Thread: Need some FEA of HEB beams for a mechmate like router

**daviddwilson** · 10-06-2010, 04:32 PM

Originally Posted by cnc2

Remember, the longer the belt, the wider it has to be in order to keep stretch at a desired value.
Remember, at 42.5 Kgf the stretch is 0.4mm/m (every meter of belt will stretch by 0.4mm) so a 3 meter long belt will stretch (0.4 x 3)=1.2mm, two 25mm 3 meter long belts will stretch 0.6mm at 42.5 Kgf, if you want to achieve 0.3mm stretch at 42.5 Kgf it will take two 50mm wide belts

OK, it's doable, we can even hope attaining 0.15mm stretch if we use a 75mm belt, BUT, is it worth the money ? How much will it cost ? what speed (IPM) & depth of cut can we expect @42.5 Kgf of belt load ? (because if we overpass this load the error will be greater & unpredictable during machining)

I'm not sure about the stretch value of 4mm/m is it for a 10mm wide belt ? then it'll be lower on a 25mm belt.

cnc2.

Hi, I think the highest forces are going to be a result of when the gantry direction reverses, which is consistent with some figures in the belt catalogs. This is an accelleration effect vs a velocity effect. I guess this means I need to go back and figure out what accelleration I will most like have.

Thank you for bringing up this stretch challenge, as the effort to do the calculations is very worth while.

_____________________
watch movies online

**harryn** · 10-07-2010, 08:09 PM

I took a look at mcmaster.com for fun and looked at their ball screw specs. While I am sure there are better ball screws out there, it was interesting to see that they spec them as having an accuracy of 0.3 mm per meter. (the spec used inches, but that is the converted number).

This is an indication that the belts are not all that bad compared to off the shelf ball screws. Of course, the ball screw will have extremely high holding force compared to a linear belt, just due the nature of a thread, but even under the design loads listed here, it looks viable.

It sure would be nice if someone were to run these numbers independently and see if my calculations are more or less right.

**harryn** · 10-08-2010, 10:43 PM

Thanks to FandZ for pointing out that if I use counter weights, I need to figure these into the calculations.

My idea is to use a counter weight on the shaft at the top of the belt, so 10 Kg on that shaft. This does not add to the stress on the belt, but does add to the mass undergoing accel.

For 10 Kg of Z section movement, then:

10 Kg x 9.8 m / s / s = approx 100 N from accel of the counter weight.

If the mass of the moving section were to double to 20 Kg, and we continue to assume 1 G accel settings, then:

- 20 Kg held vertically, so roughly 200 N from mass.
- Holding force of 400 N
- Accel of moving Z section = 20 Kg x 10 m/ s/s for 1 G = another 200 N
- Accel of Z counter weights = 200 N

Total force = approx 1,000 Newtons peak

Starting to get into some serious forces. Similar to a 250 Lb man jumping up and down on something and hoping it deflects less than +/- 0.003 inch.

Now I need to do some motor calculations, but I am pretty sure my current motors are not even close to this power level.

CNC2 - is any of this helping or making your life more complicated ?

**cnc2** · 10-10-2010, 01:42 AM

CNC2 - is any of this helping or making your life more complicated ?

Thanks harryn, it's always good as it can inspire new ideas.

+-1mm should be acceptable, OK I wanted to reach +-0.1mm/3meters (without including spindle run out...etc) at least for the no load positioning (like for a plasma cutter) but the belt has +-0.8mm/m & that's what scared me.

I'm aiming a moving gantry design like the mechmate, OK +-3mm/3meters is acceptable for wood cutting, but it'd be interesting to see the result when engraving (backlash on a lead screw almost ruined one of Irfan's engraving jobs).

For the linear encoders: I always thought they were expensive, but recently, I had an idea on how to make a cheap linear encoder, well you take a $20 encoder (AMT102) mount it on the shaft of the smallest pulley available & let it run on an unloaded belt (one narrow tracking belt per axis) that can be made out of one of the most accurate affordable belts available.

I have no idea of the kind of accuracy we could reach with such a setup, & don't know how we could use it with the CAM software.

Any ideas ?

Thanks !
cnc2.

**harryn** · 10-10-2010, 04:42 AM

Hi, I am not sure of this, but I think that Mach III cnc control software has a way to input "corrections". Please double check this, but I think that you can take multiple very accurate position measurements along an axis and input these into Mach. The software then uses this as a sort of a correction table for that axis. Obviously, this is not that helpful for backlash, but as far as absolute positioning, it is very helpful.

It is possible that I am confusing this with other software, so a check through the manual or the mach forum might be in order.

Of course, this all assumes that you can actually make these very accurate measurements on your router.

I suppose one of the most accurate methods would be a laser interference based meter, which counts wavelengths (650nm) but I don't have easy access to this. Maybe your university does

BTW - the ATL belt is more accurate than that I think.

**cnc2** · 10-10-2010, 11:53 PM

Hi,

There is no way to use Mach3 with linear encoders in a closed loop look at post #2 from H.O in this link:
http://www.cnczone.com/forums/diy-cn...ar_scales.html

Linear encoders can be used with Mach3 only for display purpose without correcting the movement's accuracy.

It might be possible in EMC2 Linux.

