stiffness measurements cnc mk3

[routercnc]

I've been making some stiffness measurements of my machine - build thread here:
http://www.mycncuk.com/forums/gantry-router-build-logs/6988-routercnc_mk3.html

Initial basic measurements were made as follows:
Y axis jogged to centre of gantry (worst case)
X axis jogged away from home, position arbitrary
Z axis lowered 50mm from home position, in a typical machining position
DTI clamped to bed with probe touching round section of collet nut (just above the flats)
Force of 50N applied (5kg) by hand using digital kitchen scales - guide figure for machining aluminium
DTI reading taken

X direction:
11575
Y direction:
11576
Z direction:
11577
(although this photo looks like I jogged the machine onto the scales I did not do that - I pushed the scales up into the collet to take the DTI reading)

For the X direction I applied 50N and measured a deflection of 0.04mm. Stiffness is then calculated from 50/0.04 which is 1250N/mm. The other axes were calculated the same way and were:

X 1250N/mm
Y 1000N/mm
Z 2450N/mm

I was hoping for at least 1000N/mm to give 50N load and 0.05mm deflection during aluminium machining (simplistically).
Although I have just about achieved that the bed stiffness has not been measured (although I think that should be pretty good), and I was hoping for a bit more than 1000N/mm.

So exploring the Y axis stiffness in more detail I took measurements at various distances on the Y/Z axis components, starting at the tool location and working back to the gantry centre. In this way I could plot the loss of stiffness from the machine gantry all the way out to the tool location. These results are below, with a photo showing a ruler to help you see where the measurements were collected on the machine.
11578

Y axis plate, distance from gantry face 10mm, stiffness 5000N/mm
Z axis rail, distance from gantry face 32mm, stiffness 3333N/mm
Z plate, distance from gantry face 55mm, stiffness 2500N/mm
Tool Centre, distance from gantry face 130mm, stiffness 1000N/mm
These were taken along a line shown by the ruler.

You can see that the stiffness quickly drops off as you move towards the tool centre, and the importance of minimising the distance and maximising the design at the concept stage.

Out of interest I also measured the stiffness of the gantry sides, where they meet the gantry and found a stiffness of 10,000N/mm. So the basic gantry is fairly good, but the losses build up as you progress across the Y and Z axes and to the tool.
11579

I've got lots to finish on the machine but it has given me ideas about where to go next to get a bit more out of it.

Thanks

[Neale]

At a quick glance, it looks as if a fair bit of the loss of stiffness as you move out from the gantry towards the tool is actually coming from play or flexibility or otherwise in the Y and Z rails, plus a component from gantry twisting. Your first figure is presumably bending of the gantry, primarily, and then the other effects become larger as you move out from the gantry. Maybe the Z bearing spacing could be increased? Reduce thickness of Z moving plate to reduce leverage? Deeper gantry? I'm playing with ideas for my own design and just finished my first pass (on paper) at the whole Y/Z assembly so very interested to hear your thoughts. I think that by reducing clearances and cutting clearance grooves in moving and fixed Z plates you can just about fit a ballscrew nut in between while still using 15mm rails but it's all a trade-off, of course, as you are weakening the plates a little.
Thanks for giving the data - it's good to see real numbers to start to get a feel for what is happening in the structure.

[routercnc]

Hi Neale,

Glad it is of interest. I'm hoping that the method and setup is simple enough for others to try and compare to. I know Jonathan has measured something similar so perhaps others have too.

Yes the first figure (10,000N/mm) is the basic gantry stiffness and everything else drops away from this due to various factors you have mentioned. What you cannot see in these photos is that the Y axis ballnut screws to a wooden panel on the Y axis (!), so this probably accounts for some of the 10,000N/mm gantry side dropping to 5000N/mm on the Y plate. An upgrade to aluminium should help here. It's a carry-over part from an old design when I had less machining capability.

I've always felt I could have done a better job, especially on the Z axis, along the lines you have mentioned above plus other ideas. This data shows there is still room for improvement although I have been accused recently of only using the machine to make a better machine, and not actually making anything with it. Easy to get drawn in . . . .

