On Tuesday 22 June 2021 02:06:45 Chris Albertson wrote: > Seriously you can't get away from it. With 40 million lines if the > shaft turns at 60 RPM or one revolution per second it is moving slow > but the total number of edges is 16 million per second and you could > find hardware fast enough but at 3,000 RPM it would be a lot harder. > > The entire reason I brought up the subject is because people started > posting about shielding and star grounds. That is not the kind of > noise we are talking about. Quantization happens even in an > electrically quiet environment and results from design decisions made > such as the sample interval, number of lines of the sensor, and the > rotation speed. It gets worse as you go slower. > Chris is spot on correct. Quantization noise is not electrical noise and never was. Its the combined effect of very few edges passing in a given sample period, phasing errors between the two signals, etc. If 3 edges pass in a sample period, and 4 in the next period, thats a huge error the pid will try to correct. The average speed will be maintained if the pid is properly tuned, but the motor itself may wind up getting full supply power for a millisecond, and virtually stopped by a 4 quadrant controller in the next millisecond, leading to noises from any gears with lash being violently slammed from one side of the lash to the oher. Been there, done that, sounded like the ball bearings in every bearing in the head have suddenly had their round balls replaced with cubes.
I installed a very hi resolution encoder. Those same bearings I thought had square balls, cannot now be heard at 500 revs, and the Pgain went from 2.5 to 2500 before any sign of instability. My encoder SCALE went from 268 in high gear to a bit over 7000, and something over 14,000 in low gear. If I was to build a tool changer that needed the spindle oriented just so, I can do it, probably without a stop pawl as the servo could hold it while an impact wrench runs the drawbolt. > On Mon, Jun 21, 2021 at 1:15 PM Thaddeus Waldner <thadw...@gmail.com> wrote: > > Okuma machine builders really don’t like quantization noise so they > > use 40 million CPR encoders on the axis motors. > > > > > On Jun 21, 2021, at 11:34 AM, Chris Albertson > > > <albertson.ch...@gmail.com> > > > > wrote: > > > Quantization noise is fundamental and present in all digital > > > systems. There is absolutely no way to avoid it. All you can do > > > is minimize it. > > > > > > To know how fast a shaft is spinning or how far it moves, you have > > > to > > > > count > > > > > edges and if you get 100 counts you just can not know if you are > > > closer > > > > to > > > > > 100 or to 101 counts. Basically the counter always rounds down > > > even if > > > > at > > > > > 100.99999 it tells you "100". So the cound is "wrong" 50% of > > > the time. This applies not only to encoders, but A/D converts and > > > every other kind > > > > of > > > > > measurement. > > > > > > This is the nature of digital data, of representing the world > > > with a > > > > fixed > > > > > number of bits. > > > > > >> On Sat, Jun 19, 2021 at 4:00 PM John Dammeyer > > >> <jo...@autoartisans.com> wrote: > > >> > > >> But doesn't the 7i92 FPGA deal with quadrature encoders and > > >> therefore doesn't really need to deal with a sampling process but > > >> instead looks at edges? And because of the way quadrature works, > > >> with the A/B phasing > > > > you > > > > >> don't get the same types of errors compared to polling a bit > > >> level X > > > > times > > > > >> per second and trying to decide when it's high/low. > > >> > > >> John > > >> > > >>> -----Original Message----- > > >>> From: Chris Albertson [mailto:albertson.ch...