[wizard]

Thanks guys for your replies. By "ball screw stretch" I mean lack of stiffness. The screw compresses or stretches when pushing/pulling the load. Some people that have done RF45 conversions have complained about screw stretch.

There is more that one way to look at stiffness. For most users stretch isn't a huge issue though the ram growth might be. In any event is should be fairly easy to calculate stretch due to mechanical loading. What may be a bigger issue is lead screw twist due to torsion applied. This is a very real issue when working on high precision machinery though I'm not convinced it is a huge issue with a router type machine.

Twist is a huge problem when trying to interpolate curves to a high degree of precision. I know of some cases where tool path modification is used to correct for the lead screw unwinding. This is done for high precision optical work but again I can't see it making a huge difference on a router with reasonable sized components.

The reason I bring this up is because if you know precisely what is causing your error you can correct for it given sufficient metrology hardware and the requisite programming skills.

I did get the High Performance Machining book. It has provided some valuable insights. I am definitely building an "under $100k" machine that the book talks about. Ha! But seriously, I will do another post about what I learned in that book.

I switched my design to using servos because the stepper's torque tops out <1000 RPM, and going with any larger ball screws puts you over what the steppers can do. If that wasn't an issue, everyone needing a powerful push could just use 1600 oz/inch steppers, but they are terrible at higher RPM's.

Servos use to be far more expensive than what we are seeing these days. Given that it isn't an automatic no in a custom router. Even then I'd still consider carefully what makes sense, higher leads or higher RPMs.

I was going to argue about total mechanical advantage including the step down pulleys, but OK, I get it. You want as much torque as you can get for quick small movements, which is what the machine is doing most of the time. I see both of you telling me the same things.

We are saying approximately the same thing because it makes sense in the general case. You have to think about your specific use case though. If you aren't convinced this is the age of YouTube, plenty of videos machining related as such search for those that are similar to your interests. For me the question comes down to is it worth the extra expense to engineer for wild rapids? Obviously you don't want to fall asleep waiting for an axis to slew across its length. What you ultimately end up with, will likely be axis specific motors and transmissions.

I will go with 5mm pitch screws. It looks like:
Y (65"): 2505, max of 220 IPM. A C7 pair from LMB2008. I haven't seen anyone stocking larger ones.
X (41"): 2005, max of 470 IPM. A single screw, so I can try for a C5, or C7 LMB2008 if I have to.
Z (18.3") 1605, max of a gazillion IPM. A single also, C5 if I can.

There are domestic manufactures if LMB2008 can't deliver what you want. Which brings up a question you are a awfully concerned about stretch and other accuracy issues but what about your lead screws?

I am planning on glass slide encoders for X and Z. The 2 60" for Y are quite expensive, so maybe not yet for Y. I will do another post showing the slides in my design. They should help precision even with C7 screws.

2 for the Y? How is that?

For a router type machine I'm not sure glass scales are really worth the investment. I guess it depends upon your goals and just how good your mechanics end up.

I did some research on maximum ball screw RPM. Looking at a PDF, they show that the max rpm varies based on the nut design.
The general rule is (root diameter in mm) * RPM < 50000. This max speed is allowed for "short periods". Here is the max speed for the common sizes based on that formula:
16mm = (50000 / 13.3) = 3759 RPM
20mm = (50000 / 17) = 2941 RPM
25mm = (50000 / 21) = 2381 RPM
At these high speeds the lubricant starts getting thrown off of the screw, and you have to monitor the lubrication and nut temperature.

I'd go so far as to say that if you don't have automatic lubrication don't even bother with those high speeds.

This answers the question for my RM2005 Y. 2000 RPM for short periods is OK.


Assuming it didn't have to slow down, It would be going 1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9 = 45 inches in 9 seconds. Traveling at 9 IPS or 540 IPM.
Yes, I have to look at accelerations. Servo sizing is next! I will be looking at gear ratios, kilowatts, brush vs brushless, motor inertia, encoders, drivers, tuning, and accelerations. I had it made when I was just looking at steppers!

Err steppers still require proper sizing if you expect them to work correctly.

[steve123]

There are domestic manufactures if LMB2008 can't deliver what you want. Which brings up a question you are a awfully concerned about stretch and other accuracy issues but what about your lead screws?

