[sillythings]

Hi all,


I have a x2 mini mill for a while now. I'll be using this thread to track my progress as I attempt to improve the rigidity of the column.

I've done a lot of reading on what other folks have done. This have given me a lot of ideas. One area that I like to add my contribution is to quantify via repeatable tests of my progress. I also hope to identify the source of the problem that is backed by test data.

I will be documenting my effort over multiple posts with photos. Each picture is full resolution. You can read my dial directly if you look at it in full resolution.

Also, WARNING, I had to modify the clamp I used to hold the weight in my test. Before my modification, the clamp will slip and drop the weight! 20 pound of dumbbell weight can definitely break bones! See the bottom of this post of the test setup.

Anyway, I look forward to hearing your idea and suggestions.


***

To start, I've measured the baseline flex in the machine. All future efforts will be compared with these:

Test load: 20lb
Torque arm (diagonal from dumbbell weight to column base joint): 2.5ft.

Front column flex: 0.004"

Attachment 249656



Left column flex: 0.0005"

I am really surprised that this is so good!! I've rechecked multiple times over multiple days this test.

Attachment 249658


Right column flex: 0.0035"

Attachment 249662

back column flex: 0.0035"
Attachment 249660




Conclusion:

Needless to say, considering the small 20lb load, the numbers are really disappointing. The one bright spot is the left side flex of 0.0005". I have not figure out why it is so much better. But it does suggest the whole thing can be improved a lot.


Test setup:

My test setup came from metabo in this post
Home Model Engine Machinist - View Single Post - Reducing Mini-Mill Column Flex (and Column Y-Axis Alignment)


This is a really elegant way to achieve repeatable tests! It is way better than pushing the column by hand! Furthermore, I don't have to pull/push the column while at the same time trying to work my camera! It is brilliant!

However, there is a real danger that the stock clamp (the type depicted in the photo) will slip when 20lb are put on one end. The stock "live jaw" will give out under high load. This has happened to me. I modified the clamp to handle much higher load by:

1. drilling a handful of of holes in the steel bar
2. drill a hole in the "live jaw"
3. used a thick pin (see picture) to prevent the live jaw from slipping. (a thinner pin can snap)

When in use, please make sure it is the thick pin that is taking the load and not the 2 small metal plates that came with the clamp.

Please see photo below for how my modification works.

Attachment 249664

Even with this modification, to be honest, I am still not 100%comfortable. Please be careful and consider other safe guards.

[Fastest1]

Interesting tests to say the least. I appreciate your efforts to quantify the flex.

I do understand establishing a baseline, your tests appear to have the weight much further out from the column than the spindle is. Did you measure the the flex with the weight at a similar point, just for the knowledge?

I look forward to your approach on a solution.

[sillythings]

For the baseline, I always put the weigh at the end of the clamp and the spindle at the same height.

I put the weigh further out from the spindle to reproduce a reasonable amount of torque on the vertical column in a safe manner. Toque=Force X torque arm. By using a long torque arm, I can use less dumbbell weight.

[Potatohead908]

Yes 20 pounds of steel falling on your foot sucks! I just had a 3/8"x6"x24" steel plate (about 20 pounds) fall on my foot on Thursday from 3.5 feet height. Landed on edge in front of my ankle in the fleshy part and now have bruising all the way around the ankle and to my toes!

[rythmnbls]

Anyway, I look forward to hearing your idea and suggestions.

Here's my solution, A bit over the top but the results were excellent.
The link is to a series of photo bucket pics of the upgrade.

Mill Photos by madluther | Photobucket

Regards.

Steve.

[Fastest1]

Here's my solution, I bit over the top but the results were excellent.
The link is to a series of photo bucket pics of the upgrade.

Mill Photos by madluther | Photobucket

Regards.

Steve.

Nice work. I would say that qualifies for overkill.

How have those acetal nuts worked out for you in a machine of that size?