[Neale]

There's a familiar theme - "Are you ever going to actually make anything?" "Well, as soon as I've finished this machine to make that tool to finish off those other bits I started a year ago last Christmas, I'll be ready to go..." OTOH, I have been able to respond to the "I need a turntable to display my embroidery in an exhibition. It has to do about 3RPM, it's Friday evening, and I need it ready and painted for Monday morning" kind of request so I just about keep my head above water re brownie points.
I know about wood bending - my current router is an MDF build (cold-rolled cow dung, as a friend unkindly called it), and you can judge the humidity in the garage by the sag in the bed. And the Z platform. And the gantry. I'm about to stiffen the bed with a couple of bits of steel tube I recently found, although given that the bed currently has the stiffness of crocheted cooked spaghetti, that won't be too difficult.
There are a number of build threads on this forum where people have gone to considerable lengths to build a really strong machine, and I have to agree that a machine that is a bit too strong is a much better result than a machine that is a bit too bendy. However, it's difficult to know where "acceptable" is on that kind of scale, which is why seeing some measured numbers helps. Is your machine currently usable? Can you relate the stiffness numbers you have obtained to actual cutting speed, DOC, surface finish, etc, results?

[routercnc]

I've machined some holes in the brackets for the bed (see my mk3 build log) but don't want to do much more until the limit switches, e-stops, extra X axis limit switch, etc is made and set up. It's too risky to run in this condition.

From that limited experience it feels like the machine is going to be reasonably OK for aluminium in terms of build stiffness, but as per my other post on 'machining aluminium with a 1.5kW spindle' I'm having problems relating to having to use too high an rpm and melting the aluminium to the tool. I was running at 0.5mm DOC, 300-600mm/min feed, and 10,000 - 12,000 rpm (!). The surface finish on the inside of the holes looked fairly good but I can't comment further without cutting doing something a bit more demanding.

I won't be machining anything else for a bit whilst I finish bits off so hopefully someone else can make similar measurements and comment further if you need a better answer sooner.

[routercnc]

double post

[routercnc]

. . . .Can you relate the stiffness numbers you have obtained to actual cutting speed, DOC, surface finish, etc, results?

I'm still playing around with DOC, F&S, but here are some photos of holes I cut in the bed to allow nuts to be added for clamping things down. Finish is not too bad - you can see the finish on the sides of the hole, and at the bottom of the hole.

11681

11682

[EddyCurrent]

@routercnc, thanks for posting your stiffness measurement method, out of interest are you using grease in the bearings ?

[routercnc]

No problem Eddy.

The profile rail bearings were second-hand (fa-system ebay) and arrived with grease in. I've never added any since (about 3 years ago!) as they seem to keep running smoothly. I don't get in the garage much, perhaps 1 hour per week, and most of that time I'm doing improvements or other non-cutting work.

I've greased the ballscrews every now and again.

[EddyCurrent]

I just tried mine now it's mostly together mechanically. Build thread here http://www.mycncuk.com/forums/gantry-router-build-logs/6565-ready-steady-eddy-23.html

Using a Mitutoyo digital DTI with 0.01 mm resolution, deflections with 5kg force applied as above.

X 0 mm
Y 0.04 mm
Z 0 mm

I think the Y deflection is due to the ball screw system because wherever I apply 5kg load to the Z carriage I get the same deflection.

[routercnc]

Hi Eddy,

Yes, been following your build - coming along nicely. So with those figures X and Z are >5000N/mm, and Y = 1250N/mm.

The lower Y stiffness could be the mounting angle brackets for the AC bearing but not sure why it is similar as you move around the Z carraige as the moment would be changing. I also notice your ballscrew is right at the back of the gantry which is another reason for it being a bit lower than the other axes as it is the furthest away from the tool with the highest moment compared to the other ballscrews. But if the limited cutting I have done on my machine (with a lower stiffness) is anything to go by it's not a concern.

