[Al_The_Man]

IMO, If you are going with high cost drives such as the Yaskawa, then there is no point in adding the cost of the Galil card, unless as you indicated in the first post you want to synchronize two axis, Then the Galil Gearing function would be needed. As Mach certainly won't do it.
Al.

[spoiledbrat]

It is as you say. I will be slaving a drive to another for the gantry. The Galil card is also required to increase the pulse count/overall speed (compare to a parallel port), and reliability of the control system. Plus, the Galil can support 8 axis down the road. With the current configuration, I am already at 4. I would like to be able to add an additional Y-Z carriage down the road. The Galil plugin for Mach3 keeps getting better and better, from what I read.
Perhaps is servo/drive prices we posted more abundantly on the internet (Panasonic, AB, Fanuc, Parker, etc.) I would have a larger selection to choose from. Everything going on this machine is meant to be premium (well, except for the ATC and the labor ). Being the little guy, that *might* buy 4 servos, does not seem to be that enticing to dealers. I just happened to receive the time of day from my Yaskawa dealer. Kudos to them.
Much more reading to go before I pull the trigger. My understanding right now is that the servo amps will be closing the torque and velocity loop, and that the Galil card will be closing the position loop via the ratio effected encoder pulses.
Does that sound feasible?

Rob

[YZF]

The Galil would normally close the position and velocity loops with the servo amp closing the current (kind of torque) loop.

Copely, Parker, Kollmorgen, Baldor (and I guess Yaskawa) all make very high quality drives. You do not have to match the make of the drive to the motor unless the motor has some really unusual features that are only supported by the specific drive (usually not the case) despite most vendors telling you otherwise (they want to sell you both).

[MrWild]

And then there is EMC2. With EMC2 you can assign two drives the same drive letter and both get the same pulse stream for two drives on one axis. Smooth.... depending on how far your axis move per revolution, 4,000,000 divisions is just bragging rights. Even an inch per servo revolution, 4M will see a pulse every .00000025 of an inch. Silly. If your servo moves more than once per inch which I assume it does, it becomes even more nonsensical. Bragging rights might sell drives, but it isn't needed for real world uses unless you're out there making silicon wafers. If you're flame cutting, wood working etc, that amount of smooth is cost overkill. Save the bucks keeping your new biz afloat. Just my opinion, but bragging up the drives as the machine goes to foreclosure auction won't get you pennies on a dollar.

[spoiledbrat]

Bragging rights? Really?

[mactec54]

MrWild

A lot of people don't understand what a real servo system are all about, what most have been used to seeing are low count encoders which relied mostly on your computer pulse (DIY stuff),a good one to watch is a Haas machine they have up to 2009 a 3,000 count encoder, the machine jerks around & is not smooth running,every movement is a jerk, Haas have increased there encoder count to around 8,000, this will improve the machine smoothness & speed

Now look at some of the new machines with the Fanuc 31ia control on, some of these machines are running encoders with 16 million pulses per revolution, this makes for a very smooth & fast running machine, when you have a system like this you can truly do HSM anything less than these big count encoders is a joke, the (20 Bit) 4million count encoders works pretty nice to, even (16 Bit) 32768 x 4 Work great for a nice smooth machine, anything less you start to get the jerky machine movement

With real servo drive LIKE the Yaskawa & other modern tec drives, they take care of the encoder in the drive, the more encoder counts you have the better your machine will perform smooth & jerk free

Accuracy is not how many encoder counts you have, although it does help,(old systems yes) most controls are set to .0001 some a little more at 5&6 decimal places if the control can handle it,

Its so nice to use a machine like this, (that you can program in 5 decimal places) were you might make 50/100 parts, once set up & you have you tool size right, you measure the first & last parts & they are the same, unless you get tool wear , the only time you really have to check your parts through a run is when you have plus /minus .0001 to hold, Cam to Control press start, make perfect parts

[mactec54]

spoiledbrat

Ok I'm going to throw another wrench in the works for you to look at, You have the drives/motors why not get the software & computer to match, this is all you would need no PLC & no Gaili card need, direct computer to drives, either by EtherCat interface or Canopen Interface, will give you a machine that rocks, 300 metre per min rapid & 60metre per min cutting (2400 ipm cutting) that if your machine can handle that speed you can set the speed to suit your machine in the control, Just go look at Softservo systems, you can get the motors & drives from them as well & you may get a better deal

[YZF]

Just FYI I've worked on the design team of several commercial machines with servo drives and have designed servo systems myself so I think I can say I understand what a servo system is. How about 2G acceleration, 2m/s velocity, 100Kg load, encoder resolution of 0.05um and absolute accuracy of 3um over 3 meters?

