View Full Version : Rigidity of machine

AB (Unregistered Guest)
08-31-2005, 10:45 AM
I seem to have a few problems with my machine, a PRT.

I did the 'gantry wiggle' test and registered about 0.005" of movement. I checked the set screw on one of the X motors - it seems to be very tight - I tightened it as much as I could anyway yet I can still very slightly move the gear without moving the shaft.. unsure why.

Secondly, the Y carriage appears to be either very poorly designed, or the guy who assembled the machine here did a poor job of putting it together. I can move the side with the motor on it by about 0.005" in either direction (same as the others) but the side without the motor can probably be wiggled side to side by anything up to 2mm!

No doubt this is showing up in my cuts, particularly diagonals which is what made me check it out in the first place.

WHat can I do to rectify the problems? I won't be doing it as I am not the most DIY-minded of people, but I can pass the suggestions on to someone who can.

Incidently my collegue here who put the machine together has convinced himself he is not at fault and the problems lie with the design of the machine.... is he right?

Cuts appear poor due to the visible vibration and look very-poor in comparison to our other machine, a 20yr old Wadkin 4-head CNC.

08-31-2005, 10:54 AM
.005 of play is about the best you can hope for in Bots that have gearboxes on the stepper motors. This is the amount of play in the gearbox and appears to be within the claimed tolerance of the machine. My machine has at least .006 of play in each axis but makes smooth cuts in any direction. Part of the blame for poor cut quality seems to lie in the structural flexibility of the machine, the rest appears to be an issue with the software and the way the computer communicates with the stepper drives (based on my unedumacated guess).

Disclaimer: my machine does not use ShopBot control software.

08-31-2005, 10:12 PM
When I ordered my PRT I made the decision to use 2 Y motors, one on each side of the carriage. I get no vibration whatsoever due to poor design of the carriage, the only play I get is that inherent in the motors themselves.
I think this is the most cost efficient upgrade you can make. One rack, one motor, one cable. The Accessory channel is already there.
A few months ago I blew one channel and had to make do with the original specs for a while, I immediately noticed the difference in cut quality.
I believe others have gone this route, and swear by it.
Personally I think this should be standard.

AB (Unregistered Guest)
09-01-2005, 08:26 AM
I'll pass the advice on, I have suspected a second motor might be the only option. We only rout plastics so the finish is probably more pronounced than in wood, particularly when any xy moves are present. Straight lines are fine.

09-01-2005, 12:02 PM
AB, you just made think of something with that last line: "particularly when any xy moves are present. Straight lines are fine."

This jittery issue has been mentioned many times before, but it has struck me that we seem to have been discussing jitters while the SB rounds a curve. i.e not a straight line. I seem to think that we havn't really discussed a straight line at an angle, as in hexagons and octagons etc. The difference between curved lines and angled straight lines is the fact that the mathematical processing for the computer is much more complex for curves than for straight moves. Is this high demand on the computer (and electronics) causing the jitters?

Come on, where are the tinkerers to run some tests? (Our machine is busy chugging through a big job and I don't want to interfere with my son's business right now.

09-01-2005, 08:29 PM
I ran a very unscientific test with my Alpha.

Here are the parameters:

3-hp spindle
13,000 rpm
0.500 inch 2-flute straight cutter (new)
0.375 inch depth
6.5 inch per second move speed
Vacuum hold-down (I could not physically move the material even when using all of my significant weight.)

I cut the following in 1-inch thick MDF. The test was run twice, once with the cuts going in the conventional direction and once in the climb cut direction:

7x5-inch rectangle
7x5-inch rectangle rotated 45-degrees
8-sided polygon with each side 2-inches in length
6x2-inch elipse
2x6-inch elipse
6x2-inch elipse rotated 45-degrees

The 0.50 inch cutter and shallow cut depth were chosen to factor out cutter flexing. The spindle's RPM and move speed were selected to give a chip load of 0.015, which is what I use with MDF. The cut was much lighter than the cuts I usually make.