[EddyCurrent]

I can't see the angle brackets having any movement at those forces but I do think the FK bearing might have play in it, I've greased it but not spent much time yet with fine tuning using shims. As you say it may not be a concern, also did Jonathan use 200N in his measurements ? and did he publish the deflections ? I didn't see them.

[routercnc]

Hi Eddy,

Yes, Jonathan used 200N but my scales were only up to 5kg so I could only apply 50N. The response will be linear anyway at these loads, and as the load is divided by the deflection it works out the same (4 times load gives 4 times deflection). The stiffness values will be the same in both cases. We also both used 50mm Z extension so the machine condition is the same and results are all therefore broadly comparible.

Jonathan has not published figures to my knowledge, just this info in the 'sufficiently strong machine' thread:
'. . . .By applying a force of 200N and measuring the deflection I found the stiffness in X and Y with the Z-axis at 50mm extension. I did the same test on my milling machine (Clarke CMD1225C column mill) and hence found that this machine is a similar stiffness to my milling machine, although a bit weaker in Y . . . '

Would love to see some figures if you have them Jonathan ! . . .and anyone else if you've measured at 50mm Z extension.

Going back to the freeplay in the FK bearings, if you only tested in one direction and took a few readings (as I did) then the freeplay would all be used up in the first test and should not be a factor in the repeat tests. It would only matter if you went back and forth. I wanted to seperate free play from actual machine stiffness so measured in one direction only.

I've currently got my AC bearings out of the housings ready to shim up as they have quite a bit of freeplay even when loaded.

[Jonathan]

Excellent thread...


I can't see the angle brackets having any movement

Put the indicator on it and you'll soon find out. The reason I suggested in your build log making your own 'better' bearing blocks is that you can make them stronger than the originals by incorporating the brackets and bearings in one solid block, or whatever seems appropriate. However, I expect the distance measured parallel to X between your spindle centreline and the Y-axis ballscrew is the main contributing factor.


As you say it may not be a concern, also did Jonathan use 200N in his measurements ?

Yes, I used 200N as the deflections would have been too small to conveniently measure using a smaller force. To get more reliable readings you should test over a range of forces then plot a graph of the deflection vs force. The gradient of that graph is your stiffness. If the graph turns out to be non-linear, that could help explain what's contributing to the compliance.


and did he publish the deflections ? I didn't see them.

Unfortunately I lost them - however I remember the stiffness measured was greater than my milling machine (by the time you include the table), so about twice the stiffness routercnc has measured. However, we need to be careful to all take the readings at the same point, as I did it by putting a 1/2" (or 10mm, can't remember) bar in the collet and positioning the indicator on the bar about 5mm from the collet. That means the readings include the stiffness of the spindle. If you just push on the spindle body or Z-axis then clearly the spindle stiffness wont be included and you'll get better readings.

Also, when I measured the 'sufficiently strong machine' stiffness it had less then half the bolts in the gantry, which might make a difference... :playful:
I've lent my friend (who has the machine) a couple of dial indicators, so hopefully we will soon have some more reliable readings.

[routercnc]

Look forward to the results from that.

You can see in my first post that the initial stiffness measurements were on the collet nut - DTI on the round part just above the flats and load applied next to it on the flats. So I did include the spindle in the measurement, but not the tool as that would depend on the tool used for each cut. You could argue the reference setup in many ways, but that's what I did. If you agree Jonathan, would you mind reading off the collet as described when you repeat? If Eddy has copied my method then we'll have 3 comparible results to get a sense of the range of stiffness vs performance vs design to add to the general knowledge on the forum.

[EddyCurrent]

Going back to the freeplay in the FK bearings, if you only tested in one direction and took a few readings (as I did) then the freeplay would all be used up in the first test and should not be a factor in the repeat tests.

Also referring to Jonathan's point about where the measurement was made, I applied the force at the end of the spindle, with the collet nut removed, and had the DTI on the other side right at the extreme of the spindle shaft. I removed the collet nut because being slack, there was play in it.
If this thread is going to be a repository for peoples measurement then maybe a set of rules should be made to ensure everyone tests in the same way.