My issue with stretch is that I haven't figured out how to calculate it yet, so I don't know how much of an issue it is. I feel like I have dealt with the issue by going up a size on X and Y screws.
On domestic manufacturers: I will check on them again when I get closer, but in the past I couldn't find one that would give me a reasonable price on the screw with machined ends. I have seen people buy Nook screws and then try to figure out how to machine the ends themselves. I just want to order them with end machining. LMB2008 does this very easily and cheaply.

As far as accuracy of the screws, I know I don't have the budget for brand new double-nut ground screws throughout. That leaves me with C7 screws, particularly for the X, which require a pair of long ones.

I messed up and switched X and Y when talking about 2 glass slides. Here are some pix.
Two 63" ones on the X @ $409.90 each (later!): Attachment 192290

One 36" on the Y @ $239.90: Attachment 192292

One 14" on the Z @ $209.90: Attachment 192294

These are from ZS SYSTEMS - Digital Readouts (DRO) & Glass Grating Linear Scales

You can see that the cost to add these to the Y and Z is very reasonable! It is certainly far cheaper than the extra cost of ground screws. I'm not saying that these are as good or better than having ground screws because I haven't seen them in action. Most people are using only Mach3, and don't have this option. It seems like a no brainer to use a Kflop and these slides. I do have to figure out how to shield the one in the Z, and I would have to build covers over the X slides, as they are facing up. I also have an option of squeezing the Y slide between the rails, above the screw. This would be more accurate.

Err steppers still require proper sizing if you expect them to work correctly.

I had spent a lot of time researching steppers, and felt like I understood them. Motorcalcs.xls helped a lot. Servos are new to me, so I am starting over. On the mechanical side, I have to find drawings of pulleys and mounts, and figure out how to mount them. I also have to figure out what acceleration I want (something I should have already done), and what size motor will work for that. I'm actually looking forward to working with servos, although I hear that tuning can be a pain.

[arizonavideo]

You can direct drive the ballscrews with a servo if you use a larger one. I used a 32mm Hiwin 7 TPI and a 90V servo and Gecko drive. I did a test video at the time.

IH table speed test 300 IPM.MP4 - YouTube

The screw is double nut and it did get quite hot. I never really go over 125 IPM in real life.

I just got my first China ballscrew and it looks really good. I have not checked it for deviation but the backlash is zero.

[erikjgreen]

You can direct drive the ballscrews with a servo if you use a larger one. I used a 32mm Hiwin 7 TPI and a 90V servo and Gecko drive. I did a test video at the time.

[link to video deleted]

The screw is double nut and it did get quite hot. I never really go over 125 IPM in real life.

I just got my first China ballscrew and it looks really good. I have not checked it for deviation but the backlash is zero.

Well, you *can* run ballscrews at high speed, but if you're worried about "stretch" you shouldn't. I believe a length change due to heating would be far greater than one due to stretching of the metal. If people converting desktop mills have complained about stretch, then they're either running undersize ball screws or just dreaming

Here's a (sales) article from Heidenhain showing information on heating of a ballscrew in use:
http://www.heidenhain.de/de_EN/php/d...-0209/file.pdf

The thing about using glass scales is that while they do improve accuracy, to do so you either need a controller that can handle dual encoder input per axis (LinuxCNC can sort of do this, some of the Galil controllers can too) or you need a very "stiff" drive train, like direct drive, precision gearboxes, or well tuned timing belts. What happens using glass scales is that the PID loop that controls positioning works based on feedback error. If your axis isn't where it should be based on scale position, it moves the axis to the correct position. If there's any backlash in your drive train (like even in a coupler between shafts) then the scale position won't change until the backlash is taken up, by which time the axis is out of position on the other side of the mark. The controller then corrects again and overshoots again in the other direction. It keeps doing this forever. This is called "oscillation" or "hunting" and it is the main thing you have to eliminate when tuning a PID (servo) loop.

The same thing happens if your drive train isn't stiff... like if your belt drive stretches a little before springing back to the correct position. Even if given time it would end up in the right place, the fact that it doesn't do so instantly means the controller corrects more, which when the belt springs back puts you in an overshot position again, causing it to keep correcting and oscillate.

So while the glass scales will work easily if you have a zero backlash gear drive (expensive) or oversize servos using zero backlash couplers for direct drive, with any drive train having backlash or stretch you'll do a lot more work to tune it or else have to damp down performance far below what you want (oscillation also goes away if you turn down gain, which is the same as reducing the amount of torque/acceleration your servo can use).

