Re: First time EMC user (introduction)

[Ray Henry <rehenry-at-up.net>]

Richard

I guessed it but I was thinking even worse -- like the 125000 ppr's that 
I ran into the other day.  Others have explained the fix.  You are 
fortunate that the Gecko pulse multiplier will get you in range.  Until 
then, you can run with the large values in the ini by reducing the 
velocity component a lot and watch the machine crawl.

The PC will behave this way (crap out) when the product of max velocity 
and input scale is way over the top.  Freqmod just loads and then 
everything gets very slow.  Steppermod kills the PC and everything near 
it so that you have to reboot.  If you get some unexpected behaviors 
after a crash like this, you may have to reload a package or two.

Ray  

On Monday 30 September 2002 11:09 pm, you wrote:
> Hi Ray,
>
> I responded to Jon with a rather long post, so I will not repeat all of
> that here.  Thanks for the tips.  You were right, although I bet you
> didn't realize my encoders are 10,000 counts/rev in quadrature!!!!  I
> need 50,000 steps to go one inch :-(
>
> Time to rethink my electronics, huh?
>
> Richard
>
>
> ----- Original Message -----
> From: "Ray Henry" <rehenry-at-up.net>
> To: "Multiple recipients of list" <emc-at-nist.gov>
> Sent: Monday, September 30, 2002 11:26 AM
> Subject: Re: First time EMC user (introduction)
>
> > Hi Richard.  Welcome to the "Chips" (our mascot) fan club.  You may
> > have seen him a few times during the install.  The micrometer is so
> > that he can prove just how accurate the EMC can be.  The dead-blow
> > hammer is to make the part fit!
> >
> > Steppermod is a binning procedure and may well produce periods of
> > inactivity or reduced activity that will be obvious with a scope.  We
> > generally speak of jitter when we consider how well it puts out
> > pulses. You might try freqmod.  There you will find an attempt to
> > generate frequencies rather than pulse batches.  Both of these are
> > limited in different ways by the ability to generate pulse trains.
> >
> > The 350MHz PC should run freqmod with a period around 0.000035.  You
> > will not get a real high speed pulse rate from it.  Better with
> > steppermod than freqmod.  Better yet with a servo card. <hint>
> >
> > The machine sounds great.  There are a few commercial machines that
> > have been retrofitted and a couple others who are working on
> > integration of the EMC to this level of tool.  Most of these use or
> > plan to use a full servo feedback system.
> >
> > Two things worry me just a bit about the Gecko 320 in this big
> > machine application.  The first is that you will sacrefice a lot of
> > speed for two reasons.  The first is that you will not get full
> > current for the motors and that will reduce acceleration a bunch. The
> > second speed loss will depend upon the number of pulses per unit of
> > the existing encoders.  The Gecko pulse multiplier board might help
> > this quite a bit.  You can think in terms of 5k to 10k pps for your
> > computer so you need to see how many pulses it will take per unit of
> > travel and then you will know the top end for rapids.
> >
> > The second concern is with open loop operation.  You should not loose
> > pulses with the step and direction Gecko, at least not like you would
> > with steppers but there is still something that bothers me about a
> > computer making assumptions about where an axis might be on this big
> > of a machine.  The Gecko will fault after 128 pulses sent to it are
> > unanswered by the encoder but that can be quite a distance compared
> > to normal following error settings for a servo system.
> >
> > It sounds like the encoders feed the drive amps and they compute
> > velocity from the pulse train.  If this is the way the machine ran,
> > those signals must also have been fed into the control for
> > positioning.  You might well be able to make a single encoder do
> > double duty with a PPMC (Jon Elson) STG,  Studio Ferraris, Vigilant,
> > or whatever servo card(s) as well.  Or perhaps you could add a second
> > encoder for positioning.
> >
> > You might look at tkio and iosh for a possible path to produce logic
> > and work io pins for your tool changer.  There has been some talk
> > about writing an interface that would allow us to use classicladder
> > and MAT but it isn't there yet.
> >
> > M2CW  Hope this helps
> >
> > Ray
> >
> > On Monday 30 September 2002 11:14 am, you wrote:
> > > Hi everyone,
> > >
> > > I have just installed EMC and linux using the BDI cdrom image. 
> > > After figuring out that I had to log in as root, and after setting
> > > the emc.ini file to steppermod and minimilltask (it would not come
> > > out of estop with bridgeporttask) I managed to get step signals on
> > > pin 3 (X axis) of the parallel port.  I monitored these with my
> > > digital oscope, and saw an interesting thing.  I selected manual
> > > control, and then set the x axis to move 1 inch.  My ini file had
> > > the max acceleration at 2, and I captured the pulse train on my
> > > scope.  It looked about right I guess, with the pulse frequency
> > > accelerating to 2 khz.  The strange thing was that after the pulse
> > > train had fully accelerated, there would be a 1 khz pulse about
> > > every 8 or so of the 2khz pulses.  I don't know if this is some
> > > latency issue, or perhaps that is when some other task is running?
> > >
> > >
> > > I am running on an AMD K6-2 350mhz.  I am going to be using this to
> > > control a rather large Shizuoka B-3V bed mill, with Fanuc model 0
> > > brush DC servo motors (60V 12A) on the x and y axis, and a Fanuc
> > > model 5 bruch DC servo motor (90V 12A) on the z axis.  I am not
> > > 100% positive, but I believe there is not a tach on these motors,
> > > just a quadrature encoder (called a pulse coder by fanuc).  I
> > > already have two Gecko G320 servo drives (take step,dir signals)
> > > which should work for the x and y, so I am going to start with
> > > those.  I would get the servo to go card, except that I am not sure
> > > if I would be able to use analog input servo amps without a tach on
> > > my motors, so why not just get one or two more parallel ports?
> > >
> > > Anyway, I want to really dig into this.  I like the idea of an open
> > > source machine controller.  I discovered the (rather large) message
> > > archives, and I will try to answer most of my questions there so as
> > > not to clog the message board with redundant newbie stuff.  When I
> > > get a bit more experience, I would like to add to the EMC knowledge
> > > base, as I am going to need to do a few special things like
> > > controlling my 20 position ATC on the cnc mill.
> > >
> > > Back to the scope,
> > >
> > > Richard