[EddyCurrent]

I improved the FK bearing today and carried out the measurement again but this time using a spring balance like this one Rolson Pocket Spring Balance | Hand Tools (http://www.toolsdiy.co.uk/shop/view/hand-tools/pocket-balance-scales/?gclid=CPiL-tqAq70CFdShtAodiWoAkA)

With the Z carriage halfway across the gantry, Z lowered 50mm from home position, force applied to spindle shaft around flats for spanner, collet nut removed and DTI 2mm from end of spindle shaft.
Force applied = 200N (20Kg approx)

X deflection = 0.11 (1818 N/mm)
Y deflection = 0.15 (1333 N/mm)
Z deflection = 0.04 (5000 N/mm)

For Z I hooked the spring balance into the collet holder and applied slightly more force than 200N because I was pulling at a slight angle, I was going to calculate a force vector diagram taking the angle into account but just estimated it.
I managed to get the DTI under the spindle shaft so got a measurement including the spindle mounts.

Edit: Sorry, sorry sorry, I later realised thet because the DTI tip was resting on collet nut threads this was changing the reading as well, so I fitted the collet nut tightly and measured off that.

So new readings

X deflection = 0.09 (2222 N/mm)
Y deflection = 0.13 (1538 N/mm)
Z deflection = 0.04 (5000 N/mm)

[routercnc]

I improved the FK bearing today and carried out the measurement again but this time using a spring balance like this one Rolson Pocket Spring Balance | Hand Tools (http://www.toolsdiy.co.uk/shop/view/hand-tools/pocket-balance-scales/?gclid=CPiL-tqAq70CFdShtAodiWoAkA)

With the Z carriage halfway across the gantry, Z lowered 50mm from home position, force applied to spindle shaft around flats for spanner, collet nut removed and DTI 2mm from end of spindle shaft.
Force applied = 200N (20Kg approx)

X deflection = 0.11 (1818 N/mm)
Y deflection = 0.15 (1333 N/mm)
Z deflection = 0.04 (5000 N/mm)

For Z I hooked the spring balance into the collet holder and applied slightly more force than 200N because I was pulling at a slight angle, I was going to calculate a force vector diagram taking the angle into account but just estimated it.
I managed to get the DTI under the spindle shaft so got a measurement including the spindle mounts.

Agree we need a standard method so everyone can compare. Jonathan - since you intend to re-measure are you OK with the method posted above by Eddy? If so I'll also re-measure to the same condition although my machine has been partly stripped for fine tuning etc. so will be a little while.

[routercnc]

Hi Neale,


However, it's difficult to know where "acceptable" is on that kind of scale, which is why seeing some measured numbers helps. Is your machine currently usable? Can you relate the stiffness numbers you have obtained to actual cutting speed, DOC, surface finish, etc, results?

I've been cutting out a few 20mm thick upgrade parts (will post in the build log soon) and have taken some photos of the edges of one of them. These were cut using the following:
6mm carbide 2 flute
1.0mm DOC
600mm/min feedrate
occasional spray of duckoil

cutting in the X direction
(there was a fair bit of what sounds like cutter resonance in this direction)
12242

cutting in the Y direction
(much less chatter/resonance sounds in this direction)
12241


My X & Y axes are currently 1000N/mm and 1250N/mm respectively at 50mm Z extension (measured on the collet nut). But these parts were cut at a longer reach, closer to 100mm as they were almost on the bed so the actual stiffness for these cuts would be lower.

I don't have an airline at home so can't clear the cutter path and suspect some of marks are chip re-cutting damage. But some look like cutter resonance as they are so regular. This suggests if you regularly want to cut a lot of aluminium, and want a nice finish, then you need a stiffer machine than my mk3 machine, perhaps more like 2000-3000N/mm in X&Y @ 50mm Z.


For the latest part, the Z axis motor mount plate, I followed the cutter with the workshop vacuum cleaner to remove as many chips as possible and got a better finish in terms of surface nicks etc. It's better but the resonance type stripes are still present although in this photo you can't see them that well. Here is the start of the job pocketing for the stepper motor:
12243


Hope this helps.