At any rate, at some point having a too low encoder resolution is a problem, but once you go past that point than having a higher resolution is pretty meaningless and is in the noise. I.e. it is no longer the limiting factor. If you're coupling a motor to a screw what you care about is how much linear motion is one encoder count and how does that compare to your accuracy/repeatability requirements. If you want to know whether or not the encoder is the limiting factor you just look at your following error/position error, if it's more than 10's of counts dynamically and a few counts statically getting a higher resolution encoder would have no impact.

[mactec54]

Hi YZF
I use linear motors also on machines that I build that run at 1.5 & 2G as well & have a absolute accuracy of plus/minus 1um so I know what you are talking about

You say it is meaningless to have a large count encoder, Tell that to Fanuc with there 16million count encoders, I have used the machines that have them, you can see that there is a very big difference with the large count encoders as to the same machine with a lower count encoder, may be 16 million is the limit

what we have been talking about here is only 20 Bit encoders. which are very common as standard for many different manufactures on there servo motors, are they all wrong & you know something they don't

[YZF]

"only" 20 bits? 20 bits is a lot. (By the way a 16 million count is 24 bits). If you look at the angular accuracy of the highest end rotary encoders from Heidenhein and Renishaw you'll it's a lot less than 1 millionth of a rotation. I.e. the encoder may have 24 bits but the least significant bits are just noise. So it may just be a "spec"manship war going on here.

If you have a screw with 0.2" per revolution then each count of a 20 bit encoder is 5nm of linear motion. I.e. 100 times smaller than the wavelength of visible light.

I'm betting there was some other difference about the machine with the large count encoder or maybe you moved from a 1024 count encoder to a 16 million one. For one thing the tuning would have to change (the velocity gain would need to be made proportionally smaller and so would any feed-forward terms) so perhaps you've noticed a difference in tuning. Maybe it impacts the auto tuning algorithms in some way.

Are you using SIN/COS linear encoders in your machine? Do you have a configurable interpolator? If so just keep increasing the interpolation ratio and see if you can see any measureable difference. On the linear motor systems I've worked on having 0.1, 0.05, 0.01 um counts made zero impact, the following error would be unchanged and no difference in performance whatsoever.

From a theoretical perspective the encoder's quantization introduces a non-linearity to the control system. That gets smaller and smaller as the encoder resolution increases and at some point it becomes negligible.

[YZF]

My rule of thumb is to look at your accuracy/velocity requirements and choose your counts base on a factor of 10-20 from those. I.e. if you want positioning to 1um choose your encoder to be 0.1-0.05um/count (and ensure it meets your repeatability and/or accuracy requirements as well). If you want to move very slowly at a very accurate velocity you need to figure out what encoder resolution will allow the controller to measure the velocity to the required accuracy.

As I said somewhere above in this endless thread, the stiffness of the system is most likely to limit performance followed by the control algorithm and the way it is tuned. High performance linear motor/air bearing systems are a bit special because they have extremely high stiffness and very low friction/stiction.

[MrWild]

MrWild

With real servo drive LIKE the Yaskawa & other modern tec drives, they take care of the encoder in the drive, the more encoder counts you have the better your machine will perform smooth & jerk free

Accuracy is not how many encoder counts you have, although it does help,(old systems yes) most controls are set to .0001 some a little more at 5&6 decimal places if the control can handle it,

I'm not Al the Man, I don't design servo systems. However i know that if your control doesn't have a look ahead feature, the machine will jerk and thrash through all of it's moves even with a high end drive. Having too many pulses don't make up for poor design in other areas. Having high resolution on a low resolution machine (the man did say he is using rack and pinion on one axis) is wasted money when it can be better utilized elsewhere. For wood routers and torch machines, large count encoders don't mean squat except bragging rights. They will NOT turn a common machine into a silky smooth machine by and of themselves. NOT the Holy Grail. They ARE an extra NEEDLESS expense. I get the feeling you're a Yaskawa salesman.

[mactec54]

MrWild

No not a Yaskawa salesman, Just have used lots of there drives & others as well, but there drives stand out above the rest, not by much Copley, Panasonic, Mitsubishi, NSK & others have great drives as well, I have used most

You say that all machines jerk & thrash around if the don't have lookahead, that is not correct,I once did a demo for some machine builders the difference a encoder can make to even a old machine or there new ones

One thing you can do in a Yaskawa drive & some of the others is you can change the encoder count, say you have a sigma II motor with a 17 Bit encoder on it which is standard, that is 32768 cpr x 4 ,in the sigma II drives you can only use it as a 16 Bit so you only have 16384 cpr x 4 you can set this in the drive to use as little as 16 of these count or as many as the 16384, So the test was done starting at 2048 a common number of counts that a lot of encoders have, so the machine jerked from move to move
& it did not start to smoothout untill we reached just over 10,000 counts so by the time we got to the max count the drive could handle of 16384 cpr x 4 it was smooth as silk with servo lag almost 0 at a rapid travel of 1500 ipm & cutting at 200 ipm with no lookahead, It ran smooth & no more jerk moves, this machine had never in it life run over 600ipm rapid & around 50 ipm cutting

If you have read the news, Haas there new machines have increased there encoder count to more than double of what they had, which is a big improvement but still not enough to get the really smooth running machine

[spoiledbrat]

I love this thread. How many people here think that precision can not be had on a rack and pinion drive? Rack and pinion are both hardened and precision ground. Gearbox is 10:1, <2 arc minutes of backlash. Pitch dia is 55.2mm on the pinion. After some calculation, I end up with 0.0004" of backlash, and I was told to add 0.001" to allow for imperfections in the mounting surface. So, total is 0.0014", worse case.

