[machworx]

I have a millport mill (similar to a BP series 2) that I have installed ground ballscrews & am getting ready to CNC it. I can buy a ready built kit from AJAX for about 4000$ but I think I would rather do it myself for the learning experience (and less $). Initial questions are:Steppers VS servos, I've been told that servos are more accurate (Servos have a built in encoder/resolver & steppers do not?) but steppers are simpler to setup & less $. I want this machine to run off of a PC type computer so can I use either steppers or servos? In addition to the PC what components do I need for each axis? If I want servos is there a company that makes servo drives for home builders? If I buy the servos off Ebay do I want brushed or brushless? AC or DC? High voltage or low voltage? Similar questions if I go with steppers? How do I know what size motors to buy, what holding torque vs running torque?

[Hood]

I went the DIY route and am glad I did, not only is it a lot cheaper it will also be cheaper in the long run. Reason for that is I put it all together myself so if anything goes wrong I should be able to fix it, another thing is when it comes time to upgrade I will be able to do that.
Now steppers V servos, I am not able to answer as I just have experience with steppers. I dont really know enough about it to argue for one or the other but I will say that I use steppers on my Bridgeport Series 1 CNC and have never had a problem. I use 916oz/in and a 2.5:1 reduction and there is more than enough power and I have never even been close to stalling them.
What you will need for a CNC control is a drive for each motor and a power supply to power them. It is also best to have an optically isolated breakout board to connect to the parallel port with.
You can get drives for servos and steppers from www.geckodrive.com and www.rutex.com does them for servos.
Breakout boards can be had from www.pmdx.com , www.campbelldesigns.com , www.cnc4pc.com etc
For the control software there are quite a few, I use Mach3 and I think its the best but thats just my opinion.

Hood

[Al_The_Man]

If I want servos is there a company that makes servo drives for home builders? If I buy the servos off Ebay do I want brushed or brushless? AC or DC?

Brushed DC are a little easier to mix and match as far as different motors and drives as you don't have to be concerned with commutation issues as you do with DC brushless or AC.
The drives already mentioned take step/dir control which you will need.
Al.

[NC Cams]

You might be able to DIY the mechanicals but what about the software???

I"d first look at the software that will support a DIY system and then use whatever componentry that they recommend. If you don't do this, you could end up buying something that is not integrateable.

If you're not confused now, keep looking, you will be.

Here are a few links you just might find helpful:

Great NC controller thread - NC vs PC’s
http://www.cnczone.com/forums/showth...852#post182852

Lathe retrofits (real good buy vs build info)
http://www.cnczone.com/forums/showth...lathe+retrofit
http://www.cnczone.com/forums/showth...lathe+retrofit
http://www.cnczone.com/forums/showth...lathe+retrofit

Research (how to - most people don't know how)
http://www.cnczone.com/forums/showth...9&page=1&pp=15
specifically post #15

Electronics books (buy them if you don't know electronics)
http://www.forrestmims.com
or
http://www.amazon.com/gp/product/pro...lance&n=283155

The "holy grail" of electronics info for the DIY CNC neophyte:
http://www.amazon.com/gp/product/052...lance&n=283155

Web-based books on electronics:
www.ibiblio.org/obp/electricCircuits/

Why parallel ports may sometimes suck at doing CNC control::

See post #5 in the following thread:
http://www.cnczone.com/forums/showth...852#post181852

Linear scales;
http://www.practicalmachinist.com/cg...3;t=003181;p=0

Ball screw basics:
http://www.cnczone.com/forums/showthread.php?t=20748

Ball screw treatise, the hard core stuff:
http://www.cnczone.com/forums/showth...?t=8813&page=1

PID tuning (for servos)
http://www.cnczone.com/forums/showthread.php?t=20927

Daisychaining ATX power supplies:

http://www.cnczone.com/forums/showth...015#post142015
http://forums.bit-tech.net/showthread.php?t=108208

Linear P/S design/construction
http://www.campbelldesigns.com/files...ply-part-1.pdf

How to properly phase a transformer in linear p/s:
http://cnczone.com/forums/showthread.php?t=14821&page1

Servo amp P/S design
http://www.elecdesign.com/Articles/A...7635/7635.html

Servo motor , servo amp, powersupply sizing for CNC
http://www.rutex.com/pdf/Mystique2.pdf