On the bright side, the glass scales will correct for positioning error due to ball screw variation (essentially "improving" ballscrew grade) and heat expansion or stretch. The scale gives the controller a reference position that's accurate and won't change.

However, getting accurate positioning isn't as simple as just using glass scales instead of rotary encoders. If the rest of your machine isn't up to par, you'll spend weeks tearing your hair out trying to make the servos move only when commanded and not oscillate while still getting decent performance.

Using a servo large enough to direct drive a heavy table axis may mean sacrificing a lot of its speed/power in order to make things work. Eg. the servos I will install on my mill are 380V/13A peak, 6000rpm @2.5nM. They could theoretically direct drive my 5 lead ball screws, but the acceleration wouldn't be great (compared to what I'd get with a gear or belt reduction) and I'd have to limit their speed to 1200rpm or so (less than 25% power) to avoid shortening the ball screw's life (ball screws will lose accuracy long before they fail, so while the balls won't fall out on the floor if you overdo it, the accelerated wear will start to affect things fairly quickly). It would work with no backlash, but I'd be giving up part of the servo power I paid $$$ for. Also, if I use timing pulleys to permit the servo to use its full rpm, then my torque and acceleration are multiplied, which gives me shorter overall time and better accuracy.

If you add up the cost of glass scales, the cost of servos and a drive train that can move the table efficiently, and the cost (in money or time) of a controller that will handle the glass scale input and tuning of same, you start to see why many people don't use them.

One more note on scales... while they will provide a non-changing reference for positioning, they won't automatically give you accuracy and precision as fine as they can measure for cutting. For example, on a machine the size of the one you're building, if you install glass scales with 0.0001 (one tenth) resolution, even if you put in the effort and cost to tune the drive train to position down to that level you still can't use it for anything but measuring parts with a probe. The reason is that cutting forces put an enormous load on the cutting tool/spindle and push it out of position. Even with accurate positioning, your mill will simply be too small to not bend under large cutting forces. My own 3800 lb cast iron mill still deflects a measurable amount under heavy cuts (defined for me as cutting near the full 5 hp power limit of my spindle).

Even if the mill column and head doesn't deflect, the cutting tool itself might! Depending on cut load, any end mill will deflect when cutting, so most CAM programs can compensate for this. The author of GWizard has a nice article on it. Essentially, you want to use the largest diameter end mill you can. I try not to use less than 1 inch diameter mills myself.

So... if you want super high accuracy for your mill, you need to plan on A) High cost for glass scales, servos, drive trains, and controller B) High cost from high mass (weight) of the machine to damp vibration and avoid deflection C) Very light cuts that take a huge amount of time and D) Lots and lots of time spent tuning the whole setup for accuracy, all of which will have to be re-done whenever the temperature of the mill's environment changes by as little as 5C.

Incidentally, the deflection issue is one reason why there's a trend toward high speed machining in current machine tools. If you cut thinner chips using more rpm and less torque you get less deflection, and if you optimize the cutting path to keep the chip load on the tool maximized and even, you get really really fast material removal rates.

If I were building a mill the size of yours, I'd go with medium sized servos driving 5 lead ballscrews via timing pulleys at 2:1 or 3:1 ratio, 1000 count per revolution encoders on the servos, the feedback loop closed in LinuxCNC or a good controller, and as high powered a spindle as I could afford. Such a setup on a surface plate should be capable of being tuned for 0.001" repeatable accuracy in aluminum parts very easily.

Erik

[wizard]

Well, you *can* run ballscrews at high speed, but if you're worried about "stretch" you shouldn't. I believe a length change due to heating would be far greater than one due to stretching of the metal. If people converting desktop mills have complained about stretch, then they're either running undersize ball screws or just dreaming

I have to concur stretch will not be a factor in most small machine tools. Thermal growth can be an issue.

The only other thing that is a potential issue is twist of the ball screw shaft. However I've only seen this documented on some extremely high precision lathes. Twisting of the screw however isn't stretch which has me wondering if the two are being confused.

