[emvali]

I am working with a TC4 with DX32 control, GibbsCam and a twisted postprocessor converted from imperial to metric.
Everything works perfectly but I have this annoying problem:

No matter what, the post is converting any circle in quads and while milling a small delay is made at every 90 deg. leaving a mark on the part.

As there is nothing inserted in the G Code (like dwell for example) I am wondering if there is anything to set on the controller to cancel this delay between arcs or something to be added to the code.

Any help will be highly appreciated.

[ikneb]

I have a Johnford that does the same thing I was thinking it was slop in the ballscrews.

[emvali]

I made some investigation, maybe someone knows how to solve the problem:

There are two issues with this machine:

No. 1

The circular interpolation G Code is divided in quadrants if using I,J parameters. The machine makes a stop at the end/beginning of each quadrant. If R is used, it makes a continous path.
I have a problem "convincing" the post processor to make R based arcs, while for short programs I can do it manually I can't immagine doing that for a long program.
This is when machining a round part, the problem is when doing complex parts, where for example there are arc after arc sequences, it again pauses between each start/beginning of a new arc.
This is not happening with linear-arc motion - if I make a rectangle with rounded corners for example it goes smooth

As there is nothing in the program to cause this pause, the problem is with the machine's controller, the question is if there is some parameter that is not set correctly?

No. 2

The machine makes weird knocks from both X and Y motors at tangent moves. Just for example, if I brake a spline toolpath in short linear segments the result is like a pick hummer knocking.
It is happening when the angle between 2 linear movements is bigger than 90 deg. If I make a rectangular part everything goes perfectly, as the angles are at 90 deg. But a hexagon for example will cause a knock at every corner.
I can reduce the intensity of this knock by inserting a deceleration at the transition point, but this only reduces the knoks, not eliminating them.

I am willing to use another CAM software if this solves the problem, but I am sure the problem is with the machine. I am afraid that these knocks and pauses are not adding to the machine's longevity.

I am really interested to know if other TC with DX32 users have the same problems and how they deal with it.

[KTD1]

The knock you are hearing is likely a free play issue in your screws. My machine when pushed on short moves like 0.01 inch 3 axis point to point moves makes a bit of a growl. You can feel the machine vibrate as the drives move, stop, move, stop, for each point.

The DX-32 controllers can only look a couple of lines ahead so they are somewhat limited on speed on these sort of programs. There are a few things you can do.

1) Using the radius or fillet code/command makes the machine look ahead in the program and that is why the dwell is reduced. Invoking cutter compensation (G41 or G42 if you are only 2D machining) will do the same thing.

2) The marks at tangent points are from the material that is being removed in the “dwell” from the deceleration then acceleration from move to move. The material it is removing is tool flex. The marks are much less if you run a dry (second) pass.

3) Reduce your feedrate. On my older DX32, 20 in/min is about all she can do smoothly. On my newer one, I can run upwards of 40 in/min.

4) Use a lower programmed feedrate and feedrate override to 150%. Don’t ask me why, I just know it works.

5) There is a deceleration override G code but I cannot remember what it is and don’t have my book in front of me. Seems like G06 or G08 (I use the modal command) will allow the machine to move through the points without deceleration but you must be careful when using the command. You must cancel the override before making any rapid moves or before making any real sharp turns. Without the deceleration, the drives can miss their mark and fault out.

Hope this helps

[emvali]

Looks like you are right, there is something loose at the screws. So far I had the impression I hear that knock at tangent moves only from X, but now that I checked is also caused by the Y.

Any idea what the screws repair involves?

Also, there is a problem with the spindle, at max. speed (8000RPM on the TC4) it is impossible to stand near the machine, the spindle makes a high pitch loud noise which is not normal for sure.

What is the recommended source for BP parts?

[neilw20]

G61 is exact stop mode.
G64 is constant velocity mode.

You might find G64 (if controller recognizes it) removes the pause in the circles.

If you have backlash compensation setup and enabled, everytime a leadscre changes directions, all others must stop and wait for the backlash takeup to finish.