Bearing literature:
http://www.bardenbearings.com/literatr.htm
http://www.timken.com/products/bearings/catalogs/
http://www.jp.nsk.com/app01/en/catal...gi?ec=bearings

Get the NSK E1102 catalog for starters

Press fits:
http://www.eng-tips.com/faqs.cfm?fid=1230

Motor torque info/defininition
http://www.merkle-korff.com/formulas.asp#con

What gage wire to use:
http://www.cnczone.com/forums/showthread.php?t=17350
http://www.cnczone.com/forums/showth...480#post136480

Servo vs Stepper (READ THIS!!!!!!!!!!!!!):
http://www.cnczone.com/forums/showthread.php?t=17419

Simply stepper info:
http://www.parkermotion.com/catalog/catalogA/A12.pdf

Stepper reverse engineering:
http://www.doc.ic.ac.uk/~ih/doc/stepper/others/

Stepper sizing:
http://www.cnczone.com/forums/showthread.php?t=17707

Small stepper P/S
http://www.campbelldesigns.com/files...ply-part-1.pdf

G code parts counter
http://www.cnczone.com/forums/showthread.php?t=22223

Haas TNC explanation
http://www.haascnc.com/customer_serv...he/96-8700.pdf

Finally, probably a majority of the DIY stuff is stepper oriented. Conversely, a preominant majority of the commercial machine tools that are "factory built" consist of servo based systems.

THERE ARE LIMITATIONS/ADVANTAGES TO BOTH.

[ger21]

Why parallel ports may sometimes suck at doing CNC control::

See post #5 in the following thread:
http://www.cnczone.com/forums/showth...852#post181852

Why post this? Parallel port control may suck based on someone's speculation of how he **thinks** the software works? Sure, if you try to use too slow of a PC, or a PC with software running that interferes with the control software, you may have problems. But people are buying $299 Dells and even $199 PC's from Fry's, and they all run Mach3 fine right out of the box.

There are over 5000 registered users of mach3, who found that it didn't suck too bad, because they all payed $150 for it, even after they got to try it out for free for as long as they wanted.


Servos are not necessarily more accurate than steppers.

For stepper and brushed servo drives, see www.geckodrive.com and for brushed and brushless servo driveres, www.rutex.com

Whether steppers or servos, you want DC, with the exception of possibly brushless, which is the most expensive route also. Voltage depends on the drives you choose to go with.

[NC Cams]

I usually edit out my "personal comments" before I post a crib note response to someone. I appologize for not doing so on this one. I figured that sooner or later he'd be asking a lot of the "usual" questions so I provided answers en-bulke. Besides, the note in question says "may suck", not "will suck" - big difference.

We've all seen cases where guys run into "issues" with LPT driven systems. They use laptops with low voltage LPT's, they use laptops with power managment strategies that mess with real time machine control. Folks even use "lamp cord" instead of shielded wire - even when TOLD to use shielded cable. Some folks just have issues, period.

As well as Mach is supported, it is not always that easy to implement. Most problems are due to use shortcuts, user deficiencies and or other things that crop up. Fortunately, there is a support base of folks who've been there and done that and know what to do to fix most anything.

Whereas at one time I thought Mach was "just another" S/D system (my stuff is commercial and servo based) I'd have to say NOW that the Mach method is probably THE ticket for lots of guys looking for a reasonable cost DIY CNC system. Just follow their recommendations for hardware and you should do fine.

Servos are a bit more difficult to get your arms around. YES, Mach will drive servos BUT you still don't have real time F/B to the PC during operation. I know of NO "Mach like" system for servos based retrofits that fit on medium to larger machines with FULL F/B - there are table top mill servo systems but these can't usually be scaled up and I dunno if they do FULL F/B.

Yes, steppers are not usuall more accurate than servos. BUT we've seen where, in spite of HUGE efforts to try, steppers can't match the performance/finish that a servo generates. WE've seen this time and time again it comes to precision form grinding ala cam profiles and/or OD grinding of journals. From this, we've simply come to expect more when WE judge a CNC - your needs may not be so stringent and probably won't be.

Besides, in our case, steppers were found to leave 'steps' and they can't hold shape adequately due to the rapid motion changes needed to achieve a cam profile shape.

To test the ultimate efficacy of ANY system (servo or stepper), simply try to cut a perfectly round OD or ID circle REALLY FAST.