NONE of the cuts were satisfactory (although none were particularly bad either - they all looked exactly like the cuts I've been making for more than one year, since the Alpha arrived). The cuts using only the X-axis or the Y-axis were the best, showing 'chatter' in the ramping sections of the cut. The elipses showed significant 'chatter' in the tight curve section of the cut and moderate 'chatter' in the portions of the cut that had the least curvature. The 8-sided polygon showed the most 'chatter' on the 45-degree sections and the least 'chatter' on the horizontal and vertical sections. The rectangle that was rotated 45-degrees had more 'chatter' than the un-rotated rectangle.

Now, let's talk about 'chatter'. If my math is right, a 2-flute cutter rotating at 13,000 RPM would have a cutter striking the material 433.333 times per second. If the axis was moving at 6.5 inches per second, there should be a cutter tool mark every 0.015 inches. The 'chatter' mark should be similar to the tooling marks I get when I feed material too fast through my joiner, or too fast when edge routing on the router table; that is to say, I should see a mark every 0.015 inches along the path of the cutter. In other words, the only deviation that I should see is a cutter mark every 0.015 inch. The curves should be perfect, except for a tool mark every 0.015 inch. The 45-degree portions of the rotated parts should be perfect except for a tool mark every 0.015 inches, etc.

The 'chatter' was very similar to the photos that I posted May 13, 2005 (Techniques for Cutting, Drilling, Machining/Quality of Cut). Rather than uploading a bunch of new photos, please just visit that post.

My conclusions from the tests I made today is that the Shopbot PRT Alpha is NOT following the tool path exactly generated with PartWizard 2. The test was too simple to determine whether the cause of the problem was machine flexing, or axis drift (caused by 'floating' the X-axis on top of the V-rails, etc.), or limited computer CPU cycles, or a combination of all three factors.

Bottom line is that I can live with the problem until someone can identify the cause of the major factor causing the 'jitter'. If the cause can be eliminated without re-designing the entire machine, I'll probably implement the change. However, if the problem has multiple sources and if none of the individual sources makes a significant improvement, I'll leave the machine alone.

09-01-2005, 09:02 PM
Mike: I don't think your test was "unscientific" at all. In fact, after having numerous posts and emails on this subject (chatter) being addressed and analysed over the last several months I have come to the conclusion that rather than it being a problem with either PW or the tool itself the major culprit is actually the sb code itself. Although those who have stiffened up their gantries and their carriages have seen some improvement, the only persons I have heard from who have ended up experiencing smooth cuts comparable to a large investment machine have been those who have abandoned SB code and gone to Gecko drives and Mach2 software. I don't really fault SB for this as the development of a system that works that smoothly AND has such easily learned language in concert with a machine priced as reasonably as a shop-bot all in one package is a big chunk to bite off. I do feel, however, that if the main thrust of SB is to produce great running entry level priced machines then they should "bite the bullet" and come to terms with the fact that other controllers and software produce a superior end product with their machine. My overall impression is that the SB controlling system is still a "beta" product and that, despite our appreciation and loyalty to the complete SB system, there is already a solution to this situation out there and if the solution is that high a priority on our agenda then it is our turn to bite the bullet and spend the "upgrade" bucks to fix it.


09-02-2005, 12:55 AM
I partially agree with you. But in my opinion, the code is probably solid but the electronics might be underpowered. I base that opinion on the fact that the code gives a very simple command to move in a certain direction (sometimes all three axes). If the CPU/electronics can't translate that command into the necessary steps to accomplish the task, with synchronization, there will probably be an error of some sort.

As an aside, I've bought some Gecko G212 drivers to drive some old Oriental PH299-3 drives that I've had for ten+ years, a breakout board and the necessary power supply components to build the electronics for a horizontal boring machine to replace the Delta unit that I have. At some point, I might try connecting the breakout box with the Mach1/2/3 software to the Alpha to see if that software package can handle multi-axis movement better than the stock Shopbot controller. However, I've learned long ago to not 'fix' something that is working (even if there is a tradeoff between hoped-for features and reality).

09-02-2005, 01:04 AM
Mike, that is the most scientific test on this issue in all of living history! Thanks!

