shop was no exception. I decided to try to do it wirelessly, thinking this would be easier than dealing with cables. So, I bought a wireless router at Staples and a USB wireless connection for the shop computer and I was off to the races. I had only a modicum of the usually nuisances that accompany this sort of job. For example, the "Quick Start" software that came with the router to make it easier completely failed to work. I don't think I've ever seen a network "Quick Start" utility that did work, which seems pretty ridiculous in this day and age. In any event, I had wireless down to the shop and I could log into my network from my shop computer. The chief advantage to this was not having to schlepp g-code files from the CAM program in my office down to the machines.
Having crossed this hurdle without singing too many tail feathers, I thought, "Why not keep going and get the lathe's GRex to run via the shop computer?" Capital idea! I had, unfortunately, screwed up royally in my approach. You see, the wireless modem on the shop computer plugs into its USB port and the GRex wants a CAT5 connector. Back to staples looking for a wireless guzinta (because I need something that guzinta that kind of jack). No such animal. How about Radio Shack? Never heard of it. Back home, I went to look at NewEgg. By now it seemed that nobody had thought it would ever be a good idea to take a wireless connection back into a CAT5. How silly is that?!??
I don't give up too easily on these network jobs. If you show fear or weakness, you're done for on this kind of thing. So, I started reading up on the Belkin web site, that being the brand of wireless router I had purchased, and discovered it has a "Wireless Access Point" mode. Great! This must mean I can extend my wireless reach with a second router and it would have plenty of nice CAT5 plugs on the back. Back to Staples and a new box was had. More configuration conundrums. As soon as I configured the second router as an Access Point, the PC immediately quit being able to communicate with it in any way shape or form. Drat!!!!!
I went through the hard reset process with it about 4 times before deciding that was for the birds. Out came a big long cable with CAT5 connectors on either end. I plugged that into a distant connection, ran it all the way to the shop, and plugged it into a cheap 8 port switch. I plugged computer and GRex into that switch, booted the computer and low and behold I was connected! I did the usual little back and forth between Mach 3 and the GRex before they were willing to communicate (just one retry, no biggie), and all was well.
Why didn't I do things the simple way to start with? Sigh...
Debugging the Lathe Electronics
Having gotten things connected down in the shop via LAN, nothing would do but that I needed to fire up the stepper motors for the lathe and verify I could spin the shafts under Mach control. Why? Well why not??? I'd seen this work before, but under much more experimental circumstances in my office and not down in the shop. If I could make it work in the shop, I could bolt the cover on, stick it in the rack next to the computer, and get going finishing off the lathe.
Alas, this was just not an easy day. It's funny, but I must have carried over some bad vibes from that last Friday the 13th or something. I hit the jog keys in Mach 3 and nothing. I could at least see the proper axes were lighting their LEDs on the GRex, so I know I had everything good to that point. What gives???
Out came the multimeter and I started tracing down how far the AC was getting to the DC power supply. The answer was, "Not very far." Doh! By this time it was 9 o'clock in the evening, so I shut down the shop and came upstairs to write this little missive. I suppose progress was made, but it certainly seemed minute. Still, it can't be all that hard to track that AC down and figure out what's gone wrong.
Dropping Out Parts in CNC
Let's say you're making a part that is going to be machined all the way around. How do you make sure it drops out nicely from the stock without hanging up on the cutter?
For example, let's take my EZ-Clamp design:
I want to machine the clamp so it is as thick as the stock. How do I make it drop out?
- Machine it to partial depth and then flip it over and clamp it in soft jaws for the vise that are machined with a "negative image" of the part. This method does require production of the softjaws, and so is perhaps better suited to production of more than one part.
- If there is a hole, bolt the part down to a fixture and machine away everything but the part and the fixture.
- Leave some tabs 0.010" thick that hold the part but that are easy to file through and then sand off. Apparently Mastercam has a feature to put the tabs in automatically. I believe some of the VCarve programs will too.
- Machine the part on the end of a piece of round stock and then part it off on the lathe. It will be helpful if the depth you cut the part is a little greater than required.
- Clamp half the part with 2 clamps. Machine the part that is free of clamps. Shift the clamps to the other side one clamp at a time. Finish machining.
- Consider super glue and double sided tape. These can work if the cutting forces and vibration are not too great.
Proper Installation of Precision Dowel Pins
I recently bought some precision dowels on sale somewhere to use for locating parts. You shouldn't be using bolts to locate parts, they're simply there to act as fasteners. Use dowel pins to precisely locate parts. The trick in having them work precisely is in how you make the holes to receive the pins. So, I went out and tracked down a thread that talks about it. There's always a thread to be found on the Internet!
