My First CNC router, Servo + rigidity questions

[esm7]

I have been working on a CNC router for a long time, and I finally have the thing built with three axis of linear motion and a router mounted, and so I am ready to take the next step. This next step seems quite complicated, and wanted to see if I could get your opinions. I can sort of try to grind through the mat, but I am not sure from an experience perspective if my assumptions / conclusions make sense.

Goal of the design:
Mainly woodworking to facilitate cabinet and furniture making. Materials include sheet materials, hardwoods, and in the long term a 4th axis for doing interesting cylindrical parts. I started this project with the idea that if at any point I stopped I would still have something useful. So the first part of the design was as a workbench with drawers.

Dimensions / travel:
X: 70" , Y: 48" , Z: 12".

Basic machine architecture / construction: Gantry design from welded steel (then machined or ground). 3 axis of linear slides with the long axis (X) slides mounted on the underside of a frame.
The gantry is on top of a 4"X4" maple legs + carcass for drawers.
Attachments:

Schematic / CAD solid model:



Y axis gantry welded, then ground:



Gantry verticals (welded then machined)



Base Frame (welded then ground)


Assembled workbench + CNC


So now the big question is how do I go from here to CNC..
So I was planning on ballscrews and servos. For the ballscrews, i was thinking of 1" X 0.25 rolled.
I was thinking that 7" / sec was probably as fast as I really needed which is ~ 1680RPM or so. From the on line calculators, it looked like the max RPM is fine for fixed one end, and supported on the other. Also it seems like from the different motors available, this is too slow, and so I though of reducing the speed by a factor of 2 with a timing belt. (is a belt a mistake?)

Then the next step is how big should the servos be?
So I have the weight for each axis, and I can calculate force due to acceleration: (0.5 sec to come up to speed)
gantry = 240 lbs
Y axis plate = 34 lbs
Z axis plate = 61 lbs
When I do the calculations (taken from some web site) it seems like the force of the cutting dominates the force required by the servo.
(I think I am making a mistake)..
But for a 50 lb cutting force, I got 2.3 in *lbf of torque.
I had no idea if this torque or cutting force made sense. I don't think I have ever put 50 lbs on a router. When I used galil's calculator (http://www.galilmc.com/learning/motorsizer-start.php)
I got 2.2 inch lbs continuous, and 4.8 inch lbs peak (assumes 2/1 reduction)
Is this underpowered ?

I tried to calculate the cutting forces by assuming that the I was using something like a full 0.5 to 3 HP of power at the different speeds, and it looked like 50lbs might happen...


Then it occurred to me that I had no idea of how rigid my setup was, and if cutting with this sort of force even made sense. So I tried to measure the stiffness of the machine by pulling on it (with a rope and pulley / weights) and measure with a dial indicator how much it moved. I held each axis in place with breaks that I had designed. (sort of like disk break calipers on the linear rails)
X axis stiffness measurement:

Y axis stiffness measurement:

Z axis stiffness measurement:

Plotted results:


So it looks like I could expect the thing to move ~0.005" for 50 lbs..
Which I guess is ok for wood. I haven't seen stiffness numbers for other machines, so I am not sure how it compares.. though I had hoped that it would be pretty rigid given the design..

Then the next question was:
What servo motor would work with these sorts of requirements?

3000 RPM, 2.3 in lbs continuous, and 4.8 in lbs peak ? One that would work with a driver and also work with Mach3 ? It looked like the Gecko 340 might be an option, but then I was hard pressed to find the motor to match with this drive. I looked on line (ebay and some others) for many hours, and found hundreds of different servos, drivers, and controller configurations. Unfortunately most seemed to be of the resolver or absolute encoder feedback type.. Or they were monsters, voltages too high, or some brand that I had never seen, and could not find pinouts or basic specifications for..

Does anyone have any suggestions for reasonably priced servo driver combos that work with Mach 3 ?

so.. should I put a motor on each side of the gantry ?? Do I need to ?

So that is where I am now. Any comments / suggestions ?

Thanks
Evan

[Dale gribble]

Hi, machine is looking good so far, I am just finishing my router which has Hiwin linear rails on all axis ( zero play ), I see you have done some stiffness tests but when testing the y axis i would also check for racking as you are using a single lead screw. I done some similar trials without my lead screws connected on my Y axis and there was a lot of racking. This can be totally eliminated using dual lead screws. There is a cost to this set up an additional motor, lead screw, ball nut, coupling and drive are needed.

Regards

Dale

[ger21]

By using a timing belt for gear reduction, you can use a smaller motor to get the same torque at the screw. Gear it so the max rpm you need at the screw is achieved at 80% of the motors max rpm.

If you get a ballscrew with a lead of .5 or more, that'll allow you to use a servo with less torque, because you'll gain the torque back in the gear reduction.