When think you've got the system right, do the test cut and make it perfect as it with with NO chatter, no flats at the direction change points, no diametral deviation at the max velocity points at 45 deg and then RETURN to X=0.0000 and Y=0.0000 coordinates at the end.

Do that with ANY system and you've REALLY got something running SWEET. We can just do it with our Eztrak to within 0.0003" but the cut takes nearly an hour to finish. The neighbor's Haas does under 0.0005" but takes 10 minutes.

I'm not belittling Mach or any other step/dircection system. You just have to realize that some aspects of CNC just are NOT that easy to do well or on the cheap.

[machworx]

I want to thank everyone for the assistance so far. There's a lot of info to digest which I am working on. But as usual good answers routinely bring up new questions which I will ask as they come up. Everone talks about this Mach program, can one of the NC programs that come with either Unigraphics or Pro-E Do the same thing? The part about stepper driven mills & the not truly round problem I experienced myself when I was running a "bandit mill" back in the 70's & would be my main reason for wanting a servo driven mill. But it seems that the servo driven option is more expensive & maybe a little more of a headache to set up. Any guesses on a typical ratio of ($ for servo driven)/($ for stepper driven) so I can deem if it's worth the extra $. I have gathered that S/D means step/direction but what about F/b or FULL F/B? Enough questions for now until I catch up. Thanks

[NC Cams]

I'll provide a oversimplified feed back (f/b) explanation.

A stepper moves based on pulses or steps sent out by the controller. THis is done in conjunction with a direction command - step - directin or S/D. Send a step, it moves, send another another move, do it fast enough, psuedo commutation occurs and you have "smooth" rotary motion. Smooth being relative in this instance - you're still stepping only real fast. Real fast steps can cause inductive heeby jeeby's that half stepping can mitigate.

The resolution for each step is a function of the motor design. Some motors have X steps per rev and some have Y steps per rev. THus, unless your controller is capable of partial step resolution, the smallest step your motor take is a function of it's design characteristics. You can research the steps per rev or deg/step for yourself.

So, the PC sends out steps and the motor moves those steps. Well, the steps were sent but did the motor move any of them, some of them or all of them???

That's were F/B comes in.

Add an encoder and feedback is sent back to the controller so you can close the control loop - move this many steps, yes sir, I moved them and here's proof.

The situation is that steppers are usually parallel port driven. They don't have enough f/b channels available to accept the encoder f/b from all the channels simultaneously. You can add more and more LPT ports but the data you put on the port ONLY is read when the port(s) are scanned by the PC. Since this may not occure every clock pulse of the PC, you don't have instantaneous f/b. (essentially true for this simple explanation).

So, you simply try to build a robust stepper system that behaves and moves when and how much you say to and not do so spasdically because you really don't have f/b to the PC.

Ok so along comes Rutex and Rogers machine. They build provisions for encoder f/b. BUT does it go ALL THE WAY back to the controller for instantaneous control? Essentially no.

They both do error verification remotely to a predetermined amount of step misses (recall this is oversimplified for conceptual, not full technical explanation of what's going on).

Thus, if you have an error window of 50 counts (250 sent but only 199 equivalent steps are reported as having been taken) the Rutex or Rogers machine thingies send out error pulses to stop program execution. Since the machine may have a resolution of 0.000002 per pulse, 100 millionths error is pretty paltry but what if you're stepping at 0.0001 resolution??? Is 0.005 error tolearable??? That's your call. This error window varies but we're only trying to explain the concept, not design a system.

The typical servo contoller has similar f/b but instead of the f/b going to some remote device for decision making, the encoder counts go back to the PC and the PC does the comparison process in real time mode. Due to more sophisticated control regimes that are available to a servo, you can now do more sophisticated feedback via PID loops and other stuff beyond the realm of this simple explanation. However, this control has to be interpreted by the PC.

In the system that I'm most famiar with, the pulses being returned are coming back pretty darn fast. Thus, you can actually move so fast during a prescribed move that the incoming datastream comes in too fast for PC to count. Or signal noise blurs the count. Or inertia can't be overcome and the stuff did't move like it should have. You can end up with error as well. So you have a choice, slow down the machine or widen the error window so you can run the SOB.

Both systems servo and stepper are taking pulses and turning them into motion. F/B is added for verification that the prescribed motion did occur. However, electrons get lost and stuff just happens. THings don't go as planned.

