On Sunday 02 June 2019 08:16:24 pm John Dammeyer wrote: > Hey Gene, > > > If you've got noises bad enough to mess with the encoder, I'll bet a > > bottle of suds your grounding system violates the single point > > "star" rule. You want shielded cables yes, but all shields and all > > ground commons come back to a single long bolt, and none of it is > > grounded any place else. Any other ground, including the 3rd static > > ground in the power cords is a ground loop and will intro noise. > > Only one master connection to the reference ground should be > > allowed. > > You haven't read the CUI data sheet then. My mouth dropped when I > read the specs and looked at the cable they make for this encoder. > It's shielded 14 or 18 conductor depending on the model. At the > encoder end there's a teensy little connector along with a loose wire > with a ring terminal The data sheet clearly says this ring terminal > is connected to the 5V ground and should be fastened to the motor > frame. (which ultimately leads back of course to your big ground > bolt). > And creates a huge ground loop.
The shielding on the motors drive cables should quit before it gets to the motor and the other end should be tied to ground at said bolt. The motor is bolted solidly to the machine frame in 99.99% of the cases unless you made the adaptors on a 3d printer. You should have as a motor ground, one 10 gauge or so wire from the machines frame back to that single bolt. The shielding of the encoders cable should be fixed to this bolt. I am assuming the encoder itself is an ABX, and that in this case it has +A and -A outputs, ditto for B and X. I have such an encoder, but its an Omron with those 6 wires plus a 5 volt supply and a supply return, which also goes to THAT bolt. Thats 8 wires and the shield. This encoder is driven by a rubber coupler, from a brass extension I drilled into the rear of the G0704's 1 HP motor, and in effect made a rear shaft for a motor that doesn't have one. Its all grounded to the rear of the motor housing by virtue of 4 brass standoffs into the motor cover. The shielding may go into the encoder, but I found no continuity from its cable shield to any metallic part of it. In my case I am not using the X outputs as that is still coming from the original optical interrupter encoder I made and put on the spindle, so the scale value now changes with the heads gear position, which I have tally switches on which go true only it the knob is within about 2 degrees of home in either position. Scale changes are done in the hal file. You should need to fool with that after the initial calibration. Now, since I wasn't aware this encoder was differential output when I bought it, and I frankly don't think the vendor knew the difference either as it wasn't in the specs he listed on fleabay. It didn't have near enough swing to feed a 5i25 encoders single ended inputs directly. So I had a dilemma to solve. So I now have a pair of $2/copy on ebay, rs485 to ttl boards, preset to disable its duplex stuff, in a small Hammond die cast box in the middle of the cable, powered by the 5 volts and ground going by towards the encoder with the rs485 green screw socket connected to the + & - of of each channel, and the ttl pins on the other end of this teeny little card are routed on to the breakout and hence to the 5i25's encoder inputs. The optical X is already a cmos rail to rail switch so it goes straight to the 5i25's x input, by way of the breakout. You'd need 3 of those little boards because you'll need to do the same for your X. I found them so handy I bought another 10 pack for spares or ??? ====== The breakout was sainsmart, but is now a 7i76's only encoder input, works exactly the same once I'd fixed the routing in the halfile. Noise has never been a problem, and the rs485 receivers have worked flawlessly. And since I was gain switching the scaling with the tally switch status, I took advantage of the neutral position when neither switch was closed, to feed about a 25 rpm signal to the motor driver. So I can have it running at 3 grand in high gear, reach up and grab the gear shift knob, turn it as fast as I want/can and its down to 25 rpm in a 200 milliseconds, long before the gears have disengaged then as the knob continues to turn, the other gear engages while its turning too slow to damage anything, and when the final switch closes it returns to 1500 rpms. ditto going the other way too. > The other end of this 36" cable has 6" unshielded and a 3 length > connected to the shield. That one of course should go to your big > bolt. > > But that they want me to connect 3 different motors 5V Gnd at > dramatically different parts of the frame to the frame boggles my > mind. Mine too John. > The reason I'm trying a replacement encoder is because Henrik Olsson > had nothing but troubles with his HP_UHU servo drives and the US > Digital encoders. He was fortunate enough to have several other > brands handy and had absolutely no overrun errors ever. Even with the > cables wrapped around the motors a number of times. But put the US > Digital one back in and constant overrun errors. Mine aren't as bad as > his but then mine are a slightly different model. So although I've > got a clean system and signals that look perfect on the scope (like > his did) I thought it was time to at least verify that the encoders > weren't the problem. > > I even went as far as buying braided shielded twisted pair 16g motor > cable. Not cheap. So if the new encoders work then we know it's > something with the US Digitals. And I am using differential pairs for > A and B. The receiver is a 75115. I've even wired in termination at > the encoder end so the cable is a properly balanced transmission line. > > So at the moment we're thinking that the US Digital units are > sensitive to magnetic interference from the acceleration or > deceleration heavy current. (105VDC power supply and a motor rated to > momentarily draw 40A but steady current at 9A). Here its a 90 volt 9.7 amp motor, getting banged by 126 volts dc and current limited to around 17 amps peak. Reversal time at 3 grand in high gear is about 400 milliseconds accompanied by s short squeek from the motor as it vocalizes on the current limit in a pico pwm-servo amplifier. > And it doesn't matter whether the PC is running LinuxCNC or MACH3. > Henrik designed and sells a replacement CPU module that plugs in place > of Uli's UHU Atmel processor. It's dsPIC based and has the hardware > quadrature support. So he doesn't use interrupts on encoder edges and > perhaps might not have the problems. But I agree with him that if I > can get overrun errors from the Atmel UHU processor then switching to > something else won't solve the problem. Might only just hide it. > > If I use the stepper motor power supply (60VDC) I can mount my two > Gecko Servo Drives in place of the HP_UHU but then I'd never get full > speed from the motors and I still wouldn't know if I was having > problems until I would perhaps see a home not agree with my Shumatech > DRO. And that equals wrecked parts. Thats NOT what you came here to do I don't think. > And be aware this mill conversion project with the HP_UHUs started > back in 2008. And just gets postponed by a year or so every once in a > while. The mill with the scales and DRO has been adequate for most > precise operations that I need to do. One of My Electronic Lead Screw > controllers driving the GECKO 950oz/in knee motor through a 3:1 > toothed belt has made raising the knee simple. Another ELS on my > rotary table as simplified much of that work. And all that logic is already part of LinuxCNC. I don't know how you are driving Z now, but theres no reason you can't divide that up in hal and drive both the knee and the quill in opposite directions to become the total Z to linuxcnc. There is very very little in the way of limits to impede that. If I had a knee mill, I likely would have that figured out already. > At the moment I don't really 'need' CNC on the mill. I just want it. BTDT, bought the t-shirt and wore it out already. ;-) > I think I posted a youtube link before with a BeagleBone Black, > Xylotex Cape and MachineKit exercising the knee and the Y axis. The > motor mounts are all castings. This link is with MACH3 running the > roadrunner.tap GCode sample. https://youtu.be/OUQzyp-3WXY > > That's when I discovered the move to home after it was done was out by > a few thou. So for now it's been two steps backwards and one step > forward. You can see the US Digital encoder on the X motor. Nice > totally shielded aluminium cover. The CUI has more plastic bits. > I'll let you know how it works out. > > John > > > > > > > _______________________________________________ > Emc-users mailing list > Emc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-users Cheers, Gene Heskett -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene> _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users