Here's a (sales) article from Heidenhain showing information on heating of a ballscrew in use:
http://www.heidenhain.de/de_EN/php/d...-0209/file.pdf

The thing about using glass scales is that while they do improve accuracy, to do so you either need a controller that can handle dual encoder input per axis (LinuxCNC can sort of do this, some of the Galil controllers can too) or you need a very "stiff" drive train, like direct drive, precision gearboxes, or well tuned timing belts. What happens using glass scales is that the PID loop that controls positioning works based on feedback error. If your axis isn't where it should be based on scale position, it moves the axis to the correct position. If there's any backlash in your drive train (like even in a coupler between shafts) then the scale position won't change until the backlash is taken up, by which time the axis is out of position on the other side of the mark. The controller then corrects again and overshoots again in the other direction. It keeps doing this forever. This is called "oscillation" or "hunting" and it is the main thing you have to eliminate when tuning a PID (servo) loop.

One interesting application I've seen used linear scale for course position with the controller switching to very high resolution rotary encoders for the high precision work. Counter intuitive but the machines worked well considering the technology of the time.

The same thing happens if your drive train isn't stiff... like if your belt drive stretches a little before springing back to the correct position. Even if given time it would end up in the right place, the fact that it doesn't do so instantly means the controller corrects more, which when the belt springs back puts you in an overshot position again, causing it to keep correcting and oscillate.

I spent a few years of my life working on machines that required perfect belt tension. Many a problem solved by adjusting the belt so it feels right.

So while the glass scales will work easily if you have a zero backlash gear drive (expensive) or oversize servos using zero backlash couplers for direct drive, with any drive train having backlash or stretch you'll do a lot more work to tune it or else have to damp down performance far below what you want (oscillation also goes away if you turn down gain, which is the same as reducing the amount of torque/acceleration your servo can use).

There are many factors here to consider, one should not automatically assume that linear scales will help. I think that is what you are trying to say. This is especially the case for router type machines.

On the bright side, the glass scales will correct for positioning error due to ball screw variation (essentially "improving" ballscrew grade) and heat expansion or stretch. The scale gives the controller a reference position that's accurate and won't change.

However, getting accurate positioning isn't as simple as just using glass scales instead of rotary encoders. If the rest of your machine isn't up to par, you'll spend weeks tearing your hair out trying to make the servos move only when commanded and not oscillate while still getting decent performance.

Not to mention the huge expense involved in linear scales, the installation and maintenance headaches and that they get in the way. The other problem is how do you close the loop. You may end up needing a higher resolution encoder than maybe you should be using if the controller uses the encoder for velocity feed back. If you want easy, the way to go is the more conventional motor and rotary encoder arrangement.

Using a servo large enough to direct drive a heavy table axis may mean sacrificing a lot of its speed/power in order to make things work. Eg. the servos I will install on my mill are 380V/13A peak, 6000rpm @2.5nM. They could theoretically direct drive my 5 lead ball screws, but the acceleration wouldn't be great (compared to what I'd get with a gear or belt reduction) and I'd have to limit their speed to 1200rpm or so (less than 25% power) to avoid shortening the ball screw's life (ball screws will lose accuracy long before they fail, so while the balls won't fall out on the floor if you overdo it, the accelerated wear will start to affect things fairly quickly). It would work with no backlash, but I'd be giving up part of the servo power I paid $$$ for. Also, if I use timing pulleys to permit the servo to use its full rpm, then my torque and acceleration are multiplied, which gives me shorter overall time and better accuracy.

If you add up the cost of glass scales, the cost of servos and a drive train that can move the table efficiently, and the cost (in money or time) of a controller that will handle the glass scale input and tuning of same, you start to see why many people don't use them.

Exactly.

On a router type machine they cold be seen as a waste of time and money.

One more note on scales... while they will provide a non-changing reference for positioning, they won't automatically give you accuracy and precision as fine as they can measure for cutting. For example, on a machine the size of the one you're building, if you install glass scales with 0.0001 (one tenth) resolution, even if you put in the effort and cost to tune the drive train to position down to that level you still can't use it for anything but measuring parts with a probe. The reason is that cutting forces put an enormous load on the cutting tool/spindle and push it out of position. Even with accurate positioning, your mill will simply be too small to not bend under large cutting forces. My own 3800 lb cast iron mill still deflects a measurable amount under heavy cuts (defined for me as cutting near the full 5 hp power limit of my spindle).

Even if the mill column and head doesn't deflect, the cutting tool itself might! Depending on cut load, any end mill will deflect when cutting, so most CAM programs can compensate for this. The author of GWizard has a nice article on it. Essentially, you want to use the largest diameter end mill you can. I try not to use less than 1 inch diameter mills myself.

On most routers everything else is deflecting too. It is a non trivial task to get a solid platform.

