I've been battling this problem for ages. And it's only getting worse because the complexity of my work is increasing.  Plus I'm really getting tired of sanding out the chops.
I have a 4' x 5' Gantry router using 600 OZ Ametech servos driven 2:1 and Rutex 990H drives.
I do fairly large foundry patterns. My last pattern had over a million lines of code and took 33 hours to run. And that was just the finish pass!

The problem is related to acceleration for sure and I've read all the posts about Tempest. I even gave Quantum a try and although it smoothed things out it was just too slow to be practical.  If I raise the acceleration the cuts get better but I start loosing position.

My plan is to upgrade the servos to something that will give me a buttload (that's a technical term look it up) of torque. In a perfect world I would change the servos, drives and power supply. But in this economy it far from a perfect world.
My thought is to switch to the Keling KL34-180-90 1125 oz servo and a new power supply. I'm hoping to retain the Rutex drives and limp it along until I can replace the drives with a set of Dugongs down the road.

So My questions are:
Am I kidding myself about using the rutex drives with the new servos.

Will this intermediate step actually help increase my acceleration.

Thanks
Derek

You may be seeing some effects from the number of parts between the signal and the cutter. you might ought to look into some servos that can be driven directly from MACH witout the need for intermediate hardware.

I have used steppers and servos with one ot the rutex boards and the servo did not perform to my expectations. On the other hand I have used Yaskakwa servos in pulse/direction mode and Omron SmartStep (pulse/direction only) from 30 watts through 800 watts and found them to be very relaible. Aren't cheap, but they are very relaible.

You might also look at you drive tuning and deadband. If the drive needs to see a few steps before it responds you are going to get poor results. Is the drive system as tight as possible? Is there any lost motion in any axis that might be causing you some grief?

I have a 4' x 5' Gantry router using 600 OZ Ametech servos driven 2:1 and Rutex 990H drives.
My thought is to switch to the Keling KL34-180-90 1125 oz servo and a new power supply. I'm hoping to retain the Rutex drives and limp it along until I can replace the drives with a set of Dugongs down the road.

I published my experience with several DC servo drives in this document: http://www.thecubestudio.com/ServoDriveReview.htm

I did not review the older Rutex drives, but as stated in the review, the Rutex model current at the time I reviewed it had the best performance of all of the drives. The issue was reliability at the drives rated output. The Rutex drives should be capable of doing the job provided they can handle the motor's max amps.

You did not state the voltage, amperage or type of power supply you are using. What I have found it that for acceleration, you need to have the motor's MAX amps available to the motor . . and set in the servo drive.

In DC servo land, a 9 Amp 'continuous' rated motor can want 40 Amps during acceleration. Having this available to the motor is the key to performance with DC servo motors. The 'continuous' amp rating is only relevant if you are running a spindle or a conveyor or other application where the torque is actually continuous. Machine tool axis typically operate as an endless series of near max amp draw to accelerate and decelerate with 'coasting' in between.

If you provide a DC brush servo motor with it's continuous rating, your performance will be quite dismal. If you provide a 40 Amp max motor with only 20 Amps, you will not get anywhere near the motor's actual acceleration capability. A common example would be running big motors with a little Gecko servo drive. They do run fine, but you are never going to have the acceleration that the motor is capable of. Size your power supply about 20% over the motor's voltage with amps at the motor's MAX. Set the servo drives to the motor's MAX (or the drives MAX if the motor exceeds the drive) and you will have acceleration beyond what you can probably use. Unfortunately, Rutex drives don't seem to be able to run anywhere near their actual ratings.

I am just completing a review of another servo drive which will be added very soon to the review document. After many requests and an absolute need for certain features, I am looking at the Granite devices VSD-XE drive. I am so far extremely pleased with the Granite drive's performance and feature set.

 

Thanks for the info guys.

The drive system is tight as far as backlash but I have a tiny bit of shake in the gantry. The gantry is pretty ridged and it only shakes in the curves so this may be related to the chopping.
Here is a shot of a piece I just finished. I sprayed a black lacquer guide coat and gave it a quick sanding to show the chops.