Before commenting on all of that, I wonder if you could try something else as well....? The issue of cross-talk between X&Y can maybe be probed. Do you think it is worthwhile to drop the y motor, block the y-car from moving, and then run a circle or ellipse. The cut will be straight while the y-motor turns in the air (unfortunately not drawing many amps).

09-02-2005, 01:25 AM
That's a good idea. But, since I'm by nature fairly lazy, I'm wondering if anyone out there has a digital storage osciliscope that can read and store the date for two axes at one time. If I understand steppers correctly, the minor axis (the one with the fewest number of steps), will pulse at a constant fraction of the major axis (including the ramp-up/ramp-down portion of the move). A digital storage scope would record that pattern so that it could easily be examined 'after the fact'. Of course, to make the test valid, I assume that it would be best to record straight line segments instead of elipses and curves. A 45-degree line segment would have equal pulses on two-axes, while a line segment with a different slope would have a mathmatically different pattern. In any case, a digital storage scope would show the ratio of the pulses as well as the width of the high and low portions of each pulse. The timing of the pulses might be perfect, but the width might be below minimum pulse-width specs, thus compromising the accuracy of the move. If the Shopbot controller passed that kind of test, it would be show that additional strengthening would be the most productive first step; however, if the shopbot controller failed that kind of test, it would show that it's time for the Shopbot engineers to upgrade the controller interface.

AntonyB (Unregistered Guest)
09-02-2005, 06:05 AM

Just to clarify, when I said straight lines I meant lines where only 1 motor is in use, when they are both on thats where I get the cutting issues which I believe (and have now been told) are directly related to the frame of the machine, particularly the 1-motor Y-Carriage.

Whether its a 45-degree line or an intricate curve the finish appears to be the same.

09-02-2005, 07:52 AM
Mike: It seems to me that the chatter is caused by a variety of ailments. I'm not sure that there is any one "major" culprit. There is no doubt in my mind that the flex and out of balance gantry is a contributing factor and a straightforward solution similar to Brady's or a complete gantry replacement such as Steve's makes excellent improvement in accuracy. PW can probably be eliminated as one of the culprits since, although it can directly output only SB code, Its' big brothers (Pro and Insignia) are regularly used by perfectly smooth cutting machines.
But it has been shown that the overall accuracy improvement does not eliminate the chatter. For that we have to go further and change the drives. Of course, once we do that we can no longer use SB code and hence my comment that the code is part of the problem. In no way, shape, or form do I consider myself either a programmer or an electronics techie, but since this whole situation has been solved by others willing to go the whole route I find it difficult not to accept their conclusions. The main thrust of my comments above are directed at SB and it would be interesting if at some point they would chime in on one of these many threads on this topic with their thoughts. After all, their people are the ones who do that for a living while I am just a woodcutter relying on my tools to help me make a living.


09-02-2005, 10:00 AM
We're in agreement on all of your points: Gantry needs stiffening and SB code works only with Shopbot's controller. Needing to replace the drives, however, is probably not necessary on the Alpha machines, since the Oriental Motor Alpha drives can be driven with a break-out box and Mach1/2/3 software. The difficulty in using Mach software is that all of the cut files would have to be changed to G-code, which means new CAM software (such as Insignia, etc.).

For those of us who are following the new Gecko G100 series of breakout-box/controller/Mach interface, it seems that a viable computer/electrical solution is almost available. However, even if all 'jitter' that was attributable to the electronics could be eliminated, there are still the mechanical issues that need some work.

As a simple example, just look at the X-axis rails and V-rollers. The V-rollers ride along the top of the V-rails and the carriage rolls along very smoothly; however, nothing controls the up/down motion of the X-axis except the motor springs and the weight of the machine. Adding a second V-rail and a second set of V-rollers would 'trap' the X-axis and eliminate the possibility of up/down motion (and also be a source of spectacular damage if the X-axis crashed, since the weakest point in the X-axis mechanicals would fail first - can anyone say 'spindle'). Also, there is no direct stabilization of the Y-motion of the X-axis. That would require a V-rail and V-rollers mounted horizontally to control that source of jitter.

Fixing the X-axis alone would add substantial cost to the machine - and - would only solve the X-axis jitter. The other two axes would also need stiffening and enhanced hold-down rollers/rails.