So... if you want super high accuracy for your mill, you need to plan on A) High cost for glass scales, servos, drive trains, and controller B) High cost from high mass (weight) of the machine to damp vibration and avoid deflection C) Very light cuts that take a huge amount of time and D) Lots and lots of time spent tuning the whole setup for accuracy, all of which will have to be re-done whenever the temperature of the mill's environment changes by as little as 5C.

Incidentally, the deflection issue is one reason why there's a trend toward high speed machining in current machine tools. If you cut thinner chips using more rpm and less torque you get less deflection, and if you optimize the cutting path to keep the chip load on the tool maximized and even, you get really really fast material removal rates.

If I were building a mill the size of yours, I'd go with medium sized servos driving 5 lead ballscrews via timing pulleys at 2:1 or 3:1 ratio, 1000 count per revolution encoders on the servos, the feedback loop closed in LinuxCNC or a good controller, and as high powered a spindle as I could afford. Such a setup on a surface plate should be capable of being tuned for 0.001" repeatable accuracy in aluminum parts very easily.

Erik

That is good advice! The only thing I see as a concern is the 0.001" accuracy, there are many factors that determine how well the machine will do. You may get good repeatability but absolute accuracy that is far worst.

[steve123]

You can direct drive the ballscrews with a servo if you use a larger one. I used a 32mm Hiwin 7 TPI and a 90V servo and Gecko drive. I did a test video at the time.
(snip)
The screw is double nut and it did get quite hot. I never really go over 125 IPM in real life.

I just got my first China ballscrew and it looks really good. I have not checked it for deviation but the backlash is zero.

If I did the math correctly, the max speed on that 32mm screw would go something like this:
I guess the root diameter is about 28.
50000 / 28 = Max RPM of 1786
1786 / 7 = Max IPM of 255

The formula indicates that you were exceeding the max speed by quite a bit, and the heating confirmed it.

The takeaway is that the max RPM is limited by ballscrew diameter due to heating and loss of lubricant.

My X's max of 220 IPM with a 2505 is something I'm still thinking about.
My Y's max of 470 IPM with a 2005 is lots more than I need, so doing < 350 should be fine.
I think my Z is too short to get up to really high speeds.

[steve123]

Thanks again for the feedback. I understand what you are saying about belts causing problems with servo tuning. I didn't know about that, but it makes sense. I'ts also interesting to hear that temperature can throw off servo tuning.
The High Performance Machining book also had lots to say about wanting to have high servo gain. I understand that the KFLOP can handle both rotating and linear encoders on one axis. That is something I have to be sure of. I will be looking at as many KFLOP threads as I can find. I also need to look at what servo drivers would work best with KFLOP.

I am mainly trying to set aside the space in the design for the encoders, so that I can add them if I want. I had to widen the Z a little, and move the Z ball screw slightly to make a spot. I could add an encoder for less than $250, so the expense wouldn't be that bad. The real expense would be in this research time (a good expense), and the setup and tuning time. I would do an encoder after the machine is up and running.

Here's a (sales) article from Heidenhain showing information on heating of a ballscrew in use:
http://www.heidenhain.de/de_EN/php/d...-0209/file.pdf

That's a very interesting article. It's funny to see at this time as they are trying to sell us linear encoders to fix the problem!

On most routers everything else is deflecting too. It is a non trivial task to get a solid platform.

I agree. I'm trying to spend the time now to design things stiff and damped.

If I were building a mill the size of yours, I'd go with medium sized servos driving 5 lead ballscrews via timing pulleys at 2:1 or 3:1 ratio, 1000 count per revolution encoders on the servos, the feedback loop closed in LinuxCNC or a good controller, and as high powered a spindle as I could afford. Such a setup on a surface plate should be capable of being tuned for 0.001" repeatable accuracy in aluminum parts very easily.

I will go in this direction. Thanks.

[wizard]

Thanks again for the feedback. I understand what you are saying about belts causing problems with servo tuning. I didn't know about that, but it makes sense. I'ts also interesting to hear that temperature can throw off servo tuning.

Must have missed the part about temperature impacting servo tuning. I never had to deal with that due to working in temperature controlled shops. It is very possible though especially on old analog servo amps than ran hot to begin with.

The High Performance Machining book also had lots to say about wanting to have high servo gain.

High gain can be important for good performance but I can also be a headache to deal with. Some of the machines I worked on had extremely high gains which made them sensitive to external disturbance. Some of the parts where only a few microns thick and any sort of glitch would ruin the part and sometimes the diamond. Your goal should be gains high enough to get the job done.

