Try increasing the pulse width to 15. If that doesn't work, try lowering the acceleration.
The test cut for "Emery" in pine came out fine. When I made the actual cut in the customers piece, it came out skewed.
I can cut beautiful pieces, as shown in the plaque of "Our Lady of Guadalupe" but several attempts at cutting the "Smith and Wesson" logo failed.
I have tried several different programs (Ucancam, Vectric Cut2d, MastercamX3), trying different output formats, and the results are the same. These are simple pocket/area clearance style cuts.
I have rebuilt my Y gantry and Z axis twice to be sure my machine was not the cause. Rewired with shielded cable, new connectors, and resistors. parts tested on an ohm-meter for proper line isolation and resistance before installing.
Running WindowsXP/SP2.
Intel Celeron 2.6Ghz processor
500MB ram
No screen saver, not connected to internet, configured to never shut off or hibernate, etc. All power saving features disabled.
Only programs installed are those required for the work.
Using Gecko G540, wired according to instruction manual and actual conversation with Mariss.
Mach3 with the Gecko G540 xml file.
________________________________
Anyone with any ideas as to what is going on and how to fix it?
Thanks.
Similar Threads:
- Need Help!- wrong ATC arm position-wrong magazine mill position on mori seiki NV4000 DCG
- Gecko G540 Bare Bones System (Enclosure, 48v 8.3a Power Supply, Cooling System)
- Gecko G540 Bare Bones System (Enclosure, 48v 8.3a Power Supply, Cooling System)
- Need Help!- Mitsubishi system Meldas 60/60s how to set G54/G55 tools coordinator system
- Need Help!- Converting DSP controller system to Mach based system.
Try increasing the pulse width to 15. If that doesn't work, try lowering the acceleration.
Gerry
UCCNC 2017 Screenset
[URL]http://www.thecncwoodworker.com/2017.html[/URL]
Mach3 2010 Screenset
[URL]http://www.thecncwoodworker.com/2010.html[/URL]
JointCAM - CNC Dovetails & Box Joints
[URL]http://www.g-forcecnc.com/jointcam.html[/URL]
(Note: The opinions expressed in this post are my own and are not necessarily those of CNCzone and its management)
It looks like it might be missed steps on the rapids. Some have tried the Sherline mode with the 540's with good results. Not sure when you got your 540. I think this was an occurance that only effected a few 540's. Not even sure if it has been figured out why that mode works better.
Soooo.......not really all that much help, am I?
Lee
Check to see if:
1) Ports and pins, STEP OUTPUT "Active High".
2) Motor tuning, STEP PULSE WIDTH "2 microseconds".
3) BIOS, parallel port "EPP mode".
Mariss
Thanks for the responses.
In Ports and Pins, the X,Y and Z home are the only things checked Active Low.
EPP checked in the BIOS.
Pulse was set at 2.
This is a new G540 purchased within the past month. Really brought out the power in my motors to where they are now overkill for the size of my machine. Pushing them to their max would probably tear my machine apart!
I switched the step pulse to 15 and it cut the Smith and Wesson logo perfectly. I won't have time to try anymore cuts for a few days, but will get back to post the results a.s.a.p.
So, what did the change do and why did it work?
Is there any way you could post a screen shot of your Ports and Pins settings? I just want to confirm visually that the settings are indeed correct. The only thing increasing the step pulse width does it mask an underlying problem, either with the computer hardware, cabling, or software.
The G540 requires a 2uS width and does not vary; if the drive works then it will work at 2uS. There has to be something else that is causing the problem, perhaps in the software.
Marcus
Marcus, there have been a lot of users here that have had issues with their G540's, and increasing the pulse width in Mach3 has gotten them to work correctly. The problem most likely has to do with the PC, but the pulse width seems to get people up and running.
Gerry
UCCNC 2017 Screenset
[URL]http://www.thecncwoodworker.com/2017.html[/URL]
Mach3 2010 Screenset
[URL]http://www.thecncwoodworker.com/2010.html[/URL]
JointCAM - CNC Dovetails & Box Joints
[URL]http://www.g-forcecnc.com/jointcam.html[/URL]
(Note: The opinions expressed in this post are my own and are not necessarily those of CNCzone and its management)
Gerry,
I have seen these before and we are not sure what it is that either Sherline mode or half-Sherline mode is masking, but we do know that it works for some. We have gotten several back here for testing that did not work on anything but Sherline mode and put them on a function generator with a 2uS pulse and they worked just fine. I am not sure what the actual problems are, but if Sherline mode works then by all means use it. I was on a different forum trying to diagnose what the cause was and could not come up with any common factors between the different cases, so I really don't know what it is.
Marcus
Perhaps you need to find a PC that only works in Sherline mode?