Of course I would prefer having a machine that totally followed the programmed tool path, showing no chatter except tool marks, but I would much rather have the Shopbot, that comes pretty close to doing all that I expect it to, than have no CNC router.

09-02-2005, 12:23 PM
Unfortunately I do not believe the same can be said of the drives for the PRT. Upgrading to an Alpha and then adding on the breakout box and Mach software is not only not cost efficient but seems contrary. One can, if you wish to retain PW, change to the Artcam release of it (Artcam E?) for a relatively low cost and you never know, they may even come up with an upgrade price or a converter add on for existing users of PW. I have Pro, but interestingly enough, find myself doing occaisional straight cutting files in PW. The very lack of many of the tools actually helps me generate some files quicker. Keep us updated on your experiments-they are very valuable.


09-02-2005, 12:59 PM
Quality of cut (http://www.talkshopbot.com/forum/messages/29/7835.html) is the thread with pics that Mike mentioned earlier.

Some more references to the jitters:

Accuracy of PRT vs. Alpha? (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=312&post=26699#POST26699)
Noisy motors + rough cuts (http://www.talkshopbot.com/forum/messages/26/1370.html)
Deflection in Z carriage (http://www.talkshopbot.com/forum/messages/27/8758.html)
Hold downs (http://www.talkshopbot.com/forum/messages/7/876.html)
turnbuckle tightness & greasing... (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=29&post=3104#POST3104) <---see last post in this thread
Should there be play in the Vexta steppers? (http://www.talkshopbot.com/forum/messages/26/4255.html)
Jerky motion on PRTalpha (http://www.talkshopbot.com/forum/messages/29/6541.html)
Welded PRT Gantry (http://www.talkshopbot.com/forum/messages/28/9164.html)

09-02-2005, 04:23 PM
In your 1st post " Quality of cut" have you been able to answer your #2 question by improving the cut quality after installing your spindle & without any other modifications?

09-02-2005, 04:29 PM
The spindle was the single most important 'accessory' that I've added to the machine. There was an improvement in the cut quality, BUT, the problems that have been discussed in this thread were not eliminated by replacing the PC7518 with the spindle.

The main advantage of having the spindle is being able to dial in the exact RPM needed and then knowing that the RPM selected will be the RPM delivered. The spindle is also MUCH quieter. Be aware that you might have a few issues with electrical noise. I pulled my hair out for almost a week before discovering that the USB to Serial Bridge was the main problem.

09-02-2005, 06:42 PM
Gerald: Thanks for the research to gather all those various threads and collating them. That is a truly mundane chore that is much appreciated.


10-25-2005, 03:11 PM
A simple thing we did by mistake today....

Disengaged the y-motor so that it was free to turn in space. Then moved the gantry using the keyboard cursor control. Right arrow moved the gantry smoothly right.

Then forgot that y was was disengaged and pressed up and right arrows together to get diagonal movement. Of course we only got x-movement, while the y-motor turned in space, but it was a rather rougher x-movement. Quite clearly the SB controller was less happy to drive the two axes simultaneously. Cross-talk?

Will see if a faster PC improves this. Maybe even try the v3 again. It is a simple enough test.

10-25-2005, 11:49 PM

If you increase your move speed by 40% and do the diagonal move again is it the same as a move in the x alone at the normal speed?

10-26-2005, 01:01 AM
Dick, a valid point, will get the stopwatch out next time.

10-26-2005, 03:21 AM
"When I ordered my PRT I made the decision to use
2 Y motors, one on each side of the carriage. "

Is this an option from ShopBot, or a user mod only? Either way how much does it cost?

Jacob K.

10-28-2005, 10:04 AM
Curious that there's been no further comment on this clever idea.
I wonder how many Botters have, or have not, done this modification.
We need a roll-call: stand up and be counted, guys!

ShopBot should indeed consider offering a modification kit to better their machines.

11-06-2005, 12:38 PM
I can't fully appreciate all the details involved but can say improving cut quality would be a very nice improvement. I know Mike's opinion and test seem very thorough and trust you others have a very good grasp of the situation. If welded gantrys, 2nd motors, or additional support of any type can be determined to be "the silver bullet fix" I would really like to know about it and what it takes to implement. Seems an obvious choice for the SB team to be thinking about.