I understand that the KFLOP can handle both rotating and linear encoders on one axis. That is something I have to be sure of. I will be looking at as many KFLOP threads as I can find. I also need to look at what servo drivers would work best with KFLOP.

KFlop thread do seem to be hard to come by.

I am mainly trying to set aside the space in the design for the encoders, so that I can add them if I want. I had to widen the Z a little, and move the Z ball screw slightly to make a spot. I could add an encoder for less than $250, so the expense wouldn't be that bad. The real expense would be in this research time (a good expense), and the setup and tuning time. I would do an encoder after the machine is up and running.

My personal opinion is that if you go to linear encoders this takes the mill out of the normal DIY solution. At that point you are talking a considerable expense over what most people do.

That's a very interesting article. It's funny to see at this time as they are trying to sell us linear encoders to fix the problem!

Thermal issues are very real. In repetitive processes warming up the machine to thermal equilibrium is one way to get around issues with thermal growth. As I've indicated above on most router type machines linear scales would be a waste due to all the twisting and deflection going on.

I agree. I'm trying to spend the time now to design things stiff and damped.

I will go in this direction. Thanks.

[clmenz]

I got some 3x3x 1/4" steel for feet, made a divot for the bolt-ends to sit in, and leveled the plate to within .0002 per 10". I spent more than 4 hours on it by myself (next time get a friend to help read the level!). The problem was twist in the plate. I could back off the high-corner (front right) bolt until it rocked on the other diagonal bolts, but it still wasn't level on that side. I jacked it up and looked at the hard rubber pads underneath. These pads have threads in them like tire rubber. I saw that the plate has 3 points of contact. Two on the left side, and one in the middle on the right side. The one that was for the middle-right was not centered (probably moved in transport). I moved it back into the box marked on the bottom of the plate, and tried again. Better, but still not quite there. I moved farther it back several times and got it all level. The pad is half-way out of the box marked on the bottom of the plate. By this time it was midnight. The next day my quads are killing me.

I'm not satisfied with this setup for several reasons:
1. There is still very little weight on the back-left and front-right legs.
2. You can put weight on any corner and watch the level move (probably the rubber pads squishing).

I'm thinking that in order to use this as part of a machine, I would want 4 hard points of contact that I can adjust. That way I can take twist out, and have it be solid with lots of weight on any corner. I'm thinking of something like four 4"x6"x1/2" steel plates epoxied to the bottom, with a connection to the structure at the middle of each plate.

[QUOTE=steve123;1306326]Here is an updated design. This has a frame that wraps around the granite surface plate. There are leveling bolts under the plate. Once the plate is level, I will apply epoxy filler between the plate sides and the metal pads there. I will bolt the sides against the plate.

Attachment 191278 Attachment 191280 Attachment 191282

This design now has 3 x 5 x 1/4" rail support beams. They will be filled with EG.

The beams don't stick up as far as before, and they are bolted into the granite. That should be very stiff and dampen vibration very well.

The gantry now has short risers on the sides.
Attachment 191284 Attachment 191288

Here is the full assembly.
Attachment 191286

The Z can slide over the x rails. There is about 35" of x travel with the 1" steel spacer blocks. Attachment 191290

I'm still working on sourcing ball screws and servos. I looked for quite a while for C5 grade screws, and there isn't much available on ebay. There certainly are no pairs. I think it will be rolled ball screws. I'm thinking of mounting glass slides on all axes (at least the Z). Servos at the moment will be kellinginc.net 1kw brushed.

There is a hacker space opening in my city. This is going to be extremely useful. [url=http://www.vocademy.com]VOCADEMY ? The place to learn and ma

[steve123]

I have been taking machining and welding classes at Vocademy. I ran my designs past two machinists there and got their feedback. I have been able to simplify the design quite a bit.

Here is a quick overall view, with details to follow. Mmmm see the Tormach spindle and motor? Cutting steel? Yes!, TTS? Yes!
Attachment 250416

I'm going to start by showing you with the base and work up.
I decided to get rid of the beam in front of the granite plate, and just use the 3x4' plate as a work area, plus about 5 inches overhang in front. This allows me to use the granite plate's existing steel base. Please note that all of the holes in the top of the granite plate are mockups for now. The only holes that really count for now are the ones into the sides. The beams are mounted with 1.5" stand-offs. This allows the cutter to reach the edges of the plate.