Gerry
UCCNC 2017 Screenset
[URL]http://www.thecncwoodworker.com/2017.html[/URL]
Mach3 2010 Screenset
[URL]http://www.thecncwoodworker.com/2010.html[/URL]
JointCAM - CNC Dovetails & Box Joints
[URL]http://www.g-forcecnc.com/jointcam.html[/URL]
(Note: The opinions expressed in this post are my own and are not necessarily those of CNCzone and its management)
This "Sherline Mode" business mystifies me. The G540 step pulse response time is 0.7uS so any pulse wider than that means nothing should happen afterward; the opto-LED is 'on' in the first 0.7uS, end of story.
What worries me is this can be an indication a parallel port isn't meeting IBM spec. If the port pin drivers can't quite muster the required 5mA, an extended step pulse (5, 10 or 15uS) 'on' time might just drag it across the threshold. This is dangerous because it means the port may be on the threshold of not functioning at all.
The parallel port is a "legacy port" so it may be the area where inexpensive motherboard manufacturers cut corners. The thinking may be no one will notice it doesn't meet IBM spec. This goes a long way to explaining why most people have no problems while a few others have to use the "Sherline Mode" solution. Nothing else makes sense to me but this troubling scenario. It takes 5mA of step pulse source current to operate the G540 and I'm guessing there are machines out there that cannot quite manage that.
Suggestions and a request:
1) Request: Is there someone out there that is (a) having a "Sherline Mode" moment and (b) has an oscilloscope? Can you please take a picture of the step pulse from the parallel port while it is connected to the G540 and post it?
2) Suggestion: For someone having a "Sherline Mode" moment, can you please get and try a PCI parallel port card? After-market cards have far better port pin drivers than the motherboard has. Can you please see if everything works then as it should with 2 microsecond Mach3 step pulses?
I really want to track this problem down and kill it once and for all. There is a free G540 in it for the first person whose help is instrumental in helping identify the cause and solution to this problem. It is about 3% of all G540 applications and it is irritating.
Mariss
Mariss: We have seen this "artifact" in some of the G251 units. In those designs we are using a buffered step & dir pulse (nice sharp step pulses that rule out any parallel port out-of-spec thoughts) into a HCLP2531 and a logic side power supply referenced to the Motor Neg. It also seems to manifest itself when you crank the pulse rate do to below about 200 PPS. In most cases the motor just stops turning at all. The turn-on waveform at the step pin of the G251 is very sharp (high to low transition) and goes to common. The trailing edge of the waveform is rounded but it mystifys me why the problem shows up in the slower pulse rates instead of faster. Moving the pulse rate to 5usec helped (moved the stall lower down the PPS scale) but not entirely. This is not a common occurrence and replacing the drive with another "fixes" the issue.
I have one 251 (maybe if Marti did not send it back already) I can put on the bench. I have just about any type of test equipment you need. Tektronix 200 MHZ DSO, Tektronix 350 mhz analog (My trusty 485); lab grade variable bench supplies, variable pulse width generators, function generators, non-contact infrared (real ones) temp monitors, computers running MACH3 and EMC2 etc, etc.
I find it helps to do the tests with real world signals from our powered BOB's as function generators and programmable pulse generators have "perfect" signals.
Knowing the input to the CPLD is 3V there may be something in the interface to the CPLD inputs that is at a tolerance point. (just an old engineers hunch).
I have a couple of units in the field that are on their way back to us. Another clue is that as the 251 unit heats up the problem becomes more pronounced. That is another indication that a component tolerance is involved. It never gets excessively hot (we have real time temp monitoring in our interface cards) In doing an infrared finger test (OUCH! test) the main filter cap on the misbehaving drive seemed warm(er).
I am seeing the same 3% number or less but before we knew to start looking for the telltale ultra low speed performance we had no way to test short of putting each 4 axis unit on a machine and run a long test for lost steps. Makes it a PITA to do volume production.
When the units come back I will see if I can profile a condition that will predict a possible problem unit. The waveform going in looks no different than for the other axi that works fine. I changed the HCLP2531 just in case (engineer's illogical hope for something simple) but the results were the same. I can try some signal condtioning AFTER the opto on our card but I predict it's not a signal input problem.
I would think with the same parallel port and same test computer and same breakout buffer chips (our same UBOB card design) as used with G203v's, and G320's with never a problem, we can look further down the line. My contention is it's something in the drive electronics and not the BOB interface but then it's always easier to point the finger at somebody else!
TOM Caudle
www.CandCNC.com
Tom,
That would be great. The CPLD is set to "Schmidt", "Float" and "3.3V LVTTL Standard" on all inputs. The step and direction input signal path is INPUT > 3K pullup to 3.3VDC > series 10K to CPLD > 100pF CPLD to GND. The 10K / 100pF provides a 1 microsecond noise filtering time constant. The CPLD clock is 2.5MHz, the step and direction inputs to D-Flops to insure fully synchronous logic. The boards are 4-layer with the inner layers dedicated to groundplane and power distribution.