[rythmnbls]

The acetal nuts have been pretty good so far, The Y axis nut is a fraction loose and has about .001" of backlash.
The X axis nut is a bit tight and has no backlash at all.
Making them has a bit of a learning curve, I expect the Z axis nut will be pretty close as I get more experience in making them.
The best part is they are easily made (and re-made).

They are a big improvement over the stock screws.
I especially like the .1" pitch, the stock screws were .05" pitch and it felt like you were cranking your arm off to move the table any distance.

Steve.

[Fastest1]

If you can get down to .001 or better, you are doing great. Backlash comp can easily address that. I get between .004-.007 on the LinearMotion screws. Never reballed them though.

I have some Haydon Kerk screws (Claimed 6mm, I think it is a 1/4-20 with a coating) and am going to use Moglice to cast them into my Sherline lathe cross slide. Along with the Kerkote already on the screws they should turn pretty freely. I have a few more tricks for it, again. I would like to get a more aggressive lead. Will start thinking about the Z soon. It is a 3/8. Should give me a few more options.

[sillythings]

Source Of Flex


The following is my analysis of the sources of flex in the mini mill Z column and the magnitude of their contribution to the total flex.


Attachment 249848


The sources of flex are:

1. The flex of the long z column itself (red box)
2. The right angle bracket (green box)
3. Everything inside the joint area (blue box)
   
   1. the big nut
   2. section of the vertical column under the big nut
   3. the surface that joint to angle bracket, etc...
   4. these are so tightly coupled that there is no easy way to separately analyze them so I grouped them together.

1) Measure Z column flex (red box)

Here is my test setup. Notice my dial indicator is attached to the lower portion of the column and measure the movement of the spindle that is higher up on the column. Any measured movement can only come from column flex

Attachment 249850

Here, the dial reads ~0. I am satisfy that the Z column flex alone is small. There are much bigger fish to fry elsewhere.


2. Measure flex of the right angle bracket (green box)


Notice my magnetic dial indicator is directly attached to the right angle bracket wall and I measured against the mill bed. This is the key to how I isolate the measurement from the rest of the system.

Attachment 249852



The dial reads 0.0015".

This means there is rotation at the base of the right angle bracket that is attached to the milling bed. The bottom of the angle bracket base is very narrow. Please see the yellow line in the first picture.


3. Measure flex at the joint (blue box)

I setup this test by directly attaching the magnetic base of my dial to the side wall of the vertical column alone.

Attachment 249830


Given the location of the dial attachment, this reading is the sum of the flex from both the blue box and the green box. The total dial reading of change in height to the bed is 0.001"

Hmm, why is this reading less than that of green box alone?!! After repeating the test multiple times, this reading is correct. I had to think hard how this happened.

This is actually caused by 2 bars in different position rotating about the same center. This picture should clarify the issue. The black line represents the measurement of the flex at the angle bracket. The green line represents the measurement of the flex at the join. My dial indicator is attached to the end of the 2 lines.

Attachment 249842

There is no reduction in flex at joint vs the angle bracket. The different readings are simply caused by different position of 2 bars rotating around a common center.

There is another significant result hidden here. Again, notice the measurement at the joint was strictly less than at the bracket and significantly so. This is a good thing.

Remember: measurement at the joint = flex from the angle bracket + the flex from the joint components.

If the joint measurement were equal to or greater than the angle bracket measurement, it would mean the components at the joint added more flex. As it stands however, since the measurement at the joint is strictly less than at the angle bracket (and significantly so), there is no evidence that the stuff in the joint added more flex.


Summary:

There is flex rotation centered at the base of the right angle bracket (green box in first picture). The above measurements are taken close to the center of flex rotation. The baseline measurements are taken 2~3X the distance away. Applying the distance multiplier to 0.0015" get us close to the baseline flex measurement (see my first post).

While I am satisfy with this finding, I recognize my test setup here is still a bit rough. It is hard to carefully isolate things, measure them, and relate the results with other results. At any rate, the numbers are rough but I think we are in the right ball park.

