Tom,

All default axis settings are hard-coded into my init.c, which was exactly the same for both runs. I also made all axis settings in both versions of KMotionCNC exactly the same. Step width is the default 2uSec, step polarity low, just as it's always been (except for that brief diversion on step polarity).

I don't see what setting could possibly cause such a perfectly repeatable loss of position, and such identically non-orthogonal, and non-straight, lines. The problem really can't be lost steps in the traditional sense, since all the paths cut are *exactly* the same dimensions - just crooked. Each path is shifted down in Y by precisely the same amount, almost as though a coordinate rotation was applied. I've never seen any kind of noise problem be that repeatable. That, combined with the self-modifying offsets convinces me there's something very wrong somewhere in there. It's also, several times, gotten into a state where the interpreter simply stops working (a single MDI line will cause the interpreter to just hang, and never give a complete callback), and cycling power seems to be the only viable recovery.

Last week, after getting the step polarity set back right, I was able to make 25 copies of a more complex part with v428, and, other than the jogging problems, everything worked perfectly.

Regards,
Ray L.

--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Ray,
>  
> It seems you missed my point.  I wasn't trying to make the point that your system suddenly has noise.  I'm thinking that one of your settings is wrong to make the system behave strange.  Something like the Step Pulse duration, or open collector drive vs TTL drive, or something similar.  That is why I specifically asked for that test.
>  
> Regards
> TK
>
> From: himykabibble <jagboy@...>
> To: DynoMotion@yahoogroups.com
> Sent: Wednesday, February 22, 2012 6:19 PM
> Subject: [DynoMotion] Re: Has My KFlop Died?
>
>
>  
> Tom,
>
> I've run extensive tests jogging, and running G-code, that beats the hell out of all the axes for as much as an hour, then returns to zero. Never seen an error there outside the mechanical repeatability of the machine (about 0.001"), other than when the step polarity was wrong. If it was noise, why have I never seen anything remotely like this before? Why does it work perfectly under v428, and work horribly under v429z? Nothing but the KMotion software is different between the two runs. I just don't see noise explaining that. Also, I would expect noise to be largely random. This is absolutely, 100% repeatable - it screws up in *exactly* the same way, and the error magnitude is exactly the same, every time. That sure doesn't smell like noise to me. I was also using the exact same DSP code in both cases.
>
> I have also, with both my app and KMotionCNC, seen several sudden, HUGE, un-commanded, changes in the DRO values. ALL axes suddenly jump to seemingly random values. It hasn't happened enough times to get a clue what is triggering it, but I've seen it about a half dozen times now, in just a few hours of working with it. It appears all of the fixture offsets are suddenly getting clobbered.
>
> Regards,
> Ray L.
>
> --- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@> wrote:
> >
> > Hi Ray,
> >  
> > Sounds like you are gitting some sort of "drift" where noise steps on an axis that is supposed to be stopped is being told to move a few steps now and then.
> >  
> > Did you double check that you moved all your settings over to the new version properly?
> >  
> > How about simple tests like Jog X back and forth and watch the Y Axis Dest and the Physical Y motor?
> >  
> > Regards
> > TK
> >
> >
> > ________________________________
> > From: himykabibble <jagboy@>
> > To: DynoMotion@yahoogroups.com
> > Sent: Wednesday, February 22, 2012 5:17 PM
> > Subject: [DynoMotion] Re: Has My KFlop Died?
> >
> >
> >  
> > Tom,
> >
> > The good news is, I discovered the Flash recovery after making my post, and resuscitated the board. The bad news is v429z is VERY sick! I went back to v428, and ran the exact same program, and it ran perfectly. I then tried it again with v429z, and it was screwed up in precisely the same manner, and *exactly* retraced the paths from the first run. Both runs were using KMotionCNC, not my app. So, there is a real, serious problem there. It almost looks like each of the parts was individually rotated a few dregrees clockwise, AND offset about 0.070" to Y-, as the first part was about 1/4" from the front edge of the stock, while the 5th or 6th part was nearly to the edge of the stock.
> >
> > By "not straight" I mean the paths are both angled, AND not straight lines. They have steps and curves in them. After running part of the program, I halted execution, and did a G0X0Y0, and every time, running v429z, the Y axis is offset about -0.350". I see no indication of servo faults. Even stranger, watching the DROS when it's supposed to be moving straight, the DROs *look* correct (either X or Y changing, not both) yet it is NOT cutting a straight line!
> >
> > Regards,
> > Ray L.
> >
> > --- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@> wrote:
> > >
> > > Hi Ray,
> > >  
> > > Very strange.  Can't think what would cause that.  When you say they are not straight - do you mean they are straight but angled?  There is a geometric correction file capability that could potential do such a thing, but I assume you don't have anything selected for that file.
> > >  
> > > Regarding the board you can probably do a Flash Recovery.  Turn off the board.  Run the V4.28 KMotion.exe.  On the Config/Flash screen push "Recovery"  Follow the instuctions.  If it succeeds in loading/running the firmware then Flash a New Version.
> > >  
> > > Regards
> > > TK
> > >
> > >
> > > ________________________________
> > > From: himykabibble <jagboy@>
> > > To: DynoMotion@yahoogroups.com
> > > Sent: Wednesday, February 22, 2012 4:31 PM
> > > Subject: [DynoMotion] Has My KFlop Died?
> > >
> > >
> > >  
> > > Today I tried once again to make some parts, this time using v429z and it's doing the most bizarre thing I've ever seen. The parts are small, about 2" x 4", more or less rectangular, 10 pieces on each blank. However, the rectangular parts are coming out all identical, but all wrong - they are basically coming out trapezoidal, and even the supposed-to-be straight sides, which should be exactly parallel to the X and Y axes, are NOT straight!
> > >
> > > Since absolutely nothing on the machine has changed since the last time I used the machine a few days ago, it seemed reasonable to assume there is a problem with v429z. So, I tried to revert to v428. I used the v429z KMotion.exe to re-FLASH the KFlop to v428, which seemed to go OK, but, once done, I get only an error message telling me a KFlop is present, but is not responding. I am now unable to re-FLASH the board. I get the same result with two different PCs, so it appears there is now a problem with the board.
> > >
> > > Any suggestions?
> > >
> > > I am also convinced there is still a problem with MoveExp - it generates infrequent, random faults, even at very low speeds.
> > >
> > > Regards,
> > > Ray L.
> > >
> >
>