Base:
Attachment 250406

Adjustability. This allows turning screws to raise/lower the ends of the beams.
Attachment 250414

Planned steps:
1. Fabricate the two X beams:
a. Purchase beams.
b. Measure actual inside dimensions, and design plugs for each end.
c. Get plugs and triangles water-jet cut.
d. Drill holes for upper bolt pipes.
d. Weld in plugs, upper bolt pipes, and fabricate angle-iron/triangle supports.
e. Get beams heat treated, put beams on a mill and true up the ends.
f. Possibly get beams skimmed flat (if I can find a shop that can do the length). Otherwise I will have to level with epoxy.

2. Drill 10 .75" holes in each side of the plate to mount the 3x5x.375" X rail beams.
3. Epoxy in 303 stainless rod, and later drill and tap the bolt holes.
4. Mount the X beams. Make basically level with surface plate (if not skimmed flat), otherwise make exactly level with plate. I can use my Noga holder and test indicator.

Things I'm figuring out at this stage:
I found places that sell the water swivel and diamond-coated bits.
I found a water jet place nearby, and will use them for lots of parts.
I plan to do the welding on the X beams (at Vocademy!), and then get them stress relieved. I have a name of a place that does stress relieving.
Find a machine shop that can mill 65.5" beams flat on one side.

That's it for now. Next time I will post the gantry design.

[chevydyl]

I have ran several of my own ideas past my machinist mentor. He's a journeyman tool and die maker. Has built alot of different machines. He keeps telling me "it's not gonna be rigid enough" Has shot down almost all of my ideas. One that he remains optimistic about is a stationary bridge like natemandoo built. He believes that EG filled steel tubes for the bridge may suffice for my machining needs. Aluminum, wood, and some steel work. I'm gonna use the Tormach 770 head. It offers 10k rpm. May upgrade the motor hp though. From what I see in the latest pics is your gonna have major issues with deflection. Your vises and work holding will probably be solid. That granite table is sweet. But I look at that gantry and the just looks like it's gonna give poor results. In the EG forum a guy did a bridge out of EG. It doesn't move but the mass and rigidity help keep the cutter where it is supposed to be. Just my 2 cents. Maybe for your work it will suffice. But what about that every once in awhile job outta stainless that you simply cannot do.... I have turned people away because I don't feel comfortable turning certain parts because of my lathes rigidity. Not that it won't do it. But I would be standing there for hours on the clock that would take the machine shop 30 minutes.

[nateman_doo]

One of my regrets was NOT filling all the beams with concrete. I filled the 4 vertical posts with concrete, but not the rest. The machine wasn't terribly rigid until I added bracing. The aluminum bracing was crap, so I had to go steel. With the extra steel bracing, it made a world of difference. I am using ALL STEEL construction on my next build. Aluminum slabs are now steel, and the Z support will be all steel. Aluminum will be reserved for things like motor mounts, and securing the ballnuts to the plates.

I do like my tormachs stationary bridge setup (first time I heard the term, so I will roll with it). Tramming is pretty easy. I have followed many peoples advise on improvements that I wish I implemented into my Tormach's design. Use of dowel pins came late in the game, so don't be afraid to use them. Like the bearing block. It will be one less area to flex if the bearing block is pinned & bolted to the plate. Then you just have to worry about your ACB's.

[steve123]

chevydyl:
My Tormach spindle will be the one from the 770 also, so 10k RPM should work for me.

I'm still considering adding EG/concrete to the beams. I am thinking I will make a bolt-on cover on the back end of the main X beams . Then I can tip them up and fill them, then cap that end later. I'm considering putting a capped pvc pipe down the middle with little spreaders to make the EG/concrete go mostly around the edges.

nateman_doo:
I have always enjoyed seeing your builds progress. You have gotten some beautiful results from "That Thing". I will keep your tips in mind about adding concrete and using dowel pins.

[chevydyl]

Idk steve. The actual bridge of the gantry looks rather flimsy for any type of steel cutting. Maybe it's just that your not showing that part of the design as a whole. But in order for that cutter to remain straight that bridge needs to be solid. And have some mass. I'm not speaking from experience I am speaking 3rd person as my mentor would say "ain't gonna be rigid enough" I seen one of his machines which was a twin spindle adjustable Gibs. He told me at 90 an hour there's was 13k in parts and labor no pics but I was amazed at the quality. Surface ground dove ways. Also had gib locks

[steve123]

The Gantry

The beams are .375" steel tube 3"x5"x45" The unsupported length is 36.125". They weigh 70lbs each. The green bar on top is a glass scale. The whole gantry will weigh nearly 200 lbs.