I know that others have suggested the joint area, especially the big dish shape washer, is a major area of flex. Still others have suggested warping of the Z column near the big nut as the problem. However, I have not seen any objective measurement that showed flex. I would greatly appreciated it if other folks can put up some measurements in this area of the mill. I can definitely be wrong in my testing and I'd love to know.



Final notes about these tests:

For the above tests, I had to disassemble the mill to properly analyze each component. The numbers on this page, since the mill has changed from baseline, should not be compared directly with the baseline numbers. The numbers on this page should be used as is for identifying the source of the flex. When I complete the upgrade of the Z column, I will put everything back and redo all measurements.

Also note that for each of the above tests, I always retest multiple times. I also verify that the dial returns back to zero after the dumbbell weights are removed each time.




[Update]

Base on suggestion from Blight, I used a better measurement method for the joint area. The new method isolates the flex to joint area only. The reading for the joint area alone is 0.0007". The summary above is no longer correct. There are flex in both the angle bracket and the joint area. This retest is documented in a follow up post.

[The Blight]

A quick note on no3. Place the magnetic base on the angle bracket, and the dial indicator tip at the top of the column. This way you have isolated the measurement to the connection points between the angle bracket and the column. You have already verified that the column has close to 0 flex, so any readings here will show how much the connection is flexing.

It's nice to see some data on this. Great work so far!

[amyers]

Summary:

There is flex rotation centered at the base of the right angle bracket (green box in first picture). The above measurements are taken close to the center of flex rotation. The baseline measurements are taken 2~3X the distance away. Applying the distance multiplier to 0.0015" get us close to the baseline flex measurement (see my first post).

many people already came to that conclusion long ago, see a number of fixes for it here. Shop Info

[sillythings]

A quick note on no3. Place the magnetic base on the angle bracket, and the dial indicator tip at the top of the column. This way you have isolated the measurement to the connection points between the angle bracket and the column. You have already verified that the column has close to 0 flex, so any readings here will show how much the connection is flexing.

It's nice to see some data on this. Great work so far!

Thanks! This is an excellent approach. I'll put up my findings.

[sillythings]

many people already came to that conclusion long ago, see a number of fixes for it here. Shop Info

You probably have not read my first post.

I already read the link in your post and many others before the start of my effort. In fact, from reading what others have done, some improvement are not reproducible and others may even hurt.

My goal is to make improvements that is backed by measured results. I also hope the data I gather will be useful for others.

[sillythings]

A Much Safer Way To Apply Test Load


I came up with a much safer way to apply test load.

Attachment 249890

Now if the clamp should slip, the dumbbell weight will drop 1" onto my bench! No more worries of breaking bones.

For those who haven't read my first post, please note that the clamp in the picture has been modified to handle larger load. Anyway, please be careful.

[sillythings]

Retest of Joint Area Flex (blue box)

This is a follow up to the my post on identifying the source of flex.

Attachment 249898


Test setup:

Attachment 249894

Notice the magnetic base of my dial indicator is attached to the angle bracket. My previous post has establish the flex in the column itself is minimum. Therefore, this measurement isolates the flex to the joint area.

The dial reading for joint area test is 0.0007


Just to double check, I retest the flex of the angle plate itself (green box).

Attachment 249896

The dial reading for the angle plate test is 0.0015. This is the same as before.

Conclusion: both the angle plate and the joint area have flex problem. The angle plate flexes approximately twice as much as the joint area.


Notice I am using the new safer way to apply the 20LB force documented in a previous post.


Thanks again to Blight for your suggestion on how to take the joint measurement.

[bhurts]

From the point of view of someone who doesn't own one of these mills I can't see how any of those measurements will help improve the cutting performance. I think the previous posts on this testing and the fact that the improvements made didn't help and sometimes hurt shows its the wrong direction for improvement.

There are 4 major factors that are being ignored that are way more important for performance.

1. How much deflection do you get at the tool when loading the spindle from each direction.
2. How much deflection of the z axis do you get when plunging and pushing on the tool in with the other axis.
3.How much deflection do you get on the x axis when it's acting against the z and y axis.
4. How much deflection do you get in the y axis while being acted upon by the x and z axis.