I will spare you the reading of the thread by cutting to the chase. There are 2 preferred methods. The first will be a little faster, and slightly less precise:
- Spot drill: And folks were at pains to point out that you should not be using a center drill for this purpose!
- Drill 1/32" under.
- Plunge an undersized endmill.
- Finish by reaming.
The second method, which is more accurate but also slower, is to use a boring head.
Consider metric sizes for the undersized endmill. For example, 6mm is 0.2362", which is a reasonable undersize for 0.250".
If the fit is really crucial and you can drill the pin bores all the way through, consider clamping the parts in their proper alignment and machining the dowel bores in a single operation. You can go back and oversize ream one of the two holes for clearance.
Unusual Edge Finding Accessories: Toolmaker's Chairs
Edge finders are important gadgets to have so you can make sure your mill is lined up on a particular feature. Normally, they are just little gizmos you stick in the spindle that work either electronically or mechanically. The classic design spins and when it just touches the edge it "kicks out" to tell you that you've found that edge.
There are more elaborate schemes available, however. SPI makes a couple of edge finder accessories that are called "Toolmaker's Chairs". They look like this:
As you can tell, they have embedded magnets to hold them in place, and they are precision ground to 0.0001"! You use them to precisely locate using a dial test indicator. To find a corner, the chair on the left is places on the corner and then you indicate on the circular hole that is precisely centered over the corner. To indicate on an edge, use the slotted tool and your DTI.
This interesting tool also serves for center punching:
Here is a similar Japanese-made commercial tool:
A Few New Goodies Arrive on My Doorstep
Kabel Schlepp cable carriers (used to keep cables from tangling on CNC machines):
A precision ballscrew for my gang slide:
A planer gage, sorry no pix! And an ER40 collet chuck in 30 taper which may someday be used on my mill belt drive conversion:
Interesting Lathe Modification: Headstock Tramming Screws
I came across one of the more unusual lathe modifications I've seen in a long time on CNCZone. This fellow has added setscrews that bear against the 4 mounting bolts for the headstock of his Asian minilathe so that he can tram the headstock for more accurate results:
Allen wrench is used to adjust the tram...
It's an intriguing idea, but I wonder how well it works in practice? It seems like he is just cocking the head on the V-shaped ways, which would have to introduce some odd side effects, reducing rigidity at the least. OTOH, someone else on the thread opined as how this is a standard feature for some 9x20 lathes from the factory. Perhaps the 9x20's don't sit on the V-shaped ways? Another remarked that its important to torque the main bolts carefully, constantly rechecking alignments lest they be knocked out again. I will also add that I read some posts on CNCZone by Widgitmaster that shows his Birmingham 14x40 lathe has bolts similar to what this fellow is using for headstock alignment.
I had heard a better way to approach this problem is to first level the lathe and then apply controlled twist the bed to counteract whatever the remaining errors are. One fellow in that thread claims this was the "official" approach advocated by Warner Swasey for their lathes, and certainly a number of the fellows I've come to respect are very supportive of this approach.
At some point I mean to try the bed adjustment. I want to go through a full "accurizing" process with my lathe after I get the CNC conversion completed to see what the maximum level of performance is that its capable of. FWIW, folks on the aforementioned bed twisting (sorry, trying to humorous) thread feel that taper of less than a thousandth over 6" on a piece with reasonable diameter so it won't flex is fine. The factory specification for Monarch 10EE's was 30 millionths of taper in 2" of length. That's 0.000030"!
Lathe Progress Continues
I had to make a little faceplate to go over the breakout board I'm using to connect the control panel. This was a perfect job for the Mini-Router. Took hardly any time at all and the end result in 1/8" aluminum was perfect:
I also did some serious design work on how the spindle control circuitry will work. Here is the schematic:
For a further exploration, please see the driver electronics page. I'm still not done on this design, but I am much closer!
Cutting T-Slots in Cast Iron for Machine Tables
My Gang Tool Slide for the lathe is going to require me to cut T-slots, so I went poking around and came up with some tips on cutting T-slots and cast iron.
First thing is you'll need a T-slot cutter. These are special milling cutters made for the purpose of cutting the bottom extra wide area and they're not cheap!
Here is what one looks like:
Note the unusual staggered tooth pattern. It's designed to help pass the chips through the confined spaces of the T-slot.
Okay, given the requisite cutters, the first task will be to square the cast iron piece if you haven't done so already. The Fidgiting Widgitmaster makes these T-slot tables for a living, and has recommended that when roughing cast iron, one should use low rpm and high feed rates. Within the limits of my machine's HP ratings, my 3" indexable face mill could be used at 764 rpm, 20 ipm, and 0.050" depth of cut. The Widgitmaster goes on to suggest that for a fine finish, try a 0.005" depth of cut and slow feed. I'm thinking I'll use my newly purchased big fly cutter to do this job.
