Ugh.... Z axis keeps getting currupt

Any thoughts? I'm am trying some test cuts with a machine that gets about 1/4 through a job and then the Z-axis sinks into the project. I am using G-code from a program called "Halftoner" by Jason Dorie, that creates lines of varying thickness based off a photograph. The job is cut using a V-bit and the depths of cut are between -0.01" and -0.06". I was test cutting into a piece of foam. About 1/4 through, the bit just plunged about 3/4" into foam and kept going, EMC was displaying the correct depth for that cut, around -.015. The feed rate is only about 17 in/min.

A followup... it looks like it always happens after a movement at safe height. The command following those movements is G00 Z0.0 to get back to surface height, then the combined x/y/z movement continues for the engraving.

On this last one, it sounded like the Z motor stalled a little bit on that move before it plunged. I wonder it is trying to move the Z axis at the feed speed? I noticed there are F40 statements for the feeds, but there are no instructions as to the plunge/retract speed. Does LinuxCNC just try to make those movements as fast as it can?

On that note: my engraving pass code looks like of like this:
N106720 G1 X0.1301 Y1.6136 Z-0.0368 F40
N106730 G1 X0.1301 Y1.6136 Z-0.0368
N106740 G1 X0.1476 Y1.6136 Z-0.0368
N106750 G1 X0.1651 Y1.6136 Z-0.0370
N106760 G1 X0.1826 Y1.6136 Z-0.0370
N106770 G1 X0.2001 Y1.6136 Z-0.0370
N106780 G1 X0.2176 Y1.6136 Z-0.0372
N106790 G1 X0.2351 Y1.6136 Z-0.0372
N106800 G1 X0.2526 Y1.6136 Z-0.0372
N106810 G1 X0.2701 Y1.6136 Z-0.0374

Is it accurate to only need to specify the feed rate once? One thing is though the G-code is set to 40, the machine is only going at about 17 in/min.

Once you set the speed with F40 all G1 moves will try to move at a feed rate of 40 until it is changed.
However if your machine cannot run at that speed it will move as close to that speed as possible.
Try going back to your configuration and reducing the velocity and or acceleration of the Z axis and retest it to make sure it is not missing steps.

Also try reducing your speed by changing the F40 to say F10 to see if the problem goes away.

Rick G

As rick says your loosing steps . this is quite a common problem and can be due to a few things ..

most common cause is trying to move the axis faster than it is capable , this is apparent most on the z axis due to the natural weight of the carriage wanting to follow the force of gravity , i.e. it always increases steps in the downward direction
causes are :

mostly too high velocity and or acceleration setting, stepgen and dir values too low , microstepping too high , people mistakenly use microstepping to increase the number of steps per rev , this is not it's main purpose .
and power supply not being able to sustain the load or voltage required .

most cases reducing values will bring you back to normal .

there are two internal values used for motion , one for none cutting moves ( G0 ) and one for cutting moves ( G01 etc ) so if you find cutting moves are ok , but you loose steps on fast none cutting moves or when returning to machine zero
one of the above will be the cause , so altering your ini values should do it , G0 moves use max velocity and max acceleration values , where G01 moves are more controlled via Gcode feedrates .

don't forget to set your touch off height at the top surface of the material your cutting , this is usualy set at Z0 , then all cutting moves are then negative .

looking at your picture you seem to have a lot of side movement and the cutting bit is being pulled off course , by being grabbed by the foam , so look for excessive play in slides etc , also when cutting slots use an slot mill rather than a endmill
and don't try to cut a 1/4" slot with a 1/4" mill , use one slightly smaller and then cut around the edge of the slot , this will give a better finish ,

The weight of the carriage certainly makes sense with regards to the missed steps always being down, however the timing of it seems unintuitive. With a Z-axis leadscrew pitch of 12 turns per inch, that's a LOT of turns to sink as far as it did.

I had to turn on microstepping on my XY axis to get any decent feed rate out of them, I still get a lot of torque (260 oz motors). I turned it on with the Z-axis, too. First I'll try slowing the max accel and vel down on the Z, then I'll try turning off the microstepping if I need to.

Regarding the cut and the piece. This is a line based halftone. A photograph is converted to lines of varying thickness. The program determines the V-bit cut depth to achieve line width. For this program, all of the cuts should have been no more than 0.06" deep or so. Only the very tip of the 1/4" engraving bit should be used.

One thing I noticed after posting tonight is that the program isn't very efficient. To convert the pixels to g-codes, the program must be creating a matrix and generating a G-code command with every grid spacing. So, even an entire pass is at the same Z height, it is breaking that up into the lots of little steps. I think that is causing the feed rate to be minimized, since it must be accelerating and decelerating in each given step without getting to target feed rate.

one thing you need to remember the torque drops off , the greater the microstepping becomes , once you get past the optium for the specific motor ,
microstepping is realy only needed to help in removing resonance and not resolution , also make sure your power supply is capable of delivering the amperage needed if all steppers are being driven hard ..
steppers draw most current when stationary and not when moving , so you find the missing steps can become amplified due to the amount the z axis drops before the motors are able to come back into holding the axis .

z axis is probably best at say 1/2 step due to the above reasons

what program are you using to produce the gcode ?

Thanks, that makes sense. I'm at 1/2 step now. Good to get this figured out, since I'm planning on mounting a bigger spindle, later. Right now, I have an old Dremel in there, but I plan on using that to cut a mount for the Craftsman laminate router I just got. The controller, motors and power supply came as a kit from A Quick CNC, it was one of their older units. I'll have to double check the size of the power supply, but it seemed fairly decent.

For the software, I am using Halftoner 1.4:
jasondorie.com/page_cnc.html

I really like the effect of this software, as well as his "Reactor" software. But, I'm still in the process of getting this machine setup, hence the foam cuts for now.

When I have time, I am planning on trying out CamBam. It might be better to export from halftoner into CamBam as a DXF, then use CamBam to generate the g-code. We'll see.

A quick update - I used the kids' naptime to make some aircuts with some new settings. I slowed the Z-axis down about half and decreased the acceleration by a factor of 3. I also doubled the acceleration of the X and Y axis, which as only set to 10 in/s/s. Running tests on single axis seemed ok, I ran them with 0.1" cycle sizes to get rapid pulsation back and forth, and they seemed fine (even with me applying force to the carriage with my finger). I also ran an air cut on the piece I was doing before, it didn't plunge the Z axis at all, and the average velocity increased from 17 to 24 in/min.

I'll try it out the job in foam later... when the kids aren't sleeping.

that's good news , now just try increasing one item at a time , till you find a setting that gives missing steps , then back off to the nearest setting that works fine .

Sounds like you are making progress now, like cncbasher stated find the limits of your axis and then back off to be safe.
I like to be conservative with the numbers, don't want to have a problem 3/4's through a long job.

Rick G