[machintek]

The DX controls did not like 3D moves. Some VMC machines sounded like they would walk across the floor in 3D. Later software on the VMC was better and required retuning the Siemens drives. (At one point there was a class action lawsuit about this but Bridgeport went belly up). The TC4 has Yaskawa drives and I do not know anyone doing 3D with that combination (Yaskawa and DX32). I know the drives can do it because I did some very fast machining of a small 2 inch square Mickey Mouse with a FANUC control with optional look ahead on a TC3. I did drip feed from a PCMCIA card.
Talk to EMI (emi-inc.net) as they were selling BMDC4 boards and other goodies to assist in 3D work for the DX32 controls.
Check the backlash on the machine. First the actual/physical lost motion. Then see what the parameter is (maintenance page, arrow down to backlash). This will give you a idea of the total backlash in each axis and if you have a problem.

George

[emvali]

Thanks George,

The problem is not with 3D, I tried several times and the result is indeed very poor. Take a look at the attached drawing:

The exterior contour of this part is made with no problems and no pauses.
The interior starts the "hammering" at every corner.

Now, if you go deeper on this part, the bottom of the rounded part, inside or outside, if machined by slices - 2D movement only the toolpath is broken in arcs and short linear movements - here the disco begins, no matter how hard I try to reduce speeds at arc-linear transitions.

Now I need to ask this:

What is the procedure for measuring the backlash and how much is allowed to be entered on the controller, I mean how much can the controller deal with.

[emvali]

I found all the disks with backups of the machine's parameters made by the previous owner and the original disk supplied with the machine.
I pasted everything in an Excel worksheet, compared them and surprise !!! nothing was ever changed, all important data is the same as what the machine arrived with. (I compared backups made in 1999, 2003, 2006 and 2007 with the original)

So the backlash was never adjusted.

Looking at the different parameters in the list I found something that maybe is related to the problem:

[77]
t=Integral Following Error Gain
d=This value is used for a limited time during a reversal in the direction of travel. It allows the accumulated integral following error to "slew" rapidly to a new stable state. Too large: causes bumping during reversal. Too low: swings the axis wide during reversal.

[78]
t=Change-In-Direction Counter Initializer
d=Initializes a counter when the axes reverse direction. If non-zero, the Integral Following Error Gain (parameter 77) is used instead of the normal gain (parameter 79) for the integral following error term. Too large: stability problems. Too low: bumps on reversal.

[79]
t=Normal Gain For Integral Following Error Term
d=Main term for the position loop. The function of the position loop is to maintain a zero following error whenever a move is not active. If this gain is too large, the axis will oscillate slowly. If this gain is too low, the system may have excessive following errors and/or may not adjust to changing following errors quickly enough.

[80]
t=Following Error Integral Term Magnitude Limiter
d=Limits the following error integral term "magnitude". If too low, the integral term may not grow large enough to push the axis into position. If the integral term grows too large, it masks severe mechanical problems until something breaks.

On my machine these parameters are set as follows:

77 = 10
78 = 3
79 = 20
80 = 40.000

Any ideas how the values are supposed to be calculated?
The way BP wrote the manual is quite idiotic.... "if set too low it may do that.... if set too high it will break something"
So how should the correct value be determined?

[KTD1]

I found all the disks with backups of the machine's parameters made by the previous owner and the original disk supplied with the machine.
I pasted everything in an Excel worksheet, compared them and surprise !!! nothing was ever changed, all important data is the same as what the machine arrived with. (I compared backups made in 1999, 2003, 2006 and 2007 with the original)

So the backlash was never adjusted.

Looking at the different parameters in the list I found something that maybe is related to the problem:

[77]
t=Integral Following Error Gain
d=This value is used for a limited time during a reversal in the direction of travel. It allows the accumulated integral following error to "slew" rapidly to a new stable state. Too large: causes bumping during reversal. Too low: swings the axis wide during reversal.

[78]
t=Change-In-Direction Counter Initializer
d=Initializes a counter when the axes reverse direction. If non-zero, the Integral Following Error Gain (parameter 77) is used instead of the normal gain (parameter 79) for the integral following error term. Too large: stability problems. Too low: bumps on reversal.

[79]
t=Normal Gain For Integral Following Error Term
d=Main term for the position loop. The function of the position loop is to maintain a zero following error whenever a move is not active. If this gain is too large, the axis will oscillate slowly. If this gain is too low, the system may have excessive following errors and/or may not adjust to changing following errors quickly enough.