Next step is to rough cut the table slot with a regular square end mill. You need to make the slot wide enough to allow clearance for the T-slot cutter's shank, but it need not be much wider than that. Some folks like to cut the table slot so it will be a little deeper than the T-slot's bottom. The allows extra clearance for the T-bolts so they don't jam up as easily. It also reduces the likelihood the T-cutter will dig in. All in all, it seems a good idea to me.
Having roughed in the table slot, the next step is to run the T-slot cutter. These cutters have a hard life--you can't sneak up on the cut at all, they have to cut full width and they have to do the job down in a hole. Be sure to make arrangements to clear the chips down in that hole! I'm told most people prefer to cut cast iron without coolant, so one could use compressed air, but a shop vac rigged up to suck out the chips seems an even better idea and will keep from spreading nasty cast iron dust all over the shop. I'm also thinking I want to run that T-slot cutter without touching the Y-axis at all after having milled the table slot. Feed and speed for the T-slot cutter are interesting. I've heard a recommendation of 1/2 the spindle speed of an equivalent sized end mill but much faster feedrate because of all the teeth.
The last step is to do the finish cuts on the table slot. You want these to be well finished and true, so now is the chance to cut only on one side instead of full width. Take care the slot stays well centered on the T-slot and use optimal finishing speeds and feeds for the cast iron as well as a nice sharp 4 flute end mill.
Lifting Heavy Fixtures and Tooling
Lifting heavy tooling and fixtures such as vises, chucks, or rotary tables can be difficult. I've come across several aids in my travels that I thought I would share. First is this nifty vise caddy that I have added to my project to do list:
Isn't that cool? It's a copy and somewhat nicer looking version of a gizmo that SPI sells for almost $400. Attaches to the base of your mill and then you can move your vise on and off the table easily. I could see keeping a shelf back behind the mill that is accessible via this arm.
Next I have a tip for owners of larger lathes wanting to change chucks. It doesn't come with pictures, unfortunately. The suggestion is to chuck up a suitable section of pipe, place a block of wood on the ways to act as a fulcrum, and then use the pipe to lever the chuck off the lathe and over to where it's going. If you keep your chucks standing on edge you can simply chuck the next one onto the pipe and use the same strategy to manuever it back onto the lathe. You should probably use aluminum pipe or conduit to avoid gouging your spindle with the end.
How is this for a portable lifting gadget for moving fixtures and tooling on and off the mill tables? It was created by a machinist on CNCZone named Geof:
That seems like a pretty handy doodad. It is set up to straddle the mill base on his Haas VMC's. Geof is not done with tricks yet, however. Consider this tidy installation on his Mini-Mill:
A similar lifting arrangement with hydraulic cylinder and arm is bolted to the floor. He's got those 2 Kurt Vises on a tooling plate...
So the whole thing comes right out and can be dropped onto a rolling cart...
But what's lurking in the corner? Rotary 4th axis hangs from a turnbuckle. Once the vises are out of the way, it can be dropped down onto the table...
Of course there are those shops with a jib crane standing next to every machine, and my all time favorite for wretched excess, a home shop with overhead travelling crane:
Handy Digital Bevel and Angle Gage
I have coveted the expensive digital levels for quite a while, but didn't know how well they would work. Along came the opportunity to buy this cheap and cheerful little brother to those tools and I decided to go for it. I forget which supplier I got it from, but the cost was minimal--$39 or something similar. I stuck it in a corner and figured I would forget about it until I needed to use it. Along came a thead on HSM which motivated me to haul it out and run a couple of tests to see how well it works. Here's a little photo review essay. Suffice it to say that the device seems pretty accurate and is a keeper!
They're cute little goobers, aren't they:
Digital "Bevel Box" shows my granite surface plate is level...
My 1-2-3 block is level...
Now there is some question. I think this arrangement may be level but the Bevel Box is cocked a little bit. You sometimes need to jiggle it to get it to sit flat...
My 30 degree angle plate is surprisingly, ahem, 30 degrees...
Lathe is level according to Starrett, what will the Bevel Box show?
You can see the awful truth on the rhs 1-2-3 block. The gage block was hidden under the Starret level and I need to get my lathe shimmed up!
When the Wrong Bearings May Work for a Spindle or Ballscrew or, How You Can Make A Mill from a Drillpress
Any casual reader of CNCZone will eventually run across one of the famous bearing rants for either spindles or ballscrew mounting. Closely allied are the drill press mill rants. Some noob will inquire with much enthusiasm how to go about converting the Asian drill press they just got for $39.95 into a CNC mill capable of slicing through solid green kyrptonite at 300 ipm with an accuracy of 10 microns and the old hands will just come unglued at the absurdity of it all. While this can be entirely entertaining to watch, one does feel a bit like the beginners are receiving an initiation flogging they don't really deserve and are ill-equipped to understand.