This is a 24” diam ring. It has a 2 degree draft. I'm cutting it with a horizontal strategy with a .010” step down. I plotted out a couple of the line segments that the G code generated. The line segments are about .090 in length with a .0004” deviation from the arc.
In the picture you'll see the vertical lines that form the chop. The distance between those lines is about .380". The smaller stuff between them is just chatter from the 8” bit and the lack of smoothness in the gantry. That I can live with as a quick sanding gets rid of it.  The chop is at it's worst where the X and Y are working together and then smooths out when it gets to parallel.
The first piece I cut was with a feed rate of 80ipm. The second one (pictured) was cut at 50 ipm.
There really was no difference in the chop. 
I'm roughing out the inner pattern tonight and I'll have it ready for the finish pass tomorrow night. I read a post by Hood that suggested reducing the max speed in the motor tuning section should help the problem so I'm going to give that a try.

Simpson I've read that test a few times before and we're in complete agreement on the Geckos and Rutex drives. I have had the exact same experiences. The geckos would fault if I farted too loudly.
The Rutex 2020 drives are really over rated  in the amp department. I've popped a couple and Tom has been really good about replacements but I just have lost my trust in the Rutex product.
Thanks for the explanation on the amps and torque. I'm going to be purchasing a new power supply to run the larger servos and I'm going to be buying bigger drives but I was hoping to do it in stages. I plan on only changing out the X and Y axis.

The servos I'm looking at are 1125 oz  90V 7.8A continuous and 40A peak. Can you make a recommendation for a power supply?


Thanks
Derek

My spindle is set up with 90V servo motor and timing belts for hard tapping, indexing and similar functions, but I have not found a suitable servo drive for it, so for the moment it runs off a 125V 10A Minaric speed controller, and I cannot do the functions intended, but on the other hand, it does not draw power from my current PS, which allows me to run the 4th axis for now.

I am using the following PS to run the X,Y, and Z motors on the mill, but it is not enough power for simultaneous 4th axis and spindle motor.
http://www.kelinginc.net/KL7220.pdf   (72V 20A)

So I am supplementing with the following to run both the 90V spindle and the 72V 4th axis. This is the one you might consider for your X,Y,and Z axis:
http://www.antekinc.com/details.php?p=345  (ignore the photo of a toroid coil, the product is actualy a full power supply)

Reducing the motor max speed in tuning will probably have no effect on your problem unless you are actually cutting at the MAX speed. You did not say what the material is you are cutting or the tool and RPM used, so it's no possible to suggest anything specific. However, I do have a few comments on what you have provided:

First, .090 seems like a huge segment length to me. I would think that is going to be visible no matter what you do. The .3xx chatter seems like a rigidity issue, but Mach anticipates direction changes and it must start a turn early if you do not have enough acceleration to make the direction change at the current feed rate. You are running a high feedrate and reportedly have poor acceleration which my be forcing Mach to 'round the corners'. This *might* be an explanation for the delta between the segment length and the chatter marks. You need to eliminate possibilities in order to zero down to the cause (or combination of causes). I would suggest that you climb mill with a slow, small DOC finish pass and see if you can get a smooth cut.  The result will give you a direction for further diagnosis.

Thanks Simpson
My motors max speed is 100 ipm with an acceleration setting of 12. This is the most I can run without loosing steps. I've done experiments previously with increasing the acceleration and the chopping really improves. I really feel the router is really under powered for what I'm asking of it.

I'm really not worried about the surface finish in that it looks a lot worse in the picture than it is. The problem with contour profiling with small step overs (as I see it) you can never get a proper chip load to get a great finish. Your always just skimming the surface of the piece. If I can cure the vertical lines I'll be a happy camper! The problem really only shows up when I'm carving symmetrical objects like  rings and such.
I looked at the power supply you suggested but unfortunately the spec sheet link is broken. I'll email them and see what I can get.

Here's a picture of the piece I'm doing the finish work on and had over a million lines of code. It's a tri-power manifold for an early Lincoln.



Thanks again for the help
Derek

Your comment about 'loosing steps' is a bit confusing. This should not be possible with servo motors.

What is your encoder resolution and what are your steps per inch in Mach. Something I have seen a lot is people converting from steppers and using stepper appropriate reductions or even direct coupling. Gearing for a 1,000 RPM stepper with max torque at zero RPM and a 4,000 RPM servo with relatively continuous torque should be very different.

Just for an example, if you have a 600 oz-in motor at 2:1 ratio,then you get 1,200 oz-in at the lead screw. If you increase the gearing to 3:1, then you get 1,800 at the screw. Huge difference. If your max desired travel is say 200 IPM, the you might look at gearing for the servo motor to be at about 80% of it's rated RPM at 200 IPM of travel. If you over size the power supply , then you can use 100% of the motor's RPM rating. That should give you the max available acceleration from the motor . . provided you give the motor it's max amps.