[80]
t=Following Error Integral Term Magnitude Limiter
d=Limits the following error integral term "magnitude". If too low, the integral term may not grow large enough to push the axis into position. If the integral term grows too large, it masks severe mechanical problems until something breaks.

On my machine these parameters are set as follows:

77 = 10
78 = 3
79 = 20
80 = 40.000

Any ideas?

From someone that knows just enough to get in trouble:

Gains and backlash are 2 completely different settings. Unless you know exactly what you are doing, I would leave gains alone. This is not to say that yours may not need adjustment, just be aware that improper gain settings can cause an axis to become unstable and runaway uncontrolled. I check and adjust my backlash from time to time. As my machine has a few miles on it, the backlash is slightly different when working in the middle of the travels than it is at the end of travels. Because we often need true positions within a 0.001” on larger plates, I try to average the backlash error for all areas in the travels.

To check my backlash:
1)I zero my test indicator against something affixed to the table (vice jaw or work piece).
2)Jog the axis so that the block moves away from the indicator
3)Jog the axis so the block moves towards the indicator until the indicator reads zero again.
4)Set machine zero.
5)Jog the axis into the indicator a few thousandths past the zero.
6)Jog the axis away from the indicator only until the indicator reads zero again.
7)The reading on the machine’s axis display is your backlash error.

Some machine’s ballsrews have an adjustment on the nut. I don’t know if your model does or not. If you have the adjustment on the nut, I would adjust that first and then set

[emvali]

Thanks KTD1,

The reason I asked is because all these parameters mention "bumps" at axis reversal, which is exactly what is happening on my machine.

Some time ago, I removed the way covers and checked for something loose visually. The knoks are sometimes loud enough to expect something really loose, that can be seen - but everything was tight in place.

Some machine’s ballsrews have an adjustment on the nut. I don’t know if your model does or not. If you have the adjustment on the nut, I would adjust that first and then set

How does this adjustment nut looks like and where it is supposed to be located?
Unfortunately, in my maintenance manual sections 4G.2.6 REPLACING THE BALLSCREW BEARINGS and 4G.2.7 REPLACING THE AXIS BALLSCREWS are both on a "Procedure to be provided at a later date" page.
Murphy's laws at their best.

[KTD1]

It would be at assembly that makes up the nut, usually affixed to the table. In essence it is two nuts assembled so that one can be adjusted to oppose the other thus eliminating clearance.

Have you checked your backlash to see if in fact that is a problem?
You would need to locate the current settings (parameters) to find if they have been set at a fairly high value. My settings are about a thousandth of an inch. I’m guessing if they are several times that, the knock is the machine compensating for the backlash and nothing besides ballscrew adjustment or replacement will remove the knock.

If you have the covers off again, check for play (radial & thrust) in the bearings that hold the screw. It might be they are allowing the screw to move and make noise.

[machintek]

The X and Y ballscrews are in tension. The machine has linear ways. There should be very little backlash.
If there is a klunk as the table reverses direction in rapid jog, that would be a problem. Make sure it is not the sliding covers that are making the noise.
As i said these controls do not like 3 D work and make a lot of noise as if hitting something depending on the positioning move. To get around dwells with my BOSS 9 I used a smaller end mill and walked around a radius instead of doing a 90 degree move.

George

[emvali]

I will check the backlash soon and post the results.

George,

The problem is not with 90 deg turns, as a matter of fact the machine is doing them ok. Also, linear movements such as zigzag facing are made with no noises at all.

This is what's strange, if you look several posts down, I attached a drawing of a part with descriptions. While I do a rectangular contour with or without radiuses it is done without any noise.

But a tangent move no matter if sharp or with radius causes this knocking.
There is no screw direction change at the tanget point if we think this way.

Suppose I go from 0,0 to X50 Y0 and then X75 Y25 at the tangent point the X continues to move up and Y is starting. This point causes a knock.

If I do a 90 deg turn instead, such as 0,0 to X50 Y0 and then X50 Y50 it is done with no noise.

I could understand a backlash problem if I go back and forth with the same axis and get the knock at direction changes, but this is not happening.

Please take a minute to look a few posts down at some paramenters I found in the list.

By the way, the noise is not from any covers, it comes from the table, sounds like a deep sound and felt even on the concrete floor.