[jsheerin]

I've been working on this kind of stuff for upgrading my router lately as well. I've been shooting for much higher speeds (~1000ipm) and higher accelerations (1g on the Z axes using ball screws, the x and y will be more like 70in/s^2 and use rack and pinion drive). So you're looking at 420ipm and 14in/s^2... I'd recommend you try to design so you can cut at 420ipm (or faster) and accelerate at least at 50in/s^2. I'd shoot for rapids of double that speed or higher (so there's no cutting load there - just acceleration and deceleration). If you look up chip load information for common wood products, you will want to at least be able to cut this fast to get proper chip loads on your cutters while spinning them at the full speed of your router.

Additionally, it's not vastly more expensive to get much more power out of a servo system, so why not?

Anyway, whatever final design numbers you end up with, in that range I'd look at a Gecko G320X drive with a brushed dc motor from Keling:
http://www.kelinginc.net/ServoMotors.html
Keling sells the G320X along with encoders.
And http://www.antekinc.com/lview.php?d[]=1&r=25 for power supplies should get you going, or one of Keling's.

Just to clarify, I haven't actually used these, but I've been looking at them to run my router. I'm going to need more power for my y axis though, so I'm currently trying to get a Fanuc A/C servo off ebay spinning.

[esm7]

Hi, machine is looking good so far, I am just finishing my router which has Hiwin linear rails on all axis ( zero play ), I see you have done some stiffness tests but when testing the y axis i would also check for racking as you are using a single lead screw. I done some similar trials without my lead screws connected on my Y axis and there was a lot of racking. This can be totally eliminated using dual lead screws. There is a cost to this set up an additional motor, lead screw, ball nut, coupling and drive are needed.

Regards

Dale

Yes, I had actually assumed that I would have to use a second ballscrew, but I didn't know if I could get away with tying the axis together mechanically, like with a belt, or weather or not I needed to add a separate motor.
For the stiffness test I had a break on each axis. however, it is a good question as to weather or not my breaks were the things that were deflecting.

If I do need to go with a separate motor, then I can see that once the system was under control, it might be ok, but is there an issue with finding home for each axis ? Will it rack right away if your home switches are not perfect ?
I'm not sure if it means anything right now, but currently if I push from one side the gantry is not racking.

Thanks
Evan

[esm7]

By using a timing belt for gear reduction, you can use a smaller motor to get the same torque at the screw. Gear it so the max rpm you need at the screw is achieved at 80% of the motors max rpm.

If you get a ballscrew with a lead of .5 or more, that'll allow you to use a servo with less torque, because you'll gain the torque back in the gear reduction.

Ok, so right now to achieve the max speed (7 inches per sec or 420 ipm) I need 1680 rpm. It looked like the speed of many of the brushless servos was on the order of 3K+ and so that is what made me think of the timing belt and the 2/1 reduction. But then it seems that there might be some slop between the motor and the ballscrew shaft as a result of the belt. It seems like for a 0.25 pitch ballscrew 0.001 accuracy would be 1.4 degrees of rotation.
Is this something that can be held with a timing belt under load ?
Are there any tricks with tensioning the belt or something?
Do you think that 7 inches per second will be too slow ?

Thanks
Evan

[esm7]

I've been working on this kind of stuff for upgrading my router lately as well. I've been shooting for much higher speeds (~1000ipm) and higher accelerations (1g on the Z axes using ball screws, the x and y will be more like 70in/s^2 and use rack and pinion drive). So you're looking at 420ipm and 14in/s^2... I'd recommend you try to design so you can cut at 420ipm (or faster) and accelerate at least at 50in/s^2. I'd shoot for rapids of double that speed or higher (so there's no cutting load there - just acceleration and deceleration). If you look up chip load information for common wood products, you will want to at least be able to cut this fast to get proper chip loads on your cutters while spinning them at the full speed of your router.

Additionally, it's not vastly more expensive to get much more power out of a servo system, so why not?

Anyway, whatever final design numbers you end up with, in that range I'd look at a Gecko G320X drive with a brushed dc motor from Keling:
http://www.kelinginc.net/ServoMotors.html
Keling sells the G320X along with encoders.
And http://www.antekinc.com/lview.php?d[]=1&r=25 for power supplies should get you going, or one of Keling's.

Just to clarify, I haven't actually used these, but I've been looking at them to run my router. I'm going to need more power for my y axis though, so I'm currently trying to get a Fanuc A/C servo off ebay spinning.

Ok, thanks for the good information and direction on the desired speeds. I also saw the vast number of Fanuc motors on ebay, but I was having some difficulty finding the documentation for them. It looked like most of these motors on ebay had resolvers, and I wasn't sure how to interface them with an appropriate driver. Were you thinking of adding an encoder ?

It seems like the difference between the AC servo motor and the DC brushless servo motor is the drive method. Is this correct ? it also looks like the benefit is reduced torque ripple, and better efficiency. How does torque ripple effect the CNC operation?

