Re: [Emc-users] Two questions.

2017-01-12 Thread Gene Heskett 
On Thursday 12 January 2017 00:27:15 Peter C. Wallace wrote:

> On Wed, 11 Jan 2017, Gene Heskett wrote:
> > Date: Wed, 11 Jan 2017 23:11:51 -0500
> > From: Gene Heskett 
> > Reply-To: "Enhanced Machine Controller (EMC)"
> > 
> > To: emc-users@lists.sourceforge.net
> > Subject: Re: [Emc-users] Two questions.
> >
> > On Wednesday 11 January 2017 22:37:46 Peter C. Wallace wrote:
> >> On Wed, 11 Jan 2017, Gene Heskett wrote:
> >>> Date: Wed, 11 Jan 2017 21:50:19 -0500
> >>> From: Gene Heskett 
> >>> Reply-To: "Enhanced Machine Controller (EMC)"
> >>> 
> >>> To: emc-users@lists.sourceforge.net
> >>> Subject: Re: [Emc-users] Two questions.
> >>>
> >>> On Wednesday 11 January 2017 13:34:50 Peter C. Wallace wrote:
>  On Wed, 11 Jan 2017, Jon Elson wrote:
> > Date: Wed, 11 Jan 2017 10:36:18 -0600
> > From: Jon Elson 
> > Reply-To: "Enhanced Machine Controller (EMC)"
> > 
> > To: "Enhanced Machine Controller (EMC)"
> >  Subject: Re: [Emc-users] Two
> > questions.
> >
> > On 01/10/2017 10:45 PM, Gene Heskett wrote:
> >> On Tuesday 10 January 2017 23:02:28 Jon Elson wrote:
> >>> On 01/10/2017 08:39 PM, Gene Heskett wrote:
>  But if I hook up the halscope, the apparent quadrature error
>  is worthless due to the 1KHz servo-loop timing limiting the
>  bandwidth.
> >>>
> >>> Well, that's why we use hardware encoder counters, as
> >>> sampling the quadrature at 1 KHz just isn't fast enough.
> >>>
> >>> Jon
> >>
> >> Plz elucidate Jon. Something has got to be better than this.
> >> See the pix I posted with the last reply to Peter & the list. 
> >> The top, white trace in the halscope screen is the quadrature
> >> noise, easily 10x the timing wibbles I see on the hitachi's
> >> screen.
> >
> > If you are looking at the quadrature signals at a 1 KHz
> > sampling rate, then unless the encoder is moving quite
> > slowly, the sampling will be too coarse.  That screen shot
> > is kind of small (you can take just a single window as the
> > screenshot, by the way) but it looks like the duty cycle of
> > the encoders is not 50%.  That may well be aliasing, as the
> > quadrature signals seem to be moving much too close to 1000
> > Hz.  You'd get much better results below 100 Hz.  Or, with
> > Mesa hardware, you can sample these with the base thread.
> >
> > I can't comment on the velocity output from the Mesa, as I
> > don't know their gear (or driver).  But, the scale is 2/div,
> > meaning the scaled velocity output is jumping around 2 user
> > units up and down at a KHz rate.  Clearly, not good.
> >
> > Jon
> 
>  The velocity output is calculated by using the number of counts
>  divided by the time between the counts (time measured via a 1 MHz
>  timestamp). If you have less than 2 counts per servo period the
>  velocity estimation accuracy depends heavily the quadrature
>  phase/symmetry since the edge spacing entirely determines the
>  estimated velocity at this speed range.
> 
>  I suspect what Genes plot shows is that he has significant
>  quadrature phase/symmetry errors at the FPGA inputs.
> 
> 
>  Again here is a velocity plot of a accurate 100 Hz quadrature
>  signal:
> 
>  http://freeby.mesanet.com/100_Hz_quadrature_velocity.png
> >>>
> >>> I've not played with the stepgen type 2 yet, but I have adjusted
> >>> the quadrature a few thousandths and reduced it by about 1/3rd.
> >>> Thats helpfull as I can filter to smooth a tach dial once I have
> >>> defeated the noise.
> >>>
> >>> But I have another concern. I have had to do a full powerdown
> >>> reset to clear a watchdog bite. The original config was I think
> >>> intended to clear it by re-enabling the machine, but I could
> >>> nolinuxcnct, so I disconnected
> >>
> >> that
> >>
> >>> reset in the hal file I started with as the logic was self
> >>> feeding, locking out the ability to reset the disabled machine. 
> >>> And I couldn't even do it from the halconfig as it claimed another
> >>> signal was driving it when I tried to setp that pin.
> >>>
> >>> Do you, Peter, have a logic diagram I can re-construct for that,
> >>> that actually works?
> >>
> >> You need to restart linuxcnc to clear a watchdog bite
> >>
>  Peter Wallace
>  Mesa Electronics
> >
> > That didn't work, I have since had to do full power downs with at
> > least 15 seconds off time. 5 or 6 times now.
>
> That probably means the  watchdog bite is not the issue but a symptom
> of some kind of system crash.

Which is a puzzle. Despite standing there looking at the "watchdog has 
bit" message, only the power led is lit.

I have it running

Re: [Emc-users] EDM machining, user interface, increments

2017-01-12 Thread Nicklas Karlsson 
On Wed, 11 Jan 2017 08:50:28 +0700
TJoseph Powderly  wrote:

> ...
> I understand the need for scaling the meters, sort of zoom to look 
> closer at the process

Yes and zoom on step length to fine tune. To scale values in widget is probably 
a very good idea since it is not real time and only need to be done once. As is 
now I added scaling to real time thread which require quite many rows and use 
real time execution then really not needed.

> in real machines, we had a switch on the ammeter to change the shunt, 
> and a switch on the voltmeter changing resistor divider
> an electro-mechanical solution so we could watch low voltages and low 
> current better

I think this is common and have often heard the noise of switches in 
measurement instruments as they switched range. If best possible measurement 
accuracy is needed this is the method to.

> we had no solution to watch midrange values closer

This is harder, an offset is needed.


I have seen a few improvements and sooner or later I should get enough time to 
look at them.

--
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today. http://sdm.link/xeonphi
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Two questions.

2017-01-12 Thread Nicklas Karlsson 
> On 2017-01-11 18:18, Nicklas Karlsson wrote:
> > Yes it is way to slow, hardware counters are cheap and work well, do you 
> > have them in you PI board?
> 
> These are hardware counters sitting in the Mesa card.

I could read that in a message further down in the thread but at that point my 
message was already posted.


M.A.S.H.
Disarming a bomb:
  ...
  Do this.
  But first do this.
  ¤£$]}$££@485##

--
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today. http://sdm.link/xeonphi
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users