These seem like much more important questions to answer if you really want to quantifiably improve performance. From your measurements the column is a minimal source of flex. I would bet you get equal to or more flex from each of the moving axis that have been removed for these tests. If it were me I would worry more about the short comings of each individual axis and the spindle before determining what will actually improve performance.

Ben

[sillythings]

From the point of view of someone who doesn't own one of these mills I can't see how any of those measurements will help improve the cutting performance. I think the previous posts on this testing and the fact that the improvements made didn't help and sometimes hurt shows its the wrong direction for improvement.

There are 4 major factors that are being ignored that are way more important for performance.

1. How much deflection do you get at the tool when loading the spindle from each direction.
2. How much deflection of the z axis do you get when plunging and pushing on the tool in with the other axis.
3.How much deflection do you get on the x axis when it's acting against the z and y axis.
4. How much deflection do you get in the y axis while being acted upon by the x and z axis.

These seem like much more important questions to answer if you really want to quantifiably improve performance. From your measurements the column is a minimal source of flex. I would bet you get equal to or more flex from each of the moving axis that have been removed for these tests. If it were me I would worry more about the short comings of each individual axis and the spindle before determining what will actually improve performance.

Ben

Please see my first post on this thread where I stated my goal and detailed the baseline measurement of flex at the spindle. To say the least, when the spindle can move around +-0.004" (or 0.008"), there is a problem!

[The Blight]

Ben, from an engineering point of view it's best to isolate the many sources of deflection before working out a solution to the whole problem. When you have enough data, you can begin to work out solutions for the various areas, or combine several areas into one and find a solution for that as a whole.

If you repeat this process for the x, y and z axis you would end up with a machine that you know how it behaves under cutting loads. Right now he is working on the base construction of the machine. Once he has data on this, he can measure the deflection of the other parts. Because he knows how much the base construction deflects under these loads, he can then calculate how much the individual component of the axis is deflecting. Measuring the total deflection leaves you with a lot of guesswork as to what is actually moving. You can see the result of such speculation on all the various sites about column flex, flex in the washer..etc.

Saying that these major areas have been ignored is correct, but he might have a solid reason for doing so. Focus on one variable at a time.

Sillythings, I have one of these myself, but I lacked the patience to do all the measurements when I converted it to CNC. I just assumed the worst, and worked over all the contact surfaces until I got satisfactory results on a test cut. I wish I had gone about it the same way you have done, because it might have saved me some time.

[bhurts]

I did read your first post and all the posts in this thread. The point I am trying to make is you are not measuring where it counts. All of your measurements are based on hanging a weight on the column. If you want to improve things all of your measurements need to be based on what happens at the tool and table. That is where all the action happens. Better performance may come from some improvement at the column. I doubt you will see those improvements without addressing the other issues first.

Ben

[Eldon_Joh]

the only measurement of interest here is the stiffness at the root of the column, where its bolted to an angle plate, which is a disaster mechanically.


torsion of the vertical column from forces on the cutting bit in the X direction will probably be your weakest axis. this is a combination of the column flexing and the Z axis floppyness.

i recommend this type
http://www.hossmachine.info/images/r...le%20plate.jpg

however, that plate the whole thing is sitting on, should be either a rechtangular box (two c channels bolted together to make a box)
or much, much thicker. this topology will also allow you to tilt the Z axis

this topology is interesting, but still not very ridgid.
http://www.hossmachine.info/images/b...s%20column.jpg


here is a google sketchup file for my mill i'm working on.
http://johansense.com/bulk/taig/milling%20machine.skp

here's an idea for you.
http://johansense.com/bulk/taig/minimillx2.skp

bolt all of those sections together.
for the inside corner you will need to either weld them together, or bolt a flange to the inside and then bolt the flange to the base.
slap a piece of metal on both sides and run a crapload of bolts through the sections, its only sheer stress.