Thanks all for the advice. I now have plenty to try. I may just replace them for something else though.
> On Aug 27, 2020, at 11:30 PM, Chris Albertson <albertson.ch...@gmail.com> > wrote: > > I have to admit most of my noise-hunting experience is with vacuum > tube-based amplifiers. But the exact same techniques apply. The first > step is to always apply the best-practice construction technique. Then if > that does not quiet things down you cut the system in half and see which > half still has the noise problem and then cut that part in half and so on. > > Using this we can make tube amps today that are better than anything > people could buy back in the tube era. All the advice about physical > separation, twisting all the wires and shields and star grounds really does > work. But it is a lot of work to do as you have to think about every inch > of every wire. > > Hey, it could be worse, be glad you were not the one who had to work out > the wiring for something like the photo below. Seriously they make up > these rules about 10X cable diameter spacing and such so stuff almost works > the first time even really hard cases like below. Same with my > Amplifiers. I apply every technique I know and they now work the first > time. Before it would take days to debug. > > [image: poza-China-roboti-1280x620.jpg] > > On Thu, Aug 27, 2020 at 6:25 PM Matthew Herd <herd.m...@gmail.com> wrote: > >> Hi Chris, >> >> The frequency of the noise as measured on the ground appears to be a >> ringing signal repeating at roughly 600kHz. I guess my previous >> measurement of 80kHz was wrong. I’m not sure why it would be different. >> I’m not 100% sure that it’s a consistent 600kHz, and might actually be >> several signals overlaid. The noise events appear to ring, as they decay >> from a larger sinusoidal spike, with about 6 sinusoidal pulses in total >> until they taper off. >> >> The wires are generally bundled, so that may be a factor. >> >> The housings on the motors, drivers, and supplies are grounded to the >> machine. Only the encoders and motor to drive cables are shielded, and >> they’re shielded on only one end. The remaining wires are simply primary >> cable and are not grounded. Only some of the signal wires are twisted >> pairs. >> >> Thanks! >> Matt >> >> >>> On Aug 27, 2020, at 1:24 PM, Chris Albertson <albertson.ch...@gmail.com> >> wrote: >>> >>> Can you measure the frequency of the noise? If this is a digital scope >> it >>> might be able to show a spectra. Lacking that you can eyeball-estimate >> the >>> period or width of any spikes. >>> >>> The other bit of information we don't have is how the wiring is laid out. >>> Best practice would be to use tightly twisted pairs where the current in >>> each of the two members of the pair is equal and also to never bundle >> power >>> and signal wires. The field around a twisted pair is close to zero >>> after about 10 cable-diameters so if a twisted power cable is 3mm thick, >>> keep other wires 30mm away. >>> >>> Are all the housing on the motors, drivers, and supplies grounded. This >>> would be to a protective ground Also are the wires shielded and grounded >>> on one and only one end? >>> >>> If the noise is determined to be a relative high frequency then ferrite >>> cores on the wires help but they no nothing for 120 Hz. >>> >>> >>> >>> On Thu, Aug 27, 2020 at 10:00 AM Matthew Herd <herd.m...@gmail.com> >> wrote: >>> >>>> I spent some time the week before last investigating the noise issue >> that >>>> I’ve been having on my Bridgeport BOSS mill. The system architecture >>>> involves a single phase 240V input, which powers a VFD for the 2HP >> spindle >>>> motor, a 120V step down transformer to power some devices on the machine >>>> (spray mister, accessory lights, cabinet ventilation fans), and a >>>> 72V/24V/12V DC power supply to power the USC board and the three Gecko >> 203V >>>> stepper drivers. First, I disconnected every device from the DC power >>>> supply by disconnecting the positive terminal and disconnected one of >> the >>>> 240V AC wires to the VFD. I measured the noise from the ground bolt to >> the >>>> ground pin of the 12V output of the power supply. Noise was nominal, >>>> perhaps 20-40mV peak to peak at 50ms/div (as I recall). Reconnecting >> the >>>> VFD gave approximately 600mV of noise peak to peak. This seemed rather >>>> surprising given the fact that I’m essentially measuring the same wire >>>> across about 18". >>>> >>>> I installed a Rasmi VFD filter and I re-wired the grounds. Instead of 2 >>>> grounds which were connected via the machine, I ran all grounds back to >> the >>>> ground bolt in the power cabinet. Previously, the 72V ground to the >> drives >>>> was only wired directly from the power supply to the drives. It was >>>> isolated from the other power and logic grounds, but is now common. The >>>> only ground that was not run to the power cabinet ground was the >> stepper’s >>>> step and direction signal ground. This ground is connected at the USC >>>> board to a ground which is common with the 12V power input ground that >>>> powers the USC board (and is therefore run back to the main ground >> bolt). >>>> Upon powering up everything I discovered that the spindle index encoder >>>> noise had increased from around 1.4V peak to peak to more like 7V peak >> to >>>> peak. I’m shocked that a ground wiring change could impact this >>>> measurement so dramatically. The VFD only noise is now measuring >>>> approximately 400mV p-p at the power supply 12V ground, which I believe >> is >>>> acceptable. However, I’m sure I could find higher spikes if I reduced >> the >>>> time scale based on my previous test results. I got some conflicting >>>> results when trying to isolate the cause, so I rechecked my measurements >>>> today. Rechecking my measurements at the 12V output ground on the power >>>> supply, I found: >>>> >>>> VFD only: >>>> ~300mV >>>> >>>> USC only: >>>> ~40mV >>>> >>>> USC and 24V devices (mostly SSRs for pneumatic valves): >>>> ~40-50mV >>>> >>>> USC, 24V devices, and Geckos with NO 72V input (i.e. only step & >> direction >>>> inputs) >>>> ~40-50mV >>>> >>>> USC, 24V devices, and Geckos with 72V input (i.e. ready to move motors) >>>> ~1-1.2V >>>> >>>> Now I tested the 72V wire at the gecko’s power input and found noise of >>>> approximately 5V. >>>> >>>> Then I removed the wire at one of the two 72V supply outputs (it has >> dual >>>> 72V outputs) and measured at the + terminal with two other geckos >> powered >>>> from the other 72V output. Noise was minimal (though difficult to >> measure >>>> as my scope wouldn’t let me scale to see the ripple as clearly — at most >>>> it’s probably 1-2V). This leads me to the conclusion that either the >>>> supply somehow struggles under load or the Geckos are introducing >>>> inordinate noise. I’m doubtful the supply is the culprit because the >> noise >>>> appears to be a much higher frequency than would be expected from a >> simple >>>> rectified power supply with filtering capacitors. The geckos are in the >>>> control cabinet with the USC. Would it help to move them closer? >>>> >>>> Thanks, >>>> Matt >>>> >>>> _______________________________________________ >>>> Emc-users mailing list >>>> Emc-users@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/emc-users >>>> >>> >>> >>> -- >>> >>> Chris Albertson >>> Redondo Beach, California >>> >>> _______________________________________________ >>> Emc-users mailing list >>> Emc-users@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/emc-users >> >> >> >> _______________________________________________ >> Emc-users mailing list >> Emc-users@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/emc-users >> > > > -- > > Chris Albertson > Redondo Beach, California > <poza-China-roboti-1280x620.jpg>_______________________________________________ > Emc-users mailing list > Emc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-users _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