When you can cut a perfectly round circle, do it real real fast, have no steps, chatter or out of roundness and return to 0.0 origin each and every time, time after time, you've got a damn good control loop - whether stepper or servo based.

Why a circle?? You're creating a sin wave velocity loop which will tend to force a system into sympathetic resonance. You're also forcing two axis to keep perfectly syncronized motion while doing so in simultaneously opposite directions (as X is going faster, Y is going slower, by exact same amount - if it don't, the circle goes out of round).

Call it a test for the controller to rub its belly and pat it's head at the same time.

See how simple it is???

[fkaCarel]

Parallel port control may suck based on someone's speculation of how he **thinks** the software works?

In the other thread this was rejected as speculation. I asked why. No answer came. If there is only denial I can only assume I'm right, till proven wrong.I have read up over the years quite a few things about Intel processors. Just go figure out where large data and program processor-caches come in handy and where you can't use them.

I programmed a G-code interpreter with a 2-Mhz (yes, TWO) processor in 1986. Maximum real time speed: 40 Ipm. Linear and circular interpolation. It's just good for my ego to see that these results can't hardly be beaten by a processor that's 1000 times faster.

[NC Cams]

I think, if I interpret Carel's post properly, that he and I are on the same page in this regard: newer faster PC's aren't providing better or faster code processing with regard to today's CNC controls. Complexity in search of a need.

I'm amazed that Bridgeport did 4 axis machining with their VMC 1000's back in late 90's and only used a DOS box and BMDC3 for control. Some were 386 and some 486 but no Pentiums. In fact, you can't even use a Pentium faster than 133 to retrofit to the "legacy" systems.

What did they know about machine control that they incorporated into their DOS based DX32 system that isn't being or can't be done today???

Hmmm. Simultaneous motion with legacy, nearly free O/S's and PC's. Talk about value and a DIY dream. In some regards, the guys who did DOS programming had to be smarter.

Oh if there were only a legacy based DIY servo style retrofit kit that worked. Probably wouldn't sell - too much reliance on windoze anymore.

[Al_The_Man]

I'm amazed that Bridgeport did 4 axis machining with their VMC 1000's back in late 90's and only used a DOS box and BMDC3 for control. Some were 386 and some 486 but no Pentiums. In fact, you can't even use a Pentium faster than 133 to retrofit to the "legacy" systems.

But a distinction has to be made as to what was doing the processing, did the BMDC3 card have its own processor for motion control?
There is a distinct difference between a PC processor doing the actual motion calculation and a motion control card, which now have 32bit RISC processors dedicated to the motion trajectory.
Galil and others where doing it this way back in the 80's for 8 axis control.
I am still using some of their DOS capable Legacy cards for simple/low cost retrofit projects.
Al.

[NC Cams]

BMDC3 used 68CH30 Cpu and and 68882 coprocessor.

It also had D-A +/-10 drive for servo drive signal to amps and a host of other chips. It was a single slot ISA board that ran mill (2, 2.5 & 3 axis), VMC (3 or 4 axis), lathe (2 encoders and 2 hand wheels plus spindlle F/B); and/or surface grinder.

I'd be interested in knowing more about what kind of canned software you use with the Galil stuff to do legacy controls. Advise via P/M if you wish....

[Al_The_Man]

The retro fits I do using the Galil DOS based cards is based on some software that Galil unfortunately no longer sell (probabally because DOS based). But I can still buy the s/w key for each system for $200.00.
It was really intended for an operator interface, and it is quite powerfull in its own way, the down side is that it does not operate with standard G M code type programming, but when I get the chance I am trying to see if I can figure out some kind of interpreter.
Just when I think I have time for a little R&D someone come up with a request for another system.
So far some of the projects have been a 3 axis CNC line punch, Rotary tables, CNC back gauges for shears and breaks, 3 axis CNC tube cutoff machine(dedicated lathe), Cinncinnati point-to-point Horizontal mill, Hydraulic CNC pipe Bender. Some I have used Touch screen input.
The front end software screens can be drawn in a few minutes, when you get the hang of it.
The individual screens load native Galil two-word command programs, so it can be pretty powerfull.
If you look at the DXF to DMC convertor software on the Galil site, it gives an idea what the Native commands can do in the way of interpolated moves.
Al.