Stepper motors sort of pour power where servos sip, sip, sip . . . but servos want a BIG straw.  I did a lot of testing with DC brush servo motors from little36V NEMA23 to 90V NEMA34. With the bigger motors, the difference in acceleration between 17.5A max drive output and 35A max drive output is incredible.

Your comment about 'loosing steps' is a bit confusing. This should not be possible with servo motors.

What is your encoder resolution and what are your steps per inch in Mach. Something I have seen a lot is people converting from steppers and using stepper appropriate reductions or even direct coupling. Gearing for a 1,000 RPM stepper with max torque at zero RPM and a 4,000 RPM servo with relatively continuous torque should be very different.

Just for an example, if you have a 600 oz-in motor at 2:1 ratio,then you get 1,200 oz-in at the lead screw. If you increase the gearing to 3:1, then you get 1,800 at the screw. Huge difference. If your max desired travel is say 200 IPM, the you might look at gearing for the servo motor to be at about 80% of it's rated RPM at 200 IPM of travel. If you over size the power supply , then you can use 100% of the motor's RPM rating. That should give you the max available acceleration from the motor . . provided you give the motor it's max amps.

Stepper motors sort of pour power where servos sip, sip, sip . . . but servos want a BIG straw.  I did a lot of testing with DC brush servo motors from little36V NEMA23 to 90V NEMA34. With the bigger motors, the difference in acceleration between 17.5A max drive output and 35A max drive output is incredible.

The servos top out at 900-1000 rpm. I'm running them at 2:1 with .250 per turn lead screws. The best I can get out of this arrangement is 120 ipm so I make the max 100 ipm. As far as loosing steps well I guess in theory that should be correct. I know that if I set up a dial indicator and run a bunch of moves with the acceleration at 20 I loose position. If I turn it down to 12 I can run it for hours and It always returns to 0. I've always read on the boards about loosing steps if your acceleration is too high. If this only applies to steppers then maybe I have drive tuning issues.

I'm not cheap I'm poor   Since I don't know what the hell I'm doing I figure put the biggest motors and drives I can afford. I just feel like I'm asking too much from these motors. I built this machine when all Art was still working the bugs out of Master 5 (yes I've upgraded to Mach)  so I have probably gotten my monies worth out of these servos! Even If it doesn't solve my chopping problem it will help out my rapids. 

Derek
 

I don't think bigger motors are going to solve your problem. I can think of four places a servo motor can loose position due to acceleration; mechanical slippage (loose pulley,etc), the servo drive itself (unlikely), the parallel port hardware or Mach's port driver. I did quite a bit of research on all of the servo drives that I reviewed and never heard of Rutex (or any others) loosing position. I have not heard of, nor do I see where an encoder can cause such a problem (although the wiring definately can). So assuming you have checked the mechanics of your machine, that leaves the port harware, or Mach.

I have occasionally had a spooky random problem with the Z axis creeping very slowly on it's own while the X and Y are moving simultaneously and the spindle is running.  Since Mach is not commanding the move, and it is open loop, it is unaware with the result being lost Z position. I had it solved for quite a while, but I recently changed the Z axis servo drive and the problem has returned. Slowing Mach to 25 clears the problem for now.  

Some specific things you can try that cost very little is
1) if not already done, shield all control wires (encoder and step/dir) and keep them as far away from the motor power wires as possible. Turn off any wireless devices (phone, network, etc) that are anywhere near the machine.

2) an add-in parallel port. These generally work better than built in ports anyway in my experience and a good one that will work fine is only $15. I posted here the exact one to buy, where to buy it and where to get the correct drivers. Do a search and it should turn up.

Very cool manifold, BTW!

The first thing I thought of when I seen that picture of the "chatter" marks is...why is he using segmented lines for a circle, instead of arcs. That choppy/chattery outcome is pretty much typical when using segmented lines, and doesn't really indicate a problem with the machine. Then you stated that the manifold had over a million lines of code. Man, that's a lot of lines! Either that's one tiny step over being used, or there's a buttload (I didn't have to look it up!) of segmented lines. So I guess what I'm asking is, are you using arcs or segmented lines for curves. The difference between the two can be night and day as far as quality of cuts, and machine performance is concerned.