The bearing question is similar. Someone wants to mount a ballscrew or spindle in "ABEC7" skate bearings that were purchased cheaply on eBay and the ranting from the old hands starts in again. Pretty soon the fur is flying and we're talking about the need for $800 20TAC47 bearings on an Asian mill that didn't cost that much more than that and everyone wonders how we got there.
What's funny is that every now and again, someone actually manages to do what the experts have said is impossible. For example, there is a Mech E professor that has built a pretty nice little milling machine from a drill press:
You too can build a milling machine from a drill press...
Someone commented that this design was a nightmare and the guy obviously didn't know what he was doing because he had installed 5 deep groove ball bearings and a single tapered bearing that was in backwards of all things. You just can't do that--it ain't right!
"Hmmm," says I. When I hear that you can't do something, I kind of want to know why the guy did it anyway and how well it worked. It occured to me that perhaps this guy was clever like a fox. I sniffed around his site a bit more and learned he was a Professor of Mechanical Engineering with full Piled Higher and Deeper credentials. Now I am nto one who is intimidated by credentials having attended graduate school and met many of these sort of fellows. At the same time I do not immediately assume any PhD is an idiot either. This guy piled on 5 deep groove and 1 upside down tapered roller bearing for a reason, and it became my mission to figure it out.
It didn't take me too long to decide that maybe he was just stacking the bearings to make up for their inherent weaknesses. One often hears about stacking 3 or even 4 angular contact bearings to increase rigidity. So I dragged out my bearing catalogs and had a look at what this might mean.
From the NSK bearing catalog I found a nice comparison of the strengths of various kinds of bearings. We can see that deep groove bearings are primarily limited in that their load capacity is not as good as angular contact bearings:
Deep Groove Ball Bearings
Angular Contact Ball Bearings
Tapered Roller Bearings
Radial Load Capacity
Axial Load Capacity
Fair in Both Directions
Good in One Direction; Takes 2 bearings for 2 directions
Good in One Direction; Takes 2 bearings for 2 directions
Excellent: All tolerance classes available
Excellent: All tolerance classes available
So I decided to try to set up a comparison of the radial and axial load capacities for similar sized bearings of different types. There are formulas in the NSK catalog that may be used to compute the load capacity of up to 4 stacked bearings:
You can see that axial loads are additive but radial loads don't get 4x the value when you stack 4 bearings. In fact they aren't even 3x as strong radially when 4 bearings are stacked. That's going to be the weakpoint I suspect. The results are interesting.
Multiple 6204’s Back to Back
6204's are standard deep groove ball bearings, typically considered wholly unsuitable for spindle and ballscrew use. You can buy plain vanilla 6204's for $7.69 apiece while ABEC7 quality 6204's are $77. Here's what you can get by stacking them:
Consider that for the ballscrew application the load is going to be largely axial as we are trying to prevent the screw moving along its axis and introducing backlash.
So how do they compare to equivalent angular contact bearings?
The plain vanilla angular contact equivalent of a 6204 is simply a 7204. Your basic 7204 costs $23.88 so already we could have bought 3 6204's for the price of a single 7204 and for two 7204's we can surely stack up our 4 6204's. A matched duplex pair of ABEC7 7204's are a cool $200.
Here are the specs on stacked 7204's:
Guess what? The double AC bearing configuration is bested by a triple 6204 bearing arrangement! Now if we want ABEC7's, the duplex AC bearings are still a bit cheaper, but if we're fooling with more "stock" bearings, it seems like we can get a more rigid arrangement for less money using the deep groove bearings.
I'm sure it is probably not quite so easy, but it is certainly intriguing. It wouldn't cost much to build a test rig and see how well the deep grooves perform when stacked. Now I'm sure the bearing gurus are spinning up to full whirling dervish speed to jump all over this concept, but I remain unrepentant until I see someone hook them up and make them play.
What about the really expensive bearings?
A duplex pair of the much vaunted 20TAC47B purpose-built for ballscrews angular contact bearings turns in an axial load value of 26,600N. That's better than 3 stacked 7204's! However, note that the quadruple 6204's begins to approach this value at 26,400N. Also note that 20TAC47B's cost $800 the pair.
Can 4 of these cheap bearings do ballscrew duty as well as the $800 TAC's? That's a scary thing to spring on the bearing gurus.
It remains to be seen, but I would sure love to try the experiment someday!
No, but it would have been an appropriate post to keep people guessing.
Do you want to be a better CNC'er?
Get Better Tool Life, Surface Finish, and Material Removal Rates.