One thing you might want to check: the Signal GND connects to drive GND via a 33 Ohm resistor. Its purpose is to limit ground-loop currents between multiple drives when opto-isolators aren't used. We occasionally see this resistor burned open when there has been a large voltage difference between power ground and signal ground. A quick test is to measure the resistance between these two grounds, a valid reading is 33 Ohms.
I would appreciate it if you were to identify a trouble maker like you described and return it for torture testing here.:-)
Mariss
I found that the G251 toublemaker was returned last week with a note attached. You might want to intercept it out of repair.
I don't have one to test in house yet. Should have one in Early next week
TOM Caudle
www.CandCNC.com
Tom,
I retrieved the drive and I'm running tests on it. So far it's looking like a "no problem found".
The drive temperature is 75C (It's mounted on a hot-plate whose temperature is regulated to 75C +/-1C). The supply voltage is 60VDC and the set current is 2A per phase.
TEK00003.gif has the step input on the yellow trace and the motor phase current probe on the green trace. Note the drive microsteps on the rising edge of the 100Hz step pulse.
TEK00002.gif shows a 100Hz triangular waveform step input. The drive microsteps on the rising slope of the waveform at the 2.1V level. The step input is a Schmidt trigger so a slow rise time is OK.
TEK00001.gif is the same as 00002.gif except the timebase shows 20 microsteps over a 200mS time span.
TEK00000.gif shows the drive step input frequency being sawtooth swept from 0Hz to 10 kHz, then abruptly back to 0Hz again. There is no loss of motor steps. The test motor is equipped with a 500-line encoder; the encoder feedback is differenced with the sent step pulses. The difference remains zero at all times (hundreds of sweep cycles later).
Next up: Narrow step pulse tests.
Mariss
Maris,
To simulate the problem i would suggest that you get Tjskcnc's XML config and his Gcode file and run the Mach on your bench.
But they are" Smooth sweeps cycles later"The difference remains zero at all times (hundreds of sweep cycles later).
I bet its a ragged pulse train that the drive's anti-resonace rejects or can't cope with.
Also Tjskcnc's, have you tried a different computer ? Since this is a random problem you may have something interfering with Mach3.
Larry K
Mariss: Are you putting the test signals straight into the step pin? In looking at the input circuit you use (an I use as well) the output side of the HLCP2531 looks into the 3.3V supply with the 3K being the pullup. Turn on time at the step spin (High to low transition) is sharp. turn off is sharp up to about 2/3rds then rounds off by several micro seconds. I would assume the Schmidt inputs would square that up.
I think the problem may be more in the step interface than the actual Gecko drive. Why only a small percentage of drives is the biggest part of the mystery
TOM Caudle
Larry,
The drive's resonance compensation cannot be fooled. It is a phase modulator meaning every step pulse entering the modulator must exit. The modulator can only affect the pulse phase delay as it transits through the circuit. You can even use a pseudo-random function generator as a step pulse source if you want.
I wouldn't know how to use any CNC program nor do I have a computer with a parallel port.:-)
Tom,
Couple of questions:
1) Do you have a 100nF MLCC as close as phsically possible to the HCPL-2531 pins 5 and 8? The 2531 will oscillate like a banshee in the low VHF frequency range if supply bypass isn't nearly perfect. The oscillation breakout has a positive temperature dependency.
2) The 2531 has very low gain (0.15 min spec). You should have 10mA of LED current to be safe. The collector load is 1mA.
3) The 2531 has an asymmetrical delay time (LED to collector 'on' time of 0.5uS, LED to collector 'off' time is considerably longer). Is your pulse polarity 3.3V on the 2531 collector while no pulses are being sent (at idle)?
Mariss
100 WHAT Farad? Oh, you mean a .1 <g>. Yes .1 mfd mono cer very close to each HCLP2531.
The inputs use 330 ohm current limit from the 5 volt rail which is about 11.5 ma.
Pulse polarity on the 2531 collector is low-true (high when no pulses present).
Remember that MOST 251's work great. Only a few have caused problems. Out of over a hundred units we have 3 so far. You have one.
TOM C
All this techno-speak is way over my head. I just want a system that works properly. Everything seems to be working OK with setting the pulse to 15. I did screen shots of my configurations and will post them a.s.a.p.
As soon as I can I will set up a different computer on my system to see if that makes a difference.
I apologize for the babble. It's two grizzled old EE's talking in shorthand. We just started using the forum to try and pin the problem down in the shortest time. Normally this sort of exchange is off line. If it's causing any problems we can take it to private e-mails.
TOM Caudle
www.CandCNC.com