@gmail.com] > > >>> Sent: June-19-21 3:21 PM > > >>> To: Enhanced Machine Controller (EMC) > > >>> Subject: Re: [Emc-users] Machining question > > >>> > > >>> There are two kinds of noise, > > >>> 1) electrical noise superimposed on the signal. > > >>> 2) quantization noise from the sampling process. What happens > > >>> here is > > >> > > >> that > > >> > > >>> the computer counts the number of line crossings during a given > > >>> time. > > > > it > > > > >>> is just random luck when the time starts so on average there is > > >>> a plus > > > > or > > > > >>> minus one count error. > > >>> > > >>> So a one count error sounds small but lets say we have a 600 > > >>> count encoder and we are running at 500 RPM and sampling 100 > > >>> times per > > > > second. > > > > >>> This works out to 50 counts per period. a one count error is a > > >>> 2% > > >> > > >> error. > > >> > > >>> You will see a larger percent error at slower speeds or with a > > >>> lower resolution encoder. > > >>> > > >>> So even with perfect wires and perfect grounding and zero > > >>> electrical > > >> > > >> noise > > >> > > >>> the RPM speed is going to jump around randomly over a 4 RPM > > >>> range > > > > because > > > > >>> of unavoidable quantization error. > > >>> > > >>> The solutions either the use a larger sample time apply a > > >>> low-pass > > >> > > >> DIGITAL > > >> > > >>> filter to the signal. No amount of analog filters on the wires > > >>> will > > >> > > >> work. > > >> > > >>> On Sat, Jun 19, 2021 at 2:55 PM Gene Heskett > > >>> <ghesk...@shentel.net> > > >> > > >> wrote: > > >>>> On Saturday 19 June 2021 12:59:19 John Dammeyer wrote: > > >>>>> I need to do a couple of things. For one the AC Servo makes a > > >>>>> lot of electrical noise. The frame of the motor is connected > > >>>>> to earth through power line ground. But my bench setup has > > >>>>> the control side > > >> > > >> 5V > > >> > > >>>>> not isolated from the 'PC' side (optos are kind of useless > > >>>>> here) and although the Pi4 doesn't appear to have any trouble > > >>>>> the scope shows a pretty noisy encoder signal. > > >>>> > > >>>> You are about to learn the star ground system I think. > > >>>> > > >>>> Thats where all grounds go to a single bolt, and no other > > >>>> grounds are allowed anyplace. > > >>>> > > >>>> Shielding in motor cables ends without touching the motor, only > > >> > > >> connected > > >> > > >>>> to this same single bolt which is grounded. The common line of > > >>>> any > > >> > > >> power > > >> > > >>>> supply is connected only to this bolt, and the common grounds > > >>>> to every piece of pcb in the system comes from that bolt. The > > >>>> used to be green static ground wire in any power cord is > > >>>> rerouted to this bolt. And static grounds on a power supply are > > >>>> fed from this single bolt. > > > > Because > > > > >>>> the psu case is usually bolted down wherever its at, you may > > >>>> have to uncover the supply and remove any connection from the > > >>>> earth labeled terminal, and the psu's case. It may be a screw > > >>>> in the corner of a > > > > pcb, > > > > >>>> but it needs to be removed. > > >>>> > > >>>> Anything connected to ground at some other point in addition to > > >>>> that > > >> > > >> bolt > > >> > > >>>> becomes an antenna, is called a ground loop, picking up both > > >>>> magnetic and electrostatic noises. I had to learn that the hard > > >>>> way while building the Sheldon, blowing two 7i90HD boards out > > >>>> by picking up 80 volts p-p of switching noise that reached up > > >>>> to the bandwidth limits > > > > of > > > > >>>> both of my 100mhz scopes. Hell on an fpga with a 3 volt limit. > > >>>> I > > >> > > >> killed > > >> > > >>>> the power and redid it to the single bolt model, and that 80 > > >>>> volts of noise was reduced to under 100 millivolts. Took me > > >>>> about a week to > > > > find > > > > >>>> all the connections that should not have been made. Even a > > >>>> small > > > > bypass > > > > >>>> capacitor to a local ground instead of back to that bolt can > > >>>> become a lethal weapon to the elecronics. Its a noise injector > > >>>> in that case. > > >>>> > > >>>> The whole idea is to make the ground, if it bounces from a > > >>>> lightning strike on the can on the pole that powers your place, > > >>>> which may make > > >> > > >> the > > >> > > >>>> whole system bounce 100+ kilovolts, and if you are in that > > >>>> circuit at that instant it WILL get your attention. BUT, > > >>>> everything tied to that bolt should bounce in unison, meaning > > >>>> an individual board input will only see the difference signal > > >>>> its being fed with, and it should > > >> > > >> survive > > >> > > >>>> that lightning strike with no damage. > > >>>> > > >>>> My soap box as a CET just collapsed, so I'll do an Andy Capp > > >>>> and > > >> > > >> shudup. > > >> > > >>>>> ===================== HAL code =================== > > >>>>> Second is some of the HAL parameters and reducing to 0.003 > > >>>>> makes the display more stable. # Use ACTUAL spindle velocity > > >>>>> from spindle encoder > > >>>>> # spindle-velocity bounces around so we filter it with > > >>>>> lowpass # spindle-velocity is signed so we use absolute > > >>>>> component to remove sign # ACTUAL velocity is in RPS not RPM > > >>>>> so we scale it. > > >>>>> > > >>>>> setp scale.spindle.gain 60 > > >>>>> setp lowpass.spindle.gain 0.003 > > >>>>> net spindle-vel-fb-rps => scale.spindle.in > > >>>>> net spindle-fb-rpm scale.spindle.out => > > >>>>> abs.spindle.in net spindle-fb-rpm-abs > > >>>>> abs.spindle.out => lowpass.spindle.in net > > >>>>> spindle-fb-rpm-abs-filtered lowpass.spindle.out > > >>>>> ===================== end HAL code =================== > > >>>>> > > >>>>> But I'm still using the original signal for at speed and the > > >>>>> LED > > >> > > >> never > > >> > > >>>>> comes on nor does it know it's at speed. I should probably be > > >>>>> filtering the RPS rather than RPM I think? > > >>>>> > > >>>>> ===================== HAL code =================== > > >>>>> net spindle-vel-fb-rps => spindle-near-speed.in1 > > >>>>> net spindle-ramped => spindle-near-speed.in2 > > >>>>> > > >>>>> # ---Setup spindle at speed signals--- > > >>>>> # the output from spindle-near-speed 'near' component is sent > > >>>>> to spindle.0.at-speed # and when this is true motion will > > >>>>> start. The LED in the pyvcp-panel.xml is renamed to be a LED. > > >>>>> net > > >>>>> spindle-near-speed-led <= spindle-near-speed.out => > > >> > > >> spindle.0.at-speed > > >> > > >>>>> ===================== end HAL code =================== > > >>>>> > > >>>>> John > > >>>>> > > >>>>>> -----Original Message----- > > >>>>>> From: andy pugh [mailto:bodge...@gmail.com] > > >>>>>> Sent: June-19-21 2:48 AM > > >>>>>> To: Enhanced Machine Controller (EMC) > > >>>>>> Subject: Re: [Emc-users] Machining question > > >>>>>> > > >>>>>> On Sat, 19 Jun 2021 at 07:11, John Dammeyer < > > >> > > >> jo...@autoartisans.com> > > >> > > >>>> wrote: > > >>>>>>> I'm finding the AXIS Spindle speed oscillates about +/- 5 > > >>>>>>> RPM. > > >> > > >> I > > >> > > >>>>>>> thought that I was filtering it. > > >>>>>> > > >>>>>> It generally needs to be filtered, but it is possible you are > > >>>>>> filtering it, but not enough to get a steady reading. > > >>>>>> > > >>>>>> For whatever reason encoder velocity does always seem to be a > > >>>>>> bit noisy, despite the counters using timetamped edges and > > >>>>>> all the > > >> > > >> other > > >> > > >>>>>> tricks that might help. > > >>>>>> > > >>>>>> -- > > >>>>>> atp > > >>>>>> "A motorcycle is a bicycle with a pandemonium attachment and > > >>>>>> is designed for the especial use of mechanical geniuses, > > >>>>>> daredevils > > >> > > >> and > > >> > > >>>>>> lunatics." > > >>>>>> ? George Fitch, Atlanta Constitution Newspaper, 1912 > > >>>>>> > > >>>>>> > > >>>>>> _______________________________________________ > > >>>>>> 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 > > >>>> > > >>>> 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) > > >>>> If we desire respect for the law, we must first make the law > > >> > > >> respectable. > > >> > > >>>> - Louis D. Brandeis > > >>>> 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 > > >>> > > >>> -- > > >>> > > >>> 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 > > > > > > _______________________________________________ > > > 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 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) If we desire respect for the law, we must first make the law respectable. - Louis D. Brandeis 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