Attachment 251236 Attachment 251238

I'm shooting for a machine stiffness of at least 50k lbs per inch. The beams deflect 583k per inch horizontally at the center of the 36" unsupported span (wost case). The web between the beams is 1/4" steel plate. Vertically, the combined beams with web are 11" high. That gives about 5,580 lbs/inch vertical deflection (basically none). I don't know how to calculate torsional deflection on the beam weldment.

The ball screw (2005) is connected directly to the bottom beam, so there should be only a tiny sideways force on the top beam.

I don't have deflection calculations for the legs. I am considering making the pads that the bottom beam rest on out of .385 steel instead of the .25.

I found a place nearby (BK Customs) that will water-jet cut the parts for the legs and the webs. I will MIG weld it together, then get it stress relieved.

Truing the front face: I can put epoxy on it and lay it down on the surface plate. or I can get it milled if I can find a place.

Adjustability:
Attachment 251242

I have adjustment screws on the top of each side.

I am planning (so far) to direct-drive the Y ball screw with a DMM 750 watt servo. The inertia ratio is 2.0 (good!). This direct drive should make that axis stiff enough to use a glass scale for a closed servo loop.

The ballscrew critical speed is 2303 RPM. The Max linear speed (rapids) is 453 in/min (more than I need). The acceleration will be about .35 G on all axes.

[chevydyl]

now I see. those other pictures I couldn't really see what your plan was. not bad. i like that you used the 3/4 underneath, how come only 1/4 for those uprights that support the whole assembly? you've obviously done some calculating so i am wondering did you check with say 1/2in and find that 1/4 would provide all the rigidity that you needed?

[steve123]

I don't have any calculations for the legs. I figured that there were plates going the long way in the direction of all stresses, and that would do it. I looked at them this morning and said "what the heck" and beefed it up.

The base plate is still 3/4" (1/2" + 1/4" welded on top. The top one has waterjet cutouts for assembly).
The upper plate that the bottom beam sits on went from 1/4" to 1/2".
The plates inside it (with the holes in them) went from 1/4" to 3/8". I can probably take those to 1/2", or just remove the holes.
The tall outside plate and slanted inside plate went from 1/4" to 3/8".

Note: I will be welding in 1/2" plugs on the ends of the beams. There will be bolts and pins going into the ends of the beams in addition to the ones shown.

Attachment 251390 Attachment 251392 Attachment 251394 Attachment 251396

[chevydyl]

diggin the peg holes.... so are you gonna fill the beams? if you do remember to let them cure in the vertical position. idk maybe it would be fine perhaps better to clamp them to your surface table in the horizontal position with multiple clamps to ensure flatness.

[chevydyl]

another thought came to mind, you say there is no forces going in "long" direction of the beams, however wont there be stresses when the cutter is moving in that direction? trying to force the bridge sideways, that's one of the reasons i asked about the 1/4in selection, 1/4 is flimsy. i think your gonna yield MUCH better parts with your additions to thicker stronger plates. good job.

[steve123]

PEG HOLES: The holes that look like peg holes are for more bolts. I was planning to also drill through, ream and put in pins later in addition to the bolts. I have to look into how many pins are needed.

I'm still kicking around lots ideas for filling the beams. I would first get out a wooden mallet and whack it to see how it rings.
Some ideas I could try:
- Just fill them with dry sand (cheap, easy, maybe good enough?).
- Stand them up and put a plugged PVC pipe down the middle and pour the rest of the void full of EG/Concrete. I don't think they need to be filled in solid.
-Instead of filling completely full of EG/Concrete, stand them up and fill in 1-2", then put in some spacer material like foam/sand/perlite to fill 3", then fill in 1-2" EG/Concrete again. You would end up with vibration deadening partitions.
- Lay them down and fill one inside surface to 1/2 - 3/4", and cure. Whack them after curing to see how much attenuation it did. Then rotate and fill another surface. This would require a filling material that would flow. I would do this before truing the front surface of the beams.
- Little metal bb sized balls coated with polymer/rubber.
- Heavy oil

I like the sand idea best so far. I could then try something else if it doesn't deaden it enough when you whack it.