[machworx]

I was only looking for what the initials "F/B" stood for but the explanation you gave was excellent with respect to what really happens in the pseudo feedback loop used by some steppers vs the true feedback loop ( in the classical automatic controls type jargon ) used by servo's. Still looks like servo's are better for most things except cost. Another question somewhat related is when this mill is completed if I want to use it in a manual mode is there any difference with regards to stepper vs servo; IE can I freely spin the table motion handles when the CNC is off? Or do they lock in place only when the power is on?

[NC Cams]

When your remove power from the drives, they let the motors spin relatively freely.

If you remove the belt and turn the handle you see more of a difference.

But why bother? Use jog mode and crank by hand only when and if you have to.

Al; too bad about the DOS stuff not being supported. I'd love to buy a DOS based legacy system today for mill or lathe - especially a lathe. I'm not convinced you need the Windoze overhead to run a CNC - at least for my needs.

[machworx]

So as I continue to research the stepper vs servo I still think I'll probably bite the bullet & pay the extra $ & go servo. So what components would everybody out there suggest I buy to make this conversion. I'm refering to manufacturers of components such as Geckodrives, Rutex, PMDX, or any other companies that anybody knows of that make these type of components. As I see it right now on a per axis basis I need a servo motor, something like a Gecko 320, then something like the gecko 902 and then all axes tie together at something like the Grex-G100 which then goes to the PC with the software loaded on it. Can a person mix & match components from different manufacturers or is that just asking for compatibility problems down the road? Also when I did my ballscrew installation I only did the X & Y axes. Do most people out there use their table up & down as their "Z" axis or do they use the spindle movement as "Z". If one wants to install a ball screw into his BP clone head are there companies out there that make a "kit" for this "Z" axis conversion. Do people ever put a ballscrew into the table up/down axis to replace the standard acme?

[NC Cams]

Matched vs mix-n-match servo's = ultimately you have to tune the motors. If they have vastly different characteristics, you could find it real hard to tune them so that they run well together or run evenly. When you really need them to perform in a coordinated fashion, they just might not....

If you buy a hodgepodge of stuff, down the road, service could become a nightmare, especially if you buy a bunch of one-off "legacy" (obsolete/closeout) motors.

Z axis ball screw: There is a ballscrew retrofit kit available for driving the spindle quill as the Z axis. It bolts to the front of the spindle nose of a Bridgeport. For the life of me, I can't recall who makes it. Perhaps one of the other senior members will recall who it is. And, yes, the spindle is most commonly the Z axis.

Knee drive for Z = it has been done BUT, you're moving something that is pretty heavy. You'll need some real OOMPH to do it - I wouldn't plan on doing it fast.

EDIT

re: component selection= Don't buy your servo drive cards until you get your motors. The last thing you want to do is find the "perfect motors" and then find out that the voltage or amperage they need is more than the drives can handle.

Some of the older "legacy" motors fit that category - high amp/low volt vs newer higher voltage/lower amperage.

Biting the cost bullet isn't the only bullet you will eventually be biting when you start integrating servos into a CNC'd anything. END EDIT

[ger21]

As I see it right now on a per axis basis I need a servo motor, something like a Gecko 320, then something like the gecko 902 and then all axes tie together at something like the Grex-G100 which then goes to the PC with the software loaded on it. Can a person mix & match components from different manufacturers or is that just asking for compatibility problems down the road?

Assuming you're using Mach3, right now there are 2 different directions you can go. Through the parallel port, or the Gecko G100.

If your using the parallel port, the usual route is to go through a breakout board, which can add provisions for limit and home switches, relays, and other options, as well as making it a lot easier to wire everything to the parallel port.
Some of the most commonly used breakout boards are from www.campbelldesigns www.pmdx.com and www.cnc4pc.com

The parallel port under Mach3 control is limited to a maximum of 45,000 steps per second max. With higher resolution encoders, that may not be fast enough. That's where the Gecko 902 would come in. It multiplies the incoming steps and will let you run the servos at much faster speeds, or use higher count encoders. Rather than purchase both the G320 and 902, the G340 is a G320 with a 902 already installed.