I understand that Galil can map an axis, with 250 points. If I can find a large enough glass scale to set this up, wouldn't that take care of the accuracy problem?

Rob

[YZF]

I've never used or seen a rack and pinion linear motion system myself. What particular vendor/model are you using? What would be the backlash of the rack and pinion itself?

Mapping the axis (e.g. with an interferometer) can take care of most of the accuracy problem but only to the repeatability of the system.

[spoiledbrat]

Flexicam, multicam, CROnsrud all use rack and pinion on a lot of their system. Especially on axis longer that 6 feet.

I am using Atlanta gear rack and pinion. With precision ground rack and pinion, 0.001" of backlash can be attained in a perfect world. Gearbox will add half a thou (0.0004). Split pinion is available for zero backlash, but will not be used on this machine.

I will mainly be doing woodwork, and no, I do not require tolerances better than that. My emphasis on this build is to use truly industrial components, so that I can run the Dog $#!t out of it for years to come.

Rob

[rowbare]

Just go look at Softservo systems, you can get the motors & drives from them as well & you may get a better deal

What kind of $$$ are we talking about for a SoftServo system?

bov

[MrWild]

Spoiled Brat, R/P systems can be tuned to decent tolerances and if you can get there, .001 is great. If a torch machine, will you even be able to tell after the dross is removed? With a .001 router, do you really need a drive with .00000025 capabilities? (if geared down, add more zeros.) After your comment, I believe you do. Spend the bucks.

Sometimes it all boils down to if it gives you a warm pink fuzzy or not. If you can afford it, want it, and in a WPF sense, need it buy the high priced spread. If the need is high end, why are you using step direction signals and Mach?

[spoiledbrat]

Mr. Wild,

I have no need for the ridiculous resolution as you stated. I would be happy to hold 0.001", but I do not believe I will, since I have 0.0014" of backlash minimum on the rack and pinion (I was told to leave 0.001" of backlash in there by the gear company, and the gearbox will introduce 0.0004").

Mach control: I am active duty military, and I am currently home for a year before I deploy again. My wife runs our products day in and day out. We have both come to love the Mach software, for ease of use, infinite configuration possibilities, the fact that it is non-proprietary, and works just like I want it to. As my wife has been using this software for more than me for the last 5 years, she likes it very much. I have seen/used the Flexicam controller, and a few others. I have always thought that "0.0000" on the computer screen should be the surface of the table; and that a G1 Z5.0000 move should take the spindle to 5" above the table, not down. Go Mach!

With VB scripting/macros, anything is possible (without paying who knows what for additional options. With the use of a PLC, even more is possible.

I will be using a galil card on this install. I think I will be using analogue voltage from the card to the drives for control. Still learning about that, so I cannot say for certain that I will not be using step/direction control, but I am pretty sure I will have the choice.

I will be paying an additional $600 per servo/drive set compared to the like Automation Direct servos. I am gaining on-the-fly PID tuning, and smoother control capabilities. I am willing to stomach that much money, as I will only be building one of these machines in the foreseeable future.

The way I see it, when you buy a manufactured machine in the $50-$90K class, you are getting a machine with cheap parts on it, so that the manufacturer can improve their profit margins. I am trying to sensibly head the other way. Although I may be going too far with the servos, I would rather regret spending an extra $2400 than be required to spend another 7K down the road.

I am not selling machines, I am making this for myself. I need to achieve something truly industrial here, so that I can depend on it when I retire from the military.

If there are any servo dealers reading this forum, feel free to PM me with better alternatives for 1Kw servo packages.

One more question. What is the best servo system for a CNC router, if money were no object? (I need to qualify that... , lest someone think I live up to my screen name. I drive a 20 year old honda. This machine is no small thing for me; I have been saving up/preparing for some time.)

Rob

[Al_The_Man]

I will be using a galil card on this install. I think I will be using analogue voltage from the card to the drives for control. Still learning about that, so I cannot say for certain that I will not be using step/direction control, but I am pretty sure I will have the choice.

With Galil you have the choice, just that encoder feedback to the card will not be an option in step/dir (stepper mode), along with no gearing or electronic cam features.
Al.