Thanks
Evan

[multiplex]

racking while homing with two separate motors can be seen as a plus! it allows you to fine tune the squareness of your machine by adjusting the position of the limit switches.

obviously the downside is all of the added cost. In your case, the ballscrew, servo, etc. which i'm sure adds up quick!

[ger21]

I'd shoot for a higher speed than 420ipm, and go with higher pitched ballscrews. You don't want to be spinning the screw at 2000rpm.

With the correct belt profile, belts have very low backlash. Look for a GT2 profile. See www.sdp-si.com Typically, belts are tensioned by adjusting the motor position, but you can also add a tensioner if needed.

[esm7]

I'd shoot for a higher speed than 420ipm, and go with higher pitched ballscrews. You don't want to be spinning the screw at 2000rpm.

With the correct belt profile, belts have very low backlash. Look for a GT2 profile. See www.sdp-si.com Typically, belts are tensioned by adjusting the motor position, but you can also add a tensioner if needed.

ok, I went back and checked the critcal speed calculations.. and yes it looks like for 80 inches, the best I could expect for is like 1100 rpm. Yikes.. so then my ball screw and nut prices go up a bit, as I increase the lead..

Do you know of good places to get ballscrews + nuts.. on mcmaster the 1" X 0.5" ball screw is $18 per ft, and at roton the ball nut cost is $96
Strangely they each have much higher prices for the other items.. Do you know if there are any compatibility issues ?


Thanks
Evan

[hanermo]

Timing belts hold to well under 0.001 mm.
And around 0.1 to 0.001 degrees rotational (for me, in an indexed servo spindle).

I use HTD, 5 mm, 15 mm wide. I get less than 0.001 mm movement error by DTI on a 7 axis lathe.

You dont want servo speeds over 2000 rpm (much less if possible) when using brushed servos. They spark and wear out quickly in a production setup.

It´s all a balance. If your setup does not rack, buy all means use it as is.
Using 2 leadscrews easily - you can tie them together with a single timing belt. One end should be floating, ie on a bearing, to accomodate about 0.1-0.2 mm positional error.

The best option is an under or over centered-push gantry, or a very stiff gantry.
Diagonal bracing in steel will work fine.

In general, you never want screw speeds over 700 rpm.
IF your mounts are good enough, it is possible to run BS at 1000-2000 rpm, BUT they wear out 5-10x faster.

You proposed solution should run fine with any std BS, like the 1" 0.25" one you mentioned.
A gecko 320 will work fine.

A 200W servo will accelerate your setup to max speed in around 0.2-0.3 secs with typical loads/motors/components. It´s an easy goal to achieve.

As for power, even a small 3Nm nema 23 stepper will push a 200 kG gantry at over 240 kg push, ie over 1 G acceleration, enough to snap 12 mm endmills.
A 1 Hp servo (DC treadmill motor) about doubles that.

The upper limits become necessary when you get into production for $.
Once you get to the stage that you dont mind spending 1500$ for a screw, and the same for nuts and mounts, I can recommend something else. To start with, a gecko is a good solution.

[jsheerin]

It looked like most of these motors on ebay had resolvers, and I wasn't sure how to interface them with an appropriate driver. Were you thinking of adding an encoder ?

I think from reading some documentation that some of them have incremental encoders. Others have absolute encoders and others have serial pulse coders. However that's not explicitly spelled out in the documentation, so I found a 6 N*m motor for $20 and bought that. It hasn't shown up yet, but when it does I'll see what it has mounted on it. If it won't work, I'll try mounting a different encoder on it. If you'd like the data sheets, pm me your email address.

I'm looking at using a Granite Devices drive to run this. These will run all sorts of motors and have some pretty slick features but are significantly more expensive than a Gecko drive. However you can use them to run two brushed DC motors which cuts the price down to something a bit closer to the Geckos.

The reason I'm looking at the A/C motors is price (used) vs torque, plus the availability of higher torque motors in general. That's not to say you can't find high torque motors of other types.

On your limits question, I currently have a dual drive x axis using stepper motors. The motors home together and then are individually adjusted to their home switches. So this squares the gantry, but yes, you do have to have the home switches positioned correctly. It's not that hard to do though.

Personally, I've found that it's tough to find reasonably priced ball screws that give a good force vs speed curve (for steppers) or good force vs max speed for servos. I'm going with ball screws with belt reduction for my z axes but limiting the max speed to ~500ipm. I figure that will be enough for the short travel. However I'll have an excess of force at that speed plus 1g acceleration even using small motors (well, small compared to the Fanuc). I'm using rolled ball screws from China for this. But for my x and y axes, rack and pinion plus belt reduction seems to provide a much more usable force vs max speed trade off (or force vs speed curve for steppers).