Thanks !
cnc2.

EDIT:
Forget about the laser meters unless you wanna sell the house to afford the down payment

A 20mm O.D pulley connected to a 1024 CPT(count per turn) encoder, running on an unloaded belt (Zero stretch) can give a theoretical accuracy of +-0.06mm (one encoder count=0.06mm of linear movement)

20 x Pi=62.83 (linear travel of one turn of the pulley)
If one turn of the encoder=1024 we end up with 62.83/1024=0.061mm of travel per encoder count (theoretical dream) if the real accuracy is ten times lower(worse) on a three meter distance we'll end up with 0.06 x 10= +-0.6mm

Better results can be achieved with a smaller pulley and/or a higher resolution encoder.

The real question is: Which CAM software works in closed loop to make this possible ?

Well, I think it's doable with EMC2, read this: http://wiki.linuxcnc.org/cgi-bin/emc...es_On_One_Axis

**cnc2** · 10-21-2010, 08:24 PM

UP

**lumberjack_jeff** · 10-21-2010, 09:42 PM

Originally Posted by cnc2

Thanks harryn, it's always good as it can inspire new ideas.

+-1mm should be acceptable, OK I wanted to reach +-0.1mm/3meters (without including spindle run out...etc) at least for the no load positioning (like for a plasma cutter) but the belt has +-0.8mm/m & that's what scared me.

I'm aiming a moving gantry design like the mechmate, OK +-3mm/3meters is acceptable for wood cutting, but it'd be interesting to see the result when engraving (backlash on a lead screw almost ruined one of Irfan's engraving jobs).

For the linear encoders: I always thought they were expensive, but recently, I had an idea on how to make a cheap linear encoder, well you take a $20 encoder (AMT102) mount it on the shaft of the smallest pulley available & let it run on an unloaded belt (one narrow tracking belt per axis) that can be made out of one of the most accurate affordable belts available.

I have no idea of the kind of accuracy we could reach with such a setup, & don't know how we could use it with the CAM software.

Any ideas ?

Thanks !
cnc2.

If you've determined that rack and pinion drives and a tube steel gantry is the optimum solution, I don't understand why you want a "mechmate-like" design and not just build a mechmate.

A mechmate with a spindle and timing belt driven steppers is eminently capable of moderate aluminum milling.

EMC (and Mach III also, I understand) will allow you to fine tune the effective steps/inch as accurately as you want to be. In other words, it doesn't matter if the racks are 10 tpi or 10.2 tpi. You'll input your calibrated value when you do your final testing.

**cnc2** · 10-22-2010, 09:05 PM

Originally Posted by lumberjack_jeff

If you've determined that rack and pinion drives and a tube steel gantry is the optimum solution, I don't understand why you want a "mechmate-like" design and not just build a mechmate.

Well, I'm not so sure about the local availability of rack & pinion, and I don't like Mechmate's rail system with V grooved wheels...etc.

A mechmate with a spindle and timing belt driven steppers is eminently capable of moderate aluminum milling.

OK, but I never saw belt driven steppers...Motors are usually used to be driving not driven

EMC (and Mach III also, I understand) will allow you to fine tune the effective steps/inch as accurately as you want to be. In other words, it doesn't matter if the racks are 10 tpi or 10.2 tpi. You'll input your calibrated value when you do your final testing.

I already know that, but I thought that a linear encoder could be the right solution to compensate the belt stretching & have a reliable machine that will get the job done, not a machine which won't keep track of the real position of the cutter & end up with misaligned cuts because the belts have stretched & have do redo the job at a lower speed to reduce stretching effects.

Thanks !
cnc2.

**harryn** · 11-11-2010, 04:45 PM

Hi, of course it would be perfect if there was a way to use linear encoders close looped back to the control software. (EMC or Mach III)

So far, what I have found is that with servos, you can use the servo controller to run closed loop with rotary encoders, but this is not fed back to the controller software. (mach / emc) I am still a beginner, but that seems to be similar to how a standard stepper motor already works, although potentially more precisely.

Either way, feedback from the rotary position doesn't really solve the "where is the router really at" , as all kinds of things (pulleys being not perfectly concentric, belt stretch, etc) can affect the actual linear position.

I looked at both EMC and Mach III for their ability to run in a closed loop mode from linear encoders. (not rotary encoders). I have confirmed that Mach III cannot do this, and so far, it does not look like EMC can either, or at least that is what several similar questions have been answered indicate.

Not to be beaten, there is a way to pre-calibrate the linear position using the control software. If you can make position measurements precisely enough, and feed this back into the software during a calibration.

I did see some linear encoders that use a "tape" that rolls out straight for the encoder portion. If a person were sufficiently clever, they could figure out how to use an encoder signal to count out where things are and input this info into Mach. I am not ready for that level yet, so instead I am considering another method to get these measurements.

My brother has a decent quality manual mill, which can pretty accurately position to +/- 0.01mm / meter, if he goes only in one direction, (to avoid any backlash). We are in discussion of how to use it to make a very accurate ruler suitable for the task. The trick, is that I need it to measure nearly 3 meters, and he can only make items about 1/2 meter long, so some assembly may be required.

**cnc2** · 11-12-2010, 04:52 PM

Hi,