Acceleration?

[islander261]

Hello

I have come across an interesting problem while upgrading my table. I had always thought that the higher the acceleration you could run your axis at the better the cut quality because the quicker you get up to the feedrate set point. So I spent quite a bit of time and money improving my tables acceleration. I am glad I didn't just make the jump to servo motors and drives because then I would be out some real money. Previously I had used a Gecko G540 driver with a 48V 12A power supply (very similar configuration to many hobby systems). This allowed me to cut at 220ipm with rapids at 270ipm at an acceleration of 55ips^2 (about .15g). So I upgraded to using Gecko G203V drivers at a 76V 11A power supply to better match the inductance of the stepper motors. After extensive testing I settled on being able to run at 350ipm rapids and an acceleration of 116ips^2 (about .3g). Now for my cutting parameters (the same this file has always used):
14ga HR steel
Duramax Torch
Finecut consumables
Copper Plus electrode
45A
.090" pierce height
.070" cut height
220 ipm feedrate

I received an order for a part that we have made over 200 pieces of in the last couple of years, so a well tested cut file. I loaded up the cut file and material and started to cut, strange sound from cut! Hit program pause to check the cut and found this:

skip1.JPG

skip2.JPG

The cuts had skips in them! The only thing on my system that I had really changed was the acceleration on the X &Y axis. Because of all the little short segment the cutting height hold was active and the torch was at .072" height (I run LinuxCNC so I save the cut data for each cut for trouble shooting should the need arise). So I shut down my system and reset the axis accelerations in the .ini file to the old value (55ips^2). I restarted my system and started cutting. Cut the whole sheet of parts out with no skips in the cut! So after all of this what is fastest recommended acceleration I can use and expect skip free cuts?

John

[beefy]

I would ask on the Linuxcnc forum about this one. If higher acceleration is making the torch switch on and off during a cut, that's a big issue for Lcnc.

I assume the cut path is still good i.e. the steppers are not missing steps and it's just the torch turning on and off.

I did look into Lcnc some time ago (but quickly looked for greener/easier pastures LOL) and if I remember correctly there was a way of testing individual PCs for suitability with Lcnc. Could it possibly be that your PC cannot handle the higher processing rate that the higher acceleration may require. Just clutching at straws here.

It is weird though that the torch would switch on and off during a cut, it should simply be on until the end of a cut.

[Rodw]

I've seen this happen, more on that later.

Just to clear this up first. This isn't a LinuxCNC issue or a PC latency issue. I don't think many non Linuxcnc users have an understanding of latency in a Linuxcnc context. LInuxCNC requires the servo thread that does all the work to be called at precise timing (think timer interrupt), usually at 1000 times a second. If the PC hardware has poor latency, this thread might sometimes run late and LinuxCNC will report the error. This latency is a function of hardware and nothing else. Often changing BIOS settings will solve the problem but if it doesn't, a change in PC is required. Also, islander261 is using hardware stepping on external hardware capable of 10 mHz per second step generation which is independent of Linuxcnc. Its possible the higher step rates violate the step pulse timing settings but Linuxcnc will also report when that happens on startup. Shortening step pulses might help if that was happening but I doubt its the issue.

I've seen this dotted line cutting happen on my system. It appears when the arc is lost, the plasma cutter would attempt to restart it resulting in a dotted line. In one case, where it started near the beginning of the cut, increasing the pierce time solved the problem. In another case, it was a stuck arcOK relay in the plasma cutter.

I suspect its somehow related to travel speed during the acceleration because you have asked the plasma cutter to exceed its cutting limits. You can see in the first photo the arc is struggling to cut cleanly leading up to the missed part.

Could it be an airflow issue? Can you measure airflow at the torch?
Is the arc too cold for that acceleration?
Can you increase the current past 45 amps? Does that make a difference?

[islander261]

Beefy

Sorry I have left some details out, I already checked the torch on signal, before I set the acceleration back to 55ips^2. I carefully monitored the torch on signal with halscope and it does not turn off. The Ok to Move signal does work correctly as it goes low when the arc goes out and goes back high when it restarts. I use Mesa cards so the stepping is handled there by the FPGA and not by the PC. I actually don't think the torch is "switching" off, it is more like the arc gets stretched and goes out. The halscope of the arc voltage when this happens shows the voltage raising then abruptly falling then raising to open circuit voltage then going to near 0 as the power supply tries to restart the arc (air pressure shut off) then raising again to cutting voltage as the arc starts again. This all happens very quickly, less than 200ms some times. Just to be clear the THC is in corner hold and not making any vertical movements when this happened. I did extensive testing using some very mechanically challenging Gcode files trying to get the system to loose steps and even after many repetitions it didn't loose any steps. I also did the usual HT instructions of changing all the consumables including the swirl ring for new parts.

Yes it is weird and that is why I am leaning to a mechanical cause.

John

[islander261]

Ok, more info.

The plasma power supply air pressure gauge on the input is showing 110 psi steady during the cut. I have no way to insert a flow meter in the torch lead.

On both photos the cut direction is right to left and top to bottom. The torch is well clear of the pierce when this happens.

This is very repeatable, the skips are nearly in the same places each cut and can be cured or caused just by changing the acceleration limits on the X&Y axis. It will be a few days before I can do systematic testing to determine the exact point of onset.

By the known cut height, freedrate and arc voltage I should be cutting at very close to 45A if the HT docs are correct.