11-11-2005, 12:40 AM
What cost me some sleep last night.......

Stepper motors inherently have steps, or ripples, in their movement. Much is made of "Microstepping", where each ripple is divided into say 10 smaller ripples. This is great for motors carrying very little load and a micro-stepping SB sounds a lot smoother.

My suspicion was that, switch the router on and make the steppers carry load, you don't hear that the steppers have reverted somewhat back to the coarse ripples. A quick search around the web this morning gave me this (http://www.micromo.com/library/docs/notes&tutorials/Microstepping%20Technical%20Note.pdf) document. Quote.... "taking a microstep does not mean the motor will actually move!"

So, the bottom line for stepper motors is that one needs the highest possible number of full steps per inch of machine travel to get the smallest ripples. This is best achieved by mechanical (gears/belt) reduction systems, at the expense of top speed.

Or, change to "servo" motors, at a supposed large expense. However, it is being discovered that a lot of low cost permanent magnet DC motors make very passable "servo" motors. Strange, although I am of mechanical bent, I feel more optimistic for a servo drive rather than spending more headaches on gears/belts getting steppers to run smoother.....

(have been tipped to look at stuff like treadmill drive motors)

11-11-2005, 01:16 AM
Answering a question from another (http://www.talkshopbot.com/forum/cgi-bin/discus/show.cgi?tpc=28&post=29327#POST29327) thread:

Here is how someone else does a belt drive:

With reference to the ez-Router Parts(16) (http://ez-router.com/catalog/index.php?cPath=32) page:

motor fitted with 16 tooth pulley (http://ez-router.com/catalog/product_info.php?cPath=32&products_id=64)

then belt to Drive Gear (http://ez-router.com/catalog/product_info.php?cPath=32&products_id=58). The relative scale of the pic is misleading, the Drive Gear is a lot bigger than the 16 tooth pulley. One side of the drive gear is a big 60 tooth pulley, while the small side is a conventional pinion gear that runs on the rack.

ron brown
11-11-2005, 07:05 AM

I worked for a man one time who wrote CNC drivers, translation files and converted large industrial machines to CNC. He told me a cautionary tale about servos and drivers.

When a stepper fails, the motor will normally die. When an encoder fails, the machine can try to compensate by accellerating... He had pictures of a gantry, Y and Z carriages and associated machinery on a shop floor as a reminder of one such failure. A light truck would need several trips to haul off the pieces, After, it had been cut into several 1 ton size chunks.

He brought some interesting points...


11-11-2005, 10:01 AM
Ron, then I'll do something that I havn't done yet........fit limit switches.

11-11-2005, 10:07 AM
The document that you linked to in your 12:40 am post was interesting. I've heard of companies that have tried 256 microsteps per full step. (I went to a seminar at Oriental Motor several years ago where several engineers from HP had questions about the possibility of increasing the number of microsteps for application in their large paper plotters.)

Theres a stepper motor document at http://www.geckodrive.com/photos/Step_motor_basics.pdf
that has good information about 1/10th step microstepping. Page three talks about the 'corner speed' of a motor where the motor basically has full power. Page fifteen talks about the 1/10 of a full step error range (which, I believe is why Mariss choose 1/10 step for his microstepping design).

Going back to the 'corner speed', at aprox. 636 microsteps per inch of movement on my Alpha, any properly sized motor would be in it's power range whether it was directly connected to the driving gear or geared down 1:2 or 1:3.

Trying a DC servo is on my list, but first I'm going to try using an encoder on a stepper, similar to the Oriental Motors Alpha motors that Shopbot uses - which are excellent. I haven't had one error in the last 1-1/2 years that I can trace to the Oriental Motors driver or steppers. I believe that the Shopbot controller card is also working perfectly, but, since I don't have a storage oscilloscope, I can't study the two-axes pulse train to see whether the 'chatter' that I get on multi-axes moves is electrical or mechanical or some of each.

11-11-2005, 12:52 PM

My expertise is RF, not digital, but I can arrange access to a storage scope. If you can describe the measurements you want, I can send you the results.