The G100, on the other hand, can output steps at up to 4,000,000 per second. If using the G100, you would not need the G340, or the G902. The G320 would be fine. The G100 also has 16 inputs and outputs, which would eliminate the need for a breakout board. You actually can't use the parallel port when using the G100 with Mach3, it's not supported. One thing to keep in mind, is the the G100 running under Mach3 is not fully functional, and it may be some time before it is.
As of right now, though, Mach 3 will run code, home and limits work, probing works, and jogging is functional. This info is from the Mach3 support group at www.machsupport.com

As for mixing and matching, there shouldn't be any problems. Any step and direction drives should work with the breakout boards I mentioned. Any step and direction drives should work woith the G100. As for motors, make sure the drives are adequate for the motors you choose. A good source of servos (and steppers) for hoby use is http://www.camtronics-cnc.com/

[jdelaney44]

machworx - sounds like you have a lot of good advice already. I would like to share some experiences I am currently struggling with that may be of some use. I am using Mach 3 with brushed servos. I have a thread on this group with my whole machine build documented. The problem I have run into is, I think, with thousands of starts & stops the accell / decel tends to create a larger potential for error.

I've got a pattern for a mold I am working on and it's pretty curvey. No circles. Spirals. So my Visual Mill program generates something like 25,000 lines of gcode.

It takes 8 hours to run. The problem is that each line of code is a start & stop. They go by so fast that the machine never really stops of course, but there are little pauses here and there. Sometimes it errors out the X-axis. So I've had to slow it down alot to get it to run true. I think on the last run I maxed out at 9 ipm.

I've yet to implement even the most rudimentary feedback to trip the e-stop based on the Gecko err signal. So, I'm basically running open. When it fault, its a real pain to get back to some kind of a reference if not totally impossible.

So, unless you have the Gecko err wired in to trip e-stop or you get the Rogers board, the encoder feedback to the G320 does no good.

I'm not all hip on it, but the G-Rex is approaching this problem by offloading the motor control from the PC. I haven't checked in with the project lately but it sure looked like it was going to be cool a couple of months ago. Gecko was shooting for a stall proof stepper by adding an encoder for feedback from the stepper. The thought, I believe, was to get the best of both worlds.

"Servo" set ups as normally talked about here, use a really good DC motor with a shaft encoder. But, people have used steppers with encoders. I guess that would be a "stepper servo" or some such thing. (??) It's just a matter of finding all the right pieces of hardware and software to make it tick.

Just to clear up one other thing that I wondered about for a long time. The version of Mach3 I have does not have the capability to vary speed even if it could receive the feedback. I have seen Art write about doing this but am unaware of any progress due to my own ingnorance. This is the real nut of feedback on a machine in my humble opinion. The software would ideally have the ability to compare following error and adjust the speed until the following error becomes zero and then keep it there.

Still, in theory, none of this can be "perfect". As the others have said, this is all about how close is close enough.

On my machine +/- .001 is good. The machine itself can induce errors greater than that. But it's an old dog to be sure. Still for what I do, that's fine.

My other piece of saged advice is about motor and power supply voltage. Resist the temptation to get motors that are maybe more than 10% higher than your power supply voltage on a servo rig. It will work. But the motor is going to work best at or near it's design voltage. I know that sounds incredibly obvious. But, I did get some advice early on that led me to believe it would work better than it did. I have recently ordered a transformer with a higher voltage to close the gap some.

Whilst I am rambling on. You are going to see people saying things like amps / current = torque. From the math I have done, it's voltage and current together that have to be considered when trying anticipate motor performance. I think everyone assumes fixed voltage because that's the norm. But if I am evaluating power supply voltage at the same time, this assumption confuses me anyway

At a given voltage, yes more amps is going to equate to more torque. At a given motor load, yes more voltage is going to yield more speed. Also note that as speed goes up, induced resistance goes up and current goes down. A stalled motor has the least resistance and thus allows the most current to pass.

This is often hard for me to wrap my head around because a motor is constantly changing state. There are infinite power load and speed combinations each with different electrical characteristics.

Resist the urge to think of this in terms of pushing current. It's about the resistance in the motor allowing current to pass and then the resultant magnetic, uh, dynamics, after that.

I am sure someone who is better at the math will chime in..;-) Hope that helps.

Best,
-jd

[NC Cams]

JD44: Your post give good empirical insight into the realities of machining as opposed to the vaporware that code often posesses and control board/software makers often try to claim as factual.

Although machine code may ask for INSTANTANEOUS ramping of motor speed or servo response, the reality of life is that it won't/can't happen that way. These are due to mechanical and/or electrical limitations that you can't get around.

Thus although software may have "look ahead" and "ramping" or PID capabilities, you still ultimately have to deal with the mechanical and/or electical vagaries of your particular system. Perhaps your motors can respond as they have tons of torque and miniscule inertia (not hardly) to deal with.