John

[Rodw]

OK a couple of questions to rule out more of the Linuxcnc stuff.
Does your TD power supply use a relay to supply an ArcOK signal?
Have you got that connected direct to a mesa input?
Does your system respond to a loss of ArcOK while cutting? If so, what does it do?
If you have a mechanical ArcOK relay in the power supply, you could find the current draw is less than the minimum for the relay so a pull down resistor say 1k, 1 Watt on the input might be required to prevent oxidisation n the relay contacts which could result in intermittent performance (ask me how I know this?). If thats the case, your system might be turning the torch off.

Also, is it possible you have PID overrun at the higher acelleration which is lifting the torch unexpectedly? Did you retune your PID based torch control for the higher accelleration?

I think you really need to get some halscope plots to isolate what is happening in this segment.

[islander261]

Rod

It will be a few days before I can get to testing this.

The Ok to Move (ArcOK) signal on my TD is a dry relay contact closing. The circuit I have is a 180 ohm resistor connected to +24V connected to one contact and the other contact connected to an input bit of my 7i76E card with an LED with a 1k series resistor to 24V ground connected in parallel with the 7i76E input, seems to have been quite reliable up to now. So the relay is switching about .019A of current.

My system only uses the Ok to Move signal to detect the establishment of the arc prior to the pierce delay.

The halscope monitoring I have done indicates that the Ok to Move signal is functioning correctly following the presents of the arc. The torch on signal stays true during the cutting skips.

The THC is in a corner hold state when the cutting skips happen, again this has been confirmed by monitoring with halscope. So I see no reason why THC function is involved (none of the Z axis settings where ever changed).

Right now remembering what the halscope plots for arc voltage looked like this could be caused by an interruption in airflow to the torch or a momentary short (arc) inside the torch. I have inspected all parts and found no arc marks on the torch parts. I have shook, pulled and beat on the torch lead and connector trying to make this happen with no luck.

John

[motoguy]

Holy cow 116 ips^2 accel is fast. Tagging for info.

[robertspark]

What steps per unit do you have per axis?

What micro step setting do you have on the drive?

What is your drive mechanism.... Rack + pinion / belts?

Have you got some specs on the rack and pinion + belts ?

Are the drives direct, or ratio'd ..... (Geared or belted) .... What ratio?

Linuxcnc motion planner?? Trapezoidal acceleration?? Jerk / jolt?? drive mechanism backlash??

[islander261]

Hello

I have changed Gecko drives for Gecko drives so no changes to the setup other than the acceleration.

steps per unit 3183.09886
10X micro stepping
Rack and pinion drive, 24 pitch, pinion 24 teeth
5:1 planetary gear box
600oz/in 200 step 8mH steppers
LinuxCNC 2.8-master motion planner
No backlash at reasonable force levels, you can get .005" if you muscle the axis against the drive, this is a plasma table not a router or mill. The table always returns to 0,0 measured with a dial indicators at the conclusion of the torture test Gcode that really stresses things with many direction changing moves.

What are normal accelerations that others are using? I know with 4 X 4 or larger tables and usual hobby level drives you will be running in the .1g to .2g range to avoid step loss. AFAIK most of the industrial level tables normally run higher than this.

John

[motoguy]

What are normal accelerations that others are using? I know with 4 X 4 or larger tables and usual hobby level drives you will be running in the .1g to .2g range to avoid step loss. AFAIK most of the industrial level tables normally run higher than this.

I seem to recall Jim Colt saying that 35 ips^2 was the beginning of "acceptable" accel for air plasma cutting. On my Starlab, I'm running 45 or 50 ips^2, IIRC. I've got a 6' gantry, and I start to see wobble artifacts in 90 degree corners if I crank it higher.

[islander261]

Thanks, that fits in with what I recall being previously posted. So .0926g is the low end of the acceptable acceleration does anyone know what the top end is?

John

[motoguy]

Thanks, that fits in with what I recall being previously posted. So .0926g is the low end of the acceptable acceleration does anyone know what the top end is?

John

Top end will be where you introduce wobbles (due to structural issues with the table, such as gantry flex, table wobble, or torch holder wobble), or skipping steps with stepper motors.

[Rodw]

Thanks, that fits in with what I recall being previously posted. So .0926g is the low end of the acceptable acceleration does anyone know what the top end is?

John

Top end will be where you introduce wobbles (due to structural issues with the table, such as gantry flex, table wobble, or torch holder wobble), or skipping steps with stepper motors.

But surely, thats a cut quality issue not something that would turn the torch off.

[islander261]

Hello

I have done further investigation and found the following:

1. The torch on signal is not being affected in anyway by the acceleration (as expected).
2. I can run with the acceleration set at 100ips^2 with no problem but when I increase to 116ips^2 the arc starts going out again.
3. The cut direction seems to have a minor influence on the arc going out.

So right now I think this is a mechanical problem either the bearings for the floating torch action (float switch movement) or the magnetic break away moving at high acceleration (the switch does not activate when the arc goes out). The direction of movement that causes the most problem is the least stiff axis for the break away.

I replaced the slide bearings on the floating torch action with longer (stiffer?) ones as Sean P. has previously suggested. No real change in sensitivity to acceleration. Now I have set my acceleration to 90ips^2 to be on the safe side. I will rework my magnetic break away some time in the future to stiffen it up. I have also fixed my software so the the torch will be shut off and a feed hold asserted if the arc goes out so I can start the cut over at the failure point so I don't loose any more material.

John

[Rodw]

John, Have a look at the Snap 'n Cut Torch breakaway which is reasonably priced
I think I'll probably end up buying one of these as the heavier 100slv torch might be a bit too much for my home made breakaway...