Brady Watson
11-11-2005, 12:55 PM
"taking a microstep does not mean the motor will actually move!" That is 100% correct. I have explained some of this here (http://www.talkshopbot.com/forum/messages/312/8846.html)

The PRTs with 1/4 stepping drivers sound noisier, but at each 1/4 step they step at a torque detent. 1/10 stepping drivers sound smoother...and may help to smooth out the cut at a reduction in applied resonance and vibration...but you also may be losing microsteps as a result because the motor is not stopped on a detent.

DC servos would be the way to go for the ultimate setup...but it is important to choose a setup that can be tuned 'tight' enough with your gearing and encoder combo. The driver/amp is also critical. I have found that the Rutex drives are much tighter than Gecko servo drives and have a few other features that I like better...such as the ability to set how many encoder pulses before the machine is deemed a 'runaway' and then it takes corrective action. The key with a servo system is to realize that gearing is critical...think of a servo's power band as being the inverse of a stepper (for the most part). You can achieve this with belt reduction...however keep in mind that the combined error of the rack&pinion + belt reduction may introduce backlash into the system...they do have 'zero-backlash' belts out there...but I don't know for how long they are zero-BL.


11-11-2005, 02:15 PM

The easiest test that I can think of is to simulate a 45-degree cut. Both the X-axis and the Y-axis should have equal pulses. Or, in other words, both pulse trains should overlap (or nearly overlap, noting that there might be a small, but consistant, delay between the two axes due mostly to the ability of the controller board to generate the two separate pulses). If the Y-axis is not exactly following the X-axis (or vice-versa), then the cut would not be exactly 45-degrees (and it would produce 'chatter'). On the other hand, if the pulses to both axes occurred at exactly the same time, then we could rule out electronic problems and focus instead on mechanical problems.

The 'chatter' that I have seen in the actual cuts, occurs during the ramping part of the cut. Usually, once the motors are at speed, the cut is good.

A much more difficult test is to look at a circular cut; however, that test involves eight separate segments, where the x/y axis change their master/slave relationship eight times. So, in my opinion at least, if a 45-degree cut shows proper timing, an arc or circle would also show proper timing between the axes.

11-11-2005, 02:49 PM
Mike, I am seriously thinking of running a test where the x-channel runs off a standard SB controller and the y-channel runs off a second PC with Gecko/Mach. This should clear up once and for all if there is a cross-talk issue. No?

(My problem is that Sean is so busy getting Xmas orders out that I can't tinker with the machine)

11-11-2005, 06:03 PM
Running two axes from two separate computers would certainly clear up the cross-talk issue, but how would you syncronize the two computers? As far as I've read, neither the shopbot controller or Mach3 sends out a clock/pulse train that could be used to syncronize two machines. (My preliminary design for a controller, for a customer that went bankrupt before the project was finished, was based on using multiple Rabbit modules. There was going to be one Rabbit module per axis and a master Rabbit module that would produce a master pulse train to syncronize everything.)

Brady Watson
11-11-2005, 10:55 PM
Doesn't the Gecko G101G-Rex use Rabbit modules? The only thing up in the air now is the software (Mach4) to run it.


11-11-2005, 11:29 PM
Mike, this would be for a crude test only, and "precise" synchronization would not be required - I was planning to push two buttons on two keyboards. A 100mm circle at 50mm/sec takes about 6 seconds and my manual-synchro wouldn't be that bad.

Where are the guys with 2 (or more) ShopBots? Won't one of you do this little experiment?

(Just swop over y-cables)

11-12-2005, 11:41 AM
The Gecko G1xx series uses the Rabbit 3720 module. I've had the G101/G102 boards and several 3270 modules for a few weeks to play with. Art's MachIV software is comming along, but it isn't ready right now; however, given the short time that the hardware has been available, both Mariss and Art have worked wonders.

Your experiment just might work - if your fingers are sycnronized. If I remember correctly, most circles are cut as eight segments, meaning that ramping might be a major factor. Even if ramping does play an important part, it should be consistent on each of the eight segments of the circle. So, if there's a 'bobble' it should be show up at the same point in each of the eight segments.