Thanks,

Emanuel

[machintek]

I believe your problem is not mechanical. It is in software. I do not play with these parameters. No time and no place to experiment..
The person with the answer is Bart Stator at EMI-Inc.net. He has played with these for at least 13 years that I am aware of. He has written software for Bridgeport.
And yes, I did click on, and expand and study (very fancy and beautiful drawing) your attachment with all your notes the same day you posted it.

George

[emvali]

I made a quick backlash mesurement. The problem is the machine is set with an adapter jig made of aluminum for a part we are working on now and I didn't wanted to play with the 0 settings.

Apparently, the X is 0.01mm and the Y 0.014 which equals to approx. 0.0004" on X and 0.0005" on Y.

I also checked the setting on the DX32, it is set at X 0.005, Y at 0.003 and Z at 0.(in mm)

While the measurements are reasonable, they are different from the settings.
I also compared them with the settings in the original disk that arrived with the machine and they are the same - it looks like nobody ever made backlash adjustments on this machine.

Looking at the parameters list numbers 55 to 57, I have the X set on 0.00508 and the Y at 0.00254 which are imperial to inch direct conversion, of 0.0002" and 0.0001" respectively.

Maybe it is time to make this adjustment as it is not possible for a machine of this age to remain with the same parameters for 12 years.

I will make precise measurements, the question is how much the backlash can be adjusted by the computer.

Regarding the mechanical adjustment, the machine is with Yaskawa drives - maybe you remember if this setup has the adjusting nut that KTD1 is talking about?

[machintek]

Very seldom do we need to adjust the backlash on these machines. It is understandable to make corrections on a machine of this age.
Before I make any adjustment, i make checks at the center of travel and at the ends of travel. Typically, there is more wear in the center. If there is a difference between Center and the ends, you have to make a decision, how much of this lost motion to take out. If you take all of it out in the center, you will be overcompensated on the ends.

George

[emvali]

I measured today the X and Y backalsh.

While on X is more or less steady at edges and center, about 0.013mm which is some 0.0005" there is something strange with the Y.

The reading at the center 0.005mm = 0.0002" is almost half of that at the edges, 0.011mm = 0.0004".

I have no idea how is this possible.

I do have the strong feeling however that the knocks I hear are from the X, rarely I get from Y.

So.... the X is obvious, but what should I do with the Y?

With Yaskawa drives, is there a mechanical adjusting nut to begin with on these machines?

Emanuel

[emvali]

This motion knocking problem refuses to go !!!

It is almost 2 months I try absolutely everything, including 2 specialists one of them a designer of motion control systems.

The problem is for sure NOT backlash and not mechanical.

I got the Yaskawa control software for the SGDB-15ADG XYZ controllers, prepared a cable and got the in the trace window clear spikes at all the points where one axis is in motion and the other begins to move.

The spikes are there on both the X and Y.

All parameters in all 3 controls are the same and unchanged.

All the machine's parameters are unchanged too, as I checked the present BMDCPRMS file and compared to the one in the original disk followed by several backups disks made by the previous owner. (by the way, he claims he never heard any noises from the machine)

So everything is ok with the parameters, no mechanical problem, but this knock is driving me crazy.

George, by any chance you have Mr. Bart Stator's email?

Anyone with TC machines, can you please make a copy of your parameters file BMDCPRMS.SYS and attach it here or to my email?
I am almost sure something is wrong with a parameter and like all the big problems it probably has a minor repair.

I really appreciate any help

[HuFlungDung]

Since we're grasping at straws here, maybe you should check whatever couplings exist between the motors, ballscrews and encoders. A loose keyway can have strange effects, depending on where the encoder is mounted. I would even go so far as to back off the setscrews on the ballscrew-motor coupling to see how much real slop exists. A setscrew is not, in and of itself, able to compensate for a worn key or keyway, no matter how tight you tighten it.

If the machine has real mechanical slop causing backlash, you'll probably still get marks at the quadrant lines of a full circle. In fact the part may look better with no backlash compensation at all. But, it may not be accurate either. Real backlash elimination requires repacked ballscrews and/or new thrust bearings. All bearing surfaces and retainers need to be checked for looseness and wear.

If possible, observe the rotation of the motor shaft when you reverse direction at the smallest increment. See if it seems excessive relative to the real distance moved.