However, your power supply may NOT have the ability to instantaneously supply the 1000 amps for 3 milisec that the system is asking for to do that light speed move. If it remotely can, then you have inductance and resistance issues (simple line losses) which may prevent the system from moving as asked for.

You can get bigger motors and more power but that may not work because your controller card won't handle the current due to mosfet or trace size limitations. It all comes down to "realistic expectations" that you can satisfy with the "system integration and interactions" that can and will occur.

Keep in mind that the lack of direct PC feedback will limit you to what you can get away with on a stepper. Effectively, you can send out pulses at light speed. The controller will react in semi-equivalent fashion. However, the analog stuff downstream has other laws of phsyics that it has to respond to/behave in accordance with. Response is going to be a tad slower.

In servos with direct PC feedback, they too don't/can't respond instantaneously for the exact same reasons. Thus you have an "error window" that you have to operate within.

Exceed that with a servo and the system will shut down, stay within it and you'll get profile deviation only it will be limited to how bad it can get before the window gets too big and S/D occurs. Slow and precise versus fast and sloppy(ier) or too fast - oops - shut down.

Steppers either start to miss steps if they are asked to do stuff too fast or they may start to oscillate/misbehave or the forces overcome the stepper torque. Add encoder feedback and you get a chance to reduce the "error window" thingie to stop things from getting TOO far out of hand but instantaneous response and no error, sorry I think not.

The above may be an oversimpification BUT the net effect remains the same - you can't do things instantaneously mechanically even though the electronics can/may be able to to it almost that fast. Here's where reality comes into play.

A Haas VMC can run a tool path around a cam profile shape with VERY sophistacated VMC software and STOUT hardware in about 15 minutes. It can hold shape to within 0.0005 or less.

Yet the same profile with the same code, line for line, takes close to 2 hours to run on a servo based CNC BPT mill. Accuracy is withing 0.0003. Plenty close enough.

Consider the price and time factors and you be the judge if the results justify going with in-house BPT or outsourced Haas cut parts.

The point is that the EXACT same code takes nearly 6 times longer to achieve comparable accuracies on one machine versus the other. Why? Control sophistication and robustness and the net abiltity to have same to do the given task faster/quicker at comparable accuracty.

The fact that you can 2D or even 3D mill with a parallel port driven system is astounding. Considering that you can do it to accuracies as reported with a DIY hand built system are equally as phenomenal. Realizing that you have to slow down to 9 ipm to do it should not be surprising.

We have to run at 3-4 ipm to get 0.0003" accuracies on a profile and we're only dealing with 2D moves. it is amazing what you REALLY learn once you get past the sales puffery that ALL CNC stuff has associated there with. Unfortunately, some of this information/education isn't available until you start to run the machines and learn their abilities and limitations first handed.

Then you get disappointed, work on improving things using your skills to an accaptable level or live with what you have.

To me, it is hard to imagine an essentially open loop/non-feedback system respoding with the same speed and accuracy as a dedicated commercial VMC - the capabilities are simply not there. It will perform exceptionally well but it will have its limitations - you'll inevitably stumble across them as you use your particular system more and more.

Hint: to find the achilles heel of any CNC mill REAL FAST, OD mill a perfectly round flat disc plate (simple 2D moves).

Only do it AS FAST AS YOU CAN, with NO out of roundness, NO flats or chatter at the 4 direction change points and WITH as smooth a finish as you can genereate in mild, HR 3/8" thick steel. This will tell you how REALLY good your system response and hysterisis and control loops is/are.

BTW: the Haas option to miling our cam profile in this material in 16 minutes and holds stuff to w/in 0.0005 or so. Scotch brite clean up is adequate.

BPT takes 2 hours with accuracy as noted above with comparable Scotch brite clean up.

A dedicated CNC cam grinder will do it in a few minutes and hold tolearance of well under 0.0003" with a semi mirror finish. However, the price is 4-5 times that of a Haas and takes 10-14 days min. to get from an outsourced facility.

$25-$30K cnc mill, versus $240k VMC versus $2.5million cnc grinder. Speed and performance and accuracy costs money, how fast and how accurate do you want/need the part to be???

JD = good explanation and insight into the limitations that the stuff between the parallel port and motors do/contribute to with regard to CNC performane. Valuable data that I'm glad you shared.