[EddyCurrent]

I just put some figures into this, FSWizard - Free Advanced CNC Speed and Feed Calculator (http://zero-divide.net/index.php?page=fswizard), and it suggests 600mm/min is too slow, though I've not tried it myself.

[Clive S]

I just put some figures into this, FSWizard - Free Advanced CNC Speed and Feed Calculator (http://zero-divide.net/index.php?page=fswizard), and it suggests 600mm/min is too slow, though I've not tried it myself.

How did you put the spindle speed in? or have I missed it. ..Clive

[EddyCurrent]

It seems like you have to change the percentage figure in the 'SFM:' drop down and this calculates an appropriate spindle speed. The maximum spindle speed is set by entering a figure in the 'Max RPM' box.
I just selected SFM percentages until it gave me a decent spindle speed.

[Clive S]

It seems like you have to change the percentage figure in the 'SFM:' drop down and this calculates an appropriate spindle speed. The maximum spindle speed is set by entering a figure in the 'Max RPM' box.
I just selected SFM percentages until it gave me a decent spindle speed.
Eddy. How can you comment on the feed rate unless you know what Routercnc spindle speed was? in the first place as I didn't see it in his post. Just wondering. ..Clive

[EddyCurrent]

Well for a feed of about 1000 mm/min the suggested RPM was about 5000 so I guessed he would be using something faster.
So for 10000 RPM it works out at about 2000 mm/min

[Clive S]

Well for a feed of about 1000 mm/min the suggested RPM was about 5000 so I guessed he would be using something faster.
So for 10000 RPM it works out at about 2000 mm/min
Ah I see you worked it backwards, then yes that computes assuming he used a calculator in the first place:beer: ..Clive

[routercnc]

The rpm I use is around 12000 (200Hz). If I put the following into that calculator:

6061 T6 aluminum
solid end mill, carbide
6mm 2 flute
DOC1 mm
WOC 6mm
slot/pocket = yes
SFM, FEED 100%
max rpm 12000

I get a feedrate of 900mm/min. I'll keep experimenting with the next parts I cut out to see if I can get a better finish.

[Jonathan]

I get a feedrate of 900mm/min. I'll keep experimenting with the next parts I cut out to see if I can get a better finish.

600mm/min @ 12000rpm generally works well with a single flute cutter, so since one would expect the chipload to be slightly lower for a 2 flute cutter, as a first approximation the feedrate should be a little less than 1200mm/min.

What are you mounting the workpiece on? It can make a big difference to the finish - for example I found it easier to get a good finish when taking a small cut with my machine when it had the MDF bed. Presumably the MDF's damping properties helped. If the parts are small you'll obviously have problems with rigidity using MDF, but if they're a reasonable size and you add a finishing pass (take off 0.1-0.2mm from edge in one pass) you can get good results, e.g:

1225512256
(20mm thick parts and 25mm thick bed)

I've still not measured, but I'm pretty sure your machine is stronger than my 'old' one which cut the parts above (and below), so unless you have big problems with resonance you should be able to get similar results. Perhaps MDF is worth a shot? That's not to say you can't get a good finished clamping rigidly to aluminium, it just takes a bit more care/luck setting the feedrates, e.g:

12257

[routercnc]

Hi Jonathan,

If you look at the last photo in post 19 you can see that I generally use an 18mm piece of chipboard if I need to cut all the way through. The part I cut out in the last photo does actually have a better finish than the parts in the first 2 photos - possibly because it is cut out of a much larger sheet.

If you look at the bed you can see gaps between the profile sections used to make up to bed. The machine is used primarily for cutting wood and in that mode it has a large sacrificial wooden sheet which sits on the bed. This works well for wood machining.
If I end up doing a lot more aluminium I can see that I may need to either add a lot more profile sections to give less gaps, or invest in a large thick aluminium sheet to bolt to the bed, skim it, drill and tap lots of holes.

I'll keep experimenting with F&S as I machine out more upgrade parts, plus try some with a finishing pass.